Figure 1. Great circle paths that match the satellite data after 19:41. (Click on image to enlarge.)
Introduction
In the last post, I presented how the simulation data found on the home computer of Captain Zaharie Shah suggests that the recovered data were from a single flight session on Feb 2, 2014, in which the aircraft takes off from Kuala Lumpur International Airport (KLIA), flies northwest over the Malacca Strait, flies past the Andaman and Nicobar Islands, turns to the south, and exhausts its fuel in the Southern Indian Ocean. The simulated flight could have represented a diversion of flight MH150 from KLIA to Jeddah, Saudi Arabia, which the captain commanded two days later on Feb 4, 2014.
The alignment of data set 10N (the northernmost), 45S1 (just after fuel exhaustion in the SIO), and Pegasus Field (NZPG) at McMurdo Station, Antarctica, suggests that the simulation user might have selected autopilot and used LNAV mode. If LNAV mode is selected, the aircraft follows a great circle path between the starting and ending waypoints of the active leg. This raises the possibility that McMurdo was used as a final waypoint for navigation with the expectation that fuel would be exhausted in the SIO, well before reaching Antarctica.
In this article, I reconstruct flight paths with the assumption that MH370 was flown in automated flight in a similar way as in the simulation session. In particular, I reconstructed flight paths using the following two criteria:
- After 19:41, MH370 was flown in under autopilot, and was following a great circle path in LNAV mode and speed was controlled using the autothrottle.
- In addition to (1), after 19:41, MH370 was following a great circle path that leads to an airport in Antarctica.
What this article does not consider is how drift analyses based on the location and timing of recovered debris from MH370 affects the impact probabilities along the 7th arc. A review of the available drift analyses needs to be considered along with any path reconstruction studies before a new search area can be recommended with any level of confidence.
Flight Paths Along Great Circles after 19:41
Figure 1 shows the path (green) of MH370 as captured by Malaysian civil and military radar. After the last radar capture at 18:22, all we have to help us reconstruct possible paths are the Burst Timing Offsets (BTOs) and Burst Frequency Offsets (BFOs) from communication between the aircraft, Inmarsat’s F1 satellite above the Indian Ocean, and Inmarsat’s Ground Earth Station (GES) at Perth. The BTO values indicate how far the aircraft was from the satellite, and therefore possible locations that satisfy the BTO values form an arc when plotted. Shown in Figure 1 are the arcs at 19:41 (the 2nd arc) and 00:19 (the 7th arc). While the BTO values are indicative of position, the BFO values are indicative of the speed (horizontal and vertical) and track of the aircraft. The BFO nvalues tell us, for instance, that MH370 was traveling south after 19:41, and also tell us the aircraft was in an increasingly steep descent at 00:19.
The radar path ends as MH370 was traveling northwest in the Malacca Strait. Yet we know from the satellite data that the flight ended in the SIO, and the satellite data after 19:41 shows a progression of values consistent with automated flight. The details of how MH370 might have been flown between 18:22 and 19:41 is still the subject of much debate, and it is possible that multiple maneuvers occurred in this time period. (I’ll be presenting some thoughts on this in the next article, benefiting from some new insights.) For now, we start our analysis at 19:41 and don’t consider the time period between 18:22 and 19:41, but also recognize that some of the paths presented below may be eliminated by constraints imposed when the time period 18:22 – 19:41 is considered.
The reconstructed paths starting at 19:41 and ending at 00:19 are shown in white in Figure 1 for paths at a constant pressure altitude of 35,000 ft. Wind and temperature data from GDAS were used to relate Mach number to ground speed. Each path corresponds to a specific track angle at 19:41. (The track angle will in general vary along each path as would be expected along a great circle.) Initial track angles between 162°T and 192°T were considered, and these paths cross the 7th arc over a range of latitudes between 22S and 40S. For each initial track angle, the position at 19:41 was found that minimized the RMS error for the BTO at the times 19:41, 20:41, 21:41, 22:41, and 00:11. (The position at 19:41 was not constrained to fall exactly on the 2nd arc.) The corresponding RMS error for the BFO values at times 19:41, 20:41, 21:41, 22:41, 23:14, and 00:11 was also recorded.
For automated control of speed, two autothrottle modes were considered for the time period between 19:41 and 00:11: constant Mach number, and Long Range Cruise (LRC), in which the Mach number decreases as fuel is burned and the weight decreases. For the time period between 19:41 and 00:11, a speed is chosen so that the aircraft exactly crosses the 7th arc at 00:19 .
The LRC speed schedule also serves as a proxy for ECON speed, for which the speed also varies with aircraft weight. ECON offers better cost efficiency than LRC because the relative value of time and fuel can be adjusted through a Cost Index (CI) parameter, and because the Mach number is adjusted for wind. However, I chose to not consider ECON speed here because the speed profile would be similar to LRC (depending on the CI), and the exact methodology is not generally available to the public for calculating Mach number as a function of weight, altitude, Cost Index, and wind. (Bobby Ulich has just published a model that makes excellent progress in this regard.) If ECON speed had been considered, it is possible that the BTO values for the same of the paths would have marginally improved.
Figure 2 shows the BTO and BFO errors across a range of latitudes for crossing the 7th arc. Some observations are:
Figure 2. Match to satellite data after 19:41 for great circle paths. (Click on image to enlarge.)
- For latitudes south of 34S, the LRC speed is too slow, and constant Mach number results in a better BTO fit.
- For latitudes north of 34S, the speed reduction offered by LRC speed results in a better BTO fit than for constant Mach number.
- For the paths at constant Mach number, the Mach number varies between 0.801 for a crossing at 22S, to 0.842 for a crossing at 40S.
- Constraining the BTO error to less than 32 μs eliminates paths crossing the 7th arc north of around 26.6S.
- Constraining the BFO error to less than 7 Hz eliminates paths crossing the 7th arc south of 39S latitude and north of 28S latitude.
- The minimum BTO error for LRC speed occurs for a crossing of the 7th arc around 33S latitude
- The minimum BFO error occurs for a crossing of the 7th arc of around 35S latitude.
Flight Paths Leading to Waypoints Past 7th Arc
From among the family of reconstructed paths that follow great circles between 19:41 and 00:19, we consider three paths that align with three airports in Antarctica: South Pole (NZSP), Pegasus Field-McMurdo (NZPG), and Wilkins Runway (YWKS). In August 2014, I first considered a path towards the South Pole that might have occurred after a possible landing at Banda Aceh airport (WITT). Although I have long abandoned the possibility of a landing, the scenario of a holding pattern near Banda Aceh followed by a cruise on a due south course remains an interesting possibility. More recently, the path towards NZPG was investigated in a paper I co-authored with Richard Godfrey, and subsequently Richard proposed the YWKS destination in a separate paper.
The three paths to airports in Antarctica are shown in Figure 1 as black lines that extend past the 7th arc. The coordinates as the paths cross the 7th arc are also shown.
Added Nov 10, 2017: Additionally, a fourth path is shown which aligns with 45S, 104E, which are the coordinates from the final data set found on the captain’s home computer. This case is included to represent the scenario where MH370 was flown towards the location where fuel exhaustion was simulated on the captain’s home computer.
The match of the four paths to the satellite data is shown in Figure 2. Of the four, the path to NZSP crosses the 7th arc closest to the latitude where the BFO and LRC BTO errors are at their minimum values. (Lower BTO errors occur for constant Mach paths crossing at more southern latitudes. However, any path requiring Mach numbers faster than LRC is unlikely to have enough fuel.)
More on the Path Towards the South Pole
The flight towards the South Pole is interesting because there are several ways that the autopilot might be used to create this flight. In LNAV mode, a pilot could enter a custom waypoint with a latitude of 90S and any longitude. Or, he could enter the built-in waypoint for the South Pole, which is SPOLE. Or, if it’s available in his waypoint database, he could enter the waypoint for the runway serving the South Pole, which is NZSP. Any of these methods would cause the aircraft to follow a path that closely follows a great circle to the South Pole. Also, if a pilot wanted to reach as far south as possible, a path towards the South Pole using LNAV would be an obvious selection.
Another procedure would be to use TRK SEL mode, with a value of 180° and the NORM/TRUE switch set to TRUE. Although this would produce similar results as the method using LNAV, it is possible that the aircraft could at times deviate from the required track, and the path could deviate from a great circle. For instance, a wind gust or turbulence could momentarily cause the track to deviate. Although the autopilot would correct for the deviation and bring the track back to the target value of 180°T, the error in path that accumulated during the track deviation would not be corrected. By contrast, in LNAV mode, deviations from the great circle path are continuously corrected so path errors don’t accumulate.
There is also close alignment between this path towards the South Pole and waypoint BEDAX. I’ll discuss this more in a future article.
Figure 3 shows an exploded view of the search area, showing the boundary of what was previously searched (yellow), and where CSIRO proposes to search next (green). The new search area extends about 25 NM to the northeast of the 7th arc at 35,000 ft (blue), and about 27.5 NM to the southwest. The three impact sites proposed by CSIRO are also shown, ordered by their priority. In this part of the arc, the width of the searched area is about 19 NM to either side of the 7th arc. The highest priority impact site is 23 NM to the southeast of the 7th arc, and falls within the proposed search area, which extends between 19 NM and 27.5 NM from the 7th arc.
Figure 3. Exploded view of search area proposed by CSIRO. (Click on image to enlarge.)
The path that extends to NZSP (white) is also shown in Figure 3, which runs along 93.7E longitude. At the point of crossing the arc, the width searched was only about 6.5 NM to the northwest and 15.6 NM to the southeast. The fact that this part of the arc was only narrowly searched presents an interesting opportunity to search in the future.
In the future articles, I’ll present more thoughts on how the MH370 aircraft might have been flown between 18:22 and 19:41, and the implications for possible impact sites along the 7th arc.
Update on November 10, 2017
Here is a CSV file with data for the great circle paths, including the position and track at 19:41, position at 00:19, and speed mode. Included are the data for four paths that align with waypoints past the 7th arc. The four waypoints are the South Pole (NZSP), Wilkins Runway (YWKS), Pegasus Field-McMurdo (NZPG), and the fuel exhaustion position from the simulator data (45S, 104E).
Update on November 11, 2017
In Figure 4 below, I have plotted the paths to the four waypoints (NZSP, YWKS, NZPG, 45S) on the same plot that was generated by CSIRO to show the cumulative probability of detection of low-windage debris by the surface search for various impact points along the arc. The calculated probabilities include the drift that might have occurred between the time and location of the impact and the time and location of the search. It can be seen that there are impact points along the 7th arc and north of 33S where the probability of detection is significantly less than 100%, especially if the impact was to the northwest of the arc.
Figure 4. Efficiency of surface search shown with selected great circle paths. (Adapted from CSIRO.) (Click on image to enlarge.)
Update on November 14, 2017
IG Member Brian Anderson has reminded us that “There are a number of RNZAF photos of interesting flotsam (debris), in areas that may now be much more significant. Unfortunately none was ever recovered.” Two of the more interesting photographs are found below. The first photo (Figure 5) is an unknown object that could be part of an aircraft. The second photo (Figure 6) is a field of floating debris.
Figure 5. Aerial photograph (5849) of unknown object on March 29, 2014, at 28.8866S, 96.0844E. (Click on image to enlarge.)
Figure 6. Aerial photograph (5832) of debris field on March 29, 2014, at 28.3927S, 97.7750E. (Click on image to enlarge.)
Both photographs were taken on March 29, 2014, at coordinates not far from where the great circle path to 45S,104E crosses the 7th arc near latitude 28.3S.
In order to determine if these objects are consistent with the expected drift of debris from an impact near 28.3S on March 8, 2014, I used the CSIRO-generated drift results for an impact on the arc near 28S latitude, where the drift model was seeded with objects within an approximate +/- 0.5 deg square area. The drift model results are shown in Figure 7 for the calculated position of debris on March 29, 2014.
Figure 7. Drift results from CSIRO for an impact along the 7th arc near 28S latitude. (Click on image to enlarge.)
The results are shown for low windage debris (red) and high windage debris (green), and should be representative of a range of objects produced by the impact. Both the unknown object and the debris field are found in the general vicinity of where the model predicts objects would drift for the modeled impact location. This makes the possible impact site of 28.3S even more interesting.
@Victor,
Continuation from the previous thread…
Re: “I don’t understand the reluctance to recognize that the OI offer represents an extraordinary opportunity.”
We don’t know what OI’s offer and capabilities are. Fugro, in contrast, is a well-known heavy-weight player in the world of marine surveys. We used their services quite a few times, and they always produced data and reports of excellent quality. But Fugro is expensive compared to other surveyors. OI is a black horse to me, and frankly I wish both OI and Fugro be involved into the search.
Is OI’s contract a lump sum contract or reimbursable? If the former, then what OI’s management is going to do if/when they reach the limit? Declare bankruptcy? Given the length of the arc, this is a pretty high risk. If the latter, then obviously OI will be interested in searching as long as possible. Or are there some timeframes? I guess these points could be a subject of negotiation.
Anyway good to know that things are moving again, and that the OI’s strategy is what I suggested a year ago or more.
With regard to other matter. I found your WYKS route interesting for at least 3 reasons:
1. 172 deg track. This is a key feature of the “radar alerting maneuver” I suggested earlier.
2. Your/Richard’s positions 19:41 are 6.5-6.8N, 93.3-93.4E. My positions, depending on the type of the exit mistake, 93.3-93.5E, 5.2-5.7N. Pretty close.
3. The terminus around 29.7 is within my “compatibility zone” based on my drift study.
The intriguing feature about another route, NZPG, is the terminus at the origin of the “Curtin boom” should it be on the 7th arc. For some mysterious reason the ATBS has completely ignored it. I am only wondering whether you kept this aspect in mind when you suggested respective terminus to OI?
P.S. Why would MH370 be following “a great circle path that leads to an airport in Antarctica”? Logic-wise?
“After 19:41, MH370 was flown in under autopilot …”
So before 18:22 the plane flew a purposeful path, and after 19:41 it flew an even more purposeful path, yet during the one long gap in the coverage (or not, depending on how you want to treat the 18:40 phone event), the plane decided to “loiter”. Fiddle and diddle. Who knows, maybe it did. I wasn’t there. Don’t think you were either.
Were Antartica waypoints in the 9M-MRO navigation database, given the limited amount of memory storage in the AIMS-1 flight maangement system?
@Oleksandr: Have you been following the discussion about OI’s offer? They only get paid if they find the wreckage. They have zero incentive to prolong the search. As for Fugro, they didn’t make an offer of this type. They expect to get paid whether or not the wreckage is found.
@sk999: I disagree with your premise. If the plane was in a holding pattern, it did so for a reason, i.e., the delay was intended once out of Malaysian airspace. Possible explanations have been proposed here. I didn’t speculate in the post as to why a delay might have occurred.
If you have a path and associated scenario that is “purposeful” and meets all the criteria, please present it.
As for memory storage in the AIMS, I’m told that Malaysia Air only manages information on routes and waypoints they fly. Exactly what that means, I don’t know. SIDs and STARs for airports consume a lot of memory, so I can understand why not all are included. Airport waypoint coordinates do not.
@Victor
Your figure 2 above is missing the data for BFO error – M constant.
Victor wrote, “I disagree with your premise”
I did not invoke any premise. I just made an observation.
What is a purposeful path? One where the plane is intentionally directed to go from one point to another – say, in LNAV mode (but other modes are possible) – primarily via waypoints. A route BEDAX to the South Pole via LNAV or True Track is such a route. Many have suggested that portions of the turnback up to the point of last radar contact were flown via waypoints (e.g., KADAX, ENDOR, OPOVI, VAMPI, MEKAR, airwayy N571, …). So there is all this waypoint-to-waypoint travel, but then in the middle for over an hour, the plane fiddles and diddles. It makes you go hhhmmmm….
@Victor
Constraining the BFO error to less than 7 Hz eliminates paths crossing the 7th arc south of 39S latitude and north of 28S latitude.
Your use of an RMS metric for BFO error is questionable. I have beat this drum many times. Even well documented previous 9M-MR0 paths i.e. Figure 5.4 of the DSTG book show that errors larger than 7Hz occur, and that the distribution of errors is not a normal distribution. An RMS metric is inappropriate.
@Victor
Figure 3 a little confusing.
I believe the CSIRO2 and 3 are transposed.
The big red diamond I like too but that is apparently not what CSIRO likes.
@DrB
Thanks a lot for those tables. Impressed again about the amount of detailed work you put into this.
And indeed I’ve got my anwsers on my questions to you.
As expected range is considerably shorter with one engine INOP at FL240 compared to both engines operating FL350 (this is what I had in mind @Paul Thompson).
I see LRC INOP at FL240 falls only 2 minutes short on your endurance/error contraints (00:27/00:25).
And only 4 minutes more than MRC (00:22). Range though is ~300Nm shorter with LRC INOP at FL240.
LRC one engine INOP still looks like a theoretical possibility to me with those figures. Crash area would be ~300Nm north compared to MRC or CI52 as I see it.
Would you exclude LRC INOP based on your findings?
@VictorI
I have wondered about this before and like to bring it up now.
The 45S1 and 45S2 simulator points kind of show a similarity with the final BFO’s; a very steep descent from high altitude without fuel.
With the simulator 45S2 the ‘descent’ is ending at 4000ft.
Am I completely wrong or could it make sence that the end-flight scenario was rehearsed (also).
@VictorI
Interesting new article with ample food for thought & discussion.
I would be curious to know:
– The 19:41 latitudes for the different paths
– The average BFO error graph in addition to the RMS error
Could you perhaps share those numbers as well?
@Victor Iannello:
RE: “ECON offers better cost efficiency than LRC because the relative value of time and fuel can be adjusted through a Cost Index (CI) parameter, …”
What is the value of time on a flight to eternity?
@Niels 1941 lat/long and BFO rms was on the dropbox link
https://www.dropbox.com/s/quhfygiw29edfs2/LNAV%20Tracks.png?dl=0
@Dennis W and path-fitters. Asking my question another way. If there are “equally good” candidate path models (viz BTO error RMS) available as you rotate around 1941, why does the Bayesian pdf have a peak at all?
@DennisW: Your use of an RMS metric for BFO error is questionable
I made observations regarding the RMS BFO error. I showed where the minimum is. I didn’t say I know the plane is there. I still think there are lots of candidate spots.
Your figure 2 above is missing the data for BFO error – M constant.
The BFO errors for the constant M and the LRC cases are essentially the same.
@Paul Smithson: If there are “equally good” candidate path models (viz BTO error RMS) available as you rotate around 1941, why does the Bayesian pdf have a peak at all?
In the Bayesian analysis, they started with the radar position at 18:02 and reconstructed paths with a prior distribution for the manoeuvers that minimizes the number of turns. Any path that requires a “delay” between 18:22 and 19:41 would have a low probability because of the number of manoeuvers it would require.
@Ge Rijn: I don’t think the sim data shows a steep descent.
@TBill: Thank you for finding the error. I had the labels for CSIRO 2 and 3 assigned to the wrong positions. The figure should be correct now.
@sk999 said: So there is all this waypoint-to-waypoint travel, but then in the middle for over an hour, the plane fiddles and diddles. It makes you go hhhmmmm….
Yes, it should make you go “hhhmmmm”. However, the impact sites associated with the simplest of LNAV paths have already been searched without result. That should also make you go “hhhmmmm”. There is an error in the assumptions made in the Bayesian analysis.
Again, I’ll repeat my request that you present your best estimate of one or more paths and the associated scenarios.
@VictorI
Assuming SIM point 45S1 and 45S2 are related as closely as they seem, do you have any explanation/suggestion what SIM point 45S2 with altitude 4000ft could mean?
@Oleksandr asked: I am only wondering whether you kept this aspect in mind when you suggested respective terminus to OI?
I have not suggested a terminus to OI. If asked, I would likely include multiple candidates, some of which are presented above. Frankly, I am open to hearing what recommendations others here might have.
@VictorI
And another question. Why did you put your potential crash area west of the searched area and 7th arc around 34S/93.5E and not east of the 7th arc?
According the final BFO’s at the 7th arc and the simulator runs showing mostly turns to the left after flame out, would it not be more logical your potential crash area would be east to the 7th arc around 34.5S/94E?
@Ge Rijn: I know exactly what it means. In the paper I wrote with Yves, we explained that just before saving the 45S2 data set, the altitude was manually changed from 37,654 ft to 4,000 ft, as reflected in the respective values for the parameters AGL and Altitude. This can be repeated.
@Victor
Your reply to me:
I made observations regarding the RMS BFO error. I showed where the minimum is. I didn’t say I know the plane is there. I still think there are lots of candidate spots.
Actually what you said in your narrative above:
Constraining the BFO error to less than 7 Hz eliminates paths crossing the 7th arc south of 39S latitude and north of 28S latitude.
What you said in your narrative is axiomatically true. My assertion is that the statement is meaningless. My guess is that the 7 Hz comes from the Ashton et. al. Inmarsat paper. Where it is clear that it corresponds to a peak error measured on a particular flight. Not an RMS error associated with that flight. Using an RMS (or peak) BFO error to constrain flight paths cannot be justified. BFO can tell us that 9M-MRO went South after the FMT and dropped out of the sky rapidly at 00:19. Beyond that it is essentially of no value.
@VictorI
On your reply:
‘I know exactly what it means. In the paper I wrote with Yves, we explained that just before saving the 45S2 data set, the altitude was manually changed from 37,654 ft to 4,000 ft’
Yes, but it does not explain (to me) why this was done and what those 45S2 values could mean in relation to 45S1.
I don’t want to bring up old stuff I probably don’t understand well enough. I just wonder.
@DennisW: I fully admit that I don’t know whether or not the BFO error should be used to constrain southern paths. I do know that the DSTG assigned such a large standard deviation to BFO values that it only assured that paths were probable if the plane was flying in a southerly direction starting at some time between 18:28 and 18:40. We know that the NZPG path has a larger BFO error than the NZSP path, for instance. I don’t know if that indicates a preference for the NZSP path.
CSIRO has confidently said that the surface search should have detected debris from an impact north of 33S. I view this statement with an appropriate level of skepticism, but I also don’t assign zero value to it.
@Paul Smithson, VictorI
Thanks, Paul, Victor, to be more precise: I’m interested in the 19:41 positions for all 16 paths, as well as the (arithmetic) mean BFO error (the RMS values are already given in the graph, figure 2)
@VictorI
On your comment:
‘ Frankly, I am open to hearing what recommendations others here might have.’
Frankly, I think the probability of the CSIRO 1/2/3 is quite low probability-wise (and not only). All three being just a few NM outside the searched area would be too big a coincidence imo.
CSIRO/ATSB is holding to a ‘live-line’ here I think.
The area is searched till ~32.75S according a high speed descent crash scenario at 00:19. Nothing has been found.
So something must be wrong with this assumption imo.
Then your discribed flight paths all still assume an uncontrolled end of flight with waypoints or track set just before or after FMT and not altered till fuel exhaust. Still assuming a ghost-flight after FMT or a conscious pilot who just waited for the plane to crash at fuel exhaust..(?)
I think it’s essential (also to OI) to consider the possibility of an all controlled (end of) flight with a certain destination/waypoints set just before or after FMT.
A destination that could serve a possible motivation; to hide the plane and evidence of the one(s) responsible as good as possible.
The best area along the 7th arc with this purpose in mind is just outside the searched area east and west of the 7th arc between 32.5S and 33.5S and 95/96E in the trenches under Broken Ridge.
@Victor
(1) How do we fly LRC in the MS Flight Sim, other than manual changing speed?
All I see is we can set Cost Index or Mach/IAS. Perhaps PSS777 cannot do LRC.
(2) Why is the red diamond in Fig. 3 inside Arc7 (that’s close to where I would put it by descending after Arc5)? You normally have crash past Arc7 by FE (fuel exhaustion) at FL350.
(3) I tend to view matching the 2314 telcon BFO (216 BFO) as a key indicator of path fidelity. However, the 2314 telcon BFO goes off at some point, so I take the steady initial value (216).
It would be helpful to see representative tables of data (including 2314 phone call) maybe one path each for LRC (NZPG) and constant speed (SPOLE path). Frankly I use the data to try to fly the route in FS9 by manually changing winds etc at various points.
@TBill: Not having LRC capabilities is one of the many limitations of the PSS777. If you want to use MSFS as a tool, you would be wise to upgrade to the PMDG777 model on FSX.
@TBill, @Ge Rijn: The only reason that I highlighted the area along 93.7E longitude to the NW of the arc is that the area searched was closer to the 7th arc than the area searched to the SE. The search width was extended to only 6.5 NM to the NW from the 7th arc at 35,000 ft, which is within the error bounds of the location of the 7th arc. For the 7th arc at sea level, the search width was only extended to 2.1 NM to the NW from the arc. It is possible that the aircraft was traveling along 93.7E longitude, and impacted the water at the same coordinates as where it was for the last satellite transmission, but yet was missed just due to noise of the BTO measurement. It’s also possible that during the descent, the plane made a 180° turn and backtracked.
@Ge Rijn: If the steep descent was induced by a steep bank that was in turn caused by the re-start of the left engine, the turn would be to the right, not the left. Also, a 180° turn to the right from an initial 180° track would be closer to the unsearched area than a 180° turn to the left because the arc runs NE-SW.
@Ge Rijn
“I think it’s essential (also to OI) to consider the possibility of an all controlled (end of) flight with a certain destination/waypoints set just before or after FMT.”
I agree with you. There are 3 possibilities
(1) Uncontrolled/Accidental Ghost flight (majority of proposed paths)
(2) Controlled/Intentional Ghost Flight (Victor/Richard)
(3) Intentional Live Pilot Flight
Many feel it was either (2) or (3) above, but especially Option (3) is under-represented. Part of the rationale for not considering Option (3) is that it is a “Wild Card” hard to say where the crash site it. But it is conceivable that the live pilot actions were after Arc5 and they were not too extreme, in which case we should not ignore those cases.
Another way to say it, I think we can show close to exact math to BTO/BFO with intentional flight. So far all of the rigorous path prognosticators refuse to show us that case. My 180S CTH case pretty much meets all, but I am not putting myself in the rigorous path prognosticator (RPP) category. I am in the Not Ready for Prime Time Path Prognosticator category.
@Ge Rijn
P.S.- Yet another way to say it, I view Arc5 to Arc6 somewhat similar to Arc1-Arc2…that is to say they were some intentional maneuvers, presumably less drastic between Arc5 to Arc6, but also with potential to save fuel (eg; long desacent).
@TBill: I don’t think I ever said whether I believe there was a pilot alive at 00:19, and I prefer to not use the term “ghost flight” because it has been used to mean many things, as has the term “death dive”. A steep descent at 00:19 could have occurred with or without pilot input. What I think is unlikely is that a steep descent was followed by a long, controlled glide.
@VictorI
On your comment:
‘CSIRO has confidently said that the surface search should have detected debris from an impact north of 33S.’
I like to remind to the fact debris was detected north of 33S.
The flaperon sized/shaped piece detected was photographed but never recovered. It was later explained by bloggers as an aircraft pallet which makes you wonder still. There was never an official explanation.
Anyway that specific piece was drifting there in the middle of nowhere but right near the area and time MH370 is now assumed to have crashed.
@Victor
“due to noise of the BTO measurement”
I just got slaughtered last thread asking if there could be some BTO error.
I fly FS9 PSS777 to be consistent with what ZS might have seen. I realize PMDG is fantastic almost pro simulator. So far I am happy with constant Mach simulations.
Have you ever tried to match a MH370 path with PMDG 777? I assume we have some errors like impact of winds, shape of Earth etc.
@VictorI @TBill
A ‘ghost flight’ implies to me there was somewhere after FMT no one alive anymore, at least no pilot to control the plane.
This makes a big difference with possible pilot input at an end of flight scenario.
Indeed the steep descent and a high speed nose down impact could be initiated by a pilot too without a glide after.
But the big difference with a ‘ghost flight’ is the impact area could have been selected.
@TBill: All reasonable people understand that there is noise on the BTO measurement.
I don’t think it is useful for me to simulate paths after 19:41 using the PMDG777 model, except for the end-of-flight. It is too cumbersome to incorporate temperature and wind fields, for instance and the simulations take too long. (Time acceleration introduces errors.) On the other hand, it is very useful for simulating the flight before 19:00, including FMC programming, speed modes, turns, descents, and holds.
re: @TBill
As an original member of the IG and a proponent of the
“rigorous path prognosticator (RPP) category” approach, I have not been willing to throw my hat in the ring again, despite having some opinions on the subject based on further simulations.
The continued leakage of scraps of information and inferences derived therefrom is significant as the lack of transparency by governmental authorities (which is typical of all governments) encourages too much speculation.
There are apparently enough simulations that have merit so as to encourage entrepreneurs to undertake further searches and the proposed search areas take into account many of the suggestions made by the IG and contributors to this blog.
That the search will continue is the best news I have heard in a long time.
Sid
@Victor said “Yes, it should make you go “hhhmmmm”. However, the impact sites associated with the simplest of LNAV paths have already been searched without result. That should also make you go “hhhmmmm”. There is an error in the assumptions made in the Bayesian analysis.”
Yes. That is what has been the most frustrating and led to many conspiracy theories (one of which may turn out to be the truth, who knows…) But I do appreciate seeing this comment. It should be straightforward, but some haven’t been able to say it: the plane was not found in the highest probability areas (based on prior analysis). It is either not there (very high probability) or it was missed in the search (exceptionally low probability). I hope they find it this time. This is an unimaginable tragedy, a puzzle of epic proportions, and a potential threat to air safety globally. Praying for a stroke or two of luck to go with the genius from you all.
Separately, I recall something about the quicker and better search possibilities with OI from your previously in-depth report on their search technology. Am I recalling right? They will need it to zig zag all over finding the plane or (sadly) dispensing with the CISRO sites quickly.
Final comment on all the secrecy. I know some people can’t stand not to be told people’s sources. It does appear unseemly at times (thus some of the angst that flashes between IG members and others). It is very frustrating. And it does mean data/info cannot be scrutinized properly. But it is also in the nature of things, not just MH370…sometimes there are good reasons for it. Sometimes just an imagined sense of importance and ego in controlling the information. But, either way, it shouldn’t be a shock to folks that some people like to keep secrets. I hope it has not in any way impeded the work done by many here (IG and others). We will know one day as everything ends up in the sunshine sometime. I advise all to keep calm and constructive.
More searching is great news. Praying for the best.
@VictorI,
In response to your question on my endurance predictions, they all assume Air Packs are ON all the time until MEFE. If you turn the air packs OFF at 18:29, the endurance is increased by about 6 minutes.
@sk999,
You said: “You first state, “I updated the CI = 52 speed schedule based on matching the MH370 Flight Plan and MH371 ACARS data.”
You then state, “HERE is a paper demonstrating MH371 was flown with Cost Index = 52.”
Is there there circular reasoning here?”
No. The conclusion that MH371 was flown at CI = 52 has nothing to do with a fuel model. It is based on the excellent consistency of the air speeds between MH371 and the MH370 Flight Plan (which we know used CI = 52) as a function of equivalent sea-level weight Wo.
As a separate issue, I modified the CI = 52 speed prediction of my model at low weights in order to be consistent with both sets of data.
Many thanks Victor for an excellent post!
@DrB On the previous post you asked some questions:
1. What are the flight levels?
I tried altitudes between 18,000 feet and 36,000 feet. The 3 LNAV tracks were all computed at 35,000 feet.
2. How do you handle the effects of headwinds/tailwinds in computing ground speed?
SQRT(POWER(‘Calculated Air Speed (knots)’,2)+POWER(‘Predicted Wind Speed (knots)’,2)
−(2×’Calculated Air Speed (knots)’×’Predicted Wind Speed (knots)’×COS(RADIANS(Diff))))
Where Diff is ‘Instantaneous Track (degs True)’−’Predicted Wind from Direction (degs True)’
3. Are you using the GDAS temperature data in computing the TAS?
Yes
4. Have you figured the fuel consumption for any of these routes?
Yes, the fuel consumption is re-calculated every 15 seconds.
5. Are you assuming a constant Mach from 19:41 until fuel exhaustion?
I have tried various speed schedules, the 3 LNAV tracks were all calculated using LRC mode.
6. What is your justification for fitting a 3-digit Mach number? After an extended Hold, why would a pilot select a 3-digit Mach using the MCP?
The tracks were calculated using LRC mode (19:41 0.815, 20:41 0.807, 21:41 0.799, 22:41 0.791, 00:11 0.780). This is not pilot input, in this case,
7. How many waypoints are entered to define the course from 19:41 until MEFE? Which ones?
Just one ultimate waypoint NZSP, YWKS or NZPG.
The start points at 19:41 are chosen from the best fit:
https://www.dropbox.com/s/h8olw7w4moaauux/2nd%20Arc%20Limits.pdf?dl=0
8. Which BTO times are you using to compute the RMS of the residuals?
19:41, 20:41, 21:41, 22:41 and 00:11.
@Ge Rijn,
You said: “I see LRC INOP at FL240 falls only 2 minutes short on your endurance/error con(s)traints (00:27/00:25). . . . Would you exclude LRC INOP based on your findings?”
I used a 2-sigma criterion (± 8 minutes), and LRC INOP (at +10 minutes long) was just outside that limit. I wouldn’t exclude it, but is seems to be of a rather low probability.
@Victor “I have not suggested a terminus to OI. If asked, I would likely include multiple candidates, some of which are presented above. Frankly, I am open to hearing what recommendations others here might have.”
Very early on it was suggested MH370 went out into the Andaman Sea, came back towards Banda Aceh then round below Sumatra and ditched around 10S 108E. This required a PIC for at least most of the flight and only one more waypoint than the flights from ISBIX to the Antarctic.
Surely this is not unreasonable particularly in light of the motive that has been well canvased here and despite no whistleblower having come forward.
@TimR: A crash 100 NM off the southern coast of Java should have produced debris in Indonesia, which never occurred. It also makes no sense that the plane would run out of fuel after such a short distance was traveled with so many airports that were reachable.
You believe you have inside information that the disappearance was the result of a political negotiation that was unsuccessful. I ask you again to reveal more details or your information will continue to be ignored.
@Victor
Poor choice of words on my part, when I said I got “slaughtered” it was my sense of humor. I am most happy to hear a suggested location inside Arc7 at the red diamond. So as far as Arc7 deviations we have (a) +-5.3 nm error bar, (b) altitude adjustment per BTO calcs, and (c) anything else (noise?).
@Sid
Thank you for the perspective, it is very interesting to me.
@Paul,
@Dennis W and path-fitters. Asking my question another way. If there are “equally good” candidate path models (viz BTO error RMS) available as you rotate around 1941, why does the Bayesian pdf have a peak at all?
Not ignoring you, Paul. I just had to let my thoughts settle relative to your question. My opinion is that Bayesian analytics were inappropriate for MH370 flight path modeling. The priors assumed were just that – assumed, and not based on any relevant sampled data i.e. previous 9M-MRO flights were not representative of a diverted flight. Bayesian analytics only work well with relatively solid priors.
@Victor
One possible explanation for Shah’s SIM flight:
Firstly, I don’t believe he was planning to fly toward McMurdo Station until fuel exhaustion. The distance from Kuala Lumpur to the turn back 30NM in front of DOTEN, and back to latitude N7.5 was judged by Shah to be roughly equivalent to the distance MH370 would be flying from Kuala Lumpur to IGOGU, and the FMT on 8th March. He used the SIM flight into the SIO to obtain the distance he could travel at a constant Mach, on the Jedda fuel allocation. He would then be able to work out a rough estimate of the distance he could fly into the SIO from IGOGU at similar constant Mach on MH370, taking the proportionate difference in fuel load between MH150, and MH370 (estimated, from previous flights) as a multiplier.
He could then draw an arc centred on IGOGU, radius equal to the estimated distance applicable to MH370, draw a second line for sunrise at takeoff plus 7.5 hours. Where the two line cross is the point to aim for IF he wanted to synchronise fuel exhaustion with sunrise.
Thus he would have simulated the MH370 flight, using the MH150 flight, would have left no trace of the proposed MH370 flight on his simulator, and laid a false trail in the direction of McMurdo.
@IG
Based on previous (acrimonious)discussions relative to the OI offer, I am convinced that you know what the offer is. Your comments betray that familiarity.
My current opinion relative to the “snag” is that Malaysia wants to architect a fee based on time and materials i.e. the fee would depend on the area searched as opposed to a fix fee. This structure would avoid the embarrassment of a windfall by OI. Of course, OI would then be in the position of acting like a Fugro, except without any guarantee of getting paid.
Looking at the Fugro effort, my conclusion is that they received about $100M (USD) for searching 120,000 km^2. A simplistic ratioing (ignoring the fixed startup costs) puts the equivalent value of 36,000 km^2 searched at roughly $25M to $30M. Close to my previous estimate.
@Victor “A crash 100 NM off the southern coast of Java should have produced debris in Indonesia, which never occurred. It also makes no sense that the plane would run out of fuel after such a short distance was traveled with so many airports that were reachable ……. I ask you again to reveal more details or your information will continue to be ignored.”
I accept no debris is strange but not unexpected considering the direction of flow up close to Java.
Its not such a “short distance travelled” in that it is the same time in the air as the other proposed flights.
My hands are tied. I am sure you would well know that confidences need to be kept.
This is a valid flight path, despite not meeting the apparent requirement of being an uncontrolled flight after ISBIX.
From 19:41 onwards the BTO’s accurately fit right through to the ditching and the BFO’s are a reasonable fit up to 22:41.
@Tim R
Victor, is correct. We need more. If you cannot provide it, you are relegated to the “whacko” category. Of course, that includes the IG based on recent discussions.
@DennisW
I also sense something, but that’s way better than nothing happening.
@Victor
I am curious as to why you put that red circle on the “2nd Arc Limits” slide, indeed, why you titled it that way at all.
I first assumed that you were plotting the start points for the three destinations, and were thus effectively saying “these are the maximum limits of the aircraft on the second arc” that should be considerd. If that were the case, there would be no need for the red circle at all, so why did you plot it ?
Just looking at it, it appears to be centered on either Nilam or Radar End, and the radius of the red circle from either would presumably be estimated ground speed times the time interval from either position until 19:41.
Could you:
(a) nominate the centre point and time, and
(b) nominate the effective ground speed used for the radius, and
(c) nominate the actual iintersection points (north and south) with the 19:41 second arc, and
(d) explain why you plotted it.
@Rob
Re: The Captain’s sim flight as an exercise in fuel planning.
The simulation program used by the Captain models a B777-200LR, an airplane with a higher empty weight, different wing and more powerful engines to the B777-200ER. Does that sound like the sort of tool you would employ for an exercise in fuel planning?
Moreover, we know from Victor’s and Yves’ paper that the fuel used on the recovered flight differs significantly from a recreation of what was ostensibly the same flight. The recovered data shows a 45% discrepancy to recreated fuel usage at 5N increasing to nearly a 65% discrepancy at 10N. Whatever the Captain was simulating/modelling it is highly unlikely that it was related anything even vaguely connected to fuel planning.
@TBill
“I also sense something, but that’s way better than nothing happening.”
I agree. Resuming the search is good thing. I remain pessimistic relative to the outcome. The area defined by the Iannello and Godfried McMurdo path is where the aircraft is, IMO.
Not to say that that the IG is not in the whacko category.
@ventus45: The second arc is the locus of points corresponding to the measured BTO value at 19:41. The starting locations for the great circle paths fall near (but not exactly on) that arc. Each starting track angle that I considered corresponds to a different starting latitude. As you change the starting track, the 19:41 position rotates around the arc. The starting positions at 19:41 don’t converge to a single point, if that is your question.
@Oleksandr
“P.S. Why would MH370 be following a great circle path that leads to an airport in Antarctica? Logic-wise?”
Conjecture of course but here is one possible reason- DOTEN to NZPG cuts nicely around Indonesian FIR air space boundary. This could have been important if the target hijack flight was e.g. MH150 hitting DOTEN at an earlier time of day when radar was operating. Various other implications if this theory fits the reason.
@sk999
“Were Antartica waypoints in the 9M-MRO navigation database, given the limited amount of memory storage in the AIMS-1 flight management system?”
Not a big limitation right? I use 78S67 oceanic waypoint shortcut for NZPG. With a little more complexity, any pin-point on the globe can be entered as a waypoint.
There are only 2 Waypoints actually on (or very close) to the 2nd Arc: VOCX and ISBIX.
The aircraft speed required from 18:28:14 UTC to be on the 2nd Arc at 19:41:03 UTC is:
(1) 173.8 knots for VOCX.
(2) 353.1 knots for ISBIX.
Most likely, neither case was a non-stop direct flight track:
(1) For VOCX the speed is too slow and hence your loiter theory.
(2) For ISBIX the direct route passes right over the tip of Sumatra and, if Indonesia’s radar was active, it would be seen as a threat.
But with an indirect route around Sumatra, ISBIX is reachable without a loiter.
This makes a small preference for NZSP above NZPG in my view.
On top of the “No Loiter” argument, you have outlined above the ease of entry for the NZSP option into the navigation system.
I favour the hot spot priority order:
(1) CSIRO.
(2) NZSP.
(3) YWKS.
(4) NZPG.
@Andrew. In a post on Sep 1st at 3:58 AM I asked you for an estimation of the internal diameter of the fuel line from the APU DC fuel pump to the APU. You kindly obtained an answer, at 8:14 PM on the same day, “I’m told the outside diameter of the APU fuel line is one inch (25.4mm – forgive me I was brought up on metric). If we say the inside diameter is 20mm……”
In thanking you for that I commented, “I assume they mean outside diameter net of the shroud in the forward section.”
While looking into another fuel question I have delved again into this. The attached 3 pages are applicable to the -200ER (the AMM manual pages on the -300ER are identical). Page 127 depicts a flexible rubber and kevlar line inside a shroud. Also from the sense of the description on page 126 and the diagram on page 123 the shroud extends over the whole length of the flexible line.
For that line to hold sufficient fuel for the APU to power the 7th arc log on it would need to hold a minimum 4.4 lbs, 2 kg, of fuel. For that it would need to be of at least 10mm internal diameter and that supposes all fuel in it would be drawn through by the APU. Yet should the 25.4 mm line you refer to in fact be the shroud, if you scale from that on p127, the outside diameter of the flexible line is less than 10mm. One would expect that type of tubing to have appreciable thickness.
Do you have access please to a -200 or -300ER parts listing which might specify the internal line diameter? Alternately could I again prevail upon you to ask your tech gurus whether they can confirm the flexible line’s external diameter is 25.4 mm or close to?
https://www.dropbox.com/s/t0y366qqakaqnwi/APU%20fuel%20line.pdf?dl=0
@ALSM. The meanwhile please lie back and think of England.
@Oleksandr: You might be interested in this new article that (in a very simplified way) discusses the data collected by the Cape Leeuwin hydrophone station as it relates to MH370. Neither of the two relevant signals is (statistically) close to the 7th arc.
@Paul Smithson
Re “@Dennis W and path-fitters. Asking my question another way. If there are “equally good” candidate path models (viz BTO error RMS) available as you rotate around 1941, why does the Bayesian pdf have a peak at all?”
I don’t class myself as an accomplished path-fitter, however it seems to me that the DSTG “hotspot” was determined as much by the assumed range of possibility cruising speeds on autopilot (as advised by Boeing) as it was by the BTO errors. The IG found the same thing when they modeled BTO/BTO RMS errors, first for LRC
cruise, and then updated for CI51 ECON; the effect was simply to shift the 7th arc crossing point a hundred or so NM further west. If I’m wrong on this, hopefully Richard will clarify. The range of paths are also limited by the endurance on the available fuel. Hence the peak at S38.
The orientation of the 2nd arc serves to constrain the 19:41 crossing point, in the east-west direction, if you don’t consider the loiter scenario, which the DSTG analysis didn’t.
The DSTG analysis did allow for changes in direction at each crossing, but found that a straight path was preferred, which for me at any rate prompts one of those “hmmmm” moments. The BTO rings turn out to be particularly amenable to straight line paths, suggesting to me that the plane most probably flew in a straight line after FMT.
If this wasn’t what your original question was asking at all, then please pardon my intrusion.
@Rob
Maybe even more the range of paths is depending on which altitude the plane was flown most of the time till the 7th arc.
F.i with a drop in altitude from FL350 to FL300 the range decreases with ~150Nm according @DrB’s ‘endurance/range table’ in two engine operative speed modes. Decreasing further when altitude drops further.
As far as I’m aware no one still knows at which altitude the plane flew most of the time till the 7th arc.
~150Nm is about 2 degrees latitude which is quite a lot when you want to define a search area.
@Rob said: The DSTG analysis did allow for changes in direction at each crossing, but found that a straight path was preferred
The posterior distribution shows a preference for straight paths because of the prior distribution they used for the number of manoeuvers. For instance, a delay of an hour, be it a holding pattern or excursion, would have required many manoeuvers of a particular type, and would be improbable because of the distribution assigned to the number of manoeuvers. The prior distribution was based on a study of a commercial flights, where planes tend to fly in near straight lines. There is no reason to believe that MH370 was flown like a typical commercial flight. People don’t seem to understand this. They jump to the conclusions without investigating the assumptions.
@Richard
I am in general agreement with your search area priorities. I would say:
(1) NZSP (or 180S CTH)
(2) CSIRO
(3) NZPG
(4) YWKS
Re: your ISBIX discussion – note that ISBIX to NZPG seems to go directly over Dordrecht hole area, which I see as one possible flight strategy. If so, I assume the pilot got to ISBIX late or had less fuel than expected, and knew that Dordrecht was non-reachable.
Therefore the alternative strategy I see is 180S to deep water beyond Broken Ridge due south of ISBIX, BEDAX, and OK, add VOCX.
NZSP (South Pole) I see as 180S CTH. I believe there is a visible wind impact in the data. which means the “Fig 3 red diamond” could be offset from Victor’s to center around 94.2E. I also think they need to be looking inside Arc7, whereas the lofty FL350 Arc7 is a little unfair (it shifts the search zone more outside Arc7).
@Richard
I favour the hot spot priority order:
(1) CSIRO.
(2) NZSP.
(3) YWKS.
(4) NZPG.
Thx. I was going to ask you about that. Personally, I have difficulty with the CSIRO hotspot.
The CSIRO claims relative to the air search are a bit annoying to me. The air search in the area they are “discarding” was started nine days after the aircraft went missing. I have no experience in ocean air searches, but people I have talked to out of Moffett AFB claim that significant dispersion and sinking would have occurred in that time frame.
The NZPG terminus that you and Victor modeled would be my first choice since there is actually an “evidentiary” pointer to it.
@DennisW: Two questions:
1) For 26.9S (NZPG path), how far along and across the 7th arc would you search, and why?
2) If not at 26.9S, where would you search next?
@Victor
I would assign equal boundaries to LNAV (along the arc) and BTO (orthogonal to the arc), +/- 40km each. Search area ~ 6400 km^2. Not much different than the 10000 km^2 assigned by OI (which I also think is a reasonable choice, BTW).
After that I would do the same for NZSP simply based on the ease of using it as a waypoint.
@DennisW: Would you consider searching a narrower width between the hotspots?
@Victor
Yes, that also makes sense IMO.
This, is a very interesting paper.
https://www.nature.com/articles/s41598-017-14177-3.pdf
I can’t help but notice, that E2 (see Figure 4 in the paper) is virtually due south (180 True) of UPROB.
The distance from UPROB to E2 is approximately 2,656 Nm.
At an average 480 knots GS, leg time = 332 minutes = 5 hours 32 minutes.
The event Time @ E2 is calculated to be between 00:25 and 00:31.
So, the time at Uprob would be between 18:53 and 18:59 (for 480 knots GS to E2).
Position E2 obviously does not gell with either the 6th or 7th Arcs, so, for the moment, ignore them.
Position E2 obviously would be impossible to reach for any FMT near Ache off the Radar / Lido slide, so, for the moment, ignore them.
Position E2 is however, within reach of a Southern FMT, at, or in the vercinity of, Uprob, which can be reached coming off a path from the 17:52 phone call at Penang, via Medan and Merim.
@Paul Onions,
I will answer as many of your questions as I can.
You said: “. . . what is the expected fuel remaining at Igari? Is this higher than the estimate in the Factual Information report?”
At IGARI near 17:22 UTC my Fuel Model V5.4 indicates 42.1 tonnes of fuel remaining. The revised Factual Information in Section 1.6.5 gives an estimated 41.5 tonnes (but no details of how this was done). Since we know that the ACARS fuel report was 43.8 tonnes at 17:07, the Factual Information estimate was for (43.8-41.5) = 2.3 tonnes in 15 minutes or 9,200 kg/hr. Frankly, that estimate cannot be correct. For instance, at LRC at FL350 and +10C the fuel flow at 17:07 at 218 tonnes total weight is only 6,934 with PDA = 1.5%. My best estimate is (43.8 – 42.1) = 1.7 tonnes in 15 minutes, or 6,800 kg/hr averaged between 17:07 and 17:21.
You said: “If the aircraft maintained CI52 and was 6nm south of Penang at 1752:35, where would the aircraft be at 1825:27? Would it be starting its turn at Nilam towards Sanob and exposing its Right High Gain Antenna?”
The ground speed using CI52 during this part of the route is too low to match the military radar track. At FL340 the TAS at CI52 is 483 knots, whereas my best estimate is about 498 knots +/- 5 knots. At M0.84 at FL340, the TAS is 297 knots, and this is what I use. This corresponds to the maximum operating Mach. Wherever they were going, they were in a hurry.
The aircraft location at 18:25:27 depends on the maneuver performed at about 18:22 (a 15 NM right SLOP, or possibly a turn to the right, or possibly a deceleration and a turn. Victor and I are studying this maneuver independently. Whatever the maneuver was, we do know the aircraft was near the 18:25:34 BTO arc (it is more accurately located than the 18:25:27 BTO arc).
You said: “And if it flew Nilam-Sanob-Banda Aceh, and then maintained HDG True, where would it end? Would it be the Bayesian Hotspot?”
I have never been able to fit an acceptable CTH route through Banda Aceh. I have found a good fit passing just west of IGOGU. If you go to my pubs list HERE and click on the link for 9 January 2017 you can find the details. This route ends on the 7th Arc at 34.75S. This location is in the northeastern lobe of the Bayesian probability function.
You said: “Is there fuel remaining at 7th arc? If yes, could APU have been on, resulting in the aircraft being past hotspot? Or the left engine was still operating with inoperative Left GCU and Backup Converter and the aircraft is within 100 NM radius of hotspot?“
Using my CTH route at 257 KCAS at FL350, estimated MEFE is at 00:29. This is a bit too long. Having the APU running might eliminate this issue, but then how do you explain the SDU reboot at 00:18? My understanding is that if the APU were running at MEFE, the SDU would not reboot, and therefore no transmissions would have occurred at 00:19. If I understand this correctly, the SDU reboot means that the APU could not have been running when MEFE occurred. In addition, if a restart of the left engine occurred, it must have been brief (less than several minutes and probably very much shorter than that).
@DrB
1. I agree that the Factual Information report estimate of fuel remaining at Igari (41.5t) is too low. It should be 42.2t.
2. Primary Radar is not very accurate when compared to Secondary Radar, especially at long range. The Primary Radar screen shot shown at Lido Hotel has time stamp of 1813:16 for Vampi. Thus from south of Penang to Vampi is 171 NM in 20.6833 min equals 496 knots groundspeed. Subtract the ~13 knot tailwind and TAS is 483 knots. Project this forward and 1825:27 is start of Left turn at Nilam towards Sanob, at the exact time when the Right High Gain Antenna is finally exposed to the satellite!
3. I get a really good fit with CTH through Banda Aceh from Sanob ending at Bayesian Hotspot.
4. APU already on, then left engine fuel starvation leads to momentarily depowering of Left main AC bus as the left engine spools down (yes this is different to what happens normally) Autopilot remains engaged, aircraft porpoises further south.
5. It is possible that Left GCU and Backup converter are both inoperative, such that Right IDG is supplying all the power to the main AC buses. When right engine failures, APU autostarts, creates log on, but left engine is still operating. Projectory of flight path is greater.
@Paul,
You said: “@Dennis W and path-fitters. Asking my question another way. If there are “equally good” candidate path models (viz BTO error RMS) available as you rotate around 1941, why does the Bayesian pdf have a peak at all?”
@Rob,
You said: “The BTO rings turn out to be particularly amenable to straight line paths, suggesting to me that the plane most probably flew in a straight line after FMT.”
The reason there are multiple solution for straight paths is that the satellite motion during the flight was quite small compared to the aircraft motion. Each “arc” of constant BTO is centered at the sub-satellite point (assuming a spherical Earth). Therefore, if the satellite were stationary, BTO would, in fact, provide zero information on the initial route bearing. There would be, in this case, complete rotational symmetry in the arcs, and only the BFO would provide very coarse information regarding the route bearing. Fortunately, the satellite moved during the flight, displacing the arc centers. Thus the arcs are not “parallel” (more accurately, concentric) to one another. This displacement of arc centers is what allows more precise determination of route bearing. In this particular case, a “straight” path was initially found to fit the BTOs quite well, and that led to initial (high-speed) estimates of endpoints in the 39S area. If you slow down, as needed to avoid excess fuel consumption, the straight solutions rotate counter-clockwise to the points Victor has shown. The errors in BTO are sufficiently large as to accommodate a straight path over a wide range of bearings (about 20 degrees) – and thus we see numerous solutions. The DSTG Bayesian analysis was weighted to favor fewer turns, and thus, its probability peak was at 38S and was quite broad. We also know the aircraft is unlikely to be located there because of the unsuccessful search.
Now, in the case of curved paths, the “eccentricity” (i.e., non-concentricity) of the arcs is much more constraining on the route bearing, with the result that there are very few solutions (perhaps one or two at most, I think). You can visualize this by shifting the centers of the arcs in your mind and noting the strong impact on arc-to-arc separation.
This non-concentricity effect does not imply one or the other solution is more likely to be correct. It does mean that for straight paths, other independent information, such as assumed waypoints, must be supplied in order to generate a precise 7th Arc location. For curved paths, no other assumption is required. The only curved-path solutions identified to date are very narrow indeed along the 7th Arc (a fraction of a degree). In that respect they have the advantage of not requiring an addition assumption regarding bearing.
@DrB said: The aircraft location at 18:25:27 depends on the maneuver performed at about 18:22 (a 15 NM right SLOP, or possibly a turn to the right, or possibly a deceleration and a turn. Victor and I are studying this maneuver independently.
If the aircraft was not following N571 after 18:22:12, it gets harder to predict what occurred. We have to consider changes in speed, track, and vertical speed to match both the BTO and BFO values. And more than one change in track might have occurred. If the speed was constant, the plane turned right to around 320° and was descending at about 600 fpm. If the plane was flying at constant altitude, it turned right to around 307° and decelerated to 372 kn (GS). Similarly, the BFO values at 18:40 are matched with various combinations of track, speed and vertical speed. What occurred between 18:22 and 19:41 is not at all straightforward to determine.
@Paul Onions said: APU already on, then left engine fuel starvation leads to momentarily depowering of Left main AC bus as the left engine spools down (yes this is different to what happens normally) Autopilot remains engaged, aircraft porpoises further south.
You have said many times on this blog and on Twitter that a transfer of power for the left bus from the IDG to the APU causes a momentary loss of power and a reboot of the SATCOM. On the other hand, we know that the SATCOM is designed to ride through this interruption of power. Your evidence seems to be PMDG777 simulations you have run. I have explained that you can not used MSFS to study transients that occur over a sub-second time interval.
Do you have any new evidence which supports your assertion?
This is a flight path further North that has a pilot in control til the end or at least til late in the flight
MH370 reached cruise altitude on route to Beijing, turned back over Malaysia and proceeded up the Strait of Malacca.
Say passed NILAM on way out into the Andaman sea.
Turned back towards Banda Aceh.
At say NOPEK turned towards BEDAX.
At BEDAX a decision was made to go South in a loiter off the coast of Sumatra.
Turned towards ISBIX.
Crossed the 19:41 BTO at 4°22N on a heading of 181°.
Reduced speed slowing through 19:41 to 370 kts and maintained this.
The speed at 19:41 was approximately 410 – 420 kts.
At ISBIX turned towards Cocos Islands.
Used a FBT at PCCNG onto a heading for Christmas Island.
MH370 could have ditched in the area from 8°S to 10°S near the 7th arc.
There was a possibility of disruption on the flight deck from near Cocos Islands onwards.
This could have meant there was no turn at Cocos Islands leading to overflight and continuing to a ditching further South as far as 22°S near the 7th arc.
From 19:41 onwards the BTO’s accurately fit right through to the ditching and the BFO’s are a reasonable fit.
In summary, MH370 could have ditched in the vicinity of the 7th arc anywhere from 8°S to 22°S with a more likely range for the ditching 8.5°S to 9.5°S.
@Victor,
Re: “You might be interested in this new article that (in a very simplified way) discusses the data collected by the Cape Leeuwin hydrophone station as it relates to MH370. Neither of the two relevant signals is (statistically) close to the 7th arc.”
Thanks for sharing this link. As I see @ventus45 shared full version, but I did not dive into details yet.
First impression is that E1 area is surprisingly close to the 7th arc, so frankly I don’t understand your comment that it is not. Just assume that the error in determining distance is twice higher than the authors claim.
Secondly, it is indeed close to the point I indicated around 1.5 years ago based on the intersection of the 7th arc with the ‘curve’ of possible sources based on HA01 and RCS data (for example, see my TN-ATT-Rev1.0 note).
Thirdly, Crivelli’s paper nearly confirms that Alec Duncan’s hypothesis about a possible origin near Maldives was wrong. Initially I stated this based on my analysis of Duncan’s plots, but at that time I faced groundless resistance.
Fourthly, water temperature affects sound speed, and this could contribute to a larger error (one would need to accurately take into consideration 3D temperature to eliminate errors associated with this effect).
Fifthly, waves generated by a seabed impact may also propagate in the seabed, not only in the water. This may affect the overall ‘picture’.
@TimR,
Re: “In summary, MH370 could have ditched in the vicinity of the 7th arc anywhere from 8°S to 22°S with a more likely range for the ditching 8.5°S to 9.5°S.”
No, it could not. 17°S is a firm threshold based on a number of drift studies and the barnacle forensics. More realistic multi-criterion northern boundary is 25°S.
@Oleksandr: The best estimate of the impact is about 535 km from the arc. That is well beyond range error.
On the other hand, if the range is wrong, the bearing might still be correct. Extending a path from HA01 over the range of angles of 301.6 +/- 0.75, and the acoustic path crosses the arc very close to the NZPG path. A plane turning left after the final satellite transmissions would put the impact site closer to the middle of that range.
This could be very interesting evidence.
@Victor,
Re: “The best estimate of the impact is about 535 km from the arc. That is well beyond range error”
My estimates show approximately 350 km from the center of E1 area (24.9S, 98.1E) to the 7th arc (26.87S, 100.87S) based on the plot in Crivelli’s paper. How did you get 535 km?
While I do not agree with underlying hypotheses leading to the suggested areas, I believe the crash site might be in one of them. Here are my thoughts with regard to the search priority:
(1) ‘NZPG’ zone: define a segment of the arc at its intersection with the acoustic signal ‘curve’.
(2) ‘YWKS’ zone: I think the whole segment from 28 to 30.5S is ‘suspicious’, but it is a lot larger compared to the ‘NZPG’ zone.
(3) ‘CSIRO’ zone: put it on a shelf.
(4) ‘NZSP’ zone: put it in a dust bin.
