Preliminary Accident Report Released for SJ182

Sriwijaya Air-NAM Air Boeing 737-524 WL PK-CLC (msn 27323) CGK (Michael B. Ing). Image: 929016.

Today, the Indonesian transportation authority KNKT released a Preliminary Report on Swriwijaya Air flight SJ182, a Boeing 737-500 that crashed after departing Jakarta on Jan 9, 2021. From the report:

On 9 January 2021, a Boeing 737-500 aircraft, registration PK-CLC, on a scheduled domestic flight, took off from Soekarno-Hatta International Airport, Jakarta, to Supadio International Airport (WIOO), Pontianak, at 0736 UTC (1436 LT).

The flight was cleared by Air Traffic Control (ATC) to depart on a Standard Instrument Departure (SID) ABASA 2D to Flight Level (FL) 290. After taking off from Runway 25R, the autopilot was engaged at altitude of 1,980 feet. The pilots subsequently requested a heading change to 075° to enable them to deviate from weather. ATC responded with clearance for heading 075° and the flight began a turn to the right. ATC then instructed the flight to stop climbing at 11,000 feet due to conflicting departure traffic from Runway 25L.

About 10,600 feet, the aircraft heading started turning to the left. About 10,900 feet, the autopilot disengaged, and the aircraft turned to the left and started its descent. At 14:40:37 LT, the radar target of the aircraft disappeared on the ATC radar screen. Thereafter, ATC attempted to obtain information of SJY182 aircraft by calling several times, activating and calling on the emergency frequency, and asking other pilots that were flying nearby. All efforts were unsuccessful to get a response from the SJY182 pilot.

About 1455 LT, the Air Traffic Services (ATS) provider reported the occurrence to the Indonesian Search and Rescue Agency (Badan Nasional Pencarian dan Pertolongan/BNPP), and at 1542 LT, declared the uncertainty phase (INCERFA) of SJY182. The distress phase of SJY182 (DETRESFA) was subsequently declared at 1643 LT.

At the time of issuing this preliminary report, the memory unit of the Cockpit Voice Recorder (CVR) has not been recovered and the search is continuing.

The Komite Nasional Keselamatan Transportasi (KNKT) acknowledged that the safety actions taken by the Directorate General of Civil Aviation (DGCA) and Sriwijaya Air were relevant to improve safety, however there are safety issues remain to be considered. Therefore, the KNKT issued safety recommendations to address the safety issues identified in this report.

This investigation involved the participation of the National Transportation Safety Board (NTSB) of the United States of America as the State of Design and the State of Manufacture, and the Transport Safety Investigation Bureau (TSIB) of Singapore as States providing assistance. Both agencies have appointed their accredited representatives to assist in this investigation in accordance with the provisions in ICAO Annex 13.

The investigation is ongoing. Should further safety issues emerge during the course of the investigation, KNKT will bring the issues to the attention of the relevant parties and issue safety recommendation(s) as required.

Notably, there are findings related to the behavior of the autothrottle, as the thrust to the left engine was reduced during the climb:

After the aircraft climbed past 8,150 feet, the thrust lever position of the left engine started reducing, while the thrust lever position of the right engine remained. The FDR data also recorded the left engine (N1) was decreasing whereas the right engine N1 remained.

The SJY182 pilot requested to the Terminal East (TE) controller for a heading change to 075° to avoid weather conditions and was approved. The TE controller predicted the heading change would make the SJY182 conflicted with another aircraft that was departing from Runway 25L to the same destination. Therefore, the TE controller instructed the SJY182 pilot to stop climbing at 11,000 feet.

The FDR data recorded that when the aircraft’s altitude was about 10,600 feet the aircraft began turning to the left. The thrust lever position of the left engine continued decreasing while the thrust lever position of the right engine remained.

At 14:39:54 LT, the TE controller instructed SJY182 to climb to an altitude of 13,000 feet, and the instruction was read back by an SJY182 pilot at 14:39:59 LT. This was the last known recorded radio transmission by the flight.

The highest aircraft altitude recorded in the FDR was about 10,900 feet, thereafter the aircraft started its descent. The AP system then disengaged with a recorded heading of 016°, the pitch angle was 4.5° nose up, and the aircraft continued to roll to the left to more than 45°. The thrust lever position of the left engine continued decreasing while the right engine thrust lever remained.

About 5 seconds after the aircraft started its descent, the FDR data recorded the autothrottle (A/T) system disengaged and the pitch angle was more than 10° nose down.

At 14:40:48 LT, the radar target of the aircraft disappeared on the TE controller radar screen. Thereafter, the TE controller attempted to obtain information of SJY182 aircraft by calling the flight several times, activating the emergency frequency and calling SJY182 on that frequency. The TE controller also asked other pilots that were flying nearby to attempt contact with the flight. All efforts were unsuccessful to get any response from the SJY182 pilot.

The data from the FDR indicates the start of the turn to the left coincided with a reduction in thrust from the left engine, which would cause a yaw-induced bank to the left, although that was likely moderated by the autopilot. After the autopilot disengaged, the left thrust continued to decrease, the plane rolled left to a bank angle of 45°, and the plane rapidly descended. The pilots were not able to recover from this upset attitude.

Comment: The preliminary report does not discuss what pilot inputs occurred after the turn to the left began, neither while the autopilot was engaged nor after it disengaged. Correct right rudder input would have helped control the aircraft; incorrect left rudder input would have exacerbated the problem.

254 Responses to “Preliminary Accident Report Released for SJ182”

  1. Victor Iannello says:

    Questions for @Andrew (or anybody else):

    1. The plane started turning to the left when the autopilot was engaged. Wouldn’t there be sufficient authority from the ailerons to prevent the turn to the left even if the left thrust went to zero?

    2. Is it possible the pilot flying applied left rudder when right rudder was required, overwhelming the autopilot, and contributing to the increasing left bank when the autopilot disengaged?

  2. Tim says:

    Did wake turbulence add to the roll?

    I would like to know where and what the other aircraft was. If it was a ‘heavy’ that departed off 25L perhaps they had a wake encounter as well as the AT anomaly. I just feel that they would have been at about 250kts indicated, and at that speed the asymmetry should have been fairly benign, unless, of course they let the speed bleed away.

  3. Don Thompson says:

    @Tim asked about “where and what the other aircraft

    The other aircraft was an A320, PK-AZN. At the time when PK-CLC began to turn to the left, rather than right to the heading of 75º advised by ATC, PK-AZN was approximately 10NM directly east of PK-CLC and climbing on a heading of 26º.

    PK-AZN had departed on the same runway heading, but turned to the south, turning left and climbing before establishing its departure to ToC on a heading of 26º.

  4. Tim says:

    Thanks Don, so we can rule out wake turbulence. Might still have to wait for the FDR to explain the rapid departure from level flight.

  5. vodkaferret says:

    While we still need more info it’s looking like pilots struggling with weather and task loading make incorrect response to a mechanical failure.

    Because I might well be wrong (I’m no pilot!), but I would think a failure like that with the autothrottle shouldn’t be something that by itself would lead to airframe loss, but rather be something that a pilot team would be expected to deal with reasonably comfortably??

  6. Andrew says:


    RE: “Wouldn’t there be sufficient authority from the ailerons to prevent the turn to the left even if the left thrust went to zero?”

    I haven’t flown the B737, but I suspect the autopilot doesn’t have enough roll authority to prevent the aircraft turning. In other words, the autopilot can’t apply full aileron.

    RE: “Is it possible the pilot flying applied left rudder when right rudder was required, overwhelming the autopilot, and contributing to the increasing left bank when the autopilot disengaged?”

    That’s possible, but I think the more likely explanation is that the autopilot reached its roll authority limit and subsequently disengaged when it could not control the aircraft. At that point, the ailerons would have snapped back to neutral and the roll rate would have rapidly increased.

  7. Andrew says:


    As you mentioned, we only have limited information at this point, but on the face of it, yes, a crew should be able to handle an autothrottle failure without too much trouble.

  8. 370Location says:

    @Rob and WSPR:

    I’ve made a first attempt at winnowing the WSPR contacts that were detected passing through the SIO.

    Converting maidenhead grid to lat/lon in python was the easy part. More difficult was generating the great circle route arc segments and finding intersections with a polygon that would contain reasonable MH370 flight paths. Here’s a hopefully self-explanatory map and WSPRnet db subset from 2014-03-07@1600Z to 2014-03-08@0200:

    And here are the matching db entries in spreadsheet format:

    Intersections that wrapped around the poles and antimeridian were eliminated, but there was only one contact from inside the boundary across the Pacific.

    There are 586 spreadsheet rows, with the majority between Europe and NSW or Tasmania passing by Sumatra and Malaysia. I could narrow that down further by excluding those regions from the boundary polygon.

    Curiously, there are three great circle arcs seen on the map that end in the southernmost section of the boundary, as if there were three different ships running WSPR hardware out on the ocean. I haven’t sifted through the db to find the contacts yet.

    I might also point out that because my proposed flight path is low and slow, it is less likely that WSPR would catch a skip off a plane at FL150. Still, I hope someone will explore the reduced database for significant matches.

    The TX and RX locations on a WSPR contact might be exported into a .kml file for better visualization of the geodesic signal paths.

  9. vodkaferret says:


    Thank you for taking the time to confirm.

  10. vodkaferret says:

    sorry, should probably have said reply, not confirm.

  11. Victor Iannello says:

    @Andrew: If the autopilot did not have sufficient authority to control the turn due to the thrust asymmetry, then certainly there was insufficient right rudder applied by the pilot flying. It’s unfortunate that pilot inputs were not included in the preliminary report because how the pilots reacted to the thrust imbalance would be important to understanding the flight dynamics.

  12. Andrew says:


    Yes, little or no rudder input would be my guess. As you said, it’s a shame they didn’t include the relevant FDR data in the report. Perhaps they’ll leave us guessing until the final report is released.

  13. Victor Iannello says:

    @Andrew: My guess is that without the CVR, which would give us a fuller understanding, there is reluctance to discuss the human factors.

  14. DennisW says:


    The is no requirement to release FDR data to the public. It is solely subject to the discretion of the IIC (Investgator in Charge).

    Laws restricting the use of FDR data (for legal or disciplinary purposes) vary greatly by country. Often this data is considered confidential and requires a judge (of appropriate jurisdiction) to mandate release to third parties on a case by case basis.

