MH370 Families Launch Private Search Effort

Debris map from MH370 produced by the School of Civil, Environmental and Mining Engineering & The UWA Oceans Institute

Families of MH370 passengers and crew gathered on March 4 for the Third Annual Remembrance Event of the disappearance. As part of the three-hour event, Dr Charitha Pattiaratchi, a Professor of Coastal Oceanography from the University of Western Australia, presented “The Utilisation of Ocean Drift Modeling Techniques to Locate MH370”. According to a story by aviation writer Geoffrey Thomas, Dr Pattiaratchi said that UWA’s reverse-drift modelling puts MH370 “at Longitude 96.5 E Latitude 32.5 S with a 40km radius”.  This means his estimate is at the northern end of the new search zone recommended recently by the ATSB, which was between latitudes 32S and 36S, with a total area of 25,000 sq km, and with the highest probability at 35S latitude. Dr Pattiaratchi claims that UWA’s drift model is consistent with the recovery of 18 of the 22 pieces of debris found to date.

Also as part of the Remembrance Event was the announcement that Voice370, an advocate and support group for MH370 families, launched an effort to raise money to privately search for the aircraft. In a statement recently released, the group would like to search the “newly recommended 25,000 sq km”, which presumably is the same area recommended by the ATSB. The statement is careful to not cite the group’s fundraising goal, although in a previous Reuters story , Grace Nathan, daughter of MH370 passenger Anne Daisy, pegged the number at $15 million.

This blog is dedicated towards solving the mystery of the disappearance of MH370, and I fully support continued efforts to find the plane. However, I pose a simple question: “What level of confidence do we have today that the plane will be found in the 25,000 sq km of seabed now proposed?”

Updated on March 8, 2017

A recent interview with Dr Pattiaratchi appeared in The New Daily. Here is an excerpt from that article.

[Professor Pattiaratchi]  claims the research gives authorities the “credible evidence” required to restart the search.

“That’s as good of information as you can get from an oceanography point of view.

“There is absolutely no doubt about the debris that has been found.”

The ATSB spent almost two years searching a 120,000sqkm area in the southern Indian Ocean for MH370, an area the UWA model predicted would prove fruitless.

“As soon as the flaperon (part of the aircraft’s wing) was found, we were saying it was unlikely that the plane went down in the search area at that time,” Professor Pattiaratchi said.

“The ATSB did not take into account the debris that was found. And despite the flaperon being found on [Reunion Island in] July 2015, it took them until November 2016 – almost 18 months – for them to acknowledge [MH370] is not [located] where they were searching.

It is true that the ATSB and DSTG did not incorporate drift modeling into their analyses until fairly late in the game, and Dr Pattiaratchi’s remarks are going to cause them some embarrassment. But we also have to ask ourselves:  What part of the ATSB’s analysis was incorrect?

636 Responses to “MH370 Families Launch Private Search Effort”

  1. Gysbreght says:

    There are two possibilties: Either the airplane was still under human control and flew further than 40 NM beyond the 7th arc, or there was no human control and it ended up close to the 7th arc North of the search area. There’s not much to choose between those possibilities, but I’m currently leaning towards the second.

    The ATSB could help by being more open about what they know. David and I are still waiting for a reply to simple questions that could have been answered within a day.

  2. Paul Smithson says:

    I would be very keen to see directly what the good Dr has predicted and how (rather than what the media reports him to have said). For information, the “confidence limits” diameter for estimated point of origin on a *2-week* reverse drift model is in the order of 100km in the southern ocean. So quite how one arrives at such an extraordinarily accurate predication from back-tracked drift is a mystery to me.

  3. lkr says:

    @gysbreght

    I agree your first statement. However, I think it would be more accurate to say for the second end of flight scenario, “close to the 7th arc North of the search area ,, WHICH WOULD NOT REQUIRE human control”. I see NO inconsistency in thinking a PIC might ride it out til flameout, then either pitch it over, or control a guide.

  4. DennisW says:

    “What level of confidence do we have today that the plane will be found in the 25,000 sq km of seabed now proposed?”

    ~20%

  5. AlexSiew says:

    Brock,

    Regarding your comment in your previous thread about the 4600 us offset.

    The 2 ‘Log-on Requests’ exactly 1 second apart at 15:59:55:413 and 15:59:56:413, were disclosed by the ATSB in a December 2014 update to the Inmarsat datalog released in May 2014. https://www.atsb.gov.au/publications/2014/mh370-update-to-signalling-unit-logs.aspx

    At all times prior to the release of the Factual Information Report in March 2015, it was thought by all non-insiders that the plane (and the SATCOM system/s) had powered up just prior to 1600 UTC, as the first line in the original Inmarsat datalog (the Log-on Acknowledge) was timed at 16:00:13:406. This misconception was reinforced by the ATSB Dec 2014 update which said: “..There are two logs that contain BTO and BFO data prior to 1600 UTC while 9M-MRO was powered up at the gate in Kuala Lumpur…”. The FI revealed that the plane had actually powered up more than 3 hours earlier, at around 12:50 UTC and that the SATCOM system (or one of the SATCOM systems) had powered up at the same time with a first log-on at 12:50:19 UTC. So either the ATSB was ignorant of the true time of power-up or the ATSB was lying, in the December 2014 update.

    No one, to my knowledge, has come forward to explain why there were 2 ‘Log-on Requests’ as per the December 2014 update at 15:59:55:413 and 15:59:56:413, exactly 1 second apart. The FI showed a Log-on Handover by the HGA SATCOM system at 15:59:57 UTC, so conceivably one of these 2 ‘Log-on Requests’ could have been this handover. But what about the remaining ‘Log-on Request’?

  6. Sajid UK says:

    @Victor @all

    First of all, congratulations to Victor on setting up this excellent blog. The following comment I was actually trying to post on the previous thread but it closed within the last hour. I hope Victor doesn’t mind me posting it here instead…

    I’ll take liberty of not reading previous comments on here (yet) so forgive me if I’m rehashing/repeating anything that’s already been said.

    After reading the Lido Hotel radar article, one question above any other came to my mind above all others: Would the Malaysians really have been so careless and clueless as to present a slide with hidden evidence of a fighter jet track for the whole world to see and subsequently pore over, including the media and countless other MH370 experts and enthusiasts?

    If the Malaysian military were involved, I’d have assumed they would’ve taken extra care to ensure any evidence was not inadvertantly ‘leaked’ in such a way…

    Which potentially leaves the possibility that Victor may have stumbled upon something completely new, something that everyone (including the Malaysians) might have totally missed…

    But that would leave us asking: if it wasn’t a Malaysian fighter jet, then who the bloody hell was it?!

  7. Brock McEwen says:

    Even if search leadership had earned our trust (and my annoyingly stiff audit of search decision support to date suggests strongly that it hasn’t), Dr. Pattiaratchi’s result still presents a bit of a litmus test: do you focus on the 18 debris discoveries his impact zone explains, or the 4 that it doesn’t?

    Suspect the probability @Victor is surveying may hinge a bit on the answer to that question.

  8. DennisW says:

    @Victor

    “Also as part of the Remembrance Event was the announcement that Voice370, an advocate and support group for MH370 families, launched an effort to raise money to privately search for the aircraft.”

    This effort is simply preposterous. The administrative and legal nuances associated with implementing such a scheme are formidable not to mention the management of any search efforts. I don’t see anyone reputable signing up for such an activity. It will attract Milne types who have neither the experience nor the credibility to manage such an undertaking.

    Like Broc, I am critical of the tripartite group involved in the formal search, but they did have (with the exception of Malaysia) established resources they could activate for the effort.

  9. DennisW says:

    @Paul Smithson

    Like the barnacle work of De Deckker, we will see nothing from Pattiaratchi. I doubt anything even exists besides stuff in lab notebooks and the like. There is no way to know if he even actually said what is being attributed to him.

  10. Victor Iannello says:

    @DennisW: My guess is their focus is to continue where the ATSB left off. Right or wrong, there is the widely-held belief that if the ATSB had only searched the final 25,000 sq km, the plane would be found. I don’t think they will seriously consider teaming with other groups that are proposing a search in a different area. Already, they are distancing themselves from another group that is trying to raise money for a more northerly search.

  11. TBill says:

    One group wants to search 32-36S and another group wants to search 22-26S. For For certain therefore, it’s at 26.9-31S.

  12. Ge Rijn says:

    It’s still not proven MH370 dived high speed into the ocean near the 7th arc.
    This assumption is mostly based on the final BFO’s and that the outboard flap was most probably not deployed (not 100% proof either). Which is no proof on its own the plane could not have glided towards the surface.
    How seperation of the outboard flap and flaperon (or other pieces) occured has not been (officialy) reveiled and explained (yet). High speed flutter-seperation of some pieces is not more than an assumption.

    IMO the kind of debris, their positions on the plane (~90% trailing edge, wing related pieces), their general sizes and (lack of) very high speed impact damage, the (small) amount of pieces found to date, indicate a relatively low speed ~level entering of the water IMO.

    As long it’s not proven 100% (with official information on debris and data) the plane entered the water with a high speed dive near the 7th arc and with that the possibility of a kind of glide and farther travel beyond the 7th arc is still open (and likely IMO), it’s unwise to start searching the new area or any area.

    8 march today. My thoughts are with the NoK who probably suffer their most difficult day of the year.
    I wish them all possible comfort.

  13. David says:

    @Andrew. Overflow from before back a bit you posted,”With no hydraulic power, the left flaperon PCUs will operate in bypass mode, allowing the flaperon to move freely in both directions. In flight, the air loads will cause the flaperon to move upwards about 10 degrees. That too also produce a rolling moment to the left”.

    I raised this with the ATSB three weeks ago, as to whether it might have been not allowed for in the most recent end of flight simulations. My thesis was that it might well impose increasing bank, not much evident during those simulations. No response.

    Having looked into this again I have found the outcome more complex than I hoped. My conclusion is that more likely there would be roll to the left though not certainly. The left roll could would well result from residual left engine thrust at TAC removal.

    This outcome could help in understanding the Boeing simulations. I presume that one day we will learn the details of the configurations for those.

    https://www.dropbox.com/s/z6zlkc30de1e796/MH%20370%20roll%20at%20fuel%20exhaustion.docx?dl=0

  14. David says:

    @Gysbreght. About most recent ATSB responsiveness, they have told me that a fatal crash at Essendon airport in Melbourne has diverted effort which otherwise was on this job but that an answer could be expected.

    An article on the national TV here last night was critical of their timeliness generally and it may be that they are under resourced. Quite possibly MH370 has overloaded them.There will be support for fixing that if there is an enduring problem.

    @Victor. I attach today’s missive from the Minister. Past tense I note so not encouraging.
    http://www.minister.infrastructure.gov.au/chester/releases/2017/march/dc047_2017.aspx

  15. Richard says:

    Prof Charitha Pattiaratchi has a list of the 222 publications he has co-authored on his web site:

    https://www.socrates.uwa.edu.au/Staff/StaffProfile.aspx?Person=CharithaPattiaratchi

    Unfortunately there is no paper that relates to MH370 or drift analysis.

  16. Andrew says:

    @David,

    Thanks for posting your reasoning for the aircraft’s roll at fuel exhaustion. One comment:

    “The reason that TAC ‘may’ compensate comes in part from the recent @Buyerninety post, which says it will not necessarily respond to a catastrophic failure until, “the fuel control switch has been moved to cut-off͟” and the TAC cycled, which require a pilot. For now I assume it does compensate: I note that the Boeing simulations do not support TAC failure as having been encountered.”

    The highlighted behaviour occurs in situations where the TAC can’t accurately determine the difference in thrust, as sometimes occurs with catastrophic engine failures (eg turbine disc failure). In such cases, the TAC can’t accurately determine the engine failure until the associated fuel control switch has been selected off. With a straight engine flameout, as would occur at fuel exhaustion, the TAC should have no problem detecting the engine failure and should work as advertised.

  17. Gysbreght says:

    @David
    @Andrew

    Two questions:

    “In an electrical configuration where the loss of engine power from one engine resulted in the loss of autopilot (AP), …”

    (a) What is the response of the TAC to the first flame-out?
    (b) Does the Autothrust remain operative, i.e. does the thrust of the remaining engine increase to max CLB rating, or does it not change?

  18. Victor Iannello says:

    @Alex Siew said, “No one, to my knowledge, has come forward to explain why there were 2 ‘Log-on Requests’ as per the December 2014 update at 15:59:55:413 and 15:59:56:413, exactly 1 second apart.”

    Since you have a bad memory, I will remind you that we had a previous exchange on Reddit a year ago on this very topic. At that time, I explained:

    The SATCOM on 9M-MRO consisted of a single SDU. The double entry that you cite at 15:59:55.413 and 15:59:56.413 is a “Signal Unit Set”, consisting of an Initial Signal Unit (ISU) for the Log-on Request and a Sequential Signal Unit (SSU) for the Flight Information, i.e., the Flight ID. By contrast, for the log-on request at 18:25, there is only one SU (the ISU), and no Flight ID is transmitted, as noted correctly in the FI. End of story.

    Unless you have new facts to present, we have to move along.

  19. DennisW says:

    @Richard

    Yes, Pattiaratchi is a prolific publisher, but not on subjects involving funding decisions.

    The advantages of information asymmetry are well documented in both corporate and government domains. There is absolutely no benefit associated with MH370 information disclosure, and many potential negative consequences.

  20. DennisW says:

    Third “interim statement” released today. Link below to pdf. Not worth reading, IMO.

    http://www.mh370.gov.my/index.php/en/media2/transcript/category/21-3rd-interim-statement-8-march-2017

  21. Ge Rijn says:

    @DennisW

    “There is absolutely no benefit associated with MH370 information disclosure, and many potential negative consequences”.

    I strongly disagree. By now it’s of utmost importance ALL available information on the debris (kind of damage, separation, possible flight attitudes on impact), radar data, drift data, Inmarsat data, barnacles, gets disclosed soon.

    If the ISDG and ATSB (and France) have no conclusive anwsers then at least they are obliged in this stage to share all their information and data with the NoK and and the community who is working for 3 years now to solve the mystery.
    Not sharing this information is a straight offence to the NoK at first and second to all people affected by this drama and seriously making effort to find answers. Which you are one of them.

    To me it’s clear we’ll we are destined to keep on guessing too much without those information and data.

    In my view big progress has been made by the IG and the community as a whole. By @Richard’s and Pattiarachi’s latest detailed drift analyzis consensus builds strong on a crash area between ~28S and ~33S.
    @VictorI and others have found possible routes and attitudes of the plane that could explain a more northerly crash area.
    South of ~35S is out of the question by now. Nothing is found there and it’s not according the latest analyzis so no suprise anymore.

    I think we as a whole are getting closer and closer. There is reason for optimism in this way.
    But without discloser of all the information and data it will be impossible to make much further progress IMO.

    If the ATSB and DSTG really want to solve this mystery they have to share all their information by now. At least with the NoK and the IG members.
    By now they should know they have become a very valuable part of the investigation as well as all other seriously involved contributors.

    I’m confident we all are getting really close.
    We just need all the information and data to get even closer.

  22. Jerry M says:

    @ all; linked is a file from Feb 28 published by Malaysia with a part by part analysis of the debris in their possession found from April thru July 2016 – about 20 parts. I looked back at posts since Feb 28 and haven’t seen this linked so I apologize if it is old news.

    Anyway since you are all smart people, part of their analysis on parts believed to be from MH370 or at least a B777 is the structural damage. On almost all of these part they state something like: “The fracture on the laminate appears to be more likely a tension failure. The honeycomb core was intact and there was no significant crush on the honeycomb core.” Does this say anything about high speed crash into the ocean vs a pilot controlled glide scenario? If you read the report they say it several different ways but I’ve always wondered if analyzing the structural damage has any merit. I know the separation of the flaperon has gotten a lot of discussion but it is not included in this report. Thanks

    http://www.mot.gov.my/SiteCollectionDocuments/MH370/Debris%20Examination%20Reports%20-%2028Feb2017.pdf

  23. TBill says:

    @Ge Rijn
    I assumed DW meant no benefit to info disclosure for the powers that be (MY, ATSB, Boeing, airline industry, etc.). Others have alluded to an unspoken truth, that finding the plane is undesirable for some, because the cause of the accident could raise a lot of difficult questions, no matter what the outcome. Some journalist said this better than me, but that’s the idea.

  24. Ge Rijn says:

    @TBill

    I only read the things @DennisW writes. In this case it’s clear and subjective to me. Only his opinion. I see no context other then I mentioned.
    If he meant it otherwise I’m glad to hear it from him.

    In my (also subjective) understanding of his statement I only put my thoughts against his statement as I understand it.

    I just try to motivate the ATSB and DSTG to come foreward with all the information and data

  25. DennisW says:

    @Ge Rijn

    “I strongly disagree. By now it’s of utmost importance ALL available information on the debris (kind of damage, separation, possible flight attitudes on impact), radar data, drift data, Inmarsat data, barnacles, gets disclosed soon.”

    Of course you disagree because you are missing the point. I am speaking from the point of view of the people who have the information, not the people who could benefit from the information (us). There are courses taught in the Bay Area on how to value information, and how it should be disseminated.

    The tripartite group has much more interest in appearing to have done the right thing, than in finding the aircraft. Releasing information can only make them look bad.

  26. Ge Rijn says:

    @DennisW

    Now I understand your point of view. And it seems you could be right.
    It sure looks like it I agree.
    They could redeem themselves and proving the opposite by disclosing all the information and data at least to selected sources without taking any stand.
    Leave the NoK, IG and others to figure it out.

    Refusing in doing this calls upon them the suspucion of covering up known facts as you referre to.

    I cann’t believe most of the people involved in the official investigation (ATSB, ISDG, Boeing, Fugro, France etc.) can live with hiding any truth that could shed a clearer light on solving the mystery.
    I agree corrupt government leaders may but high educated, consious individuals in the search won’t.
    They will speak up sooner or later when the time is right.

  27. DennisW says:

    @Ge Rijn

    I hope you are correct, but I doubt we will ever see the data from the 20 previous flights of 9M-MRO. How useful would that have been to fine tune our path algorithms. We have had nothing for tuning. Only data from the accident flight, and no truth data to use to check our calculations.

    Likewise the radar data and the detailed operation of the SDU. Withholding of Shah’s simulator data, and the cell phone registration are other examples. This incident is the most obvious case of public obfuscation in my memory.

  28. Ge Rijn says:

    @DennisW

    I’m a sceptic in general but with the wishfull thinking and believe the great majority of people have a consious and find it quite difficult to bare such a truth unspoken.
    I rely in this sence on those people.
    We’ll see what comes up.

  29. Lauren H. says:

    Here is an edited portion of my 9/2/2015 post in response to an ALSM post:

    “Based on the FI actual burn rates, you believe the the right engine flamed out before the left engine. Agreed. However, without TAC wouldn’t that tend to have the plane yaw to the right and would that tend to push the plane to the right? The FCOM says that “the TAC does not fully compensate for the failed engine” and goes on to say, “following engine failure, the pilot can trim the airplane using additional rudder trim, control wheel input, aileron trim, or autopilot engagement.” It is not clear how the TAC will act if autopilot was engaged at the time of the first engine failure. How did the simulator react?

    I understand Brian’s position that the plane had to go to the left after the second engine failure since it probably could not have covered the required distance to the seventh arc if it had turned right but, what if the elapsed time between the flameouts was shorter and the airplane was flying higher and faster at the time of first engine flameout? Could it have gone to the right before it started a spiral dive?”

    While the ATSB said the right engine flameout could have occurred up to 15 minutes before the felt engine, ALSM calculated about 300 kg or so of fuel in the left fuel tank. At a guesstimated Left Engine FF of 4750 kg/hr, the Left Engine would have flamed out in about 4 minutes after the right Engine. Doesn’t that mean that an acceleration rate of -19 nm/min/min the a/c could have been still flying very fast AND while the single engine couldn’t maintain the normal cruising altitude, the a/c might have been above FL270 when the left Engine flamed out?

  30. Victor Iannello says:

    An update dated March 8, 2017, appears above.

  31. Mick Gilbert says:

    @Jerry M
    Re: Debris analysis

    Jerry, I’m not an engineer or a metallurgist but for me Item 22 – part of the right vertical stabilizer panel, is as sure a sign as you’re going to get that MH370 was not ditched. The vertical stabiliser is one of the most robust assemblies on the airplane and it would have taken a lot of energy to shatter it such that it liberated a piece of debris like Item 22. I think we all remember the images of AF447’s vertical stabiliser floating largely intact. If you look at the attempted ditching of Ethiopian 961, after the spray settles there’s the tail section sitting upright with the vertical stabiliser largely if not completely intact.
    It is difficult to conceive of a ditching scenario whereby the vertical stabiliser can be introduced into the water at sufficient speed to cause that sort of break up.
    For the engineers and metallurgists out there, I’m intrigued by the extensive fracturing and buckling of the inner skin and the relatively undamaged outer skin. I’ve wondered whether this suggests that the vertical stabiliser may have been blown apart from the inside in a similar fashion to JAL 123, the sort of thing that you might get with a near vertical entry of the airplane into the water which rapidly compressed the cabin atmosphere, blowing through the aft pressure bulkhead and venting into the vertical stabiliser.

  32. Andrew says:

    @Lauren H:

    It’s true that the “TAC does not fully compensate for the failed engine”. The reason for that is also explained in the FCOM, “so the pilot can recognise engine failure through roll/yaw cues”. The story goes that the system was so good during certification flight testing that the pilots could not determine which engine had failed solely from roll/yaw cues, because there were none. Consequently, it was ‘detuned’ to meet certification requirements.

    If the autopilot was engaged at the time of the first engine failure, the TAC will operate the same way as it would if the autopilot was disengaged, and will remove most of the asymmetric yaw with rudder. If needed, the autopilot would then apply aileron to counter any rolling tendency and maintain the aircraft’s heading.

  33. AlexSiew says:

    Thanks to Jerry for pointing out the Debris Examination Report of Feb 2017. The report jumps from Item 23 to Item 25, Item 24 is missing. Item 24 is the item found by Blaine Gibson in September 2016.

    https://drive.google.com/file/d/0B35tmLZHg1FEVDFDSWhYcUJWZ28/view

    https://thewest.com.au/news/mh370/uwa-pinpoints-mh370-crash-site-ng-b88407234z

    In the second link, Item 24 can be seen in the Channel 7 interview with Gibson.

  34. Andrew says:

    @Gysbreght:

    The ATSB did not specify the electrical configuration in question, but if loss of engine power from one engine resulted in loss of the autopilot, then I assume they considered the case where one engine generator was already inoperative prior to the engine failure. That scenario would have left the aircraft running on batteries immediately after the failure, pending deployment of the RAT and autostart of the APU.

    “What is the response of the TAC to the first flame-out?”

    Assuming the electrical configuration described above, the flight control system would have reverted to secondary mode as soon as AC power was lost following the engine failure. The TAC would not be available.

    “Does the Autothrust remain operative, i.e. does the thrust of the remaining engine increase to max CLB rating, or does it not change?”

    The L & R autothrottle servomotors are powered by 28V DC from the L & R DC buses. Assuming the electrical configuration described above, power to those buses would be lost when the first engine failed. Consequently, the autothrottle would be inoperative.

  35. Victor Iannello says:

    @Andrew said, “The ATSB did not specify the electrical configuration in question, but if loss of engine power from one engine resulted in loss of the autopilot, then I assume they considered the case where one engine generator was already inoperative prior to the engine failure.”

    I think for this to occur, both the IDG and the backup generator on that engine would have to be failed or electrically isolated, as either can supply to the transfer bus.

  36. DennisW says:

    @Victor

    “What part of the ATSB’s analysis was incorrect?”

    Nothing was wrong with the analysis. What was flawed was the decision to start an underwater search based on that analysis. You and I both know that the ISAT cannot be used to predict a terminus. Why spend 150M USD trying?

  37. Andrew says:

    @Victor said: “I think for this to occur, both the IDG and the backup generator on that engine would have to be failed or electrically isolated, as either can supply to the transfer bus.”

    Yes, that’s correct. Apologies for the oversight. The associated transfer bus would need to be electrically isolated, either by a fault or deliberately switching off the associated backup generator switch.

  38. lkr says:

    AJerry: Thanks for posting the report, good to see some analysis of the debris. We usually diagnose from cell phone pix, but never hear from hands-on expert analysis.

    “On almost all of these part they state something like: “The fracture on the laminate appears to be more likely a tension failure. The honeycomb core was intact and there was no significant crush on the honeycomb core.” — a caveat comes to mind: I would expect that “crushed” parts with damage to honeycomb would be vastlyunderrepresented — becoming waterlogged and sinking long before they make landfall in WIO.

  39. lkr says:

    per Dr. Pattiaratchi’s map: The white dots were from his 2014 forecast, which [I think] summed from sources along much of the 7th arc. It doesn’t give any new information behind his claim to have reverse-drifted to a 40km circle. And the linked story by Geoffrey Thomas gives no more details [and is compiled with very old news, like Blaine Gibson’s “burnt” fragments].

    One thing I do see on the Pattaratchi map: a hot-spot in the Comoro Islands. Hardly a tourist hotspot, but anyone planning to don BGs mantle might take a look 🙂

  40. Victor Iannello says:

    @DennisW: Yes, I agree that the satellite data alone cannot predict a crash site without additional assumptions that reduce the possible range along the 7th arc. There were two key assumptions made by the ATSB/DSTG: 1) The a priori distributions for the number and type of manoeuvers, and 2) The assumption of a level flight for each BFO set. I also think that Bobby is on to something with his concerns about fuel endurance, although his PDAs for the engines seem high, perhaps due to poorly calibrated fuel flow sensors.

    I am surprised that people continue to assign high probabilities to crash sites they propose, the most recent being Dr Pattiaratchi.

  41. DennisW says:

    @Victor

    I was baffled by the weight given to the ReUnion finding (flaperon) by the DSTG. I honestly could discern no difference in their heat maps before and after including the flaperon. It was if they were endorsing the ATSB claim that the flaperon simply reinforced the choice of the existing high priority search area.

  42. DennisW says:

    @Victor

    I too was quite taken aback by Pattiaratchi’s statements. It makes me wonder about the guy, frankly. Maybe he is old and near the end of his career? Don’t really know, but that is an astonishing statement for someone in his position to make.

  43. David says:

    @all. End of flight. The Malaysian Debris Examination Report posted by the Jerry M.
    Items 9 and 15, flaperon upper fixed panels. Of note:
    • Similar type of damage.,
    • From identical spots, either side.
    • No panels identified as adjacent recovered to date from either side.
    • On neither is there obvious forcing by a hard object. Both separated by pressure from beneath?
    • Symmetrical and simultaeous loading?

    These, having been examined, would make excellent candidates for drift experimentation, to compare with their route and timing and assess water absorption.

    I notice that debris analysis has now been assumed by Malaysia and I hope it has occurred to the CSIRO/ATSB to ask for them.

  44. David says:

    @Andrew. Yes I follow about TAC being untroubled by flameout thanks.
    Catastrophic failure can be coped with by the pilot shutting off fuel and recycling……
    I suppose if there is no pilot, such a failure can spoil your day little further.

  45. Richard says:

    FWIW

    A press article in the New Daily, Australia spotted by Duncan.

    The model puts MH370 at Longitude 96.5 East, Latitude 32.5 South, within a 40km radius, UWA oceanography professor Charitha Pattiaratchi said, north of the 25,000 square kilometre search area identified by the Australian Transport Safety Bureau (ATSB) last year.

    http://thenewdaily.com.au/news/world/2017/03/08/mh370-crash-site-possible-location/

    @Brock and I agree that a reverse drift analysis cannot be that accurate. A radius of 40 km, gives a search area of 5,000 km 2.

    My recent paper by contrast places the MH370 end point at 30°S +/- 1°S using a forward drift analysis, which is a search area of 20,000 km 2, assuming a search width of 20 NM either side of the 7th Arc.

  46. Ge Rijn says:

    @David

    I agree those parts 9 and 15 are quite significant. Now confirmed left and right wing flaperon closing panels. Also found only few weeks apart but the Mozambique panel earlier than the Madagascar panel. Which you would expect to be the other way around. But as we know the time of beaching can be quite different than the time of finding.

    Indeed very good examples to incorporate in further drift analyzis (like many other now confirmed pieces mentioned IMO).
    This could surely further refine the probable crash location.

    Important also (like you suggest more or less in your comment IMO) these parts indicate a level entrance on the water surface. Significant in this regard is the parts were not crushed internaly (like mentioned with almost all of the pieces in this report). All tension-damage (except the vertical stabilyzer piece).
    Same kind and amount of forces seem to have seperated this pieces from below.
    IMO this all clearly indicates again there was no high speed ‘vertical’ impact into the water but a relatively low speed level impact.

    Significant also in this regard is the right nose-gear door which seperated clean on the side the hinges were mounted.
    This is strange for you would expect this door would collaps inward when impacting the water and not survive without any crush damage.
    Also a clear indication IMO there was no high speed dive/impact.
    It seems even possible to me the nose gear was down on impact.
    This could explain the breaking away from the hinges.

    The report gives very important information but ofcourse raises quite some important questions again also.

    It’s still a shame though it took them so long to deliver this info and after the search has been called off.
    They could have delivered this info more than half a year ago.

    But lets stay positive. Better late than never.

  47. Gysbreght says:

    @Andrew: Thank you for your replies of March 8, 5:19 pm and 6:13 pm on the end-of-flight scenarios.

    RE “TAC does not fully compensate for the failed engine” – I would think that a time delay would enable the pilot to “recognise engine failure through roll/yaw cues”. The TAC calculates the thrust asymmetry from an engine trust setting parameter (probably EPR for the RR engine) and a look-up table. That may not fully account for all variables. Also the TAC provides only a rudder input. Maintaining track requires also a roll input. The autopilot would take care of that if engaged, but not in the abnormal electrical configuration.

    RE “What is the response of the TAC to the first flame-out?” That was my first thought too, but on reflection it may not be quite so simple, considering TAC operation in the time between flame-out and loss of electrical and hydraulic power.

    RE “Does the Autothrust remain operative, i.e. does the thrust of the remaining engine increase to max CLB rating, or does it not change?”Thank you for your explanation. Just a follow-on question: Would the autothrustb become alive again when the APU auto-starts in response to the loss of power?

  48. Ge Rijn says:

    Like to add to my previous comment some pieces could also have been additionally damaged depending on which kind of shores they beached on their way.
    Battering for some time against a rocky shore could crack aluminium and chip off edges from pieces. Sandy shores would leave them fairly unchanged.

    The vertical stabilyzer piece could have undergone such a treatment.
    Not cracking the outer carbon reinforced skin but only the aluminium inner skin. It was not found on a sandy shore like most of the other pieces.
    Just something to think about is my message with this comment.

  49. Ge Rijn says:

    @David

    To add something else regarding your previous comment on the flaperon closing panels:

    “• On neither is there obvious forcing by a hard object. Both separated by pressure from beneath?”

    Pressure must have come from beneath for the trailing edges with the rubber strips on both panels are intact. This cannot be explained by flutter seperation for the trailing edges would break/damage first and forces from above would have done the same.
    And these are also fixed panels not likely affected by flutter.

    Most logical (IMO only) explanation is they were seperated by water force coming from beneath during a reletively low speed level (small positive AoA) impact on the water surface.

    This latest debris report IMO should urge everyone to review their assumptions on the end-of-flight scenario at least.
    IMO it must have been a kind of glide and ditch.
    The flaperon closing panels among a lot of other confirmed debris parts tell no other more logical story now.
    I challenge others to come up with better explanations.

    The nose-gear door might well hold the key in this regard. Not collapsing inward but seperating from its hinges outward.
    IMO this can only mean the nose-gear door was deployed on impact and subsequently the landing gear was lowered.
    And subsequently the plane was piloted till the end the pilot tried to perform a ditch with extended landing gear.

    I know it sounds ridicules but I’m eager to hear another explanation.

  50. Ge Rijn says:

    Have to add something else.. The left wing flaperon closing panel is not the first left wing piece confirmed.
    Earlier the left wing outboard flap trailing edge piece was confirmed 9M-MRO.

  51. AlexSiew says:

    Below, the link for the report by the ATSB in September 2016 on a ‘preliminary examination’ of Item 24, the debris fragment showing clear signs of heat damage. As some have noted, the ATSB was super quick to come up with the report. This report was only upon a ‘preliminary examination’. The item must have been comprehensively examined by now, 6 months later. So why is Item 24 redacted from the latest Debris Examination Report?

    http://www.atsb.gov.au/media/5771521/debris-examination-report-4.pdf

  52. Mick Gilbert says:

    @Ge Rejin

    WRT item 18 – part of the the Right Hand Nose Gear Forward Door – how have you come to conclude that the door failed outwards rather than collapsing inwards. All we know is that the part was separated from hinge assemblies.

    And how do you account for the interior items being liberated from a relatively low speed glide and ditch? The forces required to separate the IFE frame from a seat back frame inside the airplane must have been truly extraordinary. That sort of damage, together with the damage to the vertical stabiliser, is not consistent with a low speed impact.

  53. sinux says:

    @Ge Rijn
    Your explanation ticks a lot of checkboxes. There is a lot of explaining to be done on the Nose Gear Forward Door!

    But to me it’s baffling that all fractures are tension fractures and that the fibers are pulled.
    There should be at least 50% compression fractures.

    If the items 9 and 15 where separated by upward force, logically the underside of the honeycomb should be compressed.
    Also item 7 (wing to fuselage body fairing) should be compressed.

    I can only see two possible explanations :
    1. The parts are torn off by over speed (negative pressure on the outside surfaces when approaching supersonic speed)
    2. some kind of explosive decompression (or explosion just outside the aircraft powerful enough to create an air vacuum)

    Another puzzling thing is that item 8 can’t be matched by serial number???
    “Further research and identification of the serial
    number 407 did not directly link the part to the MAS aircraft registration 9M-MRO.”
    So which aircraft is supposed to have this serial number?

    Item no 9 should have an identification plate as seen on the intact picture yet it’s missing on the debris…

  54. DennisW says:

    @Richard

    “My recent paper by contrast places the MH370 end point at 30°S +/- 1°S using a forward drift analysis, which is a search area of 20,000 km 2, assuming a search width of 20 NM either side of the 7th Arc.”

    I can’t get those numbers to work (area wise), but no matter. I am old.

  55. Ge Rijn says:

    @Mick Gilbert

    The hinge attachment on part 18 shows the panel was pulled off this hince not pressed inwards against the hince. In the latter case there would be no clean brake from this hinge and a clean side along those hinges.
    And there’s no compression damage of the honeycomb.

    As I showed more than a year ago with interior pictures from Aisiana flight 214 after it crash landed in San Francisco the interior is pretty messed up with many panels and other parts broken and seperated.
    The force required to seperate the IFE frame is little for it’s clamped on, not riveted or screwed. With some force you can pull it off with your hands.

    I agree the vertical stabilyzer piece is an enigma in this context that needs a different explanation.

  56. Ge Rijn says:

    @sinux

    Interesting you mention serialnumber ‘407’.
    As I remeber it well 9M-MRO had serial number ‘404’. Any way not ‘407’.

    Could this mean this piece is from another B777 with serial number ‘407’?
    Which one is this? Is it still flying or was it srapped?

  57. Ge Rijn says:

    @Sinux

    On the pulled fibers and lack of compression damage.
    I assume in a kind of ditching event the forces of the water coming mainly from the front direction and pushing upwards rip the pieces out backwards and upwards causing mainly pulled fibers.
    I think the upward water forces are too low in a low speed kind of ditching event to compress the honeycomb from the panels and other pieces.
    No compression damage indicates therefore IMO reletively low speed level impact.

  58. Victor Iannello says:

    @Ge Rijn: We have to be careful about distinguishing the part serial number from the aircraft serial number. We had the same confusion regarding the flaperon, which had a part serial number of 405, but could be traced back to aircraft serial number 404, which was registered as 9M-MRO.

  59. DennisW says:

    @Ge Rijn

    FWIW the rate of debris finds continues to follow my Poisson PMF calculated in 12/16. That is in any given month:

    probability of no debris finds: 0.60
    probability of one debris find: 0.30

    0.10 is the probability of more than one debris find in any given month. In this calculation only the debris that has been “confirmed” (3 pieces) and categorized as “almost certain” (7 pieces) was used.

    It might be an interesting exercise to use a Weibull distribution to try to back into the number of pieces that might be found in total eventually.

  60. Richard says:

    @Dennis

    “My recent paper by contrast places the MH370 end point at 30°S +/- 1°S using a forward drift analysis, which is a search area of 20,000 km 2, assuming a search width of 20 NM either side of the 7th Arc.”

    “I can’t get those numbers to work (area wise), but no matter. I am old.”

    I am old too!

    Would you agree with me that 29°S on the 7th Arc is around 98.9794°E?

    Is 31°S on the 7th Arc around 97.2207°E?

    Is the distance between these two points something like 279.5 km?

    If you are to search 20 NM either side of the 7th Arc, then this is a search width of 40 NM or 74.08 km.

    This makes a total area of 279.5 km x 74.08 km = 20,705.36 km 2.

    Or, where am I going wrong?

  61. Richard says:

    Charitha Pattiaratchi’s MH370 End Point at 32.5°S 96.5°E is just below Broken Ridge at a depth of 4,600m.

    There is significant upwelling in this area due to the massive change in depth from 4,600m to 1,450m.

    Global Drift Program buoys close to Broken Ridge show abnormal tracks and speeds compared with the rest of the Indian Ocean.

  62. DennisW says:

    I was using an earth radius of 6371km.

    For one degree the surface distance is 6371*pi/180 ~ 111km

    111*2*74 = 16400. Like I said, no big deal.

    And, I am older than you, I am sure of that.

    In the interest of full disclosure I was using +/- 20km instead of +/- 20NM. So I was quite wrong to begin with.

  63. DennisW says:

    @Richard

    BTW, I like your suggested area. Perhaps even a little further North.

  64. Ge Rijn says:

    @sinux @VictorI

    B777 line number ‘407’ is still flying:

    https://www.planespotters.net/airframe/Boeing/777/28532/VP-BDR-VIM-Airlines

  65. sinux says:

    @Ge Rijn
    from your link : it’s operated by VIM Airlines… a Russian company… I can see Jeff smiling 😉 But according to FR24 it’s still flying.

    I agree with Victor’s comment :
    “We have to be careful about distinguishing the part serial number from the aircraft serial number.”

    Still why can’t they link sn407 to mh370? Is the manufacturer on holiday… or did they have a fire in their warehouse 😉

  66. Brock McEwen says:

    Re: drift: none of the proposed locations explains zero debris on Oz shores. None. If you don’t believe me, ask Dr. Pattiaratchi how many “white dots” hit WA by end 2014.

    Re: R600 “special offset” of 4600: has anyone in the IG ever asked Inmarsat for the raw historical data they used to derive this value? If it should have been, say, 4500, the search box should have been 20km further SE. And the fit at 18:25 would be much better. I for one consider the comparison of a single R600 logon at the gate to its R1200 acknowledgement to be a very soft test, given the error bars on each.

  67. Victor Iannello says:

    @DennisW: Yes, 1 deg of latitude is 60 NM = 111 km. But the arc is not straight, and it is not along a constant longitude (more like 45°). Both these effects increase the area.

  68. Niels says:

    @VictorI: “There were two key assumptions made by the ATSB/DSTG: 1) The a priori distributions for the number and type of manoeuvers, and 2) The assumption of a level flight for each BFO set”. In addition: also the allowed speed changes were limited and based on prior “normal” flights, see assumption 5 p.58 of “Bayesian methods”. For a FMT before 18:40, IMO a slow down is needed somewhere between 18:22 and 19:41 to avoid ending up too far south. Not sure if the DSTG assumption would allow for this. I estimate that, typically, a 19:41 latitude north of the equator is needed to end up in the new proposed high probability area.

  69. TBill says:

    @Victor @Niels @BrianA
    Is there a logical rule or rationale why the 1941 Arc2 probably had to be approached from the inside (East) vs. from the outside (West)? I assume MH370 could have flown outside Arc2 (eg; to LAGOG) but it seems all serious paths assume MH370 came back inside Arc2 (eg; to BEDAX) before hitting the 1941 arc. I believe Brian may have recently advised us of this “rule” on another site, but what is the rationale?

  70. Victor Iannello says:

    @Niels: When I spoke of manoeuvers, I was referring to changes in speed and/or direction. Figure 7.3 of the DSTG study shows the prior distributions for number of turns and number of accelerations that were used to reconstruct flight paths. A slow down would be allowed. However, in general, to produce paths north of the search area, you need to slow down and have more turns; or loiter between 18:25 and 19:41 and continue at high speed. The distributions shown in Fig 7.3, presumable derived from previous flights, would rank these as low probabilities. That’s why the DSTG’s highest probability is associated with a straight flight at cruise speeds.

  71. Victor Iannello says:

    @TBill: If you fit the BTO values to a smooth curve and look at where the BTO value is a minimum, corresponding approximately to the closest approach to the satellite, and you assume straight flight, you can make some rules about how the plane flew relative to the 19:41 arc. Brian astutely observed this, and it helped us to visualize possible paths. However, as long as the position at 19:41 (and other handshake times) matches the BTO value within some error margin, it does not need to adhere to this rule.

  72. Gysbreght says:

    @VictorI: “If you fit the BTO values to a smooth curve and look at where the BTO value is a minimum, corresponding approximately to the closest approach to the satellite, and you assume straight flight, you can make some rules about how the plane flew relative to the 19:41 arc.”

    Implicit in fitting a smooth curve to the BFO values is the assumption of a straight path at constant groundspeed. Otherwise the curve would not be smooth but would show a ‘kink’ at each turn.

  73. Niels says:

    @VictorI: I am wondering how the DSTG algorithm would have treated the inclusion of the 18:22 radar point.

  74. Victor Iannello says:

    @Niels: Their prior was the radar point at 18:01:49 using the filtered speed and a distribution for position (SD=0.5 NM) and track (SD=1°). Essentially, they ignored the 18:22:12 position, although the track at 18:01:49 was consistent with the position at 18:22:12.

  75. DennisW says:

    @Victor

    “@DennisW: Yes, 1 deg of latitude is 60 NM = 111 km. But the arc is not straight, and it is not along a constant longitude (more like 45°). Both these effects increase the area.”

    Yeah, I know. Just being lazy.

  76. Mick Gilbert says:

    Victor,

    I have been looking at the Lido Hotel slide and have a few observations that may be of interest.

    As can be seen in the graphic (link below) there are five main clusters or segments of consecutive plots;

    • 2:00:25 – 2:01:20 MYT,
    • 2:02:08 – 2:05:15 MYT,
    • 2:13:12 – 2:15:25 MYT,
    • 2:16:25 – 2:17:10 MYT, and
    • 2:19:41 – 2:20:35 MYT

    https://www.dropbox.com/s/g0mrpu6gnyw5ck2/Lido%20slide%20data%20-%20transposed%2C%20segmented.png?dl=0

    It is very apparent that each segment records a different track for the target; starting at Segment 1 with a track of 300°M, they progress consecutively from 297°M to 289°M, 287°M and finally
    285°M. In other words, the target’s track is gradually drifting west and the rate of drift is gradually decreasing. That drift far and away exceeds what you’d expect for an airplane flying a constant heading given the conditions on the night.
    The gradually changing track is intriguing and I’d suggest that a not unreasonable conclusion is that the target was not navigating using a constant track as one might expect if the airplane was being flown in LNAV mode.

    If I run the rate of drift back to around 13:55 MYT when the target would have been south of Penang I get a starting track of around 306°. On the basis that a right base leg for Rwy 04 is 310°, 306° caught my attention. It was then pointed out to me that the STARS approach to Penang from the north east is BIDMO 1A, an approach that runs BIDMO-PUKAR-ENDOR-MEKAT-KENDI. BIDMO 1A terminates at KENDI on a heading of 310°.

    When I compare the PSR traces detailed in the FIR to a BIDMO 1A approach from near IGARI I get an not unreasonable fit (see next graphic). There is a sort of interesting alignment with the 13:52:35 MYT trace (which has no heading data detailed in the FIR but is shown/drawn to be around 262°); if MH370 was flying BIDMO 1A at cruise speed (ie diversion was programmed in LNAV but a descent wasn’t) the FMC would look to cut the ENDOR-MEKAT-KENDI corner by commanding a 25° AOB turn about 6 nm short of MEKAT. As shown on the graphic the start of that turn is roughly coincidental with the PSR trace (and if you think about what a 25°AOB turn would do to the target’s RCS for the Butterworth Approach PSR that would be when you’d expect that target to captured).

    https://www.dropbox.com/s/wvshw9awjzit2nm/BIDMO%201A%20track%20comparison.png?dl=0

    Of further interest are the facts that;

    A. if MH370 was flying BIDMO 1A, absent crew input before reaching KENDI, it would encounter an LNAV END OF ROUTE error upon reaching KENDI and default to a constant heading, and

    B. given the geometry of the ENDOR-MEKAT-KENDI corner, at cruise speed the airplane would not have reached 310° as it flew through KENDI so allowing for roll out consequent to an END OF ROUTE error its new heading would be somewhere between 300° – 310°.

    A scenario that might give rise to a diversion from near IGARI to Penang with a BIDMO 1A approach but with no alterations to cruise speed and altitude would be an inflight emergency that rapidly escalated such that crew incapacitation occured after the diversion was programmed but before a descent could be programmed/commanded.

    Now, on the basis that my initial analysis of the Lido plots was not particularly sophisticated (the “best fit” tracks for each segment were “fitted” by eye and measured to the nearest whole degree) my house of conclusions might be built on sand so I’d welcome any and all feedback.

  77. AlexSiew says:

    If I may point out another thing about the purported satellite data from Inmarsat.

    The handshakes from 1941 to 2241 UTC were all timed 1 hour apart at 1941, 2041, 2141 and 2241 UTC. This is an impossibility. The timer resolution, as Don pointed out long time ago, was 256 seconds (4 minutes and 16 seconds).

    The LGA Satcom system on 9M-MRO first logged on at 12:50:19 UTC according to the FI. It then had 3 handshakes, at 13:54, 14:50 and 15:54 UTC. The 3rd handshake is the 1st entry on the SITA Traffic Log in the FI. A few minutes later, the crew inputted flight information and the ADIRU was initialised, so the LGA Satcom system renewed its log on (now with Flight ID info) and the HGA Satcom system (which required ADIRU info for antenna pointing) logged on for the first time. See the FI.

    In the latest paper from Ian Holland, it was conceded that the AES was probably still operating at 1722 UTC (actually 17:21:-__, but rounded to 1722 by Holland) and it was noted the log-on at 1825 UTC was 63 minutes later (actually 64 minutes if not rounded up).

    The timer resolution for the Honeywell SDU on the plane in the SwissAir Flight 111, was also 256 seconds.

    http://www.collectionscanada.gc.ca/eppp-archive/100/200/301/tsb-bst/swissair-e/html/02sti/06aircraft/acars.html

  78. Mick Gilbert says:

    Further to the above, I have been trying to reconcile that Lido target’s behaviour with a known mode of autoflight without any success.

    While I hesitate to say this, the target is behaving as though it stable in flight but with no fixed heading, as though it rolled out to wings level as it passed through KENDI but that the FMC failed to lock up the final heading; its drift has been greatest when flying obliquely to the prevailing easterly and reducing as it gradually aligns its direction of travel to what then becomes a tailwind.

  79. Victor Iannello says:

    @Mick Gilbert: Thank you for posting this. As you know, I have provided you with my comments on a previous version you sent me. Rather than bias the group, I’ll refrain from commenting at this point.

  80. DennisW says:

    Alright, I was lazy and picked on Richard. Not only was I wrong, but I used the wrong parameters to come to a back of the envelope conclusion about his area. Mea culpa. Fortunately, I have no problem at all with being wrong. It happens to me a lot, and I ignore it.

    I mentioned earlier that it might be an interesting exercise to apply a Weibull distribution to the debris finds. It is very tedious, but I did it to make up for my Richard debacle.