@Oleksandr “…….17°S is a firm threshold based on a number of drift studies and the barnacle forensics. More realistic multi-criterion northern boundary is 25°S.”
According to an article in the German newspaper Kieler Nachrichten, scientists from the GEOMAR-Helmholtz Institute for Ocean Research in Kiel completed a detailed drift analysis of their own in collaboration with colleagues in Great Britain
This study was done soon after the flapperon was found.
“We have traced back almost two million ‘virtual’ particles over a period of 16 months and 95% of all the particles tracked pass just below Java.”
These were unbiased calculations provided by qualified people without overt pressure being applied to comply with the official doctrine of a ghost flight into the SIO.
http://www.geomar.de/index.php?id=4&no_cache=1&tx_ttnews%5btt_news%5d=3972&tx_ttnews%5bbackPid%5d=185&L=1
@TimR,
Re: “According to an article in the German newspaper Kieler Nachrichten, scientists from the GEOMAR-Helmholtz Institute for Ocean Research…”
GEOMAR did not consider wind in their study. I believe this is a primarily reason why their result is dramatically different from the results of a dozen of other drift studies. If you wish, you can compare it with MeteoFrance, which is also a reverse drift study, and it includes wind/no wind plots.
@Oleksandr Here is another drift study that shows a high probability that debris originated up towards Java
Météo France, the French national meteorological service.
After running hundreds of thousands of simulated drift trials under varying assumptions, Pierre Daniel, the author of the Météo France study concluded that if the piece floated as its Lepas population suggests, that is to say submerged, then it couldn’t have started anywhere near the current seabed search area. Its most likely point of origin would have been close to the equator, near Java at latitudes from 9S to 23S.
http://031c074.netsolhost.com/WordPress/wp-content/uploads/2016/05/Pierre-Daniel-report.pdf
@TimR,
Have you read my comment above? What does Danil’s report show for the scenario where wind was included?
@Victor, Dr B, Dennis W.
There are several reasons that I err towards an early turn, essentially over Banda Aceh. One of these is that if you fit a third order polynomial to the BTOs ring 2-7 or even 2-6, and use this to predict “what would BTO have been at 1825, 1828 if it was produced by the same pattern as produced 2-6”. The answer is that it predicts a value very close (within about 100 microsecs) of values observed 1825-1828. Even closer when you strip out the delta satellite-Perth element. To me, this suggests to me that if 2-6 was a “straight course” then that straight course should lead back nearly to 1828, not a long ways west of it.
Does that mean anything to you, or just serendipity?
:
1) There’s a plausible accident-scenario reason for doing so [if, like me, you are wedded to that line of thought]
2)
\
@Paul Onions,
You said: “The Primary Radar screen shot shown at Lido Hotel has time stamp of 1813:16 for Vampi.”
I don’t know where you get that time stamp. I have looked long and hard at the Lido images, and all the labels near VAMPI are unreadable to my eyes.
You said: “I get a really good fit with CTH through Banda Aceh from Sanob ending at Bayesian Hotspot.”
Please post the details so others can evaluate it.
Your point about the right engine IDG possibly powering the left AC Bus at right engine shutdown is interesting, but I am not knowledgeable in this area. Perhaps Andrew and Victor might provide their opinions on this scenario. If something like that happened, and if there was sufficient fuel to continue operating the left engine for about 7 minutes longer, as expected, then why did the IFE not log on at 00:21 since the aircraft should still have been in controlled flight and the APU should still have been running then?
oops. sorry for the orphans at the bottom. incomplete edit again. past my bedtime.
@Paul S
I reproduce your third order predictive fit, but I am struggling to understand what that means relative to a flight path. I need to noodle that some more I guess.
@DrB
You can see the time stamps on Victor’s earlier blog article about Lido
http://mh370.radiantphysics.com/wp-content/uploads/2017/03/LidoImageTimestamps.png
@TBill
I believe that was added to the graphic by a third party, and not included in the slide shown at the Lido Hotel.
@PaulS
“There are several reasons that I err towards an early turn, essentially over Banda Aceh. One of these is that if you fit a third order polynomial to the BTOs ring 2-7 or even 2-6, and use this to predict…”
…Paul what are your other reasons? in case that one does not work out? I await the responses otherwise you or Dennis have to show me the calcs. It has been suggested BFO would be a straight line for straight path, but I do not agree with that either. I think you have a certain trend incoming towards the satellite, and then moving away from it (after about 19:50).
@ Paul
I don’t think the 3rd order polynomial argument is solid proof of a straight flight from 18:xx. The same polynomial can also show that the derived BTO at IGARI would also fit quite nicely, and we know pretty conclusively that the path wasn’t straight from there.
I remember discussing the polynomial question with you a couple of years ago in relation to some other potential flight paths.
On the otherhand constructing the 3rd order polynomial in this manner does illustrate a couple of other points, i.e. one can derive the time of closest approach to the sub-satellite position, and then with a bit of spherical geometry one can estimate the average ground speed between 19:41 and 20:41 [assuming the path is relatively straight].
It is also possible, in my view, to reverse engineer the BTO for a hypothetical position at some time earlier than 19:41, say 19:01 [but not as early as 18:28] and then assuming this position to also sit on a relatively straight path, determine it’s location. One is left then with a lesser time interval to ponder over, i.e. the time between 18:28 and say 19:01, and to try various track and path options that will fit the geometry.
Of course, we don’t know that the path was indeed straight from 19:01, but it does provide an interesting possible point to help fill in the time gap.
@Brian. I’m not still banging that old drum.
In this case, I’m pointing out that backward extrapolation of a low order polynomial curve fitted to the BTO values 1941 onwards turns out to be a very good predictor of BTOs at 1825-1828. These are most certainly not the same as BTOs at position IGARI or even straight path from IGARI southward.
@VictorI
Assuming one of those routes was taken (NZPG, NZSP, YWKS) those locations had to be deliberately and purposefully selected by a pilot who knew there was not by far enough fuel to reach those destinations.
It makes no sence imo a pilot makes such a decision without a specific reachable destination in mind.
This destination was maybe somewhere near along one of those great circle routes within fuel range.
But if he selected one of your routes/waypoints he could as well have selected a direct route/waypoint to/near the destination he had in mind.
This makes more sence imo.
Why select NZPG, NZSP or YWKS if you know you cann’t reach them anyway?
After all the planning and effort/risk at- and after IGARI?
Imo if you take those three routes seriously you have to assume the flight was planned from beginning till end. Am I right here?
And imo if so, he must have had a specific end-destination in mind also.
Maybe along one of those routes. But more probably imo a selected waypoint near one of those routes he knew he could reach with the fuel on board.
Just flying to NZPG, NZSP, YWKS till fuel exhaustion makes no sence in a well planned scenario which it shows all the signs of imo.
@Ge Rijn: If the pilot wanted to remain in LNAV mode following a great circle without reaching a route discontinuity, he would need to select a final waypoint that was with certainty beyond the point of fuel exhaustion. NZPG, NZSP, and YWKS are possibilities. The sim data is consistent with NZPG as the destination. If traveling as far south as possible (after a delay) was the objective, then NZSP is a viable candidate. We don’t know what the intention was for the flight south, nor how he implemented that intention. I think flying towards a distant waypoint remains a viable candidate. I don’t dismiss other possibilities.
@Paul Smithson: The BTOs might follow a progression back to 18:25, but the BFOs do not. To consider a path with no major turns or changes in speed starting before 18:25 requires us to reject the radar data so that the position is much further north on the arc at 18:25 and traveling to the south. We also need to impose a climb at 18:28 to match the BFO. Considering that there are (many) paths matching the satellite data that are also consistent with the radar data, it is hard to justify rejecting the radar data.
It is noteworthy that the progression in BTO you see starting at 18:25 is related to the observation that the satellite declination (latitude) and the BTO seem to be related, i.e., the minimum of one is close to the maximum of the other. I’ve thought a lot about this, and I have no explanation other than it’s a coincidence. I know you’ve also studied this in the past.
Let me make a more generalized statement about coincidences. We’ve all studied the sparse data in so much detail that we recognize patterns that we often use to support particular scenarios. Some examples are the alignment of sim data points with McMurdo Station; the relationship of satellite declination to the BTO; and the identification of satellite images that support a particular drift model. Since flight paths that make use of these patterns suggest different impact points, we know they can’t all be correct. The challenge is to identify which patterns are likely to be correlated and not just coincidental. On this matter, there is considerable debate.
@VictorI
I also think flying towards a distant waypoint like NZPG (or the others) remain viable candidates but only because one of them brought him near to his specific destination before final fuel exhaustion.
Entering a direct waypoint to a near specific destination within fuel limits along those routes makes more sence imo.
There would be no route-discontinuity till after this final waypoint.
To me it just makes no sence why a pilot would enter NZPG, NZSP or YWKS and then wait for hours till fuel exhaustion and the plane to crash considering all the other information and what happened before ~18:40.
Imo there must have been a specific end-point/area if NZPG, NZSP, YWKS or another route was chosen.
@Ge Rijn
To me it just makes no sence why a pilot would enter NZPG, NZSP or YWKS and then wait for hours till fuel exhaustion and the plane to crash considering all the other information and what happened before ~18:40.
I am with Victor on this issue. Since the terminus on the 7th arc coincides with fuel exhaustion, I don’t see how it could have been a premeditated destination. The fuel models developed by DrB, Victor, Boeing, and others are not at all trivial. I doubt Shah could have predicted fuel exhaustion with sufficient accuracy for it to be aligned with a particular 7th arc location.
@Victor
It’s worth noting, you would have noticed, that MH150 scheduled takeoff time is mid afternoon local time. And would there be a 3 man flight crew for an 8.5 hour flight?
@Ge Rijn
@DennisW
re. “to me it just makes no sence why a pilot would enter NZPG, NZSP or YWKS and then wait for hours till fuel exhaustion and the plane to crash considering all the other information and what happened before ~18”
Dennis, I’m also with you and Victor on this issue.
@DennisW
If all planned he knew the amount of fuel he had available.
If too much before reaching his destination he could have loitered for a while to burn fuel (which is common to do before an expected crash landing) or jettison fuel at the end to avoid possible explosion and fire on impact.
And detection from space and a large debris field.
If Shah was at the controls he would have known exactly what his range was.
The 7th arc is only the outcome not an intention.
@Ge Rijn
The 7th arc is only the outcome not an intention.
I agree with that. I don’t think that there was any intended terminus.
@DennisW
You point out the difference in thoughts quite well.
The intended terminus defined the 7th arc outcome imo.
Not the other way around which makes no sence.
@Ge Rijn
I am relatively comfortable with the following:
1> Shah diverted the plane for a specific reason (not suicide).
2> There was a loiter West of Penang related to 1> above.
3> The specific reason did not work out.
4> Shah pointed the plane South per the rehearsal on his simulator.
5> The plane ran out of fuel and plunged into the ocean per Exner.
6> The surface search was started after the debris dispersed/sank.
@Ge Rijn: You are right. It makes no sense. Antarctica was not his destination.Without an intended terminus setting a true track would have been easier.
@Ge Rijn @DennisW @Rob
I agree partially with Ge Rijn. For example, instead of SPOLE, the pilot could enter a intermediate waypoint such as 3294S. This tells you if you have enough fuel to get over the Broken Ridge. No reason a pilot could not switch to oceanic waypoints temporarily to get fuel estimates. One hypothetical purpose for Z’s simulation work could be to get a feel for manual waypoints.
I personally use the shareware add-on “Flight Sim Commander”, which puts my FS9 flight path on Google Earth as I fly, so I can target features such as Broken Ridge. Then I can shift over to SkyVector once I see what I am doing.
The reason I do the Google Earth overlay is that, early on I adopted someone else’s hypothesis that the flight logic may have been to ditch in a deep. mountainous hard-to-find location.
If it was “Z did it”, I see at least two explanations:
(1) Suicide with intent to decoy on flight path and hide crash site, in which case the simulator may give clues as to the out-and-back decoy plan
(2) Per Dennis some kind of plot, in which case a hold pattern makes sense
In reality I mostly go with Option#1 as I have adopted Ewan Wilson’s book Goodnight Malaysian 370 as my hypothesis, and I merged that hypothesis with another book saying hiding in the undersea mountains was a possible strategy.
@DennisW:
RE “The plane ran out of fuel and plunged into the ocean per Exner.”
Why did he rehearse fuel exhaustion twice on his simulator?
@Rob
“It’s worth noting, you would have noticed, that MH150 scheduled takeoff time is mid afternoon local time. And would there be a 3 man flight crew for an 8.5 hour flight?”
Yes probably 3 pilots to Jeddah. You know they added 2nd later Jeddah flight, right? that one was at nite, but pretty sure it would been several hours earlier than MH370 getting up Malacca Straits, so radars might have been more active.
@TBill
No, I didn’t realise it was a night flight, and Victor was too polite to embarrass me by pointing it out. Three man crew would have made it much more difficult, if not impossible for Shah to hijack MH150 in the way he hijacked MH370.
Which prompts the question “what was the object of the exercise, choosing MH150 to simulate two days before he sat in the left hand seat on the real thing?” What aspect of the SIM exercise did he want to check out or verify on an actual flight? If no one can come up with a persuasive operational reason for why he needed to do it, then we are left with “to lay a false trail” as the most probable.
This is a very important issue as far as the search is concerned. If the object of the exercise was to lay a false trail, then it would give us an insight into Shah’s thinking as he formulated the plan. To be more specific: 1) it would suggest he assumed primary radar would reveal his route up the Malacca Straight, after the event (ie when the military got round to checking the radar). 2) He was planning to let the outside world know that he flew until fuel exhaustion with the 2nd logon event. 3) He knew we would probably deduce that he had flown into a remote part of the SIO.
@TBill
Continued: The idea was to give the impression that he had originally planned to hijack MH150, but during the flight, was thwarted for some reason. He subsequently succeeded with MH370
@TBill
Continued (2) The idea was to give the impression that he had originally planned to hijack MH150, but during the flight, was thwarted for some reason. When he succeeded on the second try, with MH370, the simulator would show us where he was intending to fly.
@Oleksandr
Why did he rehearse fuel exhaustion twice on his simulator?
The first simulation was to see how he would react to a nose wheel explosion.
@TBill
In reality I mostly go with Option#1 as I have adopted Ewan Wilson’s book Goodnight Malaysian 370
Hard to put “reality” and “Option#1” in the same sentence.
Wilson’s book was written and published before the flaperon was found, and before the information on Shah’s simulator was generally known. Wilson’s book is generally regarded as cheap shot to make a few bucks.
A suicide scenario cannot be reconciled with a loiter. The loiter is needed to explain the ISAT data between 18:25 and 19:41 as well as the absence of wreckage in the primary search area.
@DennisW
Talk about cheap shots. The only new evidence since Sept_2014, that was not discussed in Wilson’s book is yes the debris, which is consistent with his scenario.
“A suicide scenario cannot be reconciled with a loiter.”
Sure it can. Turning left at Banda Aceh is a very strange flight path, does not seem very likely to me. You are campaigning for your scenario. It would however be interesting to hear from Ewan Wilson if his views have changed.
@TBill,
You are campaigning for your scenario.
Definitely not. I could actually not care less what anyone thinks about my suggestions relative to MH370.
@Victor,
Earlier I asked you: “How did you get 535 km?”.
Either I am missing something, or there is a mistake in the full paper, link to which was posted by @ventus45.
Extract from the full paper:
“E1: 301.4 ± 0.4°, 1900 ± 200 km from HA01, centred at −23.662°, 96.676°”
Now compare it with Figure 4 in the same paper. The center of E1 zone is clearly around 25S, not 23.6S. Furthermore, the distance from HA01 to −23.662°, 96.676° is 2100 km, not 1900 km. It appears that the figure is consistent with the claimed distance of 1900 km, but the reported center of E1 area is not.
What do you think?
@Oleksandr
I also could not reconcile the propagation times in the paper with the speed of sound in salt water. Of course, the time of the E1 event is not compatible with the termination of the flight.
@Oleksandr: I averaged the coordinates of the HA1 sensors to get -34.891540,114.14130, and used the bearing of 301.6 +/- 0.75 to determine the 7th arc crossings. I did not use any data from the figure.
@Rob
Hard to say. It could be argued the simulator cases were indeed intended for a Jeddah flight plan, and so that is why the plan was later modified for the MH370 flight.
@Dennis,
Re: “The first simulation was to see how he would react to a nose wheel explosion.”
I think you confused me with Gysbreght, who asked relevant question.
But generally I think your list is a total mess:
1> Shah diverted the plane for a specific reason (not suicide).
Ok as an assumption.
2> There was a loiter West of Penang related to 1> above.
Here nonsense starts. What kind of loiter? What do others do during this time? You mentioned negotiations – by what means? What kind of negotiations? Any political claimants would attempt to attract as much media as possible. Money? Why did he need to go to the Malacca Strait for such a purpose, why take such a risk? At least 4 radars were supposed to track him there; how would he know that all of them would fail? Why did he need to switch off the SDU? Why did he need to switch it on again? Why was the first BFO in 18:25 login sequence correct? Why would he fly manually up to Penang?
3> The specific reason did not work out.
Ok. Then what? He was upset and decided to kill 200+ innocent people?
4> Shah pointed the plane South per the rehearsal on his simulator.
This is equivalent to suicide. Why South? To meet antipode of Santa Claus? Why would he need rehearsal for this?
5> The plane ran out of fuel and plunged into the ocean per Exner.
What is to do with Exner? If a plane runs out of fuel, it plunges, per Sir Isaac Newton.
6> The surface search was started after the debris dispersed/sank.
I am not sure what you tried to say. The flaperon, RR fragment were still floating, as we know. The engines would have sunk. Mangosteen would get rotten. Then what?
@Victor,
Re: “I averaged the coordinates of the HA1 sensors to get -34.891540,114.14130, and used the bearing of 301.6 +/- 0.75 to determine the 7th arc crossings. I did not use any data from the figure.”
So, what center of E1 area did you assume in your estimation? And what are your coordinates of the 7th arc crossing?
But this does not address my latest question with regard to the paper: the authors stated 1900 km from the center of E1 to HA01, while I am getting 2100 km if I use the position indicated by them.
@Oleksandr
3> The specific reason did not work out.
Ok. Then what? He was upset and decided to kill 200+ innocent people?
When a negotiation fails you have to carry out the “what if” in order to establish credibility for future negotiations. Like an employee walking into my office and threatening to quit without a raise. My reaction was always to dial for an exit interview with HR. Plus that, it was fun to fire an engineer.
4> Shah pointed the plane South per the rehearsal on his simulator.
This is equivalent to suicide. Why South? To meet antipode of Santa Claus? Why would he need rehearsal for this?
Not at all equivalent to suicide. Simply making good on the threat.
5> The plane ran out of fuel and plunged into the ocean per Exner.
What is to do with Exner? If a plane runs out of fuel, it plunges, per Sir Isaac Newton.
Agree. Simply recognizing that Exner has been strongly associated with this end of flight physics.
6> The surface search was started after the debris dispersed/sank.
I am not sure what you tried to say. The flaperon, RR fragment were still floating, as we know. The engines would have sunk. Mangosteen would get rotten. Then what?
Simply pointing out that expectation of finding a significant debris field after nine days of elapsed time is not realistic.
@Dennis,
Re: “I also could not reconcile the propagation times in the paper with the speed of sound in salt water. Of course, the time of the E1 event is not compatible with the termination of the flight.”
In this regard I am always implying seabed impact, not surface impact. I have not heard any strong argument against it. Of course E1 is compatible with such an event. Right now the question is whether the center of E1 is 350 km away from the 7th arc, or 530 km.
@Oleksandr
In this regard I am always implying seabed impact, not surface impact. I have not heard any strong argument against it. Of course E1 is compatible with such an event. Right now the question is whether the center of E1 is 350 km away from the 7th arc, or 530 km.
OK. That could well be compatible with the time and propagation speed.
@Dennis,
Re: “When a negotiation fails you have to carry out the “what if” in order to establish credibility for future negotiations. ”
Has he succeeded in establishing credibility for future negotiations?
Re: “Not at all equivalent to suicide”
Is not steering the plane to where no landing place exist equivalent to suicide?
@Oleksandr
Has he succeeded in establishing credibility for future negotiations?
Absolutely, IMO.
Is not steering the plane to where no landing place exist equivalent to suicide?
Of course not. Suicide is a completely different notion in the context of, by way of example, a suicide bomber or a kamikaze pilot. You are sacrificing your life for a cause. Not committing suicide due to some personal emotional issue. You should know better than that. You are word-smithing.
@DennisW If your scenario is true, Z must have known that failure was a very real (likely) outcome. Would he not have left something behind that would establish his intentions to others? He would have known that while the Malaysian officials may be dishonest, they are not stupid. They would be able to divert blame unless there was something concert to implicate them. As it stands now, they do not seem to be any worse off.
What did Z accomplish? A very bad plan executed by an apparently very careful man?
Except for that consideration, I think your scenario fits the ‘facts’ (radar data, satellite data, debris, behavior of the Malaysians, …). But I suspect there some big pieces of the puzzle missing (and perhaps a few false pieces in the pile).
@Gysbreght,
You said: “@Ge Rijn: You are right. It makes no sense. Antarctica was not his destination.Without an intended terminus setting a true track would have been easier.”
I agree. A magnetic track is even easier to set than a true track.
@Shadynuk
Yes, all well said, IMO.
According to the paper, the hydrophones at HA01 are arranged in a triangular shape spaced by around 2000 m and located at the following coordinates:
Channel 1: latitude −34.89299, longitude 114.15398, depth 1385 m = 4,543 feet
Channel 2: latitude −34.89848, longitude 114.13385, depth 1473 m = 4,831 feet
Channel 3: latitude −34.88316, longitude 114.13608, depth 1419 m = 4,654 feet
My Eye Fitted / Plotted “Centroid” of the array is:
Centroid: latitude -34.82344, longitude 114.14284, depth (google earth)1273 m = 4,176 feet
See Image:
https://www.dropbox.com/s/unu8a32rxr2i4ui/CTBTO%20HA01%20Location.png?dl=0
@ventus45: See my coordinates for HA01 above. Your latitude is incorrect. Perhaps you have a typo.
@TBill: I looked at the case where 45S,104E was entered as the final waypoint, which is which is close to the coordinates for simulator data sets 45S1 and 45S2. The crossing of the 7th arc is around 28.3S. This seems like a candidate path. Perhaps McMurdo was entered as the destination in the simulator session, but the coordinates 45S,104E were extracted from the session and used for MH370.
@Victor
I don’t know how thw wrong latitudes got in there (I was cutting and pasting from a list) but a major boo-boo none the less.
I hope this one is correct !
https://www.dropbox.com/s/0r01qvvyxq59v3i/CTBTO%20HA01%20Location.png?dl=0
@Victor
A lot is being built around the contention that 10N and 45S1 are joined directly despite a number of factors that suggest that that may not be the case. For instance, it appears that none of the other data points are joined directly; 2N to 3N requires a left turn off the runway heading and routes via at least one intermediate waypoint; 3N to 5N routes via an intermediate waypoint and requires a subsequent right turn; 5N to 10N ostensibly routes via at least three different intermediate waypoints and requires something like four separate changes of track.
Further, as you know, the airplane’s heading at 10N is 255.5°. As others have picked up on before, that makes 10N “unusual” in that the airplane’s headings at the preceding data points are consistent with where the airplane appears to go next; 2N is aligned with the runway heading, 3N is aligned with R467 and 5N is aligned with VAMPI. Not so 10N, when you extend 255.5° from 10N it does not take you to 45S1, it takes you pretty much directly to Katunayake/Bandaranaike International Airport at Colombo (VCBI), Sri Lanka.
Accordingly, we might well ask, was VCBI an intermediate point between 10N and 45S1?
As you’re probably aware, VCBI is not only an en-route alternate for traffic crossing the Bay of Bengal, INOP flying time from it together with Chennai (VOMM), India, Medan (WIMM), Indonesia and Phuket (VTSP), Thailand define the ETOPS segments of flights crossing the Bay of Bengal. Interestingly (or otherwise), 10N is very, very close to the WIMM/VCBI Equi-Time Point (nil winds), being 640 nm to VCBI and 648 nm to WIMM. (10N is also quite close to the WIMM/VOMM ETP being about 615 nm from VOMM.) The WIMM/VCBI ETP is an important consideration for ETOPS across the Bay of Bengal; it is referenced in two MAS B777 ETOPS simulator training exercises. Coincidentally (or otherwise) the Captain had been the Type Rating Examiner for ETOPS simulator training (ETPS77S6) on 16 January and was rostered to do so again on 13 February (ETPS77S1). In fact he was rostered on as the Type Rating Examiner for significant block of simulator training (5 sessions over 9 days) commencing on 8 February. That might raise the possibility that he was reviewing simulator training rather than MH150.
Any old how, in looking for any other indications that the airplane may have been heading to VCBI from 10N rather than directly to 45S1, I started looking at the sim data again and stumbled across the “Avionics.0” section. The data labels and values from 2N are as follows:
[Avionics.0]
CommlACtive=122.950
CommlStandby=118.000
Comm2Active=119.900
Comm2Standby=121.700
NavlActive=113.40
NavlStandby=110.60
OBS1=338
Nav2Active=113.40
Nav2Standby=109.60
0BS2=315
Transponder=1200
ADFActive=353
ADF2Active=281
ComTransmit=l
ComReceiveBoth=False
AudioNavlListen=False
AudioNav2Listen=False
AudioMarkerListen=False
AudioDmeListen=False
AudioAdfListen=False
AudioAdf2Listen=False
I’ve inferred that the labels Comms1 and Comms2 probably equate to VHF L and VHF R, Nav1 and Nav2 probably equate to VOR L and VOR R and ADF and ADF2 equate to ADF L and ADF R.
I’m guessing that all of the 2N values are “default” values as they seem to correlate quite nicely with what you’d expect for the airplane’s first flight out of Boeing’s Everett plant at Seattle;
Comms1Active 122.95 Unicom SEA
Comms1Standby 118.00 D-ATIS SEA
Comms2Active 119.90 Seattle Tower
Comms2Standby 121.70 Seattle Ground
Nav1Active 110.30 ??
Nav1Standby 110.60 PAE VOR/DME (PAINE)
Nav2Active 116.80 SEA VOR/DME (SEATTLE)
Nav2Standby 109.60 TCM VOR/DME (McCHORD)
Transponder 1200
ADFActive 353 RNT (RENTON)
ADF2Active 281 HPK (PARKK)/SZ (SEATTLE PARKK)
When you review the settings across the captured data points you find that while the Comms and ADF selections do not change, interestingly (and perhaps instructively), the Nav1Active and Nav2Active settings do, and for the most part, they make “sense” for the airplane’s location, as detailed below:
3N
Nav1Active 114.70 which is the VBA VOR/DME (BATU ARANG)
Nav1Standby 110.60 Default
Nav2Active 109.50 I haven’t been able to identify this one yet
Nav2Standby 109.60 Default
5N
Nav1Active 113.00 MDN VOR/DME (MEDAN)
Nav1Standby 110.60 Default
Nav2Active 116.20 VPG VOR/DME (PENANG)
Nav2Standby 109.60 Default
Then we come to 10N which, bear in mind, sees the airplane pointed at Colombo;
10N
Nav1Active 116.90 PUT VOR/DME (PHUKET)
Nav1Standby 110.60 Default
Nav2Active 114.10 KAT VOR/DME (COLOMBO)
Nav2Standby 109.60 Default
That makes me suspicious that VCBI had been selected as the destination airport on the Arrivals Page (doing so should invoke the Procedure Autotuning function for the Navigation Radio Page and select the arrival airport VOR/DME). However, because virtually nothing related to this bloody mystery is not conjectural, 114.1 is also the frequency for the VPL VOR /DME at Langkawi (WMKL).
45S1 is perhaps equally as curious;
45S1
Nav1Active 113.40 BAC VOR/DME (BANDA ACEH)
Nav1Standby 110.60 Default
Nav2Active 113.40 BAC VOR/DME (BANDA ACEH)
Nav2Standby 109.60 Default
I’m not sure how you get both Nav channels tuned to the same NAVAID but those settings are interesting.
Any old how, if you join all those breadcrumbs (viz 10N – VCBI – WITT – 45S1) you end up with a distance of around 4640 nm, and that distance has got to be a reasonable fit for exhausting the 62,700 kgs of fuel that the airplane had onboard at 10N.
So, a couple of questions arise;
Does the PSS B777 sim program model the real airplane’s FMC VOR autotune function? Or do NAVAIDs need to be tuned manually?
In the course of your and Yves’ recreation of a flight connecting 2N, 3N, 5N, 10N and 45S1/2, did you happen to note/record the Nav1 and Nav2 settings?
@Mick Gilbert
That is very interesting.
I am particularly curious about the 5N settings.
5N
Nav1Active 113.00 MDN VOR/DME (MEDAN)
Nav1Standby 110.60 Default
Nav2Active 116.20 VPG VOR/DME (PENANG)
Nav2Standby 109.60 Default
Why set MEDAN VOR/DME on Nav1 Active, when it is presumably behind you, and you (presumably) have no intention of flying “over” Sumartra in the first place ? It does seem a bit odd, without a “reason”.
Perhaps Victor and Yves’ could recreate the flight connecting 2N, 3N, 5N, 10N and 45S1/2 AGAIN, and SAVE ALL the Nav1 and Nav2 settings ?
@ventus45
The reason that navaids that are off track (and even “behind” you) is relatively straightforward. As has been explained to me, enroute the FMC autotune function will select in-range navaids for best position orientation (it uses the data to compare with the airplane’s calculated position). Where there are a number of navaids to choose from, the FMC will select ones that are separated by about 90° in bearing as that gives you the best resolution of your current position. 5N to MDN is 180° and to VPG is 085°; they are almost ideally located for the purpose of a positional fix.
@ventus45
Apologies, that first sentence should have read “The reason that navaids that are off track (and even “behind” you) are selected is relatively straightforward.”
@Mick. That’s very interesting. Does the frequency 113.4 tell us which radial that is? And does that make sense as a fly-by last auto-tune?
@Paul Smithson
Paul, if there’s radial/directional data buried in there somewhere I can’t spot it. Without understanding how the PSS B777 Professional flight sim program models the airplane’s actual FMC navaid autotuning function it’s difficult to infer much about flybys, overflys, etc. The dual selection of BAC at 45S1 is curious though. If the airplane flew 10N to 45S1 direct with the autotune function operating I would expect to see 114.1 (CC88 COCOS ISLAND) as a tuned value.
In the post above, Victor Iannello said:
“After some email exchanges with the ATSB, I was surprised to learn that the ATSB has additional data values from the recovered flight files that were omitted from the data sets that were provided in the RMP report. In particular, there is a section of the flight files in which the date and time of the simulation session are stored. These data values tie the date of the simulation to February 2, 2014. Also, the time values show that the chronological order of the data sets matches a flight departing KLIA, flying over the Malacca Strait, continuing past the Andaman Islands, turning to the south, and exhausting fuel in the SIO, in the same order that the progressively depleting fuel levels suggest. The time values also indicate that the flight session lasted for about one hour. This confirms that the position and fuel levels were modified during the simulation, just as Yves Guillaume and I had concluded in our paper.”
The distance from KLIA at 2N, via 3N, 5N, 10N to 45S1 is about 4250 NM, about 9 hours of simulated flight time. I don’t think the simulator session that lasted little more than one hour involved simulated flight between those locations. Rather, the airplane was positioned at those locations, fuel levels and other flight parameters were manually changed and .FLT files were saved. About nine minutes per saved .FLT file.
An interesting question is how the player could have known that location 45S1 was close to a great circle path between location 10N and McMurdo. In all files the [Autopilot.0] field indicated the autopilot was not engaged, and the [ATC_Aircraft.0] field indicated 0 waypoints with 0 next waypoints.
@Mick Gilbert: First, regarding the heading at 10N, as you know, the plane was turning as indicated by the 20° bank. What we KNOW is that the plane did not roll out of the turn at that point. Therefore the heading had to be greater than 255.5°. I have no idea why people continue to extend the path in this direction and expect it to have meaning.
I would be amazed if the pilot was adjusting the navigation frequencies. He would have to go to a screen in the CDU and make changes. There’s no reason to do this. As people here have explained many times, navigation occurs using three sources: GPS, inertial, and nav radio. At any one time, whichever is considered most accurate is used, which is typically GPS. Radials are NOT flown, for instance. And it doesn’t matter if the nav aid is behind or ahead of the flight path.
I would be very surprised if the nav radio frequencies selected by the FMC are related to a future waypoint. They are selected by the proximity to available nav aids, not the destination selected. And I am not even sure the values you report have any significance because there might be other values in the PSS section of the flight files (not part of the fragments we have) that are the true indications of what nav aids are autotuned. I’d have to check to be sure.
Yes, I have saved the data sets when I recreated the sim flight. At some point, I might go back and check, but I see no indications that links anything to Colombo. And if the plane was moved ahead along the path the way I suspect, the nav aid might not reflect the current position, but where it was before the move.
@Mick Gilbert
@Victor Iannello
Thanks for explaining the VOR/DME selection criteria.
@Gysbreght said: Rather, the airplane was positioned at those locations, fuel levels and other flight parameters were manually changed and .FLT files were saved.
Yes, I have been saying this for a year.
An interesting question is how the player could have known that location 45S1 was close to a great circle path between location 10N and McMurdo.
As explained multiple times before, the path from the flight plan appears in the MAP window. The user would simply move the icon forward along the path.
In all files the [Autopilot.0] field indicated the autopilot was not engaged, and the [ATC_Aircraft.0] field indicated 0 waypoints with 0 next waypoints.
As explained multiple times before, those values in the Autopilot.0 section of the flight file are meaningless as they are superseded by values in the PSS section of the flight file, which we don’t have.
@Mick Gilbert said: If the airplane flew 10N to 45S1 direct with the autotune function operating I would expect to see 114.1 (CC88 COCOS ISLAND) as a tuned value.
Not if the plane was close to BAC before it was moved forward on the path to 45S1.
@Victor Iannello: “As explained multiple times before, those values in the Autopilot.0 section of the flight file are meaningless as they are superseded by values in the PSS section of the flight file, which we don’t have.”
The ATSB says it has received four complete and two partial data captures. Are the ‘complete’ captures then incomplete .FLT files?
@Gysbreght: I don’t know whether or not the flight files the ATSB have are as truly complete as they believe. Either way, I’d like to get what they have, as they have more information than what is available publicly.
@Victor
Well I guess that it’s just one more of those uncanny MH370 coincidences that at 10N the airplane is pointing with uncanny precision at Colombo and has Nav2 tuned to the KAT VOR at Colombo!
Re: “I would be very surprised if the nav radio frequencies selected by the FMC are related to a future waypoint. They are selected by the proximity to available nav aids, not the destination selected.”
Then prepare to be very surprised, I guess. Short response; it depends. Long response, per the FCOM:
“FMC autotunes VORs for procedure flying and route operations. The FMC also tunes related DMEs. The tuning status displays are:
• P (procedure autotuning) – FMC selects navaids for approach or departure procedure guidance.
• R (route autotuning) – FMC selects navaids on the active route. The navaid is the previous VOR or a downpath VOR within 250nm of aircraft position.
• A (autotuning) – FMC selects a navaid for best position orientation.
• M (manual) – VOR is manual–tuned. Manual–tuning takes priority over FMC autotuning. Deletion of a manual–tuned frequency returns system to autotuning.”
That’s the real world. The question remains, how does the PSS B777 Professional flight sim program model that?
Re: “ And if the plane was moved ahead along the path the way I suspect, the nav aid might not reflect the current position, but where it was before the move.”
My thoughts exactly. In fact I think you can get a sense of how the simulator session may have played out by looking at the navaid settings and the fuel quantities. If the 10N navaids are in fact PUT and VPL (rather than KAT) then I suspect that the airplane was somewhere between TASEK and NILAM when it was dragged to 10N. If it had been west of NILAM then VPL would most likely have been dropped and replaced by BAC due to the range. When you then consider that only 1,740 kg of fuel was burned between 5N and 10N, the airplane probably wasn’t even all the to VAMPI before it was dragged.
More broadly, there is this persistent and mistaken belief that everything from one simulator session is somehow meant to be representative of one flight. That is not how simulators are used in commercial aviation. One simulator session typically incorporates a number of different exercises that are most assuredly not representative of one flight. For example, in just one of the sim sessions that the Captain examined on there is an engine malfunction on start-up, an engine failure on take-off, an engine restart, a TCAS alert during the climb out, a stall, and a continued climb to cruise followed by an emergency descent … and that’s just the first half of the session! That’s not meant to be representative of a one real flight.
The fact that the recovered data shows that the airplane was being dragged from point to point (which means that fuel burn is ignored) rather than “flown”, either in real time or by using time acceleration, is a fair indicator that we are not looking at one complete flight but rather something that may have been a series of exercises. If nothing else, the fact that the airplane was dragged means that the sim session was essentially useless as a fuel planning exercise.
@Victor
Just by the bye, at 10N you’ve got an airplane with a gross weight of 242,586 kg at FL400 in a 20° banked turn! Is that even possible? 242.6 tonne on a -200LR (-300ER) wing at FL400 strikes me as problematic in level flight leave alone in a 20° AOB turn. Did you replicate those conditions on the recreated flight?
@Mick Gilbert: I did some quick checking using the FLT files I saved from the simulator recreation, and also I just performed some other tests. The recreation was performed using FS9/PSS777 many months ago. I am hesitant to re-run tests using the PSS777 because I am now much more familiar with the FSX/PMDG777 model, but if necessary I could, or @TBill can.
First, the navigational frequencies in the saved file for 5N correspond to the values in the sim file 5N. Recall, we know that the position was not changed prior to the save for this data set, so that’s good.
Secondly, when I use the PMDG777 to fly the sim coordinates, as VAMPI is approached, the VORs (L and R) switch from VPL (114.10) to PUT (116.90). That indicates to me that before the user moved the plane to 10N, it was near VAMPI, as the DME is around 130 NM for both the VORs. There is nothing to suggest that the frequency is related to a navigational aid near COLOMBO.
@Mick Gilbert: Just by the bye, at 10N you’ve got an airplane with a gross weight of 242,586 kg at FL400 in a 20° banked turn! Is that even possible?
I just did it in the PMDG777 at M0.84. Unfortunately, I didn’t record the turn when I ran the PSS777 flight. But remember, we don’t know whether the altitude was changed along with the position just before the save at 10N.
In any event, your work on the radio frequencies was interesting, but it confirms what we already knew: the plane was moved along the path before the save. There is no evidence of a detour to Colombo.
@Mick @Gysbreght
FS9/PSS777 has no problem making those turns…I probably did it hundred times. The main implication I see are:
(1) that 10N turn probably did not start exactly on the LAGOG to DOTEN line…probably the aircraft was moved by hand to that area *or* the true path was more like DUBTA to DOTEN, whereas DUBTA is definitely a common waypoint on the Jeddah route.
(2) Given the N10 turn was 255 degrees, then 10N is not the exact upper point of the trend line to NZPG. The exact upper point of the trend like to NZPG, when the aircraft comes out of the turn, could be more like 10.5N, 90E. This is turn could suggest NZIR or NOBEY as alternate end points.
(2) FS9 is very interesting and somewhat unpredictable on how it makes turns, so one would have to hit on exactly the way ZS did it. FS9 is repeatable, but things like saving the case can apparently change how it does the upcoming turn. Of course simulation speed 1x,2x,4x,or 8x also impacts the curve.
(4) I know nothing about radio frequencies, so that is something Victor or someone has to help on. Tell you what, Victor and Yves did a quite thorough job…I have trouble seeing anything they missed, and I am trying pretty hard to gleen more out of it.
@Gysbreght
You should see from the prior FS9 screen shots I provided that the flight path route is the thick white line. I have recently done a take-off with Victors McMurdo route so I could see the forward path to NZPG, and I made the flight path peak at 10N so I could see when the turn hit 10N exactly. Previous to that I had the Jeddah route highlighted, which was a less useful guide.
Wow if ATSB has complete file, then they have a ton of missing data such as waypoints, time/date of flight and basically could answer all of the questions we are struggling with.
@Mick Gilbert: I explained that the heading is meaningless during a turn. In fact, the one thing we know is that it was NOT heading for 255.5° after roll-out of the turn. I also explained that the navigational frequency is related to Langkawi, not Colombo, which I repeated by flying near VAMPI. The NAV2 frequency indicates the active VOR, which was Langkawi just before VAMPI.(The procedure frequency is autotuned as the airport is approached.) If you want to ignore these facts and believe that the data somehow lines up with Colombo, I won’t try to persuade you further.
@Ocean Infinity followers
For the first time (to my knowledge) a fee has been publicly divulged.
http://www.thesundaily.my/news/2017/10/26/ocean-infinitys-offer-search-mh370-still-being-studied
As I suspected, the fee is not fixed (as initial reports would lead one to believe), but apparently activity based in some manner.
@TBill:
Thank you for reminding me of the files you sent back in November 2016. Yes, I see the white line on the MAP page.
At the time you also sent some .FLT files. Interesting that the [ATC_Aircraft.0] field indicates 6 waypoints and WaypointNext=1.
@Dennis,
Re: “As I suspected, the fee is not fixed”.
I had little doubts that negotiations were about the structure of a fee, but not about definition of ‘success’. How to monitor progress and man-hours if the fee is not fixed?
@Victor @Andrew
“… I looked at the case where 45S,104E was entered as the final waypoint, which is which is close to the coordinates for simulator data sets 45S1 and 45S2. The crossing of the 7th arc is around 28.3S. This seems like a candidate path. Perhaps McMurdo was entered as the destination in the simulator session, but the coordinates 45S,104E were extracted from the session and used for MH370.”
…with all my MH370 work, I find I use oceanic waypoints a lot. For example, let’s say instead of SPOLE the pilot used 180S CTH. Now if I put a waypoint in the FMC such as 3294S, and I press the PROG(ress) button on the CDU, it gives me fuel remaining to waypoint 3294S even though I am still using 180S CTH as the A/P flight mode, and not really flying exactly to 3294S. I am curious if that works on real B777?
So Victor I tend to look for oceanic waypoints closer to the end point, such as for example 2698S for the McMurdo case. If I was just trying to key in distant NZPG I would use (eg; in SkyVector) 78S67 for NZPG and 78S69 for NZIR.
@ventus45,
@Victor,
These minor errors in the coordinates of HA01 are of order of 1 km. I was talking about 200 km discrepancy. Where is the center of E1 zone?
@Oleksandr: As I said, I don’t know. It’s too far off the 7th arc to matter.
@Oleksandr
How to monitor progress and man-hours if the fee is not fixed?
I don’t see how it is possible with Ocean Infinity.
@Gysbreght
…and that was before I started using the PSS777 which extends the FLT files data quite a bit. But even if ATSB just had the full FS9 portion of the FLT files, that could be very important.
@TBill: It might be that the ATSB, or whoever else analyzed the sim data, compared the recovered file fragments to the flight files produced by FSX/PMDG777, not realizing that the FS9/PSS777 flight files are longer because of the PSS section. I suspect that they don’t have the full flight files, but that’s just my guess.
Ocean Infinity demonstrates performance at 5200 m, setting a record for multiple AUVs.
More here.
@Dennis, Oleks:
How to monitor progress and man-hours if the fee is not fixed?
I don’t see how it is possible with Ocean Infinity.
Seems ‘effort’ would be scalable to effective drone-hours/area searched.
Or am I missing your joke, Dennis?
@ikr
No joke, Ikr. In the case of Fugro they had imaged terrain as a measure. The scans needed to pass certain quality metrics to qualify. In fact, Fugro was not bashful about commenting on the difficulty. In the case of Ocean Infinity, I do not believe that is the case. No mention has been made of terrain mapping. It may well be that Ocean Infinity has methods other than visual examination to detect debris signatures, and then follow up optically. I really don’t know what their algorithms are. Simply measuring AUV hourly activity would not be a measure of coverage area or quality.
Hopefully, when an award is made we will get the details.
@Victor
” It might be that the ATSB, or whoever else analyzed the sim data, compared the recovered file fragments to the flight files produced by FSX/PMDG777, not realizing that the FS9/PSS777 flight files…”
Have to wonder if they are missing a lot like addon programs for flight planning and Google Earth linkage, magnetic table updates etc. PSS777 waypoint and airport data bases which are easy to edit for waypoints and airports you want to see but that they left out.
@Victor,
Re: “It’s too far off the 7th arc to matter.”
I strongly disagree. 150 km is not too far bearing in mind all the uncertainties.
@Oleksandr
My apprehension is your assumption of an ocean floor event. It is clear from the linked paper that it was based on the assumption of a surface event. Not saying you are wrong. What I am saying is that I don’t know. The sea water propagation speeds associated with the paper do indicate a significant depth.
@Dennis,
The paper suggests this method works for landslides and earthquakes. I did not have time to study it in detail, but if that is true, why not seabed impact?
@Dennis:
OK, I see where you’re coming from. My sense is that if this were a poker game, Malaysia and OIs interest would be divergent. OI is taking a high risk and wants the chance of a lucky hand. But OI, I think, is in here to sell their technology, so they’ll settle for a higher risk, moderate return strategy. As far as I can see, Malaysia wants to save face — if OI scores two weeks in, on their first priority, Malaysia won’t look like chumps. So let them require so many sea-days + so many vehicle hours. And let OI set the quality of coverage to what their models say is rational. Perhaps 99+% in areas that have high possibility, and 95% or less, steaming from one hot spot to another. We should be happy that Fugro didn’t do that — but assuming this is the last best chance [with present technology], I’d say let OI follow the evidence rather than filling out a data field.
As far as quality of
@Ikr,
The problem is not in Fugro, but in their advisers. What would you change: advisers or contractors?
I doubt that OI has technology that is significantly superior to Fugro’s as I have not heard anything about them except self-advertising. Remains to be seen.
@Oleksandr: If the uncertainty in position is so high, then only the bearing matters. That was my point, although I question whether even the bearing provides useful information. People much more knowledgeable than me in this area are diving into this pretty deep.
@Oleksandr: What I thought was self-evident about OI’s technology is it offers speed and economics. The season for searching in the SIO is short, and the cost of seabed scanning is driven by the cost of the host vessel. Six sensor vehicles in the water for a single host vessel should be very advantageous.
@VictorI
The details of the recent press releases are actually fairly nuanced. Cut paste below from a linked article.
link:
https://www.nst.com.my/news/nation/2017/10/295340/search-missing-flight-mh370-may-cost-us70-million
“Ocean Infinity has offered to provide the search missions on a ‘no cure – no pay’ basis (payment depends upon success and the recovery of property),” he said.
Ab Aziz added that the negotiation also touched on other aspects, including payment terms and schedule, such as whether the fee will paid after the actual wreckage of the plane is found.
The negotiation will also require that the wreckage is certified to be MH370 by the plane’s manufacturer, Boeing.
The terms “success depends on recovery” and “whether fee will be paid after the actual wreckage is found” suggest that Malaysia wants to pay on actual recovery of specific items (presumably the recorders).
Likewise third party certification by Boeing will almost certainly require the recovery of something.
Ocean Infinity would seemingly need to insist on being the party responsible for recovery (hence, the higher than anticipated fee of $70M) since their payment would presumably be coupled to a recovery of some sort.
@David
Re: “Do you have access please to a -200 or -300ER parts listing which might specify the internal line diameter? Alternately could I again prevail upon you to ask your tech gurus whether they can confirm the flexible line’s external diameter is 25.4 mm or close to?”
I don’t have access to the IPC. I was told the external diameter of the APU fuel supply line is 1 inch, or 25.4mm. That does not include the shroud.
@TBill
Re: “Now if I put a waypoint in the FMC such as 3294S, and I press the PROG(ress) button on the CDU, it gives me fuel remaining to waypoint 3294S even though I am still using 180S CTH as the A/P flight mode, and not really flying exactly to 3294S. I am curious if that works on real B777?”
If 3294S is the active waypoint, then the FMC fuel prediction will be based on a direct track to that waypoint regardless of the AFDS mode that is engaged. The fuel prediction will only be accurate if the aircraft is actually tracking towards that waypoint. If the aircraft is not tracking towards the waypoint, then the FMC will keep updating the fuel prediction based on the current distance to the waypoint, groundspeed, etc.
@Victor
Thank you for checking those navaid values and how the PMDG sim program models the autotune function. It appears that after 5N the airplane was probably flown, either in real time or under time acceleration, for a distance towards VAMPI that was consistent with the fuel burn between 5N and 10N (1,740 kg or around 15 or so minutes) and then dragged about 500 nm. That suggests that the airplane was possibly dragged to 10N before it had reached VAMPI.
Thank you also for confirming that the sim allows a 20° AOB turn at a gross weight of 240-odd tonnes at FL400. In real life that would be towards the limits of the airplane’s performance (the LRC maximum operating altitude at ISA+10 for the -200LR is about FL380 at that sort of weight).
@Andrew. Thanks for that confirmation.
@Mick
I don’t recall anyone questioning the similar turn at IGARI as far as being too close to aircraft performance limits.
@DennisW: I don’t know the financial details. I am hearing that things continue to move forward in a positive way. I hope that means a signed deal in short order.
@VictorI
I have the same sense relative to a deal. Attorneys are frequently not helpful in getting a deal done. I think only details remain. I don’t sense anything big blocking the way.
@Victor
@Mick Gilbert
Re: “@Mick Gilbert: Just by the bye, at 10N you’ve got an airplane with a gross weight of 242,586 kg at FL400 in a 20° banked turn! Is that even possible?
I just did it in the PMDG777 at M0.84. Unfortunately, I didn’t record the turn when I ran the PSS777 flight. But remember, we don’t know whether the altitude was changed along with the position just before the save at 10N.”
That’s interesting. The -200LR FCOM shows that the LRC maximum operating altitude at ISA+10 is about 37,300 ft at 243T, providing a 0.3g margin (approx 40° AoB) to initial buffet. The FMC MAX altitude displayed on the VNAV page is based on the same margin, so it should show approx FL373. It’s theoretically possible to do a 20° AoB turn at FL400, but the aircraft would be well above the MAX altitude and very close to the buffet. That’s certainly not something anyone would do in everyday operations! As you said, perhaps some of the parameters were changed.
@TBill
Re: “I don’t recall anyone questioning the similar turn at IGARI as far as being too close to aircraft performance limits.”
That’s a bit like comparing apples and oranges. MH370’s weight at IGARI was about 217T. In the -200ER, the LRC maximum operating altitude at 217T and ISA+10 is about 40,400 ft, well above MH370’s initial cruise level of FL350. The aircraft could have done a steep turn and not run into problems due to buffet or lack of thrust!
@Paul Onions.Your 24th Oct, 2:32 pm to Dr B.
“5. It is possible that Left GCU and Backup converter are both inoperative, such that Right IDG is supplying all the power to the main AC buses. When right engine failures, APU autostarts, creates log on, but left engine is still operating. Projectory of flight path is greater.”
The ATSB had this modelled by Boeing (p8, Search and Debris Examination Update). It says, “In an electrical configuration where the loss of engine power from one engine resulted in the loss of autopilot (AP), the aircraft descended in both clockwise and anti-clockwise directions.”
What I believe has not been raised before about this is that loss of autopilot, from AC loss, would result in failure of the AC powered main engine boost pumps for a minute pending the APU coming on line.
During that time if it had been the right engine still operating that would be reliant on fuel supply through its gravity feed.
Were it the left engine still operating (ie with IDG and back up gen off line) that engine would be supplied also initially by gravity. The APU DC pump would not have started to supply the APU during its autostart. That requires both that there is no pressure in the left engine fuel manifold and that the APU selector is at ‘start’ or ‘run’, whereas it would be ‘off’. Without that the APU in starting would need to draw fuel from the left main tank through its (stopped) DC fuel pump there.
The pump would cut in to supply the engine if and when the engine’s fuel manifold pressure dropped past nil*, that is suction overcame pressure due to gravity. In so doing, by happenstance it would then supply both the engine and APU. However the pump’s supply is limited to just 3150 lb/hr (1430 kg/hr) and the engine at power would need to supplement this through gravity feed. The APU draw off, by its gear wheel pump, is likely to increase that engine gravity supply requirement a little (130 lb/hr based on the ATSB’s 2.2 lb/min).
(*most likely there is also an RR engine rpm stipulation but my manuals do not cover that)
With fuel low in both tanks how much air might be sucked**in through the gravity supply by either engine still running is conjectural and whether that might cause that engine to flame out: restart at altitude can be problematic. Were there no relight or it did not last, the engines would shut down at 35% N3. However as to the left the APU pump, if it had started it would not shut down with the engine since a feature it has which will leave it selected is that it is now running.
(**Leaving aside entrained air, which should have been drawn out by then).
Bear in mind that the oddities and complexities of all this are a result of this being reaction to circumstances naturally not foreseen in design and certification: unpiloted fuel exhaustion.
Without pilot or autopilot during this minute and with uncertain engine happenings the aircraft might have rolled, yawed, or pitched, which too might complicate fuel supply; and aircraft trajectory thereby.
There are numerous uncertainties here.
Simulator verisimilitude achieved by Boeing and sufficient?
@TBill
Regarding the turn at IGARI, 220 tonnes (more than 20 tonnes lighter) at FL350 (5,000 feet lower) for the -200ER (as opposed to the -200LR) is well within the performance envelope, even for bank angles well beyond 20°. On the other hand, you may recall that talk of a climb to 45,000 feet at IGARI was quickly hosed down as being beyond the capability of the airplane.
Para 7. I say, “However as to the left the APU pump, if it had started it would not shut down with the engine since a feature it has which will leave it selected is that it is now running”.
Please replace with, “However as to the left, had the APU DC pump started it would not now be shut down with the engine. Since the APU is now running the pump will remain selected”.
@David
“Bear in mind that the oddities and complexities of all this are a result of this being reaction to circumstances naturally not foreseen in design and certification: unpiloted fuel exhaustion.”
Yes, I agree with the above, except that I would have said “piloted fuel exhaustion”.
@Rob. If a pilot were assumed I would have had a lot more to say.
However the final BFOs suggest there would be little effect on search width. These unpiloted possibilities are not as constrained by the BFOs as by simulator outcomes, and since we have little information on the simulations currently they are lesser constraint.
However where the 7th arc log-on was, piloted, is thrown to the winds. Course and speed changes, step climbs blurring endurance and range, flicking switches, re-engaging the autopilot, selecting APU on….
@David
Re: “What I believe has not been raised before about this is that loss of autopilot, from AC loss, would result in failure of the AC powered main engine boost pumps for a minute pending the APU coming on line.
During that time if it had been the right engine still operating that would be reliant on fuel supply through its gravity feed.
Were it the left engine still operating (ie with IDG and back up gen off line) that engine would be supplied also initially by gravity. The APU DC pump would not have started to supply the APU during its autostart. That requires both that there is no pressure in the left engine fuel manifold and that the APU selector is at ‘start’ or ‘run’, whereas it would be ‘off’. Without that the APU in starting would need to draw fuel from the left main tank through its (stopped) DC fuel pump there.”
The APU DC fuel pump should start as soon as AC power is lost and the boost pumps stop. ELMS uses the signals from the AC boost pump pressure switches to determine if there is pressure in the left engine feed manifold. Once ‘no pressure’ is sensed, the P310 EEU sends the DC pump command signal to start the pump and the isolation valve command signal to open the isolation valve. The aft boost pump is located ‘upstream’ of the fuel suction bypass valve.
@Andrew. Thanks for that. However the point that the pump will not be automatically selected on autostart still stands, I take it.
Either way, the APU pump will not necessarily protect the left engine from flame out through air drawn down the gravity feed outlets with thin fuel coverage that supply supplementing the APU DC pump; and of course will not have any effect on susceptibility of the right engine. I remember you saying earlier that the gravity outlets would be bare with quite some fuel remaining, which would increase risk of flame out of these engines during that minute to certain if so.
OK
@David “However where the 7th arc log-on was, piloted, is thrown to the winds. Course and speed changes, step climbs blurring endurance and range, flicking switches, re-engaging the autopilot, selecting APU on….”
David, I quite agree. But, putting myself in the pilot’s shoes (even though not myself a pilot, just an informed bystander) I personally would just keep unnecessary procedures to a minimum. I’ve reached the end of the flight, “I’m confident no one knows where I am, and reasonably confident no one will ever know where I am. I’m going to let the SDU go through a depower- repower- logon cycle to show the folks back home I kept going until the fuel ran out, then I’m just going to glide her down on RAT power, go as far as the potential/kinetic energy store will allow, then get her under the waves with the minimum of fuss”.
@David
Why am I so confident there was a pilot in control? It’s because the debris tells me so.
@Mick Gilbert,@TBill, @Andrew: You are taking this 20° turn too seriously. When the plane was moved to 10N, I suspect the altitude was also increased to 40,000 ft. You are seeing a snapshot before the simulation was re-started. The flight may or may not have been stable. The vertical speed of 3,570 fpm was induced by the change in parameters and re-start, as I’ve explained before. This also is unstable, and will cause the speed to drop because of the limited thrust at this altitude.
As for MH370’s climb to 45,000 ft at IGARI, a zoom and climb could have temporarily reached this altitude, although I doubt this happened.
@all
It’s all still about scenarios without pilot control at the end.
Imo this makes no sence at all still, regarding the whole flight and information we have.
It has brought us and no one else anywhere near the plane.
If OI takes this lead also they will maybe miss their potential goal only based on a ‘ghost scenario’.
This would be very limiting imo. I sure hope they don’t make the same mistakes DSTG/ATSB have made based on those ‘ghost flight’ assumptions.
To me it’s obvious a ditch-like controlled impact must have occured.
The flaperon and outboard flap section speak very clear language among many other pieces.
IFE must have been shut OFF long before impact.
The pilot had complete control over range, fuel, engines, APU, AC/DC power supply at choice.
Ignoring this obvious possibility has again the potential of limiting the OI search effort.
While those ‘ghost scenarios’ all staying close to the 7th arc left nothing.
We’ve seen the results till now.
Nothing. Another approach is needed imo.
@Ge Rijn
I don’t think the failure to find the aircraft has anything at all to do with a misinterpretation of the observables. The solution has been under constrained from the get-go. Where is a clairvoyant when you need one?
@Victor
“@Mick Gilbert,@TBill, @Andrew: You are taking this 20° turn too seriously”
You could be correct, but I know a lot about making turns in FS9 now.
@Ge Rijn
“To me it’s obvious a ditch-like controlled impact must have occurred.”
Apparently Larry Vance has a new book coming out to say that too, if I got that right.
@DennisW
I agree based on the sence it was all about finding the plane based on those observables at that time!
No debris was found yet before 7-2015. And this changed everything.
DSTG/ATSB never adjusted their search effort accordingly while they had plenty of time and information to do so.
Certainly after more debris got found.
To me it’s not about clearvoyance but about reasonable thinking.
The plane has not been found in a 120.000km2 priority area based on the assumption of a ‘ghost flight’, an uncontrolled high speed descent and impact.
So something else must have happened.
@Ge Rijn: Surely you realize after contributing here for many months that there are many areas close to the arc that have not been searched that are allowed by the satellite data. The priority zone was based on certain assumptions in DSTG’s Bayesian analysis that may not be true. A long glide after fuel exhaustion is just one of many possibilities. Many believe the other possibilities are more likely.
As for Larry Vance, the last time he was making statements for the media, he never bothered to explain how a long glide is consistent with the increasingly steep descent that the BFO data suggests. Rather, he said the satellite data cannot resolve altitude. What he neglects to say, either out of ignorance or because he deliberately cherry-picked evidence to promote his theory, is that while the BTO and BFO data are not sensitive to altitude, in fact the BFO is highly sensitive to vertical speed. That’s how we know the plane was in a steep descent.