    9.3. FDR data shall not be released to the public, except through the official public report of the accident investigation. Specifically, investigation unit staff, FDR group members, and other individuals with access to the FDR data (including party members) are prohibited from releasing FDR information to unauthorized individuals. However, the IIC
    may decide to publicly release information regarding FDR activity.

  15. 370Location says:

    So, two prior pilot reports of auto-throttle failure within a week of the crash, and three attempts to repair it. The equipment self-test passed despite the problem not being fixed. It would be very interesting to read the detailed failure reports.

    If SIC was on the radio as the thrust asymmetry quickly worsened, I wonder how aware he was of anything the PIC was doing. I hope they do get the CVR data.

  16. 370Location says:

    Here’s a revised WSPR intersection map and dataset:

    The odd station endpoints in the SIO were unlikely ships. Each had a 4 character maidenhead grid code, which is all that’s sent over WSPR. The full 6 char location is fetched from the WSPRnet database if the callsign has ever reported a connection. So, it appears that those transmitters are misconfigured or spoofing data, and have been removed.

    I also decluttered the map by restricting the range to 15,000 km. That’s 135 degrees around the globe, so a much higher chance of the signal not taking the shortest path.

  17. Victor Iannello says:

    @DennisW: True, there is no requirement of releasing the FDR data. That said, we’d know more if they did. In these circumstances, I tend to side with those wanting to release the all the data, even if there is the potential for that data to be misunderstood.

  18. David says:

    @Victor. In the LionAir JT 610 737Max case the Preliminary Report did release FDR data.
    Pages 14-16.

  19. Victor Iannello says:

    @David: Good point. And that was also before the CVR was recovered. I have some thoughts about this, but why do you think the two accidents were treated differently with regard to releasing the FDR data?

  20. DennisW says:


    I don’t know about why various accidents are treated differently, but the FDR data, and especially the CVR data, are tending to fall into the HIPPA domain relative to privacy. Not long ago a couple pilots were terminated and stripped of their commercial licenses based on FDR data (and the testimony of a fellow pilot along for a dead heading ride). It caused quite a stink, and led to laws being passed in the US regarding the use of FDR and CVR data.

    I think any country/operator would be reluctant to release that data based on potential liability issues it might raise.

  21. airlandseaman says:

    Note that PK-LQP (JT 610) also involved an intermittent equipment problem (AoA sensor) that was reported by crews a couple of times before the JT 610 flight, and technicians tried to fix it, but obviously, not successfully.

  22. Andrew says:


    As did the PK-AXC (QZ8501) accident in 2014. That aircraft had an intermittent flight augmentation computer (FAC) fault that was repeatedly ‘fixed’ by resetting the circuit breaker.

  23. Andrew says:


    It’s a bit old, but the following provides some background for the protection of FDR/CVR data:

    In the SJ182 case, I suspect they’ve withheld the FDR data for the time being because they need the CVR to help clarify what happened. I doubt it’s privacy related.

  24. George G says:


    On February 10, 2021 at 8:48 pm you wrote that you suspected that the B737 autopilot can’t apply full aileron. You also wrote that you think that a likely explanation concerning the loss of roll authority:

    “… is that the autopilot reached its roll authority limit and subsequently disengaged when it could not control the aircraft. At that point, the ailerons would have snapped back to neutral and the roll rate would have rapidly increased.”

    To that point, an incident in the UK in 1996 may provide confirmation.

    The aircraft was a Boeing 737-500, G-OBMX.

    It was reported upon in AAIB Bulletin No: 2/97 Ref: EW/C96/9/3 Category: 1.1.

    An extract from that report reads:
    “Enquiries of the manufacturer revealed that the maximum roll control deflection which the autopilot can achieve is, by design, about 4.5° of aileron movement (a control wheel movement of about 20°), depending on the stimulus. The maximum aileron movement which can be achieved by the pilot is about 20°, which requires a control wheel input of about 82°.”

    Although that incident concerned a wake vortex encounter the reported behaviour of the aircraft as a result of the encounter may make be relevant.

    @Victor, you may wish to make that a link.
    As is it will be a Cut and Paste into a Browser job.

  25. George G says:


    Concerning FDR data in preliminary reports:

    There was specific reason for the data to be included in the Preliminary Report KNKT. concerning Boeing 737-8 (MAX); PK-LQP.
    The significance of the (preliminary) findings from the FDR cannot be underestimated.

    “Several significant parameters of the DFDR are shown on the following figures.” from Section 1.11.1. of that report.

    A review of the SAFETY ACTIONs already undertaken and the SAFETY RECOMMENDATIONS from preliminary report concerning SJ182 indicates that the investigators considered other matters of most significance.

  26. Andrew says:

    @George G

    Thanks, I hadn’t previously seen that report.

  27. DennisW says:


    In the SJ182 case, I suspect they’ve withheld the FDR data for the time being because they need the CVR to help clarify what happened. I doubt it’s privacy related.

    Could be so. I don’t really know. Just speculating at this moment.

  28. Don Thompson says:

    Concerning “releasing the FDR data

    Annex 13 sets out that reports comprise sections on ‘Factual Information’, Section 1; ‘Analysis’, section 2; ‘Findings’, section 3; ‘Safety Recommendations’, Section 4.

    Any specific Safety Recommendation will be supported by relevant information/attention in the previous sections. CVR or particular FDR parameters may or not be relevant.

  29. David says:

    @Victor. Withholding FDR data. Another view.

    The G-OBMX report that @George G referenced indicates that aircraft, a 737-500, encountered wake turbulence, resulting in the aircraft banking left.
    As George G mentioned, the autopilot can apply 4.5˚ of correcting aileron deflection, entailing a cockpit wheel turn of 20˚. The pilots were talking with a cabin attendant, both with their hands off the wheel and didn’t notice.
    Once the aircraft reached about 30˚ of left bank, that increasing, they did so and reached for the controls. The commander (the report’s term) disconnected the autopilot and applied opposing rudder and aileron, which corrected the roll and bank. (The FDR indicated that in fact the maximum left roll rate, more than 18˚/sec, actually occurred after the right roll demand was applied).

    Following this the operating company issued an instruction requiring that, “… the handling pilot must be physically monitoring the controls below 10,000 feet AGL.”
    The report added that, “It is not known what would have happened if the autopilot had remained in sole control. In these circumstances, the manufacturers state that there would be no reason for the autopilot to disconnect automatically”.

    I wonder whether that instruction required and received wider attention.

    SJ182 was also a 737-500. From the Preliminary Report report P.1, “At 14:38:42 LT, the FDR data recorded that as the aircraft climbed past 8,150 feet, the thrust lever of the left engine started reducing, while the thrust lever position of the right engine remained.”
    “At 14:38:51 LT, the SJY182 pilot requested to the TE controller for a heading change to 075 to avoid weather conditions and the TE controller approved the request.”
    “At 14:39:47 LT, the FDR data recorded the aircraft’s altitude was about 10,600 feet with a heading of 046 and continuously decreasing (i.e., the aircraft was turning to the left). The thrust lever of the left engine continued decreasing. The thrust lever of the right engine remained.
    “At 14:39:54 LT, the TE controller instructed SJY182 to climb to an altitude of 13,000 feet, and the instruction was read back by an SJY182 pilot at 14:39:59 LT. This was the last known recorded radio transmission by the flight.”

    From that I think it reasonable to infer that the reduction of left engine thrust had gone unnoticed by the flight crew for at least 12 second: it seems quite possible that the bank increased and the autopilot ran out of authority without crew intervention. What is unclear is what would have been the rate of roll and extent of it given the thrust mismatch.
    Boeing would I hope confirm that the autopilot would not disconnect automatically in any like circumstances, since that would make matters decidedly worse.

    There are other issues. The auto-throttle unserviceability of 6 days before, which recurred the following day, being again cleared a day after that, appears to have repeated on this flight. Though the precise natures of the earlier failures remain undisclosed, whether their repetition and possible consequences were evident would be one. The causes of the failures and their possible misdiagnosis would be another.

    The report does dwell on upset recovery and what procedures to apply (and might extend to whether these and/or training are insufficient). Implicitly that is to the forefront of the investigation.

    The FDR would have disclosed whether the left engine throttle was advanced, whether the autopilot was operating correctly, when the autopilot was disconnected manually and whether corrective action was applied – and in particular whether that was appropriate.
    I think there would be grounds to withhold the FDR data if in the above context all that would do would be to confuse, particularly with respect to the action taken by the flight crew and, as @Andrew has said, needing CVR data to clarify. Alternatively it might lead to premature conclusions and speculation possibly unnecessarily upsetting to the NOK, friends and relatives.
    Again there might have been an undisclosed reaction to the release of the JT610 data that has deterred its release in this instance. I think a combination of the first and second to be the most likely.

  30. George G says:


    I have attached an Excel file I prepared earlier today.

    “C:\Users\George\Dropbox\For all\Slow Transition – at initial stage.xlsx”

    The time from when the thrust lever started to decrease (14:38:42) until when the Autopilot dis-engaged (14:40:05) was 83 seconds.

    The initial transition from normal controlled flight to something else may have been deceptively slow.

  31. Victor Iannello says:

    @David: There will be speculation whether or not the FDR data is released. Sometimes the partial release of data is a way for those with the data to control the narrative. I worry about this more than concerns about premature conclusions, but I acknowledge that it is a tradeoff.

    On the face of the evidence, during previous flights, there was a failure in the autothrottle that was reported and not properly repaired. When that failure again occurred during the climb of SJ182, the pilots failed to properly identify and react to the failure. Without proper pilot inputs, notably right rudder, the autopilot reached its authority limits, and the situation became worse after the autopilot disengaged. We are speculating about the pilot inputs because that data was not released, but even without the CVR data to add context, that data would be useful.

  32. flatpack says:


    On February 10, 2021 at 8:48 pm you noted the following:

    “… is that the autopilot reached its roll authority limit and subsequently disengaged when it could not control the aircraft. At that point, the ailerons would have snapped back to neutral and the roll rate would have rapidly increased.”

    Slightly tangental, but do you have any idea why the ailerons ‘snap back to neutral’ as you say, when in the end-of-flight scenarios the vertical stabiliser trim apparently becomes ‘locked in’.

    Maybe the cause of the autopilot disengagement makes a difference somehow.

    I am interested because a while ago I was considering possible ways the end-of-flight residual trim could be less than generally thought.