    In any case, here it is. Probably has little to no value to add to our endeavors.

    http://tmex1.blogspot.com/2017/03/mh370-debris-weibull.html

  81. Andrew says:

    Gysbreght:

    “TAC does not fully compensate for the failed engine” – I would think that a time delay would enable the pilot to “recognise engine failure through roll/yaw cues”.

    The ‘detuning’ of the TAC is achieved by limiting the TAC authority to 60 per cent of the available rudder movement. At high thrust settings (eg on take-off), the TAC does not fully compensate the asymmetric yaw caused by an engine failure. The pilot needs to apply additional rudder, thus ensuring timely recognition of the failed engine. At lower thrust settings, (eg during high altitude climb/cruise), the TAC provides almost full compensation and little if any additional rudder input is required.

    “The TAC calculates the thrust asymmetry from an engine trust setting parameter (probably EPR for the RR engine) and a look-up table.”

    The TAC function receives thrust data for each engine from the associated engine data interface unit (EDIU). The EDIUs use N1 from the electronic engine controls (EECs) to calculate thrust, irrespective of the engine type. There is also an EPR input, used to adjust the calculated thrust for EPR/N1 variations at high thrust settings. The following paper, ‘Propulsion Aspects of the Thrust Asymmetry Compensation System (TAC) on the Boeing 777 Airplane’, describes the B777 TAC system and the method of thrust calculation:

    http://proceedings.asmedigitalcollection.asme.org/proceeding.aspx?articleid=2087860

    “Also the TAC provides only a rudder input. Maintaining track requires also a roll input. The autopilot would take care of that if engaged, but not in the abnormal electrical configuration.”

    True; however, if we assume the autopilot was engaged immediately prior to the engine failure, it should remain engaged for a short time after the failure, until the RPM drops below idle. During that time the autopilot and TAC should start to apply left rudder to correct the yaw and possibly some left flaperon/spoiler to correct the roll, assuming the right engine fails first. Those inputs would be removed when the flight control system reverts to secondary mode and the autopilot and TAC disengage.

    “Does the Autothrust remain operative, i.e. does the thrust of the remaining engine increase to max CLB rating, or does it not change?” Thank you for your explanation. Just a follow-on question: Would the autothrust become alive again when the APU auto-starts in response to the loss of power?

    Good question! I think autothrottle operation will be restored when power is restored. I’ll check that in the simulator next week.

  82. TBill says:

    @Mick Gilbert
    I also took a couple looks at the data around Penang, during Victor’s “Possible Paths…” thread:

    “I tried to work backwards with SkyVector to see what straight path connects up with @SK999’s digitized flight path from Penang to the 1802 point. What I seem to see is the straight path OPOVI IGOGU goes from south of Penang and also matches SK999’s data within a reasonable tolerance…I found some other straight paths to SAMAK and 1090E but they do not seem to hook up well to the path prior to 1802.”

    I did not consider a waypoint path around Penang, thinking it might be manually flown around Penang. Victor has previously shared a PowerPoint presentation of FSX screen shots showing informally how he thinks the navigation might have been done (including the offset at that time).

  83. TBill says:

    @Mick Gilbert
    Here is a more detailed SkyVector comparison of digitized path vs. OPOVI to IGOGU:

    051442N1002050E 051339N1001544E 051309N1001041E 051315N1000542E 051401N1000048E 051524N0995601E 051717N0995120E 051925N0994641E 052139N0994202E 052351N0993723E 052558N0993244E 052800N0992804E 052956N0992322E 053147N0991836E 053333N0991347E 053520N0990855E 053726N0990302E IGOGU OPOVI

  84. Ge Rijn says:

    On VictorI’s update-remark; “But we also have to ask ourselves: What part of the ATSB’s analysis was incorrect?” I was thinking the following:

    With the latest drift-analysis (et al @Richard, Pattiaratchi) and the ATSB finally also confirming the crash area has to be North around ~32S also considering nothing has been found in the previous calculated priority area, I suggest MH370 flew slower (and lower) between FMT and end of flight according to the ATSB’s analysis.

    While putting the FMT between 18:40 and 18:25 and in general having a good estimate of the speeds, altitudes and headings flown between IGARI and 18:25/18:40 assuming all this ATSB analysis is correct I suggest the ATSB’s analysis was incorrect assuming MH370 remained flying at the ~same speed and altitude after FMT.

    To end up north of the previous priority area MH370 had to slow down considerably after FMT to fit the ATSB’s analysis.

  85. Ge Rijn says:

    correction; …to fit the latest ATSB’s analysis (their new 25.000km2 area).

  86. Mick Gilbert says:

    @Ge Rijn

    Re the IFE frame, it’s not just getting it off the seat back, it’s getting the fabric cover off it AND getting it off the seat back AND getting it out of the airplane. The latter requires a hull breach, not exactly what you’d expect from a low speed glide and ditch.

    As to the nose gear doors, an entry into the water from just beyond vertical (ie airplane slightly on its back) would probably blow those doors out as the nose of the airplane rapidly compressed. That sort of entry angle would also impart a lot of energy onto the vertical stabiliser.

    @TBill

    I’m struck by the fact that the Lido slide data is not really supportive of a straight path away from Penang. Maybe I’m reading too much into the individual segments, maybe elements of it need to be consigned to the odd-sock drawer called “the vagaries of primary radar” but if the tracks for each of those segments is even vaguely close to accurate I think we need to re-examine our assumptions about how that airplane navigated away from Penang.

  87. Ge Rijn says:

    One option I suggested a long time ago is that the flight turned into a one engine flight just before and after FMT forcing the plane to fly slower and lower at ~25.000ft after FMT till end of flight.

  88. lkr says:

    @GeRijn: “With the latest drift-analysis (et al @Richard, Pattiaratchi) and the ATSB finally also confirming the crash area has to be North around ~32S….”

    Further evidence for going North is that Tony Abbott knows in his bones that it’s South of the searched area.

  89. Ge Rijn says:

    @Mick Gilbert

    The IFE frame has been in the water for ~2 years. It’s more likely the fabric degraded, rotted, washed away during this time except for the tiny amounts that were trapped by the cloth-hook.

    I’m not/never saying it was a low speed (Hudson-like) ditch. I’m suggesting/argumenting it was a -relatively- low speed ~level, ~horizontal impact on the water.
    I surely don’t exclude the fuselage breached at at least one point. Even during the Hudson-ditch the hull breached at the bottom. Or a door blew out which happened during the crash landing of Asiana 124 (left aft door).
    The tail also could have seperated.
    It sure must have been a violent impact anyway.

    On the nose gear door. In a high speed dive event we are talking about milli-seconds of impact time. There would be no time to blow those doors out and the reacting forces of the water on impact would exceed the force of the small amount air blown out sideways so close near the nose probably a million times. And a piece this size and so close to the nose would not occure in such an event certainly not without compression damage.

  90. Ge Rijn says:

    @Mick Gilbert @ others

    I found this picture of the Asiana 124 vertical stabilyser and tail piece.
    I think it’s interesting the vertical stabilyser shows same kind of damage at the exact location as where the vertical stabilyser piece of MH370 was located:

    http://www.airlinereporter.com/2013/07/a-few-thoughts-on-asiana-airlines-flight-214-crash-at-sfo/

  91. Mick Gilbert says:

    @Ge Rijn

    As I suspect neither of us has ever witnessed an airplane crashing into the sea we could continue to speculate about this until the airplane is found without achieving much. Regarding the reacting forces of water it is worth noting that the impacting body immediately displaces water on contact causing water to initially move away. In a near or beyond vertical impact one side of the airplane is not going to come into contact with water immediately almost regardless of the speed of impact.

    In any event, the impact needs to be able to account for the damage to the vertical stabiliser; it us difficult to envisage a low AoA, low speed impact achieving that even with a rapid hull break up – Ethiopian 961 is a pretty good proxy.

  92. Ge Rijn says:

    I add another Asiana 214 vertical stabilyser picture that shows it even more clearly:

    http://www.zimbio.com/pictures/ullgbpCtOj9/Asiana+Flight+214+Moved+Crash+Scene

  93. Ge Rijn says:

    @Mick Gilbert

    At least this Asiana 214 photos seem to prove the damage to MH370’s vertical stabilyser is possible in a low AoA and low speed crash landing event.

  94. Mick Gilbert says:

    @Ge Rijn

    At least this Asiana 214 photos seem to prove the damage to MH370’s vertical stabilyser is possible in a low AoA and low speed crash landing event.

    Maybe, if MH370 had flown into a seawall at 100 mph and had its tail assembly torn off and then rapidly rotated through 360° as it careened along a concrete and asphalt runway smashing the horizontal stabilisers off as it went. You’re looking at gouges in the vertical stabiliser off OK214 presumably from where the horizontal stabs struck it as they parted company.

    The two impacts are not analogous, not even close. If you want an analogous impact it’s “the Zulu”, Ethiopian 961.

  95. Gysbreght says:

    @Mick Gilbert:

    The PSR data shown in the DSTG book show clearly that the autopilot was not engaged or inoperative between IGARI and 18:02. You are demonstrating that it was not engaged after 18:02.

    I can’t imagine a pilot single-handed flying manually head-down busily typing on the little CDU keyboard waypoints, SLOP, holding pattern.

  96. Ge Rijn says:

    @Mick Gilbert

    Flying into a wall of sea tail first with +100mph would have a similar impact as flying into a seawall.
    The vertical stabiliser rather was struck by seperating engine/wing parts on its leading edge during the skirting over the runway.
    A similar event could have happened with MH370 IMO with the difference it would have skirted the water surface.
    It’s hard to image a horizontal stabiliser could hit that leading edge when seperating.
    Then it should move forward first and then backwards against this leading edge.
    Anyway even as this happened the Asiana 214 photos still show damage on

  97. Mick Gilbert says:

    @Gysbreght

    Could I please trouble you to explain how the PSR data shown in the DSTG book shows clearly that the autopilot was not engaged or inoperative between IGARI and 18:02? I ask because I’m of the view that LNAV was engaged from IGARI to KENDI.

  98. Ge Rijn says:

    ..the vertical stabiliser like the MH370 piece also shows is possible with low AoA and relatively low impact speed.

  99. Gysbreght says:

    @Mick Gilbert: Can you indicate an autoflight mode that would produce the speed variations shown in the DSTG report? Or the track variations?

  100. Mick Gilbert says:

    @Ge Rijn

    If it’s hard to image a horizontal stabiliser hitting the leading edge of the vertical stabiliser when seperating from the tail assembly just have a look at the OK214 crash on video – https://youtu.be/zTXDalv7kNQ

    When the tail assembly separates from the fuselage it not only rolls at least 360° around its longitudinal axis it also flips through about 360° around its lateral axis.

  101. David says:

    Ge Rijn. “Also found only few weeks apart but the Mozambique panel earlier than the Madagascar panel. Which you would expect to be the other way around. But as we know the time of beaching can be quite different than the time of finding”.

    Of similar densities, their different weights and sizes might not provide an explanation. While I suspect the Mozambique item would float more tilted, increased windage from that might be offset by increased drag underwater due to form change there.

    But since they started in the same place and the same time, why are their termini so different? I can think of explanations but none very plausible.

    As to timings, yes the Madagascar finding might have been delayed…but yet also it could have been the reverse. Irreducible randomness.

    As it stands there is no explanation for either distance and disparity other than randomness, which hardly builds confidence.

    Using empirical data from actual debris like these two will lift both the accuracy of analyses and confidence in them. At the risk of being well behind events, I will check this is in hand/mind.

  102. Mick Gilbert says:

    @Gysbreght

    Can you clarify what DSTG report/book you are referring to please?

    Heading changes in LNAV are easily explained, if the airplane was flying IGARI – BIDMO – BIDMO 1A there would be four different headings flown and two, if not three, of them line up reasonably against the PSR data shown in the FIR.

  103. Gysbreght says:

    @Mick Gilbert: I’m referring (not for the first time) to Figure 4.2 of the book “Bayesian Methods in the Search for MH370”, by DSTG authors Davey, Gordon, Holland, Rutten, Williams.

  104. Mick Gilbert says:

    @Gysbreght

    Thank you. We know that the PSR data between 17:30 – 17:52:35 is incomplete. Victor et al have provided a very detailed review of the radar data and have presented a reasonable, cogent and coherent explanation of the airplane’s track back across the Malay Peninsula. It’s wrth noting that when PSR captures are brief, the target’s speed, bearing, altitude and vertical speed are difficult to calculate and generally speed is overstated.

    The heading changes in 4.2 are not inconsistent with an IGARI – BIDMO – BIDMO 1A approach.

  105. David says:

    @Andrew. You note to Gysbreght about autopilot and TAC, “Those inputs would be removed when the flight control system reverts to secondary mode and the autopilot and TAC disengage”.

    Possibly the TAC would have been removed when it dropped to 3% of thrust, assuming this would be above idle?
    Can you say what the trim positions the control surfaces would revert to? Those remaining active would not hold to their last command?

  106. Gysbreght says:

    @Mick Gilbert: “The heading changes in 4.2 are not inconsistent with an IGARI – BIDMO – BIDMO 1A approach.”

    I was primarily referring to the speeds. How about those?

    The speed variations are likely caused by altitude variations at the rate of 111 ft per 5 kt TAS.

  107. Paul Smithson says:

    @Gysbreght. 5kt TAS change for every 111ft in altitude? That doesn’t sound right…

  108. Gysbreght says:

    @Paul Smithson: Thanks for correcting me. Lease read:
    “The speed variations are likely caused by altitude variations at the rate of 221 ft per 5 kt TAS.”

  109. Andrew says:

    @David:

    The TAC function stops being active when the thrust difference between the two engines is less than 3 percent of the maximum rated thrust. I believe the scenario we were last discussing was the one mentioned in the ATSB report ‘MH370 – Search and debris examination update’, which had ‘an electrical configuration where the loss of engine power from one engine resulted in the loss of autopilot (AP)’? The assumption was that one side of the electrical system had failed or was deliberately isolated. In that event, the failure of the first engine would cause the loss of autopilot, etc. At that stage the thrust differential would be more than 3 percent..

  110. Mick Gilbert says:

    @Gysbreght

    Re speed changes, you will have noticed that measured ground speed changes when measured heading changes and that as the rate of change of heading stabilises measured ground speed starts coming back to around 510 kts. There’s probably a story there considering we’re talking about a primary radar trying to resolve speed and heading simultaneously.

    I can’t recall reading that the DSTG concluded that the autopilot was disengaged at IGARI.

  111. Gysbreght says:

    @Mick Gilbert: “There’s probably a story there considering we’re talking about a primary radar trying to resolve speed and heading simultaneously.”

    The DSTG resolved speed and heading from the 10-second radar data they received from Malaysia: ”The radar data contains regular estimates of latitude, longitude and altitude at 10 s intervals from 16:42:27 to 18:01:49.”

    They filtered out the measurement ‘noise’:
    ”It is possible to derive the angle and speed of the ground velocity from the
    radar reports by assuming a simplified almost constant velocity model and applying
    a Kalman filter. This assumption is acceptable for the primary radar data because the
    reports are closely spaced in time. Figure 4.2 shows the derived speed and direction
    obtained from this filter.1”

    The three-sigma covariance of estimates is shown by dotted lines in Figure 4.2. Please note that the 10-second data end at 18:01:49 and the remainder is extrapolated.

    The DSTG did not address whether or not the autopilot was engaged after the turn-back.

  112. Andrew says:

    @David:

    Sorry, I failed to reply to your second question above: “Can you say what the trim positions the control surfaces would revert to? Those remaining active would not hold to their last command?”

    It’s unclear from the manuals, but according to the article posted by @buyerninety in the previous thread (http://www.flight.org/the-boeing-777-thrust-asymmetry-compensation-tac), any rudder trim applied by the TAC will be cancelled when the TAC stops working. Any rudder trim applied by the pilot will remain. For example if the TAC had applied 7 units of trim and the pilot had applied an additional 2 units, then 2 units of trim would remain after the TAC stopped working.

  113. Paul Smithson says:

    @Gysbreght. Still doesn’t sound right. More like 0.5TAS for every 221ft?

  114. Paul Smithson says:

    In the previous “Possible Paths” post, @Richard provides coordinates for the penultimate DSTG radar position at 180149.

    “…from the last “10 second” radar data point 18:01:49 at 5.6248°N 99.0482°E”

    I am surely missing something. Did they actually provide coordinates for this position and their 182212? Or is the position cited derived from an overlay or extrapolation? Please clarify.

  115. Mick Gilbert says:

    @Gysbreght

    I really don’t have a dog in this fight. I have never believed that the 10 second radar data was built up from continuous coverage radar data, rather the Malays have taken the initial SSR track and the segments of PSR data they have (which almost invariably included the Lido data) and constructed a data set that is filled with estimates.

    What flight profile in terms of heading, speed, vs and altitude have you infered from thd DSTG analysis? Apologies if you’ve provided that elsewhere but I haven’t seen it.

  116. Victor Iannello says:

    @Paul Smithson and @Gysbreght:

    Comment corrected at 8:45 am EST

    Conserving energy, the relationship between a change in altitude and speed is
    V2 = sqrt [ (V1)^2 – 22.6 Δh ], with V1 and V2 in knots, and h in feet.

    So, if V1= 500 kn, a 1000 ft climb would drop the speed to V2 = 477 kn.

  117. Ge Rijn says:

    @David

    I think the most simple and plausable explanation of the difference between the finding times of the Madagascar-panel and Mozambique-panel would be the Madagascar-panel beached maybe a few months earlier but was found a few weeks later than the Mozambique panel.
    Something like the engine-cowling piece which was found twice few months apart (different offcourse but hope you get my point).
    Finding times don’t represent beaching times well.

  118. Victor Iannello says:

    @Mick Gilbert and @Gysbreght: At the time that I wrote my paper on the radar data, the DSTG report had not yet appeared. I did note in that paper that if the plane flew at constant altitude and speed (FL340 and M0.84 was proposed), then the timestamps associated with the radar segments reported in the FI were not consistent. Months later, the DSTG report appeared, and what I attributed to timing errors was explained by large variations in speed. I still question which is the correct interpretation, and that is one of the reasons I have pressed for the raw, unfiltered data to be released.

    As for the possibility that the plane was in LNAV after IGARI, that’s not indicated by the path. When a new waypoint is entered and flown “Direct To”, then the present position becomes the waypoint used to calculate the path, and the plane will turn and fly along this path. Looking at the path after the turn is completed, it does not align with the start of the turn. The turn after IGARI was either manually flown or the plane was turned using HDG SEL (or TRK SEL) roll mode.

    As @TBill mentioned, here are some slides I put together a couple of months ago to show how the plane might have been flown, assuming M0.84 and FL340 after the turn. By setting Kota Bharu and Penang as fixes with circles, the pilot might have navigated using the MAP mode of the ND to intercept the tangents to the circles, as shown.

  119. Ge Rijn says:

    @Mick Gilbert

    Thanks for the Asiana 214 crash video. Dramatic. I did not see it before.
    Unbelievable also the whole plane cartwheeled around both axis and the hull and wings stayed together without further breaching. In fact it looks a lot like the crashing of Ethiopian flight 961 on the water which didn’t had the luck of staying together that way. But comparing the two crashes you can see there is not much difference in impacting the water or the land.

    I keep my opinion though that something similar could have happened with MH370. Even more now I’ve seen how much impact that B777 hull and wings can survive and still stay together during a crash landing.

  120. David says:

    @Andrew. Thank you for your response. I was alluding in my second query to you saying that autopilot inputs would be removed when the autopilot disengaged.

    I had it in mind that when the RAT came on line the then active PCUs would respond to their last autopilot command, the left and centre PFCs having remained on line, battery powered, as would the L1 and Centre ACEs.

    However with the second engine still running, more hydraulics would have been available, the range of PCUs remaining active being widened, not that this affects the issue.

    So my question was, if the autopilot input was removed, what trim positions would these various control surfaces end up in, the autopilot’s PFC/ACE command being the last, supposing no other input? I think you are saying they would neutralise? I have been unable to find a manual statement either way.

  121. Gysbreght says:

    @Paul Smithson: At 500 kt TAS it is definitely 5 kt TAS for every 221 ft of altitude change.

    @ Victor:

    For thrust equal to drag the sum of potential and kinetic energy is constant and is given by:
    m*g*h + 0.5*m*V^2 = constant,
    where: m = mass; g = acceleration of gravity; h = geopotential height; V = speed

    Differentiation gives: dh/dV = – V/g

    I.o.w. the exchange rate of speed and altitude at constant total energy is proportional to speed.

  122. Victor Iannello says:

    @Gysbreght: You said, “The speed variations are likely caused by altitude variations at the rate of 221 ft per 5 kt TAS”. That looks right if you are linearizing around 500 kn.

  123. Paul Smithson says:

    @Victor & all. You can see from attached .kmz that the turnback is reproduced reasonably well by a speed reduction (in this model to M0.78) and 25 AOB turn on to reciprocal heading. This model “starts the turn” to produce an endpoint coinciding with first radar position. As the kmz shows, the model also shows excellent agreement with subsequent radar positions. The turn is to “reciprocal course” (180 degrees from IGARI-BITOD), then a course adjustment “left” at 172830. Model uses 1700Z GDAS wind and temp [adjusted for 35000ft).

    path model from start of the IGARI-BITOD turn at 172015:
    https://www.dropbox.com/s/2dieskbgvx7uw7n/IGARI%20turn%20172015%20turnback%20M0.78%20to%20173445.kmz?dl=0

    figure 2.1 representation of the “turnback”
    https://www.dropbox.com/s/tskaeg9077r1wyj/Fig%202.1%20turnback%20representation.kmz?dl=0

    primary radar positions/timestamps
    https://www.dropbox.com/s/pndxnznkl3crki2/Primary%20Radar%201730-1735%20Fig%201.1E%26F.kmz?dl=0

  124. Paul Smithson says:

    @Gysbreght @VI. I must have misunderstood. Thought you were talking about delta TAS for same M with change of 221ft altitude.

  125. Andrew says:

    @David:

    When the autopilot is engaged, the three AFDCs send command signals to the PFCs. The PFCs then process the autopilot commands according to the control laws and produce surface commands, which are sent to the ACEs. The ACEs relay the surface commands to the respective PCUs to move the surfaces (Note: the autopilot only commands the rudder during auto land operations).

    If the autopilot is disengaged or failed, the PFCs won’t receive any command signals from the AFDCs. Any previously applied inputs from the AFDCs are ignored and the control surfaces return to their neutral positions.

  126. Brock McEwen says:

    Re: has IG ever sought raw data behind R600 “4600” bias offset: hearing nothing, I’m glad I emailed Chris Ashton on March 7, seeking details. Here’s its text:

    “Dear Mr. Ashton,

    I hope this note finds you, and finds you well.

    I am writing to enquire more deeply about the following statement from p.7 of your Oct. 2014 paper, ‘The Search for MH370’:

    ‘Each power up sequence starts with a Logon Request message that has been found to have a fixed offset of 4600 μs relative to the LOI message exchange by inspecting historical data for this aircraft terminal.’

    1) How many historical data points contributed to this 4600 value?

    2) Was the statistic developed by subtracting R600 BTOs from adjacent R1200 BTOs? If not, what method was used?

    3) How were observations aggregated: mean, median, or mode?

    4) What was the 4600 value prior to rounding?

    [snipped: why I care]

    As you know, this 4600 estimate has the potential to affect key MH370 path considerations (speed/heading/change in heading near NW tip of Sumatra, and of course the position of ‘Arc 7′).

    Profuse thanks in advance for any time and consideration you can give to these questions – it is deeply appreciated.

    Kindest regards,
    [snipped: who I am]”

    I’ll keep this group posted on Chris’ response.

  127. Paul Smithson says:

    @Brock. Thanks for that. I had asked the same question of ALSM without obtaining clarify on the questions that you pose above. I look forward to hearing the response.

  128. DennisW says:

    @Brock

    My guess is that you will not get a response. I have a couple of questions posed to the ATSB/DSTG that have neither been answered nor acknowledged.

    Even the statement you reference is completely lame in my opinion:

    ‘Each power up sequence starts with a Logon Request message that has been found to have a fixed offset of 4600 μs relative to the LOI message exchange by inspecting historical data for this aircraft terminal.’

    Is the 4600 us unique to this terminal or is it a design issue which is applicable to all terminals of this type? Why were other terminals of the same type not examined extensively? When I read something like the quoted statement above, my regard for the information source goes to virtually zero.

  129. Victor Iannello says:

    @Paul Smithson: It looks as though you chose to ignore the speed profile published in the DSTG report. That is not a criticism as much as an observation.

  130. TBill says:

    @Mick Gilbert
    Of course, my straight line path from Penang was accepting Victor’s premise that the Lido points after 1802 are questionable. I was also accepting Richard’s definition of the 1802 location, which seems to be in good agreement with SK999’s digitized path.

    As far as a Penang Runway approach theory, I would defer to Victor I am sure. But my understanding is that the data seems to support fairly constant approx. 0.84 Mach and approx. FL350. I must admit my FS9 use related to MH370 paths has not included regular landing practice, but wouldn’t there be a slow down and altitude drop for a suggested landing so close to Penang?

  131. Paul Smithson says:

    @Victor. For the purpose of these radar track path models my objective is to find that speed that offers best fit to all of the observations/timestamps available from Fig 1.1E, 1.1F and (around Penang) RMP. I find that a constant M profile can “reproduce” the radar observation timestamps with surprising fidelity. If the speed was truly as variable as DSTG “filtered speed/heading” shows, then we should not able to do this without finding serious outliers. As you know, I found that the speed that I have inferred for turnback doesn’t work beyond about 1735 but a new constant M fits nicely from 1738 through to 1752.

  132. Ge Rijn says:

    IMO the track around the South of Penang cannot be right if the co-pilot’s cellphone was only registered by station BBFARLIM2 and not by any, more Southern station as the RMP states.
    If the RMP is correct in this regard it’s impossible MH370 turned South over sea around Penang. IMO it then must have turned crossing the island over land.
    You seem to ignore this factual data.

  133. Victor Iannello says:

    @TBill: A landing from FL350 would require that the Top-of-Descent (T/D) is around 115 NM from the airport for an efficient descent. (A steeper descent is possible at higher speeds and/or by employing spoilers.) That puts the T/D around 15 NM past Kota Bharu. After the arriving airport is selected along with the STAR and ILS04 approach, the entire descent, approach, and landing could occur automatically by selecting a 3000 ft altitude, selecting LNAV/VNAV, and capturing the ILS signal after the last turn to the runway. It seems odd to me that a pilot wishing to land at Penang wouldn’t have initiated the soon after passing Kota Bharu.

  134. Victor Iannello says:

    @TBill: You don’t have to defer to me. I am not immune from making mistakes, especially on a forum like this where we sometimes toss out preliminary ideas.

  135. Ge Rijn says:

    I think it’s also wise to consider the Lido radar tracks are not showing the actual primary radar screen(s) they were captured from.
    They are projected over another screen with additional data.
    We don’t know how precise the fit is.

  136. Brian Anderson says:

    @TBill,
    “Is there a logical rule or rationale why the 1941 Arc2 probably had to be approached from the inside. . . ”

    The logic is that all BTOs were decreasing over time to the 19:41 BTO, and then increasing over time from 20:41.

    Actually the point was first illustrated when a very fuzzy chart showing elevation angles to the aircraft was shown at the Lido to NOK. From this chart the IG was able to have a first shot at constructing the “ping rings”, calculating from the elevation angles. The elevation angles reached a maximum between 19:41 and 20:41.

    By fitting a smooth curve to the BTO points, or rather the calculated Line of Sight distance to the satellite, one can estimate the point of closed approach to the satellite [at approximately 19:52]. Then by making some reasonable assumptions it is possible, with a little spherical geometry, to estimate the speed of the aircraft a this time too. [Estimated ground speed approximately 494 knots].

  137. Mick Gilbert says:

    @Gysbreght

    Thank you for those files. That altitude speed plot is particularly interesting to me as I have hypothesised that, as part of an unfolding inflight emergency, the autothrottle may have disengaged due to an autothrottle servo motor failure prior to FLCH being used to command a descent. That would see the airplane in “speed-on-elevator” mode; with a fixed thrust setting you might expect to see some speed-altitude exchanges in a relatively small band, particularly when the airplane is making heading changes. The longer term impact of the fixed thrust setting is of course a very gradual climb as fuel is burned and the airplane pitches up to hold speed.

    @Victor

    Thank you for that explanation and for the slides. An initial turn back using MCP track/heading select followed shortly thereafter by a diversion using LNAV alternate airport would be entirely consistent with an inflight emergency. As Andrew has explained to me when the requirement for a diversion becomes apparent the first priority is to point the airplane in the general direction of the alternate and then fine tune the diversion. That would have been very much the case for MH370 as it was flying away from its nearest suitable and acceptable diversion airports; every minute it kept flying on its planned route was an extra minute it had to fly back.

    Comparing the track from near IGARI back to BIDMO for a BIDMO 1A approach to the primary PSR returns recorded for 01:36 –
    01:36:40 MYT, 01:39:59 MYT and 01:52:35 MYT,
    yields a not unreasonable alignment for the joining leg to BIDMO and the BIDMO-PUKAR legs. While there is no heading data for the 01:52:35 MYT
    PSR trace it is represented in the FIR with a heading of about 262°.
    The geometry of the final turn for BIDMO 1A, ENDOR-MEKAT-KENDI,
    is such that the at cruise speed the FMC would command a 25° bank right turn about 6 nm short of MEKAT in order to intercept KENDI. That turn is roughly coincidental with the 01:52:35 MYT PSR trace (and if the Butterworth Approach PSR was ever gong to track MH370 it would be when it rolled into full profile and turned and tracked perpendicular to it). Given the location and size of the track changes I don’t think those alignments are likely to be coincidental.

    @Ge Rijn

    You’re welcome, it’s a testimony to how Boeing build their airplanes that the hull stayed relatively intact after that sort of impact.

    @TBill

    Thank you for those Skyvector plots earlier and the other information.

    Regarding a BIDMO 1A approach flown at FL350 and M0.84, when BIDMO 1A is selected in LNAV as the approach the FMC expects that the crew will manage speed and descents in VNAV or FLCH. In the absence of crew inputs (perhaps because they have been incapacitated by an inflight emergency) LNAV just sticks to the programmed route which only really becomes tricky at the final turn – ENDOR-MEKAT-KENDI. Given the airplane’s speed the FMC would need to command an early turn and “cut the corner” in order to intercept KENDI. Having done so the FMC registers an LNAV END OF ROUTE error at KENDI, rolls wings level and defaults to HDG HOLD – the airplane flys away from Penang on a heading of somewhere between 300°-310° and up the Strait.

    @Paul Smithson

    Thank you for those turnback illustrations. I don’t think that what you have plotted is inconsistent with an emergency diversion that starts with an MCP HDG/TRK SEL and then changes to an LNAV Alternate Airport diversion via the BIDMO 1A STARS.

  138. Brock McEwen says:

    @TBill: re: why from the inside: I need a thing to be intuitive before I can understand it. Here’s the thought experiment which helped me understand:

    Start at Arc 6, and fly backwards in time along a hypothetical path. Choose a speed and heading that allows you to intersect each arc (5, then 4,…) without ever having to change either speed or heading. You will find that cruising speed accomplishes this.

    If you keep on that same speed and northbound heading “after” the 20:41 arc, then exactly an hour “later”, you will cross the 19:41 Arc. And you will cross it “outside->in” (remember, we’re going backwards in time). Since you intersect it at a very sharp angle, you will, if you keep going, intersect it again – this time inside->out. However, this would take far too long – meaning you’d have to either average an unrealistic speed over the 19:41-20:41 leg, or change direction shortly after 19:41.

    Note this “rule” flows from a principle of minimum path deviation, which is made reasonable by how perfectly the last 6 BTOs align. There is certainly nothing in the BTO or BFO values at 19:41 which PRECLUDES and outside->in crossing (I’m going forward in time, now). But if you did, you not only now need two turns (one to get back toward a straight-line path through the remaining arcs, and another to then STAY on it), but the two turns would have to be a of a degree and timing that they happened, by fluke, to lay down a BTO track indistinguishable from “straight and fast throughout”.

    In short: anyone who doesn’t like coincidences should favour “inside->out” at 19:41.

  139. Andrew says:

    @Mick,

    Regarding your reply to @TBill above:

    1. There would need to be a route discontinuity between KENDI and the first waypoint of the approach (CF04??).

    2. The FMS alerting message would be ‘DISCONTINUITY’, not ‘END OF ROUTE’.

    3. The AFDS mode would remain ‘LNAV’, but the aircraft would maintain a constant heading after KENDI.

    For what it’s worth!!

  140. Mick Gilbert says:

    @Andrew

    Thanks for that, always worth it!

  141. Brian Anderson says:

    re: the Fuzzy Chart of Elevation Angles

    Before any BTO data were available, certain information was presented to a briefing of the Chinese families of MH370 passengers, at the Lido Hotel in Beijing, on about 28 April 2014. One of the slides presented became known as the ‘Fuzzy Chart of Elevation Angles’.

    Who prepared this chart, and where did the source data come from are still unanswered questions.

  142. Victor Iannello says:

    @Andrew and @Mick: The waypoints for the BIDM1A STAR to ILS04 approach for WMKP is BIDMO-PUKAR-ENDOR-MEKAT-KENDI (discontinuity) CF04-FF04-RW04. If KENDI is chosen as the transition, there is no discontinuity, but there is a hold at KENDI. I noticed the discontinuity earlier today and was intrigued so I ran a test case in FSX at FL340, M0.84. The heading at the discontinuity at KENDI is around 304°, which does not match the Lido Hotel data. The LNAV path does not match the primary radar path except that both fly past ENDOR. My conclusion after looking at the LNAV path is that there was no match. I am curious as to why you (Mick) are so convinced that there is a match.

  143. David says:

    @Andrew and also for @Gysbreght. On autopilot disengagement, “Any previously applied inputs from the AFDCs are ignored and the control surfaces return to their neutral positions.”

    Thanks Andrew that is what I was about.
    Let’s see if I have this summary right in that light, to assist with Boeing simulations’ comprehension:

    Case 1. First engine failure leads to AC power loss. At a first engine failure (say right) at cruise a yaw-correcting level of TAC will be applied, the autopilot commanding heading holding bank (but no rudder trim), TAC and bank then increasing as the engine runs down. These will be removed/‘neutralised’ as that engine drops beneath idle and the FCS goes to secondary. The auto-throttle might have accelerated the left engine while it remained operative, that engine above idle.

    The aircraft at TAC removal and autopilot disengagement will yaw hard right, rolling to the right as a secondary effect of that. The yaw and roll will increase as the residual thrust of the right engine decays further. The left flaperon will not float up with both PCUs in bypass since the ACE controlling its outer PCU will remain powered and the PCU will continue to receive hydraulics from the left system.

    Incidentally the PMGs from the left engine will be on line still, powering the left and centre PSAs.

    Case 2. Second engine failure leads to AC loss. With the second (say left) engine failure causing FCS reversion to secondary and loss of TAC and autopilot input, the sequence will be:
    • At earlier right engine failure, TAC is applied as per Case 1 and increasing (incrementally and without delay I assume) as thrust decays to zero and the left engine accelerates under auto-throttle. Increasing autopilot roll commands hold heading.
    • Left engine fails, TAC reduces progressively and autopilot holds heading until that engine reaches 3% MRT when TAC is removed, aircraft yawing a little right, autopilot applying counter roll to hold heading. (assumes idle residual thrust is < 3%MRT).
    • Left engine drops below idle, yawing reduces but on autopilot disengagement a residual amount of yaw will lead to a roll right tendency.
    This may be countered by left flaperon float as in case 1 whenever hydraulic supply to the left system fails, when it will float to 10˚ up. When and if that occurs with engine run down to windmilling is uncertain. The right flaperon neutral I think will be 2˚ or more TED.

    Do you both agree? I see why simulations are needed but wonder about their programming for all of this and the data sources for it.

    The outcome is that with these engine failure scenarios the aircraft could only roll left when left system hydraulics failed. As I see it, that is hardly consistent with the Boeing simulations, so something is out of kilter still.

    That aside, it would be nice to know the times to idle from fuel cut at different altitudes, aircraft speeds and engine speeds. I have not seen any.

  144. Mick Gilbert says:

    @Gysbreght

    I have taken your DSTG Speed Altitude graph and overlaid the heading changes associated with WMKK-IGARI-BIDMO-PUKAR-ENDOR-MEKAT-KENDI. If you were navigating along that route with the autothrottle disengaged (ie fixed thrust) in “speed-on-elevator” mode I suspect that you might see something like that where some altitude is lost in the turn which pushes speed up with the FMC then trying to re-establish speed.

    https://www.dropbox.com/s/5gs60ivbgsakk1d/DSTG_spd_alt_hdg_chg.png?dl=0

    @Victor

    If you look at the average rate of drift in evidence between the segments in the Lido slide and run that back to Penang you get 305.9° so I’ll happily take 304° as an initial heading.

    I think that there is an alignment because if you initiated an turn back towards Penang at IGARI using MCP HDG SEL and shortly thereafter executed a diversion using a BIDMO 1A approach your track to BIDMO would be around 237-239°; that aligns in both heading and location with the 1736 UTC [0136 MYT] to 1736:40 UTC [0136:40 MYT] PSR trace with a heading of 237°. BIDMO to PUKAR is 247°; that broadly aligns in both heading and location with the 1739:59 UTC [0139:59 MYT] PSR trace with a heading of 244°. Then there’s the 01:52:35 MYT PSR trace which is roughly coincidental with the early turn for MEKAT-KENDI.

  145. TBill says:

    @BrianA @BrockM @VictorI
    Thank you for the explanation of BTO implications on Arc2 path beginnings. I run test paths on FS9. The utility FlightSimCommander hooks up FS9 with Google Earth so I can see the ping rings.

  146. Andrew says:

    @David:

    I agree, although In Case 2 I doubt there’d be much, if any, residual yaw by the time the left engine drops below idle.

    “The outcome is that with these engine failure scenarios the aircraft could only roll left when left system hydraulics failed. As I see it, that is hardly consistent with the Boeing simulations, so something is out of kilter still.”

    Case 1 results in the aircraft yawing and rolling to the right after the right engine fails, probably resulting in a spiral dive to the right. The left flaperon up-float following the subsequent failure of the left engine might provide some correcting force, but probably not nearly enough to stop the right-hand spiral. Is that not consistent with the end-of-flight simulations mentioned in the ATSB’s ‘MH370 – Search and debris examination update’ report, where the aircraft that started with an abnormal electrical configuration ‘descended in both clockwise and anti-clockwise directions’? I’ll have to think more about how to explain the anti-clockwise direction observed in some of those simulations.

  147. DrB says:

    @Brock,

    You said: “In short: anyone who doesn’t like coincidences should favour “inside->out” at 19:41.”

    I don’t like coincidences either, but you don’t have to have one to approach the 19:41 arc from the outside. You only need the wind to change direction from east to west when flying a constant heading.

  148. David says:

    @Andrew. What I meant was that in 8 of 10 simulations the aircraft rolled left and these cases prima facie have them all rolling right. If the left hydraulics dropped out at idle like the electrics and 8 simulations were Case 2, the left flaperon float could resolve that leaving the 2 rights to Case 1, the left system there obviously continuing. However that brings with it a roll not a bank, continuing until the APU or windmilling speed increase intervene.

    Having looked again I find no clue as to at what pressure, flow rate and N3 (I think) RPM the engine pump supply would be insufficient.

    I learned just now that aside from the RAT needing 115 knots to supply to its rated capacity (there has been conjecture about that), the AMM says, “When the RAT is extended and hydraulics off, the airplane rolls left. Two or three units of right control wheel rotation are necessary to hold the wings level” Its initial extension is hydraulics off for spin up.

    However I do not follow how it would contribute anything worth a mention if hydraulics are off for just a couple of seconds spin up, unless they mean RAT deployment for other than hydraulic requirements.

  149. DrB says:

    @Andrew,

    Thank you for responding to my previous questions. I have been on the road for a week and am now back home and trying to catch up.

    Let me be more specific about the post-diversion altitude choice:

    1. In this situation, would you always change from an odd FL to an even FL when turning back to a westward course?

    2. If you decided to change to an even FL, would you go up to FL360 or down to FL340?

    It is possible that a speed higher than LRC but lower than MMO was used from ~17:24 to ~18:40. I don’t think M0.87 (MMO) is acceptable because it uses too much extra fuel compared to LRC. In addition, it is difficult to reconcile the military radar positions with an air speed that high. However, something in between M0.84-M0.87 can be consistent with both fuel and observed ground speeds. For instance, a CI = 500 gives ~M0.851 and a Fuel Flow ~4.3% higher than at LRC. This gives a slightly better ground speed match than LRC (depending on the altitude assumed). It also is a better match to the engine PDAs in cruise because it reduces the difference between predicted and observed endurance from ~1% to ~0.1%. It is possible that this 1% difference is simply due to calibration errors in the Fuel Flow sensors or the fuel tank mass readings. It’s hard to say if reducing the predicted difference below 1% implies better agreement or is just an exercise in numerology, but you may be able to shed some light on this question.

    3. In a diversion scenario, would you ever choose a speed between LRC and MMO?

    If so, when and why would you do this? How would you go about setting the speed? It is easier than setting LRC? Would you set it using the MCP or the FMC? If using the FMC-CDU, would you ever just put in a very large Cost Index value (like 500 or 1,000)? Is there another way to set speeds above LRC using the FMC (or must that be done using the MCP)? Would you set a high KIAS instead? 330?

    Another scenario is that perhaps the APU could have been turned on near 17:24 if a drive generator was lost. If the APU ran until the time when the SDU power was restored ~18:22, and then the APU was turned off, that would use about 0.4% of the total fuel available after 17:07 (bringing the “discrepancy” down from 1.0% to 0.6%).

    4. Are there any circumstances you can envision (perhaps left IDG failure?) where you would turn on the APU even if the backup generator took over, and then turn the APU offline ~1 hour later after power was restored (to the SDU + ?)?

  150. buyerninety says:

    @Andrew
    There would not need to be a Route Discontinuity after KENDI if another
    previously flown stored route (or canned stored route) had KENDI added
    to one of its LEG pages, e.g. (WMKK/VABB) ..GUNIP KENDI VAMPI.etc.. and
    then that previous route had been made the active route.

  151. Gysbreght says:

    @Mick Gilbert: “If you were navigating along that route with the autothrottle disengaged (ie fixed thrust) in “speed-on-elevator” mode I suspect that you might see something like that …”

    The performance of an autopilot in “speed-on-elevator” mode is much better than the DSTG data indicate. The observed variations approach or exceed the acceptable performance criteria (+/- 5 knots) that a line pilot must be able to demonstrate each year in recurrent training.

  152. Andrew says:

    @buyerninety:

    True; there are any number of ways the FMC could have been programmed. My comment was related to Mick Gilbert’s scenario, where a crew programmed a diversion to WMKP after suffering some kind of emergency. In that scenario, a route discontinuity after KENDI would have the aircraft continuing on a heading of about 304°, as Victor discovered.

  153. buyerninety says:

    @Andrew said;
    “a route discontinuity after KENDI would have the aircraft continuing on a heading of about 304°”
    …yes..but we know it did not do that – it actually made a heading
    for VAMPI-MEKAR..

  154. buyerninety says:

    So my point is that I am suggesting a programming of the FMC as I have outlined
    matches with the route as far as it is known.

  155. buyerninety says:

    Also, in regard to Mick Gilbert’s scenario, if a crew member wished to program
    the FMC to Go Direct To KENDI (or closer waypoints) and wished the autopilot to
    navigate the aircraft (say, so they could attend to a pressing emergency), then
    it would be necessary to have an Active Route programmed, would it not?
    Without an Active Route (when autopilot is on), the error mesage ‘Not On Intercept
    Heading’ would be displayed, wouldn’t it? Therefore, an Active Route would allow
    the crew member to set the heading, have the autopilot on, and press LNAV (in this
    case, LNAV is therefore ‘armed’) so that when the aircraft intercepts the active
    route, then LNAV will engage and the autopilot will LNAV the aircraft thereafter.
    (You could allow an alternate to the scenario, that not only KENDI, but some other
    previous along way waypoint(s) could be added, so in fact the autopilot could be
    LNAVing the aircraft prior to KENDI, to KENDI.)

  156. Ge Rijn says:

    @all

    Like to mention that debris item 8 from the report is identified as a piece of the No. 1 outboard flap track fairing. This is the first (outer) outboard flap fairing on the left wing.
    The earlier identified flap fairing piece was from flap track fairing No. 7 on the right wing.

    https://www.atsb.gov.au/media/5770117/debris-examination-mh370_19april2016.pdf

    This could be important for it further indicates the possibility of a ‘wings-level’ horizontal entry in the water.

  157. Ge Rijn says:

    To add; this makes 3 identified left wing trailing edge related pieces versus 4 identified right wing trailing edge related pieces.

  158. Andrew says:

    @DrB,

    1. In this situation, would you always change from an odd FL to an even FL when turning back to a westward course?

    In general, I would say yes; the default semi-circular rule that is used worldwide has aircraft flying tracks between 0° and 179° at odd FL, and tracks between 180° and 359° at even FL. Additionally, in RVSM airspace, crews are expected to fly at a level 500 ft different from those normally assigned by ATC in cases where an ATC clearance is not available. That said, in an emergency a pilot is able to vary the rules if required for the safety of the aircraft.

    2. If you decided to change to an even FL, would you go up to FL360 or down to FL340?

    It would depend on the situation. If I were anticipating a landing within the next hour then it would probably make more sense to descend to FL340 rather than climb. However, if I needed to fly a long distance before landing and wanted to conserve fuel, then it would make more sense to climb, provided the higher level was close to the optimum level for the aircraft’s weight. In the MH370 case, the higher level would have been closer to the optimum.

    3. In a diversion scenario, would you ever choose a speed between LRC and MMO?

    Possibly, but again, it would depend on the situation. Broadly speaking, there are three speeds that might be used for a diversion:

    – MMO: If the aircraft needs to land ASAP, then a speed close to MMO would be preferable, assuming there is no structural damage that precludes increasing speed.
    – CI = 0/LRC: If the aircraft needs to fly a long distance before landing and the crew needs to conserve fuel, then Cost Index 0 would be the speed of choice. However, in that situation LRC provides a faster cruise speed for only a 1% fuel penalty. According to Boeing, LRC in the B777 equates to a Cost Index of about 180.
    – Normal speed: If neither of the above scenarios apply, the crew might elect to divert using the originally flight planned Cost Index. In the MH370 case the planned Cost Index was 52, which equates to approximately M0.82.

    In the B777, LRC is around M0.84, while MMO is M0.87. A crew might choose to fly at say M0.85-0.86 if they wanted to fly faster but did not want to sit right on the limit. The easiest way of selecting a different speed it to set it on the MCP. Selecting LRC is by no means difficult, but it does require several key presses to execute via the FMC CDU. It is certainly possible to select a higher Cost Index, but the easiest way of setting a specific Mach no. is to do it via the MCP or by entering the Mach no. in the FMC CDU. At high level you would set a Mach no. rather than KIAS.

    4. Are there any circumstances you can envision (perhaps left IDG failure?) where you would turn on the APU even if the backup generator took over, and then turn the APU offline ~1 hour later after power was restored (to the SDU + ?)?

    The checklist for a generator failure requires the crew to start the APU in cases where the generator can’t be reset. The APU generator then provides an additional power source. Once started, the APU would normally be left running for the duration of the flight.

    If the crew needed to isolate the left side of the electrical system for some reason (??) they would turn off the main and backup generators and select the bus tie off. In the process, power would be removed from the SDU. The resulting ELEC GEN OFF checklist would prompt them to start the APU. If, at a later time the crew was able to repower the left side electrics (??), they might then turn off the APU. Off the top of my head though, I can’t think why the crew might isolate the left side of the electrical system only to repower it at a later time.

  159. Andrew says:

    @David,

    Sorry, but I’ve run out of time to reply to your post and I’ll be flying all day tomorrow. I’ll try to put something together late tomorrow or the next day.