@TBill
I read the article with Larry Vance. I’m on his views from the get go (and even before soon after the disappearance on a Dutch blog).
But it’s not about having those same views to me at all.
I would find it very disappointing if OI would discard this possibilities of an intended destination and glide towards it.
@DennisW
“I don’t think the failure to find the aircraft has anything at all to do with a misinterpretation of the observables”
So what do you think the failure and the waste of 112.47 million US dollars (Malaysian Govt. estimate, Feb 2017) tax payers money was due to? Was it due to the ATSB wanting or persuading themselves to believe the IG’s unpiloted impact? Was it due to political pressure from on high, the same political pressure that also prevented disclosure of confidential information, on pain of a prison sentence?
More importantly, will Ocean Infinity, presumably a company with shareholders, now be foolish enough to waste a wedge of their own money on a continued misinterpretation of the observables? If Ocean Infinity follows the ATSB’s flawed analyses and draws a blank, it will positively deter any further search efforts.
Will they take a step back and take independent aviation expert advice before committing their hard earned money and reputation? Will they, for example insist on an independent aeronautical engineering analysis of the debris? The usual, tried and tested procedure of an air accident investigation is to assemble the parts (where and if available)on the floor of a hanger, in their respective locations on the airframe and see what can be deduced. I would venture to say that if were done with the MH370 debris, an uncontrolled, high speed, nose down impact would be shown the door. They might be able to see if the ATSB’s story hangs together. Possibly, they might smell a rat.
The Malaysian MOT Safety Investigation Team are due to publish their final report by January next year. I wonder what their conclusions will be? My guess is they will say the condition of the debris does not permit any definite conclusions to be drawn other than: 1) It confirms that MH370 ended up in the Southern Indian Ocean. 2) The impact was evidently of high energy, ruling out a controlled, soft ditching with flaps extended. What a travesty that will be. Will Ocean Infinity be gullible enough to believe them? I sincerely hope not.
@Ge Rijn
If there was an intended place to “ditch” I am not sure how the pilot would have done that, dive-in vs. glide. I tend to feel we have to give priority to Arc7, equal opportunity to inside and outside of Arc7 to a certain error bar, unless new data comes forward to suggest a specific long glide crash location.
@VictorI
We still don’t know if the steep descent continued after the 8 sec. snapshot. Maybe an active pilot pulled out after and set a glide towards a destination intended.
Maybe the 4000ft from 45S2 was a set pull-out limit in the simulator?
It does not have to have been a long glide. It could have been just outside the searched area. In fact it must have been for the plane wasn’t found there.
Just to put a managable constraint on possible search width I would like to advise to OI to take at least 50 miles on each side of the 7th arc.
The final BFO’s only tell a possible snapshot, thats all.
Not enough to draw definite conclusions from imo.
@Rob said: Was it due to the ATSB wanting or persuading themselves to believe the IG’s unpiloted impact?
Let’s get our facts straight. Members of the IG early on suggested that the BFO was consistent with a high speed descent. Later, after more analyses using data not available to the public, Ian Holland (DSTG) and others agreed.
I am not aware of any statement from the IG that says the end-of-flight was unpiloted. I know I have said that there may have been a banked descent with no pilot input, or a nose-down pilot input. Either way, the increasingly steep descent is not consistent with a long glide. I’ve yet to hear a sensible reason for why a pilot would follow a 15,000 fpm descent with a long glide.
@Victor Iannello:
“I’ve yet to hear a sensible reason for why a pilot would follow a 15,000 fpm descent with a long glide.”
Are you serious?
@VictorI
Your comment; ‘Either way, the increasingly steep descent is not consistent with a long glide. I’ve yet to hear a sensible reason for why a pilot would follow a 15,000 fpm descent with a long glide.’
It’s ofcourse hard to give a reason but the plane was not found between ~39S and ~32.7S near the 7th arc where it was expected to be found.
A reason?
There must be one for the plane was not found where it was expected on previous assumptions.
It must have flyed out of that area some way.
@ROB
So what do you think the failure and the waste of 112.47 million US dollars (Malaysian Govt. estimate, Feb 2017) tax payers money was due to?
Starting an underwater search based on the analytics. The decision makers knew or should have known the uncertainties involved.
The analytics themselves have never been shown to be flawed.
@Ge Rijn: I think any possible latitude should be searched to around 25 NM from the arc. As you can see in Figure 3 above, at some latitudes, the arc was only searched to 6.5 NM from the arc, which was not far enough. Perhaps we simply disagree on what that width should be. I think a search to 120 NM is hard to justify.
@Victor
Actually, as I believe ALSM pointed out a while ago, the Arc7 in Fig. 3 is skewed to high altitude, so the actual amount of Arc7 searched inside could even be less than 6.5 nm, could be 2 nm. So I like what I think Richard said was define Arc7 at say 17,500 ft so it’s in the middle of the possible range.
@VictorI
I think a width of search with a radius of ~50 miles is sensible depending on where you start and to include all possibilities.
If you start from (f.i.) ~33S on the 7th arc expanding outwards till ~35S and 32S then ~50 miles would be a managable limit in my view.
Covering ~30.000km2 from around a centered starting point could lead to finding the plane well within those 30.000km2.
Nothing gained, nothing lost regarding a possible glide and ditching.
When the plane gets found near the 7th arc according a high speed, descent scenario it would be splendid all the same.
My concerns are OI rejecting the possibility of a controlled glide after a steep descent too.
@TBill: You have to be a little careful. Yes, the 35K ft arc is about 5 NM outside of the sea level arc. But the possible distance between the last transmission and the impact also reduces if the transmission occurred at lower altitude. You have to consider both these effects.
@Ge Rijn: I agree w/ Victor, who stated: “I’ve yet to hear a sensible reason for why a pilot would follow a 15,000 fpm descent with a long glide.” In fact, I haven’t heard any reason why that might have happened, sensible or not.
Given that:
1. 9M-MRO reached FE ~00:17:30
2. 9M-MRO was descending at a rate of ~5,000 ft/min 2 minutes later at 00:19:29
3. 9M-MRO was descending at a rate of ~15,000 ft/min 8 seconds later at 00:19:37
4. There was no IFE logon at ~00:21 as expected
5. There were no other transmissions received after 00:19:37
…it is reasonable to conclude that:
a. an extremely high rate of descent began within seconds of FE
b. 9M-MRO’s vertical acceleration by 00:19:37 was ~0.68g
c. given a & b, 9M-MRO must have already descended 10-20k ft feet by 00:19:37
d. even if a pilot intentionally caused b, given c, that pilot could not possibly have levelled out and glided more than 30-40 nm. A continuation of the steep descent is far more likely, even if a pilot was in control.
Although 30-40 nm is perhaps a worst case max glide distance assuming a live pilot, the totality of the evidence and prior cases suggests that 9M-MRO is actually much closer to the 7th arc, probably within ±10nm of the “true 7th arc”. The true 7th arc is known to ±5.3 nm (due to BTO noise) and the altitude uncertainty adds about another ±2 nm to the true 7th arc uncertainty, so a reasonable 7th arc width to search is ±18 nm.
Most will recall, but some might wonder why the search width NW of the arc should be the same as the width SE of the arc (~18 nm each). The reason is that all the simulations of an uncontrolled descent indicate that a quasi spiral descent should be expected. In other words, the aircraft tends to descend in a increasingly tight circular track. The heading at 00:19:29 is unknown, but given that FE took place 2 minutes earlier, it is entirely possible that 9M-MRO was already headed east or even north at that time. Thus, both sides of the arc need to be searched with equal areas.
FTR, Larry Vance is an ignorant fool on MH370. His theory has been soundly discredited both by ATSB and numerous outside experts.
Some interesting information on how Ocean Infinity (OI) intend to operate their fleet of AUVs (link below). They will operate in tight formation in a single area, it seems. I had wondered if they would be assigned separate areas (perhaps adjoining) but a closer formation allows the support vessel to remain close to all the AUVs.
https://oceaninfinity.com/ocean-infinity-reaches-5000-meters-with-six-hugin-auvs-operating-simultaneously/
Currently, Seabed Constructor is operating several hundred kilometres to the west of the Azores, where the ocean depth is recorded as greater than 5000m, as would be required for the test being performed (map below). Ship positions are S-AIS data from fleetmon.
https://www.dropbox.com/s/5i3rdqdxn52c1w8/27-10-17-map1.jpg?dl=0
The ship movements from 00UT on the 17th Oct to 00UT in the 21st Oct are shown in the map below. The points where the ship was indicated by the S-AIS data as stopped are shown – the ship has to stop to deploy or pick up an AUV. The pattern of movements and stops suggests the AUVs were deployed in the eastern marked area and then recovered in the western marked area.
https://www.dropbox.com/s/eponnr7w2qyexnv/27-10-17-map2.jpg?dl=0
@ALSM,
Why do you spread false information again:
2. 9M-MRO was descending at a rate of ~5,000 ft/min 2 minutes later at 00:19:29
3. 9M-MRO was descending at a rate of ~15,000 ft/min 8 seconds later at 00:19:37
?
After that you again make conclusions based on your assumptions. That is silly.
Oleksandr: 2 and 3 above are true, and you know it. ATSB agrees with the statements. Victor agrees. Don agrees. Richard agrees. Bobby agrees. In fact, almost everyone that understands the math agrees. Why are you spreading more false statements?
@Richard Cole
Thank you for keeping up detailed AIS tracking with Seabed Constructor. This map shows the ship’s manouevres while working close to the Azores on Sep 10 before moving off north to the N43W27 area in the mid-Atlantic.
BFOs 00:19 cannot be deemed as a reliable indication of RoD until the mystery of the first BFO in 18:22 logon sequence is solved.
@ALSM,
“2 and 3 above are true, and you know it. ATSB agrees with the statements. Victor agrees. Don agrees. Richard agrees. Bobby agrees. In fact, almost everyone that understands the math agrees. Why are you spreading more false statements?”
Yours 2 and 3 might be true, or might not be true. I don’t know. There are several explanations, but not enough data. If you take BFOs 00:19 “as is”, you are dealing with an event of less than 1:1000 probability. Not talking that we are yet to see the simulations, which can reproduce such a high average acceleration. Sorry, I am not buying it regardless what the IG members think.
Oleksandr: And you also believe ATSB’s assessment is wrong? Please!
Oleksandr said: Not talking that we are yet to see the simulations, which can reproduce such a high average acceleration.
I provided you with the video of a simulation that shows how this acceleration might occur. Independent of whether there are some shortcoming in the model at transonic speeds, to say that the acceleration and vertical speeds are not physically possible is just plain false.
Oleksandr: And it’s not just Victor’s desktop sim (which is quite good, BTW). It’s Boeing’s many EOF simulations on their engineering simulator, my EOF trials in the UAL B777-200 Level4 simulator, and ATSB’s assessment of all the available EOF information. (Thanks again to PM for arranging those simulations.)
Regarding the simulator we used…
Manufacturer: Thompson
Model: 700-200
Year of manufacture: June 1997
Level: D
Configuration
AutoThrottle: Under control of FMS in ECON Mode.
CI=50
ZFW (Zero Fuel Weight): 395,000 lbs
Fuel at 1707 (ACARS reported): 96,500 lbs
Fuel at end of flight: 200-800 lbs per tank, depending on scenario
CG: 28%
Altitude: 35,000 (for most of the simulations)
TAS @ IGARI: 471 kts
Wind Velocity at 38S: 57 kts
Met Direction at 38S: 240 degrees (direction from which the wind is blowing)
Heading at 38S: 186 degrees
Track at 38S: 179 degrees
If I were in charge of selecting the search locations, I would hire our host, along with the IG and a few other contributors to this blog. Perhaps they would also have access to additional data.
A question: Others have calculated speed around 498 KTAS over the last 20 min of the radar. As DrB said the PF was in a hurry. I have no reason to dispute this. (Is this speed supported by a standard setting?) However, right around 18:22 the PF reduces the speed for the remainder of the flight. Why was that done:
To meet the BTO?
To meet the BFO?
To meet the Endurance?
Also, around this same time, the AES re-boots.
Could it be that someone other than the PF after IGARI is the PF around MEKAR? That might explain the changes made around 18:22-18:28. A struggle ensues but both become victims to hypoxia after a southern destination is entered.
I see some still use LRC to calculate endurance and/or range. @Andrew has stated that this setting is rarely used. It has to be another setting that also reduces speed as the a/c gets lighter.
@Victor,
“I provided you with the video of a simulation that shows how this acceleration might occur. Independent of whether there are some shortcoming in the model at transonic speeds, to say that the acceleration and vertical speeds are not physically possible is just plain false.”
Have you read my comment carefully? Where did I state that it was physically impossible?
I recall your scenario shows about 15 percents lower average acceleration compared to MH370. I think it could be significant in this context. I would like to see the same or higher accelerations than the one derived from BFO data. How frequently do you hit these extreme values?
Finally, almost everyone outside the IG does not believe in such a remarkable coincidence of the logon and the beginning of a rapid descent.
@ALSM,
As I mentioned, I was unable to derive acceleration from your transcript. Can you show that you achieved the same order of average downward acceleration lasting for 8 seconds as derived from BFO?
At the moment I lean to think that the second last BFO was subjected to the overshot of similar magnitude as demonstrated in Holland’s paper. That eliminates high acceleration, various coincidences, etc. In other words the plane was already falling at high RoD when the logon 00:19 occurred.
@ALSM,
“And you also believe ATSB’s assessment is wrong? Please!”
Yes, the ATSB did many mistakes. Why would I be sure they are correct this time, especially in light of certain remarks they did in their report?
Oleksandr: You stated: “If you take BFOs 00:19 “as is”, you are dealing with an event of less than 1:1000 probability.”
1:1000!!! What? That is another totally false statement. The probability is at least a number greater than 50%, if not 95%.
The probability that the plane was descending is virtually 100%. It was already out of fuel for 2 minutes, right? The probability that the BFO values were accurate to 250 Hz. Thus, it is virtually certain 9M-MRO was accelerating downward as described ad nauseum.
Let’s return to something worth debating.
Oleksandr: You stated: “If you take BFOs 00:19 “as is”, you are dealing with an event of less than 1:1000 probability.”
1:1000!!! What? That is another totally false statement. The probability is at least a number greater than 50%, if not 95%.
The probability that the plane was descending is virtually 100%. It was already out of fuel for 2 minutes, right? The probability that the BFO values were accurate to ±50 Hz is GT 99%. The vertical speed signal in BFO space was GT 250 Hz. Thus, it is virtually certain 9M-MRO was accelerating downward as described ad nauseum.
Let’s return to something worth debating.
@Victor
You are right. The final BFOs do present a problem for the extended glide camp.
I’m close to wearing out the grooves on my favourite record. I’ll turn it over and see if I can get to like the B side.
@ALSM,
“1:1000!!! What? That is another totally false statement. The probability is at least a number greater than 50%, if not 95%.”
This is absurd. Where have you been for so long time? If you assume BTO and BFO independent, and that the abnormalities in the BTO and BFO were caused by unrelated reasons during 00:19 and 18:25 logons, then the estimated probability of such an even would be around 1:3000 if I recall correctly. Pure Inmarsat data.
I see no point to further discuss your erroneous beliefs and your circular logic.
But please don’t present them as a given fact.
@Lauren H,
You asked some good questions.
You said: ” A question: Others have calculated speed around 498 KTAS over the last 20 min of the radar. As DrB said the PF was in a hurry. I have no reason to dispute this. (Is this speed supported by a standard setting?)”
This one I can answer. Mach 0.84 at FL340 is a standard speed setting that matches the radar track speed. CI52 is too slow at this weight. I think it is more likely that M0.84 (Mmo) was used instead of LRC.
You also said: “However, right around 18:22 the PF reduces the speed for the remainder of the flight. Why was that done: To meet the BTO? To meet the BFO? To meet the Endurance?”
Since the air speed was already at maximum, any change in speed would be a reduction. I think if you had asked either pilot about BTO and BFO they would have had no idea what you were talking about, so those don’t make sense. Personally I don’t think endurance factored into it either. One reason does, and there may be others (such as simply restoring a “standard” speed like ECON), and that is to slow down for a Hold. Unfortunately, we don’t know how far the slow-down was in the hour after 18:22; we only know the average speed through 00:17 was reduced. Otherwise the fuel would have been exhausted well before 00:17. It is possible the speed reduction was just to ECON with CI = 52, since that provides sufficient fuel economy to reach 00:17 with power, but it could have been much larger for a period of time (not more than an hour), followed by a higher-speed cruise. I think that there is likely to be a connection between the speed change(s) and the route followed. I would certainly expect them (speed and route) to be compatible, and most likely turns or racetracks occurred in conjunction with (not independent of) speed changes.
You also said: “Also, around this same time, the AES re-boots.”
I don’t think it is coincidence that, within a period of a few minutes of the power restoration to the SDU/AES, it is almost certain that a speed change or a turn (or both) occurred. In my mind, these are correlated elements of a changed strategy.
You also said: “Could it be that someone other than the PF after IGARI is the PF around MEKAR?”
I think the events then (near 18:22) indicate a changed strategy. I suppose that could involve a change in PF.
You then said: “That might explain the changes made around 18:22-18:28. A struggle ensues but both become victims to hypoxia after a southern destination is entered.“
That is possible, but lack of human control is only demonstrated by (or perhaps I should say consistent with) the data from 19:41 to 00:19. If there indeed was loss of human control, I think that occurred between about 18:39 and 19:29. For the 1-hour Hold scenario, you need a PF to get out of the Hold at about 19:29, so no pilot actions are required after that time to explain that route. For the scenario with a much earlier FMT, probably just before 18:40, you need a way to produce a southern route (possibly a 180-degree CTH or, perhaps more likely a CMT, which is nearly identical in this case). That heading/track command now appears more likely to be an intended route, instead of an unintended EOR/Route discontinuity error. In that case the command would have been executed close to 18:40. That is a little more than 10 minutes after the flurry of activity near 18:22. Perhaps TRK/HLD of 180 degrees was entered, with the NORM/TRUE switch in its usual NORM position, so that attention could be temporarily diverted elsewhere than to the flight deck, or possibly just to steer the aircraft away from land.
@DrB
Re: “This one I can answer. Mach 0.84 at FL340 is a standard speed setting that matches the radar track speed. CI52 is too slow at this weight. I think it is more likely that M0.84 (Mmo) was used instead of LRC.”
Mmo is M0.87 in the B777-200ER. I’ve never heard of M0.84 being used as a ‘standard’ speed, except that it approximates LRC when the aircraft is flying at the optimum altitude for the weight.
@David
Re: “However the point that the pump will not be automatically selected on autostart still stands, I take it.”
The autostart case isn’t specifically mentioned in the manuals, but it’s my understanding that ELMS also sends the DC pump command signal during an autostart if there is no pressure in the left engine feed manifold. It would make little sense to command the DC pump to run for a normal APU start or while the APU is running, but not do the same for an autostart, given that it would be far more important to ensure a successful start in the latter case. I’ll try to find some more information.
@Andrew. “It makes little sense…” Yes though as you know the manual does include specific criteria that the selector must be “in START or ON” position.
I must say that if the ATSB and others reckon the APU can draw fuel without the DC pump at altitude and with the aircraft in a dive it makes little sense to me why it is there for the APU at all.
Nevertherless, accepting that is the way of it, returning to your logic I think the rationale for APU auto-start is to support twin flameouts by assuredly getting the left engine reignited in a hurry on AC loss. I had not noticed the distinction between ‘left engine manifold’ and ‘left engine feed manifold’ that you pointed out. With the APU DC pump being started automatically at twin flameout there is no need for it to be selected on at auto-start if that is just what it its for.
If so there is no need for your further inquiry, though I must say that does not explain the “while the APU is running” other criterion. However, that might be because it might be selected off prematurely otherwise.
Assuming that there was a hold or other deliberate maneuvering around the FMT 18:22 – 18:40 time frame, which involved speed changes for some reason, would it be reasonable to consider what would happen if the auto throttle was disengaged at that time for that sequence of events, but, for whatever reason, was not re-engaged afterwards ?
If some speed was set along with a heading or course or waypoint, what would the aircraft do as it flew south with some fixed, but arbitrary power setting ?
Would it slowly “drift climb” as weight reduced with fuel consumption ?
@Dr B. The outcome so far of the discussion above on the possibility that an engine failed when the other had its IDG and back up generator off line can be simplified. Assuming that the PDA difference applies (no cross feed, no ancillary load imbalance) the engine failing first would be the right. If the left then flamed out as the boost pumps went off line the consequence to range and endurance would be that the left would shut down shortly after the right, ie with fuel left in the tank.
So in that scenario range and endurance would be shortened compared to the left engine failing when providing AC power.
That supposes the APU DC pump supply would be insufficient to keep the left engine from drawing air through the left tank’s gravity outlet pending boost pump restoration a minute later by APU generated AC.
This is a hypothetical to the extent it would need to be accompanied by a rationale as to why the left IDG and back up generators would both be selected off. That is not obvious.
Dr B. Third line. To, “….no ancillary fuel imbalance” please add, “…favouring the right”.
@Victor @ALSM
“You have to be a little careful. Yes, the 35K ft arc is about 5 NM outside of the sea level arc. But the possible distance between the last transmission and the impact also reduces if the transmission occurred at lower altitude. You have to consider both these effects.”
Hard for me to calculate, but I agree with whatever the “average Arc7” is we need to search equally both sides of it.
Since Item (4) we did not hear IFE logon at 00:21, that seems to agree with one hypothesis of mine the that sharp descent perhaps
(funny behavior of my keyboard or website)
Cont’d
Since Item (4) we did not hear IFE logon at 00:21, that seems to agree with one hypothesis of mine the that sharp descent perhaps started lower altitude such as FL100 possibly lower.
@ventus45: With A/P engaged and ALT HLD selected, the aircraft would not drift up in altitude if thrust were held constant.
Tbill: regarding altitude at 00:19:37…please see my statement at October 27, 2017 at 1:45 pm, item “c”.
@Andrew,
Thank you for correcting the value of the maximum operating Mach. I wonder then, if M0.84 is not a ”standard speed”, why Boeing went to the trouble of providing a specific fuel flow table for it?
Here are the Machs and TAS values for various flight levels at LRC using the estimated average weight from 17:22 to 18:22 (213 tonnes and +10C):
FL330 (TAS 486 kts) (M0.817)
FL340 (TAS 490 kts) (M0.828)
FL350 (TAS 493 kts) (M0.836)
FL360 (TAS 493 kts) (M0.841)
FL370 (TAS 493 kts) (M0.841)
FL380 (TAS 493 kts) (M0.840)
The estimated air speed over the radar track is 498 +/- 5 kts. This is consistent (at the -1-sigma level) with LRC at FL350-380. If one wants to assume an even flight level was used heading west, then you get FL360 or FL380. I would say that FL340 at LRC is not very likely (but possible). The Mach needed for 498 KTAS is M0.841 at FL340 and increases to M0.849 at FL380. At FL360 the speed is about 285 KCAS, but I don’t know why a fixed KCAS would have been used.
My conclusion is that we have three possibilities for the radar track speed setting:
1. LRC at FL350-380 (and more likely at FL360/380)
2. M0.84-0.85 at FL340-380 (set via MCP)
3. ECON with Cost Index ≫ 180
Can you suggest other options or opine which of these might be more likely to have been used?
@ventus45
Re: “Would it slowly “drift climb” as weight reduced with fuel consumption ?”
No, as per Victor’s comment. If the thrust were held constant with the AP and ALT (or VNAV ALT) engaged, the speed would slowly increase as the aircraft’s weight decreased.
@David,
The impact of no additional L engine run time after R engine flame-out is about 7 minutes of INOP range and endurance.
Another argument against this scenario is the fact that the average (radial) speed between the 00:11 and the 00:19 Arcs shows a marked drop. The speed reduction would not be this large, in my opinion, if both engines had operated from 00:11 until 00:17:30. In other words, the 6th and 7th Arcs are too close together to be consistent with no speed change until after 00:17:30.
@DrB
Re: “I wonder then, if M0.84 is not a ”standard speed”, why Boeing went to the trouble of providing a specific fuel flow table for it?”
The FCOM performance data is primarily intended for use in the case of a dual FMC failure. In that sense, I guess you could call M0.84 a ‘standard speed’ that the crew might use, but it’s faster than the typical ECON speeds used by airlines these days and there is certainly no requirement to fly at M0.84 in the event of a dual FMC failure. Indeed, many airlines tailor their FCOMs and some include the LRC tables but not those for M0.84.
Re: “The estimated air speed over the radar track is 498 +/- 5 kts.”
I assume you’ve already taken the wind into account when you determined the estimated air speed from the observed ground speed? If so, I would say that option 2 is the most likely, followed by 1 or 3. A speed of M0.84-0.85 would be consistent with someone who wanted to fly faster than normal, but not flat out at Mmo.
@ventus45
Just to expand on Victor’s and Andrew’s responses to your question about a disengaged autothrottle, if the “other deliberate maneuvering” you contemplate involved a change of altitude and that change had been commanded using the Flight Level Change (FLCH) function then the airplane would hold its speed and ascend gradually as fuel was burned. You’d see something like a 3,500 – 4,000 ft ascent over the course of the flight into the SIO.
@Andrew,
Yes, I took the tailwind into account in figuring the air speed from the ground speed across the radar track.
I’m a bit confused about M0.84. By calling it a “standard speed”, I really just meant that it was easily and directly settable. I understood from your first response that it was not normally used, but in your second response you indicated it was more likely to be used than LRC. Did I understand you correctly to mean that, if you wanted to speed up a bit from ECON (with CI = 52), that you would set M0.84 or M0.85 rather than selecting LRC or simply increasing the Cost Index to several hundred? Why (or under what circumstances) would a relatively small speed increase (only 2-3%, or less than 2 minutes over the next hour) be done at all?
@DrB. Thanks. Yes I follow. APU start is the datum.
@Victor
I don’t know about taking that 20° AOB turn at 10N too seriously, I was really just curious with regards to what it might mean and how the sim program modelled performance towards the edge of the airplane’s envelope.
What is becoming increasingly apparent is that what we are seeing with regards to the simulation airplane’s orientation at 10N (the 20° AOB left turn coming through 255.5° at FL400) is more likely a representation of what the airplane was doing on the previous segment of the simulation (the segment that falls between 5N and VAMPI) rather than what it is doing after being dragged to 10N. The proposition is at least in part supported by both internal and external factors, specifically and respectively, the “inherited” navaid settings and the fact that TBill can’t recreate a plausible turn off N877 that passes through 10N on the correct heading.
A few more inferences arise:
1. Sometime after 5N (and, if fuel burn is considered, most likely before reaching VAMPI) the simulation airplane most likely commenced a 20° AOB left turn off a heading greater than 255.5°.
2. On the basis that from the Malacca Strait (GUNIP) to the eastern Indian coastline (RINTO) N571’s track falls only between 286° – 311°, if the simulation airplane ever reached N571 at VAMPI, it left it again
3. The absence of a disparity between the “Altitude” (=+040003.30) and “AGL” (=40003.327639357369) in the 10N data set means that the simulation airplane was most likely at FL400 before it was dragged to 10N.
4. 10N most likely has no relationship to a deviation from a flight along LAGOG N877 DOTEN. There is, in fact, no evidence that the simulation airplane was ever on N877 (or N571 for that matter).
That makes depictions of the simulation “path” as travelling along N571 and N877 and of the turn at 10N as part of an FMT towards 45S1, quite fanciful (or, at the very least, very highly speculative). Further, it all calls into question whether the simulation relates to MH150.
@DrB
Re: “By calling it a “standard speed”, I really just meant that it was easily and directly settable.”
I thought you meant the crew might set M0.84 for a specific purpose, the same way they would set M0.82 for severe turbulence penetration, for example. There is no specific reason to set M0.84 except:
a. it approximates LRC at optimum altitude, and
b. there happens to be M0.84 performance data in some operators’ FCOMs.
Re: “Did I understand you correctly to mean that, if you wanted to speed up a bit from ECON (with CI = 52), that you would set M0.84 or M0.85 rather than selecting LRC or simply increasing the Cost Index to several hundred?”
Yes, but only because it’s very easy to ‘open’ the MCP speed window and wind it up to a specific speed. The LRC/increased CI options would require the pilot to select LRC or a higher CI, see what speed resulted, then perhaps make another change if the resulting speed wasn’t deemed to be appropriate. In practice, however, any one of those three methods could be used to increase speed; there’s not much between them. It would also be possible to set a specific speed (eg M0.84) in the FMC.
Re: “Why (or under what circumstances) would a relatively small speed increase (only 2-3%, or less than 2 minutes over the next hour) be done at all?”
On a ‘normal’ flight it might be done because of an ATC requirement to maintain separation with another aircraft, or to arrive at a particular waypoint at a specific time. In the context of MH370, however, who knows? In reality a small speed increase doesn’t make a heck of a lot of difference over the course of an hour, as you noted, except that it ‘feels’ a bit faster.
@Victor Iannello:
Ventus 45 said: “If some speed was set along with a heading or course or waypoint, …”
Where did he say that ALT HLD was selected?
Just curious how long my post will be “awaiting moderation” this time.
[VI: If this site were not moderated, it would be quickly overrun with trolls, spam, and disruptive comments. If you are not satisfied with the moderation here, you should consider contributing elsewhere.]
@ALSM. About search width, confidence in that depends on;
• the ATSB final BFO interpretation being correct and if it is;
• there being no glide or straightening of the descent spiral predicted by Boeing that would take the aircraft beyond distances based on unpiloted descent simulations. In other words, whether a pilot was present or not would make no difference, and;
• the simulations are valid representations of the aircraft behaviour with no pilot input, and;
• confident deductions as to search widths can be drawn from the simulations.
My judgement like yours and many others is that the prospects of the first two proving sound are high.
The others are weaker.
As you know the ATSB has said, “Simulations that experienced a descent rate consistent with the ranges and timing from the BFO analysis generally impacted the water within 15 NM of the arc”.
Following the first dot point, those which are inconsistent with those ranges and timing are irrelevant. So BFO compliance is effectively a selection tool of scenarios, as distinct from scenarios in any way ‘proving’ the BFOs.
In that context two issues which remain unclear are:
• what specific scenarios should receive no further consideration and is that consistent with other evidence?
• How valid are the simulations? In turn about this:
1. Which simulations other than those specially asked of Boeing most recently have been included? I think Gysbreght mentioned there were just one or two among those which met the criteria: not much of a sample.
2. Does ‘timing’ above mean the 8 secs between the BFOs or the time lapse after such as APU start/RAT deployment to the BFOs; or both?
3. What account was taken of the possibility of a break in communication with the satellite in steep bank or spirals?
4. What was the prospect of in-flight aircraft break up in any tight spiral or pull up, which might scatter debris into more than one site, as per MH17?
5. The ATSB says also that, “flight simulators are unable to accurately model the dynamics of the aircraft’s fuel tanks……” It adds that if unusable fuel, “resulted in APU start-up, it would re-energise the AC buses and some hydraulic systems. This could affect the trajectory of the aircraft. Similarly, the left and right engines may also briefly restart, affecting the trajectory”.
Since 5. was not included in the simulations does that not invalidate them?
I do not think that a search width conclusion can be drawn in view of that, as apparently has the ATSB/SSWG. Besides, lack of answers to 2, 3 & 4, vitiate validity further. Finally, noting the apparent sample size, there can be little confidence anyway in the actual results.
This joins other implicit probability assessments such as the position on the 7th arc and the assumption of no pilot, even if that quite probably is sound. My own intuitive multiplication of all the implicit probabilities of these results in a low one overall.
I see the simulations as being the weakest link, based on what has and has not been disclosed about them. It is also the issue that appears most open to potentially rewarding work.
Meantime, the search goes on (or I hope it will) but I think the outcome as it stands now will depend more on luck than it should.
PS About why a pilot would follow a 15,000 fpm descent with a long glide you say, “In fact, I haven’t heard any reason why that might have happened, sensible or not.”
I have mentioned that a pilot might have lost control at high altitude as control power was reduced during the time the RAT was deployed, that continuing until the APU happened to come on line. Or he might have tried a zoom climb, stalling and again losing control. As to the second, how that would meld with recovery lower down I could not say but then a pilot at that stage may not be consistent.
Maybe, “…or not” but there it is.
@Andrew. Wanted to check one thing with you, if I may. Is ECON descent speed always the same KIAS for a given cost index or does it vary depending on aircraft weight (or any other factors). What I’m trying to understand is: if the aircraft were “stuck” in VNAV ALT without having initiated descent and speed on ECON descent KIAS, would you expect the speed to change over the next 5+ hours, assuming no pilot inputs?
@ALSM RE YR comment of October 27, 2017 at 1:45 pm:
“a. an extremely high rate of descent began within seconds of FE”
Do you consider 5000 Ft/min an “extremely high rate of descent”? And do you mean seconds before 00:19:29?
“b. 9M-MRO’s vertical acceleration by 00:19:37 was ~0.68g”
Based on your numbers the average b. vertical acceleration between 00:19:29 and 00:19:37 was ~0.68g. If the vertical acceleration is assumed to be constant, then the rate of descent must have started to increase 2.4 seconds before 00:19:29
“c. given a & b, 9M-MRO must have already descended 10-20k ft feet by 00:19:37”
The steady rate of descent power off at cruise altitude is about 2000 ft/min. Between FE and 2.4 seconds before 00:19:29 the airplane would have lost 4000 ft or the equivalent in total energy. In the next 10.4 seconds the average rate of descent was (2000 + 15000)/2 = 8500 ft/min and the altitude would have reduced by 8500/60*10.4 = 1473 ft. The altitude at 00:19:37 would then be about 5500 ft less than at FE. The speed would have increased, and the excess speed couyld be converted back to altitude in the recovery. That recovery could be effected in the next 10.4 seconds and 1473 ft of altitude after 00:19:37.
“d. even if a pilot intentionally caused b, given c, that pilot could not possibly have levelled out and glided more than 30-40 nm. A continuation of the steep descent is far more likely, even if a pilot was in control.”
Your first sentence has been debunked above. I don’t share the opinion you express in the second sentence. If the purpose of the steep descent was to (re)gain airspeed, it is unlikely to have continued after 00:19:37.
@Gysbreght said: Where did he say that ALT HLD was selected?
He didn’t say the aircraft was in ALT mode (which would be entered if HOLD was selected). He did ask about a condition in which the plane was maintaining a heading or was flying towards a waypoint. That implies the A/P was engaged. With the A/P engaged, both roll and pitch are controlled. If the plane was not in a descent or climb, that implies it was in either ALT or VNAV ALT, as @Andrew said. If in either ALT or VNAV ALT, for a constant thrust, as fuel is burned over time, the weight decreases, the pitch decreases, the induced drag decreases, and the plane increases its speed, as @Andrew said.
@ventus45: Here’s one way a drift up in altitude might occur while the A/P is engaged:
1) The A/P and A/T are both engaged and altitude is steady in ALT or VNAV PTH modes.
2) The A/T is disconnected by unarming both engines.
3) A higher altitude is selected in the MCP window and FLCH is selected.
4) As fuel is burned, the plane will slowly drift up in altitude.
Note that for this scenario to occur, the pilot would have to manually select a higher altitude in the MCP.
[VNAV ALT was changed to VNAV PTH in (1)]
@Victor Iannello, @Andrew:
Ventus45 also said: “… consider what would happen if the auto throttle was disengaged at that time for that sequence of events, but, for whatever reason, was not re-engaged afterwards ?”
Can you have VNAV ALT with the autothrottle disengaged?
@Gysbreght asked: Can you have VNAV ALT with the autothrottle disengaged?
@Andrew: Please jump in if I have something wrong.
The quick answer is “yes”.
First, VNAV ALT implies that the VNAV altitude and the MCP selected altitude are different. Without the difference in selected altitudes, constant altitude is maintained by the A/P in ALT or VNAV PTH modes.
If the A/T is disconnected by flipping down the two A/T ARM switches to OFF, A/P modes are not inhibited, and the A/T will remain disconnected.
If the A/T is disengaged with the A/T ARMed, changing the pitch mode will automatically engage (wakeup) the A/T, which can then be manually disengaged, although the A/T will wakeup as required to protect against exceeding envelope limits.
@Paul Smithson
Re: “Is ECON descent speed always the same KIAS for a given cost index or does it vary depending on aircraft weight (or any other factors).”
The ECON descent speed varies a small amount with weight, but the main factor is the cost index.
Re: “…if the aircraft were ‘stuck’ in VNAV ALT without having initiated descent and speed on ECON descent KIAS, would you expect the speed to change over the next 5+ hours, assuming no pilot inputs?”
The ECON descent Mach no. is normally the same as the ECON cruise Mach no. The FMC will enter the descent phase at T/D, but if the aircraft fails to descend it will continue to maintain the Mach no. and won’t transition to the ECON descent ‘speed’ (ie KIAS). I suspect the FMC will recalculate the Mach no. as the weight decreases, but I’m not sure – I’ve never flown more than a few minutes past T/D!
@Victor Iannello:
I don’t consider myself a troll, nor my comments to be spam or disruptive.
Anybody interested in my comments should be aware that they appear on the blog after some delay.
[VI: I reserve the right to review all comments before they appear. This is not Twitter. If you are unhappy with how this site is moderated, you should contribute elsewhere.]
@Gysbreght
@Victor
I agree with Victor’s comments. In short, all VNAV modes are available without the autothrottle.
@Victor Iannello, @Andrew:
Thank you both for your replies on VNAV ALT
@ALSM
I accept the 0:19 BFO’s represent a very high speed descent in those 8 seconds till 0:19:37. And it makes a strong case for that descent rate continued till impact within ~15Nm of the 7th arc.
But imo no one can be completely sure yet at which altitude this 8 seconds descent event took place and what happened after 0:19:37.
Then there is the conflicting information regarding the debris. Particularry the flaperon and outboard flap piece.
They show no substantial leading edge damage but mainly trailing edge damage. A high speed nose down impact is impossible if those pieces seperated on impact.
There is still no conclusive evidence (published..) on how those pieces seperated.
Imo the debris (particularry the flaperon and outboard flap section) show clear signs of a ditch-like impact and the final BFO’s show signs of a high speed nose down impact.
Two very conflicting sets of information that are still not resolved.
My take on a wider search width is to include both possibilities as long no conclusive evidence has been presented for one of both scenarios.
~50Nm on each side of the 7th arc would limit chance of failure if this uncertainties still exist when OI starts searching.
And they could (and would I guess) work inside out from the 7th arc from areas not searched yet.
If the plane then not gets found within ~15Nm between ~32S and ~36S they’ll know some kind of glide had to be involved and they would have given themselves room to search beyond ~15Nm.
I think it would be ‘stupid’ if OI would limit themselves only to a high speed nose down scenario within ~15Nm of the 7th arc at this stage.
Too much conflicting evidence still.
Than it’s better to include both possibilities from the get go imo.
@Ge Rijn: You propose that the search extend to 50 NM from the arc without considering that it will restrict the distance searched along the arc. Ultimately, limited resources and time dictate a balance based on probabilities and judgment. As for not searching north of 32S, that seems arbitrary. Further north better explains some evidence, such as the timing of the “Roy” discovery at Mossell Bay and the lack of debris on the shores of Western Australia.
@VictorI
I admit ~32S is a latitude limit based on my own assumptions.
I would surely support an initial search along the 7th arc from ~35S till ~30S within ~15Nm. But as you say choices have to be made.
The several forward drift studies give uncertain and sometimes conflicting outcomes. North of ~30S drifters arrive a year too soon at Reunion and Madagaskar. @MPat’s study shows none of the 177 historical drifters (over 20 years) that passed the search area above 36S reached WA.
‘Roy’ had many big barnacles still attached when first found in december 2015 and it was the most distance piece ever found so far. So this piece sets a ~time marker of ~20 months drifting from the crash site.
North of ~30S this piece should have arrived much earlier imo.
I see the importance of searching along the 7th arc up to ~30S but not with the exclusion of a possible glide after a steep descent.
The ATSB based their search mainly only on the latter and failed to find the plane.
I won’t like to see OI taking the same assumptions the ATSB did and limit their search within ~15Nm of the 7th arc anywhere.
@Mick
“4. 10N most likely has no relationship to a deviation from a flight along LAGOG N877 DOTEN. There is, in fact, no evidence that the simulation airplane was ever on N877 (or N571 for that matter).”
Not sure about LAGOG, but DOTEN sure does seem to be the waypoint the sim study was heading towards. One could imagine that is consistent with a Jeddah route diversion…not sure if heading to DOTEN would be part of a hypothetical MH370 strategy.
@Ge Rijn: I won’t like to see OI taking the same assumptions the ATSB did and limit their search within ~15Nm of the 7th arc anywhere.
My guess is that OI will extend the search to 25 NM or so from the arc, based on the 25,000 km2 priority area recommended by CSIRO.
Hi @Andrew. I thought that if you are in ECON cruise and programme an approach, that descent would occur at ECON speed descent. I’d further understood that the switch to ECON speed, denominated in KIAS, occurs as soon as you reach T/D; and that the switch to VNAV ALT occurs as soon as deviation from vertical path has surpassed threshold. So that if you go past T/D you very soon end up in VNAV ALT at a KIAS speed of around 272kts. Is all of the above incorrect?
@Paul Smithson
My apologies, you are correct for the case where the cruise level is around FL350. I was thinking of a higher level cruise, where the aircraft’s CAS at the ECON cruise Mach no. is less than the ECON descent CAS. In that case, the aircraft will initially descend at a Mach no. equal to the ECON cruise Mach no. until the CAS catches up to the ECON descent CAS at the CAS/ Mach transition altitude.
For the FL350 case, if the ECON descent CAS is 272 kt and the ECON cruise Mach no. is more than about M0.80, then the CAS in the cruise will be more than the ECON descent CAS. In that case, the FMC will calculate a level deceleration segment at T/D and the aircraft will reduce speed from the ECON cruise Mach no. to the ECON descent CAS while in level flight. Going back to your original question, the last sentence in my previous reply still stands, I suspect the FMC will recalculate the [ECON descent CAS] as the weight decreases, but I’m not sure – I’ve never flown more than a few minutes past T/D!
@Andrew: A couple of weeks ago, Paul Smithson asked me many of the same questions he is asking you, and I performed a simulation of his test case. Among other things, I was surprised to find that the ECON DES speed did not vary with weight. We looked at a descent into WITT in which the MCP altitude was not changed after T/D, which produced some interesting behaviors. I sent you an email with the results.
@Victor
My guess is that OI will extend the search to 25 NM or so from the arc, based on the 25,000 km2 priority area recommended by CSIRO.
I am more interested in the trade-offs between detection of debris than the imaging of debris. Fabricated objects will generally have a much higher reflectivity than objects created by nature. When I gaze at the mountains surrounding my ranch there are many “specular” reflections that leap out at you from various man made objects – cell towers, radio repeaters, abandoned vehicles, old propane tanks,…and many other things. It has to be true that the sound energy reflected off of aircraft debris would behave in a similar manner. I have not seen (nor have I read) anything about the trade-offs relative to scanning at lower resolution (wider swath width) and looking for reflection peaks and subsequently investigating such events at higher resolution. My guess is that the scan coverage could easily be doubled or more by exploiting this effect.
@DennisW: Yes. OI’s mission is to find parts of MH370, which should be “acoustically hard”, and not to map the seabed. That will probably make the data that is generated less useful, but it should also reduce the cost and improve the speed of scanning.
@Victor
Excellent.
@Victor
I would also like to know the results of your descent into Banda Aceh (WITT) experiment.
Have you ever tried it from 6 nautical miles south of Penang with a diversion programmed to Banda Aceh (WITT) via Vampi-Mekar-Nilam-Sanob-BAC-route discontinuity at FL350/Econ Cost Index 52 cruise with Auxiliary Power Unit on and unresponsive crew?
Where was the aircraft at 1825:27 in the above example? Was it turning left at Nilam and exposing its right High Gain Antenna mounted on the right side of the fuselage?
Where did the aircraft end in the above example? Was it beyond 40 nautical miles past the Bayesian Hotspo
What was the speed profile passing Banda Aceh if your destination in the FMC is Banda Aceh? Was it 272/240/272 KIAS?
What was the speed profile passing Banda Aceh if after the route discontinuity your destination is still Beijing? Was it Econ cruise profile: Mach 0.82 reducing to M0.78?
@Victor. Scanning through I had missed Gysbreght’s 7:12 AM post and came across it by chance, which has happened before.Perhaps you could draw attention to release of delayed posts?
@Paul Onions: No, I haven’t repeated your test.
@Victor
@ventus45
Victor, in your 7.58am response to ventus45 with respect to the fixed thrust setting caused by a disengaged autothrottle giving rise to a gradual ascent in FLCH, you say,
“Note that for this scenario to occur, the pilot would have to manually select a higher altitude in the MCP.”
I don’t think that is correct. You don’t have to select a higher altitude in the MCP just a different altitude (higher or lower) using FLCH.
Using FLCH switches the AP mode to “speed-on-elevator” and it remains in that mode until the target altitude is reached when it then reverts to “path-on-elevator”. As the name suggests, in “speed-on-elevator” mode the AP controls speed using pitch via the elevator and ascents/descents are controlled using thrust typically via the autothrottle. Without the autothrottle engaged and absent pilot intervention, there’s no change to thrust and therefore no change to altitude but the AP remains in “speed-on-elevator”. Over time the airplane becomes lighter as fuel is burned, the fixed thrust setting means that it will tend to fly faster which the AP promptly corrects by pitching the nose up slightly and you get an ascent. It is a very gradual process. In the case of MH370 had the airplane been in “speed-on-elevator” mode with a disengaged autothrottle at the IGARI turnback then the total change in altitude over the ensuing 7 hour flight is only about 4,000 feet (a smidge under +10 fpm).
As an aside, failure of either autothrottle servo motor (ASM), the small electric motors that physically move the thrust levers to command changes to thrust, will cause the autothrottle to disengage. The circuit breakers (2.5 amp) for the Left and Right ASMs are mounted on the P11 Overheard Circuit Breaker Panel on the flight deck.
@Mick Gilbert: I don’t think that’s correct. With A/P engaged, with or without the A/T connected, a climb won’t exceed the altitude in the MCP window, even when in FLCH SPD mode.
To test this in the PMDG777, I went to level flight, SPD, HDG HLD, ALT at FL340, M0.84. I then disconnected the A/T, set altitude in the MCP to 33000 ft, selected FLCH, which was shown in the FMA as [Blank], HDG HLD, FLCH SPD. I then gradually reduced the fuel. As the weight fell, the angle of attack reduced and the speed increased. The altitude remained constant at 34000 ft. Essentially, it behaved just as in ALT mode. I then did the same experiment except I changed the MCP altitude to FL350 instead of FL330. Now when fuel was reduced there was a gradual climb which stopped at 35000 ft.
If you have documentation that says with the A/P engaged you can climb and exceed the altitude in the MCP window, I’d like to see it.
@Victor
@Paul Smithson
Re: “I was surprised to find that the ECON DES speed did not vary with weight.
I have not come across equivalent guidance from Boeing, but the Airbus publication Getting to Grips with the Cost Index suggests that the aircraft weight has a negligible effect on the ECON descent speed. This page shows that the ECON descent speed for the A340 only varies by about 3 kt for the weight range 160-190 tonnes with CI=50. I guess the B777 FMC uses the predicted weight at T/D to calculate the ECON descent speed and that value is then frozen once the aircraft enters the descent phase.
“We looked at a descent into WITT in which the MCP altitude was not changed after T/D, which produced some interesting behaviors. I sent you an email with the results.”
Thanks for your email Victor. I was surprised to see the target speed increase back to the ECON descent speed as the aircraft headed further away from the active waypoint (CF17). I know Paul mentioned that behaviour earlier in the discussion, but it’s not something I have seen previously. The FMC obviously keeps monitoring the vertical profile as the aircraft flies away from the active waypoint. The target speed seems to revert back to the ECON descent speed once the distance from the active waypoint is sufficient to put the aircraft above 10,000 ft if it were flying the ideal descent profile.
@Victor
Victor, as you can see here with regards to pitch modes, the B777 FCOM states,
“FLCH SPD (active) – pushing the MCP FLCH switch opens IAS/MACH window
(if blanked). Pitch commands maintain IAS/MACH window airspeed or Mach.” (my bolding).
What was displayed in PFD pitch flight mode annunciation when you ran that test?
@Victor
Further to the above, if you think about the logic of FLCH process for a descent, as soon as you press the FLCH switch after selecting a lower altitude you are exceeding the altitude in the MCP window.
@Mick Gilbert: Please read what I wrote. I said that the A/P would not allow a CLIMB above the MCP altitude (when not near overspeed, of course). It also won’t allow a DESCENT below the MCP altitude (when not near stall, of course). It doesn’t matter whether the altitude is changed using VNAV, FLCH, or V/S, or whether the A/T is connected or disconnected.
You asked what the pitch mode annunciation was. I already told you it was FLCH SPD. It remains in FLCH SPD until the MCP altitude is captured, which won’t occur if the altitude is set lower than the current altitude and the weight is reduced, all while holding thrust constant.
Rather than reduce the weight, the same effect can be observed in a more pronounced way by increasing the thrust. So we start at FL340 in ALT mode, disconnect the A/T, set the MCP altitude to 33000, hit FLCH, and manually increase the thrust. Rather than climb and maintain the MCP speed, the plane will increase in speed and hold altitude, unless the MCP altitude is increased past 34000, and then only until the MCP altitude is captured.
@Victor
Thank you for your response, Victor. Have you read the FCOM description for FLCH SPD? Apart from the reference already provided there is this (B777 FCOM, Automatic Flight – Controls and Indicators, pp.4.10.8-4.10.9) ;
“Flight Level Change (FLCH) Switch
Push –
• selects FLCH SPD pitch mode
• FLCH SPD pitch mode displays in green (active) on PFD flight mode annunciation
• when IAS/MACH window closed, it opens to the FMC target speed, if valid. If not valid, the IAS/MACH window opens to the current speed
• when IAS/MACH window open, it displays command speed
• when changing from TO/GA to FLCH, IAS/MACH window displays highest value of current speed or selected speed
• AFDS pitch holds selected speed. When selected altitude captured, pitch flight mode annunciation changes to ALT
• A/T operates in THR, followed by HOLD mode in descent. When selected altitude captured, A/T mode changes to SPD
• A/T advances or retards thrust levers to provide 500 FPM vertical speed for each 1000 feet altitude change
• AFDS attempts to reach the MCP selected altitude within two minutes if able with available thrust. Otherwise, A/T uses IDLE or CLB thrust to reach the MCP selected altitude
• with a higher altitude set in the altitude window, reference thrust limit changes to CLB when CRZ displayed or to CON with an engine inoperative.”
The behaviours that you are describing for the simulation program are not consistent with the behaviour described in the FCOM, specifically that when FLCH SPD is selected “AFDS pitch holds selected speed.”. I can’t add anything to what Boeing has written about their airplane.
@Victor
On an entirely separate and very much sadder note, Australian aviation journalist, Ben Sandilands, lost his lengthy fight against cancer last Friday evening. I know that Ben followed and often reported on proceedings from your blog (and before it, from Duncan’s). Ben was very much an “old school” journo, sometimes he wasn’t right but he was always forthright. He will be missed.
@Mick
I can’t add anything to what Boeing has written about their airplane.
Of course, Boeing was represented in the SSWG as was the manufacturer of the AES (who said nothing about overshoot transients).
These discussions are getting very old and very silly to me. It is a classic example of “when you are a carpenter, the whole world looks like a nail.” Good grief people, move on with your lives miserable as they may be.
Do we think we are smarter than the manufacturers of the equipment? Do we think the manufacturers are part of a plot to make the airplane disappear?
@DennisW
Has Ami had you splitting wood again?
@Victor
Regarding our discussion on the A/P not allowing a CLIMB above the MCP altitude, you are correct. I had misinterpreted a prior separate discussion on the matter. My apologies.
@Mick Gilbert asked: Have you read the FCOM description for FLCH SPD?…
I understand quite well how FLCH works. I also understand how MCP altitude constraints work, which you seem to neglect.
@Mick Gilbert: Regarding our discussion on the A/P not allowing a CLIMB above the MCP altitude, you are correct. I had misinterpreted a prior separate discussion on the matter. My apologies
Thank you. I’m glad we reached agreement.
@TBill
I said, “10N most likely has no relationship to a deviation from a flight along LAGOG N877 DOTEN. There is, in fact, no evidence that the simulation airplane was ever on N877 (or N571 for that matter).”
You said, “Not sure about LAGOG, but DOTEN sure does seem to be the waypoint the sim study was heading towards. One could imagine that is consistent with a Jeddah route diversion … ”
Mmm … Can you explain your rationale please, Bill? “Heading towards” implies it was “coming from” somewhere. The previous data point is 500 nm away and a track between the two doesn’t align with DOTEN.
More broadly, I think that there may be a marked difference between the way many of us conceive of how simulators are used and how a senior commercial pilot and Type Rating Examiner, such as Captain Shah, might actually use one. Many of us think each simulator session is representative of one flight (take-off, climb, cruise, descent, landing), typically either rehearsing a flight to come or recreating one that has taken place. That’s not necessarily how flight crew members, particularly instructors/examiners, might approach a simulation session. In many cases simulation sessions for commercial pilots, particularly during training, are a series of exercises, often quite unrelated (eg a TCAS alert and avoidance maneuvering followed by a stall at 5,000 feet followed by an accelerated climb to FL350 followed by recovery from a dual engine failure). For qualified flight crew, six monthly proficiency checks and twelve monthly Line Operational Simulations (LOSs) are normally constructed around “one flight” scenarios whereas six monthly recurrent training sessions tend to fall into the “series of exercises” category.
If we look at the six data points from the Captain’s flight simulator with a “one flight” mindset then that is what we see, one continuous flight from Kuala Lumpur terminating in the Southern Indian Ocean. If we look at the data points with a “series of exercises” mindset we might see something else, three or possibly four separate exercises, separated by the airplane being dragged from one location to another. I’m not saying that one perspective must be right and the other, by corollary, must be wrong, I’m just saying that there are different ways of viewing that sparse data set.
@Mick Gilbert: If you think a TRE would use the FS9/PSS777 to practice exercises of the type you mentioned, you are sorely mistaken. The fidelity for simulating abnormal conditions is not close to being realistic, nor does it have complete checklists that a trained pilot would follow. The FSX/PMDG777 is much closer to behaving as documented in the FCOM, but even that model has some important limitations. For instance, when I used the PMDG777 to simulate fuel exhaustion, I had to manually do things like force a failure to secondary mode and inhibit the IDGs from powering the AC busses at idle spool speeds. Any 777 pilot would be very aware of the obvious limitations of either aircraft model, but especially the limitations of the FS9/PSS777.
@Mick
Has Ami had you splitting wood again?
Stacking and chain saw work. Ami has been doing all the splitting.
We are all saddened by Ben Sandiland’s passing. Ben dared to ask questions about the disappearance of MH370 that few others were willing to ask. He continued to passionately contribute even while he battled cancer. He was a reader of this blog, and through his column, he helped some of us reach a broader audience. Rest in peace, Ben.
Read his articles often. No sign he was battling a fight like this in between. Great respect.
The unbearable lightness of being comes up in my mind.
R.i.p. for him. At least he and his friends and family knew what was coming I suppose. All the victims of MH370 deserve the same. To rest in peace and acceptance. The plane must be found.
Otherwise no closure will be possible for those deseaced and the loved ones left behind.
Read his articles often. No sign he was battling a fight like this in between. Great respect.
+1 Ge Rijn
Especially the respect part – a model for what a journalist should be.
@Victor. Contract negotiations re wreckage finds.
There is the possibility of an in flight breakup from a spiral, pull up or pitch up. I have mentioned MH 17 in this context: two main debris sites, well separated. That resulted from a fuselage break, the main front section ‘gliding’.
The Silk Air 737 ended in a very high speed descent only because a pilot applied full manual down trim beforehand then held the nose down. Otherwise it would have pitched up, according to Boeing.
If MH370 were in a spiral, speed becoming excessive, most likely the pitch up would translate to a tightening spiral then break up. If not in a spiral, then a pitch up (or piloted pull up) could lead likewise to a break up. (If not the aircraft could recover and possibly fly some distance, with phugoids… but that is another issue).
If MH370 broke, that might well be at a wing, but not necessarily: it could be the fuselage, the tail coming off. That could lead to well separated debris fields, the recorders ending up well away from the main wreckage.
Because the ultimate outcome would be beyond aircraft limitations it would not be apparent during simulations. However an ALSM simulation resulted in a spiral that when analysed evinced high g. That I think is relevant, despite no simulations envisaging tank residual fuel.
IMO the Silk Air analysis by Boeing is likely to be relevant to the 777, meaning the above is not just a hypothetical, to be left in the ether as hitherto. A mid air breakup, the front section not necessarily hitting the sea nose down, could help explain some ambivalent characteristics of the debris.
@David: Yes, I agree that finding the main debris field will be challenging.
@Victor
I did not say that Captain Shah was practising any of those non-normal items. I provided the example of a sequence of non-normal items (taken from an actual B777 flight simulator training session) to illustrate how typical simulator training sessions are constructed. The point that I was making is that the approach taken by commercial pilots, in particular instructors/examiners, to flight simulation sessions is much broader than the simple “one session = one flight” perspective that underpins the join-the-dots interpretation of the Captain’s simulation data. In short, there is more than one way to interpret the data.
The simple fact of the matter is that there is insufficient data to conclude that the data points are from a single complete, contiguous and coherently constructed flight, they might just as likely represent three or four separate exercises, separated by “dragging”, that are completely unrelated. That is the virtue of simulation programs, they allow the user to undertake multiple disparate exercises in one session.
We can say with reasonably high confidence that the simulation airplane was “dragged” from one location to another on multiple occasions during the simulation session. Moreover, we know that the “dragging” process is not constrained; the simulation airplane can be “dragged” from one point to anywhere else on the globe completely independently of its path at the originating point. Thus, there may be no relationship whatsoever between the location from which the simulation airplane was “dragged” from and where it was “dragged” to. Suggesting that there is a path or route between two locations under such circumstances is at least misguided and quite possibly wrong.
Had the user wanted to model one continuous and related flight the option of using the time acceleration function was available. Using time acceleration models fuel usage and preserves the relationship between two points, “dragging” does not.
Accordingly, the depiction of a “simulator path” connecting the various data point locations as being a route taken by the simulation airplane sits somewhere between inference and speculation. To illustrate such a path as following N571/N877, when there is no evidence that the simulation airplane was ever on either of those airways, is even more highly speculative.
Additionally, the fact that there is no evidence that fuel usage was modelled accurately between the data points supports the possibility that the simulation session is more likely a sequence of separate, and possibly unrelated, exercises. With evidence of such little regard to fuel usage, it is not unreasonable to infer that the simulation session was not a fuel planning exercise nor was it meant to model an actual flight, such as MH150. Further, given your experience that the far more sophisticated PMDG simulation program fails to accurately model airplane systems behaviour on fuel exhaustion, it is probably not unreasonable to also infer that the less sophisticated PSS simulation program would not have been used for that purpose.
@Mick
That relevancy of the simulator data lies in the SIO data points. Does not matter how or if the data points were connected (dragged or contiguous) or not. How in the world can you possibly ignore the SIO data points regardless of how they might have been created?
You are asymptotically approaching the whacko category.