  33. Andrew says:


    If a control input is removed, the airflow tends to force the associated control surface back to its trimmed position. In the SJ182 case, the ailerons would have moved back towards the neutral position, or close to it, depending on the position of the aileron trim. In the MH370 case, there’s a theory that a small amount of rudder trim was applied. If that were so, the airflow would have caused the rudder to remain at its trimmed position, slightly off-neutral.

  34. TBill says:

    @Andrew @flatpack
    I had the same question: it has been speculated in the past (by some IG members) that ZS may have found a way to automatically (passively) cause MH370 to dive at Arc7.

    So learning from SJ182, that might suggest perhaps the MH370 pilot turned off the Auto-Throttle and intentionally imparted unequal engine input, that somehow upon fuel exhaustion (eg; perhaps with turning off LEFT GEN) the AutoPilot would be locked into a bank at fuel exhaustion. I tried to sim it but did not get too far along.

  35. flatpack says:


    Thank you, that makes sense.

  36. flatpack says:


    I understand where you are coming from and I too was idly speculating about locked in trim causing a bank at fuel exhaustion however my version was a very small bank, essentially the minimum necessary to ‘stabilise’ the descent into the widest possible spiral. Thus giving the longest possible distance from the arc.

    Clearly, no bank at all would nominally give a longer straight path but I quite liked some slight bank because I thought it would impart stability.

  37. Sid Bennett says:


    You asserted that: “it is less likely that WSPR would catch a skip off a plane at FL150”

    Can you elucidate?

  38. airlandseaman says:

    Regarding control surface trim, and how it can cause a spiral descent…

    Pilots know this, but some non-pilots might not, so I thought this would be a good time to point out that control wheel in a car, when turned X degrees, will make a circle oy Y radius. OTOH, when the wheel (stick) is moved X degrees to the left (or right), the plane will start to roll to the left (or right) at a rate of Y degrees per second. IOW, the bank tends to be the integral of the wheel deflection. So, even what might seem like a small trim offset will be integrated over time to become a large bank angle and eventually go inverted.

  39. David says:

    @George G. “The time from when the thrust lever started to decrease (14:38:42) until when the Autopilot dis-engaged (14:40:05) was 83 seconds.”
    Yes, my 12 typo should have read 72. I did not take the extra 11 secs to A/P disengagement into account.

    However from the G-OBMX account there would have been no automatic autopilot disengagement, so that was the point at which flight crew intervention was likely, auto-throttle disengagement being 5 secs later.

    If there was confusion perhaps there was a pro-roll pilot input. What I infer from the discussion about upset recovery in the Preliminary Report is that the investigation does not understand the apparent flight crew reaction.

    @Victor. Yes there will be speculation, including mine above, but were I one of the investigators I would be hesitant to release data from which data-based, but what might prove to be premature, conclusions might be drawn.

  40. DennisW says:


    One sure way to impede progress in any domain is to get lawyers involved. I sense that in this case.

  41. David says:

    @DennisW. In which blame attribution trumps lessons to be learnt. But you make it sound voluntary!

  42. DennisW says:

    There is no valid excuse for withholding information.

    The Wall Street Journal has an editorial policy of not publishing a graphic in which the x-axis is not time. The reason is that it invites readers to draw their own conclusions. The WSJ people want to be able to dominate the conclusion drawing domain for their readers. Japanese publications have the highest percentage of non-time x-axis graphics.

    In the case of the SJ182 FDR, the data is not going to change. What point is there in withholding it?

  43. David says:

    Context. For examples, instrument failure, cockpit disturbance.

  44. Andrew says:


    RE: “So, even what might seem like a small trim offset will be integrated over time to become a large bank angle and eventually go inverted.”

    That’s true. In the SJ182 case it seems the roll was a secondary effect of the yaw caused by the asymmetric thrust. That effect is exacerbated by the aircraft’s swept wing and probably resulted in a very rapid departure from controlled flight if the pilots didn’t handle it correctly.

  45. Andrew says:

    @ George G

    RE: “However from the G-OBMX account there would have been no automatic autopilot disengagement…”

    I was interested to read that in the G-OBMX report because, anecdotally, autopilots do seem to have a habit of disengaging when they can’t control the aircraft. However it happened, the autopilot did become disengaged, followed shortly thereafter by the autothrottle. It might be the case that both systems were manually disengaged by the pilot flying as part of the standard upset recovery procedure. Upset recovery (or lack thereof!) seems to be a focus of the investigation and the CVR would be needed to determine what the crew were doing at the time.

  46. David says:

    @Andrew. Do you have any feel for autopilot rudder authority in the context of a large thrust asymmetry?

  47. airlandseaman says:

    David: In the B777-200 simulations we did, about 3-4 units of rudder kicked in automatically when the first engine flamed out. I’m not sure how those TAC units relate to the rudder deflection angle, but it was well within the normal rudder control range. I expect that would be true for most twins.

  48. Andrew says:


    The rudder itself certainly has enough authority to counter a large thrust asymmetry, assuming the airspeed is at or above Vmca. The B777 TAC is part of the flight control system, not the autopilot, and it does have sufficient rudder authority to balance a large thrust asymmetry. For certification purposes it was detuned slightly to assist the pilot in identifying the failed engine without reference to the engine instruments. Autopilots, on the other hand, normally only control the aircraft in two axes; pitch and roll. On some aircraft, such as the B777, the rudder does control the rudder during an autoland, but not during other phases of flight. The B737 does not have TAC and its autopilot does not control the rudder during any phase of flight.

  49. Andrew says:


    The second last sentence above should read “…the autopilot does control the rudder during an autoland…”.

  50. DennisW says:


    I have a difficult time with the assymetric thrust theory. Both pilots were very experienced aviators (on that arcraft type). Do you really believe that is the ultimate cause of this crash?

  51. Andrew says:


    Based on what we know so far, I’d say that it was almost certainly a contributory factor. We know there was a thrust asymmetry caused by the retarded left engine thrust lever. That thrust asymmetry went uncorrected (unnoticed?) for at least 83 seconds. We also know the aircraft began turning left, away from the crew’s requested heading, and that the bank angle increased to ‘more than 45°’ after the autopilot became disengaged.

    I think the left turn was caused by the uncorrected thrust asymmetry and the associated yaw caused the bank angle to increase significantly when the autopilot disengaged. It seems likely the crew subsequently lost control. The report states the investigation will focus on pilot performance and upset prevention & recovery training (UPRT), amongst other things, which suggests the investigators believe they might be contributory factors.

    Lots of hours on type doesn’t necessarily equate to useful experience that might be handy in unusual situations. We don’t have a full break down of the pilots’ flying experience, but the FO had a total of about 5,107 hours, about 4,957 hours of which was on the B737. That suggests the FO was a cadet who went straight from basic CPL training to the right-hand seat of the B737 and had flown nothing else. Lots of airlines do that nowadays, which is a big change from the past, when junior pilots had to cut their teeth (and gain experience) on much smaller aircraft before moving to jets.

    The problem is that many of these pilots have little or no experience at handling aircraft ‘on the edge’ (eg aerobatics) or in difficult situations, and are ill-prepared if something nasty does happen. There has been an increased focus on UPRT in recent times, but to be honest it’s a canned exercise in the simulator, where the pilot usually knows what’s about to happen. In my opinion, there’s no substitute for aerobatic training and, in particular, learning how to recover an aircraft after falling out of an aerobatic manoeuvre. Perhaps I’m just an old fuddy-duddy.

  52. David says:

    @ALSM, Andrew. Thanks.

    So while the autopilot was holding heading, thrust asymmetry increasing, the right wing would have been increasingly down a little, the aircraft yawed to the left, the rudder then countering the torque from the right engine. That yaw would have continued as the bank reversed, at an increasing rate, after the autopilot was overwhelmed?

    Slowly increasing yaw and right bank, then left roll gradually increasing and the heading change to around 60˚ off course, increased a little by yaw, went unnoticed until bank reached 45˚left. With the whole taking towards a minute and a half, in daylight, clear weather.

    The left auto-throttle malfunction tends to rule out a coincidence with some malfeasance, the beginnings of hypoxia or air disturbance. That malfunction being accompanied by a prolonged cabin disturbance or distraction would be quite a coincidence too unless associated by some common cause.

    @Andrew, while your post to @DennisW provides an explanation as to how failure of upset recovery might arise there is also the above question as to how the build-up would have escaped notice. Can you envisage a common cause between the left auto-throttle failure and something else that would attract/distract the flight crew’s attention?

  53. David says:

    Second line. ‘rudder’ should read ‘rudder and fin’.

  54. David says:

    @Andrew. The PIC’s last upset training was 6th May, 2019, the SIC’s 15th July, 2019.
    Do you know if this is training to a common standard and assessable, or are both up to the airline? Presumably simulator?

  55. 370Location says:

    @Andrew, Thanks for being an old fuddy-duddy pilot. What you say rings true about knowing the limits of the flight envelope. I suspect the crew flying IFR in bad weather were focused on reducing turbulence, and didn’t recognize the situation they were in of losing left engine thrust due to a faulty autothrottle.

    @Sid Bennett, Beware of inviting me to elucidate! I was merely noting that a lower altitude FL150 scenario would have less than half the radio line of sight distance to the horizon vs a FL390 high altitude cruise. Even though analysis of the WSPR database doesn’t appear to support a northern 7th Arc path, it’s still worth exploring.

  56. Andrew says:


    The aircraft was in a right turn when the thrust started reducing. My theory is the thrust reduction went undetected and that rudder was NOT applied to counteract the left yaw caused by the reducing thrust. The secondary effect of yaw is roll, in this case a left rolling tendency in the opposite direction to the turn. Initially, the autopilot would have applied more right aileron to keep the aircraft turning right. However, the thrust asymmetry was increasing and at some point the autopilot would have reached its roll authority limit. At that point, the aircraft would have started rolling left and started to turn left. When the autopilot subsequently disengaged, the ailerons would have gone back to neutral and the roll rate would have rapidly increased unless the pilot applied sufficient opposite aileron to stop the roll and opposite rudder to stop the yaw.

    As for how all this could develop undetected by the crew, consider the following:

    1. The autopilot was engaged. You would be amazed at the number of pilots who fail to effectively monitor the aircraft when the autopilot is engaged.

    2. We don’t know who was flying the aircraft. It appears the FO joined the airline immediately after completing CPL training and had no other flying experience. The sum of his flying career had been spent in the right-hand seat of an airliner, where it is difficult for pilots to develop ‘outside the box’ flying skills. That might also have been true of the captain; we don’t know much about his flying experience.