  160. Gysbreght says:

    @Andrew: RE your reply to DrB at 6:47 am, point 3:

    If you wanted to fly near maximum speed, perhaps you could descend to the Vmo/Mmo cross-over altitude at FL305?

  161. Andrew says:

    @Gysbreght:

    Yes, that would give you the best possible TAS, as I mentioned in a previous post. The resulting groundspeed would depend on the wind – it might be better to stay at the higher level if there was a favourable tailwind, but that probably wasn’t a factor with MH370.

  162. Victor Iannello says:

    @Andrew & @Mick: Here is a plot that shows MH370’s path (calculated by integrating the speed and track data from the DSTG report) shown along with some waypoints and airway B219. Based on this, I conclude that MH370 was not in LNAV mode following waypoints. It certainly wasn’t following the BIDM1A STAR (RW04), which would be BIDMO-PUKAR-ENDOR-MEKAT-KENDI.

  163. Andrew says:

    Thanks Victor.

  164. buyerninety says:

    There may be other waypoints that have not been considered.
    For instance, MH370 appears to have passed over or almost over (fly-by) KP406,
    (KP406 for the ‘RNAV (RNP) Z 22 (AR)’ approach, as seen on that chart;)
    http://vatmy.org/charts/airport/wmkp.pdf

    KP406 as a waypoint would allow a crew member (possibly anticipating/hoping for late
    contact with WMKP) to be on a heading to hit KP406 and then make a runway 22 (from
    north) landing at WMKP – alternatly, if no runway 22 landing is advised/directed, then
    flying on by KP406 thereafter to KENDI (or on to ENDOR & THEN KENDI) for runway 04.

    I realize this is a ‘quick draft’ picture you have made, which is why ENDOR is misplaced
    as over land, rather than offshore.

    (Unfortunately the route as known relies upon what are, as someone else put it,’cartoon’
    drawings. I would add “& marked by hand drawing”. Therefore, I must reserve judgement
    until the NOK are successful and actually obtain the full radar dataset {and part of
    that is from an uncalibrated radar…}).

  165. buyerninety says:

    @Andrew
    1.) When winds data is loaded or uploaded into the 777 FMC, do you know if those
    forecasts contain data for whole of world, or is it something like a region
    subset of data (for areas that the aircraft is planned to fly through)?
    2.) When you received your flightplans from the Operations Centre for input into
    the FMC, did you ever notice if areas outside the flight area were populated with
    wind data, or did you notice if wind data for such areas showed nothing or ‘zero’.

  166. Victor Iannello says:

    @buyerninety said, “I realize this is a ‘quick draft’ picture you have made, which is why ENDOR is misplaced as over land, rather than offshore.”

    No, in the image, ENDOR should be precisely placed at N5°14.69′ E100°24.47′ and over water. Perhaps you are confusing the shore line with the shallow water surrounding land.

    @buyerninety said, “Unfortunately the route as known relies upon what are, as someone else put it, ’cartoon’ drawings. I would add “& marked by hand drawing”.

    No, as I said, the route I show is based on numerically integrating the speed and track data in the DSTG report.

    @buyerninety said, “For instance, MH370 appears to have passed over or almost over (fly-by) KP406, (KP406 for the ‘RNAV (RNP) Z 22 (AR)’ approach, as seen on that chart;)

    No, the path came nowhere close to KP406.

    You may argue that the radar data we have is not precise enough to determine whether or not there was an attempt to approach WMKP. What I am questioning is the way in which the radar data we have is being used as evidence that there was an attempt to approach WMKP. I see no such evidence, but I am trying to keep an open mind.

  167. buyerninety says:

    Probably,the representation of land and water in that picture is less than ideal.

    What specific data set in that report are you referencing?

    Kindly state how far you measure, in your choice of km or Nm, the
    flightpath is from KP406?

  168. TBill says:

    @Victor
    I was also expecting MH370 ATSB path to be 6 nM offshore Penang. I guess you are saying the colors on Google Earth look different than SkyVector.

  169. DrB says:

    @Andrew,

    Thanks for answering my questions regarding turn-back speed and altitude. That is helpful information.

    If you have the opportunity to evaluate the SLOP/FMT scenario, here is what I would suggest you consider doing as an experiment.

    From my perspective, the two most important results are:

    (1) a determination of whether the EOR heading mode is Constant True Heading or Constant Magnetic Heading, and

    (2) how the FMC makes the 116 degree turn around IGOGU with a 15 NM R offset.

    Proposed experiment:

    a. Input parameters: B777-200ER, Trent 892 Engines, PDA = 4.7%, Wind = 15 everywhere from the East, SAT = ISA + 10C, FL360, LRC, NORM/TRUE = NORM, dry W = 174.369 MT, Total W = 209.9 MT at start at MEKAR (34.66 MT fuel at MEKAR).

    b. Begin active route with N571 fly-by waypoints: MEKAR to NILAM to IGOGU to LAGOG etc.

    c. Replace waypoints after IGOGU with ANOKO.

    d. Add Hold at ANOKO (at 180 degrees orientation of pattern).

    e. Start the simulation at MEKAR and record the time or make the elapsed time there equal to zero (this occurred at ~18:21:01 for MH370).

    f. At 1:11 (1 m 11 s) after MEKAR you will be at 18:22:12 position of 10 NM past MEKAR on N571.

    g. At 2:23 execute a 15 NM Right SLOP (~18:23:24 for MH370). This is the last pilot action.

    h. Record time when SLOP is completed and track is again at 296T (~5:41 after MEKAR and ~18:26:42 for MH370).

    i. Record time when FMT begins around IGOGU (~17:34 after MEKAR and ~18:38:35 for MH370).

    j. Note maximum bank angle during FMT. Is the (first) turn at 13 deg AOB or 65 deg or ??

    k. Does the track stabilize near 200 degrees or does it continually decrease from 296 T to 180T with no pause at an intermediate value? In other words, is there one continuous turn or are there two discrete turns? If there is a second turn, what is its AOB and what was the true track between the two turns? If possible save the lat/lon versus time for the whole route so we can understand the turn overshoot if any.

    l. Record the time when the FMT is complete and the heading is 180T toward the fix 15 NM west of ANOKO (~22:27 at the latest but probably sooner; not later than 18:43:28 for MH370).

    m. Record the time when the air speed begins dropping from LRC to best Holding (~256.2 KIAS); I would guess this occurs between about 20:31 and 21:31 after MEKAR (18:41:32 – 18:42:32 for MH370).

    n. Record the time when the END OF OFFSET message appears suggesting the offset be set to zero (~21:31 or 18:42:32 for MH370). Do not zero the offset.

    o. If possible, record the time when Best Holding speed is achieved.

    p. Record the time when the END OF ROUTE message appears (~23:31 after MEKAR or 18:44:32 for MH370).

    q. After the EOR error occurs, the FMC will begin a constant heading mode at the fix 15 NM west of ANOKO.

    r. The initial magnetic declination will be 1.1 degrees W of N (or -1.1 deg E of N).

    s. To discern between CTH and CMH it is necessary to simulate a lengthy route. After another hour, at ~01:20:00 (that’s 1 hr 20 m after MEKAR, or 19:41 for MH370) the mag dec is -1.5 degrees EofN and after 2 hours the mag dec is -2.6 deg EofN. So you have to go to 2:20:00 to get 1.5 degrees of change (from -1.1 to -2.6) in the mag dec (and 3:20:00 is even better with 3.5 deg of mag dec change). That 1.5 degrees should be (barely) discernible on the MCP, and the time required will be less if you can get the heading information displayed somewhere at 0.1 degree resolution instead of 1 degree as on the MCP (I don’t know if this is possible). One of the two headings selected by the NORM/TRUE switch will stay fixed with time and one will change.

    t. At this point, after the EOR guidance begins, I think you can toggle the NORM/TRUE switch to read each heading value. That tells you which is being held constant – TRUE or MAGNETIC, and this answers the first question above.

    u. It would be great if there was a way to read both the true and magnetic headings without toggling the NORM/TRUE switch. That is because supposedly Honeywell says that switch determines the guidance mode after an EOR error. I don’t believe it, but if this is correct, then toggling the switch might possibly change the guidance mode even at times AFTER the EOR error. Sounds crazy to me, but who knows? Is there a way to read both true and magnetic heading besides the MCP? If so, then I suppose the best thing to do is to leave the NORM/TRUE switch always in NORM. If not, one could just leave the NORM/TRUE switch in NORM and observe the displayed heading. If it never changes, that means the guidance is CMH. If it changes, that means the guidance is CTH. If it were me, the first time I tried this I would not toggle the swtich (just to make sure the guidance mode was not changed) and see if the displayed heading changed after 2-3 hours. Then I would change the switch to TRUE and see what the TH is. The starting and ending NORM headings and the ending TRUE heading are sufficient to tell us whether the guidance is CTH or CMH.

    v. If there is a way to “speed up time” in the Level D simulator after the constant heading route begins, that would be a great convenience. An alternative to the lengthy method described above is to do a second EOR experiment in a region where the mag dec changes much faster with location, but this involves setting up a similar second EOR scenario and may not save any time in the end.

    w. BTW, I would be happy to contribute some funds if needed to cover the simulator rental costs for this experiment. Let me know if I can help in this way.

  170. DrB says:

    @buyerninety,

    The winds at 1200 UTC don’t affect the MH370 route, so I don’t understand your comparison to the winds at 1800 UTC. My wind model interpolates between the 3-hour spacing of the GDAS data as well as interpolating in pressure level. Yes, the winds do change with time, and this does have an effect on the route. That’s why I use a 4-D wind model.

  171. DrB says:

    @Kirill Prostyakov,

    Do you have (links to) visible satellite imagery at 34S, 94E near 0000 UTC on 3 March 2014?

    Do you have HA08 data at expected arrival time assuming impact near that location at 00:21:00?

    Thanks, Kirill.

  172. Victor Iannello says:

    @buyerninety: You discussed making a landing at WMKP on RW22 from the north. I mistakenly referred to KP416 on the map, which is about 9 NM from the radar path, and used for approaches to RW22 from the north. KP406 is indeed within 0.5 NM from the radar path. But in the plate you reference, that waypoint (KP406) is for arrivals from waypoint NURLA from the south. I don’t see the logic in choosing this waypoint from the north.

    As for the data set I am referencing, it is Fig 4.2–the speed and track plots from the DSTG report.

    @TBill: I have two references for the position of waypoint ENDOR. One is the navigational data for WMKP as extracted from AIRAC Cycle 1309, which puts ENDOR at N5°14.74998′ E100° 24.40002′. The other is the position from Skyvector, which puts ENDOR at N5°14.69′ E100°24.47′. The two positions are about 0.1 NM apart. I am fairly confident that ENDOR is plotted correctly in the Google Earth screenshot.

  173. TBill says:

    @DrB
    If you follow @Kirill Prostyakov recent links/references it takes you to this information/satellite photos in the SIO by Tim Vasquez, meteorologist
    http://www.weathergraphics.com/malaysia/iozooms.shtml

    Not sure if that helps. BTW I looked at straight path Penang to IGOGU vs. following waypoints on N571 and I only saw 1nM difference on SkyVector. Could be some round-off error in there.

    @Buyerninety
    The wind maps are sure impressive in the SIO below 20S. It strikes me that the high winds seem to be consistent with the heavy cloud cover that sets in around 22S or so. I am wondering if the high winds could impact BTO/BFO enough to make a judgement if MH370 actually entered that high wind zone or not.

  174. sk999 says:

    What if one ignores all data from before 19:41? What I did was to start at 7 degrees N latitude, then run constant Mach routes in magnetric track, true track, and LNAV (great circle) modes, allowing the start time, start longitude, start heading, and mach all be free parameters. I also allowed a variable head/tail wind on top of the GDAS model that was allowed to change once per hour. The start times range from 18:31 for magtrack to 18:55 for lnav, start longitudes 94.4 to 94.1 (S to SW of ANOKO), and the endpoint latitudes are -34 to -36.5. The parameters are all highly correlated, so the errors on any parameter are fairly large (e.g., 1-sigma on end latitude is 1 degree). Still, the results are very similar to what one gets trying to construct an integrated route fitting all the data.

  175. Victor Iannello says:

    @sk999: For each of the three navigational modes, can you describe what your optimization criteria were? Did you study different altitudes, or did you find that TAS is relatively independent of altitude?

  176. DennisW says:

    @sk999

    “What if one ignores all data from before 19:41?”

    That has been my recommendation for some time, and I start all my flight models at 19:41. The only optimizations I use are the BTO values. I think single digit BFO errors are perfectly acceptable.

  177. TBill says:

    @sk999 @DennisW
    I agree starting from 1941 is attractive approach. But wasn’t that the original ATSB approach that gave us a more northerly search zone? SK999, your technique sounds rigorous, but why would it differ from ATSB analysis starting from 1941? The criticism was that starting from 1941 discounted the 1840 phone call which suggests Southerly flight direction by 1840 (which of course is in question).

  178. buyerninety says:

    @VictorI
    I admit, we are only ‘kicking the can’ in this matter of waypoints. Given the definitve
    proof is encoded on a piece of metal, silently corroding, km’s deep on the SIO seafloor,
    we’ll probably never know for sure about waypoints pre-KENDI.

    @TBill
    Yes, Google Earth in closeup shows ENDOR as being in the (Goole Earth) representation of shallow water, as VictorI said.
    Another way to get location data from Skyvector is to zoom in as much as possible,
    input (e.g.) ENDOR and click ‘Go’. Then click ‘link’, and the URL you see contains a
    numerical locational reference that you can strip from the URL & input into Google
    Earth.

    @DrB
    I have the view that MH370 was following a stored route retrieved from the FMC, either a
    canned representative stored flightplan or a previous flightplan stored ‘as flown’. If a
    previous ‘as flown’ flightplan, then the winds from that previous plan’s date may be the
    winds that the FMC was using instead of current date winds – (Andrew may be able to tell
    us if previous route plans, when made the current Active Route, use any previous stored
    winds).
    Incidently, I noticed you asked Kirill Prostyakov, about imagery 0000 UTC on 3 March 2014
    – did you mean 0000 UTC on 8th March 2014? (If so, VictorI may be able to correct that).

  179. sk999 says:

    The optimization was a standard least squares minimization, with BTO errors of 21 microsec, BFO errors of 2.8 hz, and N/S wind errors of 4 knots rms. I kept the altitude fixed at 31,000 feet. Increasing it to, say, 35000 feet will increase the Mach and push all other parameters around. For true track, the difference was minimal. For magnetic track, it pushed the final latitude North by 1.7 degrees, but the fit was worse. I have not tried to let the altitude be a free parameter. I did not get results that converged with the hold modes, so perhaps those were the modes that the ATSB was able to find routes that went far North before turning.

  180. DennisW says:

    @TBill

    The ASTB and others had routes at a 19:41 latitude at or slightly below the equator. Starting farther North at 19:41 leads to terminal locations farther to the North.

  181. Andrew says:

    @buyerninety:

    1.) When winds data is loaded or uploaded into the 777 FMC, do you know if those
    forecasts contain data for whole of world, or is it something like a region
    subset of data (for areas that the aircraft is planned to fly through)?

    Due to memory limitations, the FMC can only store forecast wind/temperature data at each waypoint of the active flight plan. It can’t store data for the whole world, or indeed a whole region. In-flight, the FMC mixes the actual wind with the forecast wind to determine a predicted wind ahead of the aircraft.

    2.) When you received your flightplans from the Operations Centre for input into
    the FMC, did you ever notice if areas outside the flight area were populated with
    wind data, or did you notice if wind data for such areas showed nothing or ‘zero’.

    The computer flight plans include wind/temperature data at each waypoint along the route. The only off-route wind/temperature data we have available is that contained in the forecast charts issued by the met people.

    (Andrew may be able to tell
    us if previous route plans, when made the current Active Route, use any previous stored
    winds).

    No! The only stored flight plans that can be entered are the standard company routes in the database. The FMC does not ‘remember’ previous routes flown. No wind/temperature data is held over from previous flights.

  182. Andrew says:

    @DrB:

    Thank you for your suggestions. The aircraft will be a 777-200 with RR Trent 872 engines, so some of the performance data will be different to the 777-200ER powered by the Trent 892. Nevertheless, the FMC/AFDS behaviour should be the same. Time will be the enemy – I will need to use the ground speed multiplier to speed things up at times when there is nothing much happening, so the elapsed times won’t be representative.

  183. Mick Gilbert says:

    @Victor

    Thank you for the primary radar data graphic. I note that the 10 second trace appears to have smoothed the original PSR trace data out in both bearing and location. I have overlaid the FIR traces to highlight the variations.

    https://www.dropbox.com/s/y3e5tqzobfxs7o9/2017-03-11%20Path%20to%20Penang%20w%20FIR%20PSR.png?dl=0

    If we leave aside the approach to Penang and return to the Lido slide (https://www.dropbox.com/s/g0mrpu6gnyw5ck2/Lido%20slide%20data%20-%20transposed%2C%20segmented.png?dl=0) we see a target that is drifting in heading through 15° in 20 minutes. That does not appear to be a target that is navigating with any sense of precision.

  184. buyerninety says:

    @Andrew
    Thankyou, so only a number of standard routes would be stored in the FMC database.
    Would ‘standard’ (expected) heights & speeds be stored in the LEGS of those standard
    stored routes, for certain of the waypoints e.g.;
    (at) waypoint CCCCC-470TAS 29000 feet, DDDDD, EEEEE-480TAS 33000 feet, FFFFF etc. ?

  185. buyerninety says:

    @Andrew
    Are Discontinuities also stored in the standard stored routes, between
    say, waypoints where a discontinuity might normally be desired in the
    route?

  186. buyerninety says:

    @Andrew
    Sorry, last but most interesting, question;
    A 777 is LNAVing along a route NILAM-IGOGU-LAGOG .
    By intention or mistake, the route is altered, such that it is now
    NILAM-IGOGU-IGOGU-Discontinuity-(either LAGOG or someother waypoint or nothing).

    If this sequence is possible, I assume after the first IGOGU, the FMC will
    complain with ‘Not On Intercept Course’ or like message, because the ‘second’
    IGOGU is now 180 degress rearward.

    Consider a similar scenario, but now with a 2 to 4 Nm right Offset in operation.
    After the first IGOGU+’offset to right’ is sequenced, the second IGOGU+’offset to
    right’ is situated 90 or less degrees to the left of the aircrafts current heading
    and location (the second offset is seen by the FMC as rotated 90 degrees
    counterclockwise around IGOGU, when current location is taken as reference).

    X2
    |
    IGOGU-X1 .like this, (X=offset position)

    ……▲
    ……|
    ……Initial direction of travel (offset track)

    At this instant, shouldn’t the second IGOGU+offset to right’ be
    within the FMC’s allowable capture range?
    Considering there’s a Discontinuity after the second IGOGU+’offset to right’, and
    therefore no more route to sequence, wouldn’t the FMC attempt to turn the aircraft
    left, and fly-by the second IGOGU+’offset to right’? (thereafter adopting the usual
    Discontinuity behaviour i.e TH or MagH).

    Is such a scenario possible?

  187. buyerninety says:

    (@Andrew Edit to above – consider the offset type as SLOP, input to FMC as per
    usual route SLOP programming.)

  188. Mick Gilbert says:

    @Gysbreght

    You wrote; The performance of an autopilot in “speed-on-elevator” mode is much better than the DSTG data indicate.

    Under normal circumstances I’d agree but I’d like to see how an airplane where a descent had been commanded in FLCH with the autothrottle disengaged would handle through turns; the autopilot would not be looking to hold altitude, just the opposite, it would be waiting for an opportunity to descend. Turns, where speed starts to wash off due to the fixed thrust setting would present that opportunity, the autopilot would set up a descending turn. Because the MCP target speed is a Mach number we’ll have a steadily increasing target speed as we descend. On rolling out of the turn the pitch down would translate into speed above target and the autopilot would command a pitch up to get it back to target speed.

    You would expect to see a descent-ascent oscillation associated with every turn and the amplitude of the oscillation should be proportional to a combination of the time taken to complete the turn and the angle of bank. I don’t know if it’s possible to model something like that in FSX.

    If it’s not something along those lines then are we looking at the airplane being flown by hand? Off the top of my heart, there would be only one reason for doing that; the PFCSs reverting to secondary or direct mode.

  189. Mick Gilbert says:

    @Gysbreght

    Mmmm … that would be top of my head, not heart. Damn you, autocorrect!

  190. DrB says:

    @buyerninety,

    Yes, I meant to type 8 March.

    AFAIK, the only thing the FMC does with current/forecast wind data is to make a very tiny correction the the desired Mach when in ECON mode for the expected head-wind/tail-wind. I am not aware of any mechanism by which its knowledge of cross-wind would be used to adjust the lateral navigation in any way.

  191. DrB says:

    @Andrew,

    Whatever you can do with the simulator in the time available will be much appreciated I am sure.

    Can you get a listing of position, altitude, speed, etc. versus time from the run? That would be extremely helpful. Perhaps useful timing information might still be gleaned if the ground speed multipliers were known and time corrections were to be applied after the fact.

    Thanks again and good luck!

  192. DrB says:

    @Mick Gilbert,

    In the two overlays you linked, the tracks are labelled as magnetic “headings”.

    Are these really “headings”, or are they magnetic “tracks”?

    Maybe I missed something, but I don’t see any evidence of cross-wind data being used to compute the actual aircraft headings from the measured track bearings.

  193. Mick Gilbert says:

    @DrB

    You’re dead right, they are magnetic tracks. And no, you haven’t missed anything, I don’t have the detailed cross-wind data but based on the forecast for FL340 for that evening you can add 1.0°- 1.5° to the tracks to derive the headings.

    Whatever which way, the Lido slide seems to show a target that is drifting in both heading and track by around 15° in 20 minutes. That is not what you’d expect from an airplane being flown in LNAV or on constant true tracks. Further, when you look at the target’s behaviour between VAMPI and MEKAR it does not appear to be turning right toward NILAM rather it is drifting left and further off N571.

  194. DrB says:

    @sk999,
    @DennisW,

    In the early days I did a lot of fitting starting at 19:41. It had the advantage of simplicity and allowed one to ignore the 18:22-18:40 data, which was not understood. It was also an efficient means of developing and refining models and fitting tools without long computer run times.

    Since then I have started fitting at 17:21. Regardless of method, at some point you still have to demonstrate that a post-19:41 path is (1) reachable by a continuous, flyable connecting path that is (2) consistent with the BTO/BFO/radar data prior to 19:41 and (3) there is enough fuel to fly until 00:17.

  195. DrB says:

    @Mick Gilbert,

    You said: “Whatever which way, the Lido slide seems to show a target that is drifting in both heading and track by around 15° in 20 minutes. That is not what you’d expect from an airplane being flown in LNAV or on constant true tracks. Further, when you look at the target’s behaviour between VAMPI and MEKAR it does not appear to be turning right toward NILAM rather it is drifting left and further off N571.”

    Yes, a lot of us have noticed that the Lido 18:22:12 position and the ~18:21 position are both south of N571. It does not appear to show the slight right turn needed at MEKAR toward NILAM that would be expected if the LNAV were following N571.

    I will also note that a direct path from VAMPI to ANOKO is within 1 degree of the VAMPI to MEKAR track. I think the radar data after 18:13 would fit a straight VAMPI to ANOKO leg (not using N571 at all) at least as well if not better than it fits N571. By the way, the times you have listed on the Lido overlay are not UTC as labelled. Local Malaysia time?

    Maybe that alignment is just a coincidence. I can’t see a way to get a due south track in the vicinity of ANOKO without going to IGOGU first. I suppose you could go VAMPI to ANOKO to NOPEK but why would anyone go to NOPEK? That doesn’t make any sense to me. At least there is some logic to going to ANOKO (ANOKO2C STAR to WITT).

  196. Mick Gilbert says:

    @DrB

    Yes, it should be MYT not UTC. Lido presents us with quite a confusing picture. I’ve not been an advocate for MH370 navigating in any sophisticated manner post the IGARI turnback but Lido shows a lack of sophistication so complete I can’t fathom what we’re seeing in practical terms. It is as though we have a target that must be stable in pitch, roll and speed that is holding to no fixed heading or track.

  197. Gysbreght says:

    @Mick Gilbert: ”If it’s not something along those lines then are we looking at the airplane being flown by hand? Off the top of my heart, there would be only one reason for doing that; the PFCSs reverting to secondary or direct mode.”

    It’s not something along those lines. We are we looking at the airplane being flown by hand.
    Either a hand on the yoke controlling the pitch with throttles fixed, or a hand on the throttles with the autopilot controlling the pitch, or one hand on the yoke and the other hand on the throttles.

    There could have been several reasons for doing that, one of which you mentioned.

  198. paul smithson says:

    Question. If aircraft slowed down could radar be “interpreting” this as change of heading rather than change of speed? That could “bend” apparent course to the left post VAMPI when in fact it continued straight at reduced speed?

  199. Gysbreght says:

    @paul smithson: Every ten seconds the radar receives an echo from the target that provides range, azimuth and elevation. So each echo defines latitude, longitude and altitude at a certain time. Those data were provided to the DSTG by the Malaysian authorities. The distance between two latitude/longitude positions divided by the time difference defines the speed between those positions.

  200. paul smithson says:

    @G. Radar clearly does NOT provide unambiguous position, heading and speed during transitions because it “assumes” perperpetuation of prior track until it snaps on to new one. Besides, DSTG were given only a SINGLE position after 180149.

  201. Gysbreght says:

    @paul smithson: Source (1st sentence)? Sure, a SINGLE position does not contain speed information.

  202. paul smithson says:

    I’d rather the radar experts comment, hence my original post was a question.

  203. Andrew says:

    @buyerninety:

    Would ‘standard’ (expected) heights & speeds be stored in the LEGS of those standard
    stored routes, for certain of the waypoints e.g.;
    (at) waypoint CCCCC-470TAS 29000 feet, DDDDD, EEEEE-480TAS 33000 feet, FFFFF etc. ?

    No, there are no standard altitudes or speeds associated with the stored company routes. The only altitudes & speeds that are stored in the database are those associated with the SIDs and STARs for each airport, but those are separate to the company routes.

    Are Discontinuities also stored in the standard stored routes, between
    say, waypoints where a discontinuity might normally be desired in the
    route?

    Not that I’ve ever seen, but I suppose it could be done. Route discontinuities aren’t normally ‘desired’ and are easily created by the pilot if needed.

    A 777 is LNAVing along a route NILAM-IGOGU-LAGOG .
    By intention or mistake, the route is altered, such that it is now
    NILAM-IGOGU-IGOGU-Discontinuity-(either LAGOG or some other waypoint or nothing).

    It’s not possible to have the same waypoint appear twice in sequence; there would need to be a discontinuity between them.

    Regarding your second scenario, I assume you mean something like this:

    NILAM – IGOGU – IGOGU – XXXXX

    where there is a track change of 90 degrees or less at IGOGU and an offset to the right? As before, you can’t have two waypoints of the same name in the LEGS page without a route discontinuity between them. In that case, the aircraft would maintain the LNAV track until it reached the route discontinuity and thereafter maintain the current heading. It would not turn to flyover the second IGOGU offset.

  204. Andrew says:

    @David:

    First, I don’t think it’s possible to say the eight simulations that rolled left were all Case 2. Indeed, the ATSB report “MH370 – Search and debris examination update’ doesn’t mention the Case 2 scenario at all, so we don’t know if it was tested again during these latest simulations. The report does state that, in an abnormal electrical configuration, ‘the aircraft descended in both clockwise and anti-clockwise directions’, so some of the eight simulations that rolled left must have been Case 1.

    As I said previously, in Case 2 ‘I doubt there would be much, if any, residual yaw by the time the left engine drops below idle’, at which point the flight control system reverts to secondary mode and the autopilot disengages. The left flaperon upfloat, together with the left rolling tendency at RAT deployment that you mentioned would then cause the aircraft to roll to the left. That would be consistent with the ATSB’s original simulations, where ‘in each test case, the aircraft began turning to the left and remained in a banked turn’.

  205. buyerninety says:

    @Paul Smithson
    You have no need of a radar expert. Part of the radar track is known to be incorrect
    due to an ‘uncalibrated’ radar.
    Comnparisons between the SSR track (given by the aircraft itself) and later PSR radar
    plots? Apples and Oranges.
    Also, you should keep in mind the experiences of the pilots of the Embraer Legacy 600
    business jet involved in the Gol Transportes Aéreos Flight 1907. Those pilots almost
    ended up in jail because the Primary Surveillance Radar showed their aircraft
    making excusions outside the flight envelope that the pilots swore they kept within.
    Fast forward to a check of the Embraer FDR, and sure enough, that PSR simply wasn’t as
    consistently accurate as was believed (and there was no suggestion that PSR was
    uncalibrated).
    Also, if one of the MH370 GPS units was non-functional for whatever reason, that would
    impact the navigation performance of the aircraft. Not necessarily in a significant
    manner, but enough to put it a handful of Nautical Miles off the track.
    And by MEKAR, how many Nautical Miles had the aircraft been without a winds update?
    (A bit interesting to me that the winds forecast, when I look at the 1200Z forecast,
    to notice that the winds at 1800Z aren’t appreciably different from those shown in
    the 1200Z forecast, except – the winds are different about the region of MEKAR/NILAM.)
    https://earth.nullschool.net/#2014/03/07/1200Z/wind/isobaric/250hPa/orthographic=-258.66,6.27,3000
    https://earth.nullschool.net/#2014/03/07/1800Z/wind/isobaric/250hPa/orthographic=-258.66,6.27,3000

    All that said, yes ,there is a kink in the track from Khota Bharu to Penang – result
    of concatenating the returns from two different radars (one ‘uncalibrated’) ? Possibly.
    A 18:22 single point plot, whose originator may have been at maximum range?
    Other plots, cluttered about the VAMPI -MEKAR line, which appear to have a snake
    undulation – but which basically still follow the N571 track (assuming there’re real..).
    All too tenuous.

  206. buyerninety says:

    @DrB
    You may not have taken on board a paragraph in the document I listed here;
    http://mh370.radiantphysics.com/2017/03/02/radar-maybe-captured-fighter-jet-chasing-mh370/#comment-848
    “Updating FMC Winds
    The winds provided to the FMC through ACARS uplink are updated at Nav Services every 6
    hours. The updates can take up to an hour and should be complete by 0600z, 1200z, 1800z
    and 2400z.”

    Wait… that’s not the interesting paragraph – this is it;
    “FMC Winds – Forecast vs Actual”
    …”When using wind in the calculation, the FMC uses a mixture of 99% IRS Wind (Actual)
    and washes it out over approximately 600nm to 99% Forecast wind. This washing process is
    not linear – at approximately 200 miles, less than half the actual wind is used in the
    calculation.
    As soon as a higher level is being considered by the FMC – only forecast wind is used.
    The FMC does not wash actual wind up or down through levels.”

  207. buyerninety says:

    @Andrew
    Thankyou, your responses are very helpful and much appreciated

    @VictorI
    DrB said, on March 11, 2017 at 11:51 am:
    “Andrew,” …”From my perspective, the two most important results are:”
    “(1)…”
    “(2) how the FMC makes the 116 degree turn around IGOGU with a 15 NM R offset.”

    VictorI, as DrB intends to pursue this further, it may be helpful to carry out another
    test, per the png picture. It occured to me that (regarding the MCP 34000 feet
    setting);
    NILAM and IGOGU are on N571, that air corridor has altitude constraints of 46000 to
    27500 feet.
    ANOKO is on B466, that part of the air corridor ANOKO is on has altitude constraints of
    27500 to 8500 feet.
    Elsewhere, I have read comments re; VNAV ALT (seen on the MPD) e.g.;
    “VNAV ALT {annunciaton} means that there’s a conflict between the profile and the MCP
    altitude”.

    It may be appropriate to test if the FMC may be treating the waypoints as having altitude
    restrictions compliant with the altitude restrictions of their associated air corridors.

    It would be interesting to run the test with same set aircraft parameters in the FMC,
    at a similar set of waypoints – but without a mismatch in altitude limits.
    This may allow to see if any altitude restrictions are affecting the seen represented
    flightpath on the display. May also save Andrew some time.

    Waypoints MABLI-SUSAR-MUMSO may be appropriate for such a test.
    MABLI and SUSAR are on L635, that air corridor has altitude constraints of 46000 to
    24000 feet.
    MUMSO is on N892, that air corridor has altitude constraints of 46000 to 13500 feet.
    (MUMSO is also on N875, that air corridor has altitude constraints of 46000 to 24500
    feet.)
    Obviously, 34000 feet is within the limits (& so there is no question as to whether the FMC
    could be planning the flightpath so as to allow for a descent to keep within altitude limits).
    Distance SUSAR to MUMSO is 21.8 Nm. Angle change MABLI-SUSAR-MUMSO is 121 degrees.
    https://fpl-1.caasaim.gov.sg/aip/2017-03-10/final/2017-03-02-Non-AIRAC/html/eAIP/ENR-3.3-en-GB.html#RTE-L635
    https://fpl-1.caasaim.gov.sg/aip/2017-03-10/final/2017-03-02-Non-AIRAC/html/eAIP/ENR-3.3-en-GB.html#RTE-N892
    https://fpl-1.caasaim.gov.sg/aip/2017-03-10/final/2017-03-02-Non-AIRAC/html/eAIP/ENR-3.3-en-GB.html#RTE-N875
    Cheers

  208. Mick Gilbert says:

    @Gysbreght

    Re: It’s not something along those lines.

    I’m not so sure. If you plot Mach number (that’s the MCP target speed that is going to be flown to in “speed-on-elevator”) instead of TAS those ostensibly wild variations in speed flatten out considerably. Then your left with a picture showing variability in a relatively tight range on both Mach number and altitude with almost all variation in altitude on the downside, which is pretty much what you’d expect to see if a descent had been commanded using FLCH.

  209. Gysbreght says:

    @buyerninety: “Part of the radar track is known to be incorrect
    due to an ‘uncalibrated’ radar.”

    When early reports appeared in the media that radar had reported altitudes in excess of FL410, the experts cautioned that the radar reports of altitude were not reliable because the radar had not recently been calibrated for that purpose. I don’t think those comments reflect on the accuracy of the radar measurement of latitude and longitude.

  210. Gysbreght says:

    @Mick Gilbert: “If you plot Mach number (that’s the MCP target speed that is going to be flown to in “speed-on-elevator”) instead of TAS those ostensibly wild variations in speed flatten out considerably.”No, the wild variations in speed will not flatten out considerably. The TAS is what it is. If the AP “speed-on-elevator” maintained a constant Mach, why would there be such altitude variations?

    Firstly, I can’t plot Mach number because I don’t know the altitudes. As I said, the speed variations could have been produced by large throttle movements with AP in ALT mode.
    Secondly, it wouldn’t make much difference. The TAS at constant Mach reduces by 2 kt/1000 ft.

  211. Victor Iannello says:

    @Mick and @Gysbreght: On a FLCH descent (irrespective of thrust), the speed window would change from Mach number to IAS at 310 kn. Even still, I don’t understand how an A/P mode that controls speed (like FLCH) can explain the large speed variations presented by the DSTG, even with turns and with no A/T.

    Of course, I still question the validity of the speed data presented by the DSTG. At some point, I’ll have more to say about this.

  212. sk999 says:

    buyerninety writes: “… the SSR track (given by the aircraft itself) …”

    The SSR track is normally determined the old-fashioned way, by the radar itself. The initial version of the SSR protocol predates the availability of GPS positions onboard aircraft, and not all aircraft are equipped with GPS. Position information is broadcast from the aircraft only via ADS-B (at least, that is my understanding from reading the SSR manuals and writing a program to decode SSR packets.) From FI, we know that HCM was monitoring ADS-B broadcasts; how that information is integrated with SSR-derived positions is something I don’t know.

    Wind information used for navigation is derived by the aircraft itself and, as we see from MH370, can be downloaded to the ground in ACARS messages. In fact, data from such ACARS messages are used as input to the global forecast models. (The paragraphs you cite from the V Australia flight training manual refer to inputs needed for the calculation of step climbs, not lateral navigation.) Even without GPS (highly unlikely), the ADIRU alone can derive wind speeds, albeit with lower accuracy.

  213. DrB says:

    @buyerninety,

    You seem to be saying, if I understand you correctly, that an erroneous wind forecast will affect the FMC LNAV and therefore the actual track flown by 9M-MRO. That is not the case. The only impact might be a +/-0.001 Mach adjustment to air speed in the ECON mode.

  214. DrB says:

    @Victor,

    You said: ” Even still, I don’t understand how an A/P mode that controls speed (like FLCH) can explain the large speed variations presented by the DSTG, even with turns and with no A/T.

    Of course, I still question the validity of the speed data presented by the DSTG.”

    Right on. Small positional errors cause large apparent speed errors over small baselines.

    For instance, let us compute how accurate each of two position measurements by a range/bearing radar have to be in order to estimate the speed with a measurement error of +/-20 knots. At 500 knots the separation of two measured locations that are taken 10 seconds apart is only 1.39 NM. Each position must have an error of only 0.039 NM in order to achieve a 20 knot error in the calculated speed (and a 2.3 degree error in estimated bearing). Clearly this is more than an order of magnitude better than the scatter shown in the Lido radar plot, so achieving useful speed and bearing estimates over a 10-second period is impossible.

    How about over 1 minute? Using just two points 60 seconds apart the errors are now relaxed by a factor of 60/10 = 6. Now you need positional errors of 0.23 NM or better in one dimension. That too is also clearly beyond the capability of the radars plotted in the LIDO slide. So you cant’t get to +/- 20 knots or +/- 2 degrees over one minute. Fitting all the data instead of just the two end points helps a bit, but more so for the average position than it does for the speed and bearing calculations which both depend heavily on the end points.

    Let’s try 10 minutes of data. Now you only need 2.3 NM positional error (in 1-D) to get +/- 20 knots and +/- 2 degrees. That is in the ballpark of the actual radar performance. I would say that a 10-minute section allows crude speed and bearing information to be extracted and compared. Sections of radar data shorter than 10 minutes are just an exercise in differentiating noisy position measurements, creating very large, unreal variations in speed and bearing that are caused by position measurement noise.

  215. Gysbreght says:

    DrB: How about the Kalman filter applied by the DSTG to filter out the measurement noise you are adddressing, and the three-sigma error bands indicated in its Fig. 4.2?

  216. Brock McEwen says:

    @DrB: re: wind can blow an “outside in” 19:41 intersection back onto remaining arcs: you have replaced one coincidence with two coincidences, plus what you (and thus far no-one else) would argue to be a physical impossibility:

    Coincidences:

    1) the only way your hypothetical works is if the plane intersects the 19:41 arc tangentially – which is only possible if it is going faster/higher than anyone has yet postulated, and considerably faster than MRC. In your seminal paper, your fitted true heading track is at (I believe) 503 KTAS – faster than any of your colleagues, who were minimizing BFO errors – and yet still appears to intersect “post-tangent” (i.e. inside-out).

    2) the winds need to be blowing west-to-east immediately after 19:41, or else the 20:41 arc will not be reached in time. A cursory glance at the winds to which you linked us show prevailing winds at that leg’s latitudes blowing east-to-west – the opposite direction to what is required.

    So an outside-in intersection requires the plane to speed up AND get blown upwind (or not speed up, but get blown dramatically upwind), and yet lay down a BTO track that happens, by fluke, to resemble an outside-in intersection at cruising speed, with no turns, and more intuitive wind effects.

    Physical impossibility:

    You are, I believe, by far the most vocal critic around here of your own seminal paper’s claim that such speeds are fuel feasible. These days, because [temperature effects Boeing missed], you tell us such speeds are utterly impossible, because MH370 would lack the fuel to reach Arc 7.

    (You could add “zero debris in Oz”, “timing/location of debris discoveries”, and “Arc 7 searched out for all swift-FMT true heading paths” as additional hurdles for any such path to clear.)

    So I stand by my original claim that any “outside in” path projection requires a perfectly offsetting set of compensating post-intersection path adjustments which I suggest should be rejected by any scientist with any respect whatsoever for the principle of parsimony for which you used to argue eloquently.

    More important (to me) issue:

    But I’m glad you join me in seeing the need for a rigorous public audit of search decision support. You may not be right about your temperature effects, but if search leadership is taking a zero on finding wreckage, then they owe it to those of us who are still working the problem to turn over ALL their data – including their fuel models and assumptions.

  217. Niels says:

    @sk999 regarding your March 11 12:42 pm post

    I’m interested to learn some more details of your calculation results, in particular those for TT mode. What are the 19:41, 00:11 latitudes and average GS you have found?
    In recent months I’m also focusing on “post” 19:41 paths, by minimization of the variance in the track generated by my explicit path generation tool. I find 00:11 latitudes around S34 degrees (but with a large margin if I apply realistic BFO offsets)

  218. Ge Rijn says:

    IMO all this detailed discussion about the Lido radar data, possible flight paths and attitudes and also the earlier radar data is of little use anymore by now in finding the crash area.
    It’s mainly based on speculation about data that are unreliable, incomplete, not disclosed, so not verified by independent parties.
    A mix of data projected on a screen coming from another screen or screens or who knows where they come from. Anyway no one here knows for sure.

    IMO the only more or less reliable objective data we have now are the Inmarsat data, the latest debris information and the latest drift analises based on drifters and the debris find locations.
    Those are objectively indicating a crash area between ~29s and ~33S.

    The kind of debris being more than 90% related to trailing edge, surface control, wing/engine, nose gear door related pieces. Their general big sizes over medio ~25 inch. Their kind of (lack of high impact) damage. The small amount of debris found so far indicating a small debris field.

    All this data and information is objectively indicating a crash area and an attitude in which the plane impacted the water. Which could not have been a high speed near vertical dive.

    It seems to me this objective data and information is too much ignored now.

  219. Victor Iannello says:

    @Ge Rijn: How do you “objectively” account for debris that amounts to small pieces of interior parts?

  220. DrB says:

    @Gysbreght,

    You said: “DrB: How about the Kalman filter applied by the DSTG to filter out the measurement noise you are adddressing, and the three-sigma error bands indicated in its Fig. 4.2?”

    Good question. First, the Kalman filter cannot extract any information that is not contained in the raw data. So you really can’t “filter out” measurement noise. You can minimize it, but not eliminate it. If your filter is not optimal, you can easily create misleading results (as DSTG admits for turns).

    On page 18 DSTG says: “The measurement error was assumed to have a standard deviation of 0.5 nm . . . .” They don’t say if this is 1-dimensional or 2-dimensional. They also don’t seem to provide any basis for this estimate (maybe I missed it). By eyeball I would have used a somewhat larger estimate – about 1 NM per dimension.

    The DSTG error bounds on Figure 4.2 vary from +/- 20 kts to +/- 60 kts after 17:40. The air speed here is consistent with a constant value, except for a small region near 17:52 which is when a turn occurred. As Figure 4.2 itself demonstrates, the Kalman filter does very poorly at turns (like at 17:22).
    Considering the lack of completeness in DSTG’s description of their error analysis, I don’t see any significant disagreement in the results.

    I only draw three conclusions from their Figure 4.2, and they are the same results I got by analyzing the same data without Kalman filtering:

    (1) the air speed is difficult to estimate during turns,

    (2) the air speed after the turn-back was measurably higher than it was before the turn-back, and

    (3) there is no convincing evidence for a speed change at any time except at ~17:25.

  221. Ge Rijn says:

    @VictorI

    Only 3 pieces of interior parts are identified. The Rodrigues closet piece which was clearly not damaged beyond reqocnition (due to high speed impact) and rather large. The IFE-mounting which was largely intact and complete. And a piece of panel which was badly damaged but also rather large and of light structure. All easely damaged like this in a relatively low speed impact (like Asiana 214 see interior pictures of this crash).
    A breaching of the fuselage, seperating of the the tail section or a blow out of a door (like happened during the crash landing of Asiana 214) was necessary to release this pieces from the hull.

  222. DrB says:

    @Brock,

    I appreciate your interest in my work. I see I still have work to do in making it understandable to others such as yourself. Virtually all the claims you made about the impossibility of such a route are incorrect.

    Perhaps this figure will help you understand.

    This is a plot of the route with accurate GDAS wind vectors shown periodically along it. At the EOR event, the wind is from the SW causing the heading to be slightly west by a degree or so of the 180 degree true track being followed then. So you start out in CTH navigation with a heading about 181 degrees. Then after a short time the wind shifts to the east. This blows the aircraft to the west toward the 19:41 arc from the outside. It becomes almost tangent to it near the equator. That’s the 19:41 handshake. The wind continues to be from the east, through 20:41 near 7.5 S and 21:41 near 15S. At about 22:00 the wind shifts to the west and strengthens. The 22:41 handshake is near 22.5S and the aircraft is now drifting eastward. The eastward drift continues through 00:11 near 33S.

    No high airspeeds are required – just Best Holding (considerably slower than LRC). Look at the plot. The ground speed is ~7.5 degrees of latitude per hour. That’s 450 knots. The air speed is not very different from the ground speed because the winds are predominantly cross-winds, not tail-winds.

    Nothing in this route is “physically impossible”. It’s a piece of cake to fly. You don’t even need a pilot after 18:23.

    You also don’t need the wind “to be blowing west-to-east immediately after 19:41, or else the 20:41 arc will not be reached in time.” In fact, you need the opposite – an east-to-west wind, which is what the GDAS data show.

    You said: “So I stand by my original claim that any “outside in” path projection requires a perfectly offsetting set of compensating post-intersection path adjustments which I suggest should be rejected by any scientist with any respect whatsoever for the principle of parsimony for which you used to argue eloquently.”

    In this case your claim doesn’t stand up. You don’t need either high air speed or west-to-east winds at 19:41.

    If you want to see something REALLY INTERESTING, have a look at
    this . It’s an overlay plot of my proposed route and the Inmarsat route. I’ll have more to say about this later.

  223. Andrew says:

    @buyerninety
    @DrB

    Inaccurate wind data loaded in the FMC will not affect the aircraft’s tracking, but will affect the computed ECON speed, as DrB indicated. The inaccurate data will also affect the FMC’s ETA and fuel remaining predictions at waypoints a long distance ahead of the aircraft, because of the way the FMC mixes the actual and forecast data. Predictions of the optimum point for a step climb will also be affected.

  224. Andrew says:

    @buyerninety

    The altitude ‘constraints’ you mentioned in your earlier post define a block of airspace for a particular route. ATC may clear an aircraft at standard levels within that block of airspace. The altitude limits are not defined in the FMC’s navigation database and are not used by the FMC in any way.

  225. Gysbreght says:

    @DrB: “So you really can’t “filter out” measurement noise. You can minimize it, but not eliminate it.”

    Of course you can’t filter out measurement noise completely. That’s why there is “covariance of estimates, illustrated as mean plus and minus three-sigma value”.

    ”The DSTG error bounds on Figure 4.2 vary from +/- 20 kts to +/- 60 kts after 17:40.”
    I think that with your exaggeration you are exposing the weakness of your argument. In the period of interest, between 17:30 and 18:00, the three-sigma error bounds are +/- 15 kt, and the “smoothed”speed varies by more than 50 kt. I don’t think the air speed can be said to be “consistent with a constant value”.

    ”(1) the air speed is difficult to estimate during turns,”
    I’ve no idea on what you base that statement. The DSTG makes a reservation only for the high acceleration manoeuvre performed by the aircraft during the first turn between 17:20 and 17:30

  226. Victor Iannello says:

    @Gysbreght & @DrB: According to the DSTG Fig 4.2, before the big dip in speed at the turn, the speed peaks at 491 kn at 17:18:57. At that point in time, the speed according to ADS-B data supplied by FR24 is 473 kn, representing an error of 18 kn. The ADS-B data speed is essentially flat in this time period. I think there is a possibility that the Kalman filter is introducing ringing caused by noisy data, as can be seen by this artefact. Again, I’ll present more on this in the future.

  227. Brock McEwen says:

    @DrB: this is the same plot you showed me months ago. It still curves right-to-left = west-to-east = upwind between 19:41=red and 20:41=orange. Slightly, I grant, but come on, man – UP WIND?!

    So I still think either the null earth wind plot you’ve presented is incorrect, or your path is not being modeled with fidelity.