@Mick Gilbert: Time acceleration is not as perfect as you might believe. First, even at 16X time acceleration, dragging is much faster. Secondly, time acceleration, especially around turns, causes flight instability on the PSS777, which anybody that has actually used the program understands.
There was a time not so long ago that a number of people did not even believe the sim data sets were all from a single session, and I received a lot of unkind comments for suggesting they were. There were claims that the points were cherry-picked among thousands, the chronological order could not be determined, and the simulation could have been performed months before the shadow volume data of Feb 4. Of course, this completely ignored the fact that the sim data sets were found grouped together in a single shadow volume with no other flight data. This made them unique, as I have said for many months.
With the additional information from the ATSB showing that the time stamps for the data sets were Feb 2, chronologically ordered according to the flight path, and all timestamps within an hour, we know the data sets are from a single session. But now, you propose that although it was a single session, the data sets might represent multiple flights since this is what pilots do with simulators.
We don’t know for sure why the captain was practicing fuel exhaustion exercises deep in the SIO, and why he chose to save a series of data sets with consecutive positions at KLIA, the Malacca Strait, the Andamans, and the SIO, and with progressively decreasing fuel levels. The likely explanation is that points are related to each other and represent a single flight, irrespective of whether the position and fuel levels were manually advanced during the flight, and irrespective of whether we know the exact path between the positions. We can’t definitively prove that the points represent a single flight, but in my book, it is by far the most obvious explanation. And it would be an extraordinary coincidence if the sim data sets were not related to MH370.
@Victor
@DennisW
You will note that I have never contended that the SIO data points are irrelevant. Nor have I suggested that the “multiple exercises” (as opposed to the “single flight”) perspective is antithetical to malicious intent; the two are not mutually exclusive.
What I principally have a problem with is supposition, interpretation and inference being presented as fact. First and foremost in the supposition as fact category is running a string around the map pins and saying, “That’s the flight path.” Not only wouldn’t I bank on it, I’d bet against it. Worse still is depicting WMKK B446 GUNIP VAMPI N571 LAGOG N877 DOTEN as the “Simulator Path”.
Then there’s the contention that the turn at 10N represents the turn towards 45S1. It is now reasonably clear that the simulation airplane was in a 20° AOB left turn when it was dragged to 10N from somewhere after 5N (and probably before VAMPI). That means that the turn may not even be (and I would argue, probably is not) relevant to 10N, its relevance is to a position 500 nm away from 10N (where, just by the bye, it would have the simulation airplane turning towards Banda Aceh).
I also have a problem with invoking the user fiddling with fuel levels to explain what would otherwise be premature fuel exhaustion at 45S1 if flown directly from 10N. You don’t need to invoke fiddling with fuel levels to explain the data at 3N, 5N or 10N; “fly (and burn) and drag” or “drag and fly” can easily explain the data at those points. Fuel exhaustion at 45S1 suggests either one or more intermediate, uncaptured points (a possibility that the ATSB alludes to) or fuel exhaustion elsewhere with the airplane being dragged to 45S1 in that condition.
Victor, what is the basis for your saying that it is a fact “… that the sim data sets were found grouped together in a single shadow volume with no other flight data“? There is no dispute that the RMP’s Flight Simulator Analysis report states that, “All these coordinates are found in a single Volume Shadow file
Information (VSI) named {OOd7ef6c-8bcb-1Ie3-b3f7-ee8a9181afad} {3808876b-c176-4e48-b7ae-04046e6cc752} dated February 3, 2014.” but I’ve yet to see any reference to there being no other flight data in that VSI.
As to the history of the Captain’s flight simulator data, up until 12 October 2017, when you released the information passed to you by the ATSB that the simulator session lasted for about one hour, the chronological order of the data files could not be determined. Up until only a couple of weeks ago any suggestion that all the data files were contemporaneous was pure conjecture.
Further to the information provided to you, in response to an email from me, the ATSB has stated,
“All files except the final (partial) file contained a [DateTimeSeason] field which contains values for Day, Year, Hours, Minutes, Seconds, Season.
The files with the field had the common values of Day=33, Year=2014 and Season=Winter
The [DateTimeSeason] field in the files indicated 15:26 to 16:38 (hh:mm).”
Now, while I strongly suspect that all the data points were created in one session, the fact that five of the six data points are timestamped within a period of 72 minutes does not eliminate the possibility that they were created over more than one session. The individual time stamps for each data point are what is needed.
As to your question do I “… propose that although it was a single session, the data sets might represent multiple flights since this is what pilots do with simulators“. No, what I propose is that the data sets might represent multiple activities or exercises that may or may not be related to one another. That proposition is not solely based on “this is what pilots do with simulators” but because that is what the data suggests.
Dennis, if I do happen to be categorised as “Whacko” I do hope that I manage to scrape into the “Respected” subcategory.
@David, report from Adam Harvey of ABC News has the following paragraph:
“Deputy Transport Minister Datuk Ab Aziz Kaprawi said that cabinet ministers had agreed “to prepare a special allocation to the Ministry of Transport amounting to between US$20 million up to US$70 million if MH370 aircraft wreckage is successfully found within 90 days”.
It’s the “within 90 days” that made me sit up. A bit optimistic, I’m thinking, considering the challenge ahead, onforseen operating difficulties, weather etc.R
@Rob. Thanks. For those who would like to read the article:
http://www.abc.net.au/news/2017-10-30/mh370-search-us-company-promised-millions-for-trace-of-flight/9100818
25000 sqkm as per the ATSB as I read it. Why a time limit at all is unclear. Maybe that stipulation instead should read to start within 90 days of contract? Similarly the 25000 may be a minimum requirement?
Australia to contribute just ‘technical assistance’, unspecified. China contribution nil?
Do not know of Adam Harvey or how well informed.
Quoting ‘Deputy Transport Minister Datuk Ab Aziz Kaprawi said […] if MH370 aircraft wreckage is successfully found within 90 days”.’
Kaprawi has repeatedly been behind the curve on detail but mention of 90 days, and that payment is being made through MoT/MY, is significant.
Malaysia’s previous search contracting, with Go Phoenix & Phoenix International during 2014-2015, was exercised through Petronas and DRB HICOM.
That previous arrangement, under the auspices of the two companies’ CSR activity, perhaps permitted more flexibility by contracting in the commercial realm while, ultimately, the government paid in the form of tax incentives that encourage CSR work. At the outset of the Phoenix International deployment on Go Phoenix, in Oct 2014, a Petronas exec stated Petronas had commintted funding through to Feb 2015 at which time they’d “sit back & think again”. They did continue to work until Jun 2015.
Governments can’t do flexibility in contracting. It’s fixed scope plus change control.
ATSB’s requirement of Fugro was 200km² per day for which two boats, each with one deep-tow vehicle, were deployed. 25,000km² over 90 days requires 277km² per day. I suspect Adam Harvey is ‘cobbling’ numbers together without any substantial coherent analysis, the focus of his headline is the ‘prize‘, disregarding the effort expended in attempting to attain it.
@Mick Gilbert: We are not going to agree, and I’ll let you get in the last word. I think you downplay the most obvious of scenarios, while you think I ignore other likely possibilities. Readers here can decide for themselves.
@David, @Don Thompson: Some quick math. The 25,000 km2 priority area extends to about 32.6S latitude on the arc. If the search continued along the arc to the north of this point with a continuous width of +/-25 NM (+/- 46.3 km), each degree of latitude searched along the arc corresponds to about 12,780 km2. For instance, a search from 32.6S up to 26.6S latitude would require 76,673 km2 beyond the initial 25,000 km2. Starting south and working towards the north extends the season of favorable weather.
It would be very interesting to know what a realistic expectation is for the coverage per day.
@ALSM: You may have missed my comment of October 28, 2017 at 7:12 am because it was “awaiting moderation” for several hours. Therefore I’ll post it again. I’ve also expanded the text with additional explanations in case the original was not sufficiently clear, and corrected typos.
RE YR comment of October 27, 2017 at 1:45 pm, let’s get the facts right:
“a. an extremely high rate of descent began within seconds of FE”
Do you consider 5000 Ft/min two minutes after FE an “extremely high rate of descent”? And do you mean seconds before 00:19:29 rather than “within seconds of FE”?
“b. 9M-MRO’s vertical acceleration by 00:19:37 was ~0.68g”
Based on your numbers the average vertical acceleration between 00:19:29 and 00:19:37 was ~0.65g. The vertical acceleration at 00:19:39 may have been greateer or less than 0.65g if it varied in that time interval. Since a certain deflection of the control column commands a certain vertical acceleration (by virtue of the B777 fly-by-wire pitch control law) it may well have been constant in that time interval. If the vertical acceleration is assumed to be constant, then the rate of descent must have started to increase from about 2000 ft/min to about 5000 ft/min 2.4 seconds before 00:19:29.
“c. given a & b, 9M-MRO must have already descended 10-20k ft feet by 00:19:37”
The steady rate of descent power off at cruise altitude is about 2000 ft/min. Between FE and 2.4 seconds before 00:19:29 the airplane would have lost about 4000 ft or the equivalent in total energy. In the next 10.4 seconds the average rate of descent was (2000 + 15000)/2 = 8500 ft/min and the altitude would have reduced by 8500/60*10.4 = 1473 ft. The altitude at 00:19:37 would then be about 5500 ft less than at FE. If a pilot can command a downward acceleration of 0.65g by pushing the control column forward, he can just as easily command an upward acceleration of 0.65g by pulling the control column rearwards. Recovery from the steep descent could therefore be effected in the next 10.4 seconds after 00:19:37, and requires also 1473 ft of altitude. So at 10.4 seconds after 00:19:37 the airplane could be back at a steady rate of descent of 2000 ft/min at an altitude of about 7000 ft less than at FE, but at a higher speed. Since the excess speed can be converted back to altitude, the manoeuvre would have negligible effect on the glide distance available at FE.
“d. even if a pilot intentionally caused b, given c, that pilot could not possibly have levelled out and glided more than 30-40 nm. A continuation of the steep descent is far more likely, even if a pilot was in control.”
Your first sentence has been dealt with under c. above. I don’t share the opinion you express in the second sentence. If the purpose of the steep descent was to (re)gain airspeed, it is unlikely to have continued unchanged after 00:19:37.
@Mick
Dennis, if I do happen to be categorised as “Whacko” I do hope that I manage to scrape into the “Respected” subcategory.
Yes, of course.
@VictorI
I imagine some priorities could be made also.
The CSIRO 1/2/3 high priority regions actually cover a relatively small area. If those are searched first and the plane is not found there I think the latest CSIRO theory can be abandoned and there would be no urgence to search the rest of the CSIRO area till ~32.6S.
OI could f.i. decide to search north near along the 7th arc starting form ~32.6S after those CSIRO 1/2/3 priority areas.
If then nothing is found still they could choose to search in more width at a certain area along the 7th arc.
I mean essentially OI could split those 25.000km2 up in different areas that don’t have to be connected.
This would give them flexability also if new information surfaces.
Maybe other debris get’s found during the search or conclusive evidence comes forward on how specific debris seperated.
Just searching that 25.000km CSIRO area would be too limited and a recipy for failure I’m affraid.
We already know MH370 is not near the 7th arc till ~32.6S.
It hasn’t been a direct plunge from 00:19 onwards at ~15.000ft/min RoD till impact. Otherwise the plane was found in the searched area.
@Ge Rijn: Just searching that 25.000km CSIRO area would be too limited and a recipy for failure I’m affraid.
When you hear reports that only the CSIRO-recommended area of 25,000 km2 will be searched, either that is said either out of ignorance, or to deliberately spread false information. There has been an incredible amount of false information about both the plans for the search and about Ocean Infinity.
@VictorI
I trust you know more than I do about OI and their plans. Kind of reassuring. I have only the common media on this.
Does the 90 days stick? Or is this also ingnorance or false information?
It seems a very short time to me considering OI and no one else have ever done this before.
@Rob
It’s the “within 90 days” that made me sit up. A bit optimistic, I’m thinking, considering the challenge ahead, onforseen operating difficulties, weather etc.R
The 90 day statement is not news. It was made in an article which appeared last Friday (10/27).
https://themalaysianreserve.com/2017/10/27/govt-expected-spend-additional-us70m-search-mh370/
Lots of weirdness being reported relative to the OI search “discussion”:
1> 90 days?
2> payment on recovery?
3> third party verification?
At this point I think we would be well-advised to ignore the journalism relative to OI.
DennisW said: At this point I think we would be well-advised to ignore the journalism relative to OI.
@DennisW, @Ge Rijn: I certainly wouldn’t assume that it is 100% factual.
@Victor
Even the payment range of $20M to $70M has not been explained at all.
To simplify, it appears there are just two categories of people in the world:
1. Whackos – those with whom DennisW disagrees.
2. Non-whackos – those with whom DennisW agrees.
This makes it easy to figure out which category you are in.
@Gysbreght
Re: Your October 30, 2017 at 8:22 am post:
“@ALSM: You may have missed my comment of October 28, 2017 at 7:12 am …”
I did not miss your original post. It was your typical diversionary nonsense, not worthy of serious a reply. Trying to pick fights over things like the exact numerical value of the acceleration (0.65 vs. 0.68 gs) or the correct adjective to characterize a 5000 or 15,000 ft/min descent rate is not in anyway relevant to the topic, which was, the best estimate of the 7th arc search width.
@ALSM: “… a 5000 or 15,000 ft/min descent rate is not in anyway relevant to the topic, which was, the best estimate of the 7th arc search width.”
My point exactly. The best estimate of the 7th arc search with can not be inferred from the descent rates at 00:19:29 and 00:19:37.
@Gysbreght: You tell the tale of two pilots:
1) The incompetent pilot who gives a nose up input after fuel exhaustion and causes the plane to stall and eventually reach a high descent rate of 15,000 fpm.
2) The proficient pilot that expertly recovers from a high descent rate and glides the plane for many miles at an efficient descent angle.
How do you reconcile (1) and (2)?
@Mick
That is interesting time stamp info for the simulator runs. Are you saying that is recent info to you from ATSB? Should we be asking for more data from the runs? (which we may or may not be privy to, but at least we would know the response).
As far as the time stamp, that would appear to be the date/time of the simulator cases, however, for some reason if I create a new case in FS9 right now today, my FS9 flight time setting is defaulting to 11-Oct-2017 for some reason. One can of course change to any Time/Date. Seems the Date/Time, if it was instead intentionally set, would be roughly consistent with the of arrival at 45S1 assuming MH150 take off that day.
@Victor Iannello: “How do you reconcile (1) and (2)?”
I don’t have to. The descent rate of 15000 ft/min could result from the recovery from a stall, or from a surprised over-reaction to an unexpected stall warning (stick shaker). Pushing the nose down then pulling it up again does not require an expertly proficient pilot. Either an inexperienced or an experienced pilot could have done this, although I am leaning toward the former.
Now consider a proficient pilot, seven hours after high-jacking the plane at IGARI, after nearly six hour flying at roughly constant speed and altitude, on headings roughly due south, only awaiting fuel exhaustion before pushing the nose down to end the flight as quickly as possible. Does that make sense?
@Gysbreght: After a double engine flameout, there is absolutely no way that a proficient pilot would provide a nose up input (which would be required) and get to stick shaker speed. Nor is a recovery from a 15,000 fpm descent followed by an efficient glide as trivial as “pushing the nose down then pulling it up”. That would be a challenge even for an experienced pilot.
(@Andrew: Do you think a novice could easily accomplish this?)
As for your example of a pilot waiting hours until fuel exhaustion and then pushing down the nose, I don’t know if that occurred or not, but it certainly is within the skill set of a 777 captain.
@Victor Iannello: “After a double engine flameout, there is absolutely no way that a proficient pilot would provide a nose up input (which would be required) and get to stick shaker speed. “
Andrew’s simulation showed that the plane left to itself would come quite close to stickshaker speed after a double flameout.
“Nor is a recovery from a 15,000 fpm descent followed by an efficient glide as trivial as “pushing the nose down then pulling it up”. That would be a challenge even for an experienced pilot.”
I’m sure you would not find it difficult at all if you tried it in your simulator.
@Victor Iannello: “a pilot waiting hours until fuel exhaustion and then pushing down the nose, (…) certainly is within the skill set of a 777 captain.”
The captain’s “skill set” is not the issue.
@DrB
You omitted the largest category – “don’t care”.
@Gysbreght said: Andrew’s simulation showed that the plane left to itself would come quite close to stickshaker speed after a double flameout.
The case you cite was for a single engine flameout, A/P still engaged and maintaining altitude as the speed reduced. For a double flameout with no pilot input, the A/P would disengage, the control mode would degrade to secondary mode, and the plane would descend at the trim speed.
I’m sure you would not find it [recovery from a 15,000 fpm descent] difficult at all if you tried it in your simulator.
Your certainty is misplaced. It’s challenging, as I said. It’s hard to keep the wings level, and it’s easy to rip the wings off the plane on the recovery.
@Victor Iannello: “The case you cite was for a single engine flameout, A/P still engaged and maintaining altitude as the speed reduced.”
Thank you for the correction. The airplane came close to the stickshaker before the second flame-out and stabilized at just 4 kt above the stickshaker speed until the second flame-out. The point is that at that point the pilot may have become aware of the situation and over-reacted to it.
@Victor
@Mick
@ventus45
@Rob
Re: Simulator Cases
>>For those interested in MH150 Time Tables, there are two documents.
(1) From the appropriate MAS timetable Booklet
Flight MH150 was 15:15 Take off until 18-Jan-2014 (Days 1 34 6 )
Flight MH150 was 15:15 Take off 19 to 27-Jan-2014 (Days 1 34 67)
after 27-Jan-2014 the MH150 time table was modified per below announcement:
(2) Malaysia Airlines Increases Jeddah Service in Feb / Mar 2014
“Malaysia Airlines from 28JAN14 to 29MAR14 is adding 2nd daily Kuala Lumpur – Jeddah service, on board Boeing 777-200ER aircraft. Reservation for the new MH168/169 service is now open.
MH150 KUL1145 – 1600JED 772 D
MH168 KUL2230 – 0245+1JED 772 D
MH169 JED0500 – 1835KUL 772 D
MH151 JED1730 – 0705+1KUL 772 D
MH169 begins from 29JAN14 until 30MAR14.
From 30MAR14, MH will operate 1 daily flight as MH168/169, which means the oneWorld member shifting operational schedule to night-time departure from KUL.”
>>I default to Victor re: potential significance, but it would appear that the 15:26 time in the saved simulator case is close to the 15:15 take off time prior to the schedule expansion on 28-Jan. At the moment I am thinking the Time/Date in Microsoft Flight Sim is shown as the last saved case, but I need to do some testing to verify that.
@TBill: “At the moment I am thinking the Time/Date in Microsoft Flight Sim is shown as the last saved case, but I need to do some testing to verify that.”
i your verification tests, could you also look into what happens to the Time/Date if you use the fast-forward option? Does the Time fast-forward or not?
Gysbreght: When you deliberately quote people completely out of context to mislead readers, you make a fool of yourself. You should be ashamed.
You stated at October 30, 2017 at 1:39 pm:
——————-
“@ALSM: “… a 5000 or 15,000 ft/min descent rate is not in anyway relevant to the topic, which was, the best estimate of the 7th arc search width.”
“My point exactly. The best estimate of the 7th arc search with can not be inferred from the descent rates at 00:19:29 and 00:19:37.”
——————-
That is another of many lies by omission. My complete sentence, which you butchered, was:
“Trying to pick fights over things like the exact numerical value of the acceleration (0.65 vs. 0.68 gs) or the correct adjective to characterize a 5000 or 15,000 ft/min descent rate is not in anyway relevant to the topic, which was, the best estimate of the 7th arc search width.”
In context, it is clear I had no intention whatsoever to say what you claim I said. Of course 5000 or 15,000 ft/min descent rates are very important in estimating the search width. I have no doubt you are 100% aware I was referring to your ridiculous argument that my use of the adjective “extreme” to describe the 15,000 ft/min descent rate somehow discredited the whole thread of logic.
@ALSM: My out-of-context quote was of course deliberate and ‘tongue-in-cheek’. My apologies if that upset you.
Your reply limited to the first point addressed in my comment just for completeness and for the purpose of the more significant aspects I raised later my comment ignored what I was trying to bring across. That is that those 8 seconds do not define what could have occurred in the 20-odd minutes of descent that lay ahead.
@TBill
Good work TBill. And wise of you to default to Victor on significance of the 15:26 SIM time.
CORRECTION: Please read “… later in my comments”.
@Victor
Ron is a typo. There is no Ron, trust me. There is only a Rob. My mobile has autocomplete syndrome
@DennisW
“Lots of weirdness being reported relative to the OI search “discussion””
Apologies Dennis for making something out of nothing with the 90 days thing. I don’t mind being associated with weirdness but if you should ever put me in the whacking camp with those other unfortunates, it would send me over the edge.
@Rob
No fear. You are pure of heart. BTW, I do think the 90 day thing is significant. I have no idea what it could mean in the search/payment context.
Gysbreght asked: could you also look into what happens to the Time/Date if you use the fast-forward option? Does the Time fast-forward or not?
Of course time accelerates when the time acceleration option is selected.
@DrB – A while back, someone posted the idea that the initiation of a SLOP caused the 18:22 radar return. That means we have 3 major changes around this time:
A change in direction
A change is speed
Turning the AES back on
This would indicate a change in strategy (unsuccessful negotiations, etc.) or a change of control from one person to another. It’s also possible that the original PF regained control later on but didn’t disable the AES. The change in control is one of the few explanations for the AES reboot. I wonder if the escape hatch could be breached after an hour or so?
A few of us had calculated a weight of 210 mt (so 35.6mt fuel) at 18:22. How accurate is this using your fuel model?
As for all of the calculations, I wonder about the accuracy of the following:
The FCOM gives methods for determining the “break-even” speed. In other words is it worthwhile to change altitude for a reduction in head winds? That suggests that the winds can vary with altitude (plus they gust and seem to change direction often). If we do not know the flight level during the entire flight, how sure can we be of the winds? Even your 6 kts tolerance could be 30 kts over a 5-hour flight. Correct?
I believe you found that the temperature was 10°C above Standard and this would have an adverse effect on fuel consumption. Does it make sense that the temperature could maintain this offset over the entire flight?
@Victor
We may not agree on interpretations of the data but surely, if we want readers to decide on a truly objective basis, we can agree on the importance of clearly separating fact from assumption, inference and supposition.
@Victor Iannello: Thank you for clarifying the time acceleration option. Does that mean that the time actually spent in that February 2 simulator session could be considerably less than the 1:12 hh:mm between the time stamps?
@TBill
Re: “That is interesting time stamp info for the simulator runs. Are you saying that is recent info to you from ATSB?”
Yes. Subsequent to Victor’s revelation of the roughly one hour “session” time on 12 October I wrote to the ATSB requesting further details. The 15:26 to 16:38 (hh:mm) information was provided to me in their reply.
“Should we be asking for more data from the runs? (which we may or may not be privy to, but at least we would know the response).”
I have; to date I have not received a response.
Now, to explore your observations on timestamping, are you saying that the timestamps might likely refer to the “simulation time” (ie the simulated time of day for the simulated flight) rather than the actual real world time (ie the time at which the user is actually sitting at the PC)? I had (perhaps rather naively) assumed it would be the later.
@Gysbreght: It’s possible that time acceleration was chosen for short periods of time and the actual time passed was less than the duration suggested by the timestamps. We also know that the simulation was paused and parameters were changed, which would have increased the time passed. If the user was moving the position forward, there would be little reason to accelerate time, and there would be reasons to not do it.
@TBill: I think you will see that if you create a flight using Free Flight, the simulation time will default to the current time. There are other conditions where the time defaults to the current time when the program starts. Try it.
@Victor
OK, yes if you start out in the CREATE NEW FLIGHT then you get today’s date. If you run another case first, you get the date of that last case. Of course, the user can change the default TIME/DATE to any time/date.
@Victor Iannello: “It’s possible that time acceleration was chosen for short periods of time and the actual time passed was greater than the duration suggested by the timestamps. “ (My bolding).
Did you mean less rather than greater?
To make a judgment on time acceleration versus time spent in pauses one would need access to the time stamps of the individual locations. I asked the ATSB for those time stamps but their reply indicated only the first and the last. Perhaps worth another try?
@Mick @Victor
“Now, to explore your observations on timestamping, are you saying that the timestamps might likely refer to the “simulation time” (ie the simulated time of day for the simulated flight) rather than the actual real world time (ie the time at which the user is actually sitting at the PC)?”
Yes I believe the info you quoted is the “simulated date and time” of the flight, for example, if I save a case and had set a time in 2014, it shows 2014, or in the future, I could save a 2018 case. As Victor points out, when you create a new flight, the Time/Date default setting is today’s date, or alternately as I found the default time/date can also be populated a prior case you ran. So there is a possibility the first file was saved at approximate real time.
What we would need, that I do not know if we have, is the date/time of the FLT file save.
Another point of interest.
The [DateTimeSeason] field of the recovered files indicated 2 February 2014; 15:26 to 16:38 (hh:mm). On the day the simulation was conducted the PIC was on a rostered day of leave. The following day the PIC was rostered to fly from Kuala Lumpur to Denpasar, Bali and return the same day.
The Shadow Volume that contained the recovered file fragments was dated 3 February 2014. To save the recovered fragments of deleted Flight files in the Shadow Volume on 3 February 2014, the computer must have been operating and an application must have been writing new files to blocks of disk space that had been allocated to the previously deleted files.
Odd.
@Gysbreght
As far as time acceleration in FS9:
The main thing I have noticed is the observed turn radius is wider as you go faster sim rate, and also you can overshoot and make ugly wide turns with oscillations. So I always try to take turns at 1x. Any serious work, if I am working up a real flight path, would have to be verified in 1x speed. Also I always try to move back to 1x before saving a case.
Other than turn radius and general loss of stability, as far as I know the clock and other functions operate well in accelerated time.
@Victor. You say of a piloted recovery in your simulator”…and it’s easy to rip the wings off the plane on the recovery”. Literally or do you mean exceed the flight envelope? What indicates that please?
Pilot black out possible?
@David: High g-loads on wings.
@Gysbreght said: To make a judgment on time acceleration versus time spent in pauses one would need access to the time stamps of the individual locations
No. Neither time acceleration nor pausing has an effect on the relationship between position and simulation time. Manually changing the position does.
(I changed “greater” to “less” in a previous comment.)
@Victor
@Gysbreght
Re: ”After a double engine flameout, there is absolutely no way that a proficient pilot would provide a nose up input (which would be required) and get to stick shaker speed. Nor is a recovery from a 15,000 fpm descent followed by an efficient glide as trivial as “pushing the nose down then pulling it up”. That would be a challenge even for an experienced pilot.
(@Andrew: Do you think a novice could easily accomplish this?)”
I didn’t respond earlier as it was way past bedtime in my part of the world.
Let’s say a low speed event did occur at some point. The recovery from such an event doesn’t normally require an extreme nose-low attitude. However, a novice pilot (or even a startled experienced pilot) might disconnect the autopilot (if it were engaged) and ‘shove the nose down’, causing the rate of descent to increase rapidly over a short time interval. If the speed was low at the beginning of the upset, then a novice pilot might still be able to effect a recovery if the duration of the upset were short and the speed brakes were deployed during the recovery. Unfortunately I don’t have a simulator at hand to try it out.
As Gysbreght noted, the aircraft speed stabilised just above the stickshaker during the simulator experiment. I think he has a valid point that a startled or novice pilot might over-react, as described above. However, that is difficult to reconcile with the estimated one-minute time interval between the second flameout and the log-on request at 00:19:29. One minute is an awfully long time to sit there and do nothing, assuming the low speed was recognised immediately after the autopilot disengaged following the second flameout.
@TBill
@Victor
@ventus45
@Rob
Regarding the mooted MH150/151 MH168/169 schedule changes.
Bill, I don’t think that those changes were ever implemented. When I skim the historical departure/arrival information for January – March 2014 I don’t see any change to the departure time for MH150 (scheduled departure 1515 MYT) and I can’t find any records for MH168/169 at all.
@Mick
Interesting…
@all,
HERE are two MH370 models – one for BTO and one for BFO. Just click on the link in my pubs list for 31 October 2017 to download a macro-enabled EXCEL file.
You can input the aircraft coordinates, flight level, and temperature deviation, and the BTO model calculates the true altitude and the Calibrated BTO value. There are 21 lines of inputs if you want to compute an entire route. As an example, I put in the values from my CTH route fit dated 2017.02.27.
You can also input the true course, ground speed, and Rate of Climb for each line, and the BFO model computes the Calibrated BFO values.
I also included in one table all the available BTO and BFO data from Inmarsat. For the BTOs, the table shows the “raw” reported values and adds the corrections needed in some cases to generate the “Calibrated” BTOs (which are then directly comparable to the model predictions). I also listed the errors for each BTO based on the DSTG guidance.
The table also includes the “raw” BFOs from the Inmarsat data logs. The first correction is for the channel-dependent bias values. The second correction is for the 16 Hz self-calibration step applied at 00:16. The third correction listed is my estimate of the SDU OCXO warm-up transient correction. Adding all three corrections to the raw BFOs gives the “Calibrated” BFOs. I also estimated the BFO errors by summing in quadrature the noise and drift error, the transmit channel calibration error, and the estimated OCXO transient error.
I included some notes in the data worksheet that explains the corrections and error estimates for BTO and BFO.
@Lauren H.,
You said: “@DrB – A while back, someone posted the idea that the initiation of a SLOP caused the 18:22 radar return.”
In my opinion the radar contacts continued until 18:22:12 and then stopped because the radar maximum range of 250 NM was reached.
You also said: “A few of us had calculated a weight of 210 mt (so 35.6mt fuel) at 18:22. How accurate is this using your fuel model?”
My fuel model predicts 209.45 tonnes at 18:22.
You said: “If we do not know the flight level during the entire flight, how sure can we be of the winds?”
We can’t be sure. What I try to do is find a flight level that, in conjunction with the wind and temperature data, gives an acceptable fit to the BTOs/BFOs and also matches the known endurance within the error in my fuel model. That fit synthesizes tailwind errors and, for constant-heading routes, crosswind errors which I add to the GDAS values. I expect those wind errors to match the GDAS errors if the route is the correct one.
You said: “Even your 6 kts tolerance could be 30 kts over a 5-hour flight. Correct?”
No. The GDAS data gives temperature, wind speed, and wind direction every 3 hours at 1 degree spacing in latitude and longitude, and at several pressure levels corresponding to typical cruise altitudes. I interpolate over all these parameters to get a best estimate along the route. However, the wind errors do not accumulate over time, so your 30 kts is not appropriate. 5.7 kts is the only precise, published measurement of the GDAS RMS wind error. The temperature error is probably < 1C.
I assume the GDAS wind readings over time at a single location have the same error as spatially independent readings made at the same time (both 5.7 kts RMS).
You said: “I believe you found that the temperature was 10°C above Standard and this would have an adverse effect on fuel consumption. Does it make sense that the temperature could maintain this offset over the entire flight?”
No, the temperature deviation is only near 10C near FL350 when close to the equator. At greater latitudes the temperature deviation shown by the GDAS data drops to 9C at 22S, 6C at 33S, and 3C at 34S. The declining temperature deviation means that less fuel is burned late in the flight, and the KTAS at a given Mach is also declining with time.
@Victor. “High g-loads on wings”.
The simulator indicates that or is that an impression?
RE: “…and it’s easy to rip the wings off the plane on the recovery”
One of the shortcomings of a simulator is that the user doesn’t feel the g-forces acting on the real airplane.
Returning to the Shadow Volume:
It is odd that nothing is known about the files that were saved on the simulator on 3 February 2014, nor of the time of day that the Shadow Volume was created on the day the owner was rostered to fly from Kuala Lumpur to Denpasar, Bali and return the same day.
Hi Dr.B. You have been busy! It is a great help to have this index of papers, some of which I am now seeing for the first time.
23 Oct. paper on predicted fuel/range with fuel model 5.4.
This is a tremendously useful summary of candidate speed modes and altitudes filtered by predicted fuel endurance. I had hoped that it might have a somewhat more “discriminatory” conclusion but it appears that there is still a surprisingly wide range of options that are compatible with observed endurance.
Is your MEFE for both engines pooled with PDA1.5% or time of final fuel exhaustion (R)? If the latter, please conform PDAs used for R/L for the purpose of the model.
Please also confirm your assumption on speed/altitude between 1720 and 1829 that underpin your assumed “starting fuel load” at 1828.
I couldn’t see in your tables the “southern latitude limits” [mentioned at the end] that are associated with each scenario.
Dr B. In the above, might it also be useful to list/summarise the range of constant M or KIAS speeds that are endurance compatible over a range of altitudes?
@Victor Iannello: “Neither time acceleration nor pausing has an effect on the relationship between position and simulation time.”
Thank you for your comment. I’m interested in the spacing between the time stamps, i.e. were they about equally spaced, or were some close together with larger gaps between others?
@Gysbreght: We don’t have individual timestamps.
@Gysbreght said: It is odd that nothing is known about the files that were saved on the simulator on 3 February 2014, nor of the time of day that the Shadow Volume was created on the day the owner was rostered to fly from Kuala Lumpur to Denpasar, Bali and return the same day.
Shadow volumes are automatically created by Windows 7 during periods of inactivity, and usually with a spacing of a week or more. If the captain’s computer was running while he was away, it would be a candidate time to create a shadow volume.
@David: There is an option to display the g-loading on the wings as the simulation proceeds. Without gently pulling back, the g-loading exceeded 3.5. This also assumes the flight control mode is secondary, where there is no envelope protection.
If I have time, I will create a video showing this, although I don’t know a realistic sequence of events to simulate. The stick shaker speed is reached after a single engine flameout, A/P engaged, and holding altitude, at which time the nose would automatically drop, and a steady, stable descent is achieved, all with the A/P engaged. If the pilot intervened and lowered the nose, the rapid descent would not have been timed with the log-on at 00:19, which we believe occurred some time after the APU started following the second engine flameout. There is a timing inconsistency of 2 minutes or more, depending on how long the last engine continued to run after reaching stick shaker speed.
@Victor Iannello: “Shadow volumes are automatically created by Windows 7 during periods of inactivity, and usually with a spacing of a week or more.”
I couldn’t find that on the Microsoft site. Can you provide a reference?
If it is true, then what is saved in the Shadow Volume at the time it is created?
Correction to my October 30, 2017 at 10:34 pm comments.
The last paragraph should read:
…that is difficult to reconcile with the estimated two-minute time interval between the second flameout and the log-on request at 00:19:29. Two minutes is an awfully long time to sit there…
That makes it even more difficult to reconcile!
@Victor
Re: “Without gently pulling back, the g-loading exceeded 3.5.”
Just curious, but do you recall the maximum speeds that were achieved with both the high-g and ‘gentle’ manoeuvres?
@Andrew: I would say it is two minutes or more, depending on how long the last engine continued to run after reaching stick shaker speed.
@Andrew: If you have an interest, I will make a video. However, I need to know exactly what scenario I am simulating. Previously, I simulated a dual flameout, and raised the nose to get to stall, then lowered the nose to reach high descent speeds, then recovered. I think is an unlikely series of events.
@Victor
Re: ” I would say it is two minutes or more…”
Yes.
Re: “If you have an interest…”
No, I was just curious.
@DrB
Thank you!
>Why not include the 23:14 Sat Call time slot?
>I like the path end point 34S, 94.15E
Earlier Victor posited “It would be very interesting to know what a realistic expectation is for the coverage per day.”
I expect that the six Hugin AUVs should cover 600km² per day, taking recovery cycles into consideration. The side scan sensors employed by Ocean Infinity’s Hugin AUVs are of similar specification to those deployed with Fugro’s autonomous and towed underwater vehicles: all are Edgetech systems operable at 75/410kHz “full spectrum CHIRP” side scan sonar mode.
Thanks to “TBill”, it now seems that:
MH150 Departed KUL at 11:45am local (UTC+8) = 0345UTC and arrived JED at 1600 = 4:00pm local (UTC+3) = 1300UTC.
This is a flight time of (1300 – 0345) = 0915 = 9hrs 15 minutes.
MH168 Departed KUL at 22:30pm (UTC+8) = 1430UTC and arrived JED 02:45am (UTC+3) = 2345UTC.
This is a flight time of (2345 – 1430) = 0915 = 9hrs 15 minutes.
So, a normal flight time of 9:15 can be assumed, plus mandatory diversion and holding reserves, would give an endurance of nearly 11 hours, so let’s say 11 hours.
11 hours endurance for MH150 (departure 0345 UTC) gives fuel exhaustion at 1445 UTC.
11 hours endurance for MH168 (departure 1430 UTC) gives fuel exhaustion at 0130 UTC.
Note that sunrise at 45S 104E on 4th Feb 2014 was 22:04 UTC.
Obviously MH150 was not a viable flight to reach 45S at or just before dawn, but MH168 certainly was, with 3.5 hours of fuel to spare.
So, for a 1430 UTC departure for MH168, there is only an available flight time interval of (2204 – 1430) = 0734, ie, 7 hours and 4 minutes to reach 45S 104E at dawn.
There is only one realistic flight path that meets that mission requirement, to arrive at dawn.
Using 480 knots TAS as the average ground speed nil wind yardstick, the route is:
WMKK (1430 UTC) – GUNIP (21 min 1451 UTC) – TASEK (32 min 1502 UTC) – MABIX (62 min 1532 UTC) – 45S104E (7h06m 2136 UTC) – YPPH (9h03m 23:33 UTC).
There are two very interesting points about this flight path.
The first is that the flight path Mabix to 45S104E crosses L774 at a point 41Nm NE of UPROB, only 77 minutes into the flight (at 1547 UTC) 15 minutes past Mabix. I will come back to this later.
The second is that the flight can comfortably reach YPPH via 45S104E in 9 hours and 3 minutes, the same time almost as the scheduled 9 hours and 15 minutes flight for MH168 to OEJN JED Jeddah, with standard reserves.
The question must therefore seriously arrise, did Z in fact actually plan for MH168, instead of MH150, with the intention of having himself re-rostered from MH150 to MH168 ? As a very senior captain, he would have more than enough “pull’ to do that, even at pretty short notice, should he have chosen to do so.
Indeed, it is also very interesting to consider the track from Tasek to Mabix, from a “radar” point of view, assuming Z was really wanting to “drop off the scope” so to speak.
The interesting new “fly in the ointment” here, may well be Mick Gilbert’s recent observation, that apparently MH168 never flew. Is it possible therefore, that Z’s “initial plan” was for a flight, that as things turned out, became “The Flight that never was” in more ways than one ?
But, if we take the lesson of the sim data to be that in all probability the intention was in fact to ditch at dawn, then it is a simple matter to apply the same rationalle to the circumstances he faced at Penang, at 17:52, on MH370.
The time available to fly south to dawn from there was his crucial issue. He had no time to go northwest up the Mallacca Strait. He had to fly across Sumatra, and had planned to anyway, through a civil ATC radar black hole, which also turned out (so we are told) to be a military radar black hole as well.
In the context of MH370, it is worth noteing, that Penang (17:52 UTC) direct to Mabix is 346Nm in 44 minutes approximately, arriving at approximately 18:36 UTC.
Similarly, and alternatively, Penang (17:52 UTC) to 41Nm east of Uprob on L774 is only 41 Nm further, at 387Nm in 49 minutes approximately, arriving at 18:41 UTC.
Interesting timeing for a southern FMT, for both cases.
@Mick @Victor @Gysbreght
Re: MH150 on 4-Feb-2014, isn’t it unusual that ZS flew Denpasar round trip on 3-Feb and then Jeddah on 4-Feb?
Mick- I assume you confirm 4-Feb-2014 MH150 flight was 15:15 Departure? Amy further info such as flight path?
[VI to Gysbreght: If you complain one more time about the way this site is moderated, you will be banned.]
@TBill
“Especially for you” Hopefully this link works for MH150 on 4th Feb 2014
http://de.flightaware.com/live/flight/MAS150/history/20140204/0715z/WMKK/OEJN/tracklog
@Ventus45
“But, if we take the lesson of the sim data to be that in all probability the intention was in fact to ditch at dawn, then it is a simple matter to apply the same rationale to the circumstances he faced at Penang, at 17:52, on MH370.
The time available to fly south to dawn from there was his crucial issue. He had no time to go northwest up the Malacca Strait. He had to fly across Sumatra, and had planned to anyway, through a civil ATC radar black hole, which also turned out (so we are told) to be a military radar black hole as well.”
Sorry but I must be missing something. At Penang at 17:52 MH370 had enough fuel and enough time to proceed up the Malacca Strait, turn south at IGOGU at 18:37 onto a heading of 186 and continue on a great circle until fuel exhaustion, AND arrive in time for sunrise. I agree with you on one thing: it is quite plausible that Shah planned to synchronize fuel exhaustion with local sunrise.
@TBill: I don’t see anything unusual in the rostering of ZS. Perhaps Andrew could answer.
@Victor Iannello: Is it permitted to inform readers that my delayed comment has appeared?
[VI: Comments that I deem to be informative, collegial, sincere, and relevant are permitted.]
@Don Thompson
Seabed Constructor has finished its current swing testing the six OI AUVs and is moving off to the Bahamas. No sign yet of it returning to its earlier ports of call in the UK, or deploying closer the MH370 search area.
@Joseph Coleman
Thank you so much Joseph! I certainly had you in mind. You saved me $$, amd I know @ventus45 is looking for that data
OK well that estimated path looks like DUBTA on P574 sky vector to me pending overlay on SkyVector. So I am saying the Simulator case was possibly DUBTA divert to DOTEN, and possibly the 15:26 time in the sim studies was inserted as realistic take off time for Flight MH150.
@ventus45, @TBill
Concerning the announcement you have quoted, dated Aug 2013, referring to the seasonal introduction of MH0168.
An MH0150 typical departure from KUL was 07:35UTC (15:35MYT) arriving JED 16:20UTC (19:20 local), that’s ATD/ATA.
MH0168 was announced as an additional seasonal service, to operate only 28JAN14 to 29MAR14. Its departure from KUL would have been 17:15UTC (23:15MYT) arriving JED 02:45UTC (05:45 local).
However, according to the November 2013 MAS timetable MH168 never made it as an operational service.
Hi Richard,
Thank you for that information. I’d checked earlier but missed it leaving the assumed trials area.
@Don Thompson
Doesn’t the Nov-2013 MAS Time Table seem to show MH150 changing to the proposed new departure time on 28-Jan-2014? but it seems like that change did not happen either.
TBill,
The timetable for 01 Dec 2013 thru 29 Mar 2014 contradicts the August 2013 announcement of the MH168 additional daily service KUL/JED during February and March 2014.
The recorded actual times of departure/arrival for MH150 KUL/JED after 28th Jan 2014 contradicts the timetabled KUL departure of 1145MYT, it remained as 15:15MYT
It appears MH168 did not make it onto the schedule during the period of interest.
@Don Thompson
Agreed!
@Paul Smithson,
You said: “Is your MEFE for both engines pooled with PDA1.5% or time of final fuel exhaustion (R)? If the latter, please conform PDAs used for R/L for the purpose of the model.”
I compute the L and R PDAs and use them separately to predict L and R engine fuel consumptions and flame-out times. The values are shown at the bottom of the table “Output Parameters” on the “Endurance Model” worksheet (0.45% L and 2.55% R).
You said: “Please also confirm your assumption on speed/altitude between 1720 and 1829 that underpin your assumed “starting fuel load” at 1828.“
For the endurance/range plot, I assume the period from 17:22 to 18:29 (Leg #2) was flown at M0.84 and FL340, as shown in the “Input Parameters” table in the “Endurance Model” worksheet.
You said: “I couldn’t see in your tables the “southern latitude limits” [mentioned at the end] that are associated with each scenario.“
The “southern latitude limits” are not listed explicitly in this table, but you can compute them if you want by using the polynomial equation shown at the top of the figure entitled “Range from Earliest FMT Position to 7th Arc” in the “Endurance vs Range Plot” worksheet. This uses the post-18:29 Range shown in the table in the same worksheet. You can also estimate the latitude limit by simply reading the curve in the figure. The southern latitude limits are also plotted in the MEFE/Range figure at the top of this worksheet.
Later you said: “Dr B. In the above, might it also be useful to list/summarise the range of constant M or KIAS speeds that are endurance compatible over a range of altitudes?”
Good idea. I’ll explore those two options and post my results later.
[Gysbreght is banned.]
@Victor,
@Gysbreght,
Re: “Gysbreght is banned.”
Gents, come on. Sometimes discussions are getting hot, often we have disagreements, but as long as they are not turning into personal attacks or fights of gut feelings, I think they should not be banned.
@Gysbreght,
Sometimes it is better to let it go.
@Victor, please don’t become JW-II. You know that your blog is unique, and it is focused on the science and engineering. I think Gysbreght has provided many valuable thoughts and analytics, so I sincerely hope you will calm down and take a right decision only after that. It is easy to break, but difficult to repair.
@TBill
Regarding MH150 on 4 Feb 14.
Yes, Flight Stats says it departed at 1514 MYT, one minute before its scheduled time of 1515 MYT.
Regarding the Captain’s cluster of flights on 3 – 6 Feb 14.
Such clusters were unusual but certainly not unique. He had two not dissimilar clusters in mid-2013.
6 Jun 13 – MH102/103 – KUL-DAC-KUL
8 Jun 13 – MH2 – KUL-FRA (Early AM departure, duty starts late PM 7 Jun 13)
9 Jun 13 – MH3 – FRA-KUL (Arrives KUL early AM 10 Jun 13)
6 Jul 13 – MH851/850 – KUL-DPS-KUL
7 Jul 13 – MH190 – KUL-DEL
8 Jul 13 – MH191 – DEL-KUL (Arrives KUL early AM 9 Jul 13)
Andrew is undoubtedly far better placed to comment on the vagaries of rostering.
Thinking loud. 18:25:27 BFO. If Holland’s curves always exhibit such a monotonic behaviour as presented in his paper, then BFO 18:25:27 could potentially be explained by a rapid descent occurring simultaneously with the SDU boot-up, and subsequent recovery 7 seconds later. Of course, such an explanation does not eliminate the elements of coincidence, but at a glance it appears to be physically possible.
@Tbill
@Mick Gilbert
http://uk.flightaware.com/live/flight/MAS150/history/20140204/0715zZ/WMKK/OEJN
According to this, Gate departure was 15:14 and take off was 15:29
@Tbill
@Mick Gilbert
http://uk.flightaware.com/live/flight/MAS150/history/20140204/0715zZ/WMKK/OEJN
According to this, Gate departure was 15:14 and take off was 15:29
Rob says:
October 31, 2017 at 12:35 pm
@Ventus45
“But, if we take the lesson of the sim data to be that in all probability the intention was in fact to ditch at dawn, then it is a simple matter to apply the same rationale to the circumstances he faced at Penang, at 17:52, on MH370.
The time available to fly south to dawn from there was his crucial issue. He had no time to go northwest up the Malacca Strait. He had to fly across Sumatra, and had planned to anyway, through a civil ATC radar black hole, which also turned out (so we are told) to be a military radar black hole as well.”
Sorry but I must be missing something. At Penang at 17:52 MH370 had enough fuel and enough time to proceed up the Malacca Strait, turn south at IGOGU at 18:37 onto a heading of 186 and continue on a great circle until fuel exhaustion, AND arrive in time for sunrise. I agree with you on one thing: it is quite plausible that Shah planned to synchronize fuel exhaustion with local sunrise.
@Rob
I wrote that before going to bed in the early hours after a long day. I have just had 7 hours sleep and one coffee while reading the posts since.
You are not missing something.
It is I who forgot to put in what is a crucial point in my view, and it is this.
The sim is telling us that he wanted to get to 45 South.
He chose 45S in the first place for a very specific reason in my view.
That reason, was to get into the circumpolar current to ensure there was next to no chance that any debris would ever wash up anywhere that it would be found, recognised, and recovered, and give the game away.
The plan was “to vanish”.
So, my reasoning is this.
From Peneang at 1752 he had fuel remaining for (00:17 – 17:52) = 6 hrs 25 minutes, which equates to remaing range of about 3,080 Nm.
Getting to Mabix (direct) would take him until 18:36 and he now has fuel remaining for only (3080 – 346) = 2734 Nm. Even going due south from Mabix only gets him to 42.5 S, still 2.5 degrees latitude or 150 Nm short of his “goal” or target.
Alternatively, going via Medan direct to 41 Nm East of Uprob on L774 gets him closer to his goal.
Getting there costs him 49 minutes and 387 Nm. He now has fuel for (3080 – 387) = 2693 Nm. Going due south, that gets him to about 43.6 South, now only about 85 Nm from 45 South, which is well within glide range.
But, on the night of the 7th / 8th March, the terminator for his “new” fuel exhaustion time of 00:17 has moved west, so he will burn a few miles getting there, but no many, and he still ends up well south of 40 in theory.
So I think the only solid thing we can infer from the simulator, is that his initial PRIMARY goal was ditch at sunrise. His SECONDARY goal was to do that as far south as possible, preferably at 45S or further, if he could make it, but anything south of 40 would be acceptable.
If he had flown from Penang to Igogu, he would arrive there at 17:52 + 00:46 = 18:38 UTC for 375 Nm, leaving him a range of (3080 – 375) = 2705 Nm. Even if he then flew due south, that gives him 45 degrees of latitude range, from Igogu at 7.5 N, which means he ends up at only 37.5 South at best, which puts us pretty much rite back to square one, where we have been looking for three and a half years, with no result.
If we accept that the simulator is telling us that 45 south was most probably the goal, coincident with sunrise, for the ditch, then quite clearly, N571 and all that it entails, is “out the window”.
@Oleksandr
Interesting.
Tangential note.
There has been a feeling expressed here that interest in MH370 amongst the general public has waned to the point that it no longer cares. I have thought that way too. On occassion I give talks to various audiences on MH370, always replete with lots of slides and narrative. By now, three years after the event, I figured that the story had run its course. Today, I gave a very informal talk without slides, just draowing maps of Malaysia on a chalk board. The audience was what I would call a scientific general public. The interest was intense. Afterwards, I attended a completely unrelated function. Two people, one a very well know scientist and a former division head at the NSF, the other a former director of a major national laboratory, both expressly regretted having missed my talk. I had clearly captured their attention.
Conclusion – the interest may be latent, but it is still there.
Have slides, will travel.
@all,
HERE is a new Version 5.5 of my fuel model that now includes endurance and range estimates for the cases of constant Mach and constant KCAS after the earliest possible FMT. Just click on the link in my pubs list for 31 October 2017 to download a macro-enabled EXCEL file of Fuel Model V5.5.
Here are the cases with acceptable endurance (within 8 minutes of 00:17 UTC):
M0.84 None
M0.82 FL370-410
M0.80 FL350-410
M0.78 FL340-360, FL410
M0.74 FL310-320
275 KCAS FL250-330
250 KCAS FL390-400
You can compare all the results in the new Range versus Endurance plot. I also added the southern latitude limit to the summary table.
@Don Thompson
On rechecking Seabed Constructor’s course and declared arrival time, destination is the Freeport in Bermuda (not the one in the Bahamas). Much closer to the area it has been working.
Dr B. Many thanks. That is really useful.
@Victor. “I don’t know a realistic sequence of events to simulate”. No, not in the way of MH370 descent replications, what with software limitations at flight extremes and the ‘unawareness’ of tank residual fuel in this and all simulators.
However about the pilot recovery instance you describe, the ‘g’ gives a general measure of the loadings on the aircraft when pitching up, pilotless. The NTSB in the 737-300 Silk Air investigation said of that dive, (which I take as wings level), without the use of nose down trim, or large control column forces and inputs, “the airplane would have initiated a return to a nose-up attitude due to its inherent stability.”
That sounds similar, yet both appear inconsistent with your June BFO replication. There, at the end, without a spiral and at high speed the aircraft showed no sign of pitching up.
Any thoughts?
@Ventus45
Thanks for explaining. I have a clearer picture now. I’m with you on the sunrise, and the advantage of reaching the circumpolar current, However, I am going to stay with the Malacca Strait route, following the FIR boundary, FMC at 18:37 in the vicinity of IGOGU and crossing the 7th arc at about S37.7, E89 (close go the DSTG hotspot) particularly as DrB’s fuel model indicates that constant Mach 0.81 at between FL350 and FL390 would have sufficient endurance.
If Shah wanted to get as close as possible to the circumpolar current, and the argument for it is persuasive, he would have needed every ounce of his gliding skills to do it. There is the problem posed by the final BFO’s. But bear in mind he was flying into the northern fringe of the Southern Jetstream and encountering a significant headwind as fuel exhaustion approached. If he had been cruising at say 40,000ft, a dive to a lower altitude after fuel exhaustion could make sense. He would be gaining kinetic energy at the expense of potential, and reducing the headwind component.
Do you catch my drift? Pardon the pun.
@David: I’m not sure I understand your question. The only way I produced the BFO values was with either a banked descent or nose down input. With wings level and no pilot input, a damped phugoid would develop, with the initial amplitude dependent on the initial conditions.
@DrB: That’s really useful. It’s too bad the DSTG in their Bayesian analysis did not constrain speeds in a more realistic way.
@All: With preparations for the re-started search ongoing, now is the time to succinctly express your views about where to search after the initial 25,000 sq km.
@Rob
What I understand from @DrB’s numbers max. reachable latitudes depend on where the FMT was.
He took the earliest possible FMT at 18:22. Straight from there max. reachable latitude towards the SP would be a lot farther south than a FMT later on at say 7N and maybe a loiter of some kind included after 18:22.
I think it’s not likely the FMT took place at 18:22 and so the max. reachable latitude from there is not likely either.
If I misunderstand this I’m glad to hear.
@VictorI
On a simulation-proposal I was thinking about ChinaAir 006 again. This 747 plunged vertically from 41.000ft to 9.600ft where the pilot pulled out with 5.1G’s strain on the plane.
I know it’s not quite comparable for this plane had engines working but I assume the pilot did not use engine power to recover from this dive.
I imagine it could be usefull to see how this pilot managed to recover from a same kind of descent rate/dive as MH370 seems to have made without distroying the plane.
@Ge Rijn: I didn’t say it was impossible to successfully recover from a high rate of descent. I said it would be challenging, and not as easy as simply pulling the nose up, as one person implied.
@victor. My question stemmed from your comment to Gysbreght:
G: “I’m sure you would not find it [recovery from a 15,000 fpm descent] difficult at all if you tried it in your simulator.
V: “Your certainty is misplaced. It’s challenging, as I said. It’s hard to keep the wings level, and it’s easy to rip the wings off the plane on the recovery”.
It was the recovery ‘g’ there I was talking about, which later you described as, “Without gently pulling back, the g-loading exceeded 3.5.” ie a pitch up as I understood it, without pilot input.
@DrB: Regarding your calculations of the maximum southern latitude for various scenarios, some questions:
1) For the 1 hour hold + LRC calculations, at which position did you start the flight south?
2) Did you consider roll modes and compatibility with the BTO (or BFO) values? Or were great circle paths considered without regard to the BTO?
Thank you.
@David: Now I understand the confusion. I meant that with pilot input, it was necessary to pull up slowly. Pulling up rapidly produced high g-loads.
@Victor. Thanks. I see now.
In the 737 Boeing simulator software from the report, “the
nonlinear mathematical software has been validated up to Mach 0.87, and extrapolated using computational data up to Mach 0.99”.
It is unclear at what Mach this, “nose-up attitude due to its inherent stability” would apply to the 737 but in MH370’s case a deal later than would apply at early recovery from the BFO descent rates anyway I would think.
@VictorI
I know you did not say it was impossible but it would be challenging to pull the plane out of such a descent rate.
ChinaAir 006 shows this was surely not simply done by just pulling the nose up. They had to level the plane first to the horizon which was not visible till 11.000ft and then pull out. They accomplished this within ~1500ft (11.000ft to 9500ft) from a nearly straight spiral dive to horizontal/level agian.
All engines had flamed out early in the high speed descent btw.
They restarted after leveling out at 9.500ft.
I would be curious how MH370 would have behaved in same kind of circumstances and what would be needed to level the plane and pull it out in time.
Would this be possible regarding the descent rates the BFO’s show and what altitude space would be needed?
A replicated simulation with a B777-200 could give some more insight I’m thinking.
@David, @Ge Rijn: Again, if there is interest, I can run a test case. I need to know exactly what we are simulating. For instance, the airspeed at the time of recovery will greatly impact the wing loading.
@VictorI
I like to add ChinaAir 006 shows an incident where a plane the size of a B777 virtually runs out of pilot control at high altitude with engines flaming out shortly in the dive-descent. It’s mainly on it’s own and free falling till 11.000ft when the pilot can see the horizon and can level out again.
From 41.000ft till 11.000ft it is in fact an unpiloted plane.
So possible similarities with MH370’s end-flight scenario can be made imo.
I just think a comparing simulation with the available MH370 data could give more insight in possibilities (and impossibilities) regarding MH370.
@GeRijn
I think you are possibly misunderstanding the situation.
Firstly, although I am yet to pore over DrB latest endurance figures, I believe he takes FMT to be nominally at 18:29, not 18:22.
I’m glad you asked out the endurance question. I have already established to my own satisfaction that if MH370 flew at constant M0.81, at FL350, then it would comply with the BTO, ie cross the arcs at the right times, an ndd reach the 7th arc at S37.6, E89.1, at UTC 00:19:30, IF it had enough fuel to make the journey. DrB’s latest results suggest that it did have enough fuel, give or take a few minutes.
So what I’m now saying iro the M0.81 constant Mach great circle from IGOGU to S37.6, E89.1: “If it looks like a duck, quacks like a duck itand walks like a duck, it probably is a duck”.
We now have a reason why Shah glided as far south as possible; to get a close to the circumpolar current as he could, (credit to Ventus45 for that idea)
It makes S35 look less and less likely.
@VictorI
I sure would be interested ofcourse in the simulation I propose (and other simulations too ofcourse).
I answered before your latest comment…
ChinaAir006 reached VMO speed in the dive and pulled ~5G’s in the pull out. The plane just survived but loosing landing gear doors and parts of the elevators..
Maybe that’s a limit for a simulution?
Maybe it’s an idea to let the plane (MH370) spiral out of control after flame out at ~30.000ft asif no pilot was at the controls and then try to recover before ~9000ft?
@Rob: So what I’m now saying iro the M0.81 constant Mach great circle from IGOGU to S37.6, E89.1
My calculations say that to cross the 7th arc at 37.6S latitude on a great circle, BTO constraints require a speed of around M0.83 at FL350 starting at a latitude of around 0.4S at 19:41. There would not be sufficient endurance at this speed.
Can you please show you arrived at 37.6S latitude at this (M0.81) speed?
@Rob
You’re right about the 18:29 instead of 18:22 as a most early FMT @DrB mentioned. Sometimes those close numbers get mixed up a bit by me. Awfull ofcourse but in this case not quite relevant imo.
I’ll await @DrB’s anwser to @VictorI’s question few posts back on this subject.
@Rob
The endurance/range issue was initiated by me but @Paul Smithson picked up on that further for he stated there was no significant difference between them while I was arguing there was. There the discussion started on this issue.
I’m glad @DrB picked up on this so detailed and pointed out range and endurance are two quite different factors.
You can circel for 7 hours and not make much range at all.
But this is the beauty of discussion.
@sk999
“Conclusion – the interest may be latent, but it is still there.
Have slides, will travel.”
Thank you for that perspective. To gather support for the search, we should have an MH370 speakers list, and use your slides. Perhaps Malaysia realizes that dropping the search would not solve the problem in the public eyes.