    3. We know it was daylight, but we don’t know the weather was ‘clear’. The report states there was no ‘significant development of clouds’ along the flight path. By that, I think they mean there was no significant Cb development along the flight path. However, the infrared satellite image shows the cloud top temperature was in the range -34°C to -21°C, which means the cloud tops were well above the aircraft’s altitude. It’s therefore quite possible the aircraft was in IMC rather than ‘clear’ sky, and the crew deprived of visual references.

    4. The thrust reduction occurred slowly, so it would have been somewhat insidious if the crew weren’t properly monitoring the aircraft.

    5. I don’t know what Jakarta is like since the onset of COVID, but normally it is extremely busy and ATC is often chaotic, with a lot of rapid-fire radio chatter. That, combined with the need for weather avoidance, is a powerful distraction.

    This is obviously speculation, but given the above, it might be the case that once the crew did realise something was wrong and initiated the upset recovery procedure, they might have been disoriented and failed to apply the correct control inputs to recover the aircraft. That could have been exacerbated by poor upset recovery training and poorcoordination between the two pilots.

  57. Andrew says:


    RE: “Do you know if this is training to a common standard and assessable, or are both up to the airline? Presumably simulator?”

    Upset recovery procedures and recommendations are published by the aircraft manufacturer in the QRH and the FCTM, so in that sense they are ‘standard’, and in my experience they are assessable. Training, however, is normally something that’s agreed between the operator and the civil aviation regulator. ICAO and the aircraft manufacturers publish guidance, but in this case it seems the regulator did not provide any guidance to the operator. The report states the following: “The investigation was unable to find procedures pertaining to the delivery of upset prevention and recovery training, and guidance from the DGCA to aircraft operators and/or approved training organizations (ATOs) to enable and support the implementation of effective upset prevention and recovery training.”

  58. Andrew says:


    RE: “Presumably simulator?”


  59. Victor Iannello says:

    @David: Even in IMC, in addition to the PFD showing a turn and bank to the left, the yoke wheel would have been turned to the right. Once established in a turn, the wheel should have normally been close to neutral. Of course, there was also the misalignment of the thrust levers. So there were instrument readings and other visual indications in the cockpit that there was a problem that needed manual corrections.

  60. Andrew says:


    There were undoubtedly cues in the cockpit that should have alerted the crew. The question is did they pick up on those cues and, if so, did they take any corrective action? Given the thrust kept reducing over a fairly prolonged period, it seems they either didn’t notice it, or spent far too long trying to figure out what was happening instead of taking manual control. The CVR is obviously crucial to understanding what the crew was doing.

    In IMC, pilots must obviously rely on their instruments to form a mental picture of what the aircraft is doing. Although it becomes second nature to a certain extent, it requires more mental processing than simply looking out the window. That, in turn, reduces the spare capacity the pilot has available for processing other information and can lead to tunnel vision. You’d be amazed at the things some pilots miss when they’re put under pressure. It’s quite illuminating sitting in the instructor’s seat of a simulator and watching how easily things can go wrong.

  61. Andrew says:


    By ‘tunnel vision’ I mean a loss of situational awareness.

  62. Victor Iannello says:

    @Andrew: I didn’t mean to imply that because there were visual clues in the cockpit that it would be routine to take corrective action in a timely manner. Without understanding what was going on in the cockpit, it’s not possible to judge whether or not the pilots performed at an expected level.

    I’ve personally experienced tunnel vision early in flight training when the requirements to aviate and communicate exceeded my bandwidth. As more activities become routine, the more excess bandwidth there is to handle unexpected events. Certainly training helps, but we still all have bandwidth limits.

  63. airlandseaman says:

    There seems to be confusion about the WX conditions at the time and place of the upset/accident. Clearly, it is important to know if the crew was in IMC at the time. The Preliminary report states:

    “The superimposed ADS-B-based flight profile with radar weather image at 1438 LT provided by the BMKG indicated that the radar intensity level along the flight profile was not more than 25 dBz9, which means that the flight path did not indicate any significant development of clouds.”

    The map provided in the PR showed no WX along the final flight path. However, eye witnesses in a fishing boat a few hundred meters from the POI stated there were clouds, rain and lightning in the area. Several other news sources said the same. So which was it? Seems to be a crucial fact necessary to the understanding of what was happening in the cockpit.

  64. Victor Iannello says:

    In this video, it shows that with a single engine operating at FL380 and 220 KIAS, the autopilot of a B737 has sufficient authority to maintain heading without rudder input. At lower altitudes, the thrust asymmetry torque would be higher, so at similar indicated airspeeds, the autopilot would have a harder time compensating for the loss of thrust from one engine.

    [The link should now point to the correct video.]

  65. ventus45 says:


    Pure Speculation:
    There is an interesting Antenna Array on Kerguelen Island ( French Southern and Antarctic Lands )
    Possibly an Over the Horizon Radar (OTHR) ??
    Location: 49.350434°South 70.266352°East

    ARRAY = Receiver = 16 elements = 15 spaces at 15 metre spacing = 225 metre wide array aperture.
    ARRAY = Transmitter = 4 elements = 3 spaces at 15 metre spacing = 45 metre wide array aperture.
    Array Baselines: 257 – 077 degrees True.
    Array Bore-sight: 347 degrees True.

    If it is an OTHR, and if it could scan say 60 degrees either side of bore-sight, that would give it a surveillance arc of 287 to 047 degrees True.

  66. David says:

    @Andrew. Thank you. Re your 1., failure to monitor being common. Indeed G-OBMX was an example.

    After that the airline requirement for ‘hands-on’ beneath 10,000ft in auto-pilot might have been of wider benefit, such as in this instance, had it been adopted broadly.

    Re your 3. and @ALSM later, ie IMC or not, that will be one for the investigation to resolve and perhaps is part of the reason for denying FDR data to the public. It is hard to see how upset recovery would be confusing if not. Disconnecting the autopilot leading to a sharp roll increase to the left visible would lead to the G-OBMX reactions surely, even by my limited piloting experience, once (I learnt using voice tubes in a Tiger Moth).

    Re 5, monitoring radio traffic. Again one would hope the investigation would look into whether there was anything going on at the time over the radio that was particularly distracting.

    I remain curious as to whether whatever caused the auto-throttle failure could have raised a related distraction. For the investigation of the cause of the A/T failure(s) to clarify, I hope.

    Thanks re the common upset recovery training and the Report quote. I had the nagging feeling that lack of manual instructions might be a focus, ie foist on Boeing.

    Re simulator training for that, firstly, sorry I overlooked your answer to @DennisW on that earlier.
    If the recovery action was flawed in this instance, that would raise a case for either more automatic recover assistance or, if infeasible, exposure to such as “falling out of aerobatic manoeuvres” experience that you have mentioned.
    The French with their greater reliance on automation, do they have any automatic recovery in your experience?

    @Victor. “Once established in a turn, the wheel should have normally been close to neutral.” Yes thanks. That tends to deflate the post G-OBMX instruction about hands on monitoring though those hands should pick up the movement all the same.
    Re your discussion with @Andrew re tunnel vision and loss of situational awareness. That can apply in ships too, particularly at night in close company.

  67. Andrew says:


    There were certainly Cb and TS in the area, if not directly on the aircraft’s flight path. The satellite infrared images in the report show what appears to be a large Cb to the south of the crash site and airport. The cloud top temperature in that area is in the band -69°C to -75°C, which suggests considerable vertical development. The METARs for the period before the accident also show Cb and there was a thunderstorm overhead the airport about an hour beforehand. The METAR shortly before the accident shows few Cb at 1700 ft, overcast cloud with a base of 1800 ft and light rain shortly before the accident. The aircraft might well have been in IMC even thought it was clear of any Cb.

    SA 09/01/2021 08:00-> METAR WIII 090800Z 28008KT 4000 -RA BKN016 OVC018 26/24 Q1006 NOSIG=

    SA 09/01/2021 07:30-> METAR WIII 090730Z 30006KT 5000 -RA FEW017CB OVC018 25/24 Q1006 NOSIG=

    SA 09/01/2021 07:00-> METAR WIII 090700Z 30007KT 4000 VCTS RA FEW016CB OVC018 25/24 Q1007 NOSIG=

    SA 09/01/2021 06:30-> METAR WIII 090630Z 34007KT 2000 TSRA FEW016CB OVC018 25/24 Q1007 NOSIG RMK CB OVER THE FIELD=

  68. David says:

    @Ventus45. At that Kerguelen Island spot is a SuperDARN radar for geospace scientific study. I imagine that would be it?

  69. George G says:


    David said on February 13, 2021 at 6:28 pm:
    QUOTE @Victor. “Once established in a turn, the wheel should have normally been close to neutral.” Yes thanks. That tends to deflate … END Quote.

    Please review.

    Once established in a normal turn, yes, the wheel should have returned to close to a neutral position.

    And, had the intentional turn to the right to achieve a new heading of 075 degrees been established then this would have been the case with SJ182.
    BUT, nine seconds BEFORE the pilot requested the heading change the engine thrust split had started. This beginning of the thrust split thus necessitating an autopilot command for (the then slight) aileron movement to roll right may, and probably was, initially almost imperceptible at the control wheel (yoke), even if the PF had had his hand on it.

    As the pilot requested course change, and was given approval, and then the course change command was entered into the Autopilot, the thrust split presumably was slowly increasing (i.e. the left engine thrust setting was presumably slowly reducing). AND the control wheel would have been slowly moving (rolling) to the right.

    It is very likely that when a, or the, pilot recognised that the wheel was moving right then this was just what he would be expecting as autopilot response to the commanded course change.

    NOTE: The word “presumably” has been used in the above very deliberately. We really have NO real idea just in what manner the thrust reduction occurred, but one possibility is that it was progressive throughout the incident. I choose to presume that this is what occurred.

  70. Andrew says:


    In the video, the control wheel deflection seems fairly close to the autopilot limit (20°), even at FL380!

  71. Andrew says:


    RE: “The French with their greater reliance on automation, do they have any automatic recovery in your experience?”

    I don’t know of any aircraft that have ‘automatic recovery’, but modern FBW types certainly have flight envelope protections that should prevent the aircraft reaching an extreme attitude. The B737, on the other hand is relatively ancient design with a conventional flight control system and no protections other than stall warning and bank angle alert (on some aircraft).