    And you’ve still selected a counter-intuitive speed (450) so that the (correctly modeled) eastward drift in the final hour cancels the wind drift perfectly, to lay down a BTO track which is by fluke, indistinguishable from true track at LRC/MRC and cruising altitude.

    You like your new post-19:41 path. I get it. You enjoy selling it. I get it. It is not the most parsimonious interpretation of the BTO data. I can reference either common sense, Simon Hardy’s work, the IG’s tangent work, OR your own seminal 2014 paper (which chose true track=straight line at ~500 ktas as the most parsimonious path) to prove it. Please convince all of them before having another go at me.

    But again: our time is far better spent developing and signing a joint demand that search leadership show us what it’s got in its pocketses. I grow increasingly shocked by the willingness of scientists in this community to continue to snarl over the scant table scraps of “evidence” the public has been fed thus far. We’re like a jury being denied access to any actual testimony, yet told to assess the court room sketches until we can render a verdict.

    I for one demand access to the court of law. Even if we agree on nothing else, I dearly hope you join me in this demand.

  228. Kenyon says:

    @Ge Rijn

    Re: “The IFE-mounting which was largely intact and complete.”

    The IFE interior frame recovered on MH370 is not so easily removed from the interior if the seat back. This is the interior mounting frame, not the exterior casing. This is not a piece you would find inside the cabin or on tarmac of Asiana 214 due to low impact. It could be possible to find on MH17 due to the relatively high energy of impact but even on MH17 there appears to be no interior mounting frames ejected as in MH370 as evidenced in MH17 crash photos.

  229. Brock McEwen says:

    Re: “the site where the aircraft entered the water was not between latitudes 32°S and 25°S along the 7th arc.” (per ATSB’s Dec 20, 2016 “First Principles Review”, p.18):

    The one and only reason the ATSB seems to need in order to reject this entire region? “The extensive aerial and surface search conducted between 18 March and 28 April 2014.” (IBID)

    Interesting. The “priority area” driving the mid-2014 multimillion dollar bathymetric scan of a vast region NE of Broken Ridge was determined by this same agency to be between -27.4S and -32.1S (“MH370 – Definition of Underwater Search Areas”, June 26, 2014, p.41).

    Back then, the dispositive value of their “extensive aerial search” of that region two months earlier did not even rate a mention. (IBID)

    How does the same aerial search go from irrelevant to game-breaking in the ATSB’s eyes?

  230. Victor Iannello says:

    @Brock McEwen: I have asked the same question. The standard response is that today there is more certainty of a high-speed impact leading to a large debris field, and therefore it should have been spotted during the aerial search.

  231. TBill says:

    @DrB
    Can you give us your updated flight path complete *.pdf again? I have the 9-Jan version. I am thinking you gave us an update but I cannot find the link. Thank you.

  232. Ge Rijn says:

    @Kenyon

    The IFE-piece is the exterior casing. It’s clamped on and can be removed by just pulling with your hands.
    Look at the Asiana 214 pictures in the link. Altough a different IFE casing design you can see a few of those exterior casings detached also.
    Then take a look at the picture of the detached closet. A similar closet as where the Rodrigues piece came from. Also in this case the front wall was detached and destroyed.
    And look at all the interior panels that detached and were broken in pieces:

    http://www.nycaviation.com/2013/07/photos-inside-the-asiana-214-wreckage-and-cleanup/

  233. DennisW says:

    @Ge Rijn

    The French are the stumbling block here. They know full well how the trailing edge of the flaperon was damaged – water impact or flutter. Why there has been no reports on their investigation is frustrating.

  234. DrB says:

    @Brock,

    You said: “@DrB: this is the same plot you showed me months ago. It still curves right-to-left = west-to-east = upwind between 19:41=red and 20:41=orange. Slightly, I grant, but come on, man – UP WIND?!“

    You are misunderstanding what the wind actually does. The easterly wind does not have to change direction to cause a leftward bend. It only has to slacken.

    The wind is always blowing from the east from ~19:00 – 22:00, but the strength varies. The typical windspeed from 19:00 to 20:00 is fairly high (as indicated by the length of the arrows). With a constant heading of ~181 degrees, the strong easterly winds there cause the track to be several degrees to the right (~183-184 degrees). That’s why the path curves to the right. Then the wind speed slacks off from 20:00 to 22:00, but it is still from the east. Now the true track will be smaller than 184 degrees and closer to the true heading of 181 degrees. That’s why there is a slight curvature of a couple degrees back to the left ( from ~184 to ~182). Continuing on, after ~22:00 the wind actually shifts to the west. Now the track is pushed to the left of the heading so it is less than 181 degrees. The very strong westerly winds at 22:40 to 00:11 actually shift the track from 181 heading to ~174 degrees track.

    To summarize, with an easterly wind, the track will always be greater than the heading (181), but how far past 181 depends on the (left) cross-wind speed. With a west wind, the track will always be smaller than the heading (181), with the deviation depending on the westerly (right) cross-wind speed.

    You also said: “And you’ve still selected a counter-intuitive speed (450) so that the (correctly modeled) eastward drift in the final hour cancels the wind drift perfectly, to lay down a BTO track which is by fluke, indistinguishable from true track at LRC/MRC and cruising altitude.”

    This is not correct either. A ~180 degree true track at any auto-pilot speed cannot match the BTOs. Also I guess you failed to notice that the speed required (about 450 kts down to ~22:41) is much less than LRC and MRC.

    There is nothing counter-intuitive about that kind of speed. If you fly faster you won’t make the 7th Arc going south. You’ll run out of fuel first.
    Inmarsat tried a due south track and found it did not work. They did not understand how their final route came to be, but they got it almost right just by tweaking the track to vary a few degrees on either side and also allowing the speed to decline (again, for which they had no explanation other than that’s what the satellite data indicated). Their only major error was in the last leg because they failed to understand it was the wind that was causing the track to vary and the winds were quite strong during that last long leg.

    As various people, including at least Victor and DennisW have pointed out, incorrect assumptions lead to incorrect results when fitting the satellite data. The high-speed straight track is a perfect example. It fits the BTOs (too well), and it marginally fits the BFOs is you allow sizable residuals. However it is incorrect because of two flawed assumptions that I made back in the summer of 2014: (1) a constant TAS speed could be flown by a B777 auto-pilot, and (2) there was enough fuel to reach the 7th Arc at that speed. The whole reason I built a fuel model was to prove that there was enough fuel. To my surprise, it turned out that not only can’t you reach 39S, you can’t even get to 36S!

  235. Ge Rijn says:

    @Kenyon

    Proof the IFE-casing is the exterior casing comes from the cloth-hook.

    Another interesting picture is the nose gear. You can see the nose gear doors are gone while nose gear chamber is still intact.

  236. DrB says:

    @all,

    I have significantly revised my paper on the OXCO transient effect on BFOs and other related topics. You can get it
    here .

    As always, constructive criticisms are welcomed.

    Two of the real values of Victor’s blog are the opportunity to learn from others and to get feedback on one’s own work. It’s really quite valuable to have a process something like refereeing a paper for publication, except this has a lot faster turnaround.

  237. Ge Rijn says:

    @DennisW

    The French are treating MH370 as a criminal case. They probably have good reason to keep treating it this way. Therefore they keep their evidence secret for the time being IMO. They possibly only will show it in a court trial or on demand of a court.
    I agree with you they probably know how the flaperon damage and separation occured.
    And this would be key evidence on how the plane entered the water and if this was done deliberate or not.

  238. DrB says:

    @TBill,

    No, I have not posted an update to my route file since Jan. 9th. I am preparing a (very long) paper with full details, but that won’t be out for a while. Some of the refinements I have made over the past two months have moved the end point a bit to the NE. It’s interesting actually, that this seems to happen with almost every tweak.

    I hope to incorporate a more refined FMT turning model if Andrew has a successful simulator test of this.

    In addition, I am working on an end-of-flight model that several people here have contributed their knowledge and experience to making more accurate. I hope to post it this week for feedback.

  239. Ge Rijn says:

    @VictorI

    IMO it’s more logical to conclude no large debris field was spotted during the aerial search for there was no large debris field. Otherwise they should have spotted it. The aerial search started late. A small debris field would have been spread over a wide area by the time the planes reach the search areas.
    A few items of interest were spotted but could not be found and retrieved by the time ships reached those spots.

    Griffin and @MPat predicted ~30 pieces (drifters) out of ~170 pieces of debris (drifters) to start with would arrive on African shores within ~2 years. This is the picture we see till now.

    I thought it was @Brock McEwen who calculated with a large debris field of ~10.000 floating pieces ~600 pieces of debris should arrive on African shores within ~2 years ( I’m sure Brock McEwen will correct me if I’m wrong on this).

  240. DennisW says:

    @DrB

    I do have some comments on your OCXO paper, but the format you use does not allow cut and paste. Could you please put it out as a pdf so that references to your work can be accurately retrieved.

  241. Brock McEwen says:

    @DrB: I concede you have just educated me on true heading. I get it now. Apologies for taking so long – it is fiendishly counterintuitive for true track to be impervious to wind, whilst true heading (i.e. “let the wind blow ye where’re it may”) drifts you upwind. But I now understand that the selected heading is relatively upwind to begin with – with the slight winds offsetting this in the first hour, with zero winds causing the path to approach the programmed heading. Thanks for taking the time and energy to explain it.

    But of course I noticed that 450 was less than MRC/LRC; my point has always been that the answer to any selection of speed, altitude, or mode – and/or changes to any of the above – cannot be simply, “because that’s what’s required to fit the BTO’s. One needs a better reason than that. True track at LRC/MRC at/near last known altitude, with no turns after 18:40, fits the BTO data, and every flight dynamic can be defended intrinsically as a likely a priori selection.

    I have never argued anyone’s path was “impossible”. I just think that your combination of flight dynamics (slow to 450, select true heading) has to me very little in the way of intrinsic rationale. It may be in the mix now (among ISAT data trusters), given the searched out area, but this path proposal would have been laughed out of these forums in mid-2014. (But don’t feel bad; so would all the others getting floated.)

    @Victor: Thank you for asking the question. It may not surprise you to learn I don’t think much of the answer; what was it about their own original working hypothesis – hypoxia, with fuel exhaustion at ~FL350, followed by pilotless spiral to impact – would lead them to assume anything BUT a high-energy impact? What rubbish.

  242. DrB says:

    @Andrew,

    I had another thought about your simulator run. To find out whether the FMC hold True Heading or Mag Heading after the END OF ROUTE error, being able to effectively speed up route is an advantage, not a disadvantage. That feature allows you to fly a greater distance in a shorter real time in the simulator, and this is just what we need. After all, the timing of the heading readings (true and magnetic) after the EOR error is not critical at all. We just need to see which one is constant and which one varies.

    I don’t know how much you can speed things up, but if you can cover 2,000 NM or more into less than an hour then you will see larger magnetic variation and the result should be very clear. It would be good to see one of the headings (i.e., mag and true) change by at least 2 degrees if possible, just to make sure we get it right.

  243. TBill says:

    @all
    Did some BFO calcs (@Arc6) arbitrarily assuming straight path from MEMAK to various spots on Arc7:

    Looking for 252 BFO ARC6 (measured):

    40S calc BFO = 260
    36S calc BFO = 254.5
    30S calc BFO = 246.5 (similar to Richard’s YWKS path)
    26S calc BFO = 241.0 (similar to Victor/Rickards NZPG path)
    09S calc BFO = 221

    That’s put 252 BFO at around 34-S which is similar to DrB and SK999 paths. If I invoke a turn to 90 deg to the Sun, I can remediate the 240’s to 252 for Arc6, if I thought Arc6 BFO was my only problem with a curved path.

  244. buyerninety says:

    @Andrew
    The picture of the flightpath that that post is referencing is this;
    https://www.dropbox.com/s/ehv7qma79posla1/2017-03-05%20Cockpit%20R15%20offset.png?dl=1
    We were trying to ascertain why the aircraft path is projected to to fly in that large
    curve beyond the waypoint, rather than simply turn left a little before the Active
    waypoint. Also why VNAV ALT is seen annuciated on the MPD.

    My suggestion (for checking) was that perhaps the (PMDG) FMC regards those waypoints as
    having the altitude limits of those routes they are on.

    A left turn before the waypoint then settling on the heading that the Entry to the HOLD
    requires, in concert with a reduction from the speed seen on the display to that required
    in the HOLD, seems achievable in the 22+ Nm between the Active Waypoint and the HOLD
    position (for the non-offset represented path). DrB has opined that a left turn slightly
    before the Active waypoint appears a more efficient manoeuvre than the large curve the
    FMC is planning.

    The large curve seems superfluous, unless the PMDG FMC is calculating to allow not only for
    a speed reduction, but also for a reduction in altitude (if e.g. the PMDG FMC waypoints have
    the altitude limits that apply to the routes they are on, the altitude reduction required
    would be {34000-27500=} 6500 feet.)
    Comments?

    @Mick Gilbert
    In your Lido overlay, the series of crosses you have designated with 300° are offset from
    the following series marked 297°. Given that a plot point constitutes the first (leftmost)
    element of a plot, with the designator to the right of the plot point, it is debateable
    whether your 300 series follows the plot points on the left part of that radar trace, or
    follows a line placed approximately through the centre of that trace ‘smear’. I hope you
    can understand that others can have a different perception of the plot points location
    to your perception.
    That aside, you should acknowledge when arguing this matter that there is an included
    change of angle, under Penang to VAMPI (~291°), then changing to VAMPI to MEKAR (287°) –
    therefore at least 4° change of angle is implicit when LNAVing from under Penang (say
    fly-bying KENDI) via VAMPI on to to MEKAR. At present, your assertion of target not LNAVing
    because you state it is drifting by angles >10°’s, includes this implicit 4°angle, which
    would be more correctly not included when asserting any ‘drift’ angle amount.

    @sk999
    I will correct my reference in future to be ‘the ADS-B derived track’ or close term.
    “The paragraphs you cite from the V Australia flight training manual refer to inputs
    needed for the calculation of step climbs, not lateral navigation.”
    Thankyou, I was hoping that that matter of ‘FMC winds’ would be clarified by someone.

  245. DrB says:

    @Brock,

    Good. Finally. Yes, it takes a while to figure out what happens in constant heading modes. It didn’t come easy at first to me, either.

    You said: “But of course I noticed that 450 was less than MRC/LRC; my point has always been that the answer to any selection of speed, altitude, or mode – and/or changes to any of the above – cannot be simply, “because that’s what’s required to fit the BTO’s.”

    Here we go again. It seems I have also failed to make very clear to you another point. My route does not have an adjustable speed.

    The Best Holding speed is “cast in stone” by Boeing for this aircraft configuration in their Holding table of KIAS versus altitude and weight. I simply interpolate their table to get IAS. I do have altitude as a free parameter, but that is not a degree of freedom available solely to tweak speed, because it also impacts the Fuel Flow, and there is only one altitude where the endurance is correct when using the (now known) engine PDAs. So altitude, in effect, is already constrained by the need to match the endurance. There is no degree of freedom available for adjusting speed, other than the minor tail-wind errors of ~5 knots or so.

    I think my route may actually be the first one that fits the constraints and has no adjustable speed parameter and also has the altitude constrained by the endurance (not by the BTOs). That fact that it still fits the BTOs to 29 microseconds and the BFOs to ~2 Hz without a speed parameter and with the altitude separately constrained by endurance is quite important in my view. On this point we seem to agree.

  246. Andrew says:

    @DrB:

    Thanks, I had the same thought – I certainly won’t have the luxury of sitting there for 4-5 hours watching it all unfold in real-time!

    I can certainly factor in the groundspeed multiplier to the elapsed times as you suggested previously. The only problem is I don’t know the accuracy of the multiplier, so any factored times might only be approximate.

  247. DrB says:

    @DennisW,

    Hmmm. That paper is a PDF file already, but you are correct that you can’t copy out of it. I used WORD and then just saved it as a PDF, and that’s what you get. I am not aware of any options in WORD when creating the PDF version.

    I’m open to suggestions if someone can tell me how to create a PDF which allows copying selections.

    I would prefer not to post the WORD file for several reasons. I prefer the PDF format because that does not allow editing of the document, but I would like people to be able to extract whatever pieces they want for their own use. Can someone help with this?

  248. DrB says:

    @Andrew,

    Approximate times are fine. Knowing the scaling factors will allow adequate approximations of the times.

    Again, the key questions are:

    1. How does the FMC actually fly the turn from 296 true to 180 true with a 15 NM right offset?
    2. Do the EOO and EOR errors occur simultaneously when the Holding (+ offset) fix is reached?
    3. After an EOR, does the FMC hold mag heading constant or true heading constant?
    4. A less-important question is, when does the slow-down to Best Holding occur?

    Perhaps after some introductory work, a NOK group (or perhaps Mary Schiavo) might be interested in using crowdfunding money to pay for full-length simulator runs?

  249. buyerninety says:

    @DrB
    If covering a shorter distance in the CTH test is desireable, the magnetic lines of force
    at lower Africa ‘bunch’ and allow transversing over a changing declination of from +1°
    (comparable to initial declination in Malacca Strait), to ~+28°, approaching the southern
    end of Africa by traversing , say, the 30° Easting, over a distance of 1200Nm.
    If that area is undesireable, a similar ‘bunching’ occurs, in the region of the U.S. Great
    Lakes traversing then eastwards towards the U.S. East Coast.

  250. buyerninety says:

    EDIT; ‘changing declination of from +1° from the region of Lake Tanganyika

  251. DrB says:

    @DennisW,

    The file properties say “Content Copying is Allowed.” If you view the file in your web browser, you can’t copy from it. However, if you download the file and open it with Acrobat Reader, then you can select and copy sections to paste in a second file.

    Let me know if that does not work for you.

  252. Andrew says:

    @DrB:

    I can answer Q.2 now, although I will confirm it in the simulator. The END OF OFFSET error message should occur two minutes before reaching the offset termination point abeam ANOKO. The END OF ROUTE error message should then occur as the aircraft passes abeam ANOKO.

  253. DrB says:

    @buyerninety,

    Yes there are better places to do this experiment than near Indonesia. In fact, almost everyplace else in the world is better in that regard.

    It’s up to Andrew, but I think it would be better to do all the testing using the real locations with one setup and one continuous run. I would just start at 18:22 with the correct legs, put in the offset, let it run through the FMT and slow-down until the FMC errors occur (all at real time rate), and then speed it up for the SIO route until you get a clear winner on heading. Speeding up the sim after the FMC errors occur can give up to 10’ish degrees of change in a reasonable length of time.

  254. DrB says:

    @Andrew,

    Roger that. Thanks.

  255. Andrew says:

    @DrB
    @buyerninety

    Thanks for that suggestion @buyerninety – you saved me the effort of checking the charts for a better area! I’ll endeavour to try it on the route south from ANOKO, but if I’m pushed for time I’ll try a southerly route over southern Africa. It’s easy enough to reposition the sim to a new spot and the close spacing of the isogonals in that part of the world would answer the question very quickly.

  256. Andrew says:

    @Brock McEwen
    @DrB

    Forgive me for intruding, but the 450kt true airspeed that DrB used equates to an indicated airspeed of about 253kt at FL360 and ISA+10. That indicated speed happens to be the aircraft’s best holding speed at FL360, given the aircraft’s likely weight at ANOKO.

    No doubt it’s been discussed previously, but ANOKO lies at the beginning of one of the arrival routes to Banda Aceh, so a planned hold at ANOKO is consistent with a scenario where the crew intended diverting to Banda Aceh for some unexplained reason. As DrB suggested in his paper, the crew might have become incapacitated after executing an offset and the aircraft then continued on its merry way past ANOKO, on a constant heading because that is how the aircraft is designed to behave. That probably seems bizarre, but stranger things have happened in aviation; a hijack gone wrong perhaps?

  257. David says:

    @all. The Ethiopian 961 ditching at Comoros Islands has some salient features aside from any ditching aspects.
    Although a 767 there are some parallels which might be of interest, including engine run down times and speed decrease rate after first engine fuel exhaustion, plus the 767’s APU run time on residual fuel after fuel exhaustion.
    https://www.dropbox.com/s/j7zfs6tu8z2774v/Comoros%20accident%20appraisal.docx?dl=0

  258. Mick Gilbert says:

    @buyerninety

    Re: the Lido slide, yes, I’m aware that there are different interpretations of the data; that’s why I posted on here, I am keen to hear (read) them. As I stated in my original post, my analysis is not particularly sophisticated and my “best fit” tracks were fitted by eye.

    Regarding the first two clusters of plots (in fact, all of the clusters) I worked off Bill Holland’s cleaned up version in the first instance (https://www.dropbox.com/s/87qntavpy7t5aih/Lido%20-%20Seg%201%20and%202.png?dl=0) and then transcribed to a Skyvector background for context. Accepting your point about a track change at VAMPI, that still leaves 4° of drift in the gap.

  259. Mick Gilbert says:

    @DrB
    @Andrew

    I understand that best holding speed will vary with altitude and weight. Is it the case that once the FMC has calculated a best holding speed, it is not then subsequently varied as the airplane gets lighter?

  260. David says:

    @Andrew. Yes I follow what you say about the Boeing simulation case 1 and 2 distribution, thank you.
    They say, “Reasonable values were selected for the aircraft’s speed, fuel, electrical configuration and altitude, along with the turbulence level”. That leaves somewhat open how many Case 2’s there were. As you say the two Case 1’s both going right is contradicted by their both-right-and-left remark. Even so it is unclear what precisely they mean by, “In an electrical configuration where the loss of engine power from one engine resulted in the loss of autopilot (AP)…” and how many such configurations they simulated. I hope one day we will learn rather more about what the whole 10 did simulate.

    On 25th February, 4.44 am before you started posting here, Gysbreght mentioned some information he had obtained from the ATSB which I repeat to save him finding them/you being unaware:
    “Yes, the ATSB report does not state which engine fails. Apparently the one-engine-inoperative tracks are no.s 7, 8 and 10 (see chart below). I assume that no. 9 is with both engines failed, and the airplane trimmed for one engine inoperative.”
    “ The database was exceeded only in tracks no. 3 and 7.”
    “Simulations 3, 7, 8, 9 and 10 “recorded descent rates that equalled or exceeded values derived from the final SATCOM transmission. Similarly, the increase in descent rates across an 8 second period (as per the two final BFO values) equalled or exceeded those derived from the SATCOM transmissions.”
    “I have asked the ATSB which aspects of the scenario could explain that track no.3 is so special in comparison with others in the group 1 – 6.”
    His chart is at
    https://www.dropbox.com/s/z02p174h4kdybad/BoeingSims.png?dl=0.

    Subsequently I have approached the ATSB for more information on this also. So far we have received no response though I gather it can be expected.

  261. Andrew says:

    @Mick

    The FMC best hold speed should decrease as fuel is burned and the aircraft’s weight decreases. That’s reflected in Fig. 7 of DrB’s paper, if not the discussion above! DrB??

    @ David

    Thank you for the additional information Gysbreght gleaned the ATSB; it helps clarify their report a little. It’s a shame they didn’t spell it all out in the first place!

  262. Gysbreght says:

    @Andrew: “It’s a shame they didn’t spell it all out in the first place!”

    Hear, hear!

  263. Victor Iannello says:

    @DrB said, “Perhaps after some introductory work, a NOK group (or perhaps Mary Schiavo) might be interested in using crowdfunding money to pay for full-length simulator runs?”

    To be clear, I won’t let this site be used to raise money, especially from the NOK. Relative to Mary Schiavo, she is a lawyer representing clients that are suing Boeing. You can collaborate with whom you choose, but be aware of conflicts.

  264. Victor Iannello says:

    @Andrew & @Mick: @Andrew said, “The FMC best hold speed should decrease as fuel is burned and the aircraft’s weight decreases.”

    Assuming the A/P is in VNAV mode, i.e., the speed window is closed, which is Bobby’s assumption.

  265. TBill says:

    Re: Simulator tests
    I’d give this lower priority but if it was me first thing I’d do is see how fast I could depressure by putting outflow valves on manual/open and shutting bleed air. I think the answer if pretty darn fast (contrary to Nixon’s recent book comments) although it’s possible the sophisticated simulator is not programmed accurately for that experiment either.

  266. Andrew says:

    @TBill

    Yes, it would be “pretty darn fast” with the outflow valves in MAN and fully open! Remind me again what Nixon had to say?

  267. Brock McEwen says:

    @DrB: I am glad we remain on the same page re: parsimony. You will recall I championed your seminal paper for its treatment of BFO errors in particular, as something to manage – not minimize.

    But you will also recall my primary criticism – that you were overstating your conclusions (flowery adjectives, expressions of certainty).

    Back then, you had the lowest PDAs in town – and we’re sure they were right.

    Nowadays, you have the highest PDAs in town – and are sure they are right.

    If you can demonstrate conclusively that Boeing misguided the search for three years with infeasibly low PDAs – something that joining my demand for full transparency should help you achieve – I will withdraw my criticism of your path.

    Until such time, I dispute this first domino you topple – and thus will deem the rest of the dominos (-> holding -> 450 -> true heading) unnecessary, and will instead continue to defer to the IG’s published assessment of a priori expected autopilot dynamics:

    First place: MRC/LRC at ~FL350
    Second place: Holding at ~FL100 (though BFO fit is poor, and curl is left unexplained)
    Distant third: anything else

    I will not debate this proposal further until I see evidence Boeing misinformed the search re: PDAs.

  268. Andrew says:

    @buyerninety

    Sorry, I missed your earlier post. I’m not sure about the reason for the overshoot in the turn; possibly something to do with the large track change, as Victor suggested. Perhaps the FMC treats it as a fly-over waypoint in that situation. I intend checking to see if the simulator exhibits the same behaviour – track changes like that are not something we normally see in the real aircraft.

    VNAV ALT normally engages when there is a difference between the target altitude set on the MCP and the cruise altitude in the FMC. It is also seen in the descent phase when the aircraft levels off at an intermediate altitude while in VNAV. I can’t really comment any further without knowing how the FMC was setup and how the aircraft came to be at FL340.

    Regarding the altitude limits, the FMC doesn’t know about them, so they shouldn’t be a factor in any calculations it makes.

  269. DennisW says:

    @DrB

    Slow network speeds out here in the boonies, so I will paraphrase.

    I am very skeptical of the 18:25:27 logon event. You are claiming a BFO error of some 25Hz relative to your path model. I claim zero BFO error with my path model. I am not saying your path model is wrong, but it does allow you to reject the 18:25:27 BFO as does Holland.

    Why Holland or someone in the SSWG did not perform experiments with equipment identical to that in 9M-MRO baffles me. They could have characterized the transient response versus off time and temperature doing hundreds of logons. Instead Holland limits his observations to data gathered in previous 9M-MRO flights.

  270. TBill says:

    @Andrew
    Let me take another closer look at the Nixon book. Keep in mind, we do not know if a depressurization actually happened on MH370, and if it did, we do not know if it was intentional or accidental, nor when it might have happened during the flight. There is varied speculation out there.

  271. DrB says:

    @DennisW,

    You said: “Why Holland or someone in the SSWG did not perform experiments with equipment identical to that in 9M-MRO baffles me. They could have characterized the transient response versus off time and temperature doing hundreds of logons. Instead Holland limits his observations to data gathered in previous 9M-MRO flights.”

    Actually, all of Hollands events happened on the ground, not in flight. That is because very probably there are no other in-flight log-ons caused by a power interruption. It doesn’t happen in normal operations.

    I think (but am not certain) that there were additional ground-based tests of the SDU that go beyond what Holland reported, based on my recollection of ATSB’s description of SDU investigations. However, none of these would show the initial rise in BFO at LOR because the OXCO was not initially below room temperature for any of their tests. They only show the decay because the overshoot occurs earlier in those ground-based cases. In that sense, the 18:25 log-on is unique.

    This method of fitting a transient curve is not very sensitive to 10-20 Hz errors in the two 18:25 BFOs because the transient then is so steeply rising. The 18:27-18:28 BFOs actually have more impact on the fitted curve, and thus they offer a stronger constraint against major maneuvers happening then (18:27-18:28).

    You also said: “I am very skeptical of the 18:25:27 logon event. You are claiming a BFO error of some 25Hz relative to your path model. I claim zero BFO error with my path model. I am not saying your path model is wrong, but it does allow you to reject the 18:25:27 BFO as does Holland.”

    As I said above, I don’t think one can infer with certainty from the 18:25 BFOs alone that a maneuver was happening then or not. I am not making a case for that. I am contending that (1) the 18:25:27 BFO was accurately measured (Holland says it was not) and (2) the SLOP inferred from the BTO data alone is entirely consistent with the 18:25-18:28 BFOs once the transient is accounted for. If you want to believe that 9M-MRO was still on N571 at 18:25, your problem is with the BTOs, which don’t support this. As I said, you can’t decide this based on BFOs alone. I did not reject the 18:25 BFOs because they didn’t match my expectations (you seem to want to do the opposite – keep the first one because it matches your expectations). I concluded that it is impossible now to make a judgment, based on the 18:25 BFOs alone, of whether or not a maneuver was ongoing at that time. Fortunately, we have simultaneous BTO data that tells us one did occur.

  272. DrB says:

    @Mick Gilbert,
    @Andrew,
    @Victor,

    At the onset of Best Holding speed at 18:41: FL = 360, IAS = 256.1, TAS = 453.9, W = 206.72 MT, M = 0.7737, SAT = 226.6 K (ISAT + 9.8C).

    Just prior to right engine fuel exhaustion at 00:08: FL = 360, IAS = 237.3, TAS = 420.1, W = 175.10 MT, M = 0.7217, SAT = 223.1 K (ISAT + 6.3C).

    Yes, the speed is continually varied by the FMC in VNAV as Victor indicated. All of these parameters are plotted in my Figure 7.

  273. DrB says:

    @Victor,

    You said: “To be clear, I won’t let this site be used to raise money, especially from the NOK.”

    I wholeheartedly concur. I never suggested that.

    The NOK are doing crowdfunding already. I was pointing out that such funds might initially be put to better use to perform simulator runs that potentially confirm route predictions done using computer models. Obviously it would be cost efficient to do that before restarting a new and expensive seabed search, assuming there are credible candidate routes to be evaluated in a simulator.

  274. Victor Iannello says:

    @Andrew & @buyerninety: Re VNAV ALT mode, I don’t know exactly the history of the flight prior to holding altitude at FL340 as during that flight I did many combinations of ECON CLB, CRZ, ECON DES, and FLCH. I assure you that it doesn’t matter relative to the overshoot of the waypoints, which is what I was studying.

    As for whether the waypoints were fly-over type, I assure you they were not. They were entered by me, and I don’t know how to enter fly-over waypoints other than entering fly-by waypoints with fixes at some distance before and after the waypoint you want to assure is flown over. (There is probably some sequence to enter a fly-over waypoint in the FMC, but I don’t ever use them.) Of course, SIDs and STARs could have fly-over waypoints as part of the route, but that is not relevant here.

    For a route with a sharp turn, there will be overshoot. We know that for a given turn radius (dictated by speed and bank angle), there is a place to start the turn before the waypoint is reached that minimizes the distance flown with no overshoot. However, this is not the only consideration. Air traffic considerations dictate that cross track error is kept within limits, which may dictate that the turn is started later and includes overshoot. I haven’t done enough experiments to determine what this criterion is, but I am sure there is threshold turn angle above which overshoot occurs.

    Also, overshoot doesn’t mean the waypoint was flown-over. It just means the turn initiated later than the point for minimum flown distance with no overshoot.

  275. DennisW says:

    @DrB

    ” I concluded that it is impossible now to make a judgment, based on the 18:25 BFOs alone, of whether or not a maneuver was ongoing at that time. Fortunately, we have simultaneous BTO data that tells us one did occur.”

    I think the path posted by Richard, which has no maneuvers, is perfectly acceptable relative to the BTO’s and also has an 18:25:27 BFO that is very close to the recorded BFO.

    Had Holland et. al. done tests with AES hardware they could have instrumented it to measure the oscillator frequency versus time for various temperatures and off periods very accurately instead of relying on BFO measurements.

    I also have significant heartburn with the BFO and BTO “statistics” bandied about by the DSTG. Neither metric is suitable suitable for modeling with Gaussian stats.

  276. DennisW says:

    @DrB

    “The NOK are doing crowdfunding already.”

    Do you have a link for that activity. I have been searching for updates on the “private search” status, and have been coming up empty.

  277. Victor Iannello says:

    @DrB: You said, “The NOK are doing crowdfunding already.”

    The NOK, through Voice370, have explicitly stated that they have NOT started to raise money. They are distancing themselves from any group that has started to raise money. Here is part of their statement:

    “We wish to announce that the NOKs have NOT initiated any crowdfunding for the MH370 Search. To date the NOK, as Voice370, are not associated with nor endorse any crowdfunding started by unknown parties.”

  278. Victor Iannello says:

    @DennisW: The statement from the NOK that I am aware of is the one I linked to in the post above.

  279. DrB says:

    @Victor,

    Here is the article that discusses crowdfunding by Project 370:

    http://www.ibtimes.com/will-mh370-ever-be-found-crowdfunding-campaign-aims-finance-new-search-missing-plane-2502116

    Perhaps it’s just poor reporting, or did Project370 add that last line about not initiating crowdfunding after their first statement was released?

  280. Victor Iannello says:

    @DrB: There are two groups. One is Voice370, which represents at least some NOK. That is the press release I referenced. The other is Project370, which is affiliated with the person who now calls himself Mike Chillit. Chillit’s group is trying to raise money. Voice370 is distancing itself from Chillit’s group (Project370). The confusion that has been created is very unfortunate.

  281. DrB says:

    @DennisW,

    You said: “I think the path posted by Richard, which has no maneuvers, is perfectly acceptable relative to the BTO’s and also has an 18:25:27 BFO that is very close to the recorded BFO.”

    Actually any path roughly parallel to N571 will appear to match the 18:25:27 BFO. By itself, that BFO is not a particularly useful indicator of route, even if one thinks he knows the OXCO transient effect at that time or chooses to assume it did not happen then but magically corrupted the BFO only 7 seconds later.

    If you believe the 18:22 location is correct, a maneuver did occur after 18:22. If you choose not to believe the 18:22 position, then you can change course before 18:22 and match the BTOs at 18:25-18:28. You can’t have it both ways. A maneuver is required. The only question is whether it occurred before 18:22 or after.

    You also said: “I also have significant heartburn with the BFO and BTO “statistics” bandied about by the DSTG. Neither metric is suitable suitable for modeling with Gaussian stats.”

    I agree with you for the BFOs. For the BTOs the PDF appears more Gaussian and there is no evidence of significant drift or structure in the errors, so I think in that case assuming Gaussian statistics is OK.

  282. Victor Iannello says:

    @DrB said, “If you choose not to believe the 18:22 position, then you can change course before 18:22 and match the BTOs at 18:25-18:28. You can’t have it both ways. A maneuver is required. The only question is whether it occurred before 18:22 or after.”

    I did an entire post showing there are straight paths starting from the penultimate radar point at 18:02 that satisfy the BTO and BFO data with no manoeuver.

  283. TBill says:

    @Andrew
    Nixon in his book discounts the possibility of intentional depressuring, says it is very difficult. Somewhat misleading, he then goes into a discussion about 500 ft/min max. depressure rate if the bleed air is turned off (here he is talking about natural leak rate through the hull with outflow valves closed). When Nixon finally directly addresses possible intentional case, he says that will not happen because the co-pilot would stop it. He also feels there was many spare hours of O2 in the unused masks and crew bottles, such that any such attempt would be unsuccessful. Basically he discounts the intentional depressure option as an unrealistic conspiracy theory of high complexity with little chance of success.

  284. DrB says:

    @Gysbreght,

    You said: “”The DSTG error bounds on Figure 4.2 vary from +/- 20 kts to +/- 60 kts after 17:40.” I think that with your exaggeration you are exposing the weakness of your argument. In the period of interest, between 17:30 and 18:00, the three-sigma error bounds are +/- 15 kt, and the “smoothed”speed varies by more than 50 kt. I don’t think the air speed can be said to be “consistent with a constant value”.”

    There is no exaggeration. I simply measured the peak-to-peak speed difference of the bounds shown in the DSTG Figure 4.2 speed curve. At 17:40 the bounds are +/-20 kts, and at 18:22 the bounds are +/- 60 kts. I didn’t make those numbers up. That’s what the DSTG graph shows. I didn’t say whether or not those were 1-sigma or 3-sigma. DSTG calls them 3-sigma. I don’t agree with their ASSUMPTION that the radar positions are good to 0.5 NM 1-sigma (they never did say if this was in 1-dimension or a CEP or ??), but that is not the point. The point is they believe the correct speed will be between those limits. A constant speed nicely falls between those bounds everywhere from 17:40 to 18:22 but for a few knots discrepancy during the turn around Penang, where the speed error will increase because of the curvature of the path.

    You also said: “”(1) the air speed is difficult to estimate during turns,”
    I’ve no idea on what you base that statement. The DSTG makes a reservation only for the high acceleration manoeuvre performed by the aircraft during the first turn between 17:20 and 17:30“

    My comment is based on my experience while doing it. If you tried it yourself, you would find that the unknown turn radius significantly degrades the accuracy of the estimate of the total turn path length (which is essentially the speed). The DSTG’s analysis method is oversimplified in the turns, but the data don’t support anything very sophisticated anyway. Another way to look at the problem is that the optimum Kalman filter to extract speed from a straight path is very different than the optimum Kalman filter to extract speed from a tightly turning (more curved) path. You can’t do well in both cases with a single filter. Their Figure 4.2 shows a likely speed error in their reconstruction method of ~300!!! knots at 17:23. Why would anyone be surprised at a 10-20 knot error in DSTG’s method for a (slower) turn at 17:52?

  285. Victor Iannello says:

    @DrB and @Gysbreght: My guess is there are time synchronization errors on the radar captures, possibly because more than one radar head was used to reconstruct the path. When the vector path is differentiated to obtain track and speed, the discontinuity in the path caused by the jump (forward or back) in timestamp causes a sharp peak (or dip) in speed. The Kalman filter then smooths this “impulse” and produces oscillations.

    The dip in speed to 199 kn (groundspeed) from a climb at high altitude is well beyond envelope protection limits. Even with a chandelle turn, there is no reason to intentionally put the plane in this condition.

    The peaks in speed for several intervals exceed Vmo/Mmo limits, regardless of altitude. While certainly possible, that would take some skill to fly in this condition. It is hard for me to imagine the same pilot who took the plane to 199 kn as the same pilot who is flying above Vmo/Mmo.

    Not to mention that the DSTG speed data doesn’t agree with the ADSB data before the transponder was turned off at 17:21.

    I really don’t put a lot of confidence in the accuracy of the speed data presented by the DSTG.

  286. DrB says:

    @Victor,

    Well said.

  287. Gysbreght says:

    @Victor: “I really don’t put a lot of confidence in the accuracy of the speed data presented by the DSTG.”

    You’re repeating yourself.

  288. Gysbreght says:

    @Victor: I agree with you that without access to the un-filtered radar data, and with pre-conceived notions on both sides of the argument, there is not much ground for discussing this any further.

    Regarding the the dip in speed to 199 kn (groundspeed) from a climb at high altitude, I would go even further to state that it is physically impossible as shown and therefore must represent a “mismatch between the assumed linear Kalman filter model and the high acceleration manoeuvre”.

    Paraphrasing Andrew, I would say that “it is a shame” for the authors to make that remark without attempting to explain what went wrong, and to correct that.

    My position remains that, if anything, there has been too much filtering and smoothing, and the actual speeds probably varied more than is shown in Fig.4.2.

  289. Lauren H. says:

    @DrB – Both you and I have focused on the endurance and range in trying to pinpoint MH370’s final location. The endurance to 17:30 just about eliminates major mechanical issues and the “check point” table is reasonably accurate in choosing the range. However, here are a few potential issues with some of your assumptions that might affect your calculations.

    We both used the FCOM as a source for the fuel calculations. It seems the Tables in the FCOM are used to manually calculate the amount of fuel required based on varying circumstances but some of these might be interpreted differently as follows:

    1. You used a 0% PDA for the values in the FCOM. Is that reasonable? I would think, for conservatism, the FCOM would be based on a “typical” PDA for engines with the average amount of hours (maybe ≈2-3% is reasonable). Note, the various tables show the extra fuel used during a climb is made up by the fuel saved during the descent. We all know this can’t be true so there must be “cushion” in the values.
    2. You included the 5% increased usage for Holding in a “Racetrack” pattern but did you see that some Engine INOP values include the extra fuel burned by the APU? I guess if you have an Engine INOP you might need the extra juice supplied by the APU and a pilot would start it up. Since the thinking is the APU started after the second engine flamed out, this might be an indication there wasn’t a Person Flying when the first Engine flamed out.
    3. You have used a Cost Index of 52 from one of the updates. However, the Cost Index of 52 wasn’t defined in the update and I have seen the Cost Index defined in kg/min or lb/hr where 1 kg/min=132 lb/hr. Big difference. Which did you use and why?

  290. Victor Iannello says:

    @DrB: Thank you. Another point about the DSTG speeds: If the speeds really did vary as shown, the extra fuel burn would push the end point even further north. The fuel flow is not linear with speed–it rises rapidly around Mmo, so the extra fuel consumed when flying faster than typical cruise speeds isn’t balanced by the reduced fuel consumed when slower than typical cruise speeds. In essence, fuel efficiency is greatly reduced at speeds at and above Mmo, which will reduce the possible range.

  291. Mick Gilbert says:

    @DrB.

    Thank you for the clarification on best holding speed, that’s what I thought. I was following your discussion with Brock and at one point you wrote, “My route does not have an adjustable speed.” That (coupled with the fact that I was looking at your original paper that only has 5 figures) is why I asked for the clarification.

  292. Andrew says:

    @TBill

    I haven’t read Nixon’s book, but I’ll make the following observations:

    1. The FO wouldn’t be able to stop the Capt depressurising the cabin if he was locked out of the flight deck or otherwise incapacitated.

    2. The passenger oxygen system on 9M-MRO was the chemical oxygen generator type, which only has a duration of about 22 minutes. The crew oxygen system is supplied by two oxygen cylinders and has a much longer duration, but it is only available to the pilots on the flight deck. At FL350/360, the 237 or so passengers and cabin crew would have run out of oxygen pretty darn quickly in a depressurised cabin, even if they’d used all the spare masks and portable bottles that were available in the cabin.

    I’m not saying it happened, but I don’t think it would be a difficult exercise.

  293. Gysbreght says:

    @Victor: “In essence, fuel efficiency is greatly reduced at speeds at and above Mmo, which will reduce the possible range.”

    When you talk about speeds above Mmo, what altitude are you assuming?

  294. DrB says:

    @Lauren H,

    Thanks for your inputs and suggestions on the fuel issue.

    You said: “1. You used a 0% PDA for the values in the FCOM. Is that reasonable? I would think, for conservatism, the FCOM would be based on a “typical” PDA for engines with the average amount of hours (maybe ≈2-3% is reasonable).”

    My understanding is that the FCOM numbers are what defines 0% PDA, and these FFs are designed to be used with brand new engines. The PDA values I have derived from the average Fuel Flows provided me by ATSB are calculated the same way:

    Measured Fuel Flow = FCOM Fuel Flow * (1 + PDA) * [ 1 + deltaTAT*(3%/10C) ]

    So regardless of whether or not my PDA definition is correct, I still get the correct number for the actual 9M-MRO Fuel Flow because I have “calibrated” it against the measured FFs from prior flights.

    My fuel model is only accurate for limited climbs and descents when already at a cruise altitude. It does not attempt to model FF for take-off and climb. It is not needed in this case since we have a fuel reading after the climb to FL350 is complete. The only climb I am using now is from FL350 to FL360. Not a lot of fuel is burned doing just 1,000 feet.

    The PDA is designed to be a correction to the FCOM fuel flow tables, and no part of the PDA is embedded in the tables.

    You also said: “You included the 5% increased usage for Holding in a “Racetrack” pattern but did you see that some Engine INOP values include the extra fuel burned by the APU?”

    The APU fuel burn is actually dominated by the inlet drag effect on the main engines fuel flows (if they are running) rather than by the APU engine itself. I got a chance last week to see a B777-200ER at the gate with the APU running and the inlet door open. It’s a pretty big scoop, so I can understand the extra drag. Unfortunately, the INOP Holding Fuel Flow table is useless to us because its highest altitude is FL250. The table assumes level flight at constant speed, and that’s about as high as you can fly on one engine. Previous discussions here indicate it might actually fly as high as FL290 with no maneuvering margin. So if one engine fails at FL360 while in Best Holding, the FCOM doesn’t tell you what the FF will be for the remaining engine. Andrew thinks 3500-4000 kg/hr and Gysbreght thinks maybe a little higher. That is why I am working on an end-of-flight model to try to more accurately determine when the right engine flamed out.

    You said: “. Since the thinking is the APU started after the second engine flamed out, this might be an indication there wasn’t a Person Flying when the first Engine flamed out.”

    I agree.

    You also said: “. You have used a Cost Index of 52 from one of the updates. However, the Cost Index of 52 wasn’t defined in the update and I have seen the Cost Index defined in kg/min or lb/hr where 1 kg/min=132 lb/hr. Big difference. Which did you use and why?”

    In response to my query, ATSB provided me the Flight Plan value of CI = 52. Subsequently it appeared in one of their reports. I use the Cost Index definition used by Boeing which I have made available here. See the graph on page 25 which applies to B777s.

  295. DrB says:

    @Mick,

    Right. I meant that my computer model does not have a specific speed parameter that I adjust when fitting the route using Best Holding speed. The FMC VNAV follows a schedule to adjust the IAS in real time to optimize endurance when Holding. I know what that schedule is from the Boeing manual, and I replicate it in my computer model.

  296. DrB says:

    Sorry. Left out a space so the link in my post above failed to work. Let me try again. Look for the articles on Cost Index here.

  297. DennisW says:

    @DrB

    “I agree with you for the BFOs. For the BTOs the PDF appears more Gaussian and there is no evidence of significant drift or structure in the errors, so I think in that case assuming Gaussian statistics is OK.”

    Yes, the histogram of BTO errors presented in the DSTG book does appear relatively Gaussian, however, large errors are significantly over represented in the sampled data.

    Also the distribution of the BTO errors has a significant non-zero mean of ~10 usec.

    http://tmex1.blogspot.com/2017/02/bto-error-scatter-plot.html

  298. Victor Iannello says:

    @Gysbreght: For a given TAS, lower altitude is lower M, but higher IAS. Overspeed is if either Vmo or Mmo is exceeded. At FL304, Mmo = 0.87 corresponds to Vmo = 330 kn, so that defines the maximum TAS for no overspeed, as @Andrew has said. With a +14K offset from ISA temperature, that corresponds to a TAS = 527 kn. Looking at the groundspeed from the DSTG report, and adjusting for the prevailing winds, puts the TAS higher than 527 kn for periods of time, and at times as high as 552(!)kn. Whether exceeding Vmo or Mmo, the fuel efficiency will be significantly less than ECON=52 or LRC conditions.

  299. TBill says:

    @Andrew
    I agree with you.
    Goodnight Malaysian 370 by pilot Ewan Wilson and Geoff Taylor is the book that speaks for me, pending the day we actually learn something new.

  300. DennisW says:

    @DrB

    Would you please post the values of the following parameters at 19:41 and 20:41 for your new route:

    ground speed
    track
    latitude
    longitude

  301. TBill says:

    @DrB @VictorI
    Re: over-shoot
    I see in FS9 that wind affects when the turn (at IGOGU to ANOKO) is initiated, but I do not see much impact on the shape of the turn. With a 50 knot tailwind, the aircraft starts the turn roughly 11.5 nM from IGOGO (vs. about 10 nM in the no wind base case). With a 50 knot headwind, the turn starts approx. 7 nM from IGOGU. Without recording actual tracks, the shapes looked about the same asfar as overshoot. I was hoping the 50-knot headwind could make the sharper (>90) turn work, but the FMC started the turn later to prevent that outcome.