Technical BFO question sk, I was reading some of your 2014 posts and you had offsets that you added to the BFO’s. Are you still doing that? or have the BFO numbers been updated such that no offsets are needed?
@GR. Yes, you were quite right (and I was wrong) that the one-engine range is not better while the endurance is. Having said that, I do understand the difference between the two.
@Rob. 37S is still on the N side of the large borderline area between SIO gyre and circumpolar proper. If you come down there debris ends up on S Aus and NZ. In absence of anomalies due to local eddies, this holds all the way through to about 45/46S. So “aiming for the circumpolar current” won’t cut it, in my view.
@VictorI
Your comment:
@All: With preparations for the re-started search ongoing, now is the time to succinctly express your views about where to search after the initial 25,000 sq km.
With all latest evidence and indications (simulator, debris, drift analysis etc) I think it’s essential OI takes a deliberate goal into account the PIC took the plane and wanted it to vanish with the at least possible evidence. He succeeded till now.
This motivation has to be taken into account.
Just outside the CSIRO priority area east and west of ~32.6S there are deep trenches. Especially to the east. The pits there are ~6000m deep.
This region has to be searched imo.
All signs are the flight was a planned one with a necessary calculated end imo. A deep trench would serve all those objectives.
@sk999
Conclusion – the interest may be latent, but it is still there.
Curious post relative to “scientific interest”. My experience has been almost the polar opposite – I found very little interest from the get-go. Most of the people I know who have worked in navigation and understand time of arrival and Doppler looked at the issue long enough to understand it (about 10 minutes) and the data we have to work with, and said the solution was indeterminate, and that it was a waste of time to screw around with it. That was my conclusion as well, but being retired I had nothing better to do.
Reality has come home to roost. We have no convincing analytics relative to a precise location after more than three years and a lot of looks.
@Victor
“@All: With preparations for the re-started search ongoing, now is the time to succinctly express your views about where to search after the initial 25,000 sq km.”
I assume the 25,000 sq km includes the green space in Fig 3, which I support because I think that should have been the first place to search.
After that I would simply continue up Arc7 in the same manner. It is OK if they jump around to certain “hot spots” eg; 26.9S, but the idea would be to cover Arc7. They should NOT skip the hard part of Broken Ridge. I feel Inmarsat made an enormous contribution defining Arc7, and that effort deserves follow-up.
Miscellaneous points:
> I have an issue with the *assumptions* in the models (Bayesian/fuel) to help define the search area. The reason Arc5 to Arc6 is hard to fit could be due to pilot actions (due to sunrise, winds, sat calls, etc) and descent/slow down/change of track/fuel saving steps, are all possible there, in addition to late FMT after 18:40 (which we tend to say is “loiter”). I assume an open-minded Bayesian model just says somewhere on Arc7, but if there could be an “un-opinionated” model, then it could be useful to have model predictions.
> @ventus45 has a good theory about 45S circumpolar currents, which helps give the ZS simulator cases a possible end-strategy. But I do not currently feel MH370 had enough fuel to accomplish that. So there must have been an alternate strategy, and I feel 180S past Broken Ridge at 33-35S has many of the same attributes as 45S. I am currently trying see if I can import the NASA Perpetual Ocean data to Google Earth (someone did this mapping already), so I can see if there are logical flight paths to the ocean currents as known in 2014.
@Victor
Re “My calculations say that to cross the 7th arc at 37.6S latitude on a great circle, BTO constraints require a speed of around M0.83 at FL350 starting at a latitude of around 0.4S at 19:41. There would not be sufficient endurance at this speed.
Can you please show you arrived at 37.6S latitude at this (M0.81) speed?”
Victor, I’ll be happy to do that for you tomorrow, when I can get to a PC. I’ve had a rather busy day and am getting in some much needed “quality time” with myself, walking along the river Medway, in the heart of the Kent countryside. I hope that doesn’t sound too pretentious, I wasn’t meant to. 😊
@Paul Smithson ” 37S is still on the N side of the large borderline area between SIO yre and circumpolar proper. If you come down there debris ends up on S Aus and NZ. In absence of anomalies due to local eddies, this holds all the way through to about 45/46S. So “aiming for the circumpolar current” won’t cut it, in my view.”
Paul, isn’t that being just a bit picky?
Let’s say that he actually got to S38 in the middle of the “hotspot”. Then a 100Nm glide would get him closer to S40, wouldn’t it? A glide would explain why the ATSB couldn’t find him. A flight path parallel to the sunrise terminator would solve the problem of synchronising fuel exhaustion with sunrise.
What’s not to like, apart from the revised CSIRO drift studies? I have little faith in them myself, considering how they seem to have been made to support the ATSB after they failed to find the aircraft where they assumed it would be.
The ATSB told me in late 2016, that S40 was possible, according to CSIRO.
Then there’s the previously discussed evidence from the debris,
@Rob
A glide would explain why the ATSB couldn’t find him. A flight path parallel to the sunrise terminator would solve the problem of synchronising fuel exhaustion with sunrise.
What’s not to like, apart from the revised CSIRO drift studies?
A great many things can explain the ATSB failure to find the aircraft. Why settle on your explanation? Why even bring it up? It does nothing to favor your hypothesis.
What’s not to like is the absence of debris in WA, and the incompatibility of your terminus with the simulator data. Why not pick a terminus compatible with both? Seems like a much stronger argument to me.
…and so it goes…
http://www.themalaymailonline.com/malaysia/article/not-true-mh370-search-mission-to-continue-next-year-transport-minister-says#m6wVeAfoLMD7D67t.97
As I said, it is senseless to speculate at this time.
@DennisW said: As I said, it is senseless to speculate at this time
Yup.
@TBill – As an alternative to Victor’s McMurdo, I really like continuing the track away from MEKAR and then turning to a CMH or CTH of 180° south. Could it be possible that any speed and/or direction change necessary to meet the BTO & BFO data is due to the winds?
A reprint of my 10/01/2014 post:
Per the ATSB report, Inmarsat used the March 7, 2014, BFO data to predict the flight paths of MH021 and MH009. Presumably, they were able to calibrate their BFO model to the known actual location of each of these flights. At the point where these flights crossed the 7th Arc, their predictions were off by 145nm and 200nm, respectively.
Is there a reason why current models are believed to be more accurate?
In using the sim data to correlate the accident flight, might it be possible that ZS had planned to divert the Feb. 4 flight to Jeddah but changed his mind somewhere along the Straight of Malacca and later chose MH370?
@Lauren H asked: At the point where these flights crossed the 7th Arc, their predictions were off by 145nm and 200nm, respectively. Is there a reason why current models are believed to be more accurate?
Any specific location is based on other assumptions such as the navigational mode and whether there were pilot inputs. The DSTG model used a prior distribution for number of manoeuvers that biased the results towards favoring straight paths. Any prediction is only as good as the underlying assumptions, including prior distributions.
@Lauren
In using the sim data to correlate the accident flight, might it be possible that ZS had planned to divert the Feb. 4 flight to Jeddah but changed his mind somewhere along the Straight of Malacca and later chose MH370?
I think a planned diversion on a flight to Jeddah is unlikely for a couple of reasons:
1> A three man flight crew.
2> The diversion from IGARI provided an extra hour of negotiating time. A negotiation could only be started after the diversion was known and verifiable. A Jeddah flight removes that extra hour.
People here who think Shah committed suicide and wanted to find the best “hiding place” for the plane in the nooks and crannies of the SIO are in the pathetic category. As are the various mechanical failure affectionados. The plane was diverted for a reason. Some day that reason will come to light.
A diversion on a flight to Jeddah simply does not work.
Oleksandr says:
“Sometimes discussions are getting hot, often we have disagreements, but as long as they are not turning into personal attacks or fights of gut feelings, I think they should not be banned.
@Victor, please don’t become JW-2. You know that your blog is unique, and it is focused on the science and engineering. I think Gysbreght has provided many valuable thoughts and analytics, so I sincerely hope you will calm down and take a right decision only after that. It is easy to break, but difficult to repair.”
As a longtime silent reader, I’d just like to echo Oleksandr’s thoughts, which reflect mine.
What I like so much about your blog, Victor, is
(1) its calm, informational, scientific, factual nature, largely void of personal attacks
(2) that unlike JW you don’t censor thoughts that are different from your own, but instead you are very tolerant and allow a very free flow of discussion and information, which makes your site an extremely interesting one with a large spectrum of theories and investigative approaches. Thank you for that.
I think (2) creates a positive conversational atmosphere (unlike @JW),
which leads to –> people collaborating instead of attacking each other,
which leads to –> (1).
In the early days, JW was like that, too. But then he started becoming evermore authoritarian and ban more and more people (for nothing else than differing opinions, honestly) until almost noone from the early days was left. Luckily you started your blog, so that most of the homeless JW contributors could find a new home, provided by you. Your work is an important part of finding MH370.
I don’t know what Gysbreght wrote (since you redacted it) and don’t want to get into your way, Victor, but I just hope that in the future you won’t go down the same path as JW. It started in the same way.
@Lauren H
I used M0.84 until Arc5 where I see a slow down and descent. But I was using uncorrected standard air temp, so that’s one thing I’d need to go back and fix. (I previously shared a preliminary 180S CTH path from ISBIX).
Mathematical basis: if you put the Arcs 3/4/5 on Google Earth, you are looking for the place where a vertical (180S) line has the same speed between Arc 3 and 4 and between Arcs 4 and 5. There is no exact place where the speed is equal for strictly vertical lines, but due to the winds, the Arc 3 to 4 line is actually a little slanted.
@Victor
The DSTG model used a prior distribution for number of manoeuvers that biased the results towards favoring straight paths. Any prediction is only as good as the underlying assumptions, including prior distributions.
Yes for sure. A Bayesian analysis is meaningless without solid priors. Why the DSTG took this approach without any solid priors (previous scheduled flights do not count as meaningful priors) is a mystery to me as is their statistical approach to the BFO characterization from previous flights.
My SO (PhD EE MS Math) happens to be an excellent mathematician (unusual for a PHD EE) with over a decade of GPS industry experience (in the trenches not working a budget). She looks at me like I have just fallen out of a tree when I am working on MH370.
@DennisW “Reality has come home to roost. We have no convincing analytics relative to a precise location after more than three years and a lot of looks”.
As a Dennis “Whacko” I came up with a location three years ago that had nothing going for it except that it was under pilot control and fitted the BTO’s perfectly.
Every time I raise my head above the parapet I get knocked back and yet I have frankly not seen anything more convincing.
I can only continue to make my case because an unbiased observer would, in my opinion, have to acknowledge it has as much merit as anything else that has been put forward.
@Perfect Storm: There is a minimum level of decorum required for collegial discussion. When somebody has been repeatedly warned and cannot meet that requirement, and in fact chooses to deride the minimum process of review, it is time to make a change. I could cite specific instances, some very recent, of disruptive behavior. However, I think there are more valuable things to discuss.
@TimR
I have not kept a list relative to the whacko category. I actually view it in the same light as autism. We are all affected by it to some degree. The clinical attention to it has served the raise the frequency of diagnosis. Not the actual prevalence of it (despite what my Cali anti-vaxxer friends believe). The recent wave of sexual harassment revelations is another example of the same affect. I expect to be accused of it any day now.
@DennisW
It is a privilege to be rated by Dennis as a “whacko”.
I wear it with pride, it is a badge of honour.
@DennisW “What’s not to like is the absence of debris in WA, and the incompatibility of your terminus with the simulator data. Why not pick a terminus compatible with both? Seems like a much stronger argument to me.”
Dennis, I don’t believe you’re naive enough to think Shah would leave any really useful evidence on his SIM hard drives. The data he left accidentally on purpose on his computer was intended to show us that he was toying with the idea of taking a plane into the SIO until the fuel ran out, flying in LNAV. The smoke without the gun.
And the no debris on WA shores argument is based on supposition, not science. Before the flaperon a arrived at Reunion, CSIRO were predicting that debris should be make landfall in Indonesia. When that was shown to be nonsense, they publicly admonished themselves for being so inept. The “expert” drift modelers have their methods, but the Indian Ocean obviously has it’s own, and it’s keeping it to itself.
CSIRO have now been able to pinpoint a number of prospective termini, with unparalleled precision! I rest my case on he drift modeling fiasco.
@Rob
You are taking the simulator data to the next level of abstraction. That is a dangerous thing to do, IMO.
I agree CSIRO has a checkered history relative to the drift analytics, but just about everyone concurs with the WA location for debris originating as far South as you hypothesize.
@Dennis,
Re: “Interesting.”
Yea, that is interesting. My estimates show the downward velocity component of 30 m/s (5900 fpm) required to explain the difference of 131 Hz (=273 Hz – 142 Hz). I think it is quite realistic to recover 30 m/s speed over 7 seconds. But I am sure, ALSM will not like this explanation.
@Oleksandr
Yes, I nominally use 4Hz per m/sec. I don’t have any experience on the recovery time.
@Rob,
“ to think Shah would leave any really useful evidence on his SIM hard drives”
What gives you the impression that Zaharie had a depth of knowledge of Windows 7, the NTFS filesystem, and its arcane file system state recovery capabilities, in order to accidentally, on purpose, delete files but leave breadcrumbs? You’re implying professional system administrator level knowledge.
“ no debris on WA shores argument is based on supposition, not science”
You’re drifting into unfounded opinion there.
“ CSIRO were predicting that debris should be make landfall in Indonesia”
It was not CSIRO staff but GEMS, engaged as one of six contributors to AMSA’s initial drift work, who made the error that resulted in the call for debris observation on Indonesian shores.
“ pinpoint a number of prospective termini, with unparalleled precision”
Next time you’re out at the mouth of the Medway, look to the north-east 330km towards Den Helder, NL, visualise a swathe 75km wide (Sheerness to Stansted): that’s the precision. Don’t forget, the ‘X’ is a similar distance out from port as from Sheerness to Trondheim, Norway.
@Ge Rijn,
You said: “I think it’s not likely the FMT took place at 18:22 and so the max. reachable latitude from there is not likely either. If I misunderstand this I’m glad to hear.:
Yes, apparently you did misunderstand. To clarify, what I have said is the following:
1. There is no evidence for a turn after passing VAMPI and before 18:22:12.
2. The BTOs and BFOs between 18:25-18:28 do not allow a southward FMT during this time period.
3. The earliest possible FMT southward would therefore be at ~18:29.
4. The earliest possible FMT southward would also reach the southern latitude limits I calculated assuming a great circle path thereafter.
5. FMTs later than 18:29 or which follow curved paths cannot reach as far south as a turn at 18:29 followed by a great circle path.
@Richard Cole
According to VTS Bermuda, Seabed Constructor due in Thu at 0700 and departing again, Fri 1600.
DennisW says, “My experience has been almost the polar opposite – I found very little interest from the get-go.”
It is hard to predict where a person’s interests lie. I once flew into somewhere to give the talk, and on the way out of the airport started telling the taxi driver why I was there. He was all ears. Totally fascinated. Asked a ton of questions. Became my personal taxi driver for 3 days. Best taxi driver I ever had.
Presentation matters. If you try from the start to flog the ATSB, you will lose your audience. I just try to present the facts as we know them. No finger pointing. It seems to work.
@VictorI,
You said: “@DrB: Regarding your calculations of the maximum southern latitude for various scenarios, some questions:
1) For the 1 hour hold + LRC calculations, at which position did you start the flight south?”
Good question. The answer is that for all cases I used the same position for the start of the southbound path as the estimated aircraft position at 18:29 (7.18N 95.68E). For the case of the 1-hour Hold, my calculated range assumes there was no racetrack flown, and both the Hold speed and LRC were consecutively flown along the same great circle. The range for a 1-hour leg at Holding speed at FL200 is about 309 NM from the initial descent, so you can subtract that from my indicated total range for that case to get a range for the case where the racetrack was flown at the 18:29 location. If the racetrack is flown farther out to the northwest (near VOCX), then the range south will be further reduced by the extra path from 18:29 to the racetrack position. There will be some small errors still remaining because the great circle southward would start from a slightly different position. It’s pretty straightforward to estimate where that occurs on the 7th Arc using my BTO Model. Just find the net range (= my total range less 309 NM and less the distance from the 18:29 position to the racetrack position). Then find the location on the 7th Arc where that range from the racetrack position matches the calculated net range. You can solve this iteratively by picking an approximate latitude, then adjust the longitude by trial and error to get a Predicted BTO = 18,400 at 00:19:29. (As an aside, it is interesting that the 00:19:37 BTO has a smaller error than the one at 00:19:29.) Next find the distance between the racetrack position and the 7th Arc position (with BTO = 18,400). Iterate this process until the calculated distance equals the net range. Then you know the southern latitude limit for a Hold at a fixed position.
You also asked: “2) Did you consider roll modes and compatibility with the BTO (or BFO) values? Or were great circle paths considered without regard to the BTO?
In calculating the southern latitude limit, I only considered range along a path following an orthodrome (great circle). I did not consider other roll modes, nor did I consider consistency with BTOs or with BFOs, nor did I consider tailwinds. The impact of tailwinds is expected to be rather small since the winds were primarily from the east or from the west. The major wind effect is to curve the post-FMT path eastward for constant heading navigation modes.
@ventus45
@Rob
@Paul Smithson
I agree with Paul regarding the Antarctic Circumpolar Current and the unlikely association between it and 45S1 from the recovered simulation data. It is generally accepted that the northern boundary of the ACC is defined by the eastward flow of the Subantarctic Front. In the Indian Ocean that is found no further north than (and generally much further south of) 48°S.
In the recovered simulation data, at 10N in the Andaman Sea the simulation airplane had more than enough to fuel to reach well south of 48°S. The fact that the flight appears to terminate at 45°S is antithetical to targeting the ACC.
With regards to trying to glide to the ACC from even the southernmost point of the fuel constrained 7th arc, not even World Gliding Champion Sebastian Kawa could have gotten the airplane anywhere close. Moreover, if there had been any prior study of the ACC (and you’d expect that there would have been if it was geographical goal of the whole exercise) then it would have become apparent that just missing to north is worse than missing by a long margin with regards to where and when debris is likely to come ashore.
@sk999
Could be an age thing as well. In the “valley” you are pretty much irrelevant at 40 or so.
@DrB: Thank you, although some may have thought you matched the BTO/BFO data between 18:28 and 00:19 when you reported reachable southern latitudes.
Here is one of @sk999’s presentations.
What the Malaysian Transport Minister says. Another version. Now a 5000 sq km aim.
https://www.msn.com/en-my/news/national/malaysia-hoping-to-re-launch-mh370-search-mission-in-january/ar-AAuj7y3
@David: Ignore it. OI will search at least 25,000 sq km. As some reports said, the contract with OI is being negotiated by a group headed by Azharuddin Abdul Rahman, the Director General of the DCA. Prime Minister Najib can of course kill the deal, irrespective of other negotiations. We’ll just have to wait to see whether things go forward.
Here’s a video of what Transport Minister Liow Tiong Lai actual said.
Additionally, reports say that with the technical investigation completed, the focus will be on the criminal investigation, with a report expected next year. I’d be surprised if that adds to what we already know.
DennisW,
I am well over the age of irrelevance. However, I do have a finger on the pulse of many who are under. Make no mistake – MH370 does get their attention. At least down to age 25. If not under.
R.E. “the valley”. Today I guess it means “silicon”. Back many years ago, it meant “San Fernando”. Thirty years from now it will mean … who knows?
It used to be that irrelevance kicked in at age 30. Where did this 40 come from? Age inflation, methinks.
@All
With respect to the subject of “public interest” in MH370.
I have noticed, that among my friends and associates, that the level of what I will call “active” interest has dropped off to zero, but surprisingly, the level of “passive” interest, both among them, and in the wider community, is still very much alive and well, but only if they are “prompted” by some event, ie, if it “pops up in the news”.
By that I mean, that my friends and associates know of my deep interest, since it was a daily subject of conversation up until the end of the search early this year. Since then, they don’t ask me about it any more, or only occasionally, and I don’t tell them what is currently going on (for example, here in this forum), unless asked, but they know I am still burning many hours daily on it. They sort of consider it to be an obsession of mine, indeed, when it does come up, they refer to it, not as MH370, but as “Pete’s plane”.
Recently, within the last few months, I have, on reflection, had three interesting experiences of this.
(1) David Mearns.
David Mearns was in Australia a short while ago, for the HMAS Sydney aniversary memorial ceremonies. As most of you would know, he found HMAS Sydney. He was interviewed on morning TV a number of times and was also doing “book signings” of his latest book “The Shipwreck Hunter” whilst he was here. (I managed to secure a signed copy for myself, and I recommed it, it is a fascinating read). What was interesting though, was that in his TV interviews, he made a point of raising, and talking about MH370, how it was unacceptable that the search had been abandoned etc, that it must be found, and he more than strongly hinted that he stands ready to help, but did not go any further with that. The level of interest from the interviers seemed to be both high and genuine. Later that day, when I arrived at the club, the first thing people said to me was, “hey – did you see that shipwreck guy on TV this morning”, followed by, what happens now, is “he” going to go and look for it, etc.
(2) Charts.
I had obtained some genuine navigation charts (both marine and aviation) of the Indian Ocean for doing some plotting. The trouble was, that they were physically “too small” for what I wanted to do. I wanted them “blown up”, if possible. I consulted one of the guys at the club who is a building inspector of some sort. He deals with plans and blueprints etc on a daily basis. He came in to the club after work about 6pm, as he usually does, so I button holed him. After discussing it for a while, he said “lets have a look”. Since I had the charts rolled up in the back of the car, I went outside and got them. We unrolled them, and spread them out on the pool table. All of a sudden, or so it seemed, we had a crowd form. In the end, I had to give a little “off the cuff” lecture on what this was all about. The upshot, surprisingly, was that those who were NOT previously aware of my work, were the ones who seemed most genuinely interested. So, MH370 is still there in the public mind, not on the surface perhaps, but not far below, let’s say at periscope depth. Anyway, after people drifted away, Mark pulled out his mobile and made a couple of calls. Eventually he came up with a printing company that does some work for him, who he said has a very large scanner that could take these charts, and could do what I wanted. So, now armed with a contact name and number and an address of this place, not too far away at Arncliff, I headed off next day to see what could be done.
(3) Printing Company.
I arrived at the printing company just after lunchtime. I walked in with the charts rolled up and presented at the front counter, which was unattended. I had a look around, it is a small private company, only half a dozen staff as it turned out, but there were printing machines of all types, some quite large, with lots of activity going on. Eventually, one of the staff saw me, and came over to the counter. After the initial “hello” formalities, I began to explain what I wanted. Within seconds, before the end of the opening sentence actually, “so you are the guy Mark told us was coming – come in – let’s have a look”. It turned out that they know Mark quite well, having done work for him for years. Over the course of the next hour or so, they scanned the charts to a very high resolution DPI wise, and brought them up on the computer. The software they have is amazing, and the woman doing it is a wizzard. Anyway, with me perched over her shoulder, we moved the charts around to select the areas of interest, and made new files of those sections, and enlarged them to the max for the printers, which then did the job. We went over to the printer selected, and waited for the first one to slowly come out. It was a very large, very impressive machine. I asked her how much. She smiled, and simply said, “heaps”. As the first one came out, slowly, I was more than impressed with the quality of it. After examining it closely, I decided to proceed with the others, full well knowing that I would be shelling out quite a bit, but I “bit the bullet anyway”. About this time, the “boss” came in from somewhere, and she explained to him who I was and what we were doing. As the prining was done, we examined them as they came out, one by one, and as they were all perfect as far as I was concerned, they were placed on top of each other on a large bench, and at the end, rolled up. During this process, there was time for a discussion of my project, and they did seem genuinely interested, to the extent that when it came time to “work out the bill”, the boss said that since it was a “private” effort, for a good cause, well ……..he hummed and harrred for a bit, and then came up with a price, which was about one third of what Mark had suggested (at commercial rates) I might be up for. I was more than happy with that !! Basically, I think he really only charged me for the materials, and not for their time or expertise.
So the point is, people at large, still do care about it, and some are even willing to assist in some small ways, as this small company did for me. It seems that there is a little bit of a modern day “Titanic” feel to it, in the public mind.
On the subject of the sim and 45S.
Even the ATSB indicates that they take the sim data as “indicative” of an “intent”. The real question is, was there really any “intent” in Z’s mind to begin with, and if so, what was that “intent” ? There is a wide range of views on both of those.
Looking at it now, it is all very well for many to dismiss 45S as a realistic planning intent, based on what we now know about the Indian Ocean and drift studies and all that.
But ask yourselves, idividually, what did “you” know about the SIO, it’s currents and all that, back in February 2014 ? Does anybody want to put their hand up, as an “expert”, back then ? Deafening silence.
When I first looked at it, my prior knowledge was close to zero, and I live down here, and dad was a yachtie (although I was not, any more than I had to be, which was limited to within sight of solid ground !). My initial look into it started with google, you look at the simple charts first. They all show the gyre and the main currents fairly simply. If you are a pilot, not a sailor, and you look at them with an “intent” in mind, for an idea of where to go, the first and second and third looks all tell you the same thing, south of 40 will do nicely, 45 would be better. If that is all he did, research wise, the sim has to be taken as a clue (in my view).
@Victor. On the video he did say at 2:04 that, “the three countries involved in the search wanted to narrow it down to 5000 sq km”.
Separately there has now been a Ministry statement also:”International reports claiming the MH370 search could resume next year is (sic) untrue…”
https://audioboom.com/posts/6450153-ministry-clarifies-mh370-search-report
A reasonable explanation of the apparent contradictions, allowing for interpretation, is that statements that the search WILL go ahead are premature, since Australia and China have not yet been consulted and a contract has not been finalised.
@David: It would make no sense for OI to search only 5000 sq km. Don’t believe it. I suspect the contradictions on the status are occurring because the technical personnel within DCA are making progress on the negotiation and want to proceed, but in the end, this will be a political decision.
[Other comments redacted at Dennis’ request.]
Back to 370, I don’t have a good feeling about it. The Malay activity with Ocean Infinity is confused at best. It seems they are falling back in line with their tri-partite partners. I don’t see a resumption of the search as likely.
@TBill,
You asked: “Why not include the 23:14 Sat Call time slot?”
I just posted a new version 1.1 of my BTO/BFO Models that includes the 23:14 phone call. By the way, you can add as many lines to the calculator as you want by just selecting the cells in the bottom row and copying them downward. The time order does not matter.
With repect to the charts I mentioned above.
In case anyone wants them so that they can get them professionally printed out locally, I have uploaded them to dropbox, but be warned, they are huge files.
File 1 – Is a cropped section of a marine navigation chart = 96 Megabytes.
https://www.dropbox.com/s/lubu640whepirgd/SIO-1.pdf?dl=0
File 2 – Is a cropped section of a aviation navigation chart = 113 Megabytes.
https://www.dropbox.com/s/vnucapuxn0vjrxq/SIO-2.pdf?dl=0
@DrB
Thank you! I was just looking at the worksheet in some detail. I was particularly noticing the +4-Hz offset you are adding to sat calls BFO, as I was not aware some folks are post-adjusting some of the BFOs.
@ventus45
Re: “Even the ATSB indicates that they take the sim data as “indicative” of an “intent”.
Really?! Where have the ATSB used the words” indicative” and “intent” together with regards to the recovered flight simulation data?
@Victor
I can not download @sk999’s presentation that you linked, and the audio is so low it is almost inaudible. Is there another way to get it ?
@Mick Gilbert
https://www.atsb.gov.au/media/5773565/operational-search-for-mh370_final_3oct2017.pdf
In discussing the simulator data on Page 99, they say:
“There were enough similarities to the flight path of MH370 for the ATSB to carefully consider the possible implications for the underwater search area.”
If that doesn’t indicate “indicative” and “intent”, what does ?
The ATSB did not put that comment in just for fun.
@DrB
Thanks for explaining your max. latitude from 18:29 to me and @VictorI.
I see I inderstood you well. I only mistakenly put 18:22 in my post instead of 18:29 which I corrected few posts later to @Rob who pointed me at the mistake.
All clear about your max. latitude thoughts now.
@TBill.
You said: “I was particularly noticing the +4-Hz offset you are adding to sat calls BFO, as I was not aware some folks are post-adjusting some of the BFOs.”
Yes. It was apparent early on from the raw BFOs that the transmitted frequencies varied by a few Hz among the different channels in the SDU. For an example, see Table 10 on page 41 in my Addendum 5 from November 2015. The first publication of which I am aware is the paper by Hyman and Martin in December 2014. Mike Exner also contributed significantly to understanding this effect.
Tbill:
ATSB used only a subset of the R channel BFO observations because it was evident different channel types (R, T, C) had slightly different BFO Bias Values. We want to use all the BFO observations from all channels. So, several of us in the IG worked to calibrate all of the channel types using all 597 available observations. It turns out the C channel BFO Bias is estimated to be ~4 Hz lower than the R channel bias(~150 Hz), thus C channel BFO Bias= ~146 Hz. Thus, to compare R, T and C channel BFO values directly, it is common to normalize the values to be the R channel equivalent. For C channel BFO values, that means adding 4 Hz to make them R channel equivalent. Calibration values for all channel types and channel units available here: https://goo.gl/tCWTp6
@ventus45
Normally, “inverted commas” are used to show a direct quotation, that is referencing the specific words used in context. The ATSB used neither of those words.
“If that doesn’t indicate “indicative” and “intent”, what does ?”
Well, that should be self-evident; actually using the words “indicative” and “intent” would be a good start.
@Victor. Don Thompson says, “I expect that the six Hugin AUVs should cover 600km² per day, taking recovery cycles into consideration”.
Ten days production, plenty more than that for preparation and transit time. Unrealistic as you say to both parties: risk, overheads to OI, incommensurate reward to cover those risk and overheads.
However what he has said is troubling for other reasons. What is the nature of this contract? If the contractor offers a reward contract, isn’t the extent of his searches up to him to decide when enough is enough, so long as there is no dawdling?
Besides, again from this 25000/5000 business it sounds almost as though the tripartite countries are prescribing search areas, in which case implicitly they could be accepting some responsibility for their prescriptions/advice.
Maybe the Malaysian public service is, like others, intent on precluding public (and political) uproar over value for money should the wreckage be located in a couple of days using the search information and resources provided by governments, not just Australia but also the SSWG internationals.
The press would bear down. What? $50 million for 2 days work based on others research? What is our government doing gambling and not getting its money’s worth?! A paid for search would have cost considerably less! It was obvious it was there. Etc.
Perhaps the size of the reward will increase with the area searched.
Turning though to suggestions you sought as to the search area, I put myself first in Oliver Plunkett’s shoes. In principle start at the highest probability areas and work down, quitting when probability gets too low, though with some gain from the publicity and equipment proving. Keep things simple.
To define that more, some back-of-the-envelopes. Plan on 85 days of searching this coming season @600 sq km/day, say 50,000 sq km. Part A would be the 25,000 ATSB search area but review that with ATSB/SSWG as to balance of overall probabilities with Part B, the 25,000 left. That roughly would allow a run up to 28deg and a run back, searching 7½ miles (12 km) each side of the arc, 6 AUVs of 8 down, each with 2000m swept width. 600 sqkm/day at 3knots say, (5 km/hr, 60 sq/km), 10 hrs effective sweeping a day. Adjust targets as experience is gained.
Note this does not discriminate in priorities up the arc because most likely there is insufficient reason from conjecturals to chop, transit and restart. Also aerial and surface search outcomes are of little consequence, their runs commonly being very widely spaced and the debris having drifted during the delayed start. If priorities are about equal, give some priority to getting in early the run directions which weather might make awkward later.
If at the end any time is left over, use it on the then highest probabilities remaining.
Have a taxi ferry on standby to transport people and heavy items. Air drop parts.
Have the option of extending by a season should your shareholders choose.
Note for the tripartites. All search information to be disclosed by OI and placed under public ownership. All advice by the ATSB to be for information only. Make sure there is no use of this just for publicity and proving, then cessation at shareholders’ behest: the required search area should be the 25000 minimum (unless found part way through that).
@Victor. I left off the heading of the above, “The 5000 sq km search aim statement”, though a bit late.
The below is on the simulator sequence you offer to take a look at.
The ATSB says of the most recent Boeing simulations, “In some simulations, the aircraft’s motion was outside the simulation database. The manufacturer advised that data beyond this time should be treated with caution”. They have not said where the database limits are. They also say, “Some simulated scenarios also recorded descent rates that were outside the aircraft’s certified flight envelope”. They do not say whether the simulation outcomes of those are to be trusted. They have said also that the simulations do not include the effects of APU and engine starts even though an APU start is central to the 7th arc explanation.
In view of this I think an outcome from home simulators when working in these fields might be misleading yet much of my interest lies in those esoterics. So I can propose no sequence in those fields which to me would be useful, though thanks for the offer.
Also on pitch up, when Boeing advises caution with 777 motion simulation outside the database I assume they allude to insight they have similar to the 737. The Indonesian investigation of the Silk Air crash did not accept Boeing’s view of aircraft behaviour outside flight testing.
However, in both instances there is no more expert assessment available than theirs nor is there likely to be, so that should be the default view with the 777. What I am getting at is that lack of any allusion by Boeing to 777 pitch up dumps a bucket of cold water over my assertion that it might behave like the 737. I note also the EgyptAir flight 990 Boeing 767 pilot did not need nose down stabiliser trim to hold the nose down. So my postulation about pitch up now looks shaky.
@ALSM, @TBill: I think the bias for the C-channel is different as we have no data on the ground to use for calibration. What Mike has done is to assume a linear fit of the other BFO values to infer a bias for the C-channel. Doing this makes the BFO fit at 23:14 a measure of the BFO linearity for a given path.
@DennisW: Please do yourself (and your former employer) a favor and review your comments before you post them.
@David wrote:
“6 AUVs of 8 down, each with 2000m swept width. 600 sqkm/day at 3knots say, (5 km/hr, 60 sq/km), 10 hrs effective sweeping a day. Adjust targets as experience is gained.”
Only six AUVs (and ASVs) deployed on Seabed Constructor. I will suggest the effective side scan swath is less than 2000m, more like 1600m to allow for overlap with adequate resolution. Fugro’s original intent was to survey 2000m swathes but it was reduced to an average of 1700m. Your speed and mission duration estimates are too high & too low, respectively: halve the speed and double the duration.
I understand that the Edgetech equipment employed by Fugro on their deep tow vehicles and the AUV was capable of comparable performance but the AUVs were used for higher resolution infill work so less range.
OI’s Hugins are similarly equipped, I don’t expect any significant performance difference per vehicle in wide area survey.
@David, @Don Thompson: I’ve heard mention that the scanning profile will be optimized for finding the debris field rather than seabed mapping. I imagine this could result in more economical scanning compared to what Fugro was doing.
@Tbill, @VI Re C channel BFO bias:
My intent above was not to defend the method used to derive the channel dependent bias, but merely to explain why Bobby made the 4 Hz adjustment.
The BFO channel dependent bias estimates were derived in the fall of 2014. As Victor noted above, the C channel bias was estimated from the 23:14 data as shown on the last page of the following, published in Dec 2014 by me. https://goo.gl/S454AY.
The estimate does depend on an assumption that there was no change in altitude at the time, and the linear trend. I still think those assumptions are reasonable given the full record. That said, the C and T channel bias differences vis a vis the R channels do not have much impact on any of the important questions.
@ALSM @Victor @DrB
“The estimate does depend on an assumption that there was no change in altitude at the time”
That’s why I am so tunnel-vision focus on it…without the correction, I had come to the conclusion of a descent starting at maybe 22:50 (twilight).
Thanks to all for the explanation.
By the way, also note the values start jumping around about half-way through the reading period. So I tend to use the initial value (216). My interpretation of the mini-upset is a little turbulence in the high winds below 22S. Other thoughts?
Bad markup above, only first para should be in emphasis.
@Victor
@DennisW: Please do yourself (and your former employer) a favor and review your comments before you post them.
Good point. This whole endeavor has me spun up badly.
@DennisW: If you are so “spun up” that you are making injurious statements about you and your employer, it’s time to take a break.
@Victor wrote: “optimized for finding the debris field rather than seabed mapping”
Certainly, any search has to be optimised to find the debris field. Given that the debris field will be scattered it may be feasible to comb, rather than mow, each sector. By comb, I imply that survey of an area might involve interlaced sweeps, leaving gaps to incrementally work to 100% (or the deemed maximum) coverage.
The equipment must be operated in a mode that is capable of discriminating potential objects in a debris field. That is where I expect similarity between Fugro’s configuration and OI’s.
@Don Thompson: I wasn’t referring to the search pattern. Rather, since the search is for hard objects with significant acoustic signatures, might this allow the sweeps to be either wider or at a higher speed?
Don: I agree. Combing, rather than mowing, could effectively double or triple the search rate, depending on the estimated size of the debris field.
@VictorI
Here are the calcs I promised you yesterday, applicable to constant Mach 0.81 great circle.
The following calcs apply to the IGOGU/ISBIX/7th arc great circle that crosses the 6th arc at S36.60, E89.24, and the 7th arc at S37.62, E89.08. Calcs assume the aircraft flew at a constant altitude of 35,000ft until right engine fuel exhaustion.
Distances and arc crossing coordinates obtained from Moveable Type Scripts online Vincenty calculator for WSG84 ellipsoid.
The great circle distance between IGOGU and the 6th arc is measured at 2,651.5Nm, the corresponding journey time 5.5625hrs (5hrs, 33.75mins) measured from the mid-point time of FMT 18:37:15*. When a time averaged 2.2Kts headwind is factored in, the airspeed works out at M0.81, ie. 476.67Kts + time averaged 2.2Kts headwind = 478.87, divided by 590.50 = M0.81. (590.50Kts is the estimated average value for Mach 1 between IGOGU and the 6th arc, for estimated average temp 42.5C at 35,000ft altitude).
* Between completion of initial turn-back manoeuvre at approx 17:25 and last radar contact at 18:22, the aircraft flew at a relatively constant airspeed M0.84. Between 18:22 and arriving at the FMT mid point (estimated time 18:37:15), the airspeed is assumed to have gradually reduced from M0.84 to M0.81.
When calculating the average airspeed for each leg of the IGOGU/ISBIX/7th arc great circle I found that, assuming it flew at a constant altitude of 35,000ft, the aircraft maintained an almost constant Mach 0.81 throughout its journey south, subject to the limits of accuracy of the wind speeds and temperatures determined from the earth.nullschool.net global map of windspeeds and temperatures. applying specifically to altitude 34,000ft, for the date/time in question – via link on Brian Anderson’s paper “Determining the Mid- Flight Speed of MH370, 20/03/2015, Duncan Steel website.
IGOGU to 2nd arc: 512Nm, 480Kts average groundspeed, 1.8Kts average tailwind = Mach 0.808 (Mach 1=592Kts, Temp -41C) 2nd arc crossing point S1.0, E93.53.
2nd arc to 3rd arc: 478.7Nm, 478.7Kts average groundspeed, 1.6Kts average tailwind = Mach 0.806. (Mach 1=592Kts, Temp -41C) 3rd arc crossing S9.0, E92.7.
3rd arc to 4th arc: 488.6Nm, 485.6Kts average groundspeed, 4.6Kts average tailwind = Mach 0.812. (Mach 1=592Kts, Temp -41C) 4th arc crossing S17.09, E91.82.
4th arc to 5th arc: 480Nm, 479.6Kts average groundspeed, Zero average head/tailwind = Mach 0.81. (Mach 1=592Kts, Temp -42C) 5th arc crossing S25.06, E90.87.
5th arc to 6th arc: 692.2Nm, 461.5Kts average groundspeed, 13.5Kts average headwind =Mach 0.81. (Mach 1=586.5Kts ave.for leg, Temp -46C ave.for leg) 6th arc crossing point S36.6, E89.24.
6th arc to 7th arc: 61.5Nm, 434Kts average groundspeed, 33Kts average headwind = Mach 0.804. (assuming Mach 1=581Kts, Temp -50C, at altitude 35,000ft. However, aircraft would have been descending)
Perhaps interesting to note that a 100Nm (17:1 ratio) piloted glide initiated nominally at 35,000ft at MEFE, 14Nm prior to crossing 7th arc, and maintaining the same heading, would have terminated at approx S38.9, E88.9 when taking estimated headwinds into account, or 15Nm outside the southern boundary of 120,000sq.km. search area. For comparison, a 13:1 ratio glide would have terminated just inside the search area. Possibly a case of so near yet so far.
@VictorI – Originally, it was decided that pilot input after the FMT would give a search area so large it could take 10-20 years to complete a search. The only things that have changed since then is the recovery of debris and elimination of the previously searched areas. Has that changed the “no pilot input” assumption?
@Lauren H: Now we have a higher level of confidence that the final BFO values indicate a progressively steep descent. Regardless of whether there was pilot input, that suggests the impact was not far from the arc.
@Don Thompson. With 6 AUVs down there be at least one in transit and another being charged, maybe 2 depending on charge rate. Then there are repairs and replacements. More than 8 I would say to support the 6 deployed for 3 months, though I had read that they already had 8. Not just the items of course but the stowage for them. Likewise the surface vessels.
Some more idle thoughts. I imagine they would run staggered, not abreast, a replacement starting behind that being withdrawn. It would be awkward if running staggered interfered with their efficiency, time being lost reforming an abreast formation.
Then there is the like withdrawal and need for back up of the surface vessels.
Sea states present and predicted could affect AUV recovery times and damage rates. If caught out by weather do you put them all in loiter? What of the surface vessels? Put them in mother’s lee?
Their trials will have given them clues how to do all this but getting it all fully and efficiently operational might take a couple of weeks.
Re combing as opposed to mowing.
I don’t like the combing idea. I think the logic is flawed.
If the aircraft ditched, the debris field is the 4 corners size of the aircraft, say 80 by 80 metres.
If the aircraft went in nose first, the debris field should be relatively compact, like AF447’s debris field, which was small, only about 200 by 600 metres.
I think mowing with confirmed swarth edge overlaps is essential.
@David,
Six Hugins are deployed on Seabed Constructor, undoubtedly the ship carries spare parts. AUVs, and ASVs, 7 & 8 have been described as a separate package to be deployable on ships of opportunity.
The Hugin battery packs are swapped rather than recharged in-situ, endurance depends on speed but 72hrs is possible for side-scan sonar survey. Operational methods remain to be seen in detail but the AUVs will operate simultaneously during a mission.
Two stingers launch & recover the AUVs off the ship’s stern, directly from a purpose built, covered, hangar. The ASV LARS (launch & recovery system) operates with the ASV alongside the ship, each LARS is designed and built to launch and stow three ASVs. Many photographs have been uploaded to sites such as Marine Traffic since July, when trials began, showing the working configuration. The ship is presently in Hamilton, Bermuda, after working two areas in the mid-Atlantic over the last four weeks.
Sure, sea state and weather will dictate operations, but that’s what forecasting is there for.
@Don. As of June 15th they had 8 AUVs as 8 USVs and presumably they could get these all stowed, if not more, if they expect to have 6 down continuously.
http://subseaworldnews.com/tag/ocean-infinity/
@Don. Sorry. Overlapped.
@Victor,
I have assumed that the AUV operations for side-scan sonar survey will be conducted in similar configuration to Fugro’s work, the ‘state-of-the-art’ hasn’t progressed significantly. The Hugin’s Edgetech side-scan sonar operates at two frequencies to optimise cross-track range and resolution, while delivering a compromise along-track resolution. Fugro elected to operate at 75kHz for a 1600m cross track swathe and 410kHz for near range detail. That configuration, in turn, dictates the AUV speed and along-track resolution. 75kHz and 410kHz are the extremity of choices available.
Appendix D of ATSB’s Operational Search/Final Report summarises Fugro’s results. Debris fields and a lone object were discriminated, albeit lying on seabed that provided good contrast (high backscatter).
The Kraken SAS sensors are not spec’ed to deliver the range achievable with side-scan.
Of course, there’s the possibility that an undocumented, ground breaking, technique has been developed that exploits MBES, over a wider swathe than side-scan, to discriminate man-made objects on the seafloor …
@Don. Impacts on initial securing will be the main limit to acceptable sea states IMO. Yes if the seas remain benign and the USVs can be refuelled and serviced without grappling the damage rate would be minimised.
You have more faith in SIO weather predictions though than do I and with likely recovery times of USVs and benign conditions needed for that there might be little scope for taking chances.
With battery exchange, if quick and easy, recharging AUV downtime is overcome though there is transit and recovery time still for AUVs and transit and refuelling of USVs. I have my doubts that staggering would overcome that or would be efficient.
Also I think expecting to keep on line a total of 12 vehicles for 3 months say without needing to replace any, including during repairs, would be optimistic.
Thanks for that info. Obviously and necessarily I post when far from being fully informed and am really indulging in some mind clearing, though ceasing that for now.
@David,
On 4th Sept, ASV Global posted on social media that ‘build of the final Ocean Infinity SeaWorker ASVs is well underway’ together with a video.
There are only six deployed on the host vessel at this time.
@Don. Thanks. Maybe they will increase that, or they have a secret recipe.
A correction to my earlier please. By “weather predictions” I should have included sea state predictions. Seas come from wind, fetch and time, but swells of course can be independent and from a distance.
@ventus45
IF the aircraft ditched do you not think at least one engine would have detached as it skimmed across the waves? Wouldn’t the main fuselage drift before sinking and then not sink vertically to the bottom?
At least three recovered fragments originate from an engine inlet cowl and a cowl door, that suggests that engine debris was created at impact.
@Don. Late final. ‘Simultaneous’ as the recipe. Maybe. I have had a look, though getting off topic.
https://www.dropbox.com/s/9glvhnn6lr426nh/Simultaneous%20vs%20serial%20AUV%20deployments.docx?dl=0
@David @Don Thompson
Imo it’s all about the assumptions.
If the plane made a high speed nose down impact the plane was smashed in small pieces scattered along the seafloor.
Biggest pieces would probably be remains of the engines and landing gear.
High resulution scanning would be needed closer to the seafloor with a smaller width taking more time.
If the plane ditched or made a ditch-like impact (extreme like AF447) there would be much bigger pieces or even the hull with wings attached in a small area.
No need for a very high resolution scanning.
Three to six meters of resolution would be enough which would save a lot of time imo.
@Ge Rijn
My understanding is that the requirement in the Fugro search was to detect (with high probability) an engine-core sized target at the edge of the 1.6km wide sonar scan. This was presumably the assumed worse case of a very high-speed sea impact that shattered the airframe. A target of this size was dropped to the seabed just off Fremantle and used occasionally for testing. I doubt the requirement for the OI work would be much different since a high-speed impact cannot be ruled out, and that Fugro requirement seems a minimum for that situation.
“made a ditch-like impact (extreme like AF447)”
When AF447 impacted the Atlantic Ocean the last recorded FDR values showed a vertical speed of -10,912 ft/min, a ground speed of 107 kt, pitch attitude of 16.2 degrees nose-up, roll angle of 5.3 degrees left. Not even remotely “ditch-like“.
Assumptions, yes, but those based on credible and plausible reasoning, not “imo”.
@Richard Cole
I agree a high speed impact cannot be ruled out. But imo a ditch-like impact cann’t be either.
Maybe it’s a idea to make high resolution scans close to the 7th arc first relating to high speed impacts and lower resolution (and less time consuming) scans beyond those regions after.
@Don Thompson
That’s why I called AF447 an ‘extreme ditch’.
Result was a relatively small debris field on the ocean floor with big pieces of fuselage, wings and engines close together.
Easy to spot with three to six meters sonar resolution.
@Ge Rijn
“Easy to spot with three to six meters sonar resolution.”
Really?
The successful deep ocean search for AF447 operated Remus 6000 AUVs with side-scan sonar operating at 120kHz so as to ensure adequate discrimination of a likely target at a maximum range of 700m. The Airbus debris field straddled the overlap of two adjacent passes by the AUV. The seabed in the area was abyssal plain. That was fortunate as the sonar backscatter from the debris field contrasted well against the uniform return from the seafloor. Discrimination is not a simple ‘pixel’ based resolution issue.
This presentation delivered by Jan Ingulfsen at a recent conference in Perth, WA, might help illuminate some of the points @David & I discussed above.
@Don Thompson: Thanks for finding that and sharing it. I think advanced technology that drastically reduces the cost of a wide search area offers hope that we can find MH370.
@Victor
Re: ZS Sim Cases
I wonder if Yves has any ideas about to prevent the >90 deg turn over shoot we see in FS9/FSX? I am finding the turn thru 10N may have an unusual quality. Pending fine tuning, it is the exact correct place to start a very sharp >90 deg turn that ends exactly on the DOTEN to NZPG great circle without over-shoot or under-shoot.
@Victor
I think advanced technology that drastically reduces the cost of a wide search area offers hope that we can find MH370.
“Hope is the currency of the poor.” – Cesare Maestri. Probably the best climber in the history of the world.
Sorry, I could not resist. I know “hope” is all we have at the moment.
@DrB, Your fuel model is proving extremely useful, particularly the Endurance v Range table. I had been working on what I thought was a promising candidate path, a great circle, constant M0.81 at FL350, with FMT at IGOGU at 18:37 and crossing the 7th arc at S37.6 but I can see from the table that FL350 has insufficient range to reach latitude S37.6.
However, I notice that M0.82 at FL380 meets the desired range and MEFE time requirement very closely. Range requirement from 18:29 to MEFE at 00:17 is 67NM (estimated) to IGOGU, plus 2,651NM to 6th arc crossing at latitude S36.6, plus 47NM to MEFE at 00:17 = 2,765NM. By comparison, from the table the range for M0.82 at FL380 is 2,774NM from the 18:29 point, and MEFE is 00:16. It could be that ground-speed of M0.82 at FL380 is comparable to ground-speed at M0.81 at FL350, (assuming the same windfield) because the lower temperature at the higher altitude has the effect of increasing the Mach number, for the same ground-speed. Is that a reasonable assumption?
I know that S37.5 has been searched already, that the final BFOs indicate a rapidly accelerating descent rate between 00:19:29 and 00:19:37, and the latest drift results run counter. But the BFOs represent an 8 second snapshot, and if a piloted glide had begun at 38,000ft, a steep descent at one point should be considered less unlikely perhaps?
All,
I am trying to understand better just what factors drive the DSTG’s various solutions and what happens as assumptions are relaxed.
As an initial investigation, consider Bayesian Methods, Figure 10.3, top. This figure shows the probability distribution in final latitude along the 7th arc using the BTO data alone. Surprisingly, there is a
well-defined narrow peak centered around -39.2 deg latitude (plus another one at -36.7 – presumably magnetic track routes.) How can that be? If we consider the BTOs from 19:41 to 24:11, they can be fit with either an LNAV or true track route with constant Mach, but the starting location is ill-constrained. (More specifically – even though the satellite was moving and thus the “ping rings” are not concentric, one can pick any starting latitude from 0 to 20 degrees, and the fits are all essentially identical.)
The DSTG started it modeling at the 18:01:49 position. If one is free to choose the speed and heading at that point up to the FMT, a second speed and heading after the FMT, and the time of the FMT, there is a 1-dimensional family of routes that match all the BTOs (including the ones from 18:25-18:28). For reasonable speeds both before and after the FMT, there is a spread in the final latitude ranging from -37.5 to -40.2 deg.
Why is the DSTG distribution in final latitude so narrow? It is because the DSTG imposed two additional constraints – the speed and heading were constrained to be close to the values that its filter had output for the 18:01:49 point based on the primary radar data up to that point (Figure 4.3 – 508 knots ground speed and 289 deg true heading.) Formally, it used these values as “priors”, with stdev of 17 knots and 1 deg respectively (Table 8.3). Now, I am not running Monte Carlo’s like the DSTG, but if I make a least squares fit with these free parameters and priors, I find a final latitude of -38.8 deg with a stdev of 0.6 deg. This latitude is still 0.4 deg North of the DSTG value, but the stdev is small, in agreement with the DSTG result.
The time of the FMT is 18:36, in agreement with (but not at all driven by) the BFO of the 18:39 phone call. However, all is not well. The BFO residuals are big (5.5 hz stdev). The time that the final radar position is reached (whether my value or that of B. Ulich) is approx. 18:23:38, more than 1 minute later than what the Lido Hotel radar image showed.
Comments welcome. More to come.
Minor edit – “Table 8.3” => “Table 8.2”
@Don Thompson
With lower resolution scanning I did not mean 3 to 6 meter p/pixel but to scan with minimal resolution to find 3 to 6meter (and bigger) objects.
The main idea is to scan possible high speed impact areas along the 7th arc with high resolution and areas beyond/outside those areas with lower resolution to save time and money.
For if the plane is not found in the possible high speed impact areas it probably made a ditch-like impact with much bigger pieces remaining close together on the seafloor.
@Rob,
I am glad you find the Endurance Model useful. I wanted to allow people to try various scenarios on their own, and I am glad you are doing that.
You said: “It could be that ground-speed of M0.82 at FL380 is comparable to ground-speed at M0.81 at FL350, (assuming the same windfield) because the lower temperature at the higher altitude has the effect of increasing the Mach number, for the same ground-speed. Is that a reasonable assumption?”
The TAS is shown in the Fuel Flow Model worksheet. If you put in M0.81 at FL350 and +10C, the TAS is 477 kts. At FL380 the same TAS gives M0.815. So, yes, that is a reasonable assumption, that a higher Mach in colder air gives the same TAS.
You also said: “But the BFOs represent an 8 second snapshot, and if a piloted glide had begun at 38,000ft, a steep descent at one point should be considered less unlikely perhaps?”
A piloted descent at 15,000 fpm does not seem reasonable or likely to me, but I am not a pilot. I am quite convinced that the two final BFOs accurately indicate the RODs at 00:19.
@sk999,
Regarding the DSTG analysis, you said: “Surprisingly, there is a well-defined narrow peak centered around -39.2 deg latitude (plus another one at -36.7 – presumably magnetic track routes.). . . . However, all is not well. The BFO residuals are big (5.5 hz stdev). The time that the final radar position is reached (whether my value or that of B. Ulich) is approx. 18:23:38, more than 1 minute later than what the Lido Hotel radar image showed.”
My first solution back in July 2014 was at 39S at a constant track and constant speed. You are correct that the BFO errors there are a problem because their standard deviation is too large. Even worse, they have a mean which is not close to zero. That indicates there is a systematic error in the course southward, and both these issues are resolved for courses close to 180 degrees.
Despite much looking, I have never gotten a magnetic track solution at S36.7. I don’t think a peak there is a magnetic track route. More likely it is a slower straight path.
I am surprised by your computed time for the last radar position. 18:23:38 seems awfully late. Could you please explain the method you used to derive this time?
@DrB
Even worse, they have a mean which is not close to zero.
I told Victor I was done posting, but your statement above demands a comment. Why in the world would you think the BFO errors should have a zero mean? You have made this assertion before. It is just plain wrong. Does a random walk have a zero mean? Does figure 5.4 of the DSTG book have a zero mean? Were you asleep when ergodicity was discussed in your stats class?
@Don. With PAX capacity of 102 (less I suppose the search staff, ie other than ship’s crew), plus lift from 6000m and ROVs, maybe Seabed Constructor could do wreckage recovery; if equipped with some bits and pieces the ATSB mentions in its Operational Search report. There might need to be some USV offloading for deck space. A part of contract discussions maybe, together with the site survey and information needed before starting the recovery?
This was a presentation to unstated Perth WA attendees I notice, not at the ATSB’s Canberra but I suppose that unless there was another presentation they went across.
Various interesting practical points like the effectiveness of the hook-up gathering ‘funnels’ on the USVs with falls blowing and swaying about and the USVs rolling, pitching and heaving are of interest but must work quite well, in limited seas anyway.
It is worth them pointing out a relationship between telemetry/data transfer and distance and also the challenge of timely processing of the data, given its quantity. However why they would depict USV at the p13 distances they do not make clear. I assumed the ship would be in the middle with USVs abreast. Perhaps her underwater noise might be a reason why not?
I note for the first trial (p25) they describe area vs time, the outcome being almost 200 sq km/day, which is also the average over the “last month”. Both presumably had the various gaps that would come with trials.
Their use of the word ‘simultaneous’ applies to the searching being 6-at-once I think, not whether they are recharged/refuelled in a search halt (as distinct from serially). There look to be (p26) 2 AUV recovery doors, allowing for recovery and battery exchange in pairs, applicable to both the ‘one hit’ and serially.
There must be a high potential demand for deep ocean searches to warrant the level of expertise, investment and enterprise at work here and I suppose competitors will be committed likewise. What will search capability in 10 years look like???
Bobby Ulich,
Much thanks for your commentary. My spouting r.e. magnetic track to -36.7 was just a conjecture – I didn’t study it one iota, so it could be wrong.
R.E. time to last radar position, what I do was to assume a constant Mach starting at the 18:01:49 position and continuing up to the FMT, the exact value of said Mach being found by least squares. I then plucked off the time when this particular path passed the longitude of the last radar point. Remember, I am just trying to recreate something close to what the DSTG did – I’m not representing that it is what actually what happened.
@sk999 said: another one at -36.7 – presumably magnetic track routes
@DrB said: Despite much looking, I have never gotten a magnetic track solution at S36.7. I don’t think a peak there is a magnetic track route. More likely it is a slower straight path.
Or a slower path with another turn or two.
@DrB, Excellent work on the Fuel modeling. Thank you for sharing. Please check the following Version 5.5 model data points against your source data.
Workbook Tab [LRC INOP FF] ( 80FL, 270 MT)
Workbook Tab [Holding INOP FF] (200FL, 140 MT)
Workbook Tab [CI=52 FF] (270FL, 180 MT)
Cheers, Tom
@sk999. Bless you. I was hoping that someone with the requisite skills could properly address my earlier incredulous enquiry on how, exactly, the DTSGs probability peak should be so clearly defined, yet Victor’s BTO and BFO RMS error curves appear to provide a gigantic and nearly equally-good spectrum of candidate paths from 19.41ff. Since all parties appear to be in full agreement on the BFO and BTO interpretation methodology, the only reason for such a difference must lie with DSTG’s priors – as Victor has pointed out. I’m not entirely convinced that this is wholly down to a model bias towards fewer turns. If I understand them correctly, their point is rather that with reasonable priors a solution can be obtained with a single manoeuvre. Other solutions are also possible, but they require multiple changes in track/speed. And in the absence of a sensible reason for supposing the more complex solutions to be credible, a parsimonious solution is preferred. I’ll be following closely the discussion among experts on this topic as I am yet to be convinced that DSTG’s clearly-defined PDF peak is a mirage.
@VI and sk999 re bi-modal peak in PDF. If memory serves, I think that there reason behind this result was clearly explained but I can’t recall the details. Something to do either BFO uncertainty intervals or how much weight is put on veracity of 1822 vs taking 1802 as the LKP?
@Paul
If memory serves, I think that there reason behind this result was clearly explained but I can’t recall the details. Something to do either BFO uncertainty intervals or how much weight is put on veracity of 1822 vs taking 1802 as the LKP?
So why not look up the details? What is wrong with you people?.
@David
“presentation to unstated Perth WA attendees”
The PDF is hosted by SUT, the (Australian) Society of Underwater Technology). The site also hosts the October conference agenda, etc.
The cone-bullet latch mechanism you note on the ASVs is also typical for ROVs. You’ll note that the AUV hangar includes cradles for six Hugins, two of which are the launch and recovery ‘stingers’.
@Dennis W. I’m simply pointing out where to find the answer so that the person asking the question can go and look it up for themselves.
Alright Dennis. Since I’m in neighbourly mood this fine morning, the answer to the bimodal or not is found in Bayesian methods, figure 10.3 page 87. Using BTO only, you get a bimodal PDF with one peak at 39S and a second at 37S. With BFO included you get more of a single peak at 38 with a bit of a shoulder at 36S.