    Boeing and Airbus have taken a different approach to flight envelope protections on their FBW aircraft. Boeing’s protections resist the aircraft exceeding certain limits, but they can be overridden by a determined pilot. Airbus protections, on the other hand, prevent the aircraft exceeding the limits, irrespective of the pilot’s control inputs. For example, the protections on Airbus A330/A350 aircraft will not allow the bank angle to exceed 67° (45° if nose down pitch attitude protection or high speed protection is active). If the bank angle is more than 33°, the bank will return to 33° if the pilot releases the side stick (0° if high speed protection is active). Similarly, the pitch attitude cannot exceed 30° nose up or 15° nose down, and the high speed protection will prevent the aircraft exceeding Vmo/Mmo plus a small margin.

  72. David says:

    @Andrew. My thanks again.
    I see. So the A330/A350 would have prevented the seriousness of this.

  73. Andrew says:


    Yes, provided there are no failures that cause the flight control law to degrade. In addition to the protections I mentioned above, the Airbus flight control system automatically reacts to a sideslip and should stabilise the bank angle at about 5° if the pilot does not apply rudder to correct an asymmetric thrust condition. The TAC and flight envelope protections on the B777 (& presumably the B787) would also protect the aircraft if the pilot failed to apply rudder, assuming no flight control law degradation.

  74. George G says:

    Question for Andrew: If it comes about that flight envelope protection, other than pilot intervention, would “have prevented the seriousness of this” as David says, then would not both (or either of) the Boeing or the Airbus systems have provided adequate protection, which is their intention as I understand it. In other words would not a B777 or B787 system have done the job of protection as well as an A330/A350 system, excepting that a “determined pilot” might be able to take the Boeing aircraft out of the normal flight envelope if they deemed it absolutely necessary.

  75. Andrew says:

    @George G

    Our posts crossed paths. The answer is yes; see my post above.

  76. George G says:

    @Andrew, Noted. Thanks.

  77. Victor Iannello says:

    @Andrew said: In the video, the control wheel deflection seems fairly close to the autopilot limit (20°), even at FL380!


    When I first saw the video, my first reaction was it demonstrated that the autopilot had sufficient authority as applied to SJ182. Then, I looked more carefully at the PFD and saw the altitude was FL380. So I believe the question is now settled about whether the autopilot had sufficient authority to handle the thrust unbalance at around 10,000 ft. It does not.

  78. Don Thompson says:

    @David and @ventus45

    Yes, SuperDARN.

    Per SuperDARN at JHU/APL, “Each of the radars has two arrays of antennas. The primary array consists of sixteen antennas. The secondary interferometer array consists of four antennas” which is consistent with the layout apparent on the satellite imagery of Kerguelen Island.

    Unfortunately, it appears to be directed to the south.

    However, it may be worth @Rob’s attention.

  79. David says:

    @George G. Herewith is the review outcome, for my part.
    Whether the right wheel continued to increase after the right course change was entered depends on how much wheel was applied already and how much extra the turn required. However once the right bank turn rate was realised that wheel would reduce to just the amount needed to continue hold that bank and the yaw needed to counter the then thrust asymmetry, probably still small as you say. That though would then increase as the asymmetry increased until the wheel reached its 20˚ limit. Clearly that was reached well before the heading passed back through 046˚ in what was now a left turn. This right wheel increase from close to neutral to 20˚ should have been odd to any pilot keeping an eye open. Even if that was not a wake-up at that early stage the wheel staying would have stayed there while the aircraft bank reversed to the left plus the course reversed, that continuing for some time, should have prompted a reaction.

    I detail this more extensively below, together with some observations. Thank you for prompting this review.

    At 14:38:42 the left engine thrust started reducing. With that, right bank would start, the wheel moving gradually right, left yaw increasing, so heading moving slowly left.

    Nine seconds later a change of course from 290˚ to 075˚ was approved. Once dialled in the autopilot would increase the right bank and wheel from an unknown amount to something more, to realise the required turn rate. Having achieved that it would return towards neutral to continue the turn, the difference to neutral being that to overcome the aircraft’s natural stability and to hold, then gradually increase, yaw as thrust asymmetry rose.

    56 seconds later the heading was 046˚ and decreasing meaning before that the aileron deflection had already reached its autopilot limit (the wheel reaching 20˚ right) as had yaw due to that. The aircraft may or may not have reached its new course before that. So sometime during those 56 seconds the thrust asymmetry had then started to yaw the aircraft yet further so that its heading change reversed. That reversal would have induced left bank, the right wing being more exposed to direct airflow so realising more lift, that leading to left bank and a course change to the left: the secondary effect of yaw that @Andrew mentioned. (I notice the investigation report equates heading with course but that overlooks the yaw.) As thrust asymmetry continued to increase so would the left bank, roll rate being inhibited just by the aircraft’s natural roll stability. The wheel would be at 20˚right still, where it would remain, still trying to counter the still increasing thrust asymmetry, heading and course change continuing to the left. Thus it is evident that at that speed and altitude thrust asymmetry was sufficient to overcome the autopilot aileron limit, confirming @Victor’s observation above.

    Seven seconds later was the last recorded transmission from the aircraft apparently with no sign of concern or alarm in the pilot’s voice so presumably the 20˚ right wheel, left bank and left course change remaining unnoticed.

    It was another 11 secs before action was taken by which time the aircraft had
    rolled to 45˚ left though how much of that was after autopilot disconnect is unclear.

    Even before that the aircraft was in something akin to the classic crossed-controls situation, so much warned against in many aircraft for risk of a spin.

    Observations: A simple torque balance around the z axis would suggest that the side load on the rudder plus fin might be around a quarter to a third of the right engine thrust at that point, tending to bend the fin and just conceivably stall it. Also, the yaw to the left would tend to lift the right wing such that the roll rate might be increasing excessively at intervention, adding further to difficulties in controlling bank on autopilot disengagement ie over and above that experienced by G-OBMX, where bank had increased at autopilot disengagement even without that.

    @Andrew. The A330/A350. Understood. The 737 design reflects its older generation

  80. ventus45 says:


    Damn. Since the French are known to be very protective of their southern territories, and with the rocket / satellite tracking station close by (-49.351669°S, 70.256918°E) I thought it looked possible, even probable. But – too good to be true. Shot down in flames yet again.

  81. David says:

    @ventus45. Worth raising.

  82. DennisW says:


    I stopped concerning myself with that 30 years ago. I often say and do things that are just plain stupid in retrospect. If someone cares, it is their problem. Not mine. I’m not going to change.

  83. George G says:


    Am in general agreement with your summary.
    After reviewing Figure 2 of the KNKT Preliminary into SJ182 and then looking at GERRY SOEJATMAN’s comparison of the final flight of PK-CLC with the previous one, and also AirLandSeaMan’s 3D/2D comparison on Google Earth, I got more curious concerning track data..

    The link below is to a chart of information concerning the track of the final flight from adaption of the SJ182_Granular_data_release which airlandseaman shared in late January. I have blatantly adapted his Excel file; Thank you alsm.

    It is a work in progress. I was initially looking at aspects of the altitude rates during the climb, but have got sidetracked.

    Some points:
    Two relatively stabilised turns to the right can be “seen”. The first is the “usual” departure turn during initial climb. The second is the requested and agreed by ATC weather diversion turn. This was the one intended to reach a heading of 075 degrees.
    A “stabilised” turn was achieved for somewhere between 30 and 45 seconds.

    You can almost “put your finger on” where the autopilot ran out of authority.

    My initial purpose was to see just how much radio traffic was going on and how much this may have been a distraction.
    But time has got ahead of me and I am wary of preparing something too cluttered.

  84. Victor Iannello says:

    @George G: That’s a very illustrative figure.

  85. David says:

    @George G. “You can almost “put your finger on” where the autopilot ran out of authority.”

    Thanks for your diagram. Near enough 14:31:40 was when the right turn started slowing so that would have been it. While the investigation has it that at :47 the aircraft was turning left, from your diagram it had just stopped turning right then and was transitioning to left.
    Heading would have preceded – @Victor’s video indicates yaw of 3deg at 38,000ft to give an idea.

    At 14:40:05, maximum altitude, they have the ‘heading’ at 016 deg (I take it they mean course) whereas yours is at about 024 deg so pretty close.

    Pitch angle they have at 4.5 deg nose up but that is with 45 deg bank.

    A surprise is that the turn rate decreases steadily I see. Head scratch.

    Five seconds later the pitch was 10 deg nose down, a change of almost 3 deg/sec. That to me suggests increased yaw under the thrust asymmetry, wing loading decreased. A/P disengagement at that point.

    @Andrew has suggested that the flight crew might have been intent on diagnosing the problem, presumably being of two minds as to what to do. It was 5 secs later before the A/T was disengaged, the cause apparently being identified at that point.

    I hope the search for the CVR continues.

  86. David says:

    @George G. I see it reached 047˚ (their Fig. 2 about 040˚), so well short of 075˚

  87. Andrew says:


    RE: ”It was 5 secs later before the A/T was disengaged, the cause apparently being identified at that point.”

    Possibly, but it might simply be the point where the crew began the upset recovery procedure. The first steps of that procedure are to disconnect the autopilot and autothrottle. We won’t know unless they find the CVR.

  88. David says:

    @Andrew. Fair enough, thanks. Maybe they never did identify it.

  89. David says:

    @George G. Wait one. You have the right turn from about 285 deg at about 14:38:05. Likewise, their Fig. 2 has that right turn starting before 14:38:16LT.

    Yet their History of the Flight record has that turn being approved at 14:38:51. Does that mean the aircraft turned before that was approved?

    That record has the left engine run down starting 9 secs before that.

  90. David says:

    @Victor. Further on autopilot authority being exceeded and your video being at 38,000 ft, I notice that also the aircraft was in a descent to 28,000. So S & L the engine asymmetric thrust would have been more.

    In SJ182 the good engine was at climb thrust when that authority was exceeded.

    @Andrew. I hope pilots would not leave this aircraft in autopilot after engine failure, irrespective of whether there had been an upset.

  91. Victor Iannello says:

    @David: The plane was drifting down to FL280, but the left thrust was N1 = 98.1%. Essentially, the thrust asymmetry was at its maximum value at FL380.

  92. Andrew says:


    RE: “I hope pilots would not leave this aircraft in autopilot after engine failure, irrespective of whether there had been an upset.”

    Leaving the autopilot engaged after an engine failure is recommended and is not a problem IF the pilots fly it correctly. That would involve manual rudder input to counter the asymmetric yaw, followed by rudder trim. It’s easy to do, but the pilots obviously need to recognise the engine has failed, which comes through appropriate monitoring of the instruments.