  302. Kenyon says:

    @ Ge Rijn

    “The IFE-piece is the exterior casing”

    I recommend a deeper study of the recovered MH370 IFE frame and its installation location within the seatback. As you point out Asiana 214 seatback and IFE are not the same model as MH370. In addition the MH370 interior frame (not exterior bezel) has the integral coat hanger which protrudes from the inside to the outside and gathers fabric. I also believe from memory that all the visible exterior plastic, frame and bezel on MH370 was black (not critical to understanding the construction but of interest). The debris recovered is not exterior but interior to the seatback (and white).

    I venture that if all you had was a baseball bat and 20 minutes of swinging power you could not dislodge the IFE Frame as recovered on the beach. It requires series of significant blows and forces.

  303. Victor Iannello says:

    @TBill: Thank you for following up on this. I haven’t run the series of experiments to determine under exactly what conditions FSX produces the overshoot because I’ve already determined that overshoot occurs for the IGOGU-ANOKO turn and we don’t know how accurate FSX is for big turns. I suspect FSX is modeling it correctly as its “track” record has been pretty good for predicting the behavior of automated flight–PMDG receives continuous feedback from B777 pilots. For turns less than 90°, I don’t see overshoot.

  304. DrB says:

    @TBill,

    Thanks for testing the turns with FS9 in various head-winds. The results are strange. It appears like the turn start time depends more on the ground speed, not the air speed. For a constant maximum bank angle, the rate of turn depends inversely on the (air) speed. Higher speed means lower turn rate which means an earlier start time. Big tail-wind = high ground speed = lower turn rate = early start time. Big head-wind = low ground speed = higher turn rate = later start time. That’s what your test showed. Maybe the turn rate equation it uses incorrectly applies ground speed instead of air speed?? Well, at least we know the B777 is different from FS9.

  305. DrB says:

    @DennisW,

    Here are the parameters you requested (and a few more):

    Route: CTH at Best Holding

    Date UTC Event Lat Lon Track GS

    2014/3/7 18:43:28 End of FMT +7.269 94.166 180.00 451.3
    2014/3/7 19:41:00 2nd Handshake +0.018 93.719 183.54 452.7
    2014/3/7 20:41:02 3rd Handshake -7.540 93.191 184.01 452.1
    2014/3/7 21:41:24 4th Handshake -15.091 92.898 182.19 448.7
    2014/3/7 22:41:19 5th Handshake -22.487 92.939 179.70 442.8
    2014/3/8 00:11:00 6th Handshake -33.046 94.099 174.43 424.9

    Don’t forget the aircraft paths between these positions ARE NOT STRAIGHT, so they are not accurately represented by either rhumblines or geodesics. You have to do a step-wise path integral over the leg to get the correct path length.

    I re-read the DSTG description of the BTO errors. My memory was faulty about the BTO drift. It does occur, but apparently over 24 hours it is small enough to essentially ignore. You can see the BFO mean wander about in Figure 5.3. It would have been nice also to see a PDF for each day.

    Does anybody understand why the BTO mean changes over a couple of days?

  306. DrB says:

    @Lauren H,

    Here are my equations for Fuel Flow and Mach relative to LRC as functions of Cost Index:

    https://drive.google.com/file/d/0BzOIIFNlx2aUX3VXZm15bHZ6X2M/view?usp=sharing

    These are required to build an ECON fuel model. They are based on my polynomial fits to the data in the 2 Boeing publications at the link I previously posted.

  307. DennisW says:

    @DrB

    Thx. I have a very specific reason for asking.

  308. Andrew says:

    @DrB:

    Does it not make sense for the FMC to use groundspeed rather than airspeed in turn calculations, given that it is navigating the aircraft according to a specific ground track? In the bad old days of pilot navigation we always used groundspeed when determining turn points, etc. According to ‘The Avionics Handbook’ (Chap 15, Flight Management Systems): “Lateral turn construction is based on the required course change and the aircraft’s predicted ground speed during the turn”.

    http://www.ohio.edu/people/uijtdeha/theavionicshandbook_cap_15.pdf

  309. Andrew says:

    See pages 10 and 21 of the above document.

  310. David says:

    @TBill. “Keep in mind, we do not know if a depressurization actually happened on MH370, and if it did, we do not know if it was intentional or accidental, nor when it might have happened during the flight. There is varied speculation out there.”

    About that and later discussion between you and Andrew, there is negative evidence that this was not done for long periods by shutting off pack air, the aircraft thence unmanned say. FCOM notes that turning off recirculating fans will increase fuel consumption by 0.7%, each, because ventilation from bleed air is increased. The Maintenance Manual quantifies the increase in bleed from shutting down a fan as 67lb/min. What I extract from these two statements is the relationship between bleed air rate and fuel consumption.

    As I mentioned previously at 35,000 ft the pack air supply (including trim air) will be around 255 lb/min (Training Manual). Assuming fuel consumption increase to be about proportional to bleed air flow, shutting off bleed air packs (which also shuts off trim) would lower fuel consumption by above 2½%. That would increase range, so a prolonged decompression would be incompatible with fuel consumption to those termini where otherwise it is spot on.

    Re Nixon and a max 500ft/min cabin altitude rise rate, I have not read the book. It might be a coincidence that the FMC climb mode, usual schedule, limits the rate to that. Opening the outflow valves and turning off the packs would outflank the schedule.

  311. David says:

    @TBill, %th line. To,”….not shutting off pack air”, please add, “(as well as outflow valves)…”

  312. David says:

    @TBill. Trying again. 5th line. To, “…by shutting off pack air”, please add, “(as well as outflow valves)…”

  313. Gysbreght says:

    @Andrew: “Lateral turn construction is based on the required course change and the aircraft’s predicted ground speed during the turn”.

    Thank you for the link to The Avionics Handbook. Turn radius at a bank angle is a function of airspeed. Groundspeed varies with track angle in the presence of wind. Therefore the description on pages 10 and 21 would imply that the FMS commands a variable bank angle during the turn to follow the planned circular ground track.

    Back in January 2016 (on the JW blog) Oleksandr drew attention to the holding pattern of EY440 on January 7, 2016. In that pattern the airplane maintained a constant bank angle in the turns, and the radius changed with the wind component.

    So does the FMS logic for track changes enroute differ from that for a racetrack holding pattern?

  314. Gysbreght says:

    For some reason the embedded link does not work. Here it is in the old-fashioned way: https://www.dropbox.com/s/zxfaxzmt9eqvhl8/EY-440-UTM47.png?dl=0

  315. sinux says:

    @DennisW
    Interesting!
    Correct me if I’m wrong, but an other approach to what you are doing is to say that the average distance from the plane to satellite (or GES) has to increase at 10m/s between 19:41 and 20:41 (35975m /3600s).

    If the plane is flying straight and level and at constant speed then this will be the same as averaging the runaway speeds at 19:41 and 20:41 respectively.

    Does it still work in spherical geometry? or if the plane is turning or on a curved path?

    To bring it back to the way you are calculating :
    For which types of flight path is the average Doppler compensation equal to the average of deltaFcomp1941 and deltaFcomp2041? What assumptions must we take for this to be true?

  316. Gysbreght says:

    @DrB: “Here are my equations for Fuel Flow and Mach relative to LRC as functions of Cost Index”

    Perhaps it would be prudent to add that the equations are valid for the range of CI shown on the graph.

  317. sinux says:

    @DennisW
    As you mentioned, what you are doing is extremely sensitive to the heading. DrB gives you a track of 183.54° for 1941. How do you estimate the heading to be 183.0 ? Same for 2041 he gives you 184.01 and you use 182.0?

    To recap :
    1. Is deltaFcomp really varying linearly (for which assumptions)?
    2a. How did you derive the headings?
    2b. Is the SDU using the track or the heading to calculate deltaFcomp?

  318. DennisW says:

    @sinux

    “Correct me if I’m wrong, but an other approach to what you are doing is to say that the average distance from the plane to satellite (or GES) has to increase at 10m/s between 19:41 and 20:41 (35975m /3600s).”

    That is correct. However, after sleeping on it, I suspect that the contribution of satellite motion to the BTO’s is not insignificant relative to small BTO differences. The approach definitely needs more work.

  319. DennisW says:

    @sinux

    I had estimated DrB’s headings earlier using a graphic he had posted. I simply overlooked changing them with the values he provided. When I use DrB’s provided headings the average Doppler drops to around 37Hz which is very near the value calculated for other paths i.e. McMurdo or Cocos paths. My conclusion at this moment is that the satellite motion is not negligible relative to the BTO difference value of 240 usec, and affects all paths between the 19:41 and 20:41 range rings by about the same amount, ~20Hz.

  320. sinux says:

    @DennisW
    You’re right, it’s not insignificant. It’s already included in the BTO measurements. But not in deltaFcomp. All you need to do is to compute a deltaFcomp’ that is calculated for the actual satellite position instead of a fixed one.

    To answer my question 2b above :
    SDU uses IRS label 314 as an input (and not 313).
    As per page 21 of :
    https://aerocontent.honeywell.com/aero/common/documents/myaerospacecatalog-documents/BA_brochures-documents/Laseref_VI_FINAL.pdf

    label 314 is True Heading (Primary).

    To help visualize your approach, one can imagine a string going from the plane via the satellite to the GES. The speed at which the string is unrolled at the GES should be 10m/s on average (between 1941 and 2041). If that speed varies linearly between those times, one can calculate the theoretical speed at 1941 and at 2041 from a proposed path and check if the average is right.

  321. sinux says:

    @DennisW
    As per 2b above.
    I first thought you were back calculating the heading of the plane from the wind data provided by DrB. Which is what one should do as the SDU is using heading and not track to compute detaFcomp. So you might have been not far off actually 😉
    Now we just have to hope that the wind was relatively constant at the ping times…

  322. Andrew says:

    @Gysbreght:

    “Turn radius at a bank angle is a function of airspeed. Groundspeed varies with track angle in the presence of wind. Therefore the description on pages 10 and 21 would imply that the FMS commands a variable bank angle during the turn to follow the planned circular ground track.”

    Correct. If there’s a large track change in the presence of a strong wind, the FMC should change the angle of bank during the turn to maintain a constant radius of turn with respect to the ground.

    “So does the FMS logic for track changes enroute differ from that for a racetrack holding pattern?”

    No, the FMC will still attempt to fly a race-track pattern with constant radius turns with respect to the ground, up to a point. According to Honeywell: “The holding pattern turn radius is calculated by assuming a 25° bank angle at a groundspeed equal to the true airspeed equivalent of the FMS holding command speed, plus the absolute magnitude of the wind vector.” The FMC then varies the bank angle during the turns to maintain the calculated turn radius.

    “Back in January 2016 (on the JW blog) Oleksandr drew attention to the holding pattern of EY440 on January 7, 2016. In that pattern the airplane maintained a constant bank angle in the turns, and the radius changed with the wind component.”

    My guess is the aircraft was heavy for the altitude it was at and the FMC limited the bank angle to avoid exceeding the available thrust limit. In that case, it would have flown the holding pattern with a constant bank angle at each end of the pattern to keep the pattern as small as possible.

  323. Victor Iannello says:

    @Sinux: To be clear, the true track angle is used for Doppler compensation in the AES, not heading.

  324. Victor Iannello says:

    @Andrew: Thank you for your knowledgeable responses. I hope we don’t exhaust you with questions. I really appreciate the contribution you are making here.

  325. sinux says:

    @VictorI
    Do you have any reference for that?

    Please see the Honeywell document linked above :
    label 313 is “Track Angle True”
    label 314 is “True Heading (Primary)”

    from MSC7200 manual page 5-63 L(2) : […] are connected to an ARINC 429 source of IRS label 310, 311, 312, 314, 324, 325, and 361 information (although label 361, Inertial Altitude, is not required for SATCOM).

    @DennisW
    DrB’s route is CTH so heading will be constant at 180 (pending the above)… which gives me a 84Hz average on his route (with static satellite).
    Perhaps it’s easier to calculate a BTO’ for each time. BTO’ being the hypotetical BTO that would have been recorded if the satellite was at its fixed point.

  326. TBill says:

    @Victor
    On the BFO calc how do we calc the AES Track?

    @David
    Thank you for the fuel consumption info on bleed air. I am trying to recall if that’s directioanlly what I was expecting, or the opposite of what I was expecting.

  327. Andrew says:

    @Victor: You’re welcome!

    @Gysbreght: The other possibility for the shape of the EY440 holding pattern is that the aircraft’s weight and altitude meant the calculated size of the pattern exceeded the ICAO protected airspace limits. The FMC would then have used a constant bank angle of 25° at each end of the pattern in a bid to constrain its size.

  328. Victor Iannello says:

    @sinux: Page A-27, Table A-10, “ARINC 429 Data Requirements”. Label 313 is track angle and is available to the SDU with a minimum rate of once per 55 ms.

  329. Victor Iannello says:

    @Andrew said, “the calculated size of the pattern exceeded the ICAO protected airspace limits”.

    I think we often neglect this requirement when thinking about flight paths. I suspect that is why we see the overshoot for the IGOGU-ANOKO turn.

  330. Victor Iannello says:

    @TBill asked, “On the BFO calc how do we calc the AES Track?”

    I’m not sure I understand your question. For our BFO calculations, we assume that the AES is using the true track in its calculations. That data is available to it over the ARINC 429 bus. On the other hand, heading is probably used for antenna steering (although I doubt that antenna pattern of the HGA is narrow enough that it makes much difference if heading or track is used for steering).

  331. TBill says:

    @David
    Less bleed air increases thrust, I thought.

  332. sinux says:

    @VictorI
    Thanks for the reference! However this applies to the SCU (Signal Conditioning Unit). Which does the following :

    “The MCS system requires ARINC 429 data for antenna pointing, antenna stabilization, and Doppler frequency correction. These requirements are defined in Table A–10. If the aircraft does not have an IRS that supplies this ARINC data, the SCU can be used to supply the data.”

    We don’t know if one was used. But even if it was, the fact that the SCU uses label 313, doesn’t mean that the SDU will. As per my original quote above, it doesn’t…

  333. TBill says:

    @Victor
    OK true track. Thank you.

  334. Victor Iannello says:

    @sinux: Table A-10 defines what inertial data is required by the SATCOM, whether or not it is supplied over ARINC 429 or using an SCU. We know that the SATCOM on a B777 is supplied the inertial data over ARINC 429, and so the SCU is not necessary. Either way, Table A-10 says that track data is required at an update rate between 22 and 55 ms. I don’t know why label 313 was not included in the excerpt you cite, but considering that is included in Table A-10 along with technical requirements for update rate, I think it is very unlikely that it is not used by the SDU to calculate Doppler compensation.

  335. DrB says:

    @DennisW,

    Here are the average True Headings for my CTH-BH route for the same legs as the previous table I posted:

    Date UTC Average True Heading for Previous Leg

    2014/3/7 18:43:28 180.33
    2014/3/7 19:41:00 181.31
    2014/3/7 20:41:02 181.31
    2014/3/7 21:41:24 181.31
    2014/3/7 22:41:19 181.31
    2014/3/8 00:11:00 181.31

  336. sinux says:

    @VictorI
    It’s a shame we can’t even have consistent manuals 😉
    Is it perhaps possible to distinguish which is used with the sparse data we have? Before the ADS-B stopped transmitting?
    It may have a very small impact around the Malaysian Peninsula as the winds are not that strong. But the more south it goes, the bigger the impact. Specially as it’s cross track in this area.

  337. sinux says:

    @DennisW
    With the latest figures from DrB above : 68Hz … Getting closer!

    @DrB
    How do you explain the change from 180 to 181.31° ? I thought the plane was entering a Holding pattern at 180 and staying on that heading upon EOO and EOR errors. What did I miss?

  338. David says:

    @TBill. “Less bleed air increases thrust, I thought”. Bleed air costs fuel consumption in that the engine compresses it with other non-bypass fan flow but since it is then diverted to bleed gets no ‘return’ of that energy when the rest of the non-bypass portion is expanded in the turbines. Turning the bleed off therefore will increase thrust a fraction and auto-throttle can adjust fuel and revs down a touch.
    Come to think of it though I know not what offsetting drag increase would result from the outflow valves being wide open. It would be a coincidence if the two about balanced though.

    PS I gave up. “(as well as opening outflow valves)” it should have been as you will have guessed.

  339. TBill says:

    @sinux
    I do not want to answer for DrB but it’s close enough to 180 to wonder if 180T could be made to work as a version.

  340. Richard says:

    There has been discussion of the MH370 flight path up to 19:41 UTC and whether MH370 approached the 2nd Arc from the inside or outside.

    Brian Anderson answered this question with his observation above, where he concludes from the inside: “The logic is that all BTOs were decreasing over time to the 19:41 BTO, and then increasing over time from 20:41.”

    Brian goes on to estimate the point of closest approach to the satellite at approximately 19:52 UTC and then to estimate the ground speed using spherical geometry as approximately 494 knots.

    Based on Brian’s observation, here is a paper defining the range of the MH370 End Point:

    https://www.dropbox.com/s/pdyzruozkzgr1t1/The%20End%20Point%20of%20MH370.pdf?dl=0

  341. Richard says:

    CORRECTION: “concludes from the outside” or from the East

  342. Victor Iannello says:

    @sinux: In @DrB’s path, the track is 180° at the EOR, not the heading. Whatever the heading is then held constant.

  343. Ge Rijn says:

    @Richard @Kenyon

    Interesting paper. Besides the always arguable assumptions it’s based on, it independently fits with some earlier and the latest drift-analysis crash areas.
    Between 28.5S and 32S fits very well with the approach you took based on the Inmarsat data and possible FMT’s independent from those drift analysis.
    They end up in the same area.

    I think this is the way to go. Possible flight paths have to fit this drift analysis based area. I think we have a rather well defined end-area now which can serve as a ‘calculating back’ area for a most probable FMT.

    @Kenyon

    The IFE-casing is the clamped-on exterior casing. No baseball bats needed to seperate them.
    Do some research on MH17 seat-pictures (I warn you for those pictures can be horrifying)

  344. Victor Iannello says:

    @Richard: Thank you again for your hard work. You’ve provided more for us to think about.

  345. TBill says:

    @Richard
    I like SANOB at 3394S because lately I’ve been thinking of “user-defined” Waypoint paths like 1094E to 3394E. Maybe 1094E start would be True Heading analogous to DrB path, whereas I’d have to see if the winds would blow MH370 over to 93.xx which is the eastern extent of Arc2. If its Track or Waypoint then I need 1093E start to intersect Arc2.

    Anyways just saying it is so easy to enter those “user defined” waypoints, I certainly do not hesitate to consider that technique might have been used. So there is possibly an added set of Waypoints for your analysis (sorry but maybe there are not a whole of them?)

  346. TBill says:

    EDIT: maybe there are not a whole *lot* of them?

  347. Richard says:

    @TBill

    You are absolutely right and I will extend the study to include the user defined waypoints.

    @Victor

    Thank you for your hard work providing this excellent forum, full of outstanding ideas from great people.

  348. Richard says:

    @TBill

    Here is a link to an updated table of results:

    https://www.dropbox.com/s/s3ijs8cwjjbbfan/MH370%20Flight%20Path%20Model%20V16.0%20RG%20Table%20V2.png?dl=0

    Despite there being 7 user defined waypoints additionally identified (0498E, 0496E, 0696E, 0895E, 0894E, 1296E and 1592E), as you will see from the table, the MH370 End Points either give a high residual error for the BFO satellite data or are firmly in the area already searched by the ATSB.

  349. DrB says:

    @sinux,

    You said: “How do you explain the change from 180 to 181.31° ? I thought the plane was entering a Holding pattern at 180 and staying on that heading upon EOO and EOR errors. What did I miss?”

    At the end of the FMT (~18:43), the track is 180.000 degrees true. The heading is 180.33 degrees because there is a slight right cross-wind.

    At 18:44:32, the EOR occurs. The wind then is a little bit stronger right cross-wind, so the true heading then was 181.31 degrees. The FMC now begins to hold that true heading constant for the remainder of A/P powered flight.

  350. DrB says:

    @Richard,

    Just to make it clear, my CTH route does not comply with two of your assumptions: (1) waypoints (without offsets) were used, and (2) the path was great circles. It is therefore not surprising that my predicted terminus does not fall within your “high probability” range.

    In addition, my CTH route also does not comply with Brian Anderson’s assumptions: “One substantive assumption is necessary: that the aircraft was travelling in a straight line between these time, or nearly so. A slightly curved track would result in a greater distance being travelled, and hence a slightly greater speed being appropriate (than the lesser speed obtained from the method described here). Of course it is also necessary to assume a more-or-less constant speed between these times/between these ping rings.”

    Again, my CTH route is not straight nor is does it have a constant speed. Therefore Brian’s results do not apply to it.

    I have found it best to minimize assumptions. I think the only one I have left now is that the route be flyable via auto-pilot and auto-throttle after 18:22. I’m not counting BTO/BFO as an assumption here because without those we have nothing.

  351. Kenyon says:

    @Ge Rijn

    “The IFE-casing is the clamped-on exterior casing.” Disagree

    “Do some research on MH17 seat-pictures” I’ve spent countless hours studying photos including MH17.

    I’m tapping out of this discussion, I tried, we simply and respectfully disagree.

  352. TBill says:

    @Richard
    Thank you I am impressed so fast.
    Hmm no hits there.

  353. Mick Gilbert says:

    @DrB

    In your earlier discussion with Brock you posted a post-FMT winds diagram (March 12, 2017 at 2:42 pm). If it’s not too much extra work for you could I trouble you for that diagram with wind strength and direction for each vector labelled on the diagram please?

  354. Brock McEwen says:

    @Richard: I join the chorus thanking you for your continued efforts. I respect those who refuse to give up.

    To help me understand what you are doing, I tried to recreate one of your paths. I picked SANOB. I started at your 00:19 end point, and worked backward through the arcs at 494 KGS. My path was great-circle (if I squinted), and ended up intersecting the 19:41 arc at [4.2N, 93.7E] – so I reconcile to your path, to within the approximation error of my own personal “Google Earth Picasso”.

    But even if I back up to your 4.0N, I am only 217 nmi crow’s flight distance from what you’ve used as the last known radar position (6.5774°N 96.3407°E at 18:22, 79 minutes earlier. Even if I instead connect 18:22 to 19:41 in the way I thought you would have done – heading almost due east from 18:22 to SANOB, then continuing east on that same heading until ~[6.64°N, 93.577°E], then turn south onto a path that dovetails perfectly into the 494 path at [4N, 93.7E], I still get a path only 300nmi long. 300nmi in 79 minutes is only 228 KGS.

    What am I missing?

  355. Brock McEwen says:

    @DrB: re: “I’m not counting BTO/BFO as an assumption here because without those we have nothing.”

    I do consider the BTO/BFO values an assumption. That’s why I’ve been TESTING it. My guiding principle: “if these values are authentic, here are some other things we should expect to see; do we see them?” It tends to fail these tests.

    Also interesting is the stack of foul-smelling “corroborating” evidence that oozes out from he shadows at regular intervals, only to be robustly debunked upon actual scrutiny. It smacks more of opinion management than an honest effort to unearth and assess helpful evidence.

    So if we did ever reject the signal data, we’d not only have more than nothing; we’d have more than we have now: a logical suspect, in the form of the state(s) trying to sell us the signal data.

    I’m not there yet – all I want to do is force investigators to demonstrate a good faith search, via full transparency. But the conclusions of the most exhaustively researched publications to date seem to be there already.

  356. DennisW says:

    @Richard

    My personal opinion is that the speed was lower at 19:41 than 494 knots. 494 knots is hard to reconcile with the BFO value for reasonable headings.

  357. DennisW says:

    @Brock

    I believe the BFO and BTO data are legitimate as reported. I do not agree with the statistical inferences that are continuously made relative to this data. Neither BFO nor BTO are Gaussian distributed (I do not know why the BTO is not Gaussian.) BTO should be very good. The “fur” in the tails of the error histogram is a mystery to me.

    There is merit in Victor’s view that the outliers in BTO are rare enough to be reasonably discarded. Still. I do not like it because I cannot explain it.

  358. DrB says:

    @DennisW,

    Suppose the BTO outliers are caused by good BTO measurements coupled with bad BTO predictions (caused by errors in the assumed aircraft location)?

    If DSTG used ACARS GPS position data (spaced some minutes apart), and a turn occurred between the time of the last position reading and the SDU handshake, DSTG would have some position error when predicting the BTO (even if they tried to interpolate or extrapolate using ground speed). That will create an apparent outlier in BFO error. If you think about minor turns at waypoints along airways, a few of these large BFO errors could occur every flight.

    I have always ignored the large BFO errors, assuming DSTG didn’t always get the correct aircraft lat/lon.

  359. Richard says:

    @TBill

    I have updated the paper with your kind suggestion to include user defined waypoints.

    https://www.dropbox.com/s/pdyzruozkzgr1t1/The%20End%20Point%20of%20MH370.pdf?dl=0

    If you would like an acknowledgement in the paper, please send me your name to r.godfrey@web.de and once again many thanks.

  360. Richard says:

    @DrB

    I agree with you that we should try and minimise assumptions. I am testing various sets of assumptions and in this particular case adding Brian Anderson’s observation to the list of assumptions.

    I very much appreciate your work on the SDU OXCO and demonstrating a fit to a Constant True Heading and your desire to reduce assumptions. As you say yourself, (1) you have to define an aircraft starting point (or aircraft prior), (2) a continuous path within the B777-200ER Trent 892 performance envelope, (3) a fit to the BTO and BFO satellite data and (4) the fuel range and endurance.

    I agree with your statement “I have started fitting at 17:21. Regardless of method, at some point you still have to demonstrate that a post-19:41 path is (1) reachable by a continuous, flyable connecting path that is (2) consistent with the BTO/BFO/radar data prior to 19:41 and (3) there is enough fuel to fly until 00:17.”

    In my view, a MH370 End Point around 33.9°S 94.2°E (just after 00:19:37 UTC) is unlikely as this area has already been searched by the ATSB to a level of 97%. This leads me to want to test the assumptions.

  361. Richard says:

    @Brock

    You ask “What am I missing?”

    The short answer is a loiter.

    The long answer is:

    Let us take your calculation that at 19:41 UTC the aircraft was at 4.2°N 93.7°E.

    The BFO at a Ground Speed of 494 knots is satisfied by a track of 216.2798°T from outside the 2nd Arc coming from the East.

    Following this track back, it passes within 12 nm West of waypoint SANOB.

    So SANOB is a possible candidate waypoint, albeit 185.4 nm from 4.2°N 93.7°E and therefore a more marginal candidate than other waypoints closer to the 2nd Arc.

    In order to satisfy the time difference between 18:22:12 UTC and 19:41:03 UTC of 78 mins 51 secs, it is not possible that the aircraft flew directly from the last known position to waypoint SANOB, as these points are only 40.1 nm apart. As you point out, the overall speed is too slow. A significant excursion toward waypoints such as IGOGU or ANOKO must have taken place, before returning to waypoint SANOB.

    Mick Gilbert has outlined a number of possible holding patterns in this area, that fit the BTO and BFO satellite data. Victor has pointed out a possible holding pattern in the Car Nicobar area.

    That is why I drew the 2 red circles on the Google Earth map in my paper. The larger red circle makes it obvious that even at a maximum speed of say 511 knots, the aircraft can travel a long way between 18:22:12 UTC and 19:41:03 UTC. The smaller red circle shows it is possible to almost reach the 2nd Arc at 18:41:03 UTC, so you could end up having an hour to kill.

  362. Victor Iannello says:

    @DrB said, “I think the only one I have left now is that the route be flyable via auto-pilot and auto-throttle after 18:22.”

    If you consider that you can enter speed restrictions for a waypoint, that is not as constraining as you might imagine. In fact, it is possible to fly with no pilot intervention starting at 18:02 (with no pilot intervention for the lateral offset) and meet the BTO, BFO, and fuel requirements until fuel exhaustion. Do I think that occurred? No. I suspect there was pilot input after 18:22.

  363. TBill says:

    @Richard
    Thank you for the paper and your efforts…you have already acknowledged.

  364. DrB says:

    @Richard,

    I don’t agree the area along the 7th Arc near 34S has been “searched by the ATSB to a level of 97%”, because the search distance normal to the arc is reduced in that area, particularly so inside the arc. By my estimate they have covered about 5 NM inside the arc. It is quite possible the aircraft is farther inside the arc.

  365. Richard says:

    @DrB

    You state above that you estimate at 00:11:00, that MH370 was at 33.046°S 94.099°E tracking 174.43° at 424.9 knots.

    Even slowing to 377.7 knots by 00:19:29 and 375.2 knots by 00:19:37, I estimate your MH370 end point is around 33.9°S 94.2°E. This is not 5 nm inside the 7th Arc, but just 0.35 nm inside the 7th Arc.

    According to Richard Cole’s maps, this area has definitely been scanned:

    https://www.dropbox.com/s/hd435m7o4286vhq/Richard%20Cole%20Map.pdf?dl=0

  366. TBill says:

    @Richard @Victor
    I am thinking the terminology for those whole number waypoints are sometimes called “Oceanic” although I am thinking they also work on land.

    Key new point for me: I just commanded FS9 PSS777 to fly to:
    N0645.0E09315.0
    which the FMC readily accepts and goes to. This answers my question if a specific exact target could be entered (YES). Relax though, I shall not ask you to calculate infinite waypoints.

    One theory I arbitrarily adopted was a possible crash site in a target undersea hard-to-find location. That is part of what I am trying to explore with Flight Simulator. So now I see that answer is “yes” it is easy. Previously I learned how to place the FS9 flight on Google Earth with the undersea bathymetry map, and now I see how to intentionally fly to any feature I see on Google Earth.

  367. Victor Iannello says:

    @Richard: Of the many paths I have reconstructed, the one that seemed most probable to me is BEDAX-SouthPole, ending at 34.2S along the 7th arc, which I proposed in July 2014. The fit to the BTO and BFO is excellent, and the plane would be in LNAV until fuel exhaustion, which makes the terminus relatively well-defined. I was not bothered by the need for a loiter to the north of Sumatra (others were). My hopes were dashed when GO Phoenix scanned the area back in early winter of 2014, and I stopped working on this path. Drift studies like yours also seem to be pushing things further north. Also, the interpretation of the BFO data at 00:19 suggests a terminus close to the 7th arc, which decreases the probability that the area scanned was not far out enough from the 7th arc. So, my take is that @DrB’s terminus is not impossible, nor is the BEDAX-SouthPole route. However, in my opinion, at this point, there is stronger evidence suggesting a terminus that is further north.

    One conclusion that I think is inescapable is that the plane did not fly at speeds close to LRC from 17:21 to fuel exhaustion, even if the PDAs are not as bad as Bobby believes. To me, this increases the chance of a descent to a fuel-saving holding pattern (which also explains the BFO values at 18:40) OR an extended flight after 19:41 at speeds much less than cruise speeds (as Bobby suggests).

  368. Brock McEwen says:

    @Richard: thank you for clarifying.

    If distance covered in those 79 minutes doesn’t even matter in this process, then by what logic do you deduce a 4°N intersection at 19:41 from a “to SANOB” bearing at 18:22?

  369. TBill says:

    @Victor
    That is valuable background info re: BEDAX to South Pole and how you now see it based on that perspective. In general seems to me your BEDAX path and DrB current path point to ~34S as the “correct” place that the orig search should have been centered in hindsight based on BTO/BFO and maybe fuel. So that sort of supports the NOK thought of giving the 32.5 to 36S area a wider search area in case there was a way for the plane to deviate from Arc7 presumably to the inside as DrB points out.

  370. DennisW says:

    @Richard

    “The BFO at a Ground Speed of 494 knots is satisfied by a track of 216.2798°T from outside the 2nd Arc coming from the East.”

    What altitude (distance from earth center) are you using for the imaginary stationary satellite at 64.5E?

  371. Victor Iannello says:

    @TBill: Several drift analyses also support a terminus further north.

  372. Richard says:

    @Dennis

    What altitude (distance from earth center) are you using for the imaginary stationary satellite at 64.5E?

    42,186.932 km

  373. Ge Rijn says:

    @Kenyon

    I respectfully agree to disagree on the IFE-casing.
    Although I’m sure you are wrong.
    But it’s not alone about this casing. It’s more about +90% trailing edge, surface control, left and right wing related, flap fairings, engine cowlings, nose gear door pieces found. Their sizes, damage and positions on the plane. The small amount of debris found so far.
    Everything just points to a relatively low speed horizontal level impact.
    To me these are facts by now. There was no high speed dive-like impact.
    The plane must have been pilot controlled till the end IMO.

    IMO efforts should be undertaken to fit the Inmarsat-data to such a scenario.

  374. DennisW says:

    Are you aware of the statement made on page 29 of “Bayesian Methods…” referring to the AES Doppler compensation algorithm?

    “The compensation also assumes a motionless satellite at its nominal satellite location of 64.5◦E. The satellite altitude used in the correction is 422 km higher than the nominal 35788.12 km value.”

    The 35788 km is from the surface of the earth. From the center of the earth you get 42581 km.

    It makes about a 1.6 Hz Doppler comp. difference at 19:41.

  375. Richard says:

    @Brock

    If distance covered in those 79 minutes doesn’t even matter in this process, then by what logic do you deduce a 4°N intersection at 19:41 from a “to SANOB” bearing at 18:22?

    I work backwards from 4°N on the 2nd Arc, knowing that the instantaneous track at 19:41:03 UTC that gives a residual BFO of zero is 216.3123°T given a ground speed of 494 knots.

    The reverse of 216.3123°T is 36.3123°T, so I look along the line from the 2nd Arc at 36.3123°T to see which waypoints I hit. SANOB is a close hit.

    Of course, this assumes that the track is not changing, that is why the process is better for waypoints close to the 2nd Arc, unlike SANOB.

  376. Brock McEwen says:

    @Richard: I think I have it now; thanks again.

    The implications are

    A) as we go back in time from Arc7, the great circle assumption ends at precisely Arc2, because I’ve confirmed your track from Arc2 to Arc7 is ~180°, not 216°.

    B) all post-18:22 loitering is thus presumed to occur prior to hitting the listed waypoint – an assumption that fits a little more awkwardly than usual, as you say, for my randomly selected sample.

    If so, I guess I’m left wondering about the intrinsic reasonableness of a ~35° left turn at 19:41. A turn is not required by the upstream paper on which your work builds, if I read it correctly. As such, there seems no reason to add this extra turn in, other than that is what is required if one arbitrarily chooses to minimize BFO error in two silos: the 19:41 error singly, yet all the rest jointly.

    I think it might greatly facilitate our assessment of each of your paths if you could add two columns to your table:

    1) number of degrees of turn required at 19:41
    2) number of minutes each path must* spend in a loiter between 18:22 and 19:41

    (* assuming 494 KGS from 18:22 onward, for ease of computation and comparison)

    Thanks again for considering my feedback.

  377. Mick Gilbert says:

    @Ge Rijn

    There is a persistent misconception that high speed impacts with water liberate large amounts of floating debris whereas low speed impacts have less debris. A review of accidents where airplanes have flown into water suggests that, in fact, the opposite is true; high speed impacts produce relatively less floating debris.

    Consider three well documented high speed impacts into water:

    – Adam Air Flight 574, a Boeing B737 flying between the Indonesian cities of Surabaya and Manado crashed at high speed into the Makassar Strait near Polewali in Sulawesi on 1 January 2007. In all, only 194 pieces of floating wreckage were recovered.

    – Flash Airlines Flight 604, a Boeing 737-300 crashed into the Red Sea at high speed after taking off from Sharm el-Sheikh International Airport, Egypt on 3 January 2004. Only 55 pieces of floating wreckage were recovered.

    – EgyptAir Flight 990, a Boeing 767 that crashed at an impact speed of over 1000 km/h into the Atlantic Ocean about 100 km south of Nantucket Island, Massachusetts on 31 October 1999, also produced very little floating debris, less than 200 pieces in all.

    Constrast those with a notable lower speed impact into water:

    – Air France Flight 447, an Airbus A330, enroute from Rio de Janeiro to Paris which crashed into the Atlantic Ocean on 1 June 2009. The impact speed was estimated to have been around 200 km/h. Searchers located the first piece of floating debris on 5 June 2009. In all well over 700 pieces of floating debris were recovered.

    You’ll note that the recovered wreckage falls into two broad categories;

    First there are the relatively smaller items that are consistent with a high energy impact occasioning a hull break up. As alluded to earlier, the piece of the vertical stabiliser is instructive here. The vertical stabiliser is one of the most robust assemblies on any airplane and is usually found largely intact after water impacts that break up the rest of the airplane (Air France AF447 and Ethiopian Airlines ET961 are but two notable examples – the latter is particularly relevant as it was a controlled ditching attempt). The energy required to break up the tail is significant, pointing to a very high energy impact. Moreover, for that energy to be imparted on the tail, the airplane would need to be either inverted or nearly vertical on impact; neither of those orientations is consistent with a controlled ditching. (OZ214 is neither appropriately analogous nor is the damage to 214’s vertical stabiliser (strike damage occasioning gouging) even vaguely similar to the tail section recovered from from Linga Linga, Mozambique on 26 August 2016.)

    Then we have the three relatively larger items, the flaperon and wing flap sections, that do not show signs of a high energy impact. The pieces have come away from the airplane in such a fashion that they are largely undamaged, apart from damage to their trailing edges (the section of port outboard flap has one piece of strike damage suggesting that it may have struck the forward outboard tip of the horizontal stabiliser). The flaperon is instructive here; it sits immediately behind the engine on a B777. In a ditching the engines will come into contact with the water first offering considerable protection from high speed water for the flaperon. Virtually the only way that the flaperon could be torn away in a ditching is for the engine to come away and strike it. However, there is no strike or impact damage to the underside of the flaperon. It is most likely that the flaperon and sections of flap were torn from the airplane in the sort of high speed dive that would be associated with an uncontrolled descent after fuel exhaustion.

    Returning to the nose gear door, there is simply no logical reason for lowering the gear when ditching; it makes no sense and not only is it not an included item on the ditching checklist, the checklist requires that the gear be confirmed as UP.

  378. Ge Rijn says:

    @Mick Gilbert

    None of the identified items shows significant crush-damage only tension-damage. The honeycomb in all pieces was structuraly intact.
    Even the nose gear door item. This door would be one of the first items to impact the water. In a high speed dive-like impact this door would have been crushed completely and blown into smitters inwards the hull.

    The fact of no crush-damage on any piece already counts on itself for evidence there could not have been a high speed dive-like impact IMO. Especialy the nose gear door item supports this evidence. I think the lack of crush-damage on this item and the clean undamaged side where it was attached to the hull and its size tell the door was deployed when it hit the water with relatively low speed and ~low AoA.

    I agree there is no logical reason to lower the landing gear in a ditching attempt. But I think we have to consider the pilot (if there was an active pilot) had no intension to survive a ditching.
    Maybe it was done to lower the impact speed? Or to speed up the sinking process?

  379. Gysbreght says:

    @Mick Gilbert: “Returning to the nose gear door, there is simply no logical reason for lowering the gear when ditching; it makes no sense …”.

    Agreed, so that is probably not what happened. However, when an airplane lands on its belly like AF447 did, or the B777 in Victor’s video visualisation of ZAH’s sim point “2016-12-15 45S2 no pilot input.wmv” the impact forces are likely to pull the nose gear out of the uplocks, pushing the doors open.

    Of course, this can also occur under high-g loads inflight. In the Boeing simulations published the ATSB report of 2 November 2016 only those with A/P loss at first engine flame-out were likely to have produced high g forces in flight.

  380. Ge Rijn says:

    @Gysbreght

    With China Airlines Flight 006 the plane pulled ~5G’s when pulling out of a high speed near vertical dive.
    The landing gear stayed where it belonged but two actuator doors of the main landing gear where lost. The nose gear doors did not detache:

    https://en.wikipedia.org/wiki/China_Airlines_Flight_006

  381. Paul Smithson says:

    @Ge Rijn. Thanks – that is an interesting reference you posted – and I think it demonstrates how it is possible for various parts (gear doors, flaps etc) to be separated before impact from an aircraft that is in uncontrolled descent.

  382. sinux says:

    @Ge Rijn, @Gysbreght,
    From the manual :
    “The forward doors of the nose gear wheel well operate hydraulically during gear retraction and extension. The aft doors operate by mechanical linkages that connect to the nose gear. The aft doors close only when the gear retracts.
    The nose landing gear uses center hydraulic pressure to retract and extend.”

    So it’s possible that with loss of hydraulic pressure once APU stops, and high G’s, only the forward doors will open.

  383. Mick Gilbert says:

    @Ge Rijn

    When an airplane makes a high speed, high angle of entry impact with water it isn’t crushed, it’s blown apart. Have a read of the accident investigation report on Adam Air Boeing 737–4Q8 PK–KKW, Flight 574;
    from Section 1.12 Wreckage and impact information:

    “The high speed and steep impact angle are considered to have generated a huge pressure wave in the inside of the aircraft fuselage causing the fuselage to explode. This phenomena would have similarly affected other major aircraft structural components. The combination of the pressure wave, and high kinetic energy at impact, caused the aircraft to break up and become fragmented.”

    You’re obviously convinced that the airplane was ditched; I don’t believe that there is one iota of evidence to support that, be it physical wreckage or satellite data.

  384. Ge Rijn says:

    @Paul Smithson @Mick Gilbert @Gysbreght

    Have to mention no flaperon, outboard flap sections, trailing edges or other wing related parts seperated during Flight 006 high speed dive and pull out.
    Mainly elevator parts where ripped off due to excessive loads during the pull out. Not by flutter. The same loads caused both wings to be pushed 5cm upwards permanently. Nothing seperated by flutter.

    @Mick Gilbert

    I’m not convinced the airplane was actively ditched. I’m only convinced the plane entered the water ~wings level with relatively low speed and low AoA. I regard a AF447 kind of impact could be possible also but such an impact would have caused a lot more floating debris (as you previously mentioned) which would have resulted in a lot more pieces beeching and found by now.
    IMO a lot of debris-evidence now is pointing to some kind of ditch-event.
    I think it’s unwise to keep ignoring this evidence and with it refusing to consider the possibility.

  385. Mick Gilbert says:

    @Ge Rijn

    We need to be careful when looking at potentially analogous accidents to ensure that we are in fact looking at comparable events. The damage to China Airlines flight CI006 was not due to high airspeeds (it is unlikely that it exceeded 255 KIAS) or high rates of descent (did not exceed about 19,000 ft/min), it was due to high G loadings (two separate 4.8G and 5.1G loads).

    There was no aerodynamic flutter with CI006 because the control surfaces remained powered and stabilised throughout the various excursions from normal flight. On fuel exhaustion MH370 would have lost power to the left and right hydraulic circuits (the centre circuit would have been powered by the Ram Air Turbine). Without hydraulic power, the Power Control Units (PCUs), the servos that move and maintain the position of the control surfaces, automatically go into by-pass mode, allowing full and free movement of the attached control surface. Thus, the loss of two hydraulic circuits would have left control surfaces either completely or partially free to move under the influence of relatively high, and almost certainly fluctuating, airspeeds and rates of descent.

    It is generally unwise to consider items of evidence in isolation; when viewed together the physical wreckage and the satellite data suggest a high speed/high rate of descent impact.

  386. DrB says:

    @Richard,

    You said: “Even slowing to 377.7 knots by 00:19:29 and 375.2 knots by 00:19:37, I estimate your MH370 end point is around 33.9°S 94.2°E. . This is not 5 nm inside the 7th Arc, but just 0.35 nm inside the 7th Arc.”

    You are being silly now. Anyone who thinks they can predict the 00:19 position from the 00:11 position and speed with sufficient accuracy to distinguish between 0.35 NM and 5 NM inside the 7th arc is vastly overstating the case. VERY UNCERTAIN ASSUMPTIONS must be made in order to obtain an intersection with the 7th Arc. You don’t say what your assumptions are, but no matter. They are likely to be wrong. The track from 00:11 to 00:19 is highly uncertain because (1) the speed profile after fuel exhaustion is currently unknown, (2) the exact time of right engine fuel exhaustion is unknown (although likely to be several minutes prior to 00:11), and (3) even the general turn direction after the autopilot stops at 00:17:30 is unknown (but shown to be highly variable by ATSB/Boeing simulations). In short, no one, not even you, can now predict whether or not 9M-MRO ended up 0.3 NM or 5 NM inside the arc.

    A predicted 00:19 location can only hope to identify a local region of the arc to be searched, not a predicted specific location of 9M-MRO wreckage.

    Richard Cole’s map shows that area has been scanned only to about 3 NM inside the surface arc. I would not call that a 97% probability.

  387. Ge Rijn says:

    I like to add I still consider a pilot-controlled kind of ditch-attempt as the most likely. Based mainly on the assunption the flight never turned into a ghost-flight.
    This ghost-flight assumption is not logical at all and not based on any evidence at all. The chance a regular normal passenger flight turns into a ghost-flight along its route must be 1 in many millions of flights.
    The Helios-flight was the very, very rare exception and the only one of this kind in hystory I know of (not counting the Lear-jet).

    IMO the ghost-flight assumption has been the major basic mistake which led to the wrong search area and the conclusions of an uncontrolled end-of-flight scenario. All confirmation bias orientated based on no evidence at all except for the still not well understood and regarded unreliable final BFO’s.

  388. DrB says:

    @Dennis W,

    For my post on March 15 at 1:12 am on BTO outliers, I must have been tired. Obviously the last two references should have been to BTO, not BFO.

  389. DennisW says:

    @DrB

    I took all of the references as to BTO.

    Your comments on the underlying cause of the BTO outliers could well be true. If that is the case, then there would be no BTO outliers for MH370 data.

  390. Ge Rijn says:

    @Mick Gilbert

    You give sensible arguments. Indeed some control surfaces would be in bypass-mode and free floating if only powered by RAT. This would concern the left wing flaperon i.e. But not the found right wing flaperon which remained actuated by one actuator in such a case.

    The outboard flaps are actuated by screw-jacks driven by hydraulic motors. When hydraulic pressure fails they stay locked in the position they are in. They won’t float freely. The outboard flaps are considered to have been retracted and thus locked in place. No way flutter could have seperated them for this reason. And high speed dive impact would not have left those pieces undamaged (especialy their leading edges) the way they were found.
    That’s why both found flap-sections could not have seperated due to high speed flutter or high speed dive impact. You give the arguments yourself in a way regarding Flight 006.

    Also considering the trailing edge damage of those pieces (and the kind of damage to many other pieces and their positions) there is only one solution IMO.
    I won’t repeat myself for I’ve made my points clearly by now I guess.

  391. DrB says:

    @DennisW,

    Furthermore, if DSTG did not precisely fit every turn to the prior flight routes, which I suspect they did not, they would see errors for both BTO and BFO as a result. That would contribute to, but not totally explain, their high value for BFO standard deviation.

  392. Ge Rijn says:

    @sinux

    Interesting info on the nose gear doors. I suspect the forward door actuators will be in actuated-mode provided with centre pressure from the RAT.
    I think it would be a major design flaw if you would allow those door-actuators to go in bypass-mode in such a case.
    In any case you should be able to lower the landing gear under RAT alone and with it those forward nose gear doors deploying without danger of damaging the nose gear.

  393. ALSM says:

    There seems to be confusion here on some of the facts related to the possibility that aeroelastic flutter occurred. Rigid control surfaces held in place by jack screws are not immune to flutter. In fact, rigid structures may be more likely to flutter. The entire wing structure of an airplane can go into flutter. The horizontal stabilizer (fixed part) can go into flutter. So the existence of full or only limited hydraulic power (produced by the RAT) doesn’t tell us much about the flutter theory.

    I first discussed in-flight separation due to flutter in a paper released 2 days after the flaperon was found. After nearly 2 years, and many theories that have come and gone, I remain confident that the very high speed descent indicated by several lines of data and evidence, resulted in some external separations prior to impact (including the flaperon). Some parts may have separated due to classic flutter, while others simply succumbed to the extreme g forces and vibration induced by near Mach 1 speed (like 006).

  394. George Tilton says:

    @DrB
    “Furthermore, if DSTG did not precisely fit every turn to the prior flight routes, which I suspect they did not, they would see errors for both BTO and BFO as a result.”