It’s also worth noting that probability north of 34S is practically zero in both cases, and 35S is practically zero on the “BTO only”. When BFO is included, 35S lies on the tail of the PDF with relative frequency of about 0.07 (about a fifth of the peak value).
Here is a short video showing the final sea trails of the Unmanned Surface Vehicles (USVs) that Ocean Infinity is using on Swire’s Seabed Constructor.
The DSTG in their Bayesian analysis considered these kinds of manoeuvers: changes in course, changes in speed, and changes in altitude, where each change of a given type is considered to be one manoeuver, and the time between manoeuvers is described by a Poisson distribution.
I have long believed that another manoeuver that should have been considered is a holding pattern, which could be described by a random variable indicating the duration in that pattern. When you consider that holding patterns are automated, holding patterns are commonly found as part of “normal” commercial flights, and holding patterns are also common in “abnormal” commercial flights, such as when the crew in an emergency situation is assessing and controlling conditions while in the vicinity of a suitable airport, it seems there is ample justification for including a manoeuver of this type.
Of course, I also believe the possibility that the plane was descending during the sat call at 18:40, and the FMT occurred afterwards. In the time since I described that possibility in a previous post, I’ve done additional work that suggests to me that rate of change of the BFO during this descent would be less than previous models suggest.
I think that if you allow these two possibilities (descent at 18:40 and holding pattern manoeuver), and you eliminate trajectories crossing the 7th arc further south than around 36S based on previous search data, you would see a posterior distribution that is relatively flat up to around 26N latitude, as the curve for BTO error shows in Figure 2 above. That doesn’t help us define hot spots, but it does justify a fairly large search area.
An animated clip showing how the Denith designed LARS operates with the C-Worker ASVs.
All,
Slight correction to above post r.e. BTO-only solutions. Turned out I was using constant airspeed, not constant Mach. Switching to the latter, I now get best fit for LNAV mode going to latitude -39.1, which matches Figure 10.3, top. Turn is a minute earlier – 18:35.
The second peak in Fig 10.3, top might be true heading routes. I can get a decent fit that turns at 18:39 and goes to -36.6 provided I incorporate wind errors. The final latitude depends on altitude, since winds are involved; this particular run used FL280 and Mach 0.758.
@Victor
Bottom line for search zone model – stick with Arc2 as the starting point and it is possible the only arcs that might be without pilot input are Arcs 3-5, and even that assumption I would give up on if necessary.
@TBill: I think solutions where no pilot input is required after 19:41 (or sooner) are preferred. The flight between 18:22 and 19:41 is much less likely to be automated.
@sk999. Thanks for that. Your 39.1S route in LNAV with turn at 18:35 is not going to be fuel feasible, I suspect. What does your fuel model say?
@Victor. Agreed. Would have been nice if the modelling had permitted a holding pattern in the mix of possibilities since this appears to be the chief means by which more northerly terminus becomes a decent path fit. I recall Dr B was running models to reach the 35S area but I’m not sure if he found a path model that produces a good fit without inclusion of a holding period?
I wonder if they (DSTG) might consider re-running the models with that variant and seeing if the PDF it produces indicates any clear peaks. I shouldn’t think that would be a terribly time-consuming or costly thing to do with the modelling they have already done?
testing
@Victor,
“I think solutions where no pilot input is required after 19:41 (or sooner) are preferred. The flight between 18:22 and 19:41 is much less likely to be automated.”
Ironically I arrived to this idea 3 years ago, but I was heavily criticised for pursuing it.
These days my thinking is concentrated on triangular patterns with descent. A wrong exit would result in the heading of approximately 172 deg. At the moment I have 4 such patterns, which fit 18:22-18:41 data.
Perhaps Andrew could tell whether repeated triangular patterns can be used for some other purpose besides indication of navigational or radio failures.
BTW, I believe 23:14 BFOs provide strong indication that the plane was flown in one of the three possible constant heading modes (true, magnetic or hygroscopic), all being subjected to the wind.
@Oleksandr said: Ironically I arrived to this idea 3 years ago, but I was heavily criticised for pursuing it.
I certainly don’t recall criticizing it since I considered a loiter followed by a BEDAX-SouthPole path since July 2014.
@Victor,
I did not mean someone specifically, but in general.
General consensus was that the FMT took place between 18:25 and 18:40, wasn’t it?
@Victor
re my great circle constant M0.81: I’ve just remembered that I was using ground miles as per the Vincenty calculator for the calcs, rather than air miles. ie miles covered at altitude. the additional distance from IGOGU to the 7th arc is 4.5NM for FL380. And a 7th arc at 38,000ft would be further south than the 33,000ft arc. I seem to remember reading in an INMARSAT paper, that they assumed the arcs to be nominally at 10,000 metres altitude, although I’m not sure if that figure is still used for calculating the arc positions. If the 6th and 7th arcs are 2NM further south, then the total distance from 18:29 to MEFE becomes 67.5NM (estimated) from 18;29 to FMT at IGOGU, plus 2,658NM from IGOGU to the 6th arc, plus 47.5NM from the 6th arc to MEFE at 17:30 = 2,773NM. Compare this to DrB’s 2,774NM for M0.82 at FL380. There’s more, M0.81 at FL350 equates approximately to M0.82 at FL380; average TAS for the journey south is now increased to 480Kts from the previous 478.69Kts, which equates to average M0.813. Now using the simple conversion factor I borrowed from DrB (ref his previous post to me) to convert Mach at FL350 to Mach at FL380, you multiply by 0.82, and divide by 0.815, and get M0.818, or very close to M0.82.
Thus the distance covered at constant M0.82,FL380 turns out to be within 1NM of the IGOGU/ISBIX/7th arc great circle distance, and the MEFE time is within 1 minute. An exceptional result. One should remember that DrB has gone to great lengths to ensure his fuel model for MH370 is as accurate as possible. It is a fuel model for those specific engines and the specific atmospheric conditions on the night. This cannot be dismissed as a coincidence. This has to be none other than the actual flight path.
@Oleksandr said: General consensus was that the FMT took place between 18:25 and 18:40, wasn’t it?
Before the search of the first 60,000 sq km failed to find the wreckage, and before any debris was recovered from Eastern Africa, an impact along the 7th arc at around 38S was the consensus, corresponding to an FMT between 18:29 and 18:39.
@Rob: By my calculation, a great circle path crossing the 7th arc at 37.5S requires a speed of M0.828 at FL350 to match the BTO data, which is well above the M0.81 you have calculated.
@Victor
Could you possibly show me your calcs?
@Rob,
Where, in your calculations, do you account for the wind. It has a significant effect in more southerly latitudes.
@Oleksandr
“BTW, I believe 23:14 BFOs provide strong indication that the plane was flown in one of the three possible constant heading modes (true, magnetic or hygroscopic), all being subjected to the wind.”
What is your logic about 23:14? I tend to agree there could be a wind/turbulence effect at some point in time the BFO jumps from steady 216 up to 220-222 with bouncing around. I like 216 as the base point.
@Brian Anderson
Re wind calcs, probably best if I send you the relevant section of my earlier post to Victor. You can then see how the wind was accounted for for each leg. I took the wind values from the earth.nullschool map linked to your “Calculating the Mid Flight Speed” paper. I took multiple readings along each individual leg, then averaged them to get a single wind speed/direction for each leg. Same for the temperatures. Here is the relevant section:
“When calculating the average airspeed for each leg of the IGOGU/ISBIX/7th arc great circle I found that, assuming it flew at a constant altitude of 35,000ft, the aircraft maintained an almost constant Mach 0.81 throughout its journey south, subject to the limits of accuracy of the wind speeds and temperatures determined from the earth.nullschool.net global map of windspeeds and temperatures. applying specifically to altitude 34,000ft, for the date/time in question – via link on Brian Anderson’s paper “Determining the Mid- Flight Speed of MH370, 20/03/2015, Duncan Steel website.
IGOGU to 2nd arc: 512Nm, 480Kts average groundspeed, 1.8Kts average tailwind = Mach 0.808 (Mach 1=592Kts, Temp -41C) 2nd arc crossing point S1.0, E93.53.
2nd arc to 3rd arc: 478.7Nm, 478.7Kts average groundspeed, 1.6Kts average tailwind = Mach 0.806. (Mach 1=592Kts, Temp -41C) 3rd arc crossing S9.0, E92.7.
3rd arc to 4th arc: 488.6Nm, 485.6Kts
average groundspeed, 4.6Kts average tailwind = Mach 0.812. (Mach 1=592Kts, Temp -41C) 4th arc crossing S17.13, E91.82.
4th arc to 5th arc: 480Nm, 479.6Kts average groundspeed, Zero average head/tailwind = Mach 0.81. (Mach 1=592Kts, Temp -42C) 5th arc crossing S25.12, E90.87.
5th arc to 6th arc: 692.2Nm, 461.5Kts average groundspeed, 13.5Kts average headwind =Mach 0.81. (Mach 1=586.5Kts ave.for leg, Temp -46C ave.for leg) 6th arc crossing point S36.6, E89.24.
6th arc to 7th arc: 61.5Nm, 434Kts average groundspeed, 33Kts average headwind = Mach 0.804. (assuming Mach 1=581Kts, Temp -50C, at altitude 35,000ft. However, aircraft would have been descending)”.
I have since found, from Dr Bobby’s endurance table, that M0.81 at FL350 has insufficient range, but that M0.82 at FL380 appears to have exactly the required range and endurance.
However, according to Victor, the overriding problem with this particular constant Mach path is that it does not comply with the BTO. I find that a bit surprising, but then I admit I’m no expert on path fitting.
Curiously, Bayesian Methods Fig 10.3, bottom, “with BFOs,” has a broader distribution than the top, “without BFOs.” Why does adding data cause the distribution to broaden? A guess is that the top plot was not generated with all degrees of freedom in the model turned on, while the bottom plot was.
Next exercise – suppose I only consider the data (BTOs and BFOs) from after the FMT. I will assume that the plane followed a constant LNAV route at constant altitude and Mach starting just before 18:40 up to the end of the flight. What can we say?
I have four free parameters – the starting latitude, longitude, heading, and speed. (I’ll take a starting time of 18:38, just before the 1st phone call.)
The DSTG report assigned errors of 29 microsec to the BTO (with a couple of values higher) and 7 hz to the BFO. Given those errors, I find that the best-fit starting coordinates are 8.6N, 94.3E, and a best-fit end latitude of -36.9 deg. However, these values are very poorly constrained, and the 1-sigma error on the final position (measured along the 7th arc) is 3.4 degrees. Note also that the best-fit starting coordinates are well beyond where the aircraft could have flown by 18:38 (and curiously is near the 19:12 NW point, although that should be regarded as a coincidence.)
Final exercise. What if I include the data from before 18:38 along with the starting position and priors assumed by the DSTG, as I previously did for the BTO-only case? I end up at -38.4 degrees, with a small error – only 0.23 deg. For comparison, a true track ends up at -37.9 deg.
The overall spread in the final distribution must come from combining at least three individual narrower distributions (1 each for LNAV, true track, and true heading) plus accounting for other elements that have been left fixed here – different altitudes, variable winds, etc.
@TBill,
Re: “What is your logic about 23:14? I tend to agree there could be a wind/turbulence effect at some point in time the BFO jumps from steady 216 up to 220-222 with bouncing around. I like 216 as the base point.”
When I developed a so-called CTS flight model (already outdated) I noticed that wind pattern ‘centered’ around 96S, 22S, FL200-FL250 was just right to alter ground velocity to fit the BFOs in 23:14 cluster. Even the trend during the 2nd call was reproduced correctly. I observed the same effect in later models, where I managed to connect 19:41-00:19 section to 18:22 radar position via triangular patterns.
I guess the same is applicable to CMH models. I am currently looking at that issue, but perhaps @sk999 can already comment on it.
In summary, nothing to do with the turbulence; just the plane needed to fly through such a wind pattern in the mode, which could alter track angle.
@Victor,
“Before the search of the first 60,000 sq km failed to find the wreckage, and before any debris was recovered from Eastern Africa, an impact along the 7th arc at around 38S was the consensus, corresponding to an FMT between 18:29 and 18:39.”
Exactly.
My 3-year old prediction that the FMT did not occur during this interval was based on simple logic: the plane was actively piloted till 18:22; some unorthodox actions, which resulted in the SDU logon 18:25, were in process. How could the plane become unpiloted in just 5-15 minutes?
@VictorI
You wrote: “@TBill: I think solutions where no pilot input is required after 19:41 (or sooner) are preferred. …”
What are your arguments to support this thought / statement?
(If you look in detail to the BTO/BFO data as @TBill is doing, I also find there are some indications for “maneuvering” still going on between 19:41 and 20:41. However: given the unclear BFO errors, not something we can say with certainty.)
@Kenyon,
I checked the three cases of Fuel Flows you mentioned. The first one was a dyslexic typo I made in manually transcribing the Boeing table.
The second one seems to match the Boeing table. Perhaps you might explain why you thought there was a problem with it. Be sure to use the Flaps Up table value.
In the third case, there was an incorrect entry which I have fixed.
LRCINOP FF FL80, W 270 tonnes: Was FF 8,436, Corrected to 8,346
Holding INOP FL200 W 140 tonnes: Was FF 3,780, Appears correct
CI=52 FL270 W 180 tonnes: Was FF 2,566, Corrected to 2,891
These two corrections are included in a newly posted V5.6 fuel model. You can use the same link as before.
Thanks for pointing these things out, Tom. You have a keen eye.
@VictorI
At the beginning of this thread I indicated I am interested to know the 19:41 latitudes of the sixteen great circle paths you have analyzed (fig. 1, fig. 2). Would it be possible to share those?
@DennisW,
You said: “Why in the world would you think the BFO errors should have a zero mean? You have made this assertion before. It is just plain wrong. Does a random walk have a zero mean? Does figure 5.4 of the DSTG book have a zero mean?”
Actually, yes it does have essentially a zero mean. Here is why:
1. The mean of the BFO Probability Density Function of BFO errors shown by DSTG in Figure 5.4 is listed immediately above the figure in Table 5.1
2. The mean observed BFO error was between 0.10 and 0.27 Hz. That’s VERY close to zero.
3. To quote the DSTG on the previous page, “The mean BFO error was close to zero in all cases . . . .”.
4. Therefore, the expected value of any size ensemble of BFOs is also within a few tenths of a Hz from zero.
5. Therefore an ensemble with a mean significantly different from zero (say several Hz), is possible but is less likely than an ensemble with a near-zero mean.
I think you were asleep when you read the DSTG report.
Ocean Infinity. A couple more on how it all works:
AUV storage (including hoists), launch, recovery:
https://spark.adobe.com/video/z8cksY38bal7z?w=_2913
AUV innards:
https://spark.adobe.com/video/4e6XIf5pGDBJL?w=_5634
@Paul Smithson,
You said: “I recall Dr B was running models to reach the 35S area but I’m not sure if he found a path model that produces a good fit without inclusion of a holding period?”
Yes, I have found a CTH route ending at 34.75S that uses a single FMT without a Hold. I published this in January 2017. A link is in my pubs list, and a PDF can be found HERE . This route is an excellent fit to the satellite data, except that the endurance is a few minutes too long using my new fuel model. Another issue is why a pilot would set the NORM/TRUE switch to TRUE at MH370 latitudes.
A more likely possibility, in my view, is a 180-degree Constant Magnetic Track. This closely mimics the CTH route and perhaps it is a more likely setting. My first attempts gave good (but not perfect) results fitting a CMT route using a 2014 magnetic declination table. Now we know 9M-MRO used a 2005 table. I have these 2005 data, and I will see if that improves the fit. The end point won’t move much.
@Oleksandr
Not convincing re:2314 without a plot or something so show it. At some point the 2314 BFO reading sequence instantaneously went from steady, to jumping around a bit.
@Dr B Thanks for posting the link to your path. I’ll take a closer look.
I recall you saying previously that CMT looks similar to CTH and I was surprised. I thought that CMT bends quite a lot more than CTH (in the last 1-2 hours) but I guess I must be wrong. I’m using a modified version of Barry’s model that offers CMH but not CMT as a mode so I can’t actually simulate it. Could you post the .kmz of the two tracks so that we can see them side by side?
@TBill,
I’ve just taken a fresh look at my old CMH model. My prediction was indeed correct: BFOs 23:14 also agree well.
I guess I know why the DSTG missed this batch of routes:
“An initial Mach number was selected
from a uniform prior between 0.73 and 0.84; this was chosen on the basis of expert advice to ensure that the required flight endurance is achievable.”
My experiments show corresponding Mach of 0.70-0.71, so the next task will be quite delicate: the existence of a CMH route depends on fuel.
@TBill,
Re: “Not convincing…”
I posted plots and tables quite a few times… including recently. You may also check my CTS note (rev 02.12.2015), where you can see what wind does around 23:15.
I don’t have plots for CMH models (yet), as I abandoned this class of models long time ago. I believe effect is similar.
@DrB
“Now we know 9M-MRO used a 2005 table.”
How do we know that? I have been using 2005 table ever since ALSM said ATSB intimated that to him a while back. But I also use circa. 1995 mag tables as that was the FS9 default, and sometimes I use 2017 to bracket.
You will find 2005 harder to fit, I would think, as it is more curve.
@Oleksandr
Re: “Perhaps Andrew could tell whether repeated triangular patterns can be used for some other purpose besides indication of navigational or radio failures.”
I’m not aware of any other purpose. I’ve already given you my view on the likelihood of MH370 flying a triangular pattern; I really don’t think it has ‘legs’ (if you’ll pardon the pun).
Bobby Ulich,
The DSTG discussion regard the BFOs in general and Table 5.1 in particular is quite confusing, and I am usure of what it really means.
The text discussing Fig 5.4 (the BFO residuals for the Mumbai-KL flight of Mar 2, which is what DennisW frets over), states as follows:
“The residual error is clearly not zero-mean, and the mean varies with time.”
The bias value used for this plot was derived from on-tarmac measurements. The DSTG then goes on to say:
“The potential variations were incorporated by modeling the BFO bias as an unknown constant with a prior mean given by the tarmac value and a standard deviation of 25 Hz.”
In other words, the in-flight BFO bias was basically forced to give zero mean aside from a slight “tinting” introduced by the prior.
What, exactly, does “Mean BFO error” in Table 5.1 refer to? Mean of what relative to what? Are the on-tarmac residuals relative to a bias determined by the on-tarmac values themselves, and the in-flight residuals relative to a slightly tinted bias? Any insight would be welcome.
On a related note, the DSTG had no example of a cold restart of a SATCOM while in flight. Will the BFO bias come back to its original value? It is worth noting that the BFO bias offset for MH371 was 25 hz bigger than that for MH370 (pre-IGARI).
@Victor: I am deeply appreciative of the efforts of those who work tirelessly, intelligently, and endlessly to uncover MH370’s true fate. And I continue to appreciate your efforts in particular; in the hypothetical that the ISAT data we were fed is, despite clear appearances, authentic, it is always useful to nail down the Bayesian pdf of possible impact points.
Re: “best fit ending at 34S”:
1) in your analysis, you’ve chosen to fit BTO and BFO values for 6 arcs from 19:41 onward – not 3, as the JIT did in April of 2014. If you had been heading up the JIT – and had to decide how many arcs to fit paths to – would you have chosen 6, or 3?
2) I can happily wait until the post with which you’ve teased us, but hopefully this is something you can let slip: for this best fit path, at what precise coordinate is it at 19:41?
I also reiterate (and hopefully focus) my polite request for the names of all the people from whom you’ve received data you’ve used to shape your theories and models since March 8, 2014. As we all know, the likelihood of this data being suspect grows with each passing day, and is something the “discovery” of deep sea “wreckage” is unlikely to quell.
Even if one illogically sets at zero the possibility that something is amiss, here, proper data sourcing is the heart of good science. I’m interested mostly in the Météo and RMP “leaks”, but a comprehensive list of full names and titles of all “data” used as inputs to all of your advertised models works best – it would minimize the amount of my pestering you’d have to endure. Thanks in advance.
In other developments:
Re: hydro-acoustic study to which you referred last month: have you read any public response from Dr. Alec Duncan? In a critique he emailed me (originally sent to the journal’s editor), he was characteristically diplomatic and open-minded – but still left me with the strong impression there was limited rigour to this new science, and limited value to the experiment.
He also agreed with me that the likelihood the Scott Reef and Leeuwin sound blips were of the same event was somewhere in between how Senator Watling’s committee characterized it (0%) and how the ATSB’s final report characterizes it (100%). I have promised not to divulge his own gut feel on the likelihood, but all who actually read his Scott Reef work can see for themselves that he basically stays on the fence. So neither the Senate Committee nor the ATSB are properly characterizing the work. It is bizarre to me that they would each slant the science in opposing directions. Throw it on the pile of rabbit holes down which we’ve been coaxed over the years…
@TBill,
ATSB has confirmed to me that 9M-MRO used 2005 magnetic variation (declination) tables. They are unsure of the exact day of the year used. I am using 1 January 2005.
@Paul Smithson,
I have previously posted a plot a long time go that shows what you want. You can get it HERE .
There are 3 plots in this file. The first plot shows the wind vectors along the southern route.
The second plot shows CTT, CTH, and CMT route fits I have done. The CTH and CMT are essentially identical down to 22:41, and then the CMT curves more eastward than the CTH (your memory is correct on this).
The third plot compares the Inmarsat route to my CTH Route. Again, they are virtually identical down to 22:41 where the CTH route curves eastward whereas Inmarsat assumed it stayed at 180 degrees (because they had no FMS navigation model). They just fiddled the speed of the last major leg so that they could fit the 00:11 BTO with a straight south path.
Still, the very close agreement of the CTH, CMT, and the Inmarsat routes from 18:41 to 22:41 is actually quite striking. All in all, I think Inmarsat did an admirable job at the time, except for the last leg where they were unsure what the bearing should do.
@sk999,
You said: “What, exactly, does “Mean BFO error” in Table 5.1 refer to? Mean of what relative to what? Are the on-tarmac residuals relative to a bias determined by the on-tarmac values themselves, and the in-flight residuals relative to a slightly tinted bias? Any insight would be welcome.“
I can only guess what DSTG did. I suspect they picked a BFO Bias value for each flight so that the in-flight mean error for each flight was very close to zero. Therefore, the mean BFO error for the ensemble of flights (shown in Table 5.1) was also very close to zero. Because the best-fit bias values were somewhat different from each other, DSTG correctly concluded that the mean BFO error over a long time period would not stay at zero because of drift (i.e., “the mean varies with time”). I don’t recall DSTG discussing how many automatic calibration events occurred over the span of the 20 flights. We do know one occurred for MH370 at 00:16:27 (-16 Hz).
Adding in the Tarmac values shifted the mean BFO errors for each flight as well as the ensemble of flights only very slightly away from zero, as expected.
You also said: “On a related note, the DSTG had no example of a cold restart of a SATCOM while in flight. Will the BFO bias come back to its original value? It is worth noting that the BFO bias offset for MH371 was 25 hz bigger than that for MH370 (pre-IGARI).”
Who knows? That is one of the reasons why I directly requested Dr. Holland and ATSB to arrange for additional SDU thermal tests when cooled to in-flight temperatures. I believe Dr. Holland explored the feasibility of such testing by the manufacturer, but I have not received a clear answer as to whether or not those tests have occurred or will occur. At this late date, I suspect they aren’t likely to happen.
I don’t seem to have any BTO/BFO data for MH371. Where is it available? I’d like to run some examples using my BTO/BFO model.
If we had access to the bias values determined by DSTG for each of the 20 flights as well as the self-calibration adjustments, we could plot the long-term drift over a period of weeks. That would be helpful in estimating the possible drift from flight-to-flight as well as the drift during a single flight. Unfortunately, DSTG did not list those values in their report. I’ll ask Dr. Holland for them.
Bobby: Re: “I don’t seem to have any BTO/BFO data for MH371. Where is it available? I’d like to run some examples using my BTO/BFO model.”
Check your email.
@Niels: I emailed you the results for the 16 constant Mach and 16 LRC cases.
@Brock McEwen:
1) I think a fit to 6 arcs is better than 3.
2) I’m not sure what you are referring to as best fit. If you mean the great circle to the South Pole, the coordinates at 19:41 are (2.6,93.7).
3) I have not and will not compile a complete list of everybody I’ve interacted with since March 8, 2014. If you have a specific question, ask it, and if I am free to answer it, I will.
4) As we all know, the likelihood of this data being suspect grows with each passing day, and is something the “discovery” of deep sea “wreckage” is unlikely to quell.
The only thing I know is that the assumptions that went into the DSTG Bayesian analysis were incorrect. Please don’t ascribe your opinions as something “we all know”. And if the wreckage is found, only fringe elements will believe the evidence was fabricated.
5) I have no special knowledge about the information from Meteo. As for the RMP report, I have stated many times that the report was leaked to the French media, and subsequently Florence de Changy shared large portions with members of the IG.
6) I have seen comments regarding the recent paper in Nature discussing AGWs and MH370, including comments from Alec Duncan. It appears that experts in the field doubt the methodology and conclusions from that paper. I claim no expertise in this domain.
7) Has Florence de Changy backed up her claim that the simulator data “has been dismissed as a clumsy fabrication based on several simulator routes flown by Shah, not one”?
@DrB: In addition to what @ALSM is emailing you regarding the satellite data from MH371, there were two posts and many associated comments from contributors that discussed these data at length:
The Unredacted Inmarsat Data for MH370
Some Insights from the Unredacted Satellite Logs
@All
Reading @DrB’s analytics and others on possible flightpaths and FMT’s it comes up to me again the FMT probably did not took place between ~18:25 and ~18:40 or after with holding patterns, race tracks or triangle emergency legs.
All much too complicated and thought out to fill the void till 19:41.
Much less complicated and logical imo is the plane flew straight on after ~18:25 (possibly with a descent and off-set) till ~8N/~92E and made a major left turn there flying back till the 2d arc and made a second smaller right turn before or after till it’s final destination.
This will put the crash area further north ~35S also which imo fits better with the forward drift analysis.
I’m not in to the specific scientific details. It just seems more logical to me.
Is a route like this possible considering the BFO/BTO’s and all other known data?
@Ge Rijn
Yes that route is very possible. That is basically the 180S CTH path I shared a while back, and that is similar to other 180S paths. DrB just made the statement that 180S paths are the best at fitting BTO/BFO. The only question is: are we looking for paths that do closely fit BFO/BTO data, or are we instead looking for paths that don’t fit the data exactly (needed if the BTO/BFO data is imperfect).
I personally favor close match to BTO/BFO at the moment. Assuming live pilot allows almost exact match to BTO/BFO on a 180S path. Yes I think that could mean MH370 went to 92S (Indonesia FIR boundary) and came back to ISBIX/BEDAX area for the real FMT.
@Victor: thank you for responding. I’m still awaiting a response fro FdC on her source.
Specific question #1: can you please tell us the name of the person who gave you the report whose inspection caused you to publicly dispute FdC’s reporting that 11 of 12 flaperon serial numbers did not match – and that the 12th was a handwritten number (?). If I recall, you were impressed by the fact that “only” 9 of 12 serial numbers were wrong.
@Brock McEwen: Don’t attribute statements to me I didn’t make. I never said I was “impressed by the fact that only 9 of 12 serial numbers were wrong”. When you distort statements, sources are less willing share information.
Here are the facts: Florence asked me to look over the evidence that she was able to obtain related to using the serial numbers from the flaperon to link it to 9M-MRO. She was doubtful that a positive ID could be made. It was also clear that she was confusing the serial numbers of the parts and the airframe.
Based on my assessment, I told her I had little doubt that the flaperon was from 9M-MRO. There were serial numbers from 4 part types: front spar, rear spar, cleat, and fittings. Of these, only the front and rear spars had recordable numbers. That made the serial numbers for the cleat and fittings useless for identification, but not “wrong”.
For the front spar, the part number and serial number were clearly legible, and these were related to flaperon serial number 405 installed on airframe 404, which was registered as 9M-MRO. The numbers on the rear spar were harder to read in the image she had, but by comparing the fuzzy numbers to what would be expected for a rear spar from this flaperon from 9M-MRO, a match was plausible. I am not aware of any “wrong” serial numbers.
@Victor
Regarding 6) I have seen comments regarding the recent paper in Nature discussing AGWs and MH370, including comments from Alec Duncan. It appears that experts in the field doubt the methodology and conclusions from that paper. I claim no expertise in this domain.
Do you have a link or correspondance your willing to share as regards to the comments from the experts about that particular paper?
Thanks in Advance
@Joseph Coleman: None that I could share without permission.
@Victor
Ok thanks anyway. I won’t bug anyone for these comments. As a layman to most things it just seemed there was slightly different results as to distances and directions of source compared to the previous HA01 study in particular for SIO. It does seem interesting if just co-incidental that the recent paper points to a source further south close to where a lot of searching was done and found nothing related to MH370. It’s sounds like an interesting study as most others we have seen but still very vague as to at least a near close to confidant place to start looking again.
@Lauren H @all
I am embarking on an exercise to look at “exact solutions” to the 180S paths. Just getting started, but tentatively looking like:
Exact Solutions to 180S Paths (Work in progress)
CTH: M0.810 from about ISBIX (final Arc5 longitude 93.15E @ 478 GS)
CTT: M0.845 from about ISBIX (final Arc5 longitude 93.70E @ 498 GS )
CMH: TBD
CMT: TBD
LRC: TBD
where: Exact solution defined based on assuming Inmarsat Arcs 3-5 BTO’s are accurate and that the aircraft autopilot settings and behavior (altitude/speed/heading) was constant during this period.
Does this make sense?
Joseph Coleman: I’ve been a party to a new discussion recently about the SIO hydroacoustic data, stimulated by Usama Kadri’s paper in Nature. The “E2 event” described therein does not appear to have any possible connection to MH370. The E1 event appears to have occurred too far NW of the 7th arc, but the HA01 bearing (301.6 degrees) to E1 raises some questions. When A. Duncan et. al. first identified this HA01 signal in 2014, it appeared to be coming from a direction too far north of the search area identified from the Isat data, and the distance was too far. Since then, we have debris analysis and Z’s sim data pointing to locations further north east, potentially consistent with the 301.6 bearing. And Kadri’s distance put’s it closer to the 7th arc, though still too far NW. This is why it is being reconsidered now.
The experts in the discussion (not me) are skeptical about this E1 observation being related to MH370, but everyone agrees that if the range estimate is wrong for some reason TBD, the bearing is potentially important. So we are looking into that possibility. It is a long shot, but we are working on it.
Table 1 SWAMS (43.0 South 75.6 East), and NEAMS (31.6 South 83.2 East) ceased operations 01-01-2014. What a shame.
Then of course, I seem to remember that there was a comment by former French President Nicolas Sarkozy, that France had some information, but he would not elaborate.
Perhaps what he was referring, to came from the OHASISBIO-5 network ?
@Andrew,
“I’m not aware of any other purpose. I’ve already given you my view on the likelihood of MH370 flying a triangular pattern; I really don’t think it has ‘legs’ (if you’ll pardon the pun).”
Thanks for your comment. Well, the FMT was born without ‘legs’. But this did not prevent the ATSB, DSTG and IG and many others from using respective assumptions in their models for more than 3 years.
Triangular patterns are interesting because they have a discrete set of possible headings, one of which is consistent with the heading into the SIO, thus offering a trivial explanation of how the plane ended up in the SIO.
@DrB,
“This route is an excellent fit to the satellite data, except that the endurance is a few minutes too long using my new fuel model. Another issue is why a pilot would set the NORM/TRUE switch to TRUE at MH370 latitudes.”
I think it is time to admit that “the best fit” is not “the most likely”.
In addition, the “best fit” depends on the metrics you introduce to quantify errors. I trust you are using RMS, correct? If you use the smallest absolute error, you will get somewhat different result. Which one would be preferred? Also, in general case you need to assign weights to the BFO, BTO and fuel errors, so that the “best fit” trajectory would be dependent on these weights.
@TBill,
Can you enter 2D/3D wind field in your simulator?
@Oleksandr,
My route fitter minimizes an objective function by simultaneously varying numerous route parameters. The key to effective route fitting is actually in selecting the form of the objective function. My objective function forces close matches, if possible, between the calculated RMS values of key error parameters and their independently-determined expected values. Currently I use 29 microseconds for the RMS BTO error, 4 Hz for the RMS BFO error, and 5.7 kts for the RMS tailwind and crosswind errors. I also require the absolute endurance error to be < 8 minutes (2-sigma). I don’t fit the temperature deviation error since the temperature error using GDAS is small (probably < 1 C).
I always start with equal weighting factors for the RMS components. Sometimes I adjust those weighting factors to improve the overall fit, taking into account the expected deviation of a RMS metric when the ensemble size is small (which it is for all the parameters fitted using a RMS figure of merit). Generally we have about 5 numbers to work with, so the estimate of the RMS error is itself "noisy" due to small-number statistics. That error in the calculated RMS statistic (based on 5 samples) is about half the calculated RMS value. So if we calculate an RMS of 30 microseconds for the BTO error, the uncertainty in that estimate is about 15 microseconds (1-sigma). Thus one should really only require the calculated RMS BTO error to be in the range of 15 to 45 microseconds in order to be consistent with the expected value of 29 microseconds (which is based on large-number statistics from 20 flights).
I am not forcing RMS errors to be zero. That is an incorrect method.
I believe my approach causes the "best fit" to be at/near a peak in the probability function. Note that I said "a peak", not "the peak". It remains to be seen exactly how many "peaks" the satellite data allow, but this number is not large. For the routes without an extended loiter (which I have not modeled in detail), I think it is 1-3.
@sk999,
After re-reading the DSTG report for the nth time, it seems they set the BFO bias value for each flight in order to make the (non-outlying) BFO errors zero using the Tarmac data, except for the log-on requests, which they ignored for this purpose. In other words, they used R1200 data but not R600 data (used for log-on requests only). However, the R1200 events used R8, R11, and R4 transmit channels, which have bias offsets with differences up to 4.6 Hz. There is no indication they corrected the BFO error data plotted in Figure 5.5 for channel-to-channel bias differences. Doing so would obviously produce a narrower probability density function.
I think the width of the Figure 5.5 PDF is caused by two effects: (1) the uncorrected channel-to-channel bias differences and (2) the frequency drift during a single flight. The problem is, we don’t have enough information to accurately separate the contributions of each effect. We do know the channel-to-channel bias differences are significant, and if those were removed, the PDF should be noticeably narrower (with a smaller standard deviation). For that reason, I think it is proper to require a somewhat smaller RMS BFO error than the DSTG calculated when one is fitting BFO data which has already been corrected for the channel bias difference, as I do.
@DrB, @sk999, @DennisW: In the DSTG report, there are disconnects regarding the BFO error that I have never been able to reconcile. On one hand, the DSTG advises that the BFO bias should be modeled as a prior mean given by the tarmac value, but with a standard deviation of 25 Hz due to bias drift. On the other hand, the standard deviation is 4.3 Hz for the 20 previous flights using in-flight values and excluding outliers, including BOTH noise and bias variation, and with the mean BFO error only 0.2 Hz. Then they say their model uses a standard deviation of 7 Hz to be conservative. Those three statements don’t appear to be consistent.
I agree regarding the inconsistent/incomplete description of BFO errors. It is also troubling they did not mention the automatic calibration procedure effects. Perhaps they ignored or corrected the Tarmac data taken before an auto-cal occurred (which I think should have occurred on some of the 20 flights and always on the ground before taxi).
Bobby: Re “Perhaps they ignored or corrected the Tarmac data taken before an auto-cal occurred (which I think should have occurred on some of the 20 flights and always on the ground before taxi).”
My understanding from ATSB is that those OCXO calibrations only occur on boot if at least 26 hrs have past since the last OCXO offset adjustment. So they do not automatically happen prior to every flight.
Victor,
The description of the use of the tarmac value as a “prior” on the in-flight BFO bias is what throws people. In Bayesian analysis, a “prior” encapsulates what knowledge you have about a parameter in the absence of new information. In practice, the BFO bias is determined the way you might imagine – make the mean of the BFO residuals zero, but other parameters are affecting the mean as well, so the prior is treated like a direct observation of the BFO bias, but with a large error, so it doesn’t skew the answer too much. Anyway, that’s how I implement them. All parameters are assigned priors, even if “diffuse”.
ALSM,
Has the self-calibration procedure been described here before? It keeps getting mentioned, but how it works, how long it takes, and what impact it would have (e.g., are communications suppressed during the process) are mysteries to me and any clarification would be appreciated.
@sk999,
I would like to extend my BTO and BFO models to include the time frame of MH371. That would allow us to compare calculations for that flight and perhaps understand the “BTO drift” you reported earlier.
What I need are lists of X/Y/Z and Xdot/Ydot/Zdot and δf(sat) + δf(AFC) at time intervals of an hour or less. Perhaps more time points are needed for the EAFC term if it is changing rapidly.
I can then post a revised model that covers both MH371 and MH370.
The inputs to my model are time, lat, lon, flight level, temp deviation, true course, ground speed, and ROC. We also need a list of those parameters for the MH371 BTO/BFO measurement times. You must already have generated this from the ACARS reports.
Perhaps others also have this information and can provide it. Thanks.
sk999: I asked ATSB about the 17 Hz jump circa 16:00 UTC. They confirmed it was a routine recalibration of the OCXO offset. Note that adjustments can only be made in increments of 4 Hz. Thus, the adjustment at 16:00 was 16 Hz (~17 Hz observed). I then asked ATSB to inquire w/ Inmarsat about the details of the calibration process. But ATSB never got an answer to my questions.
From my prior experience with similar mobile satellite systems, and the Inmarsat system in particular, I can make an educated guess that the system uses the GES (Perth) clock to calibrate the offset. As far as I know, there are no independent clocks on the aircraft better than the SDU OCXO. If that is true, then there is no way to calibrate the SDU OCXO using another independent standard on the aircraft. Of course, there is a GPS on board, but I’m not aware of any mechanism to compare the GPS clock to the SDU clock. OTOH, it would only take a simple algorithm at the Perth GES to measure the BFO bias on the ground and send a command back to the SDU to bump the local standard 4*N Hz.
@DrB: I asked Inmarsat for the EAFC + Satellite terms during MH371. Inmarsat said they would check to see if that data was available (which was a bit odd considering they needed this data to determine the BFO error for previous flights). I never heard back from them. The best we can do is approximate these terms based on the periodic data we have from March 7 – 8, 2014.
@sk999: I think of the standard deviation on the prior value of the BFO bias in the same way as the prior value and standard deviation of the speed and track at 18:02.
So, I’ll give you a specific example. For LRC speed, FL350, great circle path, and with initial conditions at 19:41 of a track angle of 168° and position of (9.0,82.9), the path crosses the arc at 26.6S latitude. With a bias of 150 Hz, the RMS error for the BFO is 7.8 Hz and the mean error is -7.3 Hz. If I increase the bias to 157.3 Hz, the RMS error is now 2.8 Hz and the mean error is 0. Is this an acceptable BFO error? (The RMS BTO error is 32 μs.)
@sk999: Here is response from the ATSB regarding the SDU frequency calibration:
There is an automatic SDU calibration process to correct log-term OCXO drift. Analysis of the GES logs from 9M-MRO from 04/03/14 to 08/03/14 shows three such probable calibration events at: 04/03/14 11:43:32 (-15Hz), 06/03/14 00:35:25 (+16Hz) and 07/03/14 16:00:28 (-17Hz). The actual adjustments will have been +/-16Hz, and these three calibrations appear to show a system adjusting around nominal as seen in SDU logs from other aircraft. This calibration is performed after logging-on and, if necessary an adjustment is made. For an adjustment to be made, one hour must have elapsed since power on, and 26 hours of on-time must have elapsed since the last adjustment. It is not possible to know from the GES logs, when and for how long, the SDU on board 9M-MRO had been powered on prior to flight MH370.
@Oleksandr
“Can you enter 2D/3D wind field in your simulator?”
As it pertains to MH370, all I know how to do is adjust wind manually. I can set different winds in different levels of atmosphere. You can also set weather at weather stations, but all I got is COCOS.
Bobby Ulich,
You say, “I would like to extend my BTO and BFO models to include the time frame of MH371. That would allow us to compare calculations for that flight and perhaps understand the ‘BTO drift’ you reported earlier.”
Yes. I’m all ears!
But then you request, “What I need are lists of X/Y/Z and Xdot/Ydot/Zdot and δf(sat) + δf(AFC) at time intervals of an hour or less. Perhaps more time points are needed for the EAFC term if it is changing rapidly.”
And further, “The inputs to my model are time, lat, lon, flight level, temp deviation, true course, ground speed, and ROC. We also need a list of those parameters for the MH371 BTO/BFO measurement times. You must already have generated this from the ACARS reports.”
First of all, more than half of the data that you request do not exist in the form that you need – I would have to generate them (e.g., satellite parameters.) That introduces a hazard. Second, most of the rest of those data (e.g, ground speed and true course) are not fundamental but are derived from primary sources. Yes, I have computed them. But what if my derivation was faulty? I would be more than happy to provide and/or generate any numbers needed for a cross-check. I am less than happy to generate numbers simply to be used as “fundamental” input.
As usual, for reference, my index of reports is here:
https://docs.google.com/document/d/14hleZyx1pUPL44yaeHKt6jnSQ3DbgRq2zibbKkFLq2c/edit?pref=2&pli=1
I wrote multiple reports r.e MH371 back in June and July (2017), including one on the AFC+SAT correction. That was a squirrely effect to figure out. For MH371, there is not such thing as “δf(sat) + δf(AFC)” as Inmarsat provided for MH370 – I had to use a different method.
Finally, all of my calculations and results are based 100% on publicly available data.
ALSM & Victor,
Thanks for your input, but it is still rather unclear what is used as the calibration reference. Yes, one can use a clock to calibrate a frequency, but one would need to have a coherent source and count cycles. (I’ve done it with pulsars.) I highly doubt that the OCXO allows one to do that. Alternatively, the SDU might simply monitor the Psmc frequency (a la Rockwell Collins), which is known, As long as you are on the ground and the satellite orbit is known, most other extraneous effects are removed. Or something else.
Sk999: Every GES has a freqency standard (aka a clock) good to 10^-15 or so. Usually it is a cesium or rubidium oscillator. The down conversion chain and all the CUs are phase locked to the station standard. So the CUs can measure the incomming carrier frequency to a fraction of a Hz, ignoring the pll noise. That is how it can and has been done in some systems. The SDU does not need to trim the ocxo frequency. It simply uses the NCO divider chain to add an offset.
@Victor
re: BFO bias:
>In the DSTG report, there are disconnects regarding the BFO error that
>I have never been able to reconcile. On one hand, the DSTG advises that
>the BFO bias should be modeled as a prior mean given by the tarmac value,
>but with a standard deviation of 25 Hz due to bias drift. On the other
>hand, the standard deviation is 4.3 >Hz for the 20 previous flights
>using in flight values and excluding outliers, including BOTH noise and
>bias variation, and with the mean BFO error only 0.2 Hz. Then they say
>their model uses a standard deviation of 7 Hz to be conservative. Those
>three statements don’t appear to be consistent.
My reading of the situation is as follows:
i) The BFO offset (delta f bias) is modelled as a slowly varying function of time, with a prior mean of the tarmac value and a prior standard deviation of 25Hz. Each flight’s BFO data is then analysed (“its distribution conditioned on the other states can be estimated with a Kalman filter”), adjusting the BFO offset distribution, to minimise the mean of the BFO residual noise (omega k BFO) for that flight. The BFO residual noise is then as far as possible the random noise at each measurement.
ii) Fig 5.1 shows the values of the BFO residual noise, but _excludes_ the BFO offset noise. The last paragraph of p30 (of the published book) is clear on that.
iii) The paper does not give the output values of BFO offset distribution analysis for each flight. It states that since the BFO measurements on the accident flight are far apart in time, the noise from the bias variation is uncorrelated, so can be added to the residual (random) noise. The adjustment from 4.3Hz to 7Hz would suggest that the bias noise is evaluated to have a standard deviation of ~5.5Hz.
@Richard Cole: Again, for the flight I generated above, would you say that the BFO error falls within acceptable bounds?
@Victor
Assuming that you are fitting 5 data points (BFOs at 5 arcs) and for the first case only the starting point (so not the on-tarmac bias), then the first case is not excluded at the 90% level (using the chi-squared test – my experience). The second case looks like an overfit, reduced chi-squared is much less than one.
@Richard Cole: These are not two cases. They are both the same path in which a great circle path at FL350 is found (after imposing the starting track angle) starting at 19:41 which minimizes the RMS BTO error at 19:41, 20:41, 21:41, 22:41, and 00:11, and the BFO errors are examined only after the path is generated. Using a constant bias of 150 Hz results in an RMS error of 7.8 Hz and the mean error is -7.3 Hz. By incorporating the mean error into the bias, the mean error is eliminated and the RMS error is reduced to 2.8 Hz. Again, using your interpretation of the BFO error statistics, would you say this is an acceptable path? By what logic?
@Victor
The basic point, made before, is that the noise in the data must be preserved in the fit. So the BFO noise of 7Hz, or something like it, must remain. An RMS error of 2.8Hz on a set of five datapoints, when the dataset has a noise of 7Hz, will happen in less than 5% of cases. It’s not excluded at very high significance, but the data suggests it’s unlikely and that the extra parameter (changing the bias) is not justified, or indeed required as the first set of parameters is not excluded.
The 7Hz BFO noise is a big number and makes it hard to exclude a lot of tracks.
@Richard Cole: I apologize for repeating the question, but I wanted to be sure that we fully understood your interpretation. For MH370, if we use a BFO bias of 150 Hz (the tarmac mean), and if we had many samples during the flight, we would expect the BFO error to have a standard deviation of 7 Hz, of which 4.3 Hz is due to the (short term) BFO noise and 5.5 Hz is due to (long term) bias drift, and the variances add since the uncertainties are not correlated.
@Victor
That’s my interpretation. Formally, the figure of a standard deviation of 7Hz only applies to datapoints that are well separated in time (which applies to the accident flight), with separations greater that the minutes timescale of the drift. It is probably not the standard deviation of the whole set of BFOs, including short term noise and long term drift, since the latter errors are correlated.
We haven’t been given the distribution of the long-term bias drift so we have to take their word that it can be represented as a normal distribution (for well separated samples) and can be added to the short-term noise and the sum remain normal.
@Victor
erratum
…probably not the standard deviation *of the errors* on the whole set of BFOs including short term noise and long term drift…
@Victor
I was wondering…
The BFO bias of 150 Hz was after the 16:00:27 recalibration sequence.
Can we have confidence that the 16Hz adjustment was saved to nvram and reapplied after the power-on reboot at 18:25 or did the BFO bias revert to the pre 16:00 value?
Is there any way to tease out the answer from the post 18:25 data?
@George Tilton: As the calibration is meant to compensate for long-term drift, I would be surprised if calibration value was not retained. A more general question is whether the BFO bias changes after a shutdown of an hour.
@Victor
Would any bias change be the same for all channels (incl. the sat phone channel)? Inmarsat did not seem too concerned about drift it when putting their JoN path together.
sk999, Victor, Don, George…
Re: method of OCXO ground calibration
Don and I had a chat this morning about the OCXO calibration method. Don took a look at the SU packets sent to the AES immediately preceding the -16 Hz adjustment at 16:00, but found no evidence of that any of those SUs looked like a “OCXO calibration command”. I am now more inclined to believe the calibration procedure uses the P channel carrier frequency as the reference, as suggested by sk999.
Although the P channel carrier frequency was an obvious technical possibility from the start, I was reluctant to go there for several reasons. First, I thought the use of the P channel as a reference would be covered under the Rockwell patent, thus not permitted. Second, the P channel has some uncompensated C band uplink Doppler due to the Miteq firmware bug. Third, the L band downlink has uncompensated s/c motion induced Doppler. However, after further review of the patent, I don’t think the use of the P channel for OCXO calibration purposes would be covered, since the patent is all about Doppler compensation, not OCXO drift compensation. Perhaps the residual C band uplink Doppler and uncompensated L band downlink Doppler (with the plane on the ground) account for the non zero BFO Bias observed (~+150 Hz). This would also explain why the Bias changes a little from one cal to the next (depending on s/c and a/c state?).
Re: “Can we have confidence that the 16Hz adjustment was saved to nvram and reapplied after the power-on reboot at 18:25…?”, I think that is a safe assumption. Once calibrated, the calibration should not change until the next calibration occurs, which requires at least 26 hrs of operating time to pass. Obviously, the AES power could be interrupted several times during that 26 hr period, so the design would need to account for that.
TBill:
Re: “Would any bias change be the same for all channels (incl. the sat phone channel)?”
The BFO bias is approximately, but not exactly the same for all channels. Although it is always expressed as a frequency offset *term*, it is actually a *factor*, not a term. Since the bandwidth (~40 MHz) is small compared to the carrier frequency (~1600 MHz), applying a correction factor to a carrier at, say, 1630 MHz gives approximately the same offset in Hz as applying the same factor to 1640 MHz.
@ALSM
Thanks…
I suspected that the recalibration is internally triggered by the SDU power-on time when I did not see any SU command from the GES.
In my BFO model the 18:27 and 18:28 BFO values are 19Hz higher than predicted.
Obviously your OCXO warm-up transient would account for much of that even ~2 minutes after power was restored. The manufacturer of my bench frequency counter recommends ~10 mins warm-up to meet spec.
George: Yes, it takes several minutes for equilibrium to be reached. The last few hZ of settling can take much longer than the first 100 Hz.
The devil is in the detail. It is important to understand the detail wrt the settling at 18:25-18:30. Once the thermal control loop settles in at the setpoint temp, the precise physical point at which the temperature is measured is held within a few milli-Kelvins of the setpoint temperature, but 3D thermal gradients throughout the OCXO thermal envelope continue to settle. Those gradients continue to settle for several more minutes depending on the thermal mass distribution properties, which act like a low pass filter to the heater input.
The residual temperature coefficient for an OCXO exists mainly due to the finite gradients inside the envelope. As the external ambient temperature changes, the setpoint temperature stays constant, but the difference between the wall temperature and the point of measurement changes due to imperfect insulation. Thus, the gradients change. This is also why a double wall OCXO has much better temperature stability than a single wall.
Richard Cole,
I was waiting for DennisW to say “balderdash”, but if he won’t, I will.
The DSTG BFO residuals plot shows a standard deviation of 4.0 Hz without even correcting for channel-to-channel bias differences. Those are up to 4 Hz, That means that the standard deviation is actually noticeably smaller than 4 Hz after these corrections are made, and probably near 3 Hz.
The calculations of residual errors illustrated by DSTG is exactly the same as what we do when we find a BFO bias value that gives zero mean BFO error on the Tarmac at zero speed. Their published distribution is what is left if one analyzes the rest of a flight using that one value for BFO bias.
Your notion that the BFO error has 7 Hz standard deviation after Tarmac bias calibration is obviously incorrect. You will also recall that Inmarsat used a PEAK error limit of 7 Hz, which for small ensembles is consistent with ~3 Hz standard deviation.
@Victor: Thanks for the reply. So is it fair to say, then, that FdC has looked at the same evidence you have, but come to an opposite conclusion re: flaperon authenticity (your extreme confidence vs. her extreme doubt)?
Perhaps we should all have a look at this evidence, so we can judge for ourselves.
@Brock McEwen: Exactly where has she expressed “extreme doubt” that the flaperon is from 9M-MRO, other than her questioning of the missing ID plate?
@Brock: You managed a comment with no scare quotes [“evidence”], no demand of names and only one insinuation that our host is an agent of disinformation… You’re slipping!
@DrB,
“My route fitter minimizes an objective function by simultaneously varying numerous route parameters.”
As you know I am also using similar approach. My point is that a resulting trajectory depends on the exact formulation of your objective function. Why would minimizing RMS error be better compared to, say, max abs?
Also, take a look at DSTG Fig. 5.4. To me it is obvious that we miss some term/formulation. Gysbreght noticed that period of ‘structured’ oscillations resembles Schuler’s oscillations, but amplitudes would be smaller as per paper found by Dennis. We don’t account for the cause of the ‘structured’ variation (aka “geographic dependence”), so what are the real BFO errors in our simulations? Can be as high as 25 Hz by the end of flight.
Another issue. When I added fuel into my minimization functional (in the form of the final mass error, which is accurately known), a new problem popped up: the amount of fuel which was consumed between 17:30 and 19:41 is not known. Of course, if one considers only straight constant level path, this problem does not exist. But any maneuvering (e.g. triangular patterns, racetracks, etc.) would result in more fuel consumed during 18:25-19:41. On the other hand, descent would save some fuel.
@TBill,
Re: “As it pertains to MH370, all I know how to do is adjust wind manually. I can set different winds in different levels of atmosphere. You can also set weather at weather stations, but all I got is COCOS.”
If I understood you correctly, you cannot feed a 2D/3D field (GDAS was in my mind) into your simulator, because this would require enormous effort, correct?
@DrB,
@TBill,
P.S. Btw, deviation of 23:14 BFOs from the “trend line” could also be caused by the same unknown effect as seen in DSTG Fig. 5.4. This is another explanation.
@Oleksandr, @TBill: There are commercial packages that will generate a file with the high altitude wind and temperature fields for the entire earth, and those can be imported in FSX and FS9. Some of those packages also offer the capability to load historical data, and not just the present conditions. I haven’t taken the time to do this because I don’t use FSX for path reconstructions. I have other tools much better suited for this purpose (MATLAB and EXCEL programs).
@DrB
> Their published distribution is what is left if one analyzes the rest of a
>flight using that one value for BFO bias.
That’s not what the report says. Selected phrases from section 5.3:
– The bias term δ f bias k (xk , sk) is time varying. In the BTO case the variations in bias were slow enough to be ignored within a single flight and we were able to assume that T channel was constant for each flight. This is not the case for the BFO bias term. The mean bias is different between flights and even within a single flight there is evidence of structured variation.
-The variations in [BFO] bias shown in Fig. 5.4 happen over a timescale of minutes rather than hours.
– However, the drift of the BFO bias means that it is not sufficient to assume that δ f bias k (xk , sk) will be the same in flight as on the tarmac before takeoff.
– The mean BFO error [in table 5.1] was close to zero in all cases, indicating appropriate δ f bias k (xk , sk) values were chosen for each flight.
Fig 5.4 shows the BFO errors for the 2-Mar-2014 to Kuala Lumpur. There are dozens of datapoints with errors between 8 and 20Hz. If the data plotted in fig 5.5 is the total BFO error (rather than just the residual measurement noise after removal of the time varying bias) where are those datapoints with large values in the figure?
>Your notion that the BFO error has 7 Hz standard deviation after Tarmac bias calibration is obviously incorrect.
I know you want the BFO error to be small, but the report does not sustain that position.
ALSM, Victor –
It would seem that monitoring the P-channel frequency is the likely calibration mechanism. Are adjustments limited to +/- 16 hz in any single adjustment? I agree with ALSM that there is residual uncompensated Doppler in the frequency that is received, and it can be sizeable, particularly if Inmarsat does not compensate the uplink C-channel Doppler shift. [I doubt that the Miteq pilot receiver is involved here, not that it matters.] Otherwise huge jumps are possible, which would explain why the few steady-state BFOs in the Holland paper are so widely spread.
Bobby Ulich,
I, too, apply channel-dependent bias corrections to the MH370 data. However, I couldn’t get them to work for the MH371 data. At that point I gave up.
@Victor,
Re: “I have other tools much better suited for this purpose (MATLAB and EXCEL programs).”
Yes, indeed.
Air/Ground signal in Boeing 777-200 turned out to be a rather sophisticated product of various sensors (WOW (weight on wheels), TILT-UNTILT, airspeed, ground speed, Mach number, altitude and pitch) analyzed by dedicated hardware in the EE-Bay:
http://www.google.com/patents/US5826833
So, what is supplied to the SDU/AES: Air/Ground or WOW?
And the most interesting question: what does it do to the SDU/AES?
@Oleksandr,
You posed some good questions. You said: “Why would minimizing RMS error be better compared to, say, max abs?“
We know from DSTG that the probability distribution functions are approximately gaussian. When this is the case, the RMS statistic is much more reliable than a peak metric. The reason for this is simply the fact that the RMS statistic uses all the data, whereas the peak metric only uses one data point. If the distribution were uniform, then a peak statistic would be preferred.
You also said: “. . . what are the real BFO errors in our simulations? Can be as high as 25 Hz by the end of flight.”
No. The DSTG probability density function for PDF includes ALL sources of error occurring after setting the bias so the mean Tarmac error is zero. That plot includes noise, quantization error, structured variation (≈± 1 Hz and which is contributed to by the use of low-accuracy trig functions in the frequency compensation algorithm), drift DURING the flight, and channel-to-channel bias offsets. The ONLY effect not included is the drift from the time of the Tarmac start-up for one flight to the time of the Tarmac start-up for another flight, which DSTG removed by fitting (for each flight) the BFO bias term to zero the mean errors on the Tarmac. The fact that, using this method, the distribution of residuals measured over 20 flights also has very-near-zero mean proves that the MEAN in-flight drift is actually quite small. Otherwise the mean error would be clearly non-zero. In other words, the drift over a period of days/weeks is not monotonic in the DSTG data set. That said, on any given flight, in-flight drift can occur, and it can be monotonic, but the magnitude of that drift cannot be very large (I would guess a few Hz) because the overall error distribution is not that wide. Errors are large as 25 Hz are not possible based on the DSTG method (which I also use) and their analyses. Another indicator of the expected drift over days/weeks is the automatic self-calibration step, which typically is 16 Hz. Drift of that magnitude is not expected to occur in 24 hours or less, since the auto-cal does not apply another correction unless a time interval of at least 26 hours has passed.
Using the 4.3 HZ “in-flight only” standard deviation of BFO error from DSTG, and (roughly) removing the effect of the channel-to channel bias offsets (which itself is 3 Hz 1σ), we are left with 3.1 Hz as an approximate standard deviation of the in-flight BFO errors with the channel bias corrections removed. That is my best estimate based on the data available today, and that is what I am now using to assess “Calibrated” BFO residuals (which have the channel offset differences removed in the CBFO values).
You also said: “When I added fuel into my minimization functional (in the form of the final mass error, which is accurately known), a new problem popped up: the amount of fuel which was consumed between 17:30 and 19:41 is not known.”
Exactly. It is not known, but it is predictable given assumptions. That is why my Endurance Model actually begins at 17:07. Knowing the fuel then, and the flight conditions shortly afterward, I can compute the fuel remaining at 17:22. The Endurance Model I posted has a default starting time of 17:22, but the fuel then is found using the same model from 17:07 to 17:22. Using my Endurance Model you can find the fuel available at any point from 17:22 through fuel exhaustion, but of course you have to assume the speed mode, altitude, and temperature deviation between at least 18:29 and 19:41.
You also said: “Of course, if one considers only straight constant level path, this problem does not exist. But any maneuvering (e.g. triangular patterns, racetracks, etc.) would result in more fuel consumed during 18:25-19:41. On the other hand, descent would save some fuel.”
I suspect the effect of the triangular turns on fuel consumption would be similar to that for the racetrack turns, which is a 5% penalty. You should be able to model the triangular pattern fuel consumption well using the racetrack numbers. That option is built into my Endurance Model. There is a flag to set to include the fuel penalty during Holding turns. It is called “Switch to set for racetrack pattern fuel flows” which is applied only in HOLDING or INOPHOLDING. As I demonstrated previously, the savings for a descent to a Hold, followed by a climb back to the stratosphere, produces a net fuel savings, but it is rather small.