  93. David says:

    @Victor, Andrew. OK, thank you.

  94. Victor Iannello says:

    Another part potentially from MH370 was recovered from South Africa.

  95. Victor Iannello says:

    @David, @Andrew: It’s getting a bit down in the weeds, but it’s interesting that the amount of rudder action that completely removes the slip is not precisely when the wings are level. A single engine will produce thrust (parallel to the axis of the plane) and a torque with a moment arm equal to the displacement of the engine from the centerline. This is countered by the rudder, which produces a side force (perpendicular to the axis of the plane) and a torque with a moment arm equal to the aft displacement of the rudder from the COG. Of course, because the moment arm of the rudder is much larger than the moment arm of the engine, the rudder force is a small fraction of the engine thrust. Nonetheless, without some bank to introduce opposing side force, the side force of the rudder is unbalanced, which would result in side slip. So, some bank is required to balance the side force of the rudder for the flight path and the axis of the plane to be aligned.

  96. Andrew says:


    Yes, the best engine-out performance is achieved with the control wheel centred and a small angle of bank towards the live engine. A couple of useful references:

  97. George G says:


    YES, “Near enough” to 14:39:40 is about as close as you can get to when the autopilot ran out of authority..
    Two seconds later, even marginally one second later, the data indicates that the right turn rate had reduced.
    The right bank would now also be reducing.

    Between 14:39:42.680 and 14:39:49.554 the granular data has 23 samples with “47” being recorded for “Track”.
    And twelve of the next fourteen samples have Track values of “46”.
    This means that over the six seconds from 14:39:43 through to 14:39:49 the aircraft transitioned from a right turn to the start of a left turn.

    This is pretty consistent with the report statement that “At 14:39:47 LT, the FDR data recorded the aircraft’s altitude was about 10,600 feet with a heading of 046 and continuously decreasing (i.e., the aircraft was turning to the left).”

    You say: “At 14:40:05, maximum altitude, they have the ‘heading’ at 016 deg (I take it they mean course) whereas yours is at about 024 deg so pretty close.”

    The apparent difference approximates the data scatter bandwidth at this point.

    You say: “A surprise is that the turn rate decreases steadily I see. Head scratch.”

    ? You confuse me here. ?

    The turn rate to the left is in the order of 35 degrees in 10 seconds, 3.5 deg per sec, once it gets going. The drop-off on the screen, or in the chart, is that high turn rate.


    Auto-Pilot and Auto-Throttle Dis-Engagement:

    The Auto-Pilot dis-engaged at 14:40:05 when at maximum altitude. The high left turn rate, mentioned above, had been reached by then. Also by now the aircraft was rolling left at a high rate. That is my thought.

    I guess that the Auto-Pilot auto disengaged due to either the degree of left roll already reached and/or the roll rate.

    We yet have no evidence that the dis-engagement was pilot action.

    Finding and downloading that CVR memory module is a must. It must be giving the investigators some sleepless nights.

    However, when the A/P did dis-engage the, as I see it already high, left roll rate would have been critically exacerbated.

    Five seconds later at 14:40:10 the Auto-Throttle dis-engaged. No comment other than I note Andrew’s comments.

    In the Preliminary Report it is stated that at 14:40:05 and at the highest altitude reached (10,900 feet),
    The “FDR data recorded that the A/P system then disengaged at that point with a heading of 016°, the pitch angle was about 4.5° nose up, and the aircraft rolled to the left to more than 45°.
    Ever since first reading this statement I still cannot reconcile to my own satisfaction whether this meant:
    EITHER that the aircraft had rolled to 45° left at that time;
    OR that the aircraft then proceeded to roll 45° to the left.


    You say: “@George G. Wait one. You have the right turn from about 285 deg at about 14:38:05. Likewise, their Fig. 2 has that right turn starting before 14:38:16LT.”

    In Chart 2.2 I show two different relatively stabilised turns to the right. Figure 2 of the Preliminary report is similar.

    The first stabilised turn to which I have fitted a straight line extends from approx 38:09 and 289 deg to approx 38:43 and 341 deg. This is the 1.6 deg per sec right turn.
    I termed this as the “usual” departure turn during initial climb, for the purposes of clarification in my original post.

    The second is the one of interest, from approx 39:07 and 372 deg to approx 39:37 and 403 deg. This is the 1.05 deg/sec turn.
    This straight section of the flight representation on Chart 2.2 actually starts at about 39:00, or even a bit earlier, but later than 39:51.


  98. George G says:

    ….. but later than 38:51.

  99. TBill says:

    Exciting about the possible debris…fingers crossed we need something,
    usually each anniversary brings up something good.

  100. TBill says:

    @For all the HAMs apparently out there:

    Space Weather News for Feb. 17, 2021

    HAM RADIO SIGNALS FROM MARS: Ham radio operators are doing something that until recently only big Deep Space Networks could do. They’re listening to radio signals from Mars. The “loud and audible” transmissions come from spacecraft orbiting the Red Planet. Some hams are hoping to detect NASA’s Perseverance rover as it touches down later this week. Find out how it’s done on today’s edition of

  101. Victor Iannello says:

    @Andrew: Thanks for those references. It’s a bit comforting when observations from force and torque balances are confirmed by others.

  102. Victor Iannello says:

    @TBill: At this point, the debris is unlikely to have any probative value.

  103. Tim says:

    That debris found on a beach in SA is looking like one third of a B777 left inboard spoiler panel. First piece found in ages. Good news to see large, and significant parts are being found.

  104. airlandseaman says:

    We have a short paper coming together on the new 9M-MRO/MH370 debris. It has been identified as one of the 4 Inboard Spoilers. It is premature to discount the probative value. It could provide additional evidence for (or against) the In-Flight Separation Theory. Should be done soon.

  105. George G says:


    Following on from the above discussions concerning SJ182 I have tidied up the previous chart.

    The straight line fit to the last right turn is now applied from virtually 14:39:00 through to 14:39:40. This has resulted in a slight adjustment in the calculated turn rate from 1.05 to 1.04 degrees per second.

    I have also added vertical indicator lines at the times 14:39:40 and 14:39:42.

  106. David says:

    @George G.
    A most useful figure thank you.

    14:39:47 on your figure was when the rate of change of course was nil, ie a tangent to the course was horizontal, in which case the aircraft had not yet begun the left turn but had ceased the right. Maybe I am being pedantic in the sense it probably matters not but I believe the Report’s statement that it was turning left to be untrue.

    Re my, “A surprise is that the turn rate decreases steadily I see. Head scratch.” That confused you. Yes that was a misinterpretation by me and I now cease scratching. At 14:40:05 it was now steeply banked and turning at about 3½˚/sec.

    You say, “We yet have no evidence that the dis-engagement was pilot action.” From your G-OBMX report Boeing indicted the autopilot should not disengage automatically and @Andrew has said that would be the first step in upset recovery, followed by A/T disengagement. The A/T was disengaged 5 secs later so I think it fair to say that a pilot was the source of the disengagement.

    Whether the autopilot disengagement would have “critically exacerbated” the roll rate, the increase in that would be what aileron deflection from 4.5˚ to neutral would contribute. That depends on what the roll rate was otherwise. On that and your following uncertainty about whether the bank had reached 45˚ by then, they say after mentioning A/P disengagement and pitch angle, “…and the aircraft rolled to the left to more than 45˚.” I think that had meant by then they would have said, “ …..had rolled…”.

    Yes you show two turn rates to the right but between them the left engine thrust starts to reduce. That to me is what causes the change in the previous turn rate that stemmed from the original turn left, reducing it, ie there was the one turn selected. The left asymmetry had a greater effect apparently than the right bank increase, as they increased together.

    So to me that started at 14:38:42 but initially that was imperceptible. There are a few slightly errant pale blues before 38:51.

    PS your latest just noticed. Thanks for the changes though I have not got to that yet.

  107. George G says:



    (not necessarily about the your opinion that the aircraft had not begun it’s left turn by 14:39:47 which I will explain later if need be)

    You are so right when you say:
    “THERE WAS ONE TURN SELECTED.” (my capitals)

    When the pilot requested a course change to a heading of 075° he was not asking for an immediate heading change but for permission to continue the existing turn until he had reached a heading of 075° instead of ceasing the turn at a heading of 035° as had been done for the previous flight.

    The AVGS Playback in the following by Gerry Soeatman on his site helped appreciably in this conclusion.

    AND I missed it !
    In an earlier comment above I wrote: “The first stabilised turn to which I have fitted a straight line extends from approx 38:09 and 289 deg to approx 38:43 and 341 deg. This is the 1.6 deg per sec right turn.”

    In fact I fitted the line up to a data point at 14:38:42.298. Right at the point where the FRD recorded that the left thrust lever started decreasing.

    So, the track data confirms it when you say: “you show two turn rates to the right but between them the left engine thrust starts to reduce. That to me is what causes the change in the previous turn rate that stemmed from the original turn RIGHT, reducing it, ie there was the one turn selected.” (after I changed your typo from “left” to “RIGHT”)

    The plotted data shows a period of transition and decay in turn rate from the initial intended turn at a rate of 1.6 deg/sec for a period of approximately 18 seconds, to a more stable turn rate at “1.04” deg/sec for another 40 seconds before the autopilot ran out of roll control authority.
    The data, could also be validly described by “shows a period of transition of approximately 12 seconds followed by a more stable turn rate for another 46 seconds”.
    Where “one” puts the intersection of these two is arguable.

    But the data does appear to reflect the iniation of thrust lever reduction.

  108. Victor Iannello says:

    Re SJ182: I think we can see with high confidence what happened – Right-to-left thrust unbalance from errant operation of the autothrottle, which was not corrected with sufficient right rudder, which led to a large bank to the left and eventual loss of control. One important question is why the rudder inputs were not correctly applied. Possible reasons are failure to monitor instruments, work overload, spatial disorientation, unfavorable weather, or other mechanical or instrument errors. Without the CVR, we may never know, unless the FDR data has other illuminating information that has not been released.

  109. David says:

    @Andrew. To me that is behind a paywall, maybe to others.
    The article is also at:
    It could be seen as pre-emptive, though the Malaysian Ministry of Transport apparently has taken earlier action. I see that is mentioned in an amendment to the original Bloomberg article.

  110. David says:

    @Victor. On 10th February, in your earlier post, I posted an article by Geoffrey Thomas in the West Australian. That contained an interview, the URL for which is:

    I have today asked the Australian Minister for Transport, who also has other portfolios and is also Deputy Prime Minister, some questions referring to that.
    For some reason my computer did not retain a copy though I have a receipt and am told it has been received, but I have not yet received a promised copy back from them. Maybe I will still.