    What is not appreciated is that if a plane flies exactly the same route at exactly the same speeds but commencing the flight at a different time of day and messages are exchanged at exactly the same points in the flight…both BTO and BFO values will all be different than the prior flight.

  395. Andrew says:

    Ge Rijn:

    The landing gear hydraulic lines are only pressurised during extension and retraction. At other times the system is depressurised and the various components are held in position by mechanical locks. The RAT only provides hydraulic power to the primary flight control components of the centre system; it does not power the landing gear. If the centre hydraulic system is powered by the RAT, then the landing gear must be extended using the alternate extension system. That system has its own hydraulic pump (powered by the hot battery bus) that is used to unlock the landing gear doors as well as the main and nose landing gear. The landing gear then extends under gravity.

  396. Richard says:

    @DrB

    I do not appreciate your rude remark!

    I am sorry you do not believe the accuracy of the Inmarsat satellite data, but that is your problem and not mine.

  397. Victor Iannello says:

    @Andrew: We may have another hint about what Performance Degradation Allowance (PDA) should be using for fuel flow calculations. I’d like your opinion.

    On a page from the Flight Plan labeled “Flight Brief”, it says “FF FACTOR P1.5”, which I interpret as fuel calculations were performed using a +1.5% change (increase) in fuel flow due to performance degradation of the engines. This parameter is similar to the FF parameter from the DRAG/FF line on the IDENT page of the FMC. (Drag and FF need to be separately specified because on parts of the flight, such as during climbs and descents, thrust doesn’t equal drag.)

    On that same “Flight Brief” page, it says “AVG ISA:P6”, which I interpret as average temperature conditions of ISA+6K during the flight. I suspect that this temperature offset is also used to adjust the fuel flow calculations for flight planning, but it is separate from the FF parameter discussed in the previous paragraph. During the flight, the adjustment to calculated fuel flow due to temperature offset would be automatically calculated by the FMC using actual and forecasted conditions.

    If this is true, we would have our answer for the (average) PDA of the engines. It would be 1.5%.

  398. DrB says:

    @Richard,

    You said: “I am sorry you do not believe the accuracy of the Inmarsat satellite data, but that is your problem and not mine.”

    You are missing the point I was trying to make. I believe the satellite data, but it does not tell me (or you or anyone) how to compute the path from the 00:11 position to the 00:19 position. The result is that the 00:19 position is even more uncertain than is the 00:11 position.

    Here’s why this is so. There are effectively three unknowns in determining the 00:19 position (relative to the 00:11 position). These are:

    (1) the BTO error at 00:19 (which defines an arc on which the 00:19 position lies),
    (2) range from the 00:11 position, and
    (3) track bearing from the 00:11 position.

    The range variable is equivalent to the average speed during the leg.

    To assist in solving for these three unknowns, we have only one known parameter value – the 00:19:29 BTO. That means we must assume two of the three unknowns in order to reach a solution.

    I assume these two: BTO error and average speed. For BTO error, I assume it is zero at 00:19. I also assume average speed based on an end-of-flight model. In my opinion, the most uncertain of the three is the track bearing for the 00:11 to 00:19 leg, because the auto-pilot has shut down during the leg and no longer maintains lateral navigation. So the track bearing for this particular leg has significant uncertainty.

    Which two parameters do you assume in order to calculate a predicted 7th Arc location, and why did you choose those two?

  399. DrB says:

    @Victor,

    You said: “If this is true, we would have our answer for the (average) PDA of the engines. It would be 1.5%.”

    If “FF Factor P1.5” means 1.5% PDA, then we would simply know what PDA was used for flight planning. That number may be different than the actual values for 9M-MRO. MAS may, for convenience, simply use a fleet average or a typical value for PDA to avoid the necessity of keeping track of specific values for each aircraft.

  400. DrB says:

    @George Tilton,

    Thank you for commenting on the BTO/BFO errors. You said: “What is not appreciated is that if a plane flies exactly the same route at exactly the same speeds but commencing the flight at a different time of day and messages are exchanged at exactly the same points in the flight…both BTO and BFO values will all be different than the prior flight.”

    I am certainly well aware of that fact as is Dennis. I am sure DSTG is aware of it also.

  401. Victor Iannello says:

    @DrB: Are you aware of ANY airline (especially one with B777s) that uses a fleet average PDA for convenience? If the PDA is too high, you load extra fuel, fly heavy, and waste fuel. If the PDA is too low, you eat into safety margins. Surely MAS is sophisticated enough to track PDAs for individual engines and use that information for planning, flight performance, health monitoring, and optimization.

    Again, I’d be interested to hear what @Andrew says.

  402. DrB says:

    @Gysbreght,
    @Andrew,
    @Victor,
    @Mick,

    I have made the first run of my new End-of-Flight Model. I call it Straw-Man #1.

    You can get it here .

    If you are so inclined, please provide your comments and constructive criticisms. This model is based on numerous inputs and suggestions made previously by others on this blog.

    In particular, I need help in modeling the phugoids at the end. In addition, the EOF Model predicts R engine flame-out very early. If this is correct, then the slow-down must be at a lower rate than the -19 kts/min seen by ALSM in the simulator when starting at a higher speed. Otherwise the 00:11 BFO would show the descent, which it does not for my CTH route. I suspect the slow-down rate is nonlinear, but I don’t know how to characterize it.

    The plot is very busy, but I find it easier to understand if all the parameters are aligned on the same time axis. Caution: you have to “stare at it” for a while before it begins to make sense. The line colors match the label colors, and all fuel parameters are plotted on the right-hand vertical axis.

  403. DrB says:

    @Victor,

    Nothing definite. There was a PPrune post: “To account for aerodynamic degradation, all Live2Air flights are flight planned with a PDA (Performance Degradation Allowance) of 1.2%.” Live2Air is an IFE add-on.

    A Boeing Aero magazine article says “Validate performance degradation for extended twin-engine operations (ETOPS) critical fuel reserves planning (in lieu of the regulatory requirement of 5 percent fuel mileage deterioration allowance).” This seems to imply the airline has a choice.

  404. Brian Anderson says:

    DrB,

    It seems to me that you have manipulated the data to match your desired scenario. Too what end I wonder? For example, others have observed a small descent rate of 200 ft/min at 00:11. The speed reduction upon one engine failing has been observed in a simulator to be linear at 19 knots/min [TAS]. There is no evidence of a heading [or track] change at 00:11, but it seems that you need one in order to reach the 7th arc with your scenario. On the other hand, it may be that the track immediately following the first engine failure remains straight, but the interception of the 7th arc is a result of the aircraft entering a left turn, developing into a spiral. That possibility would seem to mitigate against phugoids, well, a series of phugoids in a straight line at least, because it is of course possible for an initial phugoid to then roll off into a turn. The spiral scenario, with turn rates increasing would seem to more easily fit the BFO observed descent rates.

    I have previously shown that if the speed is too low at 00:11 it is very difficult to reach the 7th arc.

  405. DrB says:

    @Victor,

    Here are a several relevant quotes from the ICAO Procedures for Air Navigation Services – Air Traffic Management Document 4444 dealing with “Special Procedures for In-Flight Contingencies in Oceanic Airspace” :

    15.2.2.3 If prior clearance cannot be obtained, an ATC clearance shall be obtained at the earliest possible time and, until a revised clearance is received, the pilot shall:
    a) leave the assigned route or track by initially turning 90 degrees to the right or to the left. When possible, the direction of the turn should be determined by the position of the aircraft relative to any organized route or track system. Other factors which may affect the direction of the turn are:
    1) the direction to an alternate airport, terrain clearance;
    2) any lateral offset being flown; and
    3) the flight levels allocated on adjacent routes or tracks;

    15-4 Air Traffic Management (PANS-ATM) 12/41/111//0015 No. 4
    b) following the turn, the pilot should:
    3) acquire and maintain in either direction a track laterally separated by 28 km (15 NM) from the assigned route; and
    4) once established on the offset track, climb or descend to select a flight level which differs from those normally used by 150 m (500 ft);
    e) turn on all aircraft exterior lights (commensurate with appropriate operating limitations);

    15.2.2.3.1 When leaving the assigned track to acquire and maintain the track laterally separated by 28 km (15 NM), the flight crew, should, where practicable, avoid bank angles that would result in overshooting the track to be acquired, particularly in airspace where a 55.5 km (30 NM) lateral separation minimum is applied.

  406. DrB says:

    @Brian,

    You said:” It seems to me that you have manipulated the data to match your desired scenario. Too what end I wonder? For example, others have observed a small descent rate of 200 ft/min at 00:11.”

    I’m not sure what “manipulated data” you are talking about. If a particular route has a predicted BFO that does not agree with the measured, you have two choices: (1) assume a climb or descent to fix the disagreement or (2) look for a different route. I don’t view either approach as being superior. They are simply different. Of course, if one has to do this at 19:41 – 22:41 then that probably signifies a route error. However, at 00:11 we simply don’t know if a slight descent was occurring then or not. If you have to add one to make your predicted BFO agree with the measured BFO, that’s OK unless and until there is concrete evidence a descent was not ongoing then. We certainly don’t have that now. That’s why in the notes I pointed out the fact that the 00:11 BFO and the assumed speed slope are related.

    With regard to the magnitude if the INOP speed slope, I also pointed this out. I believe the simulator tests were done at a significantly higher speed that Best Holding. That’s why I asked for help in characterizing how the slope changes with air speed so that I can accurately model it at the lower speed. I doubt it is linear all the way down to 370 KTAS.

    You said: “There is no evidence of a heading [or track] change at 00:11”. There is also no evidence that the track did not change. Remember Duncan Steel: “Absence of evidence is not evidence of absence.”

    You also said: “it may be that the track immediately following the first engine failure remains straight, but the interception of the 7th arc is a result of the aircraft entering a left turn, developing into a spiral. That possibility would seem to mitigate against phugoids, . . . .”

    I agree with the first part. The heading/track should have been controlled until the auto-pilot shut off (at 00:17:29 I believe). Then for the next two minutes a significant turn is possible. That will limit the potential course deviation when averaged over the 00:11-00:19 leg. Still, II don’t think 10 degrees is out of the question. Based on my recollection of what others have said here, I believe that phugoids and a spiral turn are possible together. Perhaps Gysbreght will enlighten us on this point.

    You also said: ” I have previously shown that if the speed is too low at 00:11 it is very difficult to reach the 7th arc.”

    It is not difficult for my CTH route because of the angle with which it intersects the 7th Arc. It has a 00:11-00:19 leg that is 50 NM long with an average air speed of 354 KTAS, ending exactly on the 7th Arc at FL300. Its track is 165.6 degrees true, which is ~9 degrees (roughly 8 miles) to the left of the previous leg extension. Therefore it is not difficult to reach the 7th arc (if you start at 33.0S, 94.1E at 00:11).

  407. Andrew says:

    @Victor
    @DrB

    Victor, I agree with your observations regarding the DRAG/FF factors and the average ISA deviation included on the Flight Brief. Airlines have Aircraft Performance Monitoring programs (APM) that monitor the performance degradation of individual aircraft against the book figures as the aircraft age. APM produces DRAG and FF factors (IDLE/PERF factors in Airbus parlance) that are used by the airline’s computer flight planning system and the aircraft FMCs. The DRAG factor is used to correct the climb/descent predictions, while the FF factor is used to correct the expected fuel flow against the book figures. In my experience, the factors are included in the computer flight plan given to the crew before departure. Both factors are regularly updated and inserted into the FMC by either the flight crew or engineers, depending on airline policy. The average ISA deviation is used only by the flight planning system; the FMC uses the forecast temperature data inserted by the crew with the forecast wind data.

    I can’t speak for Malaysian, but I would be very surprised if they used a fleet average – as you said, it could cause a safety issue in cases where the FF factor was too low, especially on long-haul flights.

    Further reading:

    http://www.cockpitseeker.com/wp-content/uploads/goodies/ac/a320/pdf/data/PerfoMonitoring.pdf

    https://www.researchgate.net/publication/277180673_Aircraft_performance_monitoring_from_flight_data

    http://www.boeing.com/commercial/aeromagazine/articles/qtr_4_06/AERO_Q406_article2.pdf

  408. Gysbreght says:

    @Andrew: Sometimes there is a difference between theory and practice.

    As a former 777 pilot, have you actually seen the Performance Factor in the Flight Plan change from tailnumber to tailnumber?

  409. Gysbreght says:

    @Brian Anderson: “That possibility would seem to mitigate against phugoids, …”.

    Why is that?

  410. Andrew says:

    @Gysbreght:

    Yes. I work for a large international airline and I can assure you the DRAG/FF factors are different from airframe to airframe.

  411. Victor Iannello says:

    @DrB: First, thank you Bobby for your meticulous documentation of all your work. It always impresses me the way you collect feedback, make incremental improvements, and continue to evolve your models.

    I am curious why you chose to use the minimum manoeuver speed as the speed at which descent begins with an engine out and under autopilot. I believe the speed should be the stick shaker speed, which is closer to around 163 KIAS. I believe @Andrew confirmed this based on what he has observed in Level D simulators.

  412. Gysbreght says:

    @DrB: Thank you for offering your Straw-Man #1 for comment.

    I would like top offer following comments for your consideration:

    Note 9: As stated by Brian Anderson, the OEI deceleration observed in the simulations witnessed by ALSM was about 19 kt TAS/minute. The Minimum Maneuvering speed at MH370 ZFW is 208 kt IAS. Holding speed is constrained to 209 kIAS.

    Note 10: I have earlier estimated the rate of descent corresponding to the horizontal deceleration of 19 kTAS/min as 600 fpm. I have since discovered that I under-estimated the acceleration factor at constant calibrated airspeed, which reduces the rate of descent to 480 fpm initially. Please accept my apologies.

    Note 13: At 00:17:29 the Left tank empties and the autopilot disconnects. The steady rate of descent changes from about 480 fpm to about 2000 fpm. The phugoid parameters with wings level could therefore be estimated as initially a rate of descent of 2000 fpm +/- 1520 fpm, average speed equal to the trimmed speed of 209 kIAS, period 82 seconds. These parameters change slowly with altitude.

    Note 14: For a more detailed modelling of the phugoid the bank angle needs to be specified. The ATSB first stated that in simulations of uncontrolled descents with both engines out, the airplane always slowly rolled into a descending left turn at low bank angles. The probable explanation for that behaviour is the asymmetric power supply of the left and right flaperons by the RAT. In November 2016 the ATSB published more recent Boeing simulations for end-of-flight scenarios that the ATSB specified to Boeing but has not yet published. From the results only the trajectories are shown in Figure 6 of the Nov.2, 2016 report. When the tracks are numbered from left to right at the top of the chart, then tracks 1 through 6 (A/P disconnect at 2nd flame-out) all turn left and cross the 7th arc approximately at a right angle. Simulation #3 is the only one in that group that “recorded descent rates that equalled or exceeded values derived from the final SATCOM transmission”.

  413. Gysbreght says:

    To add: Simulation #3 is also the only one where the track exhibits evidence of phugoid motion.

  414. Mick Gilbert says:

    DrB,

    Thank you for the detailed work. It is a busy graphic but it does paint a clear picture. One parameter that is not modelled and that is probably unknowable is angle of bank; I suspect that it would play a not insignificant role in determining the amplitude and period of the phugoid. As modelled, the phugoids appear to add 2 minutes to time aloft or potentially roughly up to a 10 nm shift to point of impact so at the end of the day it may not be worth investigating.

  415. TBill says:

    @DrB
    Nice work and I am not ready to understand end of flight details. But I do have some philosophical ideas. 34S Arc7 was already “thin-slice” searched, so you are actually arguing that the flight ended up in a spot other than Arc7 at 34S. Lately I wonder why we ignor BFO, instead of asking what BFO is saying. Arc6 BFO at 252 I think says about 34S on Arc6 at full speed and 180S heading. If we are slowing down and descending, and I think that shifts location maybe to 32/33S at Arc6 to hit 252 BFO.

  416. George Tilton says:

    @DrB: Your work is impressive.

    I am no slouch when working with Excel.

    But I really do need to get off of Office 97.

  417. Gysbreght says:

    @DrB:

    “With regard to the magnitude if the INOP speed slope, I also pointed this out. I believe the simulator tests were done at a significantly higher speed that Best Holding. That’s why I asked for help in characterizing how the slope changes with air speed so that I can accurately model it at the lower speed. I doubt it is linear all the way down to 370 KTAS.”

    ALSM posted a simulation that he had witnessed on the JW blog. In that simulation the airplane decelerated at FL350 from 272 kIAS to 207 kIAS in 5’26”. At ISA+10 that corresponds to 472 – 367 = 105 kt TAS, or 19.3 kt TAS per minute on average. In the last 1’36” the speed reduced by 18 kIAS, 29.6 kTAS, or 18.5 kTAS/min. Note that the end speed of 207 kIAS is close to the holding speed.

    “Based on my recollection of what others have said here, I believe that phugoids and a spiral turn are possible together. Perhaps Gysbreght will enlighten us on this point.”
    Phugoid motion is a slow pitch oscillation at constant angle of attack. The pitch oscillation can occur with or without bank, in level flight or in descent or climb. I have demonstrated that in an EXCEL model that reproduced ALSM’s simulation. The pitch oscillation was not noted by the observer. Perhaps Victor can confirm it from the GoPro file.

  418. Victor Iannello says:

    @DrB, @Andrew, et al. I think we can be fairly certain that MAS had a fairly sophisticated methodology for tracking the performance of each aircraft. The hint is that the fuel calculation used a “3% Enroute Alternate (ERA)” contingency rather than the 5% contingency that would otherwise be required, as per ICAO regulations. This is explained in the Flight Planning and Fuel Management Manual on page “5 App 3-10” (160/204 of the PDF), where it says:

    The operator should: employ a hull-specific [Fuel Consumption Monitoring (FCM)] program to monitor the actual fuel consumption rates of the specific aeroplane utilizing 3% ERA contingency fuel.

    I also noticed on the “Fuel Analysis” sheet for MH370, there is a line that says “PER FACTOR P1.5”, which likely is a performance factor. This value is identical to the “FF FACTOR:P1.5” found on the “Flight Brief” page.

    Based on this, I think it is very probable that the fuel consumption rates of the engines of 9M-MRO were monitored and analyzed, and the average PDA for the engines was +1.5%.

  419. Brock McEwen says:

    Update re: error bars on R600 incremental offset of 4600us:

    I cc’d both Peter Foley and Chris Ashton – and both responded.

    Peter’s response left the distinct impression that he did not understand the questions. (Aside: there should not be a single informed person left in the MH370 community who does not yet recognize the need for a stiff audit of search leadership.)

    Chris’ response did acknowledge that The DSTG book to which Peter referred me answered none of my questions, and at least addressed the correct subject matter. Here is the crux of Chris’ response, verbatim:

    “I cannot tell you the exact number of data points that were used to determine this value, however the number was large and based on all flights in the IOR for at least two weeks prior to the accident flight (based purely on my memory here). The analysis was performed by another member of the Flight Reconstruction Group, and was a good piece of work, based on a considerable number of flights from a variety of aircraft terminals: sufficient for us to determine this as a fixed offset value. More than this I am unable to tell you.”

    (I will assume IOR stands for “Indian Ocean Region”, but welcome expert opinion.)

    My reactions:

    1) there is no reason to withhold the exact number of data points – yet withhold it, they still do
    2) reasonable interpretations of the word “large” still end up adding concerningly large error bars to Arc7
    3) a definition of “large” that adds SMALL Arc7 error bars leaves the 2 zeros in 4600 as an unexplained coincidence
    4) “more than this I cannot say” could be construed as referring to something other than the limits of his knowledge

    Also of interest are potential schisms between

    -“fixed” nature of this differential over a 2 week period, & offset variability over time reported elsewhere (eg DSTG book)
    -“multi-terminal” sourcing he cites now & “THIS terminal” description in his Oct. 2014 JoN paper

    But since I didn’t want to swamp Chris with questions, I replied only with thanks for his time and consideration, and explicit confirmation that the two zeros were just a huge fluke.

    And to confirm that this raw data was provided to the ATSB in 2014, to ensure they had everything they needed to build a proper search box.

    I’ll keep this forum posted on his response, if any.

  420. Brock McEwen says:

    For clarity: I requested that HE explicitly confirm the double zeros were a fluke.

  421. Victor Iannello says:

    @Brock: Thank you for the follow-up with Chris Ashton.

  422. TBill says:

    @Victor
    On your BEDAX to South Pole, if you had a prior write-up I’d be interested in the link to it.

  423. DrB says:

    @Victor,
    I noticed an item in the Updated Factual Information that is related to the engine PDAs. In Section 1.6.3.7 on page 47 in Table 1.6D – Deferred Defects it says “3 07 Nov 2013 From Daily Engineering Operations Report (DEOR) – Right engine consumes average 1.5T more fuel per/hour compared to left engine”.

    This confirms the fact that in 2013 the right engine consumed considerably more fuel than the left engine – so much so it was considered a defect. Its repair was deferred, presumably for economic reasons.

    However, I can’t make any sense of the number given. At a typical Cost Index of 52, the nominal Fuel Flow at FL370 is about 3.00 metric tons per hour per engine. An additional 1.5 MT/hr would bring the right engine to 4.5 MT/hr, and implies a PDA ~50%!! This must be very wrong. Even shifting the decimal point one place is still too large an effect.

  424. DrB says:

    @Victor,

    You said:”Based on this, I think it is very probable that the fuel consumption rates of the engines of 9M-MRO were monitored and analyzed, and the average PDA for the engines was +1.5%”

    The one thing that is most certain in the analysis of prior flights is the average measured ratio of Right and Left Fuel Flows in cruise given me by ATSB. I can tell you that even if the Left engine were 0% PDA (which it cannot be), then the average is still greater than 1.5% because the right engine is so bad.

  425. DrB says:

    @Mick,

    You said: “One parameter that is not modelled and that is probably unknowable is angle of bank; I suspect that it would play a not insignificant role in determining the amplitude and period of the phugoid. As modelled, the phugoids appear to add 2 minutes to time aloft or potentially roughly up to a 10 nm shift to point of impact so at the end of the day it may not be worth investigating.”

    The AOB is a crucial parameter but at the moment we can only guess the general direction. It looks like an initial turn to the left is more likely than a turn to the right, based on speed to reach the 7th Arc.
    One purpose of this modelling is to get a better feel, if possible, for how far inside the 7th Arc the aircraft could be. 10 more miles could be important.

    The reason I included the phugoids is that I wanted to see if the change in 8 seconds from 4,400 fpm ROD to 15,000 fpm could be due to the phugoid.

  426. DrB says:

    @TBill,

    You said: “34S Arc7 was already “thin-slice” searched, so you are actually arguing that the flight ended up in a spot other than Arc7 at 34S.”

    Well certainly I am not arguing it is within the thin strip already searched. I suspect it is between 3-30 NM inside the arc near 34S, or possibly just outside the area previously searched. My predicted BFO at Arc 6 at 33.1 S is 254 Hz (with no ongoing descent). Measured is 252 Hz.

  427. DrB says:

    @Victor,

    You said: “I am curious why you chose to use the minimum manoeuver speed as the speed at which descent begins with an engine out and under autopilot. I believe the speed should be the stick shaker speed, which is closer to around 163 KIAS. I believe @Andrew confirmed this based on what he has observed in Level D simulators.”

    That was a mistake on my part, I think. I do recall the discussion of the correct IAS when descent begins. I think Gysbreght first suggested Minimum Maneuver speed. This is an important parameter because it determines how long after INOP the BFO should show the descent, and whether or not this happens before or after 00:11. If there is no disagreement, I will change it to 163 KIAS for the next run. That’s the value of “group-think” in studying problems like this. The whole is smarter than any part.

  428. Brian Anderson says:

    @Gysbreght,

    “That possibility would seem to mitigate against phugoids, …”.

    Well, phugoids in a straight line . . .

    SImply because if the track is held constant through 00:11 until the second engine fails, as it should [and not manipulated as DrB has done] and the deceleration is about 19knots/min [TAS], then it is unlikely that the aircraft can reach the 7th arc.

    It is very possible that an initial phugoid rolled off into a left turn. The turn may have been shallow to begin with [a la ATSB, but note that the ATSB dd not say it stayed in a shallow banked turn]. A left turn makes it possible to intercept the 7th arc and a continuing spiral with increasing turn rate produces the right conditions for the calculated descent rates of around 14,000 ft/min.

    If the intercept with the 7th arc results from a left turn in this fashion, then joining this point with the intercept at 00:11 and calling it a change in the track at 00:11 [and a shorter distance] is not, in my view, the correct answer.

  429. Richard says:

    @DrB

    You are arrogant and please do not tell me that I am either “being silly” or “missing the point”.

    If you had bothered to read any of the 25 papers, that I have published on MH370, you would see that I am genuinely interested in helping find MH370. Whether I got it right or wrong, time will tell.

    You equally deride others whom I respect, such as Brian Anderson, who also pick up on your changing views, designed to suit your bias, prejudice and preconceived ideas.

    Unfortunately for Victor, this is no longer a scientific forum, but a typical American style slanging match.

    “All truth passes through three stages. First, it is ridiculed. Second, it is violently opposed. Third, it is accepted as being self-evident.” – Arthur Schopenhauer

    I have had enough of your ridicule and opposition. I am bowing out.

  430. Gysbreght says:

    @DrB: “I think Gysbreght first suggested Minimum Maneuver speed. (…) If there is no disagreement, I will change it to 163 KIAS for the next run. “

    Not so quick. In his first post here on March 2, 2017 at 12:23 am Andrew says:

    “Regarding the autopilot behaviour, we demonstrate the aircraft’s envelope protection in the full flight simulator by reducing the thrust at high level and waiting to see what happens. The autopilot maintains altitude as the speed decreases and continues to do so until the speed is well below the minimum manoeuvring speed, but just above stick shaker activation. At that point the autopilot gives up maintaining altitude, as indicated by an AUTOPILOT EICAS caution and an amber line through the pitch mode annunciation. The aircraft then descends and accelerates back to the minimum manoeuvring speed and continues descending at that speed.”

    That makes more sense to me than whatever Microsoft Flight Simulator does. So let’s wait until Andrew has determined what a Level-D simulator does.

  431. Gysbreght says:

    @Brian Anderson:

    “Well, phugoids in a straight line . . . “
    Not sure what you’re trying to say.

    “SImply because if the track is held constant through 00:11 until the second engine fails, as it should [and not manipulated as DrB has done] and the deceleration is about 19knots/min [TAS], then it is unlikely that the aircraft can reach the 7th arc.”
    How do you explain that all the Boeing simulations shown in Figure 6 of the Nov.2, 2016 report of the ATSB reached the 7th arc?

    “The turn may have been shallow to begin with [a la ATSB, but note that the ATSB dd not say it stayed in a shallow banked turn]. A left turn makes it possible to intercept the 7th arc and a continuing spiral with increasing turn rate produces the right conditions for the calculated descent rates of around 14,000 ft/min.”
    Except simulations 7 through 10 which represent an entirely different scenario, none of the recent Boeing simulations ended in “increasing” turn rates. Only simulation #3 exceeded the rate of descent derived from the final SATCOM transmission. Wy it did remains to be explained.

  432. DrB says:

    @Gysbreght,
    @Victor,

    Thank you for evaluating my Straw-Man #1 End-of-Flight Model. Your comments and suggestions are helpful.

    Here are my specific responses and follow-up questions:

    1. I’ll put in a constant 19 KTAS/min speed slope at INOP until descent begins, and we’ll see what happens.

    2. I’ll also put in 480 ppm instead of 600 for the ROD during OEI and 2,000 fpm after 00:17:29 (but see my question below).

    3. I am still wondering if the descent will begin at the highest speed in the INOP Holding table, which is 209 KIAS, or at the stick shaker speed of ~163 KIAS that Victor observed in his test. I can’t recall if his test case was using INOP Holding or INOP LRC? Victor and Gysbreght, please opine on the CAS at which the descent should begin. I assume then that the CAS will be held constant during the descent until L engine flame-out. I also assume (perhaps wrongly) that that speed during descent might depend on whether the VNAV is trying to maintain INOP LRC or INOP Holding.

    4. Victor, with the autopilot still operating during INOP, I don’t understand why it would let the speed drop below Minimum Maneuver (+1 kt) in order to maintain altitude. Somebody please explain the logic behind this choice (if there is any).

    5. Gysbreght, you said something I didn’t understand (that’s not the first time, but the problem is usually on my end). You said the ROD corresponding to the -19 KTAS/min deceleration was 480 fpm. But I thought the whole point of the descent was to maintain a constant CAS. Why should it depend on the deceleration rate that only occurs before the descent begins?

    6. In ALSM’s test, what happened when the IAS reached 207? Did descent begin then, or did the deceleration continue to a lower air speed?

    7. You said: “. Simulation #3 is the only one in that group that “recorded descent rates that equalled or exceeded values derived from the final SATCOM transmission. . . . Simulation #3 is also the only one where the track exhibits evidence of phugoid motion.” One can infer from these observations that it appears likely that a phugoid was underway in MH370 at 00:19. Otherwise none of the other non-phugoid simulations were capable of achieving the high ROD observed at 00:19:37. That is why I included the phugoid in this model – to possibly match the 00:19:37 BFO while still at a fairly high altitude. So far it appears that the altitude at 00:19 was still quite high, maybe FL300 or thereabouts. And yet you see large ROD values and large changes in 8 seconds. I was thinking maybe these two RODs could be fit by a phugoid on the steep downslope. Otherwise how could the ROD be 15,000 fpm at FL300?? Maybe I am missing something here.
    8. If the phugoid is modelled as a mean ROD of 2000 fpm with a sinusoidal variation with amplitude 1520 fpm, you can never get to 4400 fpm, much less to 15,000 fpm. Either the phugoid ROD fluctuation is very much larger than +/- 1520 fpm, or some other mechanism is required to explain the 00:19 BFOs. Simulation #3 indicates that phugoids are capable of much larger RODs, since it got to 15,000 fpm. I don’t know if this was due to a very high mean rate or a very high sinusoidal term, or both. With a longish 82 second period, the sine term must be huge in amplitude to change the ROD by 11,000 fpm in only 8 seconds. In fact, the sine term amplitude has to be at least (15,000 – 4400)/2 = 5300 fpm to generate two RODs 10,600 fpm apart.

    8. Does anyone have a guess as to what happens to the air speed after 00:17:29?

  433. Victor Iannello says:

    @DrB: I think we have agreement that with an engine out, the plane will decelerate until the stick shaker speed is reached before altitude is no longer maintained, and not the minimum manoeuver speed. After the descent begins, whether the plane descends at the stick shaker speed or accelerates to the minimum manoeuver speed is in question. I’m fine with going with whatever @Andrew believes is correct. I trust the Level D simulator more than the PMDG 777 model. In any event, it may be a moot point because it may be that the second engine flames out before reaching the stick shaker speed.

  434. Victor Iannello says:

    @DrB: I believe there was a typographical error in the FI. I think the 1.5 MT/hr was meant to be 1.5%. If we believe the 1.5% for the average PDA for the engines is correct, that would put one engine at 0.75% and the other at 2.25%. When you consider that the performance of an engine is subject to statistical variation, that seems reasonable. I bet there are some engines that roll out of the factory with a PDA less than zero.

  435. DrB says:

    @Victor,

    That is quite a stretch to go from 1.5T/hour to 1.5%, and it looks to be more than a typo. In addition, 0.75% for an engine built in 2004, rebuilt in 2010, and with 40,000 hours on it is a similar stretch. Based on the numbers from ATSB, the L engine is in the middle of some quoted ranges of PDAs. It is nowhere near 0.75%. The right engine was declared a defect in 2013 but was not sent to the shop. It’s still puzzling to me how the take-off and climb ratios aren’t too far apart, but the cruise numbers clearly are. Maybe somebody who knows gas turbines can explain why.

  436. DrB says:

    @Gysbreght,

    You said: ” “The aircraft then descends and accelerates back to the minimum manoeuvring speed and continues descending at that speed.” That makes more sense to me than whatever Microsoft Flight Simulator does. So let’s wait until Andrew has determined what a Level-D simulator does.”

    I understand now, thanks to the explanations from Victor and you/Andrew. Do you have a guess as to the acceleration rate from stick-shaker to Min Maneuver? Also, what will the ROD be during this acceleration, and what will it be once the speed is steady at Min Maneuver? Presumably the latter is smaller than the former, since for the latter there is no ongoing air speed acceleration, just air speed maintenance.

  437. TBill says:

    @Victor @DrB
    For what its worth, the fuel worksheet in the Folder 5 says:
    PER FACTOR P1.5

    On another subject (my pre-occupation with jet fuel quality analysis for which there is no disclosed data – to my knowledge) I see that the Boeing jet engine fuel flow equations have a fuel flow correction factor for LHV (Heating Value per pound). So I am back to thinking there is a need to know analysis of the actual jet fuel, which my perception was standard step#1 to analyze fuel after an accident. Some of the fuel came from day before 9M-MRO flight to Peking.

  438. Andrew says:

    @Victor: Are you able to share the MH370 Flight Briefing/Fuel Analysis?

    @All: I have had a session in a B777-300 full-flight simulator (RR Trent 892 engines) and was able to test the scenarios suggested by DrB. I also had a look at some of the other aircraft behaviours that have been discussed. I’ll write up the results over the next week and post a link, but in the meantime, I’m sure you’ll be interested to see the findings, some of which might force a change of thinking:

    DrB’s Questions:

    1. How does LNAV mode fly a turn with a track change of 116° where a 15nm offset has been inserted?

    The FMC flew the turn with a constant bank angle of 22° to intercept the 15nm offset track to ANOKO. The aircraft did not cut the corner and remained at least 15nm to the right of IGOGU at all times. The following sequence of photos shows the path the FMC calculated. The dashed circles are range rings centred on IGOGU at distances of 5, 10 and 15nm.

    Approaching IGOGU offset:
    https://www.dropbox.com/s/r38bkvoeelxij4j/Approaching%20IGOGU%20R15nm%20.jpg?dl=0

    FMT abeam IGOGU offset:
    https://www.dropbox.com/s/twfrlyqhi7kokqf/FMT%20at%20IGOGU%20R15nm%20.jpg?dl=0

    Approaching ANOKO offset:
    https://www.dropbox.com/s/r5159vv69zsjitj/Approaching%20ANOKO%20R15nm%20offset.jpg?dl=0

    2. Do the FMC END OF OFFSET and END OF ROUTE error messages occur simultaneously when the offset holding fix is sequenced?

    No. As stated previously, the END OF OFFSET message occurs two minutes before reaching the end of the offset track. The END OF ROUTE message occurs when the last offset waypoint is sequenced, ie abeam ANOKO. This behaviour was confirmed in the simulator.

    3. Does the FMC maintain a constant true heading or a constant magnetic heading after passing the last waypoint of the route while LNAV is engaged?

    The FMC maintained a constant magnetic heading after passing the last waypoint with LNAV engaged. This behaviour was tested by repositioning the simulator to overhead the Johannesburg VOR (JSV), an area where the isogonals are closely spaced. A waypoint was created 20 nm due south (180°T/199°M) of JSV, with a hold inserted at that point, together with a 15nm offset. The simulator was then positioned on the offset and allowed to proceed with LNAV engaged. On reaching the offset, the aircraft maintained a constant heading of 199°M for 20 minutes with the ground speed multiplier set at 5X. During that time, the variation increased from 19°W to 33°W and the aircraft’s true heading changed by the same amount. The wind was increased from a 5-knot crosswind to a 100-knot crosswind to determine if the aircraft was maintaining a constant heading or track. The increased wind strength caused the track to change, but the heading remained constant at 199°M throughout.

    4. When does the aircraft decelerate to the BEST SPEED prior to crossing the offset holding fix?

    The aircraft did not decelerate to the best holding speed; it maintained the LRC Mach no. after the aircraft passed the offset holding fix. The FMC does not seem to recognise the hold as part of the offset path, which probably makes sense, given that the path terminates at the offset holding fix.

    Other Questions:

    1. Deceleration rate after first engine failure.

    The right engine was failed while cruising at LRC/FL360 with a gross weight of 176,000 kg. The aircraft took 11 minutes 52 seconds to decelerate from 264 KIAS to the minimum manoeuvring speed of 208 knots. Note: The cruise reference thrust limit during the exercise was CLB not CRZ. The default limit (ie CLB or CRZ) is selectable according to airline policy. I do not know which thrust limit was used as the default by Malaysian Airlines. If Malaysian used CRZ as the thrust limit in their B777s, then MH370 would have decelerated at a faster rate.

    2. Envelope protection descent speed.

    The aircraft maintained level flight in VNAV PTH until the stick shaker speed (168 KIAS). The aircraft then pitched down and initially accelerated to about 180 KIAS before stabilising at 172 KIAS with a descent rate of 850 ft/min, as shown in the following sequence of photos:

    Stick shaker:
    https://www.dropbox.com/s/4gw2tvbaj0mqro5/Envelope%20protection%20-%20stickshaker%20speed.jpg?dl=0

    Initial pitch down:
    https://www.dropbox.com/s/eebssnbsluqi8ku/Envelope%20protection%20-%20initial%20pitchdown%20and%20acceleration.jpg?dl=0

    Stable descent:
    https://www.dropbox.com/s/jd1ews1ywpeotuh/Envelope%20protection%20-%20stable%20descent.jpg?dl=0

    3. Descent behaviour after second engine failure.

    The aircraft initially remained wings level after the second engine failure and began descending at a rate of about 2,500 ft/min. The aircraft gradually began rolling to the left and initially established a bank angle of 5°. A gentle phugoid motion then developed, with the speed varying from 212 to 235 KIAS, and the rate of descent varying from 500 to 3,500 ft/min. The period of the phugoid motion was about 45 seconds. During this motion, the angle of bank increased to 7° and then stabilised. Unfortunately, I ran out of time and had to stop the simulation as the aircraft descended through 15,000 ft. At that point, the phugoid motion seemed fairly stable, as was the angle of bank. The aircraft did not appear to be entering a screaming dive!

  439. Victor Iannello says:

    @TBill: Your preoccupation with jet fuel quality is wise. I just don’t know how to get the answer you seek.

    @DrB: The wording in the FI is “Right engine consumes average 1.5T more fuel per/hour [sic] compared to left engine”. On my keyboard, the “T” key is just below the “%” key. That is an easy typo to make, and is not a “stretch”, as you claim.

    As for your doubting the 1.5% average PDA, that is what MAS’ Fuel Consumption Monitoring Program says should be used for 9M-MRO. We know there was a FCM program because the fuel contingency was based on a 3% ERA methodology. You may prefer to believe your numbers, which nobody can verify because you haven’t shared them, but fuel flow measurements collected at short periods during a cruise will not be as accurate as measurements of total fuel consumed for a trip.

    I don’t know what the PDA of an engine is after an overhaul, but if worn seals are replaced, tip clearances are back to nominal values, and flow passages are cleaned, the performance should be almost good as new.

  440. Andrew says:

    @DrB
    @Victor

    It’s worth remembering that aircraft tech log entries are all handwritten in the first instance and subsequently entered into the airline’s maintenance record system. It is not at all unusual to see typos, especially when the data entry is done by clerical staff who have little understanding of the terminology and/or can’t read the engineer’s writing. I would be absolutely amazed if a fuel flow discrepancy of 1.5T/hour was signed off as a deferred defect and allowed to go unrectified for four months!

  441. DrB says:

    @Andrew,

    Thank you very much for performing the simulator tests and for obtaining definitive results on the first try. I think there was at least one surprising result for everyone, and several for me. This will take a while to digest. I can’t help wondering why the high-end PC-simulator results are so different. I’ll have more comments later, after my head stops spinning.

  442. DrB says:

    @Andrew,

    Two quick questions. Do you recall what the altitude was when the second engine flamed out? Also how long after the second engine stopped did it take to get to 15,000 feet?

  443. DrB says:

    @Victor,

    Now I see what you mean about the typo. Maybe it was just a typo. I’ve sent some questions on this to ATSB to see if they can shed some light on it.

  444. Andrew says:

    @DrB,

    Sorry, I didn’t take note, but the second engine failed not long after the aircraft was stabilised in the ‘envelope protected’ descent. If I recall correctly, it failed at about FL340. It then took almost 10 minutes to descend to FL150.

  445. David says:

    @Gysbreght. With an engine out and under autopilot, you quoted @Andrew. “The autopilot maintains altitude as the speed decreases and continues to do so until the speed is well below the minimum manoeuvring speed, but just above stick shaker activation. At that point the autopilot gives up maintaining altitude, as indicated by an AUTOPILOT EICAS caution and an amber line through the pitch mode annunciation. The aircraft then descends and accelerates back to the minimum manoeuvring speed and continues descending at that speed.”

    During the slowing and consequent aircraft pitch up, elevator offload will result in the stabiliser assuming the pitch up trimming load. After the elevator nosed the aircraft down, I assume the stabiliser nose trim up would reduce, the elevator control law having changed. Second engine fuel exhaustion timing and thence AC failure could be either side of the nose down (I note Victor’s comment to that effect above. Andrew’s above are well separated). At this point the autopilot would disengage, the stabiliser remaining trimmed nose up, the amount depending on timing. The elevator will neutralise (Andrew advised that prevailing autopilot commands would be disregarded) when quite possibly it had been nose down.

    Certainly if the aircraft elevator and stabiliser are both in neutral by the time the AC fails, having adjusted during run down for the pitch change on loss of the second engine thrust, the aircraft subsequently will not have an inbuilt pitch in the offing, but otherwise most likely it will pitch up as residual stabiliser nose up is applied.

    In the unlikely extreme of the second engine failing at or close before the point of nose down, there could be an AF447 parallel. However since this has not been raised hitherto and evidently is not encountered in simulations I presume that extreme has been ruled out.

    Even so we may need to take a look at not just left roll initiators, the left flaperon and RAT. There could be a TAC of sorts in pitch, though not removed at AC loss. The effect on ensuing descents will depend on how close together engine failures are assumed to be.

  446. Ge Rijn says:

    @David @Andrew

    Maybe something to consider in regard of elevator behavior after second engine flame out.
    As mentioned in an earlier linked paper only the left elevator will be actuated under RAT and the right elevator will be in damped-mode.
    The damped-mode will prevent this right elevator from fluttering but will allow it to move under the various aerodynamic loads.
    Could this have any stabilysing or de-stabilysing effect (i.e. in (de)stabilysing bank angles/roll or pitch)?

  447. Gysbreght says:

    @DrB: “But I thought the whole point of the descent was to maintain a constant CAS. Why should it depend on the deceleration rate that only occurs before the descent begins?”

    Newton’s 2nd law: F = m*dv/dt = m*a

    Assuming still air and ignoring an inclination of the thrust vector to the flight path, the equation of motion along the flight path can be written as:

    (T)hrust – (D)rag – (W)eight*sin(FPA) = (W)eight*a/g

    or: (T-D)/W = sin(FPA) + a/g

    In level flight: FPA=0 and a/g = (T-D)/W
    At constant speed: a/g = 0 and sin(FPA) = (T-D)/W

    An airplane descending at constant IAS is decelerating in terms of TAS. The FPA is therefore less than (T-D)/W. That difference can be expressed in the acceleration factor at constant CAS, which is about 1.2 at the CAS and altitude considered.

    “In ALSM’s test, what happened when the IAS reached 207? Did descent begin then, or did the deceleration continue to a lower air speed?”
    In ALSM’s test the second engine flamed out when the IAS reached 207 kt, and the descent began.

    “Otherwise how could the ROD be 15,000 fpm at FL300??”Without active controls the ROD cannot attain 15,000 fpm within 8 seconds after the time of the 7th arc log-on request message. The final BFO is bogus. No one (not even ALSM) has been able to demonstrate or even suggest a scenario that matches the timing of the last two BFO’s.

    “One can infer from these observations that it appears likely that a phugoid was underway in MH370 at 00:19. Otherwise none of the other non-phugoid simulations were capable of achieving the high ROD observed at 00:19:37.”There are no non-phugoid simulations. Without active controls the airplane maintains a certain AoA. Therefore all uncontrolled descents are phugoids. The amplitude depends on the scenario. The parameters I gave apply to your particular scenario. I have asked the ATSB for information that could help me understand why the phugoid amplitude in simulation #3 was so large that 15,000 fpm was exceeded. I’m still waiting for their reply.

  448. Gysbreght says:

    @Andrew: Thank you for doing the simulator tests and for excellent debrief. I’m looking forward to your detailed write-up and trust that it will indicate the deviation from standard temperature, if any.

    @David: I hope you’ll agree that we have to digest Andrew’s observations.

  449. Joseph Coleman says:

    @Andrew

    Thanks for your interesting observations. I particularly find the 180/199. Did the approaching ANOKO Stay at 181 as seen in the illustration. Plus can true be selected at any point before after or durning FMC’s manoeuvres.

  450. Ge Rijn says:

    @Andrew

    You stated: “The FMC maintained a constant magnetic heading after passing the last waypoint with LNAV engaged”.

    Would it make any difference if the TRUE/NORM switch was set to TRUE before the EOR/Route-discontinuity? Would it still be a constant magnetic heading or would it be a constant true heading?

    Thanks also for this simulation. Very interesting and a possible game-changer as far as I can understand it.
    A pity you had no time to go all the way down under 15.000ft.
    Hope you’ll have the opportunity to finish this simulation till the end soon.

  451. Gysbreght says:

    @Andrew: “The period of the phugoid motion was about 45 seconds.”

    That is much shorter than I would expect. Please excuse a silly question, I’m really interested. Did you perhaps measure the time between one extreme to the opposite extreme, which would be a half-period?

  452. Victor Iannello says:

    @Andrew: You crushed it! Thank you for using the Level D simulator to answer so many questions we’ve had.

    Here are the Flight Plan documents for MH370.

    In the simulation, how did you fail the first engine? Fuel exhaustion? Fuel cut-off? Did the autothrottle (A/T) remain engaged after shutdown? I don’t see indication of the A/T in the PFD. How did you fail the second engine?

  453. Victor Iannello says:

    @DrB asked, “I can’t help wondering why the high-end PC-simulator results are so different.”

    PMDG uses the FCOM to model the behavior of the aircraft. In addition, it gets feedback from 777 pilots, which allows PMDG to incrementally improve the model. However, the types of questions we have often require knowledge beyond the FCOM and beyond a typical pilot’s experience. It may be that without access to the actual flight management and controls software, a Level D simulation is the only way for us to get answers. That’s why @Andrew’s work is so valuable.

  454. Andrew says:

    @Ge Rijn:

    The damped-mode will prevent this right elevator from fluttering but will allow it to move under the various aerodynamic loads.

    I haven’t read the earlier paper you mentioned, but the elevator PCUs only have three modes; normal, bypass and blocking. There is no damping function. My understanding is that when two adjacent PCUs fail (as they would have done in this case), then both PCUs will switch to blocking mode as soon as the air loads move the elevator close to the faired position (within 2° of neutral with flaps retracted). From that point on, the elevator will be kept in the faired position by a hydraulic lock within the PCUs.

  455. Andrew says:

    @Gysbreght:

    The temperature deviation was ISA+10 throughout.

    The period of the phugoid was measured as the time taken from one extreme back to the same extreme.

  456. Andrew says:

    @Joseph Coleman

    Did the approaching ANOKO Stay at 181 as seen in the illustration. Plus can true be selected at any point before after or durning FMC’s manoeuvres.

    The heading remained at 181°M approaching ANOKO. True can be selected at any time via the HDG REF switch. My understanding is that the FMC will then use true as the heading reference, but I did not have time to test that scenario.

  457. sk999 says:

    Andrew,

    Many thanks for your report on the simulator results. It has squashed many “urban legends” about how the FMC/AFDS systems work. As an aside, the fact that CRZ vs CLB thrust after one engine-out is airline-selectable was mentioned previously by someone on another board, but also it was also mentioned that on other 777 models only CLB was available (not that it matters here.)

    Regarding your EOR test around Johannesburg, presumably you had the the HDG REF switch in the MAG/NORM position. I agree with Ge Rijn that there is stil this dangling question of what happens when the switch is in the TRUE position. (I have an opinion, not that it is worth anything).