You also said: “P.S. Btw, deviation of 23:14 BFOs from the “trend line” could also be caused by the same unknown effect as seen in DSTG Fig. 5.4. This is another explanation.“
That “unknown effect” might be possible. In my opinion, the one effect we do KNOW must occur is the difference in channel bias, and I think that is the most likely cause.
@Richard Cole,
DSTG’s report is confusingly written, but I think Figure 5.5 is straightforward to understand. It directly shows the residuals from 20 flights when ONLY ONE BFO “pre-bias” value was used for each flight (and that value was determined using the Tarmac data). In other words, the drift in bias value DURING each flight was (necessarily) uncompensated, because there are no means to do so. DSTG acknowledges this contribution (i.e., in-flight time-varying bias) to the BFO residuals, but it cannot be compensated with the sparse BFO data available. It is “included” in the Figure 5.5 residuals because it is uncompensated.
DSTG later decided, arbitrarily, to ASSUME a 7 Hz standard deviation in their modeling, but this is not substantiated by their empirical data (Table 5.1 and Figure 5.5, which includes Figure 5.4).
To quote DSTG: “The statistics show that even when outliers are discarded a standard deviation of about 4.3 Hz is applicable.” Also: “Figure 5.5 shows a histogram of the 3392 in-flight BFO errors. On-tarmac BFO errors were excluded due to the pre-biasing described above.” The emphasis is mine. This “pre-biasing” is just a fancy way of saying they found the bias value using the pre-flight (i.e., Tarmac) BFOs, and they used those bias values to determine the residuals for the in-flight data. DSTG didn’t include the Tarmac residuals in Figure 5.5 because they had already forced the mean of those to be zero by finding the appropriate bias value for that pre-flight, and they didn’t want that inclusion to mask a possible non-zero mean result for the in-flight data. They used that same “pre-bias” value for the in-flight data. That is quite a reasonable approach, and probably the best one can do in this case. If there was monotonic drift of the bias during the flight(s), the density function of the in-flight data would have non-zero mean. For the 20 flights, it did not have a non-zero mean, so the mean in-flight drift (change in bias) is essentially zero. That means the expected value of the in-flight bias drift, averaged over many flights, is zero. It is likely, however, to have a noticeable variance, and therefore it contributes to the width of Figure 5.5.
You said: “If the data plotted in fig 5.5 is the total BFO error (rather than just the residual measurement noise after removal of the time varying bias) where are those datapoints with large values in the figure?”
The short answer is, they are on the left end of Figure 5.5. Here’s the long answer. In general, there are no data or other means for determining and removing the time-varying bias drift during a flight. That is the whole point behind the method they did use, which did not attempt to do that. I suspect the one example given in Figure 5.4 with large errors was a fluke, possibly caused by a computational error or by small uncompensated altitude changes. Perhaps that possibility is one reason why DSTG analyzed the residuals with and without the outliers. Whatever their cause, the larger negative errors in Figure 5.4 appear to be the negative outliers (at -10 to -20 Hz) shown in Figure 5.5. Obviously, this (Figure 5.4) was an unusual case because there aren’t that many outliers altogether in Figure 5.5.
You also said: “Your notion that the BFO error has 7 Hz standard deviation after Tarmac bias calibration is obviously incorrect.” I know you want the BFO error to be small, but the report does not sustain that position.”
It is not what I want that counts. It is what Table 5.1 and Figure 5.5 show. Including the one Mumbai flight, which dominates the outliers, the RMS is 5.5 Hz. Without the outliers, the standard deviation is 4.3 Hz (including uncompensated channel-to-channel offsets). It is even smaller (~3 Hz) when the channel offsets are compensated.
The 7 Hz BFO standard deviation is NOT A CALCULATION. It is NOT BASED ON DATA. It is a “conservative” ASSUMPTION that DSTG made early on when BFOs were a new concept. I remind you also that Inmarsat used a 7 Hz peak error criterion, which is consistent with BFO errors being much smaller than the 7 Hz standard deviation assumed by DSTG in their Bayesian analysis, and which you incorrectly portray as a real statistic.
Sk999: A Miteq eafc is used to pre compensate for the C band uplink Doppler. It has the same software bug that the corrupts the inbound link.
@sk999,
You said: “I, too, apply channel-dependent bias corrections to the MH370 data. However, I couldn’t get them to work for the MH371 data. At that point I gave up.”
Unless someone can give me a table/equations of POR/IOR satellite XYZ and Xdot/Ydot/Zdot for the time period of MH371, I don’t plan on pursuing those BTO/BFO data either. It’s quite a bit of work, especially the BFO prediction, with no certain benefit. The BTO predictions are much simpler since you only need aircraft time/lat/lon/FL.
Bobby: Why not use Richard’s sat data I sent to you?
ALSM,
You say, “A Miteq eafc is used to pre compensate for the C band uplink Doppler. It has the same software bug that the corrupts the inbound link.”
Doh! Is there a Miteq Technical Note that describes how this bug … I mean compensation … works?
ALSM –
Addendum – in the uplink direction. I have all the technical notes for the downlink direction.
Bobby Ulich,
FYI, My analysis of channel offsets for MH371 was all based off the raw Inmarsat data, for which Victor has now provided links to more than once. My analysis of offsets for MH370 was also largely based on the raw data as well – just can’t remember. (C channel is the exception). Regrading MH370, the analysis basically mirrored what the IG had already done, so I take no credit.
However, I have abandoned any idea that channel-dependent offsets can somehow reduce the BFO errors.
sk999,
You said: “However, I have abandoned any idea that channel-dependent offsets can somehow reduce the BFO errors.”
I have not and they do.
@ALSM,
Thanks. I will have a look at the stuff you sent me after I do a few other things first.
sk999: There is no Tech Note for the firmware bug I was referring to. I was referring to the residual (uncompensated) C band Up-link Doppler (caused by the inverted Perth Latitude). That error “leaks” to the L band P channel downlink frequency. That will cause a small error in the OCXO calibration, if in fact, that is how the OCXO is calibrated.
@DrB @Oleksandr
When Oleksandr and I are talking 23:14 deviation from the trend line, I think we are both talking about the scatter in the raw readings, before correction:
216 217 216 216 216 216 216 216 216 216 216 216 217 216 216 216 216 (Steady trend) all of a sudden now we get next:
220 218 222 220 220 218 218 217 219 219 217 219
As far as Sat Call Channel Bias correction you are making (+4 hz), you’ve convinced me the BFO’s from the Sat Call Channel are not calibrated the same as the hourly BTO/BFO’s of the ping rings. But because I think pilot maneuvers could be happening at both 18:40 and 23:14, I am not comfortable assigning a +4 correction which basically assumes level steady flight at those times.
@Brock
If you are interested in the positive identification of the flaperon, this has been discussed here with the post of the photographs. (http://jeffwise.net/2017/10/03/mh370-mission-accomplished/comment-page-3/#comments).
The chance of the debris being from Mh370 is definitely more than 0%. However, the residual weaknesses in the “positive” identification argument are:
* the only number linking the part to 9M-MRO is handwritten and I am not sure if the handwritten number has been associated positively to the person in charge at that time.
* even though it is traced to 9M-MRO, i may have missed that but as far as i know, it was not positively identified by Malaysian Airlines to be on the actual flight MH370. There were actually reports in the press by the NYT and by FdG that modifications made did not appear on the recovered flaperon but i am not sure if this issue was further clarified with Malaysian Airlines. The full set of maintenance records for this part is not publically available and I personnally think that only Malaysia Airlines is in a position to positively identify the debris, linking it to 9M-MRO is only 90% of the argument. This would be usefull to convince the unconvinced.
@HB: thanks. I’m well aware FdC isn’t the only person who has publicly cast doubt on the flaperon’s authenticity. It’s what makes publication of this “evidence” (@lkr) so important. It is interesting to me that no one who refers to it can actually produce it…
@Victor: when an author, giving an MH370 book tour, calls attention to 11 of 12 identification marks not matching, with the 12th – the only match – being handwritten, I take this as an expression of extreme doubt re: flaperon authenticity. Do you disagree? Do you dispute that this is how FdC characterized the report? I’m honestly confused.
But hey: I don’t want to be a ping pong ball in this bizarre dispute among folks who’ve seen a report the rest of us haven’t. Let’s all just review it together, and draw our own conclusions. I ask you please to publish it.
@Brock
The link to the report is posted in the link i gave you.
This is for the front spar. Other parts are not mentioned.
@Don Thompson, Victor. As earlier, search rate IMO could well be dominant in looking into Victor’s query as to OI search options, if these are constrained by season. This aims to contribute towards that.
Don, about your, “I expect that the six Hugin AUVs should cover 600km² per day, taking recovery cycles into consideration”, under your 1600m individual AUV search width assumption, a 1.5 knots average speed-of-advance would be required, as you imply.
Supposing battery replacement per AUV every couple of days, that taking 3 hrs from off-task to back on-task, and also around the clock searching, AUV search speed would need to be lifted 0.1 knots to realise that advance.
Presumably ‘simultaneous’ battery replacements after a 6 fold AUV joint off-task would be impractical and anyway would extend the turnaround. Assuming that just 6 AUVs are active (any others being just spares), that would leave serial withdrawal for battery replacement by pairs as the approach most likely adopted.
AUVs serially rejoining would trail 4.8nm behind where otherwise they would have been, that is pending others joining them in turn on that line and that cycle repeating. The 4.8 nm would increase maximum telemetry distance only a little and presumably the stagger would not affect performance otherwise.
The search distance loss by not supplementing those 6 AUVs would be 2.4nm/day or 6.3%.
The 3hrs would depend on the overall closing rate from depth. Allowing 1¼ hrs for both up and down, to achieve that from 6000m would require a vector component of 1.6 knots in the ship’s direction. A vertical vector of 2.7 knots would be added. Overall more than 3.2 knots water speed would be required to meet these alone. Hence an AUV speed increase would be needed to meet recovery plus redeployment time.
A second horizontal component perpendicular to the ship’s track would be required also, that becoming larger as AUV stationing distance out increases.
From 5000m in 1¼ hrs would require 2.8 knots closing speed towards the track. The resultant overall average water speed needed for recovery and redeployment for the outermost pair would be 4.2 knots, entailing higher battery consumption during that total of 2½ hrs; that is if timing and the stagger cycle between pairs are to be preserved. In other words the inner pairs would have higher reserve battery capacity when withdrawn, all pairs having the same on-task time.
This basic model might be useful depending on any changes to its underlying assumptions and also on search speed being central to OI search options.
@Oleksandr,
From the abstract of the patent you reference, “provide a signal to a flight control system of the aircraft to indicate whether the aircraft is in the air or on the ground, particularly during transition from air to ground or from ground to air.”
Note the qualification, “particularly during transition“. The requirement of the PFCS to sense conditions throughout air-ground/ground-air transitions is significantly more complex than a simple WOW, binary, discrete.
Weight on wheels, WOW, is primarily detected by load sensors on the Main Landing Gear beans, the sensor outputs are processed by electronics ‘cards’ which then forward discrete signals by means of relays in ELMS and datawords on the L & R Systems buses.
ARINC 741 Part 1 (SATCOM, Aircraft Installation Provisions) states that a single bit WOW status is included in the A429 SDU-HPA command word. The same document states that ‘standard interwiring’ provides two discretes to the SDU. Multiple references describes that a ‘simple’ discrete binary WOW signal is presented to the SDU for use within the AES (at least SDU & HPAs).
What remains explicitly undefined by the ARINC 741 specifications, or the Thales SDU System Description, is how WOW is exploited in the operation of the the AES. However, the observations stand that 9M-MRO was powered, on the ground, with the ADIRU not aligned, while a SATCOM Log On was maintained from 12:50:19 until 15:56:36 using the LGA. That the LGA was specifically selected to be the preferred antenna is denoted by the Class Of AES field in the Log On Requests (ie 00b: Class 1, LGA, packet data only) for the Log On sessions 12:50:19 thru to 15:56:36. It’s entirely consistent that the LGA is employed while the SDU receives ‘IRS’ datawords containing NCD ‘IRS’ attitude data AND while on the ground.
@Oleksandr,
I’m sure you’ll note that I omitted to reference condition of ‘IRS’ position data during the period 12:50:19 thru 15:56:36. After power up and before alignment the ADIRU has position available, “the last stored position comes from the nonvolatile memory of the ADIRU”.
@David
Kudos for your curiosity and tenacity in estimating the performance of the Hugin AUV fleet! While OI appears to continue its open ocean trials at this time, I would be surprised if some more detail is not released to further describe its deep ocean search coverage performance. However, I’m content with ‘about 6x’ at this stage.
The AUV speed is variable through the mission, dive/ascent to and from the survey depth is likely to be at higher speed than when on task, scanning. The ASVs provide the command & tracking telemetry relay between the host ship and the AUVs. The mission may be modified on while at depth.
I’m particularly interested to know if there are any other developments in sonar/backscatter processing, using available sensor techologies, that will improve the search performance. The key factor in performance is speed! It’s reported that OI has invested in Kraken’s SAS technology for deployment on AUVs, my understanding is that improvements in AUV navigation accuracy has contributed to adoption of SAS on deep ocean AUVs (however, the useful range remains less than dual freq SSS). What else? is an open question.
@Brock McEwen: when an author, giving an MH370 book tour, calls attention to 11 of 12 identification marks not matching, with the 12th – the only match – being handwritten, I take this as an expression of extreme doubt re: flaperon authenticity. Do you disagree? Do you dispute that this is how FdC characterized the report? I’m honestly confused.
Please produce what exactly what she said and not what you recall. Then I will say whether or not I dispute it, and the reasons why.
@Richard Cole, @sk999, @DrB: I have closely read your comments regarding BFO error, and re-read (many times) what was said in the DSTG report. Although I do believe it is confusing, in my opinion, the interpretation of Richard is correct. The standard deviations and means tabulated in Table 5.1, and shown graphically in Fig 5.5 for the population of in-flight BFO errors, reflect the residual error after the time-varying bias is applied, which was estimated using a Kalman filter and using the tarmac-derived value as the prior for the filter. The fact that the mean values in Table 5.1 are close to zero is an indication that the calculated time-varying bias for each flight was appropriately estimated, and not that that the drift in bias was close to zero. The calculated time-varying bias was not reported, but we do know that after incorporating the time-varying bias, the residual BFO error due to noise is 4.3 Hz, and we also know that the DSTG estimates the standard deviation due to both BFO noise and drift to be around 7 Hz, which would imply that the drift contributes about 5.5 Hz, as Richard has said. It would have been helpful if the DSTG chose to show the actual time-varying bias for some flights, but they did not.
@Don Thompson. Thank you.
@Victor: please present the report so we all can see it. Thanks in advance for your transparency.
@Brock McEwen: Please present the public statements you claim that FdC has made so I can state whether or not I agree with them rather than reacting to your characterization of the statements. You have a history of distorting statements made by me and others. Here is a recent example.
As for the report on identification of the flaperon, as @HB has said, some of the source materials were previously made available elsewhere.
Here is the message I wrote to Florence in March 2016 regarding the identification of the flaperon. I do not know how she chose to interpret these statements:
I reviewed the flaperon report. In summary, I see no mistakes and nothing wrong with the conclusions that the serial numbers show that the part is from 9M-MRO. Here is why I make this claim:
1. The photographs show identification numbers from 4 part types: The front spar, the rear spar, a cleat, and fittings.
2. Only the front spar and rear spar have recordable serial numbers. I think it is reasonable that cleats and fittings are not uniquely recorded.
3. On the front spar, the Part No. 113W6142-2 and Serial No. 3FZG81 can clearly be read.
4. On Product Sheet No. 2, the front spar part number and the (hand-written) serial number are shown to be from Order No. 38G175.
5. On Product Sheet No. 1, Order No. 38G175 is linked to the flaperon Serial No. 405.
6. On the Master Schedule of Flaperon, it shows that flaperon Serial No. 405 was installed on airframe Line No. 404.
7. This Line No. 404 was registered at 9M-MRO. See for example: https://www.planespotters.net/airframe/Boeing/777/28420/9M-MRO-Malaysia-Airlines
8. The numbers on the rear spar are recordable but difficult to read.
9. Based on Product Sheet No. 2, for the rear spar, one would expect Part No. 113W6144-2 and Serial No. 3FZQ16.
10. As you can see in the figure below, the numbers on the rear spar may indeed be these numbers.
In my mind, there is little doubt that the recovered flaperon is from 9M-MRO.
@Victor: have a listen to minutes ~14-17 for FdC’s research into the flaperon:
http://www.radionz.co.nz/audio/player?audio_id=201793735
Victor, FdC was your source for this report; are you saying you did not know until now what she concluded from its analysis, and said about it publicly?
@Brock McEwen: I don’t claim to know everything that she or anybody else says about MH370. I disagree with a lot of what she says. I disagree with a lot of what most authors on MH370 say.
The particular interview you cite was recorded on March 18, 2016, which was before I supplied my analysis to her on March 31. If you want to pursue discrepancies between what she has said publicly and what I have told her, you are free to do so. I don’t try to correct every incorrect statement that is made. If you falsely equate not being able to make a match because information has not been recorded with recorded information not matching, that’s a distortion of the truth.
@Brock McEwen
My God. Where does your paranoïa and stubborness end? Please settle down.
The flaperon is 9M-MRO. And it’s not only the flaperon you know.
Many other pieces have been found that are definitely 9M-MRO too and totally fit the context of the flaperon find.
Please stop looking for nails on shallow water and come back to reality.
The flaperon is a total non-issue in this regard and within the whole context.
Your assumptions on reverse drift studies and all that came with it are out of date a long time. You’ll find no proof for it anywhere no matter how much you ask for it to anyone in your sincerely way.
Put your obvious intelligence to current reality I suggest.
You could be a lot more helpfull that way I think.
Don’t want to offend you, just wanted to speak out.
I always respect your independent input.
@VictorI,
@Richard,
I will pose a question to Dr. Holland regarding the interpretation of Figure 5.5. Due to certain current events, he may not be able to respond for a while. Hopefully he will settle the issue regarding subtracting the Kalman filter outputs, but he may or may not be able to settle immediately the issue regarding the BFO error variance (see below).
If Richard and you are correct, and a Kalman-filtered time-varying bias was derived by DSTG using only prior BFOs, and then subtracted, we also need to know what the typical drift rates/amplitudes were for the 20 flights. If those are small, then the two interpretations converge to the same result. If they are large, then the higher variance might be appropriate (but see below) unless one does the same process for MH370 of removing the time-varying bias as the one used by DSTG (but the filter used in this process is not defined in the DSTG report).
The fact that Inmarsat used a 7 Hz peak BFO error criterion, with no removal of time-varying bias, seems totally incompatible with your/DSTG’s prediction of BFO error statistics (7 Hz standard deviation). But what does Inmarsat know? They just designed and built the system.
A back-of-the-envelope calculation can offer some insight into the expected drift rate. If the auto-cal is designed to apply +/- 16 Hz corrections not more frequently than every 26 hours, the worst-case drift rate is 16/26 = 0.62 Hz/hour. Over an 8-hour flight like MH370 the worst-case drift would be 4.9 Hz, and the typical drift should be a fraction of that, or a couple of Hz. That low drift-rate estimate is also incompatible with “. . . the drift contributing 5.5 Hz . . .” standard deviation, as you and Richard (and DSTG) claim.
An additional and telling fact is that the auto-cal corrections are not applied every day or two, but typically every week or so. This confirms my statement above that the typical drift rate (several Hz over one flight) is a fraction of the worst-case estimate.
There is also a different interpretation to consider. Suppose DSTG actually did what you claim by subtracting an estimated time-varying bias. However, if that estimator were imperfect, which it certainly must be considering the paucity of BFOs and their inherent noise, then it could increase the residuals instead of decreasing them, as you and Richard (and DSTG) have assumed in computing the 5.5 Hz (1-σ) contribution due to drift. Get my drift? Sorry, I could not resist the pun.
Basically, what the Kalman filter does is fit a slowly-varying function to all the available data up to, but not including the BFO in question. When that BFO point is close in time to lots of prior BFOs, one would expect the estimator to be relatively good with low noise. However, when the data are few and are sparse (well-separated in time), the estimator is noisy. That is because you are trying to fit at least a slope, and possible a parabola, to data which itself has noise. The noise in each BFO error is substantially magnified when a slope (or higher order term) is fit. As an example, take 3 BFO errors (with equal noise error bars) evenly spaced in time, and fit a line to the first two. Then see how well the 3rd error is predicted. The slope is noisy so that the spread of the prediction at point #3, using all lines passing within the error bounds on the first two points, is twice the noise at each point. In other words, with few and sparse data the Kalman filter can add noise in the process of attempting to remove drift.
What if Figure 5.5 is actually broader than the true BFO error distribution because DSTG applied a noisy process to remove the time-varying baseline? You (and DSTG) have made an assumption that the true BFO error PDF is broader than Figure 5.5, and therefore by applying the time-varying bias removal, the distribution was made narrower. In other words, you/DSTG have assumed:
Variance (Fig. 5.5) = Variance (BFO error) – Variance (time-varying drift estimator).
This assumption is equivalent to assuming the calculated bias drift is +100% correlated with the BFO errors.
Assuming this, naturally you found the BFO error variance using this rearranged equation:
Variance (BFO error) = Variance(Fig. 5.5) + Variance (time-varying drift estimator).
That assumption led you and DSTG to conclude that the BFO error variance was larger than the Figure 5.5 variance. I say assumption because DSTG did not present any data to determine the validity of the equations above.
I don’t think this assumption (that the drift estimator is completely correlated with the BFO errors) is correct for the BFO data. More likely, I think they are mostly uncorrelated with a small correlation coefficient (i.e., closer to zero than to one). In that case, the true BFO error PDF would be narrower than Figure 5.5. This scenario is equivalent to this assumption:
Variance (Fig. 5.5) = Variance (BFO error) + Variance (time-varying drift estimator).
Note the sign change (compared to the first equation above) from plus to minus. In other words, maybe we are both partly right and partly wrong (and what DSTG did actually made it worse not better)?
In the scenario I just described, the time-varying bias estimate is removed from the BFO data. On that point Richard and you would be correct. The Kalman-filtered bias estimator is noisy because there is very little data to work with (the data are few and sparse). DSTG’s bias removal process adds the process variance (due to noise in the Kalman filter estimate of time-varying bias) to the true BFO error PDF. It does not subtract variance. In this case Figure 5.5 is wider than the true BFO PDF. In calculating the true BFO error PDF, perhaps you and DSTG have made the same error in assuming the process will reduce the variance instead of increasing it (by the process of subtracting a noisy estimator for the time-varying bias instead of a low-noise, 100% correlated estimator, which you assume and which I don’t think is even possible for these data).
We don’t have any data to determine the variance of the drifting-bias estimator. Apparently DSTG made a conservative guess that the standard deviation would be increased from ~4.3 Hz to 7 Hz. It’s too bad DSTG did not (later on) produce a PDF similar to Figure 5.5 without the time-varying bias removal (if that indeed is what they did). If my theory is correct, we know the variance of the BFO errors will be less than shown in Figure 5.5, but we can’t calculate exactly how much less.
In summary, I am suggesting that the assumption that the time-varying bias estimator is completely correlated with the BFO errors is possibly incorrect, and that an assumption that it is closer to uncorrelated is perhaps more appropriate. (the “truth” may be in between the two cases). By the way, if we had the exact formula for the Kalman Filter DSTG used, we could run it over the MH370 BFO data and see if removing the time-varying bias using the DSTG method increases or decreases the BFO residual variance for a number of good-fit routes. That would be interesting to do. In other words, we could measure the correlation coefficient between the bias drift estimator and the BFO errors. If it is +1, the BFO error variance is larger than shown in Figure 5.5. If the correlation coefficient is much smaller than 1, then the BFO error variance is less than shown in Figure 5.5.
The apparent “disconnect” between the expected bias drift rates (based on the auto-cal) and your derived bias-subtraction variance can be resolved by assuming a low correlation coefficient between the BFO errors and the estimator. This also removes the obvious “disconnect” between the Inmarsat 7 Hz peak error bound and the DSTG prediction of 7 Hz standard deviation.
@Victor: thanks for responding. Did FdC, to your knowledge, reverse her position after speaking with you, or with anyone else? I believe her book puts forth the same position indicated in the interview (I do appreciate that its publication likewise pre-dates your email to her).
Also: can we please see all the images you reviewed? I think we’d all appreciate the chance to judge for ourselves. Thanks in advance for your transparency.
@Ge Rijn: thanks much for your contribution. Wonderful to hear that you don’t want to offend me.
I will continue to work hard to ensure that only verifiable evidence shapes my views on MH370’s fate. I am trying my best, at present, to verify the endoscopic evidence; two of the only people on the planet who claim to have actually seen it have expressed to me radically different views. I have adopted neither extreme position, but rather asked – in my sincerely way – that you and I – and the NoK, and everyone else – get a good look at these images, so we can all judge for ourselves.
My last paragraph was inadvertently left off my previous post:
Please don’t think I am criticizing what DSTG did. I am not. This was new stuff when they started, and they were pretty rigorous in their methodology. Looking back with the benefit of several additional years of study, one can say some of their assumptions might be different if one had the chance to do it over again. All in all, they did a good job using the tools they had to produce some useful predictions.
@DrB: First, the methodology is DSTG’s, not mine. I am simply trying to understand what they did based on what is written in the report together with comments from you, Richard, and sk999.
If you are trying to persuade me that a bad Kalman filter can create noise, no persuasion is required. However, I have to believe that the DSTG is experienced enough to know whether the variance reported in Table 5.1 is greater or less than the total variance of the BFO error.
I believe that the Kalman filter essentially separates the low frequency error (due to drift) from the high frequency error (due to noise). You can see this in Fig 5.4, where there are clusters of values occurring over a short time (aligning almost vertically in the plot), and the clusters are irregularly spaced from each other over time. The mean value of each cluster would represent the drift.
You can also see in Fig 5.4 that the drift is not monotonic, but appears to have some periodicity, albeit with an offset. My guess is that the data before 22:00 is representative of most flights, and the data after 22:00 shows abnormally high drift due to the “geographic dependency”. None of this drift would need to be corrected by the auto-cal feature because it is not cumulative, so I don’t think you can use arguments about the resolution and timing of the auto-cal to invalidate the 5.5 Hz standard deviation of the drift.
@Brock McEwen: FdC can speak for herself. I have no idea what her current position is. Last I heard, she believed the plane crashed in the South China Sea, and Inmarsat, the US, the UK, and Australia were complicit in a cover up. If she still thinks this is true, I would be in serious disagreement with her.
You keep asking me for the materials I used to make my assessment about the provenance of the flaperon. When somebody provides me with materials in confidence, I only disclose that material with permission. I don’t have that permission from Florence. I told you the source. You can get that material directly from Florence if she wishes to disclose it to you. Stop asking me.
@Don,
Thanks for your comment.
The patent link to which I provided, appears to be somewhat inconsistent with AMM, chapter “AIR/GROUND SYSTEM – INTERFACES”. For example, WOW Card Inputs. Citation from AMM, 32-09-00 (pdf p. 4398):
“The WOW cards get inputs from:
– AIMS (time and date, flight phase/leg, airplane registration number, ICAO code)
– Bus power control unit (external power available)
– ELMS (air/ground relay status)
– Ground test enable switch
– Load sensors
– Radio altimeter (3).”
Nothing is mentioned about air/ground speed, mentioned in the patent.
On the other hand, the SDU does not seem to be listed in the users systems, to which output from WoW cards is sent.
Re: “It’s entirely consistent that the LGA is employed while the SDU receives ‘IRS’ datawords containing NCD ‘IRS’ attitude data AND while on the ground.”
Yes, consistent. However, remember that ADIRU’s internal logic prevents its from switching off when ground speed exceeds 30 kts. Respectively, the SDU does not need WoW to judge whether the plane is on the ground or not, as the absence of the data from the ADIRU is already indicative of the plane on the ground. Of course, except ‘exotic’ cases.
RE: “After power up and before alignment the ADIRU has position available, “the last stored position comes from the nonvolatile memory of the ADIRU”.”
This is interesting. So, in theory the SDU may transmit in the air without updated data from the ADIRU, unless WoW bit prevents it from doing so.
@DrB,
Re: “Errors are large as 25 Hz are not possible based on the DSTG method (which I also use) and their analyses. ”
I disagree with this. Take a look at DSTG Fig 5.4. To me it clearly indicates a superposition of a structured error and random error. The random component generally has a relatively small sigma, somewhat around 5 Hz by naked eye (I don’t have data to derive accurate numbers; DSTG does have). The structured ‘drift’ could reach 20Hz amplitude in 4 hours on that particular flight. It appears the amplitude of this structured deviation growths exponentially starting from zero. Without understanding a reason and accounting for this effect, how can you state that the BFO errors can not be as high as 25 Hz in more than 4 hours of flight?
@TBill,
“…all of a sudden now we get next…”
In the reality the plane could be affected by gust wind, and it would take some while for the plane to adjust to new conditions (stabilize yaw). This could result in temporal oscillations of the heading, which could consequently result is the sequence of BFOs: 216-216-…-216-202-218-222-220-…
The spatial resolution of wind data I am using is 1 deg, so modeled transition occurs smoothly.
@Brock,
Perhaps you could switch to somewhat closer to your profession? There are plenty of things within your domain (BTO and BFO statistics, expected number of fragments of 9M-MRO / MH370 objects; size distribution of washed ashore debris, etc.). But you are wasting your energy doubting that the flaperon is genuine. I don’t quite understand your motivation.
Oleksandr,
You wrote “The patent link to which I provided, appears to be somewhat inconsistent with AMM”
Why would you even expect a generalised patent to be in any way consistent with the specific description of a B777 system?
Be careful not to conflate “absence of data“, no data words passed over a bus, with data words beinh passed but communicating no computed data (NCD).
“in theory the SDU may transmit in the air without updated data from the ADIRU” if one chooses to ignore, contort, or misrepresent the preceding statement, i.e. “after power up, and before alignment of the ADIRU… the last stored position comes from the non-volatile memory of the ADIRU”.
Thanks for the expert discussion on BFO stats that has been illuminating.
Now, a related question on the size of BTO errors that we would expect to see from the “correct” path. If it’s the case that:-
– 1 sigma BTO bias estimate is 29 microsecs
– RMS true wind error vs GDAS model is about 6 knots crosswind and 6 knots tailwind (*is that correct or is it 6kt on wind strength rather than 6 knots zonal, 6 knots meridional?)
– heading accuracy of the autopilot is +/- 0.1 degree or so
– and we are implicitly assuming that our satellite orbit model is perfect
then…
How big would we expect RMS BTO residuals to be on any given path (before baking in the wind and heading uncertainties)? The reason that I ask is that I’d expect it to be bigger than the 20-30 microsecs shown in Victor’s charts above. The 6 kts crosswind error alone can shift you 10kms in longitude over one hour, equating to a BTO error of ~70 microsecs? 0.1 degree heading error over an hour is another 5km lateral. So when we have all of these various error noise going on in real life (observed), what do we really expect the BTO error RMS to be for a flight path model generated using an imperfect wind model and “perfect” heading accuracy (predicted)?
Could someone with stats in their blood please put me out of my misery?
@Don,
“if one chooses to ignore, contort, or misrepresent the preceding statement, i.e. “after power up, and before alignment of the ADIRU… the last stored position comes from the non-volatile memory of the ADIRU”.”
The ADIRU alignment can be started in the air. But it cannot be successfully completed. There is time in between.
@Oleksandr wrote “The ADIRU alignment can be started in the air.”
You tried that one previously.
I will add “if the airplane moves during the alignment, the align reset mode occurs until there is no movement.“, how can that be construed as starting alignment?
Further, consider that normal ADIRU shutdown is an orderly process, not the same as removing all power (which requires six cct breakers, located on three power panels in the MEC, to be opened).
@Paul Smithson,
The GDAS wind errors are given (using the only published value I could find) as 5.7 kts per horizontal axis (out of two possible horizontal axes). So long as the axes are perpendicular, you can use any axes you want. I find it convenient to use tailwind and left crosswind.
The prediction of the BTO only depends on the satellite and aircraft positions in ECEF coordinates.
The BTO measurement errors do not depend on the wind or on the heading or track (or on their errors), just lat/lon/altitude at a given time.
The BTO errors can be 29 or 43 or 62 microseconds 1-sigma depending on the channel/message type. Most of the BTO data have 29 microseconds 1-sigma.
The satellite location error is mostly removed by selecting a BTO bias value that matched the KLIA gate location. The Inmarsat/ATSB value works just fine in my model.
I normalize all my residual fitting errors to the expected 1-sigma value for that particular data point. That way I don’t have to compute some weighted standard deviation of points with various standard deviations for each point. So my expectation for a good fit is a RMS residual of 1.0 sigma. When the number of data points is not large, there is an expected standard deviation of the calculated RMS value, and it is approximately equal to the RMS divided by the square root of the number of data points – 1. So if you have 10 data points, the sigma of the calculated RMS is approximately RMS/3. So one should “accept” fits (at the 1 sigma level)which have RMS values of 1.0 sigma +/- 0.3 sigma, or from 0.7-1.3 sigma.
Pual Smithson,
As an initial matter, the BTO errors in the measurements are what they are and do not care about the wind. What I think you are asking about is errors in the prediction (i.e., a constant heading path).
Wind errors seem to be uniformly quoted on a “per-axis” basis – thus, 5.7 knot rms from Bayesian Methods is 1-d. One thing that BM makes obscure is the spatial correlation of the wind errors. BM treats correlations as a decay in time with the time constant of 16 minutes (roughly 2 degrees of arc at typical aircraft speeds), but that is incomplete. From a study I did ages ago, the rms seems to be independent of direction, but the correlation length is roughly twice as great in the longitude direction as in the latitude direction. Since the flight path of MH370 was along a predominantly N-S line, it is the latter that matters most. Over 1 hour, the mean error in wind speed is less that the instantaneous 5.7 knots rms, since some averaging of errors will occur. However, another factor is at work that goes in the opposite direction. Most of the flights analyzed by the DSTG were in areas well-trafficed by aircraft, and aircraft-derived wind data are incorporated into these global models. MH370 went South into a region that is poorly sampled (since very few commercial jets pass through that region), so the models may have errors in those regions that are worse than where the DSTG made its comparisons.
In my constant heading models, I allow for independent N/S and E/W wind error parameters, one per hour, and I impose a prior of 0 knots with an error usually of 4 knots rms, but sometimes bigger.
Back to your question – the best approach is to recast the wind errors into components parallel to an arc and those perpendicular. Thanks to Pythagoras, the rms errors in each direction are the same 5.7 knots. Only errors in the direction perpendicular to the arc affect the BTO. Further, these errors tell you directly how much the plane is shoved in that direction, independent of the heading or speed of the aircraft. Thus, a 5.7 knot wind will shove the plane 5.7 nm, and contribute an error to the BTO prediction by whatever the conversion factor from nm to BTO microsec is at that point.
Now, this calculation is only good for the 1st arc. Distance errors are cumulative, so at the next arc, the BTO error due to wind at that arc is correlated with the BTO error at the 1st arc. Ain’t statistics wonderful?
Oleksandr,
You said: “It appears the amplitude of this structured deviation growths exponentially starting from zero. Without understanding a reason and accounting for this effect, how can you state that the BFO errors can not be as high as 25 Hz in more than 4 hours of flight?”
DSTG had no explanation for that odd behavior shown in the one flight in Figure 5.4. They don’t specifically characterize that one flight as being extreme or typical. My conclusion was, and still is, that it was an extreme event. I concluded this based on the fact that it contributes almost all of the residuals in Figure 5.5 from -10 Hz to -20 Hz. Most or all of the other data sets must have had much smaller amplitudes. The dip might be a satellite eclipse event. The amplitude is about right (the same 10 Hz as the eclipse at 20:00 UTC on 7 March 2014), but I am not sure about the sign convention. The duration of the Figure 5.4 event is consistent with the data provided by ATSB for the 7 March eclipse, which shows only that the event was completed from start to finish in less than two hours and was ~10 Hz at maximum effect. I don’t see the “exponential growth” you described in Figure 5.4.
The reasons I think the BFO errors on a typical flight, or on MH370, cannot be as high as 25 Hz are as follows:
1. I think the Figure 5.5 distribution is contaminated by uncompensated and misunderstood effects such as computation errors and possible eclipse events. DSTG acknowledged their existence, but nonetheless included them without understanding their cause(s).
2. I think the time-varying bias subtraction process added noise to Figure 5.4. (This is straightforward to test. All DSTG had to do was show the same plot as Figure 5.5 without the bias subtraction.)
3. There is nothing in the MH370 data up to 17:07 indicating errors outside +/- 7 Hz. In addition, it is possible to fit auto-pilot routes with 3 Hz RMS BFO errors. This is consistent with, but not proof of, small BFO errors.
4. Inmarsat made a sincere (and I think a reasonably accurate) effort to predict a route using the BFO data. They applied a +/- 7 Hz peak criterion. Clearly, Inmarsat doesn’t believe the BFO error could be anywhere near 25 Hz.
Unfortunately, DSTG did not provide the detailed data used in their BFO analyses. If we had that data, we could resolve most of the current issues.
@VictorI,
Thank for commenting on the BFO time-varying bias process. It’s too bad DSTG did not show diagrams similar to Figure 5.5 for (a) the uncorrected BFO errors, and (b) the bias corrections. That would have provided a lot of information and still avoided the difficulty of providing thousands of data points. Unfortunately, they chose to show only one diagram which is a mixture of two parameters (BFO errors and time-varying bias), and we have no means of separating them.
Regarding the size of BFO errors, my money is on Inmarsat.
Ask yourself why the “reconstructed” flights using BFO and BTO were not accurate when there was a turn involved. The larger the turn was, the larger the track error. I suspect that is a result of the Kalman BFO filter not being able to follow accurately the BFO steps when turns occurred. That will cause errors in the BFOs after the time-varying bias is subtracted (because no Kalman filter can follow a step accurately with sparse data). That introduced BFO error leads to track errors in the reconstructed path, and the reconstructed track error will grow as the size of turn (= size of BFO step) grows.
This effect could also contribute to the width of Figure 5.5. Perhaps the Kalman filter worked well for straight paths but poorly at and immediately after turns (I don’t see how it could do otherwise). I don’t think DSTG segregated the data in Figure 5.5 between turns and no-turns (although they did remove climbs and descents).
@Victor
Here is a link to a comprehensive review of the MH370 Floating Debris finds. The results compare well to your present post regarding “Possible MH370 Paths Along Great Circles”:
https://www.dropbox.com/s/ma9pt8zf3b0whc7/MH370%20Drift%20Analysis%20-%20Calibration%2C%20Results%20and%20Conclusion.pdf?dl=0
@Dr B – you misunderstood my question (and I must have framed it poorly). sk999 is a bit closer to what I was getting at. Thanks to both for initial response, but let me try again.
I do understand that the BTO error stats are just that. Noise arising mainly from jitter and quantisation creates some uncertainty in the derivation of the bias from a given position (or, conversely, uncertainty in a position from a given BTO) as long as our satellite orbital model is perfect. If that was all, then a “perfect” path model ought to have BTO residual RMS somewhere in the mid 20s. But that’s not primarily what I’m asking about.
Path models predict position and associated BTO using an underlying model wind and also presume 100% accuracy in aircraft speed and heading with respect to modes. In practice, the winds encountered by MH370 did not perfectly match the wind model and whatever the mode, the aircraft heading and speed would have varied according to the tolerance of the autopilot.
Let’s say we happen upon the actual path flown by MH370 (start time & position, altitude, speed & heading modes) and now reproduce it with our path modelling tool. In addition to the BTO noise attributable to its underlying statistical characteristics, we will also have additional error (BTO predicted vs observed) attributable to errors on wind x,y aircraft heading accuracy and aircraft speed accuracy. So our “correct” path should have bigger BTO residuals RMS than would be the case if we had perfect wind/speed/heading. And a “typical” wind error alone can contribute error of at least 10km in lateral position, or +/- about 70 microseconds over the course of an hour.
So I ask again, how big would we expect our BTO residual (predicted vs observed) RMS to be acknowledging not only the error stats associated with BTO per se, but also errors in wind, speed and heading?
@sk999. Thanks also for the explanation on the correlated nature of wind errors. I do appreciate that this is more likely to give us a “one sided” or cumulative error rather than random “noise” from hour to hour.
I did once go to the trouble of comparing the predicted wind along a candidate MH370 path using current (ie on the day I did the calculation, not historic model of 7.Mar.2014) wind models GFS and ECMWF and found RMS difference between the two of about 7kts and peak difference of >15kts.
@DrB: I don’t know the details of the Kalman filter, so I won’t predict whether or not it was properly implemented, but for prior flights, if we know the state, we know the BFO error, whether or not in a turn. Of course, imperfect knowledge of the state could increase the error in the extracted bias drift. I would think that a Kalman filter that simply sets limits on the drift rate would do fine in extracting the bias drift from the BFO noise, assuming the state is known at each BFO time. As I said, the mean error of any BFO burst is essentially the bias drift.
As for betting on Inmarsat versus the DSTG, I don’t have to bet, so I won’t. I’m simply trying to objectively understand what the expectations should be for BFO error, a matter which should have been settled a long time ago. There is an obviously strong disagreement among the modelers here. I do observe, however, that much knowledge has been gained since Inmarsat published the JON paper in September 2014, including understanding effects of vertical speed and power up transients. Much of that knowledge was advanced with efforts led by Ian Holland of the DSTG, not Inmarsat, although Inmarsat probably had input. I’m also not sure how comprehensively the BFO error for prior flights was studied and understood in September 2014. We’ve all learned a lot since then.
@DrB: A couple of points from your recent posts:
“I think the time-varying bias subtraction process added noise to Figure 5.4.”
Figure 5.4 shows the data before the subtraction process – bias error and statistical noise have not been separated. I assume you are referring to your idea that the Kalman filter added noise to the data between figure 5.4 and 5.5.
“Ask yourself why the “reconstructed” flights using BFO and BTO were not accurate when there was a turn involved. The larger the turn was, the larger the track error. I suspect that is a result of the Kalman BFO filter not being able to follow accurately the BFO steps when turns occurred.”
The discussion around page 30 of the DSTG book is about the attempt to separate and analyse the bias drift from the historic flight BFO data and a Kalman filter is used in that process. In the end, that analysis was unsuccessful and unseparated BFO data, with the larger 7Hz overall error, had to be used for subsequent BFO analysis. In the flight reconstruction a very small sample of BTO and BFO data was used for each flight to match what was available on the accident flight. There can be no measurement of time varying bias in that exercise, so no use of that particular filter.
As Victor has said, I cannot verify the DSTG analysis is correct since their data is not published (nor could I replicate it even if the data was available!). I am just trying to understand what the report states.
If we did an unconstrained model (allowing pilot input) but holding as closely as possible to the BTO/BFO data, my assertion is that probably gives “x marks the spot” search area. And it is probably fairly close to the Inmarsat JoN hot spot.
In any case, please give me that location first, and then I can accept that the BFO might be off, which indicates other search areas.
We keep saying the BTO/BFO data does not tell us where the aircraft is, but it probably does indeed give an answer. To me, what we are really saying is that we do not believe the data should be used directly, so we are not even going to calculate it. This is my criticism of the search process to date.
In my view the drift-data are as important as the other data now.
@Richard’s latest paper again suggests a crash area around 30S but anyway north of 36S.
Others point between 33S and 28S. The ones that predict areas between ~15S and 40S don’t predict anything at all imo and are useless.
Imo it’s remarkable still no other debris finds are reported. Imo still indicating a small debris field.
Still the 177 historical drifters that passed the previous priority area over 20 years are a fairly realistic indication of what the debris field looked like shortly after impact imo. It fairly acurate shows the current finds to date.
31 reached African shores and islands within 22 months and 13 reached WA shores.
But the 13 drifters to WA shores only passed south of 36S, none north this latitude.
With annual and seasonal fluctuations you would expect at least one of those drifters north of 36S would have beached on WA.
But none did in those 20 years.
33S to 28S is imo the best estimate regarding possible drift and arriving times of all found pieces within the window of reasonable opportunity.
This window started with the flaperon and closed with the RR-piece imo.
Everthing (or at least most) found later arrived earlier than RR I think.
TBill. See victor’s article above. If I understand him correctly it shows that the aircraft could be pretty much anywhere between 20S and 40S in one or other autopilot mode from 1941 onwards and obtain excellent BTO fit and reasonable BFO fit. If you believe that BFO error must be 7Hz or less then you can still pick any spot between 25S and 38S. And that’s even before you start allowing additional pilot inputs after 1941. In other words, there isn’t any X to mark the spot unless our starting point on the 1941 arc is quite tightly constrained in one way or another.
Personally I was surprised that there is such a wide spectrum of ISAT-compatible routes and I’m a bit staggered that the BTO residual RMS is as low as this over such a wide range of paths.
@Paul Smithson,
You said: “So I ask again, how big would we expect our BTO residual (predicted vs observed) RMS to be acknowledging not only the error stats associated with BTO per se, but also errors in wind, speed and heading?”
There is no single answer to your question because different routes intersect the arcs at different angles. That means that wind errors, and navigation errors, and speed control errors all produce varying BTO errors. Crosswind errors are a bit more complicated, because, in this case, the BTO error also depends on the navigation mode (more particularly, whether it is a constant heading or a constant track).
In constant heading modes, the impact of wind errors and navigation errors will be larger than for constant track modes because there is no reference except the “now” readings. What I mean is that the path errors will accumulate over a leg and from leg to leg. There is no way to know exactly what the cumulative effect is at a given point in time. For constant track and LNAV routes, the errors do not accumulate because the aircraft is following a calculated path. It wanders left and right of that path as it flies along it, but the errors do not accumulate (it always has a current reference position it is trying to maintain), and so its end position is much more accurately predicted and achieved than the constant heading mode. For constant heading modes, the effect of varying crosswinds is much larger than the aircraft’s ability to know a constant heading (i.e., the navigation error). In other words, the aircraft is blown laterally by varying winds, and we don’t know exactly what the winds were. We do know roughly what their mean value was (from GDAS, etc.).
For constant heading modes, the “navigation error” (a small fraction of a degree) is swamped by the “crosswind error” (on the order of 1 degree), so I just fit the crosswind error alone, understanding it also contains a small contribution from navigation errors.
The “tailwind error” is much larger than the speed control errors in its effect on ground speed, for all flight modes. So, I ignore the speed control errors, understanding they make a very small contribution to my fitted tailwind errors (I fit 1 per leg).
For navigation errors in constant track modes, we are only affected by the mean navigation error over one leg. The common-mode portion of the navigation error (i.e., common to all legs) is compensated in fitting the track bearing, so it does not need to be included separately. One can allow possibly ± 0.1 degree errors in the track angle from leg to leg, and I have done that as well.
For LNAV (waypoint) navigation, the crosswinds are accommodated without significant errors in predicted/achieved aircraft location in the cross-track direction. Tailwinds and tailwind errors directly affect the ground speed and therefore the predicted aircraft location along the great circle path. Therefore, in this case, tailwind errors must be allowed for, either as a separately fitted leg-average tailwind, or by increasing the allowable BTO errors in the fit. The two methods are essentially identical in result. I fit the tailwind errors so that I can see them separately from the BTO errors. I found it is very convenient to express all errors in units of standard deviation (sigma) for that particular data point. So, I just take the ratio of the error to its expected standard deviation. That way it does not matter how many data points have particular values of standard deviation. It is self-normalizing for all data sets.
You also said: “In other words, there isn’t any X to mark the spot unless our starting point on the 1941 arc is quite tightly constrained in one way or another. Personally I was surprised that there is such a wide spectrum of ISAT-compatible routes and I’m a bit staggered that the BTO residual RMS is as low as this over such a wide range of paths.”
Actually, this is only true for “straight” paths such as waypoint navigation to Antarctica. It is not true for “curved” paths such as CTH, CMT, and especially CMH. I have yet to find any CMH paths matching the satellite data, and I don’t think they exist. I did find one CTH route and one possible CMT route. For those solutions the range of acceptable bearings is small (< 1 degree).
Apparently, the non-concentricity of the BTO arcs has less of a deleterious effect on straight paths than it does on curved paths. It is fortunate the satellite moved during the southern route. Otherwise no autopiloted route would have a single solution. They all would be redundant in bearing (i.e., they would be rotationally symmetric about the sub-satellite point). For LNAV routes from 19:41 to MEFE, the only way to obtain an “X” spot location is to make an assumption about the target waypoint.
@Richard Cole,
Yes, your assumption was right. I meant to say Figure 5.5 instead of 5.4 in that instance.
@Richard Cole,
You said: “The discussion around page 30 of the DSTG book is about the attempt to separate and analyse the bias drift from the historic flight BFO data and a Kalman filter is used in that process. In the end, that analysis was unsuccessful and unseparated BFO data, with the larger 7Hz overall error, had to be used for subsequent BFO analysis.”
I think you are saying the BFO bias Kalman filter was not used in creating the probability distributions shown along the 7th Arc for MH370. If that is what you meant, then why does “Section 8.2 Algorithm” on page 71/128 list step 3 as “Initialize the BFO bias Kalman filter” and step 4c as “Use the sampled trajectory to update the BFO bias Kalman filter”? It appears to me that the Kalman filter was used in the algorithm that created the probability density functions near the 7th Arc.
If that is not what you meant, then please define what you meant by “subsequent BFO analysis”. Thanks.
@Paul S
I would say Victor’s article above shows what I am saying, that the SPOLE direction is where the BFO/BTO data best point to. Victor assumed no pilot input, so the BFO fit is not as obvious. I believe there is an “exact” solution in there, if we open up the (eg; Bayesian) model to very low BFO error and allow pilot input to South Pole (or 180S variant). Instead we told Bayesian model to assume BFO error was relatively large, and the pilot was not taking any control. I am thinking 180S is the only place the fits the data closely. Yes the data could be wrong, I get it. But I don’t like the idea of not even showing the solution.
@DrB (others)
The DSTG fumbled the BFO interpretation badly. I called them on it, and they (Holland) pretty much admitted that issue, but stood by their approach.
BFO error has two components – a component due to measurement and hardware implementations such as the NCO used to generate the Doppler compensation in the AES. A second source of error has to do with oscillator drift. The first component is a candidate for a “normal” distribution with the usual application of mean and variance. The second component is not. It truly is analogous (not identical) to a random walk. You do not model a random walk with a Kalman filter. That would be ridiculous – future states have no relationship to past states.
The oscillator drift (like a random walk) will have a zero ensemble mean and a variance that grows with time. Any single sample of a random walk would not be expected to exhibit a zero mean (hence, non-ergodic). The fact that the variance grows with time means that the sample statistics are also not stationary. The oscillator drift is not a candidate for Gaussian statistics which is why Dave Allan (and others at NIST) developed the Allan Variance approach which is now an industry standard. You never see oscillator specifications in terms of a standard interpretation of mean and variance.
What the DSTG has done is to compile ensemble statistics using the previous flights. These statistics are essentially meaningless for predicting what might happen on any given flight.
DennisW: Within the context of one flight (or a time period of 1 day), I agree with your BFO statement above. (Shocking, isn’t it?) Of course, there is also a long term aging term which is not a random walk. It is typically more like (but not exactly) a constant monotonic drift towards higher frequency. But this LT drift is negligible over a day. Here’s a short tutorial https://goo.gl/iSv3xK
@TBill – S Pole / 180S path may well be close to minimum RMS on BTO and BFO (basically 33S-35S). That would make it the peak of a very wide distribution in which there is very little difference in probability across a very wide range of latitude end points. Thus we cannot say with much confidence that this “peak” is much better than neighbouring latitudes across quite a large swathe. And, as previously stated, we seem to have “acceptable” BTO/BFO residuals all the way from ~25S to ~38S. @VI – please correct me if I am interpreting your article above wrongly.
@ALSM
Hugo is a great guy. He was president of Efratom (Rubidium supplier) when the company I was working for at the time, Datum, bought Efratom. I got to know both him and his wife quite well. Hugo was very much into physics, and pretty much coined the opinion that “microprocessors would never replace physics”. He liked to annoy me with that position since I was more of a math/computer guy. Like my concert violinist neighbor at the beach house enjoys referring to me as a “propeller head”.
Nice to see you are back from hibernation @Dennis.
Now, if I may poke you a little. We have heard (again) your refrain on BFO error, the mistaken approach of DSTG et cetera. Its clear from the explanation of experts that this oscillator drift aspect should be anticipated, that we don’t know which direction it was moving, that it has not been properly characterised for 9M-MRO’s SDU, and that the DSTG’s characterisation of BFO errors wrongly lumped this in with other errors with gaussian or quasi-normal distribution.
Fine. But having repeated this over and over again, do you have a notion of how large this oscillator drift might be? Do we effectively throw BFO out altogether as a discriminatory data sources and say the error could easily be 20+ Hz?
Dr B above has made clear his own understanding of the error characteristics and his reasons for inferring a magnitude/rate of oscillator drift error. Would you, as an oscillator “fundi” [Kiswahili], care to hazard your own estimate of drift magnitude over the period of a single flight?
@Paul
Yes, I know I have been a bit of a broken record on BFO issues. Certainly BFO tells us the plane went South and came down rapidly at the end. Beyond that it is dicey to use BFO other than as a broad path qualifier.
I do agree with DrB that figure 5.4 of the DSTG book probably represents the worst case of all the logged 9M-MRO flights. That is why it became Figure 5.4. I do claim, however, that other flights exhibit similar behavior, simply not as extreme. What the behavior was on the “accident” flight is anyone’s guess.
@Victor. Could I trouble you to post path characteristics and BTO errors for each of the ping rings for one example where you have BTO residual RMS of low 20s? I’m still having trouble understanding how you can find path solutions with so little error unless you have “absorbed” some of the prima facie residuals by assigning a chunk of them to wind error or navigation track error.
@Path Modelers,
I have briefly revamped my old CMT (constant magnetic track) model, and here are some preliminary results (so far I looked at SPD MACH only):
https://www.dropbox.com/s/3ib90mxp1l7ffya/stats_cmh.pdf?dl=0
I used my fuel flow model, Rev.2 (discussed earlier this year at Victor’s blog, it seems in April or so). In brief – it appears fuel is sufficient at FL above 9 km altitude (approximately). Found Mach are in the range 0.70-0.73, which explain why the DSTG did not have these paths in their study. Headings and positions at 19:41 are generally consistent with a wrong exit from a triangular pattern. Interestingly, the terminuses are between 27S and 28.5S – ‘hotspot’ of my drift study.
@Victor, @sk999, I recall you did some CMH/CMT fitting attempts; could you pls. remind your conclusions?
@Victor. For your “best BTO path fit”across all arcs 2-6, can you please tell us what was the BTO error at 1941 for a) LRC b) constant M?
@Don,
Re: “You tried that one previously.”
Yes.
Re: “I will add “if the airplane moves during the alignment, the align reset mode occurs until there is no movement.“, how can that be construed as starting alignment?”
How alignment cannot be completed if it is never started? Perhaps you differentiate between the align mode and alignment?
Re: “Further, consider that normal ADIRU shutdown is an orderly process, not the same as removing all power (which requires six cct breakers, located on three power panels in the MEC, to be opened).”
Let’s do not mix up things. How the ADIRU could be shut down is a different issue. Right now my question is whether LGA could be engaged at 18:25:34, and what is the role of WoW bit in SATCOM system.
@Dennis,
Re: “I do agree with DrB that figure 5.4 of the DSTG book probably represents the worst case of all the logged 9M-MRO flights. That is why it became Figure 5.4. I do claim, however, that other flights exhibit similar behavior, simply not as extreme. ”
I guess that depends on what the DSTG implied under the “geographic dependence”. If flights on the same/similar routes using the same aircraft exhibit similar ‘structured’ components of the BFO error, then it has nothing to do with the oscillator, but with the navigation system. Otherwise, why would the DSTG use the term ‘geographic’, but not, say, ‘time’ dependence?
@Oleksandr
1. Your sheet heading is CMH. Which is it, CMH or CMT?
2. Have you discriminated wind models at the various altitudes posited? If not, which pressure altitude winds are you using?
3, Likewise ISAT delta. Are you using the (forecast) delta at one pressure altitude and then interpolating temperature at another with lapse rate?
4. Which magnetic variation table epoch are you using?
5. (Most importantly) I’m also struggling to understand how any of the magnetic hdg/trk options can be such a good fit. Can you please provide one path example (speed/altitude, magnetic heading/track) that produces such small BTO errors?
Oleksandr,
I have never looked serious at CMT routes that are initialized on the 2nd arc. In the past I have always assumed (as did the DSTG) that the plane flew straight line segments, and in particular excluded the possibility of a hold. Your starting latitudes of 5+ degrees at 19:41 are considerably North of the straight-line routes, which generally reached the vicinity of ISBIX (0.4 deg N latitude) at that time. My final latitudes tended to be in the range -32 to -34 deg. The speeds were in your range however. The BTO fits were never very good (60 microsec rms) although the BFO fits were often improved over routes that went further South.
Note that even if the DSTG had included speeds down to Mach 0.70, it would not have found your routes, on account of the straight-line constraint going back to the penultimate radar point.
@Oleksandr
I guess that depends on what the DSTG implied under the “geographic dependence”.
I don’t know what the DSTG implied by “geographic dependence” or how they arrived at that possibility. They do say that other flights exhibited “structured variation” in the mean bias.
@Oleksandr
OK thank you. I give you extra credit for breaking 23:14 sat call into two points. But what are the base BFO’s you are assuming to calc the error?
Interesting how the sat call happens right on the L894 flight path by POLUM. In one of my paths that I might revisit, I use McNurdo path and then exit on L894.
Finally, some word out of Malaysia about the August 2017 Madagascar debris
https://goo.gl/AZ7AvH
Any apparent “geographic dependence” is purely coincidental. It did not happen.
@ALSM
I am curious as to why figure 5.4 was even included in the book. It does not add anything to the conclusion or other analytics.
@ASLM: The ‘crime’, I presume, that prevents handing over to the Malaysian, refers to the murder of Zahid Raza, and not the downing of MH370. Hard to see what evidentiary value the pieces would be to their investigation. But also hard to see how much value they’d have for MH370. But an echo of the flaperon and the tendency for Napolenonic-trained investigators to hold onto their trophies and just about all the information they prise from them.
lkr: The vortex generator debris is potentially very important to the new search. The Madagascar mafia may have a financial stake in stopping the debris from leaving Madagascar. Watch this space.
Dennis: Fig 5.4 is important not because it helps to quantify the BFO noise, but becauze it helps to quantify the limits to using BFO data to ID the POI.
ALSM. The Malaysians have the photos, the inference being those are insufficient.
@Oleksandr,
You asked, ““Right now my question is whether LGA could be engaged at 18:25:34””
The LGA could be engaged at any time. However, the expanded GES Log (published at this site) shows that, in its Log On Request at 18:25:27, the AES defined itself as HGA equipped and the Log On proceeded accordingly.
I draw your attention to the ‘SU Contents’, associated with that Log On Request:
18:25:27.421 1F D0 10 75 00 8F C5 D0 FC 05 82 09 00 00 00 00 00 B4 06
Now, consider octet #11 where the content is ‘0x82’. The most significant two bits of this octet are defined as a data field to communicate the ‘Class of AES’. A Class 1 AES (LGA, packet-mode only) is indicated by ’00b’, A Class 3 AES (HGA, packet & circuit-mode operation) is indicated by ’10b’.