    This is a long winded way of explaining that the below is the sense of what I asked but is not verbatim:
    “The Hon. Michael McCormack MP,
    as Minister for Transport,
    Parliament House,
    Canberra, ACT

    Dear Minister,

    Malaysian Aircraft Flight MH370 (was underlined)

    As you know a Malaysian Boeing 777, Flight MH370, was lost in the southern Indian Ocean with all crew and passengers seven years ago; the anniversary being in 17 days. There were Australians on board.

    An underwater search for the wreckage was conducted under ATSB management but was unsuccessful. That was followed by another, also unsuccessful, by a private venture, Ocean Infinity. That was at its own cost, having been approved by Malaysia under a no-find, no-fee basis.
    Presumably Malaysia, which conducted the investigation, consulted with others including Australia beforehand.

    A group of experts has proposed a new search, gratuitously. Geoffrey Thomas, the West Australian’s Aviation Editor and a reputable commentator, in an interview with Sky news reported recently that Ocean Infinity is willing to conduct a new search on the same public-spirited terms as before but awaits Malaysian approval.

    My questions, please, are these; has Australia been consulted by Malaysia about this and if so has it put a view, and if not what is its view?

    Yours sincerely,

    the word “interview” forwarded the interview.
    I will let this blog know the outcome, when and if.

  111. Andrew says:


    Boeing’s FOTB is simply a reminder to operators and pilots to follow the well-established procedures for upset prevention and recovery, and “to reinforce the importance of active monitoring of the aircraft state while managing the airplane flight path”. The preliminary report states that Sriwijaya Air issued similar advice to its pilots shortly after the accident and that both the DGCA and the airline are looking at the airline’s upset and recovery training. It seems fairly clear the investigation is focused largely on the pilots’ actions in response to the thrust asymmetry and subsequent upset.

  112. Victor Iannello says:

    @Andrew said: It seems fairly clear the investigation is focused largely on the pilots’ actions in response to the thrust asymmetry and subsequent upset.

    And I believe no Airworthiness Directives (ADs) have been issued.

  113. David says:

    @Andrew. Thanks.
    @Victor. Re my e-mail to the Minister for Transport, I have forwarded a copy to Geoffrey Thomas.

  114. David says:

    @George. Thanks for correcting my turn direction and for describing how the course change approval whilst already turning came to be.

    In your Gerry Soejatman article he describes likely cockpit busyness at a crucial time.

    Also he refers to an article published by and which is drawn from a book written by Chris Brady, ‘The Boeing 737 Technical Guide’, last updated November last year.
    In discussing a flameout the article says, contrary to what has been attributed to Boeing,”The autopilot will compensate for the asymmetrical thrust up to its limits and may then disconnect. Autopilot disconnect must then be accompanied by prompt, appropriate control inputs from the flight crew if the airplane is to maintain a normal attitude. If no external visual references are available, such as when flying over the ocean at night or in IMC, the likelihood of an upset increases.”

    Illustrating what @Andrew has said about the piloting steps to take Soejatman also quotes from a later part, which I think is worth quoting here more fully:
    “A ‘No Thrust Lever Response’ type of malfunction is more subtle than the other malfunctions previously discussed, so subtle that it can be completely overlooked, with potentially serious consequences to the airplane.

    If an engine slowly loses power – or if, when the thrust lever is moved, the engine does not respond – the airplane will experience asymmetric thrust. This may be partly concealed by the autopilot’s efforts to maintain the required flight condition.

    As is the case with flameout, if no external visual references are available, such as when flying over the ocean at night or in IMC, asymmetric thrust may persist for some time without the flight crew recognizing or correcting it. In several cases, this has led to airplane upset, which was not always recoverable. As stated, this condition is subtle and not easy to detect.

    Symptoms may include:
    …….Large unexplained flight control surface deflections (autopilot on) or the need for large flight control inputs without apparent cause (autopilot off).
    Significant differences between primary parameters from one engine to the next.
    If asymmetric thrust is suspected, the first response must be to make the appropriate trim or rudder input. Disconnecting the autopilot without first performing the appropriate control input or trim may result in a rapid roll maneuver.”

    @Victor. Yes, the CVR memory is sorely needed.
    The only other contributions in the offing are; IMC-or-not, radio traffic distractions-or-not and the cause, detection and handling of earlier A/T failures. As to the last in particular I would like to see whether the cause of the A/T failure might have led to associated distractions.

    There is also the question as to whether rudder was applied before autopilot disconnect but that should be disclosed when the in the Flight Data Recorder information is released, in the final report.

  115. Victor Iannello says:

    @David: Thanks for sharing the letter. If a there is a response, it could be interesting.

  116. George G says:


    We must thank alsm for the Gerry Soejatman article; I may not have found that one by myself.

    I cannot guess how many times I have accessed Chris Brady’s “The Boeing 737 Technical Site” ( in the last 28 months. It is very useful.
    (General warning: it comes up as “Not secure” to me.)

    IMC, or no external visual cues, – Highly Likely.

    You, and others,
    may or may not have noticed that:
    The Pilot’s (or a Pilot’s, see quote) response – “and the instruction was read back by an SJY182 pilot at 14:39:59”
    to the ATC instruction – “At 14:39:54 LT, the TE controller instructed SJY182 to climb to an altitude of 13,000 feet”,
    was five (5) seconds later than the instruction as recorded in the preliminary report.

    This was the final radio interaction.

    For all three previous radio interactions mentioned in the preliminary report only the timing of initiation of the radio interaction is reported.

    Six (6) seconds after the recorded time of final Pilot response the Autopilot dis-engaged, at 14:40:05.

    My suspicion is that not long after the autopilot ran out of roll control authority (estimated at 14:39:40) and after the aircraft direction changed from right turn to left turn (certainly before the final radio interaction) the roll change would have be discernible to the pilots. (This is before the dramatic roll change when the autopilot disconnected.)

    At time of the final radio interaction they may have now noticed what the instruments and thrust lever split were telling them and brain activity levels may have been very high.

    was the preliminaryfive (5) seconds later, AND

  117. George G says:

    Unintentional transmission of “was the preliminaryfive (5) seconds later, AND”

  118. Don Thompson says:

    Harking back to recent exchanges about amateur radio

    the judge who could code […] yet another of his many hobbies, ham radio

  119. Victor Iannello says:

    @Don Thompson: Any judge that can code (even in DOS/QuickBASIC) and has an interest in understanding radio propagation is ok with me.

  120. Don Thompson says:

    @George G

    General warning: it comes up as “Not secure” to me.

    That web browser message infers that a web server is not serving content over an encrypted session (URL prefix ‘http’ rather than ‘https’). In order to encrypt the server-browser session, a trusted Certificate Authority must provide a TSL or SSL certificate that verifies some level of identity for the site, often this may be simply be Domain Verification.

  121. George G says:

    @DT Ta

  122. David says:

    @George G. Perhaps the voice tenor in an ATC recording of the final transmission will give a clue.

  123. David says:

    @George G. For the record, about the Report’s ambiguity as to whether at 14:40:05 the aircraft was already banked to 45˚ or alternatively was going to roll to that, at a TAS of 300 KTAS and turning at 3½˚/sec (my measure from your graph) the bank angle would have been 44˚. If TAS instead was 280 KTAS that would have been 42˚.

    That suggests it was about there already.

  124. George G says:


    “about there already” Thank you. Consistent with the data we have.

    Two more Chart Prints, being altitude plots. Not yet sure if of any help other than data record.

  125. David says:

    @George G. Could come in handy.
    However I maintain still that in your 2.2.v2 at 14:39:37 the aircraft wings were level, the course change being zero at that point.

  126. David says:

    Should read 14:39:47

  127. Victor Iannello says:

    Flight 328, a B777-200 with P&W engines, had an engine failure soon after departing Denver bound for Honolulu. Cowling parts were found in nearby Broomfield. A passenger took this video of the engine still flaming and windmilling.

  128. George G says:

    Will prepare visual explanation.

  129. Victor Iannello says:

    Here’s the ADS-B data together with the pilot mayday call.

    No doubt the TAC in the B777 helped the pilot to control the plane.

    Here’s some of the engine parts found on the ground:

  130. George G says:


    Before I go and cut grass, which I must do …

    (David says: February 20, 2021 at 5:52 pm) : “However I maintain still that in your 2.2.v2 at 14:39:37 the aircraft wings were level, the course change being zero at that point.”

    The wings may well have been level at 14:39:47.

    I’m not sure that we have enough information, in absence of full relevant FDR information, to to determine actual timing of when the wings were level.

    Earlier, you said: “While the investigation has it that at :47 the aircraft was turning left, from your diagram it had just stopped turning right then and was transitioning to left.” (David says: February 16, 2021 at 5:04 pm).

    And, later, you said: “14:39:47 on your figure was when the rate of change of course was nil, ie a tangent to the course was horizontal, in which case the aircraft had not yet begun the left turn but had ceased the right. Maybe I am being pedantic in the sense it probably matters not but I believe the Report’s statement that it was turning left to be untrue.” (David says: February 18, 2021 at 4:22 am)

    To which I answered that you were necessarily right in the your opinion that the aircraft had not begun it’s left turn by 14:39:47, and I wrote “which I will explain later if need be”.

    I had earlier written “This means that over the six seconds from 14:39:43 through to 14:39:49 the aircraft transitioned from a right turn to the start of a left turn.”, based upon there being 23 samples all with a recorded “Track” value of 47. (February 17, 2021 at 10:39 am)

    During that time period of six seconds as the track change from below, through to return to below “47” degrees, the actual transition from right to left turn must have occurred. And, I am reading the word “Track” literally, as distinct from any interpretation that it was actually “Heading”.

    As the aircraft, by the time of track reversal, would have had a significant Right Engine to Left Engine thrust imbalance, there would have been significant yaw to the left.
    When the wings became level during the transition from right to left turn, and with that yaw, the aircraft would have already been into it’s left turn.

    Only if we knew the actual amount of thrust imbalance might we be able to estimate by how much time the turn transition preceded the wings level. I suspect that the thrust balance was truly significant by the time of track reversal, so significant that, as we now know, there needed to be more than (an estimated) four and half degrees of right aileron to counter it.