  458. Andrew says:

    @Victor:

    In the simulation, how did you fail the first engine? Fuel exhaustion? Fuel cut-off? Did the autothrottle (A/T) remain engaged after shutdown? I don’t see indication of the A/T in the PFD. How did you fail the second engine?

    I artificially lowered to the fuel remaining to 2,500 kg (L = 2,000 kg, R = 500 kg) and let nature take its course, ie fuel exhaustion for both engines.

    For some reason the autothrottle disengaged, which was unexpected. I didn’t have time to try it again.

  459. DrB says:

    @Andrew,

    You said: “The heading remained at 181°M approaching ANOKO. True can be selected at any time via the HDG REF switch. My understanding is that the FMC will then use true as the heading reference, but I did not have time to test that scenario.”

    When approaching “abeam ANOKO” in LNAV, wouldn’t the track always be a great circle (geodesic)? In that case, the NORM/TRUE switch should only affect the display, not the aircraft path.

    However, AFTER the EOR error occurs, the path leaving “abeam ANOKO” depends on the NORM/TRUE switch according to Honeywell. So if you had the opportunity to do the same experiment again, it would be interesting to set the switch beforehand to TRUE and verify the lateral navigation after EOR is constant true heading.

  460. DrB says:

    @Andrew,

    Do you have any numerical data from your run with time information?

    If not, you can possibly retrieve the time each photograph was taken by viewing the file properties of the raw images. It might be as simple as viewing the properties and looking for the “Date Created”. Or you could email me the unedited pictures and I can extract the embedded time information. This would be especially useful for the descent phase.

  461. TBill says:

    @Victor
    As far as jet fuel, we could ask ATSB if there is any quality info, or at least what standard heating value would be typical of MAS airport, that they would assume for MH370.

  462. Ge Rijn says:

    @Andrew

    Here’s the paper on the elevator actuator modes. It tells there is a damped-mode. It’s a bit confusing. On the one hand it tells both elevator actuators will go in the damped-mode if both actuators on one elevator surface fail. On the other hand it tells the actuators will go in blocked-mode in this case like you say.
    See chapter 11.5 especially the elevator example mentioned:

    http://www.davi.ws/avionics/TheAvionicsHandbook_Cap_11.pdf

  463. Victor Iannello says:

    @Andrew: Do you recall when the A/T disengaged? It must have occurred after the remaining engine thrust advanced to CLB.

  464. Ge Rijn says:

    To add; what possibly is interesting, the paper tells there’s no blocked-mode available on the rudder.
    Only active-mode, damped-mode and bypass-mode.
    Are the rudder actuators also pressurised under RAT?

  465. ALSM says:

    Victor & Bobby:

    Re: “…Right engine consumes average 1.5T more fuel per/hour compared to left engine”.”

    I spotted this obvious error shortly after the document was published. ATSB confirmed in an email it was in error, but I can’t located the email now. I don’t recall if they said it was meant to be 1.5% or something different. But 1.5T/hr is definately wrong.

  466. DrB says:

    @Victor,
    @Andrew,

    Please consider the following scenario:

    1. The Airway N571 is put into the Active Route.
    2. The (15 NM R) lateral offset is selected.
    3. The waypoints after IGOGU are deleted and replaced with ANOKO.

    Will ANOKO have the offset?

  467. Ge Rijn says:

    @Andrew

    It seems in your simulation-test the bank-angle did not go beyond 7 degrees and stabilysed or was stable at 15.000ft when you had to end the simulation. Also the amplitude of the phugoids where not extreme it seems to me. No extreme descents and climbs and no extreme banks that forced a spiral dive. And a fairly stable AoA througout the descent (if I understand it well..).
    You seem to suggest this attitude and behaviour probably continued after descending under 15.000ft till the end.
    Am I correct in interpretating your debrief this way?

  468. TBill says:

    @Andrew @Victor
    Nice simulator work!
    The photographic quality of the screen shots is excellent.
    Just out of curiosity, what city did you have to go to do this work?

    Re: 119 deg (>90 deg) Turns- It looks to me (but correct me if wrong) that the simulator had no problem taking that turn exactly without over-shoot like Victor and I are seeing in FS9/FSX.

    Re: waypoint discontinuity behavior is what FS9 PSS-777 model also does magnetic heading I think. PSS777 seems to do magnetic heading even if TRUE is switched on (but of course no guarantee that PSS777 is correct).

    End of flight, that is a little “controversial” I think maybe (lots of hedge words). That could say if the dive was steep, it was piloted.

  469. TBill says:

    PS- In the MH370 case, if the crash happened around 30-34S there was one helluva stiff wind from the west, possibly 75 knots but I am not sure widn speed vs. altitude

  470. Brock McEwen says:

    @Richard:

    1) 19:41 BFO error minimization:

    When I try to replicate your new method using your 16.0 Excel file (the one whose true track path has heading=184° at 19:41, and BFO error of an impressively low 1.0 Hz), I am able to zero out the 19:41 BFO error at 193°.

    Yet you explained to me the new path of yours I sampled (SANOB) requires a 216° heading at 19:41 to zero out its BFO. This 216° bearing is in fact the result on which your paths literally hinge.

    Why the difference?

    2) Ensuring I have your latest 16.0:

    The 16.0 sub-versions to which you linked me last week (thanks again) didn’t have the “Fixed Bias” column in the Key Points tab. All previous versions had this column. Any chance I could see the version without that column crunched out? Huge thanks yet again.

  471. DennisW says:

    @all

    I have a serious question for the group. Consider the Wiki cut-paste below:

    begin Wiki cut-paste//

    In May, the search strategy working group was established by the ATSB to determine the most likely position on the aircraft at the 00:19 UTC (08:19 MYT) satellite transmission. The group included aircraft and satellite experts from: Air Accidents Investigation Branch (UK), Boeing (US), Defence Science and Technology Group[j] (Australia), Department of Civil Aviation (Malaysia), Inmarsat (UK), National Transportation Safety Board (US), and Thales (UK).

    end Wiki cut-paste//

    Does anyone really believe that the collective on this forum is better qualified to determine how the aircraft came down than the search strategy working group defined above?

  472. Joseph Coleman says:

    @Dennis W

    Not even a collective understanding of Gödel’s Incompleteness Theorem will help this Mystery.

    “Anything you can draw a circle around cannot explain itself without referring to something outside the circle – something you have to assume but cannot prove.”

  473. sk999 says:

    The operational flight plan (OFP, which Victor recently reposted; source is RMP, folder 5) contains the detailed routing with coordinates, ground track and speed, weather data, and fuel profile. I entered the data and ran them through my fuel model (not guaranteed to be correct) to see how well it would match the fuel profile in the OFP. The one thing missing from the OFP is temperature (or if it’s there, I am missing it), so I used the GDAS prediction. I ran from IGARI to the Top of Descent (TOD). I have not tried to model is fuel usage during climbs and descents, of which there are several, and since the flight plan called for an altitude differential of 4000 feet from IGARI to TOD, I am certain to under-predict the fuel usage. To compensate, I determined what “PDA” I needed to match the OFP.

    Result? I need a “PDA” of 3.5%. The total fuel burn is 28.7 tons. Remember, the PDA of 3.5% is accounting for all the extra fuel needed to gain and lose altitude, so the true PDA is less.

    The above is all for my fuel model. YMMV. For convenience, I have uploaded the data from the OFP, should anyone else wish to have a go. As usual, here is a link to an index of all my files, with the OFP being at the top:

    https://docs.google.com/document/d/14hleZyx1pUPL44yaeHKt6jnSQ3DbgRq2zibbKkFLq2c/edit?pref=2&pli=1

  474. Brock McEwen says:

    @Dennis: I will give you a serious answer:

    Technically, none of us knows enough even to answer your question, because we don’t know who the actual representative were, and they chose to disclose neither key data nor key models.

    If we go by the REPUTATIONS of that group – and assume they did their best to find wreckage – then of course, the answer to your question is, “no; none of us can improve upon their experience and expertise”.

    However, if we go by the RESULTS of that group:

    1 month needlessly flitting among 5 disparate search zones, followed by
    2 months needlessly scanning at s21, followed by
    4 months needlessly surveying s26-s33, followed by
    8 months finally falsifying 1 theory from s33-s40 only, followed by
    20 months needlessly pushing the search zone SW, and wider

    – then the answer to your question is, “yes; ALL of us can improve upon their strategy and success. A collection of moderately enlightened pocket lint could improve upon them”.

    My call for an audit – for which your support has been noted, and appreciated – has arisen precisely because of the mind-boggling vastness of the gap between how that group SHOULD have performed, and how that group DID perform. Not only did they not find the wreckage, they failed to falsify a single one of the official hypotheses. Not one.

    The SSWG must now identify themselves by name, and show us their work.

  475. Mick Gilbert says:

    @Brock McEwen

    To be fair to the SSWG, the decisions relating to “1 month needlessly flitting among 5 disparate search zones, followed by 2 months needlessly scanning at s21” weren’t theirs. The surface search and the underwater towed pinger sweeps and sonar scans around 21°S 104°E were “managed” jointly by the Australian Maritime Safety Authority and the ATSB.

  476. gecina says:

    @Dennis
    “Does anyone really believe that the collective on this forum is better qualified to determine how the aircraft came down than the search strategy working group defined above?”

    Perhaps not better qualified but more motivated, tenacious, curious, driven? There’s a good argument for “yes”. It’s a shame though that the bottom line is still that there’s not enough information.

  477. AlexSiew says:

    [VI note: Alex is correct that the alternate airports in the DETRESFA message are different than those in the final Flight Plan that was used to calculate the fuel load. I think this was the result of a change in the flight plan, perhaps due to changing weather, and the flight plan change was not carried forward throughout the ATC system. I do think this discrepancy needs to be understood because it points to a flaw somewhere between MAS and ATC, although I don’t think there were doctored documents, as Alex claims.]

    If I may point out that the purported MH370 flight plan documents in Folder 5 of the ‘RMP Report’ are not the authentic flight plan documents relating to MH370 but are doctored documents or bogus altogether.

    The real Filed Flight Plan is at page 11 of the Factual Information Report (Figure 1.1D). See also the DETRESFA message issued on the morning in question at page 100 of the Factual Information Report. The fuel endurance in terms of time as stated in the DETRESFA message was 7 hours and 10 minutes. Both the Flied Flight Plan and the DETRESFA message had the alternate airports as ZBTJ (Tianjin Airport) and ZBSJ (Shijiazhuang Airport) which are airports not far from Beijing Airport. That the 2 alternate airports were ZBTJ and ZBSJ and that the fuel endurance was 7 hours and 10 minutes were confirmed in a paper presented by Malaysia at an ICAO meeting on MH370 in January 2015 (less than 2 months before the FI was published).
    http://www.icao.int/APAC/Meetings/2015%20APSARTF3/WP06%20MH370%20SAR%20Operations%20and%20Lessons%20Learnt%20(Malaysia).pdf

    In contrast to the real documents, the bogus documents in Folder 5 of the purported RMP Report had different alternate airports (ZSJN [Jinan], ZSHC [Hangzhou])and different fuel endurance. However, much like for Folder 4, the people doing the fabricating/doctoring did not do a very good job. Page 1/9 had the fuel endurance at 7 hours and 31 minutes while Page 3/9 had it at 7 hours and 30 minutes. Obviously these bogus fuel endurance figures are there to support the official narrative of a plane still airborne at 8.11am , the time of the original final (bogus) ping which was 7 hours and 30 minutes after takeoff at 12.41am. As for the bogus alternate airports, ICAO rules require that the fuel required to fly to the alternate which is farther away to be calculated when there are more than one alternate (as was the case for MH370) but in these bogus documents, the fuel calculations had the figure for ZSJN (Jinan – purportedly 46 minutes flight), the closer alternate, instead of ZSHC (Hangzhou- purportedly 1 hour 45 minutes flight). So the main text of the FI, which was obviously prepared by the same people who came up with Folder 5 of the RMP Report, reads as follows:

    ” The first alternate airport, Jinan Yaoqiang International Airport (China), was estimated to be 46 minutes from the diversion point with 4,800 kg fuel required and the second alternate airport, Hangzhou Xiaoshan International Airport (China) was estimated to be 1 hour 45 minutes with 10,700 kg fuel required…”.

  478. Paul Smithson says:

    @sk999. Thanks for doing that – I was intending to do the same thing today and will review your figures with interest. I thought that we believed that PDA was already included in the fuel forecast? If so, then the only thing not included would be the additional fuel required for step climbs in the plan.

  479. Paul Smithson says:

    @sk999. I suspect that the flight plan/fuel forecast [pp9-14] contains more info than you have mentioned and we can see exactly what the weather assumptions are. My guess for the cell labels below
    WPT – waypoint
    AWY – airway
    LAT/LONG – self-explanatory
    TTRK – true track
    MTRK – mag track
    SAT – static air temperature (celsius) with M prefix being “minus”
    TDV – ISA delta / temperature deviation; P = positive
    WS, TP = ??
    WC = wind correction, presumably: tailwind (P), headwind (M)
    WIND -direction/speed (kts)
    GS – groundspeed
    FL – flight level
    ZEET – time of segment (rounded minutes)
    TOTM – total/cumulative time (measured from take-off?)
    FUELRM – fuel remaining (metric tonnes) * the RQFUEL and FUELRM labels seem to have been reversed, assuming RQFUEL means required fuel for each segment

    Note also that in climb phases the FL is substituted by CLB and duration/fuel required is indicated.
    Wind and temperature at various flight levels as ENRT WIND FORECAST (presumed “entered wind forecast) is also shown on pp8/9 [15/16 of the pdf].

    I will also have a go now at extracting forecast fuel consumption for the various legs and post result.

  480. David says:

    @Ge Rijn. You posted, @David @Andrew, “As mentioned in an earlier linked paper only the left elevator will be actuated under RAT and the right elevator will be in damped-mode.
    The damped-mode will prevent this right elevator from fluttering but will allow it to move under the various aerodynamic loads.
    Could this have any stabilising or de-stabilising effect (i.e. in (de)stabilising bank angles/roll or pitch)?

    For my part, RAT operative, the left elevator will have its inboard PCU active and the outer will be in bypass. The Davi (2001) description of the elevator PCUs having a damped mode is incorrect/out of date: there are three elevator modes, “Normal”, “Bypass” or “Blocking”, according to the Maintenance Manual 2005.

    The right elevator will be blocked (locked) as Andrew describes once it floats, both PCUs in bypass, to the faired position.
    At about the same time, the left inner PCU will move the left elevator actively to the faired position at autopilot disengagement so there should be no destabilising effect in roll. Any destabilising effect in pitch will result from the shift from what their position had been to faired.

    I intrude about another query you have put, “Are the rudder actuators also pressurised under RAT?” The top PCU of three will be active. As to changes of position without a pilot, the rudder would lose TAC and it might lose any manual trim settings. I leave aside secondary yaw damper commands.

    The other PCUs will be in damped mode (if there had been one inactive that would have been in the bypass mode).

  481. Andrew says:

    Ge Rign:

    “Would it make any difference if the TRUE/NORM switch was set to TRUE before the EOR/Route-discontinuity? Would it still be a constant magnetic heading or would it be a constant true heading?”

    The HDG REF switch selects the heading reference for the PFDs, NDs, Autopilot/Flight Director System and the FMCs. My understanding is the aircraft would maintain a TRUE heading after the EOR if the HDG REF switch were selected to TRUE. Unfortunately I did not have time to test that in the sim.

    “Here’s the paper on the elevator actuator modes. It tells there is a damped-mode. It’s a bit confusing. On the one hand it tells both elevator actuators will go in the damped-mode if both actuators on one elevator surface fail. On the other hand it tells the actuators will go in blocked-mode in this case like you say.”

    Thanks for posting the link; I think some of that information is incorrect. According to Boeing’s technical training notes for the B777, the elevator PCUs don’t have a specific damping mode, although there is some damping when the PCU operates in bypass mode. In the dual engine failure scenario, where the RAT powers the centre hydraulic system primary flight controls, both right hand elevator PCUs will be inoperative and will be in blocked mode. The elevator should be locked in the faired position.

    “…the paper tells there’s no blocked-mode available on the rudder. Only active-mode, damped-mode and bypass-mode. Are the rudder actuators also pressurised under RAT?

    That’s correct; the rudder PCUs have three modes: Normal (Active), Bypass and Damped. The rudder has three PCUs, each powered by a different hydraulic system. The centre system PCU will be powered by the RAT and controlled by the C ACE.

    “It seems in your simulation-test the bank-angle did not go beyond 7 degrees and stabilysed or was stable at 15.000ft when you had to end the simulation. Also the amplitude of the phugoids where not extreme it seems to me. No extreme descents and climbs and no extreme banks that forced a spiral dive. And a fairly stable AoA througout the descent (if I understand it well..).
    You seem to suggest this attitude and behaviour probably continued after descending under 15.000ft till the end.
    Am I correct in interpretating your debrief this way?”

    Yes, that’s how I saw it. The bank and phugoid showed no signs of becoming unstable. I suspect the aircraft would have continued gently turning and porpoising until it hit the water. That could change of course, depending on the exact failure sequence and any atmospheric disturbances the aircraft encountered during the descent, so there’s no guarantee that MH370 ended the same way.

    sk999:

    “…presumably you had the the HDG REF switch in the MAG/NORM position.”

    Yes, that’s correct.

    @DrB:

    “When approaching “abeam ANOKO” in LNAV, wouldn’t the track always be a great circle (geodesic)? In that case, the NORM/TRUE switch should only affect the display, not the aircraft path.”

    Yes, that’s my understanding. The path would only be affected after the EOR point.

    “Do you have any numerical data from your run with time information?”

    Unfortunately I didn’t take detailed notes for the end-of-flight scenario. I didn’t have a lot of time left at that point and was mainly interested in having a look to see what happened after each engine failure. I’ll check the raw photo files and see what I can find.

    Please consider the following scenario:

    1. The Airway N571 is put into the Active Route.
    2. The (15 NM R) lateral offset is selected.
    3. The waypoints after IGOGU are deleted and replaced with ANOKO.

    Will ANOKO have the offset?

    The inserted offset is applied to the entire route, up to the end-of-route waypoint, discontinuity, start of a published STAR or approach, etc, whichever occurs first. I’m fairly sure the route would be offset until abeam ANOKO in the scenario you described.

    @Victor:

    “Do you recall when the A/T disengaged? It must have occurred after the remaining engine thrust advanced to CLB.”

    Yes, it disengaged about the time the VNAV mode dropped out, just before the aircraft started to descend. I don’t think that should have happened and I raised it with the sim engineers after the session. I’ve subsequently had some feedback and they weren’t able to replicate the problem, so I suspect it was a hiccup with the sim.

  482. Kirill Prostyakov says:

    @DrB There are okay images at 23:25(DMSP IR) and 0:30(Elecro L vis), at 0:00 not so okay from FY2B(vis). Of course there are things to be seen around your new dot as well. I want to invest time into general framework for extracting linear features from image strips along synthetic ‘arcs’ using BFO calibrated direction kernel. Acoustics wise, I’m looking at using Hough transform to directionally separate arrivals on HA08 array. There is overwhelming noise from two seismic surveys. Until noise is separated out, HA08 data is unusable. No easy way to look things up, sorry.

  483. Andrew says:

    @Victor:

    Scratch that last reply; my memory’s failing me. I’ve rechecked my notes and the autothrottle dropped out at the same time as the first engine flamed out.

  484. Ge Rijn says:

    @Andrew @David

    Thank you for your detailed clearification. So both rudder and elevators would be fixed in their neutral position on RAT (and without pilot input). And in your simulation-test the plane is expected to descent to the water surface rather gently under a low AoA, (minimal manouvering speed?) and a slight left bank angle. I understand this is no proof MH370 did the same. But at least it seems a clear possibility now which IMO could explain my ‘debris-assumptions’.

    Just one more question on the ailerons if you don’t mind (to both of you).
    I remember (if I remember it well) under RAT the left wing aileron is actuated (like the right wing flaperon) and the right wing aileron isn’t (it is like the left wing flaperon in bypassed-mode). Could it be the left or right wing aileron takes a compensating position against the flaperons asymmetry?

  485. Gysbreght says:

    @Andrew: 1. Deceleration rate after first engine failure.

    The right engine was failed while cruising at LRC/FL360 with a gross weight of 176,000 kg. The aircraft took 11 minutes 52 seconds to decelerate from 264 KIAS to the minimum manoeuvring speed of 208 knots. Note: The cruise reference thrust limit during the exercise was CLB not CRZ. The default limit (ie CLB or CRZ) is selectable according to airline policy.”

    So far our reference has been a test in a simulator witnessed and reported on by ALSM. We do not know whether the reference thrust limit during that exercise was CLB or CRZ. We also don’t know whether that exercise was conducted at standard temperature or something else.

    2. Envelope protection descent speed.

    The aircraft maintained level flight in VNAV PTH until the stick shaker speed (168 KIAS). The aircraft then pitched down and initially accelerated to about 180 KIAS before stabilising at 172 KIAS with a descent rate of 850 ft/min, as shown in the following sequence of photos:”

    I’m struggling to understand the values of deceleration and RoD you observed:

    The deceleration rate after first engine failure at FL360/ISA+10 is 466.7-375.0 = 91.7 kt TAS in 11’52”, or 7.73 kt per minute.

    The indicated (pressure) rate of descent of 850 ftm at 172 kIAS, 33850 ft/ISA+10 corresponds to a horizontal deceleration rate of 37.7 kt per minute.

    Am I doing something wrong?

  486. Andrew says:

    @Victor

    Continuing on, I had to advance the thrust lever to CLB thrust manually, something the autothrottle should have done in the first place!

  487. sk999 says:

    Paul Smithson –

    Thanks – your interpretation of the column headings seems correct.

  488. Andrew says:

    Ge Rijn:

    I remember (if I remember it well) under RAT the left wing aileron is actuated (like the right wing flaperon) and the right wing aileron isn’t (it is like the left wing flaperon in bypassed-mode). Could it be the left or right wing aileron takes a compensating position against the flaperons asymmetry?

    The left aileron will be operative, but the right aileron will be inoperative, as you said. The way I see it, the left aileron will be locked out, because in secondary mode the ailerons are locked out any time the flaps are retracted. The right aileron, on the other hand, will have its outboard PCU in bypass mode due to the loss of the L hydraulic system; and its inboard PCU in blocking mode due to the loss of the R ACE. In blocking mode the actuator is able to retract but not extend. Consequently, the aileron is prevented from floating up, but can move down. The air loads, however, will prevent the aileron from moving down, so it should therefore remain in the faired position.

  489. Ge Rijn says:

    @Andrew

    Clear. So also the ailerons will both be in faired/neutral position.
    It suprises me then that only the left wing flaperon is allowed to deflect 10 degrees upwards under RAT and this upward movement is not blocked like in the right wing aileron by one of its actuators.
    It seems not logical to me the design would allow just one control surface (the left wing flaperon) to destabilise/bank the plane in those circumstances.

    Is it absolutely sure the left wing flaperon will not be blocked like the right wing aileron and can deflect 10 degrees upwards under RAT?

  490. DrB says:

    @Gysbreght,
    @Andrew,

    Gysbreght said: “I’m struggling to understand the values of deceleration and RoD you observed: The deceleration rate after first engine failure at FL360/ISA+10 is 466.7-375.0 = 91.7 kt TAS in 11’52”, or 7.73 kt per minute. The indicated (pressure) rate of descent of 850 ftm at 172 kIAS, 33850 ft/ISA+10 corresponds to a horizontal deceleration rate of 37.7 kt per minute. Am I doing something wrong?”

    Maybe I am missing something here, but is the -37.7 KTAS/min you calculated assuming no thrust? ALSM’s sim showed -19 KTAS/min, and possibly the reference thrust in that case was set to cruise. In Andrew’s test it was set to CLB (climb), which is a higher thrust limit. Thus in Andrew’s case the deceleration will perhaps be less than in ALSM’s case.

    Maybe the deceleration is -37.7 with no thrust, -19 with cruise thrust limit, and -7.7 with climb thrust limit? Did I miss something?

  491. Victor Iannello says:

    @DrB: The difference between CRZ and CLB thrust limits is only several percent, and cannot explain the relationship between horizontal acceleration and descent rate.

    There seems to be an inconsistency somewhere, perhaps related to the disengaged autothrottle, which is another aspect of the simulation that is a bit troubling.

  492. Gysbreght says:

    @DrB: “Did I miss something?”

    Perhaps you have missed my post of March 18, 4:23am, in reply to your post of March 17, 7:25 pm point 5. I referred to Newton’s 2nd law to explain the relation between horizontal acceleration/deceleration and ‘corresponding’ rate of climb/descent at constant speed.

  493. Ge Rijn says:

    @VictorI @Andrew

    Yes, a disengaged auto-throttle would make quite a difference I think too. In a no-pilot end-flight there would be no one to manualy ajust the throttle.
    Like @Andrew stated he did this during the simulation, though it should remain auto-throttle but it didn’t.
    Why not? Or why? It would make quite a difference IMO.

  494. Ge Rijn says:

    To add. I mean with no interference (like @Andrew did) would the throttle remain in the selected thrust position?

  495. Gysbreght says:

    When the A/T disconnected, Andrew pushed the power lever(s) forward manually. If he had not done that, the airplane rate of deceleration would have been higher, not less.

  496. Victor Iannello says:

    @Ge Rijn and @Andrew: According to the FCOM, the autothrottle would disconnect if a “fault in the active autothrottle mode is detected”. Although LRC speed can not be maintained at FL360 with an engine out, that should not in itself cause an autothrottle disconnect.

  497. Victor Iannello says:

    @Gysbreght: We have a reduction in speed over a longer than expected time period, and an unexplained disconnect of the autothrottle. We all understand that lower thrust means a faster deceleration. I am simply wondering whether the two anomalies we see are related somehow. There might be other aspects of the simulation we don’t understand.

  498. Gysbreght says:

    @Victor: We also have a rate of descent that is higher than expected. Therefore “I’m struggling to understand”.

  499. Gysbreght says:

    We also have Andrew’s Descent behaviour after second engine failure:
    “A gentle phugoid motion then developed, with the speed varying from 212 to 235 KIAS, and the rate of descent varying from 500 to 3,500 ft/min”

    If I interpret that as 2000 +/- 1500 ft/min, then the average value is equal to ALSMs.

  500. Paul Smithson says:

    Re fuel flow in the flight plan forecast.

    I calculated FF (kg per minute) for each constant FL “cruise” section of the flight plan (excluding the climbs). Ranges indicated allow for rounding errors on the time and fuel measurements.
    I then recalculated, using Dr B’s fuel model inclusive temp compensation, the predicted FF for respective segments using the starting weight for that section and sub-section speed*/ISA delta.

    Results were as follows:

    FL330, 42 mins, initial weight 217.8: flight plan 121.4 (118.8-124.1); Dr B 113.9
    FL350, 30 mins, initial weight 212.1: flight plan 116.7 (113.1-120.3); Dr B 110.3
    FL370, 23 mins, initial weight 206.2; flight plan 113.0 (108.5-117.8); Dr B 105.4
    FL371, 102 mins, initial weight 198.9; flight plan 102.0 (101.0-103.0); Dr B 100.0

    Overall average flight plan (constant FL sections only) vs weighted average same weights/FL/WX for Dr B’s model

    flight plan 109.6, Dr B 105.2 (4.3% lower).

    *TAS translated to M number for that FL and ISA delta.

  501. Paul Smithson says:

    As regards the PDA difference between the engines, I’m sure that this has been noted elsewhere previously but bears repeating. In the FI we have engine health reports (16:41:58 and 16:52:21) that include fuel flow by engine (parameter WF) for take-off and climb. These indicate 1.9% higher fuel flow R vs L for Take Off and 1.30% higher on R vs. L in Climb.

    If we were to accept the “overall PDA” of 1.5% implied by the flight plan fuel forecast, then this would suggest PDA_Left of 0.6%-0.8% and PDA_Right of 2.1%-2.5%

  502. DrB says:

    @Paul,

    Thanks for doing the comparison of fuel flows. I have several questions:

    1. You are assuming the Flight Brief includes the effect of a 1.5% average PDA, right?

    2. What average PDA did you use for the calculations with my model?

    3. Was the ending weight for any of the segments sufficiently different than the starting weight to noticeably affect the average fuel flow? (is that what causes the variation in the Flight Brief FFs during each leg – changing weight? At least that is what I am assuming you mean by the numbers in parentheses above). If you only used starting weights in my equation you will be biased toward higher FF.

    4. You used the delta TAT equation to increase my fuel equation FFs for the same delta SATs in the Flight Brief, right?

    5. You used the continuous FF equation I had fitted to the Boeing tables (as a function of altitude weight, and Mach), correct?

    6. Does anyone know if the Flight Brief includes the 3% fuel contingency in every leg, or is it simply added on at the end when figuring fuel load?

  503. Victor Iannello says:

    @DrB asked, “Does anyone know if the Flight Brief includes the 3% fuel contingency in every leg, or is it simply added on at the end when figuring fuel load?”

    In the Flight Plan, the takeoff fuel is 49.1 MT and the remaining fuel at ZBAA is 11.9 MT, which means the trip fuel is 37.2 MT. This is also what is used for the trip fuel in the summary analysis BEFORE the 3% contingency is added. So, it looks as though the fuel calculations for each leg do NOT include the 3% contingency.

  504. DrB says:

    @Kirill,

    Thanks for the ststua report on satellite imagery and acoustic data. That’s about what I was expecting.

  505. Niels says:

    @Mick Gilbert, Brock McEwen

    “The surface search and the underwater towed pinger sweeps and sonar scans around 21°S 104°E were “managed” jointly by the Australian Maritime Safety Authority and the ATSB.”

    They got support from US Navy and Royal Australian Navy, probably not the least in terms of expertise in the rich and complicated field of deep sea acoustics.

    Two related links:

    http://edition.cnn.com/2014/05/28/world/asia/malaysia-airlines-pinging/

    http://www.navy.mil/submit/display.asp?story_id=80169

  506. DrB says:

    @Gysbreght,

    On my first reading the part of your post of about the TAS decreasing at constant TAS didn’t register. Now I get it. Thanks.

  507. DrB says:

    Sorry. Should read “TAS decreasing at constant IAS”.

  508. Gysbreght says:

    @DrB: The main part is that sin(FPA) at constant TAS is equal to a/g at constant altitude (for given thrust and drag). The acceleration factor (TAS decreasing in descent at constant IAS) is rather a detail.

  509. Niels says:

    @DrB

    Sorry to ask a question which is probably generated by my lack of knowledge of aircraft control and related terminology:

    I’m trying to understand the consequences of below quote from Andrew’s posting for your scenario where the best holding speed remains selected after say 18:44. Could you please explain?
    Do you consider the possibility that there was no offset involved?

    “4. When does the aircraft decelerate to the BEST SPEED prior to crossing the offset holding fix?

    The aircraft did not decelerate to the best holding speed; it maintained the LRC Mach no. after the aircraft passed the offset holding fix. The FMC does not seem to recognise the hold as part of the offset path, which probably makes sense, given that the path terminates at the offset holding fix”

  510. Paul Smithson says:

    @Victor. Re 3% contingency. Yes, that is also my reading: it is separate and additional to the flight plan fuel forecast. The trip fuel matches the flight plan forecast. The 3% contingency is shown separately in the total fuel breakdown.

  511. Victor Iannello says:

    @Paul Smithson: When looking at the fuel flow for individual segments, I think you have to be careful that the entire segment is at the same flight level. I believe this only occurs if the flight level at a given waypoint is the same as the flight level at the previous waypoint. For FL350, for instance, this only occurs for two legs (AC-BMT and BMT-PCA) for a total of 26 minutes, not the 30 minutes that you show. I think the fuel flow numbers you calculated from the flight plan might be contaminated by climbs, but I can’t be sure.

  512. Paul Smithson says:

    @Dr B. I’ve mailed you the calcs separately. But in brief:

    1. You are assuming the Flight Brief includes the effect of a 1.5% average PDA, right?
    The “flight plan” FF figures are as shown (tonnes /time). My assumption is that the 1.5% is already baked in.

    2. What average PDA did you use for the calculations with my model?
    I used it “as provided” without any custom PDA change

    3. Was the ending weight for any of the segments sufficiently different than the starting weight to noticeably affect the average fuel flow? (is that what causes the variation in the Flight Brief FFs during each leg – changing weight? At least that is what I am assuming you mean by the numbers in parentheses above). If you only used starting weights in my equation you will be biased toward higher FF.
    I re-started weights to match the flight plan for each segment. But yes, because of the difference in FF there is a growing difference in weight and therefore in spot FF as you progress through the leg. The parentheses simply show “hi/lo” interpretation of the flightplan numbers if fuel measurement (decimal tonnes) and time (minutes) has been rounded favourable/unfavourable.

    4. You used the delta TAT equation to increase my fuel equation FFs for the same delta SATs in the Flight Brief, right?
    I applied the temp compensation factor directly to the ISA delta shown in the flight plan

    5. You used the continuous FF equation I had fitted to the Boeing tables (as a function of altitude weight, and Mach), correct?
    Correct

    6. Does anyone know if the Flight Brief includes the 3% fuel contingency in every leg, or is it simply added on at the end when figuring fuel load?
    Answered in previous post

  513. Paul Smithson says:

    @Victor, unless I have made a mistake, the segments shown are constant FL and deliberately omit the climbs/transitions. There should be no “contamination” from climbs, except for the “hot exhaust” aspect that I understand leads to higher consumption shortly after top of a climb.

  514. Gysbreght says:

    P.S. For 172 kIAS at 33850 ft the acceleration correction is a factor of 1.135. Another detail is that in an ISA+10 atmosphere the change of true height is about 4.6% greater than the change of pressure altitude.

  515. AlexSiew says:

    @Victor,

    In addition to what has been pointed out in my previous comment, please note that the story in the FI and Folder 5 that Hangzhou was the second alternate aerodrome makes absolutely no sense. Hangzhou is almost 1200km from Beijing and a nearly 2 hour flight. The usual practice for selecting alternate destination aerodromes is to pick the closest airports to the destination aerodrome. The closest international airports to Beijing Airport were Tianjin and Shijiazhuang, the alternates listed in the Filed Flight Plan, the DETRESFA message and the 2015 ICAO paper presented by Malaysia.

    On March 13, 2014 when the story about the pings first broke, US officials said the pings lasted and the plane had flown on, for up to 4 hours after disappearance from civilian radar (which was at 1.30am). At the same time, these officials said the plane had 4 more hours of fuel. So right from the outset the period of the pings was being matched with fuel endurance. Why only 4 hours, when takeoff was 12.41am, the original time of disappearance from civilian radar was 1.30am and the flight duration was 5 hours and 34 minutes? Answer: the US officials did a back of the envelope calculation from the ACARS data which showed fuel at 1.07am was only 43,800 kg down from 49,100 kg at takeoff.

    The next day (March 14, 2014), the narrative was corrected to say the pings lasted up to 5 hours (to match flight duration). The day after, on March 15, 2014, the narrative was amended again to have the final ping at 8.11am which was exactly 7.5 hours from takeoff (to match fuel reserves). This time was selected as Fuad Sharuji, the MAS operations manager had on March 8, 2014 told CNN that the plane had 7 hours of fuel but for some reason the reports of this interview reported Fuad saying 7.5 hours instead. The video is still available online and so are the reports.

    https://www.bostonglobe.com/news/world/2014/03/13/debris-spot-shown-china-images-says-official/r1Riv5CsBoqfK3Ah2gysQK/story.html

  516. Andrew says:

    @Gysbreght
    @Victor
    @DrB

    The deceleration was timed from LRC (264 KIAS) back to the minimum manoeuvring speed (208 KIAS). The deceleration from the manoeuvring speed to the stick shaker speed (168 KIAS) was not timed. Further, the deceleration is not linear, because of the shape of the drag curve. The aircraft’s drag increases significantly as the speed reduces towards the stick shaker speed. Consequently, the absolute value of (T-D)/W would also increase significantly and is probably what’s responsible for the rate of descent being higher than you expected, based on the deceleration from LRC to the manoeuvring speed.

    In my opinion, the fact the autothrottle disconnected is neither here nor there in terms of the relationship between the deceleration and the rate of descent. If I’d left the thrust alone, both the deceleration and rate of descent would have increased. In any case, the autothrottle should not have disconnected in this scenario and was probably caused by a glitch in the simulator.

  517. Andrew says:

    @Ge Rijn

    s it absolutely sure the left wing flaperon will not be blocked like the right wing aileron and can deflect 10 degrees upwards under RAT?

    I can only go on what’s published in the books. Boeing’s technical training notes state that the flaperon PCUs only operate in two modes: Normal and Bypass. There is no Blocking/Damping mode. The notes also state:

    If both PCUs of the same flaperon are in bypass mode, the flaperon can move freely in both directions. In flight, the airloads then cause the flaperon to move up a maximum of 10 degrees.

  518. Victor Iannello says:

    @Andrew: Thank you. It very well may be that even though the speed is reduced, the drag coefficient rises so quickly that the drag force increases towards the stick shaker speed.

    Would you by any chance have the drag polar curve for a B777-200?

  519. Victor Iannello says:

    @Paul Smithson: To calculate the fuel flow at FL350, I believe you used the three legs TSN-AC, AC-BMT, and BMT-PCA as the time for those three legs add to 30 minutes. The flight levels for the four waypoints are CLB, 350, 350, 350, respectively. I suspect that the first leg, TSN-AC, is not at FL350 for the entire leg. I suspect the flight levels listed with a waypoint are not for the entire leg ending at that waypoint. Rather, they reflect the altitude AT the waypoint.

    Look at the fuel flows per engine for these three legs. Using the groundspeed and the distance to calculate the time (the listed times do not have enough resolution), and then calculating the FF/eng, I get 4373, 3214, and 3388 kg/hr for the three legs. The first leg is suspiciously high. I think the top of climb to reach FL350 occurred during that leg, i.e., at some point between waypoints TSN and AC.

  520. sk999 says:

    Paul Smithson,

    By comparing my fuel usage estimates with what the OFP has around times of step climbs, I get a rough estimate that 0.1 tons of fuel are used to climb 2000 feet. I assume that a like amount of fuel is saved during a descent. By doing so (and filling in the two missing ground speed events by assuming roughly constant air speed), I am able to model the fuel burn and timing for the entire segment IGARI to TOD with no gaps. I now use the OFP tabulated offset in static air temp and convert to true air temp to adjust for deviations from ISA. The agreement is better than one might expect – the total predict duration is 4:27, exactly what is in the plan. The predicted amount of fuel at each waypoint matches exactly to the nearest 0.1 tons (the precision with which it is tabulated) for all but 2 waypoints (off by 0.1). The PDA to make all this happen is now 3.3%.

    The fuel flow model is implemented pretty much the same way everyone else is doing, starting with the Hold and LRC charts that Bobby Ulich unearthed a while ago (and many thanks!) I interpolate in mass, altitude and Mach. The dependence on altitude is highly nonlinear, so I fit the table values with a high-order polynomial. For interpolation in Mach, I use Figure 1 from Boeing’s document “Fuel Conservation Strategies: Cruise Flight”, assume it has the same shape at all altitudes and masses, and scale it, anchored by the Hold and LRC Mach and fuel mileage values. It’s all implemented in the code I posted a while ago – not something one can plop into a calculator or spreadsheet, sorry.

    The real question is how well the OFP matches the actual performance of 9M-MRO. That I cannot answer.

  521. DrB says:

    @all,

    Following Paul’s example, I have done a detailed comparison of my Fuel Model with the MH370 Flight Brief with the same assumptions: CI = 52, ISAT+6C, ZFW = 175.000 MT, and PDA = 1.5%.

    You can get it here.

    I think Victor is correct about some contamination from climbs in certain legs, and I have tried to avoid those.

    The comparison demonstrates agreement within better than 1%. The details are listed in the Notes at the bottom of the page. It can also serve as a reference check for anyone doing a fuel model on their own.

    It’s nice to finally confirm my Fuel Model with independent calculations made by the airline.

    For the PDAs, for now I will use the 1.5% average (from the Flight Brief) and the 1.5% FF excess of the R engine compared to the L engine (from Factual Information) in cruise. On these points I think Victor is correct. That means the R PDA is ~2.26% and the L PDA is ~0.74% (in cruise). The ATSB average of the R/L ratio of FF sensor readings is quite a bit higher than the 1.015 from the FI report, but it has unknown flow rate sensor calibration errors. I have asked ATSB to provide some differences in fuel readings during prior flights. If I can obtain those, perhaps they will shed some additional light on this subject. Another argument that the ratio of PDAs is not much higher than 1.015 is the ratio measured during take-off and climb. These were near 1.01.

  522. DrB says:

    @Niels,

    Based on the PC-simulator results of Victor {and based on what I had been told by another B77 pilot (not Andrew)}, I was expecting the FMC to reduce the air speed from ECON to Holding several minutes before arriving at the Holding fix even with a route offset present. This slow-down does indeed happen when there is no route offset, but Andrew’s run demonstrated that the FMC apparently won’t decelerate prior to the Hold when a route offset is present. Presumably (but not tested), if you set the route offset to zero when the END OF OFFSET message first appears on the CDU two minutes prior to reaching the Holding fix, the FMC would then immediately begin to slow down to Holding speed. But, of course, if you remove the offset the Hold is not aborted and the racetrack will be flown.

    The implication of Andrew’s test is that you cannot get both Holding speed AND End of Route (EOR) Error navigation mode simultaneously with an aborted Hold. You only get the EOR Error. That leaves me with several unresolved issues:

    (1) Can the observed constant magnetic heading (CMH) route be fitted to the satellite data? So far I have been unable to do this, primarily because the 22:41 to 00:11 leg overshoots the 6th Arc. CMH and CTH routes are not very different except for this last long leg from Arcs 5 to 6. CTH fits very well at Holding speed. In order to make CMH fit, you need a rapidly declining air speed at this time.

    (2) Is it possible to obtain CTH navigation by some other means than an EOR error? The answer is yes, it is possible. There are two methods. In one you have to set the NORM/TRUE switch to TRUE and then manually dial in an integer number of degrees heading using the MCP. While possible, I don’t think it is a reasonable explanation. Why would a pilot ever do this? In the second method, the pilot changes the NORM/TRUE switch to TRUE, and then an EOR Error occurs, producing CTH navigation. This is untested, but that’s what Honeywell says the FMC will do. Again, I can’t think of any reason why a pilot would set the switch to TRUE on this flight.

    I make the assumption of a single FMT, and I will continue doing that until I am convinced that no solution is possible, or I find a solution. I will begin by using CMH and some unusual combinations of speed and altitude. If we accept that the average PDA is 1.5%, then Holding is actually too fuel efficient and MRC is too fuel inefficient to achieve the known endurance. We need something in between. I’ll go back and look again at One Engine Inoperative scenarios to see if they might be able to match both PDA and endurance. In addition, MRC at FL385 – 390 comes pretty close to 1.5% PDA, so I will look again at that possibility.

    Regarding the Flight Level, the ICAO guidelines for this situation call for making a 15 NM offset and then adjusting the altitude to a “XX,500-foot” level. Based on that, I think there is a possibility of a climb immediately following the completion of the 15 NM offset at ~18:27 (but prior to the 18:40 phone call). The optimum cruise FL then was FL380, so maybe FL385 makes sense to achieve the 500-foot vertical offset. Still, even if the PDA works out for MRC at FL385, I have not been able to make CMH at that speed fit the BTOs (so far, at least).

  523. sk999 says:

    Bobby Ulich,

    In your comparison with the Flight Plan, it seems that you are using Mach numbers derived from your CI 52 model – is that correct? The Flight plan lists ground speeds, wind corrections, flight levels, and temperature offsets, which combined also yield a Mach number. Have you compared the two?

  524. Andrew says:

    @Victor:

    “Would you by any chance have the drag polar curve for a B777-200?”

    No, unfortunately!

  525. Mick Gilbert says:

    @DrB

    You wrote:”… Holding is actually too fuel efficient and MRC is too fuel inefficient to achieve the known endurance. We need something in between.”

    Might I suggest two scenarios, each is a variation of the autothrottle disengaging during or shortly after the turnback was commanded near IGARI.

    1. Fixed thrust setting equating to around M0.83 at FL350 (flight level and most probable speed at time the turnback was initiated) in “speed-on-elevator” mode (ie constant Mach but gradual climb of about 4,000 feet as fuel is burned to exhaustion); and

    2. Fixed thrust setting equating to around M0.83 at FL340 (flight level and most probable speed a few minutes after the turnback was initiated) in “path-on-elevator” mode (ie constant altitude but gradual acceleration as fuel is burned).

    I’ve hypothesised that the autothrottle may have disengaged due to an autothrottle servo motor circuit breaker being tripped. If that occured during an autothrottle commanded change of thrust setting it could give rise to a very minor thrust asymmetry (the thrust lever associated with the tripped/failed ASM would stop moving immediately whereas the other thrust lever would momentarily continue movement until the autothrottle sensed the ASM failure and disengaged).

  526. Gysbreght says:

    @Andrew: “The deceleration was timed from LRC (264 KIAS) back to the minimum manoeuvring speed (208 KIAS).”

    That time was given as 11 minutes 52 seconds, more than twice what ALSM observed for similar speeds. Is it possible that the timing was incorrect? Do you have a photo of the PFD in that period? The airspeed trend vector may provide a clue.

  527. Mick Gilbert says:

    @DrB

    Apologies, I should have said CAS = 278 kts not M0.83 as the speed.

  528. Paul Smithson says:

    @sk999. If I understand you correctly, you are getting near-perfect match across the whole flight plan from IGARI to TOD, with your model output being 3.3% below the OFP – so requiring “PDA” of 3.3% to align? If we are correct in thinking that 1.5% PDA is already “baked in” to the OFP numbers that would suggest that on “zero PDA” your model would be about 5% lower than an OFP with zero PDA. Please correct me if I’m wrong…

    How does your model adjust for temperature variation from ISA standard?

  529. Gysbreght says:

    @Andrew: “Further, the deceleration is not linear, because of the shape of the drag curve. The aircraft’s drag increases significantly as the speed reduces towards the stick shaker speed. Consequently, the absolute value of (T-D)/W would also increase significantly and is probably what’s responsible for the rate of descent being higher than you expected, based on the deceleration from LRC to the manoeuvring speed.”

    That is essentially correct, but doesn’t explain your observations. In your photo “Envelope protection – stickshaker speed.jpg” the airspeed trend vector shows the speed reducing at about 2 – 3 kt IAS in 10 seconds (12 – 18 kt/min). The drag-to-weight at that speed (170 kt) is about equal to that at the mean speed during the initial deceleration from 264 to 208 kt, as illustrated here.

  530. Andrew says:

    @Gysbreght
    @Victor
    @DrB

    My apologies – I’ve had a look at the metadata of the photos I took and it seems I must have stuffed up the timing somehow, although I’m not sure how. The metadata reveals the following data points:

    6:22:23AM 240 KIAS
    6:22:32AM 234.5 KIAS
    6:23:22AM 211 KIAS
    6:24:12AM 207 KIAS
    6:26:19AM 191.5 KIAS
    6:27:54AM 172.5 KIAS

    According to my calculations (FL360, ISA+10), that gives a speed reduction of 113.5 KTAS over 5:31, or 20.6 KTAS per minute between 240 KIAS and 172.5 KIAS.

  531. Niels says:

    @DrB

    Many thanks for explaining. If you would continue to follow option (2) CTH navigation, you would still need the right IAS “profile” to fit the BTO/BFO data. Is there a thrust/speed setting other than “Best Holding” which could give approx. the same IAS profile?

  532. Victor Iannello says:

    @Andrew: Thank you for the update. The universe is aligned once again.

  533. Oleksandr says:

    Just a test post… more to come.

    This seems to be the place where all the reasonable folks moved to.
    Victor, thanks for launching it. Posting links – does it work in the same way as at JW blog?

  534. Gysbreght says:

    @Andrew: Thank you for your update. Here is your data in graphical form:
    https://www.dropbox.com/s/y1kn1x37jqveh46/Deceleration.pdf?dl=0

  535. Victor Iannello says:

    @Oleksandr: Welcome to the discussion. If you wish insert text with a hyperlink, you can use the standard <a href=> HTML tag sequence.