Further, consider octet #12 where the content is ‘0x09’. This octet is defined as communicating the Data Bit Rate Capability for the Log On. The lsb (bit 1/8) implies P-ch capability for 1,200bps and bit4/8 implies P-ch capability for 10,500bps.
The 18:25:27, and 00:19:29, Log On Requests both indicate the AES requesting capabilities for Class 3 AES operation, via HGA, and to be allocated a 10,500bps P-channel. These two Log Ons were completed as requested.
I’ll maybe return WoW later.
@ALSM
“Madagascar mafia”
Does the use of the word ‘mafia’ merely imply gangsterism, i.e. the use of intimidation, violence, in pursuit of criminal ends. Or is a broader, widely federated, coalition of criminal enterprises inferred?
@ALSM
It’s disappointing the vortex-generator piece won’t be available for detailed examination by Malaysia any time soon probably.
It could really hold some important clues about impact forces/speed and attitude of the plane on impact imo.
On the other hand it could be a possitive the Madagaskar authorities are going to investigate the piece in detail first and to see if there is any faul play linked with the MH370 investigation.
They probably have some good reason to suspect something like this otherwise they would have had no problem with handing the piece over to Malaysia.
France also still refuses to hand the flaperon over ( and it’s information). There might be good reasons for acting like this.
Not necessarily ‘maffioso-practise’.
Don: I think it is the latter. We will eventially know more about the connection between the debris and the murder.
@VictorI
Something I wonder about.
All your flight paths in your graphic above (your topic) seem to start from the second arc around ~7N/93E. I know you consider a loiter, descent and/or some kind of holding pattern somewhere between 18:28 an 19:41 (if I understood you well enough).
I wonder, did you also consider the plane flew past the second arc position and turned around later like @TBill suggested also?
Maybe till ~8N/92E or even further north/west? Then the turned flight producing the second arc on it’s way back and made another smaller turn to the right into the SIO?
In my imagination no loiter or other complicating manouvres would be needed in such a case.
I probably miss something essential here but I cann’t see what and why yet.
@Ge Rijn
One element I have not looked closely at is fuel consumption. The reason I do not look too closely at fuel (yet), is because I can think of ways to save fuel, having to do with an active pilot. The air packs can be turned off, I can easily envision a slow down/descent or even a hold after Arc5, tentatively looks to me like a CTH or CMH heading, that works with the winds, may save fuel over a CTT or CMT or waypoint setting that fights the winds (specifically in the MH370 path scenario where we have winds from East and then winds from West). So I am not too worried (yet) about fuel. I am instead worried about close match to BTO/BFO. At some point there is a day of accounting for fuel, to help decide on real life paths.
Victor I think is assuming inactive pilot after Arc5, so he needs to take his fuel savings via a hold before Arc2, to over-simplify.
@sk999
@Paul Smithson
@VictorI
@TBill
@Oleksandr
I have been trying in vain to find a candidate constant Mach great circle path into the SIO, one that used known waypoints, and I’ve come to the following conclusion; I don’t think the aircraft used any of the known waypoints to navigate from N571 into the SIO. I think the pilot flew basically same way that he flew his simulator on 2nd March 2014. For MH370 I believe he chose in advance, a specific set of coordinates that suited the purpose (Not an airfield in Antartica, BTW) that was beyond the fuel range of the aircraft, and used these to set up a manual waypoint he could designate the FMC to fly toward just prior to reaching a pre-determined waypoint on N571. On the MH370 flight, this waypoint was IGOGU. The purpose was to make absolutely sure his flight-path could never be reconstructed.
DrB’s range v. endurance table has proven particularly useful. In my opinion, a constant Mach 0.82 at FL380 has probably the most promising combination of range and endurance if one wants to avoid loiter, fly as far as possible into the remotest part of the SIO and arrive at the terminus in daylight.
I find that if the pilot timed the FMT to begin 20NM before reaching IGOGU, and then flew toward a manual waypoint located beyond S37.62, E89.08 – at this point, temporarily suspend your disbelief – it would comply hopefully with the BTO, comply with the total range from DrB’s start point at N7.18, E95.68 (adjusted accordingly for the no offset case) and comply with the endurance.
FMT of 12NM radius beginning 20.3NM before IGOGU is initiated at 18:36:30, completed 18:39;00, at N07.25, E94.528, bearing 186.1364.
Range from the 18:29 point onwards works out as follows : estimated 70NM from 18:29 point to the FMT, plus 2,655.5NM up to the 6th arc crossing, plus a further 47.5NM to MEFE = 2,773NM (fuel model 2,774). Fuel model MEFE time is 00:16.
The great circle distance between FMT and the 6th arc is 2,655.5Nm, and the journey time is 5.5542hrs (5hrs, 33.25mins) taking the FMT mid-point time as 18:37:45. When an average 2.2Kts headwind is factored in, the airspeed works out at M0.8184: 478.1Kts +2.2Kts headwind =480.31/587* = M0.8184. (taking 587Kts as average value of Mach 1 between FMT and the 6th arc, at average temp 42.5C). As you can see, the calculated airspeed is slightly less than an exact M0.82. It is possible the aircraft cruised south at slightly above FL380. After the 5th arc, altitude could have slowly risen to FL400. My estimated headwind for the leg could be a slight overestimate.
FMT to 2nd arc: 512Nm, 485Kts average groundspeed, 1.8Kts average tailwind = Mach 0.822 (Mach 1=588.4Kts, Temp -41C) 2nd arc tangent point S1.00, E93.65.
2nd arc to 3rd arc: 479.51Nm, 479.51Kts average groundspeed, 1.6Kts average tailwind = Mach 0.812.(Mach 1=588.4Kts, Temp -41C) 3rd arc crossing S9.00, E92.799
3rd arc to 4th arc: 489.43Nm, 486.42Kts average groundspeed, 4.6Kts average tailwind = Mach 0.819.(Mach 1=588.4Kts, Temp -41C) 4th arc crossing S17.12, E91.89.
4th arc to 5th arc: 480.82Nm, 480.41Kts average groundspeed, Zero average head/tailwind = Mach 0.816.(Mach 1=588.4Kts, Temp -42C) 5th arc crossing S25.113, E90.916.
5th arc to 6th arc: 693.68Nm, 462.3Kts average groundspeed, 13.5Kts average headwind = Mach 0.816.(Mach 1=583.0Kts ave. Temp -46C ave.) 6th arc crossing point S36.60, E89.25.
6th arc to 7th arc: 61.5Nm, 434Kts average groundspeed, 33Kts average headwind = Mach 0.81.(Mach1=577.46Kts, Temp -50C)
@VictorI
Re my post 2 hours ago. Please could you tell me why it’s taking you so long to moderate?
@Rob: I was on a plane. Patience, please.
@VictorI
Phew, that’s a relief!
@Rob
Well any point on the map can be a manual waypoint. Your path projects closely to whole number waypoint 3889S…but with a little more complexity the pilot can also add minutes and seconds, to fine tune it. I find I use the whole number waypoints quite a lot when working with MH370 flight paths.
@Paul Smithson:
1. Your sheet heading is CMH. Which is it, CMH or CMT?
CMH – constant magnetic heading.
2. Have you discriminated wind models at the various altitudes posited? If not, which pressure altitude winds are you using?
Yes. I am using 3D GDAS. In the model I assumed that a constant pressure altitude corresponds to constant altitude (the difference varied within approximately 50 m along the path).
3, Likewise ISAT delta. Are you using the (forecast) delta at one pressure altitude and then interpolating temperature at another with lapse rate?
No. My script directly interpolates GDAS data at a location, altitude and time.
4. Which magnetic variation table epoch are you using?
A good question. According to my records, I generated 2D magnetic declination from one of the following links:
http://www.ngdc.noaa.gov/geomag-web/#igrfgrid
http://www.ngdc.noaa.gov/geomag/WMM/calculators.shtml
But it was in the mid of July 2014, before we learnt about magvar tables. I believe I generated magnetic tables for the actual date of the disappearance (2014-03-08).
5. (Most importantly) I’m also struggling to understand how any of the magnetic hdg/trk options can be such a good fit. Can you please provide one path example (speed/altitude, magnetic heading/track) that produces such small BTO errors?
Perhaps it flew in that way?
3.5 years ago I did some simulations of CMH (it was posted at Duncan’s site), and the fit was rather good. Then someone claimed that magnetic hdg/trk is used for display only, and I abandoned CMH/CMT class of simulations. It seems nobody seriously modeled magnetic routes: I am aware only of @sk999 work, and, if I am not mistaken, one Victor’s route. The DSTG imposed a-priory limit of 0.73M on the speed, while all the CMH are slightly slower.
Here is a link to the plot of the flight path at 9,000 m:
https://www.dropbox.com/s/g5usvhgwjxn0p3d/mh370_cmh_alt9000_rev2017_11_09.jpg?dl=0
I could not resist to include my earlier estimation of the intersection of the 7th arc with “Curtin boom” bearing (HA01+RCS).
@Rob
I like the idea of a manually entered final waypoint “at the fuel limit or beyond fuel range” lat/lon to head for, since the sim point 45s104E suggests that. I also like the idea of 45S as his preferred goal lat to reach, for reasons previously stated.
So, as a suggestion, why not run a whole series of paths from your FMT to 45S, for each longitude from 104E back to say 84E, ie, 104, 103, 102, 101, 100, 99 etc, and see where you get the best BTO/BFO fits ?
@TBill,
Re: “But what are the base BFO’s you are assuming to calc the error?”
19:41:02.906 111Hz 2nd handshake
20:41:04.904 141Hz 3rd handshake
21:41:26.905 168Hz 4th handshake
22:41:21.906 204Hz 5th handshake
23:13:58.407 216Hz 2nd satcom call start
23:15:02.032 219Hz 2nd satcom call end
00:10:59.928 252Hz 6th handshake
The difference is defined as a sampled BFO minus modeled BFO.
@sk999,
Thanks for your comment.
RE: “In the past I have always assumed (as did the DSTG) that the plane flew straight line segments, and in particular excluded the possibility of a hold. Your starting latitudes of 5+ degrees at 19:41 are considerably North of the straight-line routes, which generally reached the vicinity of ISBIX (0.4 deg N latitude) at that time.”
Yes. If you assume the final phase starting at 18:40 or earlier, then 5N+ is not possible. In this regard triangular patterns are the most intriguing: I already see that it will likely be possible to connect at least one of them to the post-19:41 CMH path. However, I am somewhat reluctant to do this bearing in mind recent Andrew’s comments. I don’t know about racetrack pattern.
Re: “Note that even if the DSTG had included speeds down to Mach 0.70, it would not have found your routes, on account of the straight-line constraint going back to the penultimate radar point.”
That is true.
@Oleksandr
Re: “…what is the role of WoW bit in SATCOM system.”
The only information I have come across relates to PAST (Person-Activated Self-Test). PAST is inhibited when the aircraft is airborne, except when initiated via the CMT (Commissioning & Maintenance Terminal). WOW is one of the criteria used by the SDU to determine the airborne/on-ground status.
Re: ‘The patent link to which I provided, appears to be somewhat inconsistent with AMM, chapter “AIR/GROUND SYSTEM – INTERFACES” For example, WOW Card Inputs. Citation from AMM, 32-09-00 (pdf p. 4398):
“The WOW cards get inputs from:
– AIMS (time and date, flight phase/leg, airplane registration number, ICAO code)
– Bus power control unit (external power available)
– ELMS (air/ground relay status)
– Ground test enable switch
– Load sensors
– Radio altimeter (3).”
Nothing is mentioned about air/ground speed, mentioned in the patent.’
I don’t believe the patent information is inconsistent with the AMM. The patent relates to the air/ground sensing for the flight control system, which uses a variety of data inputs to determine the air/ground status, including WOW, TILT-UNTILT, airspeed, groundspeed, Mach number, altitude and pitch. Some aircraft systems only use the WOW data to determine the air/ground status, while others use a combination of data such as WOW, airspeed, altitude, etc. The reference you cited only describes the interfaces to the WOW cards, which do not require airspeed, groundspeed, etc to determine the WOW status.
@Don,
Re: “I draw your attention to the ‘SU Contents’, associated with that Log On Request:
18:25:27.421 1F D0 10 75 00 8F C5 D0 FC 05 82 09 00 00 00 00 00 B4 06
”
Thanks for your detailed explanation. Now I admit it looks like the ADIRU failure can really be ruled out because the Inmarsat data indicate transmission via HGA, and not via LGA.
Btw, do you have a link to complete Inmarsat logs? Somehow I missed it: in the “Update 2014” SU content was still cut out as not important. Do all other records indicate the use of HGA after 18:25? It could also be interesting to take a look at RX-power fields.
Re: “I’ll maybe return WoW later.”
I am looking forward for your comment with regard to WoW.
@Andrew,
Thanks for your comment.
Re: “WOW is one of the criteria used by the SDU to determine the airborne/on-ground status. ”
Why is it so essential for the SDU to know whether the plane is on the ground or in the air? Why is the ADIRU insufficient?
Re: “I don’t believe the patent information is inconsistent with the AMM.”
I don’t know. The patent states a lot of input information is used to define the status. AMM does not say anything how TILT-UNTILT, airspeed, groundspeed, Mach number, altitude and pitch information reach the WoW cards; only 4 WoW sensors are mentioned. Am I missing something?
@Dennis,
Re: “I don’t know what the DSTG implied by “geographic dependence” or how they arrived at that possibility. They do say that other flights exhibited “structured variation” in the mean bias.”
I also don’t know. Why “geographic dependence”, why not “time dependence”? The latter seem to be more logical, unless they did really observe some kind of geographic dependence of the structured deviation.
@Ge Rijn. “It’s disappointing the vortex-generator piece won’t be available for detailed examination by Malaysia any time soon probably”.
As reported, the equanimity of Minister Liow’s acceptance of the indefinite delay, already coming up 3 months, hardly accords with the Malaysian task of identifying the cause of the crash. We could have expected at least a work-around initiative by now or signs of one. Purportedly a search renewal is being investigated though with no sign of the urgency needed to meet the forthcoming search season.
Why could the Malaysians not ask to examine the item under Malagasy supervision? Interview the finders? As a minimum ask for more photographs if needs be and use those as the ATSB did those of the flaperon obtained from the French. Then they could at least come to a tentative finding on whether the main item is of 777 vortex generator origin based on the photographs, the Blaine Gibson description and Boeing’s opinion.
What to me increases interest in the origin of that item is that it has features needing explanation. That might also be feasible now even if its fracture mechanics need to wait.
It is in the interest of the tripartite countries and search refinement that they glean what they can. The appearance however is that the Malaysians, at least, are sanguine about the delay. That joins the tenor of Liow’s remark quoted on the 1st November, “If we are convinced, with credible evidence, only then will we continue the search mission..”
The only legitimate reason I see for downplaying the need for a report on this item, and for a new search, would be that finding the wreckage is now less important than it was, implying that there is a likelihood of finding the crash cause from other sources.
To the French view that the flaperon is evidence of malfeasance now add Madagascar’s reason for withholding these two items. Add Blaine Gibson’s withdrawal ostensibly under threat and I am left wondering why the media are not into all this.
@Oleksandr
“23:13:58.407 216Hz 2nd satcom call start
23:15:02.032 219Hz 2nd satcom call end”
OK I agree with those 23:14 BFO values, some are adding offsets to those. Apparently we need to think of those as having higher error bar.
Like Paul I am surprised at how good the fit is. I will try to see if I can match it in FS9 when I get time.
@Oleksandr
Re: “Why is it so essential for the SDU to know whether the plane is on the ground or in the air? Why is the ADIRU insufficient?”
As I said previously, PAST is inhibited when the aircraft is airborne, except when initiated from the CMT. As I understand it, one of the following criteria must be met to satisfy the ‘on-ground’ status:
– Both WOW discretes indicate ‘on-ground’; or
– IRS data indicates a ground speed of less than 40 knots; or
– IRS data is invalid.
Re: “The patent states a lot of input information is used to define the status. AMM does not say anything how TILT-UNTILT, airspeed, groundspeed, Mach number, altitude and pitch information reach the WoW cards; only 4 WoW sensors are mentioned. Am I missing something?”
Yes, you are missing something. TILT-UNTILT, airspeed, groundspeed, Mach number, altitude and pitch data is not used by the WOW cards. It is used by the flight control system in conjunction with the WOW data to determine the air/ground status for flight control system purposes only.
@Andrew,
Re: “TILT-UNTILT, airspeed, groundspeed, Mach number, altitude and pitch data is not used by the WOW cards. It is used by the flight control system in conjunction with the WOW data to determine the air/ground status for flight control system purposes only”.
Thanks for clarifying this.
Re: “As I said previously, PAST is inhibited when the aircraft is airborne, except when initiated from the CMT. As I understand it, one of the following criteria must be met to satisfy the ‘on-ground’ status:
– Both WOW discretes indicate ‘on-ground’; or
– IRS data indicates a ground speed of less than 40 knots; or
– IRS data is invalid.
”
Why would WoW be required to activate self-test or by PAST?
@Oleksandr
“3.5 years ago I did some simulations of CMH (it was posted at Duncan’s site), and the fit was rather good. Then someone claimed that magnetic hdg/trk is used for display only, and I abandoned CMH/CMT class of simulations. It seems nobody seriously modeled magnetic routes…”
That is a key point, I am thinking it was maybe only 6 months ago @Matt over on JW started to change “our” collective thinking that Magnetic Heading (not True Heading) is the default A/P mode (after waypoint discontinuity).
It is still quite controversial to me that the ATSB 38S hot spot was basically a straight (True) path pretty much based on assuming a hypoxic-style (unintentional ghost) flight. That is a illogical as it would be a Magnetic Heading in that case.
Anyways that means CMH paths are more important now, and at least @Rob still embraces an intentional True straight path flight to 38S.
@ventus45: In a previous comment, I mentioned that I found the great circle path that satisfies the satellite data that also passes the point 45S, 104E, which are the final simulator coordinates. The position at 19:41 is 8.0N, 93.1E, and the path crosses the 7th arc near 28.3S latitude. For an LRC speed profile, the RMS BTO error is 33 μs and the RMS BFO error is 6.6 Hz. I believe this is a candidate hot spot.
Can anyone explain jitter in the course reported by FlightAware FlightAware ADS-B for Frankfurt and Mumbai flights of 9M-MRO, typically reaching 3 deg by the end of flights (beginnings appear to be much more stable)?
https://www.dropbox.com/s/18rrb60ecygnc4k/tracklog-mas006-frankfurt-20140307.pdf?dl=0
https://www.dropbox.com/s/k913zw0m07cik17/tracklog-mas195-mumbai-20140302.pdf?dl=0
@Ge Rijn said: All your flight paths in your graphic above (your topic) seem to start from the second arc around ~7N/93E.
It might look like that in the figure, but in fact the position at 19:41 rotates around the 2nd arc for various paths as the position at 00:19 also rotates around the 7th arc.
Since several people have asked for it, I’ll add a table to the post that summarizes the great circle results, including position at 19:41, position at 00:19, and BTO/BFO errors.
@TBill: No pilot input after 19:41 does not require non-response due to hypoxia from an accident. There could have been deliberate actions to guide the plane to the SIO that required no pilot input.
@TBill,
Re: “I am thinking it was maybe only 6 months ago @Matt over on JW started to change “our” collective thinking that Magnetic Heading…”
If I understand it correctly, Andrew says the same.
As you know I was banned at JW site when I opposed JW’s idea that NOK and Blaine planted debris, so I did not read his blog since then.
Re: “It is still quite controversial to me that the ATSB 38S hot spot was basically a straight (True) path pretty much based on assuming a hypoxic-style (unintentional ghost) flight. That is a illogical as it would be a Magnetic Heading in that case. ”
It is a mystery to me why the ATSB and DSTG avoid discussions of magnetic routes, especially keeping in mind all the uncertainties in the BTO and BFO data.
Re: “Anyways that means CMH paths are more important now, and at least @Rob still embraces an intentional True straight path flight to 38S.”
It is not possible to reconcile 38S with the drift studies, including measured data. I don’t know why @Rob cannot understand this.
@Olekksandr
It is not possible to reconcile 38S with the drift studies, including measured data. I don’t know why @Rob cannot understand this.
@Rob is a whacko.
@DennisW
“It is not possible to reconcile 38S with the drift studies, including measured data. I don’t know why @Rob cannot understand this.”
Really ?
Mike Chillit’s latest chart and tabulated data say otherwise.
https://pbs.twimg.com/media/DOO0o-BU8AAd0pF.jpg
https://pbs.twimg.com/media/DOO0pAoVoAAAzk8.jpg
The bottom 6 red dots for Arc Latitudes 34South to 39South inclusive are very interesting, MOST PARTICULARLY for Latitudes 36, 37, 38 and 39 South, which account 55 buoys, (out of 141 for the WHOLE arc) and of them, ONLY 31 went west, and a significant number, 24 buoys went EAST.
Near enough to a 60 / 40 split.
Where did those 24 buoys go ?
In fact for the bottom 3 red dots for Lats 37, 38, and 39 south, for 44 buoys, 26 west, 18 east, it is a 60 / 40 split.
Moreover, those east going buoys are too far south to make landfall anywhere on Australia, except perhaps Tasmania, so obviously they didn’t, and similarly for any wreckage debris.
When 40 percent of those buoys could go east and never beach anywhere, it is false logic to say they didn’t because we never found any. If it was 95% to 5% I might swallow it, but at 60 / 40 – no way. Absence of evidence is not evidence of absence.
The drift studies have all (in my opinion) not been extended far enough south to give a true picture. Even Mike stopped at 39 South. I think the drift studies should be extended both sout (to 45S) and west (to 80E) at a minimum.
@Ventus
The lack of debris found in WA speaks volumes relative to terminal location at 38S and beyond. Why are you beating on a dead horse?
@Dennis: Yes, you’re right, a lack of debris in WA, where there was a search. Can’t say the same for ANY area not visited by BG on the western shores!
@Ventus45 @Mike Chillit
I wonder on what data @Mike Chillit’s numbers are based. Global Drifter?
I referre again to @MPat’s study based on Global Drifter data spanning 20 years of drifters that past the previous priority area between 32S and and 39S.
177 past this area in those 20 years. 39 of those are listed to have beached. 31 on East African shores, 7 on WA shores and one on Sumatra.
Those 7 that landed on WA shores all passed the area south of 36S.
North of 36S none of those 177 drifters reached WA in those 20 years.
This reflects the situation we have now; no debris finds in WA and around 30 finds in Africa and it’s islands.
And this study indicates north of 36S debris is not expected to wash ashore in WA (which I deduced from @MPat’s study back then mentioned in the follow-up article).
Have a read again please:
http://jeffwise.net/2016/07/07/guest-post-where-mh370-search-area-debris-has-historically-gone/
http://jeffwise.net/2016/07/15/how-we-know-where-mh370-went/
(with thanks to Jeff Wise for borrowing his articles again)
@TBill
On your comment;
‘One element I have not looked closely at is fuel consumption. The reason I do not look too closely at fuel (yet), is because I can think of ways to save fuel, having to do with an active pilot.’
I don’t quite understand why you would want to look into saving fuel if the FMT shifts north-west past the second arc toward 10N/90.
Wouldn’t the range at a specific altitude (conform @DrB’ tables) stay the same? Would not only shift the final reachable latitude to the north compared to a ~18:40 FMT?
@TBill
to add another question:
Wouldn’t an active pilot make shure he could reach a specific latitude (if this was intended) without complicating things by saving fuel by complicated means?
He could pick his FMT/route at will with the available fuel to make shure he could reach his latitude/destination without complicating matters I think.
@DennisW
“It is not possible to reconcile 38S with the drift studies, including measured data. I don’t know why @Rob cannot understand this. @Rob is a whacko.”
@DennisW
“It is not possible to reconcile 38S with the drift studies, including measured data. I don’t know why @Rob cannot understand this. @Rob is a whacko.”
DennisW. I am sorry you see it this way. I am not offended, btw. I will wear my whacko badge with pride. As I told you before, many times, the drift studies simply cannot be considered reliable. Be honest, you are not rejecting S38 because of the drift studies. You rejected S38
@Rob
It’s not only the forward drift studies that are decicive.
As @DrB has calculated ~38S is only reachable with the available fuel from the earliest possible FMT at ~18:25.
I think no one thinks the FMT was at ~18:25. Do you?
@DennisW
Continued: You rejected S38 a long time ago because it had no place in your “thwarted negotiation” theory. The pilot of MH370 was no expert on the ocean currents of the Indian Ocean. He simply figured that the further south he flew, the less likely any debris would come ashore. That’s why he flew south at optimum altitude between 38,000 and 40,000ft, to get as as far as possible on the fuel, then when CV the fuel was exhausted, glide as far as possible on the same heading. The final BFO’s don’t rule out a glide from 40,000ft, imo.
@GeRijn
“It’s not only the forward drift studies that are decicive.
As @DrB has calculated ~38S is only reachable with the available fuel from the earliest possible FMT at ~18:25.
I think no one thinks the FMT was at ~18:25. Do you?”
Ge Rijn, please check your figures. DrB’s assumed earliest FMT was taken to be 18:29, NOT 18:25. And his FMT was also deemed to have begun on an offset flight path. The offset at around 18:25 is still very much a moot point. For myself, I can see no evidence to support it. What I am seeing is a slowdown from M0.84 to M0.80, possibly accompanied by a descent, between 18:22 and 18;29, the aircraft remaining on N571. After 18:29, I think the aircraft began to climb to FL380 in readiness for the flight south.
@TBill
“Well any point on the map can be a manual waypoint. Your path projects closely to whole number waypoint 3889S…but with a little more complexity the pilot can also add minutes and seconds, to fine tune it. I find I use the whole number waypoints quite a lot when working with MH370 flight paths.””
TBill, thank you for keeping an open mind on the subject. I gives me encouragement. Yes, I agree than any point on the map can be a manual waypoint. What our particular pilot was looking for was a waypoint that ensured his plane flew in darkness until the last half hour or so. The recovered debris is telling me that the pilot was in control until the final moments. He was preoccupied with the need to minimise the amount of debris while at the same time, sinking the plane as efficiently and quickly as possible.
I think he miscalculated slightly iro synchronising with sunrise. Not the easiest thing to do in advance, when you also want to keep to a straight flight path to maximise distance covered into the SIO, and you are not sure exactly at what point the fuel is going to run out. I think he would have preferred the Sun no more than 2 degrees above the horizon. What he ended up with was approx 6 degrees.
@Rob
Yes, 18:29.. But it’s about the earliest possible FMT and the maximum range I was talking about. Conclusion is the same.
Update on November 10, 2017:
Here is a CSV file with data for the great circle paths, including the position and track at 19:41, position at 00:19, and speed mode. Included are the data for four paths that align with waypoints past the 7th arc. The four waypoints are the South Pole (NZSP), Wilkins Runway (YWKS), Pegasus Field-McMurdo (NZPG), and the fuel exhaustion position from the simulator data (45S, 104E).
@Victor. Many thanks – will have a look at these with interest.
@Oleksandr. Thanks for posting the depiction of your magnetic route. For a path that finishes around 25S I don’t have too much of a problem, particularly if it is mag track rather than heading. By this point your maximum magnetic declination is only about 6 degrees. But by 30S 100E it is 9 degrees and by 35S its 15 degrees. So M paths with more southerly end points have a lot of bend towards the end – and I think its difficult to reconcile these with the BTO unless this is accompanied by much slower speed for the last ~2hrs.
@Ge Rijn
Re: Fuel
Assuming a 180S heading, the only obvious target I can see is trying to get over Broken Ridge at 31S and 93-94E. And it looks like MH370 made that with fuel to spare, and went even further south.
Correcting my 180S CTH path for outside temp: Now M0.81 (180S CTH) fairly slow
I see many “events” after Arc5 that may have required pilot input/descent: >Twilight
>Sat Call
>Sunrise
>very high wind from the West
>Cloud cover
>Stretching the fuel (Ge Rijn gives me another list item)
Using Flight Sim, I am not inactive when these things are happening out the window. Rather I envision descending to closer to the cloud level, possibly a heading adjustment to keep going southerly in the wind, possibly other maneuvers if there was a specific flight end plan.
This is a long winded way of saying, I do not see descent to stretch fuel as a complex action, I was already descending to reduce the aircraft visual profile in the Sun.
@TBill
I agree and I can imagine a long gradual descent between 22:41 and 00:11 could meet the objectives you name and save a lot of fuel too as a circumstantial result of a long gradual descent (maybe only flying on idle a long time..)
The time laps between arc 5 and 6 is also 30 minutes longer as the previous. Maybe this shows the gradual drop in speed and altitude while descending till 00:11?
@Ge Rijn
I like a slow 300 ft/min descent starting at twilight at 22:50 and this makes the 23:14 BFO fit (which may not be as important factor as I orig thought), by the time the Sun comes out the altitude is much lower but above the clouds. But the only people can see me are anyone flying low altitude above the clouds like me.
@Victor,
The information you posted on your great circle routes is interesting but insufficient for anyone to duplicate and to verify your results.
I suggest that you post, for the 4 particular routes you have proposed, the following numbers for each Arc (beginning at 19:41 and ending either at 00:11 or at 00:19):
Time
Latitude
Longitude
Flight Level
Temperature Deviation
True Course
Ground Speed
Rate of Climb
Please also list your predicted BTO and BFO and Mach air speed at each Arc location.
The first list is the minimum information needed to independently compute BTO and BFO. The second list (predicted BTO/BFO) allows us to compare BTO and BFO models to look for possible discrepancies. Knowing the Mach you used will permit checking the speed model.
Thanks, Victor.
@DrB: Actually, for the constant Mach number flights, starting with the position at 19:41, the flight level (FL350), the Mach number, and the track angle, you should be able to duplicate the results, as the path is uniquely determined for given meteorological conditions. (Vertical speed was assumed to be zero except at 00:19.) As for ground speed, it varies continuously along the path, so I’m not sure whether it is useful to report results only at times we have BTO or BFO data.
Let me think about whether or not it is useful to compile more detailed results.
@ventus45,
Re: “The drift studies have all (in my opinion) not been extended far enough south to give a true picture. Even Mike stopped at 39 South. I think the drift studies should be extended both sout (to 45S) and west (to 80E) at a minimum.”
There several things to consider.
Firstly, you need to differentiate between 9M-MRO fragments and MH370 debris. Absence of 9M-MRO fragments in Australia does not mean absence of MH370 debris. The MAS unopened towelette could be from MH370, for example.
Secondly, you need to consider spatial distribution of found debris. W/E of the 7nd arc or absence in Australia are too coarse ‘binary’ indicators, so your appeal to these factors look at least childish in comparison with a dozen of drift studies.
Thirdly, the areas corresponding to origins in 38S-40S were well-covered during AMSA aerial search March 18 – March 27. Nothing.
Fourthy, the flaperon barnacles could not experience the temperatures derived from ‘forensic’ analysis by Prof. De Deckker if the origin was south of 36S based on NASA’s measured SST data.
Fifthly, you need to consider actual currents, winds and leeway factors of debris. N-year buoy data are useful only for understanding of general flow patterns, perhaps seasonal features.
In summary, if you look at various criterion, none is met for the origins south of 38S.
I hope this will help you to understand why your suggestion above makes no sense.
@Oleksandr
In your post to Venture. You said: “Fourthy, the flaperon barnacles could not experience the temperatures derived from ‘forensic’ analysis by Prof. De Deckker if the origin was south of 36S based on NASA’s measured SST data”.
This is not the case. The barnacles could well have colonised the flaperon after it had drifted north into warmer waters. The flaperon could have started its journey from S38.
This is a bit off the topic, but I would really appreciate if someone can provide an answer. It is related to SDU power off and re-logon.
In Holland’s paper it is explained how 7 previous SDU re-logons (p7, p8 and table 3) were recorded and further used for the BFO analysis. Particularly, Holland states that:
“The periods of SATCOM outages followed by log-on events were identied from Inmarsat-provided ground-station logs by identifying sequences of three or more unsuccessful log-on interrogations to 9M-MRO (suggesting the SDU was likely powered off) followed by a log-on to the satellite system initiated from 9M-MRO at some later time.”
Yet, there is a common opinion that isolating (intentionally) the left AC bus was the reason for the SDU power outage and reconnection brought the SDU back on followed by the re-logon.
However, data in table 3 suggests that the SDU in 9M-MRO was occasionally staying without a power before the Flight 370. If so, is it more probable that there was a occasional technical issue with the SDU in this particular airplane?
@Marijan: First, welcome to the discussion.
The previous power downs were most likely related to shutdown of the APU and/or disconnect of ground power after both engines were shutdown, which would be normal if there was an extended period between flights. Those power downs would be considered normal, and not due to a problem with the SATCOM. If something was abnormal in the signaling logs, it would have been flagged long ago.
@Rob,
“The barnacles could well have colonised the flaperon after it had drifted north into warmer waters.”
No. The temperature would not drop back to 18C, or time would be insufficient to reach La Reunion. Compatible area is between 26S to 36S, subject to the flotation characteristics of the flaperon (French or Australian). French data result in the peak probability at approximately 28S; CSIRO’s – 30.5S. Whatever you like more among these two, but not 38 and even not 36.
Re: “The flaperon could have started its journey from S38.”
No, it could not. That is what barnacles, NASA and at least two drift studies tell.
@Victor,
@sk999,
Do you have actual magnetic declination table used in 9M-MRO?
@Oleksandr: I don’t. Others do.
@Victor,
Thanks. I recall you modelled magnetic+FPA path, if my memory is correct. What was your source of magnetic declination data?
@Oleksandr: I looked at 180°M paths (heading and track) with a constant rate descent and constant FPA, with a M0.84/310 KIAS speed setting. This produced the right ground speed profile, which in general needs to decrease with time to match the BTO data. I used the actual magnetic declination on Mar 8, 2014, and not the magnetic declination table that was likely loaded.
@Victor,
Thanks; so did I with regards to the magnetic declination.
I think you are in a good standing to easily replicate my latest CMH experiments starting 19:41 as you have all the same forcing. It would be interesting to see if you can obtain similar results.
Oleksandr,
I doubt anyone has the precise table used by 9M-MRO. Some time ago I downloaded a source code program geomag70.c and two data files, IGRF12.COF and WMM2010.COF and computed magnetic deviations from those for several years (1980, 1995, 2005, 2014). My choice at the moment, based on advice from ALSM, is 2005. I store the information in the form of coefficients to polynomial equations, not in tabular form. These were all included in a code dump that I posted several months ago.
@sk999,
Thanks. I have also seen this script, but I simply generated 2D field 0.25×0.25 deg resolution using online calculator from the same website (link is above)
The magnetic declination tables for Jan 1, 2005 and March 7, 2014 are here:
https://goo.gl/ugF2GL
Covers:
0 to -50S
80 to 110E
0 to 10 km
9M-MRO had 2005 tables installed, according to an ATSB source in late 2015. I choose Jan 1, 2005 is a reasonable assumption. The largest change over the 9 years was 2 degrees, so it is not very sensitive.
Were the magnetic tables of 9M-MRO corresponding to Jan 1, 2005, or some kind of averaged, say, over the interval 2005-2010 (or 2005-2015)?
The table in 9M-MRO was not an average. It was as of a date, which I assume was jan 1, 2005. The average for all of 2005 is virtually the same.
BTW…The NOAA software used to generate the tables also output a rate of change value to estimate future values. I asked ATSB to check with Boeing, and they reported back that the B777 software does not use the rate of change information to compute a current value. Thus, I left the rate of change values off in the tables.
ALSM,
R.E. rate change not being used – thanks for checking. It was a nagging issue, although after reading much literature, I kind of figured out it had to be that way. It would seem that, rather than make any value be as accurate as possible, it is preferable to make the error in the value be as accurate as possible. The latter is easier to accomplish if the values are fixed.
ALSM,
Nice sequence of posts. Thank you.
@ALSM,
Thanks for this info.
I will ‘update’ magnetic declinations to Jan 1, 2005 in my CMH model (currently March 8, 2014). I don’t expect much difference, but let’s see.
@Victor: thank you for replying. I understand completely. I will not ask again for this imagery. The audit will reflect that you refused to share evidence you thought offered strong corroboration of the claim the ISAT data is authentic – but it will duly add that you claimed to have a perfectly good reason for refusing. Thanks anyway.
New topic: did the IG – as a group, or as individuals – ever publicly express or imply support for the absurd official claim that one or more of the acoustic “pings” detected at the Wallaby Plateau stood a material (i.e. >30%) chance of having been emitted by MH370’s black boxes? This once-official claim is now widely recognized as, at best, a time-wasting blunder; if any IG members ever supported the absurd decision to go underwater at 23s for April and May (à la R. Quest, e.g.), I’d like to develop a crystal clear understanding as to why, so that we all might learn from this honest mistake. (I don’t recall anyone in your learned group ever having made so basic an error – but if so, i’m still keen to document this explicit and immediate opposition, for posterity).
Second topic: same request, but re: absurd decision to halt the surface debris search, what with the debris having “mostly waterlogged and sunk”.
Third topic: same request, but re: absurd official claim that surface debris had “drifted West”, with first landfall expected on Sumatran shores.
Fourth topic: same request, but re: absurd official claim that the flaperon vindicated the s38-centred search zone (when it demonstrably and emphatically did not).
Profuse thanks for your time and consideration.
@Brock McEwen said: The audit will reflect that you refused to share evidence you thought offered strong corroboration of the claim the ISAT data is authentic – but it will duly add that you claimed to have a perfectly good reason for refusing.
A more honest statement would be that I gave you the source of information that was confidentially shared with me regarding the provenance of the flaperon, and I encouraged you to ask the source for the information directly. I also provided you with the exact language that I provided the person that asked me to analyze the information.
Based on the many times that you have misrepresented information, I am done communicating with you.
@Brock
Fourth topic: same request, but re: absurd official claim that the flaperon vindicated the s38-centred search zone (when it demonstrably and emphatically did not).
That is a bit harsh, IMO. The differentiator being what is probable versus what is possible.
@Victor
Thank you for the clarification. I didn’t realize that all of them were recorded on tarmac.
@VictorI
Maybe a silly question but are all those flight paths till the 7th arc within fuel range (according @DrB’s endurance/range tables)?
And if not which fall short?
@TBill
What I like is you are one of the few seriously looking into an all active piloted flight.
I mentioned several times before this is the most sensibel and logical thing to assume.
There is no proof at all even no clear indication the flight went into a ‘ghost flight’ anywhere along it’s route.
The chance of something like that to happen must be one in several millions of flights.
Therefore it’s kind of strange so many took the road of an ‘unpiloted flight’ somewhere after IGARI.
They rather take the one in a million chance than the obvious statistical normal without any proof to do so.
It might be more convenient to take this approach for else it becomes a lot more complicated.
But this should be no reason to reject normal reality.
Imo out of millions of flights this one was actively piloted from beginning till end also.
There’s no sensible reason, no proof and no clear indication to think otherwise.
@Ge Rijn said: There is no proof at all even no clear indication the flight went into a ‘ghost flight’ anywhere along it’s route.
We know there are reconstructed flight paths for MH370 that match the satellite data without pilot input after 19:41. That fact alone means we have to consider this possibility.
I don’t know for sure whether or not there was pilot input after 19:41. However, if there was a decision to fly south to the SIO, I would prefer to soon after die of hypoxia (turn off the air packs and open the outflow valves) rather than die in a plane crash after flying for 5 hours into oblivion, or worse, survive the crash and die of starvation, injuries, drowning, hypothermia, shark bites, or other terrible ways.
..to add. For this reason I think there could have been no NORM magnetic heading after FMT. It must have been either a track towards a specific waypoint or a true heading with the NORM/TRUE switch set to TRUE after FMT.
@VictorI
Yes, I would preferre hypoxia any time too when I had no other goal in mind and was forced to fly into the SIO by something or someone.
There’s no indication (yet) the pilot/plane was forced to fly into the SIO.
All indications and circumstantial evidence point to a deliberate choice of actions. All manouvres, sim-data. Nothing similar against those facts.
The point I make is the chance of a flight turning into a ‘ghost flight’ is one in a million. This chance is not a chance to build a case on when there is no actual proof or clear indication for it.
Most long haul flights go without pilot input between getting on altitude and starting the descent. This is not a legitimed criterium I think.
I agree about considering the one in a million possibility.
But why not put more attention to the most obvious?
In my latest paper on the MH370 Drift Analysis, I show how the South Equatorial Current at around 18°S 80°E can be reached from 30°S on the 7th Arc. Reaching the South Equatorial Current is essential, if MH370 Floating Debris is going to reach all 27 locations within the timeframe, where finds have been made.
I also show, by how many months in advance or behind the nominal time of April 2015, the location of all 27 MH370 debris finds can be reached, from Rodrigues to Mozambique, or from Tanzania to South Africa, by only varying the start month from 18°S 80°E from the nominal start month of April 2015 +/- 5 months.
https://www.dropbox.com/s/ma9pt8zf3b0whc7/MH370%20Drift%20Analysis%20-%20Calibration%2C%20Results%20and%20Conclusion.pdf?dl=0
..to correct; ‘when I had another goal in mind..’
@Ge Rijn asked: Maybe a silly question but are all those flight paths till the 7th arc within fuel range (according @DrB’s endurance/range tables)?
First, be careful using the value for the maximum southern latitude after a hold as the assumption is the flight to the south begins at the 18:29 position. In fact, the plane could have flown south before 19:29.
Bobby’s model predicts that there is sufficient fuel endurance for a descent to a one-hour hold at FL210 followed by LRC flight at cruise altitude. This is consistent with the great circle paths reaching 34S and further north, especially if the LRC flight occurred with air packs off, which I estimate results in a fuel flow reduction of 2.5%. As end points further south have already been searched, I’d estimate that any potential impact point along the 7th arc and north of the previous search area would not be excluded based on fuel endurance.
@Ge Rijn
I am thankful we have Victor and Richard offering flight path scenarios closer to my thinking. The avoidance of pilot input scenarios is (1) partially reluctance to “go there” but also (2) gives the models problems as there is no easy way to model that scenario. The lack of finding MH370 at 38S implies that a more complex flight path with pilot inputs is possible, which Victor and Richard pointed out a long time ago in their critique of the ATSB Bayesian model’s assumptions.
Ge Rijn: The probability of a ghost flight may be 1 in a million as you say, but the probability of a suicide flight is far lower. There are approximately 100,000 scheduled commercial flights per day now. There have been 6 cases of a suicide flight on a commercial flight in the last 35 years. That works out to be ~1 in 100 million.
@ALSM
But, we have a solid Bayesian prior with respect to MH370 – the flight went missing. 🙂
In the 35 year period approximately 3000 (85 per year average) commercial flights experienced an accident (not all fatal crashes). So the suicide probability is approximately 6 in 3000 or about 1 in 500.
@Richard: Thank you for sharing your most recent work with us.
If I understand you correctly, you believe there is much more variability in the timing to travel between the 7th arc and the South Equatorial Current (represented by the position 18S, 80E) than there is variability in the path after reaching the current at a particular time.
Your results suggest an impact point along the 7th arc between 27S and 31S latitudes, which is significantly further north than the small search area recommended by Griffin of around 35.6S latitude. I think the main reasons for the discrepancy are:
1) You put more weight on the timing of the discovery of “Roy” in Mossel Bay, South Africa, in December 2015, which moves the impact point well north of 35.6S.
2) Griffin assigns a higher level of confidence to the efficiency of the surface search for latitudes north of 33S.
I’d say that neither the timing of the discovery of the flaperon on Reunion Island nor the lack of debris on Western Australia can discriminate between 35.6S latitude and impact points further north along the 7th arc.
Another observation is that even if we accept CSIRO’s calculations of probability of detection of debris during the surface search (which some do not), there are significant holes in coverage at latitudes north of 33S, especially if the impact was to the northwest of the 7th arc.
Based on this, Ocean Infinity is wise that it is planning to search beyond the 25,000 sq km search area that was prioritized by CSIRO.
@TBill
Yes, also @Victor’s latest tables point to an area between ~29S and ~34S with the smallest (BFO/BTO) errors.
Endurance and range beyond ~34S I read as not very probable from any point at 19:41 according his flight paths.
@Richard has always interesting papers.
@Ge Rijn: Paths with smallest BFO/BTO errors are not necessarily the most likely. Depending on what we expect the BFO error to be, acceptable paths can extend significantly further north than 34S.
I don’t know the exact cut-off latitude for what impact points are not allowed by fuel considerations, as it depends on how the plane was flown before 19:41 and whether or not air packs were on after 19:41. However, I do estimate that latitudes further north than 34S are allowed.
DennisW: Re: Yes, you could say that ~1 in 500 crashes involved a suicide. Either way you crunch the numbers, there is no purely statistical basis to suggest that a suicide was likely involved in MH370. Based on these statistics, the probability is somewhere between 1 in 100 to 1 in 500…either way, very unlikely. That does not rule it out. We certainly know more than historical stat’s.
My point is that the statement:
“…the chance of a flight turning into a ‘ghost flight’ is one in a million.”
… is totally wrong. Given that MH370 did in fact disappear, the odds of a ghost flight or a suicide flight are roughly equal at this point, and together, that covers >90% of the possibilities.
@ALSM
So a straight suicide flight can be excluded I agree.
9/11 were no straight suicide flights also. They had a well thought out purpose and destination.
This was also the case with MH370 I believe.
Or the PIC was interupted in achieving his end goal, or his end goal was a deep trench in the SIO just under Broken Ridge.
Update on November 11, 2017:
In Figure 4 above, I have plotted the paths to the four waypoints (NZSP, YWKS, NZPG, 45S) on the same plot that was generated by CSIRO to show the cumulative probability of detection of low-windage debris by the surface search for various impact points along the arc. The calculated probabilities include the drift that might have occurred between the time and location of the impact and the time and location of the search. It can be seen that there are impact points along the 7th arc and north of 33S where the probability of detection is significantly less than 100%, especially if the impact was to the northwest of the arc.
@ALSM
To add. We only have the Helios flight out of those million flights as an accident without suicidal intentions.
That’s not enough to build any ‘ghost flight’ assumptions on regarding MH370 imo.
Victor- Fig 4 is very good.
If we say the area of most interest is 20-40S, we could calculate the % of Arc7 adequately searched so far. This could help push back on the bogus argument that Arc7 has already been totally searched.
Trying to think if there is some way we could superimpose a drift study fuzzy area from 28-35S to have all 3 indicators (Sat data/Surface debris/drift data) all on one plot.
Ge Rijn: At least 5 ghost flights, not just one…
https://goo.gl/6K7R2E
Victor:
Based on your chart added this morning, the 180 route to NZSP looks like the most likely route, if there was any route selected and the cumulative search data are accurate. It is also very close to the area suggested by CSIRO. Perhaps the first area to be searched should be expanded slightly to cover both areas.
@ALSM: Yes, of the four great circle paths I selected, the NZSP route has the highest probability of non-detection of surface debris. As shown in Figure 3 above, that impact point is part of the 25,000 sq km that will be searched first by Ocean Infinity. However, there are also indicators that points further north are possible. The path crossing the 7th arc at 28.3S latitude and ending at 45S, 104E, for instance, better fits the timing of the discovery of “Roy”, as Richard has shown. Also, as Figure 4 shows, the surface search had a reasonable probability of missing the debris for a 28S crossing if the plane turned and circled back, impacting to the northwest of the arc.
@VictorI,
You said: “As for ground speed, it varies continuously along the path, so I’m not sure whether it is useful to report results only at times we have BTO or BFO data.”
I believe it is not only useful, it is also sufficient to validate all proposed routes with acceptable accuracy. As you well know, it is very much easier to evaluate a route than to generate one.
The “usefulness” of the ground speed is two-fold: (1) it a necessary parameter to compute the BFO at the arcs, and (2) it is necessary to know it along the track to generate (but not necessarily to evaluate) latitude and longitude points along a path. To minimize the amount of data I suggested be provided for any proposed route, including yours, I only asked for values of ground speed at the arcs. Primarily, that allows one to duplicate the BFO calculation as a means of comparison to confirm those results.
I only requested the minimum amount of information that would allow reasonably and sufficiently accurate verification of a proposed route. Anything less than that defeats the purpose of verification. Anything more than creates a lot of work that is generally unnecessary. What I mean is that one can take the ground speeds at the arcs plus the temperature deviations and the altitudes/lats/lons at the arcs, and from them generate a very good estimate of the instantaneous Mach at the arcs and also of the average Mach and average ground speed along a leg. This will be of sufficient accuracy to determine consistency with the satellite data.
You really only need 3 things for verification: a calculator, a weather database for the arc locations and times (wind and temp from NOAA GDAS online, for instance), and a distance calculator (SkyVector, etc.). Since the wind data are of limited accuracy (~6 kts RMS in two directions), small computational air speed and course errors are unimportant since they will be absorbed as a wind error anyway. For this reason, for many parameters, one can use the average of the arc values at the leg ends as an estimate for the (integrated) leg average value, without introducing significant errors.
I was thinking about modifying my posted BTO/BFO calculator to add the calculations of the following parameters:
Both at the arcs and average values for each leg:
Ground speed
Mach
True course
Magnetic course
True heading
Magnetic heading
At the arcs only:
Predicted BTO
Measured BTO
BTO error
Mean BTO error (all arcs)
RMS BTO error (all arcs)
Predicted BFO
Measured BFO
BFO error
Mean BFO error (all arcs)
RMS BFO error (all arcs)
For the leg averages only:
Average database tailwind (for each leg)
Average route tailwind (for each leg)
Tailwind error (for each leg)
RMS tailwind error (all legs)
Average predicted crosswind (for each leg)
Average route crosswind
Crosswind error (for each leg)
RMS crosswind error (all legs)
Initially, one would have to manually enter temperature, wind speed, and wind direction at each arc location from the online database, although eventually I might automate this using the GDAS weather tables. Similarly, perhaps eventually I could integrate my fuel model so that endurance may be simultaneously determined.
Such a tool would allow a larger number of people to evaluate (and perhaps even to generate) routes, both published and unpublished, and provide a standard method for this process. If you provide the data requested, I will create the tool for everybody to use.
@Ge Rijn,
I am surprised by the series of your recent posts. You have kicked yourself into the “whacko” category.
Re: “There is no proof at all even no clear indication the flight went into a ‘ghost flight’ anywhere along it’s route.”
There is strong indication: BTO and BFO data post 19:41 are consistent with the simple modes of the autopilot. Chances of such a coincidence assuming a piloted flight are minimal.
Re: “They rather take the one in a million chance than the obvious statistical normal without any proof to do so.”
Exactly the opposite is true: chances of an actively piloted flight past 19:41 are one in a million: that is because of the “trend line” in the BTO and BFO data.
Re: “Imo out of millions of flights this one was actively piloted from beginning till end also.”
How did you arrive to this nonsense?
Re: “For this reason I think there could have been no NORM magnetic heading after FMT. It must have been either a track towards a specific waypoint or a true heading with the NORM/TRUE switch set to TRUE after FMT.”
Why would NORM be less likely than TRUE? It was in the shadow for long time because there was confusion with regard to whether NORM is used for display or navigation. In addition, the DSTG a priory implicitly eliminated magnetic modes, without conducting a proper fuel analysis. I actually suspect the ATSB knows about the existence of the magnetic mode routes, but they are reluctant to publicly admit this.
So… be realistic.
@DrB: I don’t think that it is prudent to simply average the meteorological conditions at the start and end of the legs to represent the average conditions during the leg. I calculate all parameters at a time resolution of 1 minute (except near ping times where the time step is chosen to match the ping time to the nearest second) and I don’t average.
Here are some more detailed results for the great circle path ending at 45S, 104E. You should be able to replicate these results.
@Ge Rijn
““For this reason I think there could have been no NORM magnetic heading after FMT”
Although I have found a 180S True Heading case I like, that implies there could be 180S Magnetic case I like too. Since I am taking the liberty to pilot the aircraft, I can certainly can and do watch the POS longitude tick off along the way and I can double-click and switch over to True or say 185 when the winds start pushing me East past 94E.
@Victor,
Your Fig. 4 should be treated with great caution as it may give false impression.
CSIRO’s result in background correspond to low-windage objects (1.2%) only. In comparison, the DGA has experimentally found the leeway factors of the flaperon of 2.76% and 3.29% corresponding to the two stable configurations. The latter is close to my theoretical estimates of 3.33% for a thin flat-floating object. I was also involved in some drogue studies, and we estimated the leeway coefficients of 5-6%. So this CSIRO’s particular plot is unlikely representative.
@Oleksandr: I agree with your caution in using Fig 4. However, this is what CSIRO has used to justify that debris should have been detected for an impact along the 7th arc that was further north than 33S latitude. Using their own results, there are significant holes in the coverage, especially for an impact to the northwest of the arc.
@Brock,
I don’t know what you are trying to achieve with your endless requests, but you seem to be off on a tangent, a wild goose chase, with no prospect of ever reaching an end-point.
Just for the record, I am not aware of any IG support for the “acoustic pings”. In fact I dismissed the claims as rubbish [as I think others did too], because the technical evidence could never support such a claim.
You seem to be labouring under the misapprehension that the IG had some direct influence over the official investigation and the decisions made by the ATSB and other official parties directing aspects of the search. Emphatically, the IG had no such influence.
Papers published collectively, and individually, by the IG were all based on publicly available information at the time, and expressed the views of the IG at that time. The papers were produced, always, to provide interpretation and comment in the public domain. Official organisations may have chosen to make use of this work, just as they may have taken into account work done by a number of other commentators and contributors both private and in the public domain.
What “audit”? Your “audit”? No one is interested. Please stop asking dumb questions.
@Victor
I have zero confidence in the CSIRO surface search probability graphic. The S3/S4 search zones (the area of interest for the NZPG path was not even started until 28 March 2014. Furthermore, I have never seen a surface search probability graphic published anywhere else. Where are they getting this stuff? What is the basis for the graphic? Frankly, I think they just made it up.
@DennisW: I think questioning of the probabilities is justified. But even assuming they are correct, there are a lot of holes.
Indonesia.
Some pretty strange things can happen completely unnoticed down there…
http://www.bbc.co.uk/news/world-asia-37997640
@Dennis,
Re: “I have zero confidence in the CSIRO surface search probability graphic.”
I have obtained similar results, so I am confident this CSIRO’s plot is correct. However, I noted ‘dramatic’ dependence of the debris coverage on the leeway factor. At higher leeway factor our models show some discrepancies: I think this is because I used different wind data for forcing.
Re: “Where are they getting this stuff? What is the basis for the graphic?”
Based on coverages. Probabilities assumed to be proportional to the coverages of the modelled ensembles. AMSA provided daily surveyed areas.
Of course, there is a factor of ‘detectability’. I don’t know how AMSA detected debris besides using ‘brutal’ manpower, but say we used Fugro’s LiDAR bathymetric surveys of 0.2 m resolution in 10×10 km areas – I trust floating debris could also be detected using LiDAR or similar technology.
You may also read Garcia et al. “A dynamical systems approach to the surface search for debris associated with the disappearance of flight MH370”, Nonlin. Processes Geophys., 22, pp. 701-712, 2015.
Oleksandr,
Nice find of the Garcia et al article. From the abstract: “Perhaps the most relevant aspect of this mystery is that not a single piece of debris from the aircraft was found during the intensive surface search carried out for roughly 2 months following the crash.” The article was published on July 27, 2015, which happened to be 2 days before Johnny Begue discovered the flaperon on Reunion Island.
@Oleksandr
Interesting reference. Thanks for that.
Of course, there is a factor of ‘detectability’.
That was the basis for my skepticism of the CSIRO probability map. That and the amount of time that passed between the incident and when the areas were searched. No matter. I don’t get the sense that the aerial search is a big factor in future search decision making.
@all
Generally no one likes to look stupid, but I done so many stupid things over the course of my life that are part of the public record that I no longer have any fear of it (although I did refrain from making this post for some time now).
The subject is integrated Doppler. Integrated Doppler is path independent. It is only a function of the end points of the path. A path difference of -1km with a transmit wavelength of 1 meter will produce will produce an integrated Doppler of 1000 cycles. If the time of the path transit is 1000 seconds, the average Doppler will be +1Hz.
So looking at the measured Doppler residuals, that is the difference between the actual aircraft Doppler and the Doppler compensation necessary to produce the measured BFO, one gets an average Doppler residual between IGARI and 00:11 of around 11Hz. If one computes where the integrated Doppler residual is 11Hz based on positions on the 6th ring relative to IGARI, one finds a latitude of about 30S. Of course, all the uncertainties about AES oscillator drift apply.
I will write this up in some detail, but I just wanted to toss it out there for comments.
The geometric algebra solution also homes in on 30S, but I have no explanation for that which is why I took down the link to it. I am still fiddling around with that, but have no clarity as yet. As near as I can tell Gibbs should work just fine. No one on this blog dove into that, and I don’t blame them. The learning curve is steep. (Steeper with the age of the reader.)
The annimated graphics in the supplement to the report are very interesting.
https://www.nonlin-processes-geophys.net/22/701/2015/npg-22-701-2015-supplement.zip
Obtainable on the right – in very small print under the PDF download button for the main report.
https://www.nonlin-processes-geophys.net/22/701/2015/npg-22-701-2015.html
@Ventus45
Interesting article and animations indeed.
Particulary interesting I find the 6 drifters in the 32S ‘above Broken Ridge’ animations are not moving to the east towards Australia.
Actualy the two closest to the 7th arc move to the west and north-west. The others stay more or less stationary during the period.
None show tendency to drift towards Australia.
It’s a pity there is no animation of the area between ~36S and 32S.
This would possibly show CSIRO/Griffin’s narrow drift-flow to the west and east around 35S where they based his latest priority area on.
Then finaly in their conclusions they suggest also other areas should have been search and some could have been avoided without naming specific areas. That’s a piy too imo.
..a pity too..
@DennisW @all
Straw man to develop realistic/approximate expectations for future OI bottom search.
Prelim. Assumptions/Guesstimates:
95% Chance MH370 is between 20-40S +- 200 nm of Arc7
90% Chance MH370 is between 20-40S +- 60 nm of Arc7
80% Chance MH370 is between 20-40S +- 30 nm of Arc7
15% Chance MH370 is Unfindable due to Undersea Mountains/Trenches/etc.
Results derived from the above assumptions:
20% of Arc7 searched to date between 20-40S +-30 nm of Arc7
68% Chance of finding MH370 between 20-40S +- 30 nm of Arc7
77% Chance of finding MH370 between 20-40S +- 60 nm of Arc7
5% Chance finding MH370 outside of 20-40S
10% Chance finding MH370 between 20-40S but further away than +-60 nm
Less than 68% chance finding MH370 in CSIRO next 25000 km search area
@Oleksandr: I briefly looked at your CMH results at various altitudes. Are you sure you are properly calculating BTO errors? I looked at a couple of points and my calculated errors were about twice yours.
@All: I will repeat my question. After the first 25,000 sq km is searched, where should OI search next? Please provide the logic that supports your answer.
Have we come full circle back to Inmarsat’s original prediction?
http://www.bbc.co.uk/news/science-environment-27870467
@Victor
For OI
> In general, cover Arc7 +- approx 30 km2 between 20-40S
> Priority to 180S paths which extends East to 32S very difficult area, but please try. I feel ATSB says below 32.5S just to avoid the BR area. Some feel MH370 may have turned east at the end, and some feel BR was the target, so that could possibly be a 180S flight path outcome.
> Priority to McMurdo option
Correction above +-30 nm of Arc7
@Victor,
Possible sources of discrepancy:
1. Do you compute round-trip error, or single-way error (i.e. distance error divided by the speed of light)