  536. Brock McEwen says:

    @all: has anyone been able to replicate either my 193° heading – or @Richard’s 216° – using his v16.0 model to zero out the 19:41 BFO error?

  537. TBill says:

    @DrB
    Re: magnetic heading
    Is everyone using 2005 magnetic reference which ALSM says is believed to be MH370 database? I presume the high winds below 22 South give quite a push in heading mode.

  538. ALSM says:

    @Andrew
    @Gysbreght
    @Victor
    @DrB

    I went back and checked the results from one of our Nov 2, 2014 Level D simulations and found that the plane slowed 12.6 kts/min starting from 2nd engine flameout at 35kft/273KIAS/460KTAS. It was nearly a linear speed reduction. I believe the speeds indicate a temp deviation of 0C. So there were some differences in the simulation, but the rate of decent is much smaller (39%). Spreadsheet here: https://goo.gl/crQ9BP

    I wonder if the rate of descent is that sensitive to temperature or initial conditions.

  539. TBill says:

    ..are we using real life winds in the simulations?
    I am thinking they were very strong winds

  540. ALSM says:

    TBill:

    Winds have no material effect on the IAS or TAS at a constant altitude. That said, yes, we did use the GDAS winds in our simulation back in Nov 2, 2014.

  541. Ge Rijn says:

    @TBill

    It looks to me the winds below 22S would have pushed MH370 gradualy to the east but also have been partial head winds.
    It seems to me also magnetic variation would have caused MH370 to fly more eastwards in comparison to a constant true heading.
    This would mean two combined factors that would have caused the plane to divert to the east compared to a straight flightpath.
    Do I see this correct?

    Here a maybe interesting link to calculate magnetic variation at every spot and (past)time on earth:

    https://www.ngdc.noaa.gov/geomag-web/#declination

  542. Gysbreght says:

    @ALSM: Your file MLE_sim_decelerationObservations.xlsx 18.9 kTAS/min starting at first engine shutdown?

  543. Gysbreght says:

    If your minute has 60 seconds.

  544. ALSM says:

    Gysbreght

    For TAS reduction rate, I get 19.2 kts/min speed reduction =(G5-G24)/(C24/60)

  545. ALSM says:

    Gysbreght

    To clarify, Andrew’s numbers were for IAS, so I was comparing his IAS rate to our IAS rate (20.6 vs. 12.5). Our rate of slowing in TAS units was 19.2 kts/min.

  546. Gysbreght says:

    @ALSM: The regression coëfficient is 0.315 kt/s = 12.6 kts/40s = 18.9 kts/60s

    Cheers!

  547. TBill says:

    @Ge Rijn
    Re: Magnetic vs. True – Yes you are correct, the flight could push towards the East fairly strongly at the end of flight due to the wind from the West, because the combined effect of the wind and natural magnetic curve, both to the East. So I am thinking DrB is saying this causes, when he tries Mag Heading, the Arc6 crossing happens too soon for BTO.

    If the FMC was set to Magnetic *Track* for some reason, that might help cancel the wind. Also the assumption of 2005 vintage magnetic setting in MH370 is important as there has been some change over the years (at least in FS9). But I take it 2005 basis was thought to be commonly used by MH370 or standard industry assumption at the time of MH370.

    There is a fundamental issue in that the BTO/BFO are not easily fit by magnetic heading. Victor has shown how it could work, but it violates the ATSB ground rules of a simple FMT and straight flight at constant speed and altitude. I guess we are left with 3 possibilities: (1) flight was not a simple path, (2) FMC was intentionally set to True or an oceanic waypoint (which tends to contradict the assumption that the flight south was unintended), or (3) I suppose if the Magnetic tables in MH370 were updated beyond 2005 then that could explain why it is hard to fit with 2005 magnetic curves.

  548. ALSM says:

    TBill:

    Re #3 above…I can’t find it now, but I am sure ATSB confirmed to me a long time ago that the 9M-MRO data base had the 2005 tables. They change very little over 9 years (~1.7 degrees east ITVO 7th arc).

  549. TBill says:

    @ALSM
    I should probably broaden Item 3 to say more generally some kind of unknown technical issue in the overall data set makes Magnetic Heading hard to fit.

  550. DennisW says:

    @Brock

    @all: has anyone been able to replicate either my 193° heading – or @Richard’s 216° – using his v16.0 model to zero out the 19:41 BFO error?

    I am assuming you are referring to Richard’s values at 4N 93.7E, 294 knots, and 216.3 degrees.

    For Richard’s nominal satellite altitude the BFO error I compute is ~0.3Hz. For the higher nominal satellite altitude used in “Bayesian Methods…” page 29 the BFO error I compute is 1.4Hz.

    I do not know the details of your speed/track/location. Publish them below, and I will give them a go.

  551. DennisW says:

    @Brock

    If you were referring to Richard’s terminal locations, I get very similar results. My current preference being a 19:41 location of ~8.5N, a track around 170, at a ground speed of 480 knots.

  552. Ge Rijn says:

    @TBill

    On the constant magnetic heading, wind and magnetic variation you name 3 options.
    @ALSM states with certainty the 2005 tables were in the data base of 9M-MRO. With ~1.7 degrees of change in 9 years it still would be a ~100 miles difference in longitude that don’t fit the BTO/BFO data well. Quite a lot actualy in terms of defining a possible crash area. So IMO option 3 is rather unlikely compared to the other 2 options.

    Option 1; ‘flight was not a simple path’ assumes complicated actions by an active pilot. Holding patterns, race tracks, descents etc. Overly complicated and not proveable by any means at this time. So also unlikely IMO compared to the one left option you name;

    “FMC was intentionally set to True or an oceanic waypoint”
    This option can fit the BTO/BFO data without the complicated complications the other options need to comply with the BTO/BFO data.

    IMO your logic further understates the most probable option is an active pilot selected the TRUE/NORM switch to TRUE before an EOR/Route discontinuity or selected a particular oceanic waypoint.

  553. Andrew says:

    @ALSM
    @Gysbreght

    To clarify, Andrew’s numbers were for IAS, so I was comparing his IAS rate to our IAS rate (20.6 vs. 12.5). Our rate of slowing in TAS units was 19.2 kts/min.

    The deceleration I calculated was 20.6 KTAS /min based on ISA+10°C @ FL360. The IAS rate works out at 12.2 KIAS/min.

  554. ALSM says:

    Andrew:

    Thanks for clearing up my confusion over the units in you post above. I now see that your IAS and TAS rates are nearly identical to ours.

  555. TBill says:

    @Ge Rijn
    Well if the PIC was still flying, it’s easy to envision hypothetical final maneuver(s) between Arc5 and Arc6 or early maneuvers before Arc2. Or it could have been a ghost flight but a pre-programmed flight plan. But the unplanned ghost flight without an inserted destination is hard to imagine (implication of Andrew’s test work).

    We have not discussed much about maneuvers after Arc5, but I like a turn east to perhaps pick up L894, but I got BTO/BFO probs there too.

  556. TBill says:

    P.S.- We should also credit to Matt who also did simulator training session and did tell us a few months ago that he thought it was Mag Heading after discontinuity, and that’s what Honeywell told him (it depended on True/Mag button push – which by the way is hard to accidentally push because you have to first take the cover off). So we knew maybe that that was the case.

  557. DennisW says:

    @TBill

    My understanding is that MH370 ran out of fuel between 00:11 and 00:20.

  558. AlexSiew says:

    @Dennis,

    “My understanding is that MH370 ran out of fuel between 00:11 and 00:20.”

    MH370 had crashed/ditched many hours earlier. That is why no country has been able to come with any radar data showing MH370 from 1.31am MYT (1731 UTC). A Boeing 777 is a big plane and would show up as a continuous track on radar, not as a few intermittent extrapolated dots over a distance of several hundred nautical miles as some on this forum seem to believe.

    Just like the pings from 1941 to 0011 UTC, the purported 0019 signals are bogus. If they are genuine, we would not have to wait until March 25, 2014 to hear about them. We would have known about them on March 7, 2014 UK time.

    When the official narrative was unveiled on March 15, 2014, as read out by the Malaysian PM, that version of the narrative had a flaw, it had the last ping as a regular handshake, at 0011 UTC. On this narrative, the plane could have flown for another hour before the AES was logged off, so the people doing the cover-up had a problem, how to set up a site for a fake search?

    So this bogus set of signals at 0019 UTC was injected into the narrative subsequently, as a ‘partial ping’ transmitted by the plane in its last throes and on its way down, so the site can be narrowed down to a feasible area for purposes of the fake search. So the 7th arc was born.

    Once again I would pose this question to all readers on this forum, how did Inmarsat miss this set of signals at 0019 UTC? It was not just a signal from the AES on the R channel (the bogus log on request) but also several purported signals in response from the GES to the AES on the P channel to acknowledge the log-on. So when Inmarsat printed out the GES logs on hearing the plane’s disappearance and handed over the logs to SITA within hours, why did these signals not appear on the logs? As have been said many times, all signals from the AES to the GES (on the R or T channels) and from the GES to the AES (on the P channel) would have the AES ID. So it was a question of typing on the computer the AES ID to have a print-out of all transmissions from (and to) the AES on MH370.

    https://www.nytimes.com/2014/03/23/world/asia/a-routine-flight-till-both-routine-and-flight-vanish.html?_r=0

    http://www.corpcommsmagazine.co.uk/features/3724-what-lessons-can-communicators-learn-from-malaysia-airlines-flight-mh370

  559. ALSM says:

    Based on the latest ATSB/Thales info, FE would have occurred between 00:17:00-00:17:30 (allowing for 60-90 sec for APU startup time and 60 sec AES boot time to LOR transmission at 00:19:29).

  560. Marion Ravenwood says:

    This Astro Awani piece for the 3rd anniversary may be of interest here. It contains interviews with Blaine Gibson, Chari Pattiaratchi and two Professors who discuss the Inmarsat data. It is presented in a mix of Malay and English language but much of the Malay parts are explained in English so well worth watching in full: http://www.astroawani.com/video-malaysia/agenda-awani-mengenang-mh370-146070

  561. DrB says:

    @sk999,

    Yes, the Machs in my table are based on my ECON model which uses a relative speed curve I fit to a Boeing Aero magazine graph. I don’t expect “perfect” agreement because the equations Boeing uses do not seem to be published. However, I would be surprised if the difference was more than ~1 %.

    That is a good suggestion to compare the Mach derived from the Flight Plan parameters. I will do that and post the comparison. There are several other comparisons I am doing as well – incremental cruise climb fuel, endurance, etc.

  562. buyerninety says:

    VI note:”…”alternate airports in the DETRESFA message are different than those in the final
    Flight Plan that was used to calculate the fuel load”….”I do think this discrepancy needs to
    be understood because it points to a flaw somewhere between MAS and ATC, although I don’t think
    there were doctored documents”

    Well then, perhaps the discussion on this page may be helpful.
    http://archive.is/Nf31g

  563. Paul Smithson says:

    @Dr B. In the spreadsheet I mailed you I produced equivalent M for every step (using GS+wind correction = TAS, converted to M for the indicated ISA delta) employing the Hochwarth calculator. I’d like to hope that my numbers agree with yours.

  564. Lauren H. says:

    sk999 & Paul – I don’t know for sure but I would say “ENRT” means “Enroute” rather than entered.

    @DrB – Normally a change in Flight Level requires ATC approval so the AutoPilot would not change FL without pilot input. That could be why it was trying to maintain altitude after REFO.

  565. DrB says:

    @Lauren H.,

    REFO = ?

  566. AlexSiew says:

    @buyerninety,

    Re the comments of the ‘Andrew’ on the page linked.

    The ‘diversion fuel’ of 7,700 kg stated in Folder 5 was calculated from the average of the listed alternates (4,800kg + 10,700 kg) = 7,750kg or 7,700 kg rounded up. Likewise the ‘diversion time’ of 1 hour and 16 minutes was calculated from the average of the time to reach the 2 listed alternates (46 mins + 1 hour and 45 mins) = 75.5 minutes or 76 minutes rounded up (1 hour and 16 minutes).

    The real stated fuel endurance of 7 hours and 10 minutes was probably calculated as follows:

    1. Trip = 334 mins
    2. Final Reserve = 30 mins
    3. Farthest alternate destination = around 22 mins (Shijiazhuang)
    4. Contingency 3% = 10 minutes
    5. Additional Reserve 10% of Trip (see FAA regulations) = 33.4 mins

    The Andrew on this forum should be able to confirm the fuel reserve calculations in Folder 5 make no sense.

  567. TBill says:

    I hear several contacting ATSB, if there is a contact I could ask if there is any jet fuel quality data.

  568. DrB says:

    @all,

    I have done three additional comparisons with the MH370 Flight Brief predictions.

    You can get a description of this work and my results here.

    The major findings are:

    1. My predicted Mach numbers agree well with those derived from the Flight Brief parameters.

    2. The time of Main Engines Fuel Exhaustion (MEFE) can easily be predicted using only the Flight Brief fuel data assuming the aircraft stayed at FL371 instead of descending near the end of the planned flight.

    3. The predicted MEFE time using the Flight Brief speed (ECON 52) is ~23 minutes before the actual elapsed time for MH370 MEFE.

    4. Taking into account the somewhat higher temperature (+10C) of the actual flight than the +6C assumed in the Flight Brief, its predicted MH370 endurance at +10C is ~7% low at ECON 52 and FL371.

    5. A different speed mode was likely used in MH370 that had ~7% lower average Fuel Flow than ECON 52.

    6. The extra fuel required for a climb in the Flight Brief is shown to be consistent with my climb fuel equation.

  569. Mick Gilbert says:

    @DrB

    My apologies if this has been covered elsewhere or previously but what is the rationale for modelling a stepped climb to FL390 and then a stepped descent to FL371?

    What does the time to MEFE look like without those climbs and descents?

  570. DrB says:

    @MIck,

    I don’t know the logic. That climb to FL390 and then a descent to FL371 is what is shown in the Flight Brief. I suppose it has something to do with air traffic in China.The difference in Fuel Flows from FL390 to FL371 is quite small. We’re talking less than 1% for the appropriate weights. That translates to less than 4 minutes in endurance.

    At lighter weights, from FL310 to FL390, ECON mode has flat or very slightly declining Fuel Flows (but of course the air speed varies with altitude because the temperature varies). With respect to endurance, there is very little change (on the order of 1%) at any altitude above FL310. The problem with ECON for MH370 is that there is a only a small window to have a PDA = 1.5% and for MEFE to occur at 00:17. This is near FL390. However, at that altitude the air speed is too high to match the BTO data. That means that ECON mode at all Cost Indices can be eliminated for a southerly route with a FMT near 18:40. Maybe Victor has a loiter route with some time at Holding speed and some time at ECON that would also work from an endurance perspective.

  571. Gysbreght says:

    @DrB: I don’t understand your clinging to CI=52. MH370 started cruise at M.82. What is the CI for M.82?

  572. DrB says:

    @Gysbreght,

    I am not “clinging” to ECON 52. Let me explain. It’s just that ECON 52 is what the Flight Brief called for throughout the flight, so I must use that Cost Index when I compare my Fuel Model results with the Flight Brief. We also know that the speed up to IGARI was consistent with ECON 52.

    As I have shown, a continued flight after the diversion at ECON 52 does not match the observed endurance. We also know from the observed ground speeds that the air speed was increased from the turn-back to at least 18:22. This caused higher Fuel Flows during that hour than would have occurred if ECON 52 had been kept. This means that the Fuel Flows after the FMT must have been considerably LOWER than ECON 52 in order to reduce the overall fuel consumption to match the known endurance. Thus there is a need for Holding speed or thereabouts after the FMT.

    The paper I posted this morning at 12:05 AM has the Machs for CI = 52. If you look at the first table at FL350 you see it is very close to M0.82.

  573. Gysbreght says:

    @DrB: “We also know that the speed up to IGARI was consistent with ECON 52.”

    Do we?

  574. Gysbreght says:

    According to my data:
    CI=52; M=0.833
    CI= -17; M=0.82

    FI ACARS Position Report says M.82 at 17:02 and 17:07 UTC.

  575. Lauren H. says:

    DrB – REFO=Right Engine Flame Out – Someone else had previously used that acronym.

    You say MH370 flew at CI=52 before IGARI. I believe Victor has it at LRC between IGARI and 18:22. So you are saying that the air speed was increased after the diversion? (The Ground Speed after IGARI was around 500 KTS due to a tail wind, correct?)

  576. DrB says:

    @Niels,

    You said: “Is there a thrust/speed setting other than “Best Holding” which could give approx. the same IAS profile?”

    No speed profile other than Best Holding has as large a speed decline as far as I know.

    I have not studied thrust settings. If VNAV is also maintaining a constant Flight Level, the a constant thrust setting should increase the air speed as the aircraft gets lighter. This is in the wrong direction to improve the BTO fit. For the speed to drop as needed, I think the thrust needs to decline faster than the weight declines as the aircraft lightens (but I may have this wrong).

  577. DrB says:

    @Lauren H.,

    Yes, after REFO the FMS VNAV will try to maintain the current Flight Level.

    See my post from 10:56 this morning. After turn-back the air speed was increased to approximately LRC. The observed ground speed through 18:22 was 500-505 kts.

  578. DrB says:

    @TBill,

    You said: “Is everyone using 2005 magnetic reference which ALSM says is believed to be MH370 database?”

    I checked, and the magnetic variation table I am now using is not from 2005. It has values appropriate for 2014. This makes a small difference, and unfortunately it makes fitting a constant magnetic heading even more challenging.

    @all,

    Does anyone have a 2005 magvar table they will share?

  579. DrB says:

    @TBill,

    You said: “Yes you are correct, the flight could push towards the East fairly strongly at the end of flight due to the wind from the West, because the combined effect of the wind and natural magnetic curve, both to the East. So I am thinking DrB is saying this causes, when he tries Mag Heading, the Arc6 crossing happens too soon for BTO.”

    Yes, the difficulty is that the path overshoots Arc 6 and is well past it at the time of the measured BTO. The path is quite strongly curving eastward. The only way I can see to match Arc 6 BTO with CMH is for the air speed to drop significantly between 22:41 and 00:11 (even more so than Holding).

  580. Paul Smithson says:

    Following Dr B’s important post.

    If we accept that the FF predicted in the flight brief is representative of 9M-MRO’s fuel consumption (which in turn is consistent with the predicted FF from Dr B’s fuel model), then the logical conclusion must indeed be that the actual flight flew at a slower speed.

    I’m not sure that the difference must necessarily be as large as 7%. Extrapolating forward the flight brief exemplar at FL371 we would expect FF to decline as the plane gets lighter rather than a “straight line” extrapolation. By this method, I obtain expected fuel exhaustion at 251 minutes after start of that segment (03:18 after take-off), making total flight time 07:29 as compared to “observed” 07:36:30 – a difference of 7.5 minutes, or 2.6%.

    Nonetheless, given the efficient profile of the flight brief, the higher ISA delta, the higher speed between Kota Bharu and (at least) Penang, I think the inescapable conclusion must be that a considerably lower FF than CI-52 must have pertained for the majority of the time post-disappearance. Moreover, it suggests that most of the priority search area would not have been fuel feasible with a constant azimuth path (which requires speeds of M0.82+).

    That is extremely puzzling, to say the least. How on earth could $120m have been committed to a search box that was (mostly) not fuel feasible? Surely there must be a disconnect, an assumption error or a mistake somewhere in my logic?

  581. Victor Iannello says:

    @Paul Smithson said, “If we accept that the FF predicted in the flight brief is representative of 9M-MRO’s fuel consumption (which in turn is consistent with the predicted FF from Dr B’s fuel model), then the logical conclusion must indeed be that the actual flight flew at a slower speed.”

    I’m not sure that’s correct. I believe a loiter near holding speed, preferably at a flight level near FL200, followed by a high altitude cruise at MRC speed along a great circle would have the required endurance as well as match the satellite data. The descent to FL200 would also satisfy the BFO at 18:40 along a track of 296°T.

  582. Paul Smithson says:

    @Victor. Would you mind posting this path (BTO/BFO residuals and predicted fuel)?

  583. Victor Iannello says:

    @Paul Smithson: Following Richard’s lead, I’d like to explore whether a descent at 18:40, a loiter at lower altitude near holding speed, followed by a great circle path to YWKS at high altitude can meet all criteria. Preliminary analysis says it would work, but it is a work in progress. Unfortunately, I’m working on other things so that is on the backburner for now.

  584. TBill says:

    @PaulS @DrB
    I am little confused about the fuel statement 7%. In other words, the ACARS report before IGARI had an endurance number, quite consistent with observed I am thinking. So I just want make sure we reconcile or explain the ACARS fuel prediction number as well.

    Yes overall I agree…I have suggested controversy ever since the ATSB issued the first principles review. Ghost planes do not fly straight (True) paths to 38S, and as per Matt/Andrew the discontinuity default should be magnetic/curved path. DrB is implying no fuel to get to 38S anyways. So why were we there? I have trouble getting out of my mind somebody wanted to bathy map that area for reasons other than finding MH370. Seems to me DrB is defining where the orig search should have been centered, but DrB’s path still has a philosophical issue (True/straight flight path).

  585. DrB says:

    tbILL,

    You said: “Seems to me DrB is defining where the orig search should have been centered, but DrB’s path still has a philosophical issue (True/straight flight path).”

    Would you mind explaining what the “philosophical issue” is?

  586. DrB says:

    @Paul,

    You said: “Extrapolating forward the flight brief exemplar at FL371 we would expect FF to decline as the plane gets lighter rather than a “straight line” extrapolation. By this method, I obtain expected fuel exhaustion at 251 minutes after start of that segment (03:18 after take-off), making total flight time 07:29 as compared to “observed” 07:36:30 – a difference of 7.5 minutes, or 2.6%.”

    A straight line extrapolation is an approximation. The slope of the Fuel versus time plot gets smaller as the aircraft gets lighter, so the plot will eventually slope a bit less steeply, giving rise to a later intercept of the time axis at zero fuel, just as you have indicated. Based on my results with the 1-minute stepping fuel model I use, the curvature is small and the shift in the intercept reduces the shortfall (as you have pointed out), but it does not go away. This integrating fuel model is the most accurate method of predicting endurance. The point I wanted to make with the straight-line extrapolation is that it also demonstrates in a very simple way the shortfall in fuel I have been talking about for some time now.

    The higher LRC speed flown from turn-back to ~18:40 will use additional fuel not included in the Flight Brief (an additional 1.7% for 1.3 hours), so the predicted MH370 endurance shortfall will be somewhat larger than the 2.6% you estimated from the Flight Brief alone.

  587. DrB says:

    @Victor,

    You said: “@Paul Smithson said, “If we accept that the FF predicted in the flight brief is representative of 9M-MRO’s fuel consumption (which in turn is consistent with the predicted FF from Dr B’s fuel model), then the logical conclusion must indeed be that the actual flight flew at a slower speed.”
    I’m not sure that’s correct.”

    Which part may not be correct? I think it is quite clear that the average Fuel Flow after 18:40 must have been well below that of ECON 52 and also below that of ECON 0 at cruise altitudes. That means the average speed would also have to be reduced. In that sense of average (not instantaneous) speed, I think Paul’s statement is accurate.

    It is possible, as we both have pointed out, that part of the post-18:40 flight could have been at Holding (and possibly at a low altitude), saving some fuel to allow a higher speed and higher Fuel Flow later in the flight. The feasibility of this has not yet been demonstrated to my knowledge. The climb back to high altitude to allow fuel-efficient high speed also uses up some of the fuel savings garnered while at Holding. I think one has to run some actual durations for the “low and slow” portion after 18:40 plus the “high and fast” portion to see if this scenario is actually feasible for both available fuel and fitting the BTOs.

  588. Victor Iannello says:

    @DrB: I was referring to the part of the flight after the possible holding pattern. The flight from 19:41 through 00:11 might have been at an efficient cruising speed and altitude without the need to cruise at holding speed. As for the additional fuel used in the climb after the hold, don’t forget that you save fuel during the descent to the hold, which tend to offset each other. If the flight after 19:41 was at MRC and high altitude, you don’t need to save a lot during the hold to make the fuel numbers work.

  589. Mick Gilbert says:

    @DrB

    Thank you for your reply, I can see the approach that you have adopted now.

  590. TBill says:

    @DrB
    Forgive me, DrB, I am thinking the True Heading is not consistent with an unintended flight south. I definitely could see ~180T being intentional setting, however.

  591. Paul Smithson says:

    @Victor. I’m with Dr B on this one. If you refer to figures 2 and 3 of ATSB Flight Path Analysis Update Oct. 2014, they modelled max “performance boundary” on MRC assuming a) turn at 1828, FL 250-350 b) turn at 1840 FL 250-400. In both cases, they can barely make it to 38S. I have difficulty imagining a descent, loiter of any significant duration, then climb and MRC making it to [much of] the search box. Besides, MRC is a max range, not max endurance setting and I am doubtful that most of that spectrum of MRC paths would be BTO compliant. I’d be keen to see your loiter-then-onward path model when you have time to return to it.

  592. DrB says:

    @TBill,

    You said: “Forgive me, DrB, I am thinking the True Heading is not consistent with an unintended flight south. I definitely could see ~180T being intentional setting, however.”

    There is nothing to forgive. 180 True Heading could have been set intentionally. It could also have been an unintended route after an EOR error, but apparently only if the NORM/TRUE switch were set to TRUE beforehand. So far, nobody has come up with a rational reason for doing so on this flight. I think a pilot would normally use magnetic for a heading. Perhaps it is possible that a pilot wanted a due south route, but if he did, why not set 180 true track instead of 180 true heading?

  593. sk999 says:

    Sorry if this point has been raised already (been sidetracked here).

    The take-off fuel weight is supposed to have been 49.1 tons. Yet the very first point of the flight plan gives a weight of 47.5 tons, 1.6 tons less. The first point does not even have name – it is just
    N02 46.8 E1010 42.1. What is that location? It is the NW end of runway 32R. At this point the plane should be barely off the ground. Is it reasonable to have used up 1.6 tons of fuel just to get off the ground? By the time it has used another 1.6 tons, it has reached PIBOS. Maybe that’s what really happens. Boy, are these planes thirsty at the beginning!

  594. TBill says:

    @DrB
    Good question. Not sure. I suppose it could have been set as a temporary heading fix, in which case somehow the plane kept going that way. Or if intended for long term setting, the winds from the East would have drifted the plane to West farther away from Indonesia and given a more random flight path with the winds.

  595. DennisW says:

    @all

    Totally “off the wall” post below. Hey, we all need some humor once in awhile.

    So I was whining to my daughter about sk999 using tcl which is not something I am more the vaguely familiar with. Daughter is a professor of computer science at Cal Poly, San Luis Obispo. So she sends me the following link.

    https://erikbern.com/2017/03/15/the-eigenvector-of-why-we-moved-from-language-x-to-language-y.html

    It is fun and humorous to be sure. However, the author makes the following statement:

    ///
    Finding the first eigenvector is trivial, we just multiply a vector many times with the matrix and it will converge towards the first eigenvector.
    ///

    I had never heard that before so I had to give it a try with several different matrices. I would make up a matrix, use Matlab to find the eigenvectors. Then I would create an arbitrary vector i.e. [1 1 1]’ and repeatably multiply it by the test matrix. It is true that the resulting column vector quickly converges on an eigenvector.

    I am still annoyed with sky for using tcl, but at least I learned something in the process.

  596. Victor Iannello says:

    @Paul Smithson said, “In both cases, they can barely make it to 38S. I have difficulty imagining a descent, loiter of any significant duration, then climb and MRC making it to [much of] the search box.”

    I never said that the path ends in the search box. I was referring to Richard’s path that ends near 30S latitude, which is well north of the seabed that was scanned. And I never claimed that MRC speed offers maximum endurance. However, the endurance at MRC speed is greater than any other ECON speed. If endurance was to be maximized, then holding speed at FL200 would be chosen and you would gain about an hour of endurance.

    When I look at all the facts, it is hard for me to believe that this wasn’t a deliberate diversion. It is also hard for me to believe that a pilot would turn south and fly at holding speed for 6 hours.

  597. DennisW says:

    @all

    BTW, the iterative method described above converges on the eigenvector with the largest eigenvalue which is why it is so effective for Goggle. I am so jazzed just learning something. Sorry, I will now step back to the problem at hand. Sorry, Victor.

  598. DennisW says:

    @Victor

    Hey, it was sk999’s fault for using tcl.

  599. buyerninety says:

    @Alex
    That is your understanding. Here’s mine;
    The Flight Plan with alternate airports ZBTJ and ZBSJ, shown in Figure 1.1D
    of the FI, may be explained as an initial filing of that document.
    The date/time 070444 seen therein may support this view. This document may
    have been superceded by the later filing shown in the RMP Report, but was
    incorrectly included in the FI by the FI’s authors, the Malaysian ICAO Annex
    13 Safety Investigation Team for MH370.

    The URL I refered you to in my previous post suggested reasons why different
    airports were listed as ‘alternates’ elsewhere in the FI & in the RMP Report.

    Regarding the flight plan details seen in Folder 5 of the RMP Report,
    rather than viewing them from the point of view of a mathematician, attempting
    to account for every decimal and percentile, instead consider them from the
    viewpoint of ‘absolutely must satisfy the ICAO requirements‘ & thereafter
    experience and/or ‘rule of thumb’ can determine magnitude of remaining details.

    Here is how I account for those RMP Report details;
    ____________________TIME________FUEL
    TRIP________________5:34_______37200

    CONT 3% ZSPD________0:11________1200
    ……..(3% of 37200=1116,{must not be less than 1116}, so round up
    ………and remember that it is China being flown in, THEREFORE; 1200).

    ALTN / ZSJN_________0:46________4800
    ………(for the alternate ZSJN, simply consider that 0:46 is only slightly
    ………more than four times the flight time of 0:11 {each time being flight
    ………from ZBAA airspace}, and so then simply calculate four times the fuel
    ………amount of 1200, resulting in THEREFORE; 4800).

    FINAL RSV___________0:30________2900
    ETOPS / ADDNL_______0:00___________0
    COMP FUEL___________0:30________3000
    BLST_______________________________0
    TOTAL__________________________49100

    Looking further down the page, we see the listing;
    MINIMUM DIVERSION FUEL (time)1:16 (amount)7700
    This reference;
    /http://www.skybrary.aero/index.php/Fuel_-_Flight_Planning_Definitions#Reserve_Fuel_.2F_Minimum_Diversion_Fuel
    informs us of these two variant names used for the same definition;
    “Reserve Fuel / Minimum Diversion Fuel
    Reserve fuel is the sum of Alternate fuel plus Final Reserve fuel.”

    Looking at the figures above, we see that;
    Alternate (ALTN) Fuel 4800 + Final Reserve (FINAL RSV) Fuel 2900 = 7700 .
    (As a check, we can note that those times, 0:46 + 0:30 do = time 1:16 ).

    Seen in the light of the above definition, it can now be understood that
    (the page of the RMP Report that contains the listing of) ‘DIVN FUEL 7700’,
    is derived from the calculation above. The ‘2nd alternate’ airport ZSHC
    is now understood to not contribute to that figure of 7700.

    Regarding the nearness or otherwise of the alternate airports in China to
    the destination airport in China, this URL suggests a basic reason;
    https://web.archive.org/web/20130830064434/http://www.journalgazette.net/article/20130826/BIZ/308269989/1031/BIZ
    that Chinese airports are run much closer to full capacity, and booking a
    landing slot close to your original intended airport is not so easy as it
    is in the West.

    @sk999
    At least the RMP Report suggests a less thirsty figure;
    Takeoff Fuel_____________ 49100
    Taxi_______________________ 500
    Block Fuel Requested_____ 49600
    A comment seen in the URL I cited in my previous post;
    “A 100 kg discrepancy wouldn’t be unusual given the accuracy limitations of
    the refuelling system.”

  600. TBill says:

    @sk999
    Re: Fuel
    Take Off weight was 49.1 (calc) but actual total fuel (I think) was 49.6 including 0.5 for runway Taxi period.

  601. TBill says:

    @Buyerninety beat me to it

  602. TBill says:

    @DrB
    Re: Magnetic corrections, Ge Rijn had posted this link in case it helps you:
    https://www.ngdc.noaa.gov/geomag-web/#declination

    I have magdev correction files for FSX/FS9 for 2005 and 2017 and the orig FS9 which was probably 1990’s vintage.

  603. AlexSiew says:

    @buyerninety,

    Thank you for the reply.

    1. The Filed Flight Plan issued at 070444 at page 11 of the FI showed the alternates as ZBTJ (Tianjin) and ZBSJ (Shijiazhuang), the 2 international airports closest to Beijing Airport.

    2. That there was no change to the alternates can be proven by looking at 2 subsequent documents, firstly the DEFRESFA message issued at 072232 at page 100 of the FI and secondly the ICAO paper presented by Malaysia in January 2015, 10 months later, both listing the alternates as ZBTJ and ZBSJ.

    3. In addition, both the 2 later documents, the DETRESFA message and the ICAO paper, had the fuel endurance at 7 hours and 10 minutes.

    4. The alternates given in the main text of the FI at page 30 and in Folder 5 of Jinan and Hangzhou, and the stated fuel endurance in the main text of the FI at page 1 of 7 hours and 31 minutes, are contradicted by the actual documents issued on the day/night in question on March 7, UTC time, ie the Filed Flight Plan and the DETRESFA message, as well as the January 2015 ICAO paper.

    5. The fuel reserve components calculations in Folder 5 are clearly bogus. As regards alternate fuel, under ICAO rules it is the fuel required to fly to the farthest alternate which under Folder 5 would be Hangzhou but the calculations used the fuel to Jinan instead. No pilot or dispatcher would make such a basic error.

    6. Also according to Folder 5, the enroute alternate was stated as ZSPD (Shanghai Pudong). Hangzhou is near Shanghai and it is common for people flying to either to visit the other. No one in their right sense of mind would pick Hangzhou, 1200 km away, as a destination alternate on a flight to Beijing, especially when the enroute alternate happens to be Shanghai.

  604. Ge Rijn says:

    @TBill

    With the NOAA mag/var calculator I found it interesting to find magnetic variaton from True North starts at ~15S than gradualy increasing to ~13 degrees at 32S. This suprised me for this would mean a deviation from True of ~600 miles to the East only by magnetic variation.

    Am I interpretating things wrong here?
    I conclude from this a magnetic constant heading flight path after an EOR before ~15S would end up at ~32S far more further East than earlier estimates. Impossible to fit the BTO/BFO’s too(?)
    And I can imagine a pilot would not opt for a magnetic constant heading in these regions and change it in time to a True heading or track.

  605. Ge Rijn says:

    To add. If the above is correct implication would be MH370 did not fly a constant magnetic heading after an EOR.

    Or the TRUE/NORM switch was set to TRUE before an EOR resulting in a constant true heading or a distant oceanic waypoint was set and there was no EOR/Route discontinuity.
    Both latter options demand active pilot input for choosing one of those two options.

  606. David says:

    @Gysbreght. You said, “The final BFO is bogus. No one (not even ALSM) has been able to demonstrate or even suggest a scenario that matches the timing of the last two BFO’s”

    I give it a try below, integrating some of my earlier posts; and also raise some more issues affecting descents.

    At page 11, table 2, of its Search and Debris examinations update the ATSB equates these BFOs to descent rates in two postulated cases embracing estimated minima and maxima possibilities. They assess the minimum case’s 00:19:29 descent rate at 3,800 fpm. Eight seconds later at 00:19:37, the estimate is 14,600. The maximum case’s descent rate at 00:19:29 is 14,200 fpm, 8 secs later being 25,000.

    At around stall, 165 KIAS at 36,000 ISA, 295 KTAS, the 00:19:29 descent angle to produce the minimum starting descent rate would be 7.3˚, that for the minimum at 00:19:37 being 29.1˚. The average descent angle increase rate over the 8 secs therefore would be 2.7˚/sec.

    At 00:19:29 the maximum descent angle would be 28.2˚, that at 00:19:37, 56.4˚, the average rate being 3.5˚/sec.

    Naturally if the speed at nose down were higher, as in the ALSM instance where it occurred at left engine failure at 215 KIAS, the descent angles and increase rates would be lessened. Gravity and thrust acceleration even of 1g will make little difference in 8 secs at about 260 fps, so in this context can be disregarded.

    The 165 KIAS descent angles and rates might result from a stall if the left engine failed just before that, FCS then being in secondary and thus stall protection inoperative. As to the left engine failing well before or after stall, to explain the BFOs another contributor would seem needed. This is suggested by the minor ALSM/Andrew transient descent rates – and Andrew’s steady state descent of 1.63˚ (right engine stopped) to 3.5˚ (both stopped).

    That contributor might be the effect of residual trim in the stabiliser. I posted about this shortly after Andrew told us of his findings supposing it could be relevant to part of those. Stabiliser movement normally relieves elevator loading, continuously. This offload ceases on FCS reversion to secondary, that is at aircraft AC power failure and autopilot disengagement on left engine fuel exhaustion, which are followed by RAT deployment.

    The effect of residual stabiliser trim when frozen depends of course on its extent and direction. The direction may well be aircraft nose up at the end of a prolonged autopilot-commanded elevator-up period, as speed decays after right engine failure. If on the other hand this is after a prolonged elevator nose down to obviate stalling at the end of this decay, then it might be frozen in a nose-down position. This would depend not just on left engine failure timing but the time after failure for the engine to reach idle and AC power to be lost.

    Another thing which might affect aircraft trim at the point where any out-of-trim correction has ceased is APU inlet door opening, which commences at AC loss and takes 30 to 40 seconds. (By the way the heavy drag from this might intrude into descent rates after left engine fuel exhaustion. This has not been in discussion of these rates to date, presumably because the effect is supposed common between the two examples. Likewise RAT drag, though I would expect this to be small in comparison).

    There is disparity in simulations, including Boeing’s, to which all the above might contribute, even should it prove that they do not explain the BFOs. For example, in the electrical configuration where right engine failure resulted in AC loss, the APU inlet would open then, rather than at left engine failure.

    As to an explanation of those BFOs, given the expressed Boeing and ATSB confidence in them, something does. They should not have projected such outcomes without a reason apparent from simulation interpretation and I much doubt they would.

    One added point: could any or all the above complicate the character of the “phugoids”?

    @Andrew. My assumption is that the stabiliser will not go to neutral on loss of autopilot or FCS reversion to secondary. Even if it did, it would need sufficient hydraulic supply from the RAT. A stabiliser hydraulic motor consumes 8 gpm (engine driven pumps produce 48, air driven pumps 53: the RAT produces 10). Any comment please?

    Finally, Dr B’s separately calculated descent rates for the BFOs yielded a 00:19:29 4,600 fpm descent, rising to 9,100 at 00:19:37. The initial BFO would be at a flight path descent 1.2˚ more than the 7.3˚ minimum above, the pitch rate much lower at about 1˚/sec.

  607. Andrew says:

    @David:

    The stabiliser will remain frozen at its last commanded position when the PFCS reverts to secondary mode – it will not go to neutral. The only electrical control of the stabiliser in secondary mode is via the pilots’ pitch trim switches. No pilot input = no stabiliser movement.

    Following the first engine flame-out, the autopilot maintains level flight via the elevators and the PFCS then trims the stabiliser via the elevator offload function. The stall protection function stops any further nose-up pitch trim once the speed reduces to the minimum manoeuvring speed and supplies a pitch down command via the elevators at the stick shaker speed. The PFCS then keeps the speed just above the stick shaker via elevator pitch commands.

    As I understand it, the stabiliser remains trimmed for the minimum manoeuvring speed unless there is pilot intervention to trim it further nose down. I assume that trimmed position is partly responsible for the phugoid motion observed in the simulator.

  608. Ge Rijn says:

    @TBill

    Small correction. The magnetic deviation towards the East from True has actually already started North of the equator.
    At the equator 0/92E it’s 1.74 degrees.
    At 15S 95E it’s already 3.4 degrees. ( all march 2014).
    And at 32S 98E it’s 12.91 degrees.

    Look how far those red magnetic declination lines (first graphic) bend towards Australia coming from the West of Sumatra (starting at ~0/92E):

    http://www.ga.gov.au/ausgeonews/ausgeonews200503/geomag.jsp

    I think if MH370 flew a constant magnetic heading since ~0/92E till ~32S along those red magnetic declination lines it ended up far more further East than the 7th arc permits at ~32S.
    I wonder where do I go wrong..

  609. Gysbreght says:

    @David: “Stabiliser movement normally relieves elevator loading, continuously.”
    Firstly, stabilizer movement relieves the elevator with delay, after sustained deflection of the elevator for some time. Secondly, I believe that this only applies as long as the speed is above minimum maneuver speed, and below Vmo/Mmo.

    “One added point: could any or all the above complicate the character of the “phugoids”?”
    Andrew determined that the phugoid period was 45 seconds at 226kIAS average airspeed. The phugoid period for constant angle of attack is 94 sec at 405 kTAS (684 fps). Therefore the airplane motion observed in Andrew’s simulation was not a pure phugoid, but resulted from an AoA that was modulated by the FCS control law. In FCS normal and secondary modes the B777 longitudinal control law is a variant of the C-star control law which is responsive to pitch rate and normal acceleration.

  610. Andrew says:

    sk999:

    Boy, are these planes thirsty at the beginning!

    Yes, they are! Appendix 1.6B of the Factual Information Report contains the engine health monitoring data for MH370. The take-off report shows the total fuel flow at take-off was 46,335 lb/hr or 21,017 kg/hr.

  611. Oleksandr says:

    May I insert my 2 cents with regard to the fuel? If both IDGs and APU were off (i.e. if only transfer L&R buses were powered), the plane would save around 1 ton of fuel on the leg 19:40-00:19. Just of note.

  612. Ge Rijn says:

    @TBill @Oleksandr

    Want to mention also the South magnetic pole is quite different from the North magnetic pole in location compared to their True poles.
    In 2015 the coördinates where;

    South Magnetic Pole 64.28°S 136.59°E[3]

    Deviation from True would start much earlier in the Southern hemisphere.

    @Oleksandr

    Nice to see back here.
    Just another cent..
    Won’t you think at least one IDG was on line again at 19:40 after the SDU re-boot at 19:25?

  613. Oleksandr says:

    @Ge Rijn,

    I would say that the operating SDU implies that the main L AC bus was powered after 18:25 (otherwise there would not be any handshakes). But were the cabin utility switches (FCOM 6.10.1) on? To say more accurately, the uncertainty in the fuel consumed during 19:40 to 00:19 associated with the electrical load is around 1 ton.

  614. DennisW says:

    @Oleksandr

    Hey, welcome back.

    I am still of the opinion that the SDU was never unpowered. Despite DrB’s and ALSM’s arguments to the contrary, the LOR at 18:25 was too accurate with respect to BFO, and did not conform to any of the other LOR’s in Holland’s paper.

    Why the SSWG/DSTG did not do hundreds of LOR’s with similar equipment remains a mystery and a disappointment.

  615. TBill says:

    @Ge Rijn
    The magnetic heading gives a fairly modest bend to the east in MH370. Victor has a paper showing the 180S magnetic trend line. The actual location of magnetic south is not too important (unless you actually try to fly to it) because the magvar correction tables correct that. FS9 reportedly south pole location is way off, but it does not matter because the correction tables adjust to real world curves.

  616. Ge Rijn says:

    @TBill

    I assume the FMC-stored mag/ver correction tables are only used to keep a True heading or track. It seems to me in a constant magnetic heading the plane follows the actual magnetic variation lines (or the tables stored?).

  617. DennisW says:

    @Ge Rijn

    I found the following link to be very helpful.

    http://www.boeing.com/commercial/aeromagazine/articles/qtr_04_09/4/

  618. DrB says:

    @TBill,

    Would you please send me your 2005 magnetic declination file for FSX/FS9. Thanks.

  619. Victor Iannello says:

    @Ge Rijn said, “I assume the FMC-stored mag/ver correction tables are only used to keep a True heading or track.”

    Actually, it is the opposite. The ADIRU determines the track and heading as true directions. Tables stored in the FMS and ADIRU convert the true directions to magnetic directions for navigation and display purposes.

  620. Niels says:

    @DrB “Perhaps it is possible that a pilot wanted a due south route, but if he did, why not set 180 true track instead of 180 true heading?”

    From my path generation tool I get the indication that a 180 TT path may exist that fits the BTO/BFO data. 19:41 position in the N1 – N2 range. I haven’t been able yet to transform GS in TAS and check the speed profile.

  621. Victor Iannello says:

    @Niels: I studied that possibility back in July 2014 when I considered a loiter or landing at Banda Aceh Airport. (We’ve all been doing this way too long.) You’ll find that there is a path due south at 180T that is aligned with waypoint BEDAX and crosses the 7th arc around 34.2S. One of the interesting things about this path is it can be programmed using the waypoints BEDAX-SouthPole. The problem is that part of the arc was searched, albeit not as wide as further south along the arc. Another problem is it doesn’t seem to match the drift data.

  622. buyerninety says:

    @Alex
    You merely restated your opinions in this matter, and did not address the information
    I directed you to. Perhaps it would be more appropriate for you to communicate your
    concerns to the authors of the FI, and thereby increase the liklihood they will address
    this matter in their Final Report.
    As you expressed your viewpoint twice prior to my post to you, your viewpoint on these
    particular points is well known, and there is no need for you to rehash them again,
    unless you uncover new facts.

    @Oleksandr
    I’m curious, you meant ‘ton’, not ‘tonne’?

  623. Oleksandr says:

    Dennis,

    Thanks.

    Re “I am still of the opinion that the SDU was never unpowered.”.

    If the SDU was always powered, then why the two following BFOs were anomalous? I have only one answer: the loss of the satellite link could happen if the ADIRU continued feeding wrong orientation data into the SDU, so that both the satellite antennas were improperly steered for a while.

    Btw, I have set up my own drift model and screened a number of locations along the 7th arc. If Prof. De Deckker’s estimation of the temperature based on the barnacle analysis is accurate, then the most likely crash area appears to be around 30S. The probability of a crash site to be located north of 25S is almost zero, likewise the probability west of 96E is also negligible.

  624. Oleksandr says:

    @buyerninety,

    1000 kg.

  625. Victor Iannello says:

    @Oleksandr: We look forward to seeing your drift analysis that appears to limit possible end points to those falling along the 7th arc between 25S and 32S. It’s interesting that both you and Richard, using different methods, arrive at the same best estimate of 30S along the 7th arc.

  626. DennisW says:

    @Oleksandr

    “If the SDU was always powered, then why the two following BFOs were anomalous?”

    I have no credible explanation for that.

  627. ulric says:

    I think that from BEDAX the 180 Mag heading crosses the 7th arc near 32S. This would seem more consistent with the drift data.

    Endpoints depend critically on the geometry of the final turn and I have read the recent discussions of possible paths after 18:25 with a great deal of interest.

  628. TBill says:

    @DrB
    My understanding is: “Flight Simulator X uses 2005 magnetic variation.” You can update magnetic variation in FSX/FS9 with a freeware data from this site: http://www.aero.sors.fr/navaids.html …so that’s where I got the 2017 files.

    For the older FS9 (PSS777) you can update to the 2005 files at AVSIM.com go to the AVSIM Library and download “magdec2005.zip”.

    As far as I know the FSX/FS9 files are slightly different file structure, so you need magdec.bgl files specifically for FS9 or FSX. I assume mainly only useful for FSX/FS9 purposes.

  629. TBill says:

    @Ge Rijn
    On JW blog someome posted real world GPS methodology overview, basically starts with an ideal sphere and everything is calculation overlay onto that, accounting for earth shape, topography and True or Magnetic headings, so everything is a calc.

    In FS9, I think it just uses an ideal square grid. If you delete the magdec.bgl file, which is a trick to get rid of the magnetic corrections, you just get true headings on a straight grid. This allows flying near the poles in small planes. I do this (delete the FS9 magdec file) to force the A/P to go True Heading after a discontinuity, if I am trying to match that possibility.

Leave a Reply