    So, as above, “The wings may well have been level at 14:39:47.”
    I actually suspect the thrust imbalance was sufficiently significant that some right bank still existed, and the aircraft had not yet returned to wings level condition..

    I had been hoping to be able to estimate just how much thrust imbalance with the data we have, in absence of data from the Flight Data Recorder. No chance. But, someone with knowledge of the aircraft performance of a (suspected lightly loaded) B737-500 should be able to come up with an approximation just on what I just realised I wrote two paragraphs above.

    Boeing presumably have a full copy of the FDR data and already know.

  131. Andrew says:


    RE: “Flight 328, a B777-200 with P&W engines, had an engine failure soon after departing Denver…”

    It seems to be another case of fan blade separation, of which there have been several in recent times. Judging by the ATC recording, it sounds as though it was well handled by all concerned.

    RE: “No doubt the TAC in the B777 helped the pilot to control the plane.”

    Probably, but that’s not always the case. The TAC is very good with failures where there is a marked reduction of thrust on one side. However, the TAC can have trouble with failures where the thrust keeps surging, in which case it tends to drop out. The pilot then has to revert to the old-fashioned engine-out control technique. It pays to remain proficient at using the rudder!

  132. airlandseaman says:

    UA328 engine failure happened 12 mm south of my house.

  133. airlandseaman says:

    nm, not mm!

  134. David says:

    @George. From your Feb 17th, 10:39am, “Between 14:39:42.680 and 14:39:49.554 the granular data has 23 samples with “47” being recorded for “Track”. And twelve of the next fourteen samples have Track values of “46”.”

    Looking at your graph 2.2.v2 those 23 samples indicating a course of 047˚extend to past 14:39:47. Thus at that point the course had not changed, according to those data. The Report’s statement that the heading was, “046 and continuously decreasing” to me remains in conflict with that as does the like comment on your graph.

    Even so, again quite probably this will not be of consequence.

    I do agree that there was bank then. My recent embellishment to my theme, ie my assertion that wings were level, was mistaken – as you have pointed out, thanks.

    As to resolving what the thrust asymmetry was without the FDR data, looking at your new graphs of climb rate the degradation in that from the reducing left engine thrust plus the inefficiencies from yaw was pretty steady, though it did increase towards maximum altitude as bank increased. That suggests a steady decrease in thrust, overwhelming of autopilot authority indicating that it was decreasing at that point, still.
    In flight replication or gaining data from initial Boeing flight testing might not be feasible: please see my fourth last line below.

    @Andrew. Loss of control. You have addressed the likely sharp roll rate increase on A/P disengagement, were there no prior rudder application. In one of your references on asymmetric flight it says, “Attempting to control an airplane in asymmetric flight without using the rudder is potentially dangerous, since the large sideslip angle required could, on some airplane types, lead to fin stall and subsequent loss of control………… In practice, no attempt should be made to counter the asymmetric yawing moment by banking towards the live engine and leaving the rudder central or free.”

    Yet that is what the two axis autopilot does.

    If they mean at low speed flight they do not say so.

    In your other reference, about the no-rudder option it describes what might be the consequence at low airspeed and finishes, “This option is never flight-tested and not recommended to fly.” One could interpret that to mean that applies not just to low airspeed.

    With the above in mind the question arises as to whether, this situation in SJ182 was within the bounds of the foreseeable, if pilots are inattentive (eg on long haul) or get distracted, or are confused in IMC.

  135. George G says:

    Just to tidy something up:

    My Comment#30764 contained a basic error in that the word “not” was missing from a critical sentence:

    The sentence as in the comment erroneously read:
    To which I answered that you were necessarily right in the your opinion that the aircraft had not begun it’s left turn by 14:39:47, and I wrote “which I will explain later if need be”.

    The sentence SHOULD HAVE read:
    To which I answered that you were not necessarily right in the your opinion that the aircraft had not begun it’s left turn by 14:39:47, and I wrote “which I will explain later if need be”.

    And, with even more time to consider: In preparing a document that I will now have no need to send, I wrote:
    The wings of PK-CLC may have been level at the time of transition from right turn to left turn (BUT this is considered unlikely due to significant thrust imbalance causing significant yaw to the left)

  136. George G says:

    We agree that there was (still) bank. Good.

  137. Victor Iannello says:

    Jon Ostrower examined the video of the failed engine of UA328 and found frames that showed a fan bladed broken near the tip and another missing.

  138. Andrew says:

    Here’s another screen shot where the broken blade and the missing blade next to it can be seen at the top of the fan:

  139. TBill says:

    On the video, I was wondering what is burning since the the fuel is presumably off. Also was wondering if Halon would be used to put out the fire.

  140. Victor Iannello says:

    Considering that the highest rotational stresses occur at the root of the fan blade and in the disk, it looks like one blade was thrown and the adjacent blade was damaged by the thrown blade.

    Here’s a video showing the cowling parts tumbling down.

    And here’s a video showing many of the parts after landing. Notice the parts are not shattered upon impact, despite falling from 13,500 ft. In fact, based on the small amount of damage to the parts, the impacts look relatively gentle. For instance, you can see the vortex generator still attached to the cowling fragment, and the nacelle caught by the tree is relatively intact.

    Wing parts are not the same as engine cowling parts, but the lack of impact damage to the flap, flaperon, and spoiler of MH370 is consistent with in-flight separation of a tumbling part.

  141. Victor Iannello says:

    @TBill: Perhaps residual fuel that burned off after cutoff? In the video, the flame seems to be extinguishing.

    [I see others are suggesting hydraulic fluid, which might be a better explanation.]

  142. Andrew says:


    RE: ” I was wondering what is burning since the the fuel is presumably off. Also was wondering if Halon would be used to put out the fire.”

    The fire is in the area where the thrust reverser cascades are located. It might have been caused by leaking hydraulic fluid from the actuators, which ignited and set fire to the composite components of the thrust reverser. There is no fire extinguishing in that part of the engine; only the core is protected. It’s likely the damage would have rendered it useless even if it were available, because the halon would have been immediately dissipated by the airflow.

  143. Don Thompson says:


    The flames, evident in the clip posted at Twitter, appear to subside as the time progresses. I’ve been curious about that.

    The flames are escaping from the ‘mesh/grid’ that forms the cold-stream thrust reverser ports. Bypass air is directed out of these ports, not flow from the ‘hot’ section of the engine where the fuel is burned.

    The fuel spar valve shuts off supply to the engine when the fuel control switch is set to off, or the fire switch is actvated. The FSV is located on the rear wing spar. Therefore, one might suspect some residual fuel may escape from damaged plumbing that is routed around the engine and down to the main fuel pump (accessory gearbox, underside of engine) and hence to the FMU that distributes fuel to the (burner) nozzle manifold. Pressurised fuel is also used as actuation power for certain valves in the engine (I don’t have PW4000 detail).

    Extinguishant is stored in the forward hold & deployed to the engine.

  144. airlandseaman says:

    I agree with Victor. The UA328 photos and especially the videos of falling debris provide a good calibration on the descent rate of and impact damage to similar falling debris. The MH370 R inboard spoiler (8 or9), R flaperon (#1), R flap section (#19) and R aileron (#26) all may have separated in flight, hitting the water at a relative low impact speed.

    The UA328 vortex generator came down in one piece attached to a large piece of the cowling door (~9000 in^2). In contrast, the MH370 #2 vortex generator debris consisted of only ~250 in^2 of the ~9000 in^2 door and about 26″ of the 57″ long baseplate. That baseplate snapped under extreme compression in the forward direction, causing the chine to separate. The chine was completely missing. This implies (to me) that the engine was still attached to the airframe at the time of the primary airframe impact, and at a very high speed.

  145. Victor Iannello says:

    The NTSB has released an update on UA382. From that report:

    – The inlet and cowling separated from the engine
    – Two fan blades were fractured
    – One fan blade was fractured near the root
    – An adjacent fan blade was fractured about mid-span
    – A portion of one blade was imbedded in the containment ring
    – The remainder of the fan blades exhibited damage to the tips and leading edges

    So no surprises.

  146. Victor Iannello says:

    And the FAA has issued an Emergency AD to inspect the PW4000 engines on B777s.

  147. airlandseaman says:

    Here are a couple of photos of the UA328_#2_Vortex_Generator:

  148. TBill says:

    I have a pilot friend who was one of the orig United B777 pilots. I understand from him that United took the lead in flying the B777’s when they first came out. Not seeing my friend in the COVID era as our community activity is suspended, maybe I will send him email. OK yeah but he thinks like me on MH370.

  149. Andrew says:

    The following Twitter post has a few photos that show the fire damage to UA328’s engine:

  150. Victor Iannello says:

    It looks like the thrown fan blade was contained from outward projection, but the cowling and other parts flew out and struck the fuselage. Luckily, those projectiles did not penetrate into the cabin. In addition to solving the problem of the thrown fan blade, P&W needs to address the other projectiles that were generated.

  151. David says:

    @Andrew. At my comment 30768 I raised whether a two-axis autopilot having the roll authority of the 737-500 might expect too much of pilots in some circumstances.

    Bearing in mind the lack of inhibition of bank extent also, possibly this might become an issue in the SJ182 investigation as did expectations of pilots in the 737-8 Max.

    Questions raised in the investigation could be whether;
    • these possible circumstances (IMC, confusion, problem recognition and diagnosis delay) were taken into account in the aircraft design and certification,
    • thrust increases in 737 upgrades to the -500 were looked into from the A/P roll authority perspective,
    • there was in fact too much A/P authority (ie the effect at A/P disengagement without prior rudder application could be excessive).

    Then there are others beyond the scope of this investigation:
    • What reviews have there been at subsequent 737 upgrades?
    • Is the design, including the autopilot, suitable for a modern aircraft?

    In case you missed my post, the UA328 engine failure naturally having been the priority since, for any remark you may have please in due course. I will be hors-de-combat for a couple of days.

  152. George G says:

    The Excel File is still a work in progress.
    The accompanying is not simply an exercise in pedantry.
    It, or something like it, is a necessary preparation for my next step.

  153. Victor Iannello says:

    More pictures from the NTSB of UA328, including a close up of the fan blade fracture and the damage to the wing fairing.

  154. Andrew says:


    Did you mean ‘fan’ blade fracture?

  155. Victor Iannello says:

    @Andrew: Yah!

  156. TBill says:

    Re: your comments, I understand the design requirement is a “blade-off” event is supposed to be handled by the engine design without damage to wing, fuel, or fuselage systems. Not sure if that rule covers 2 blades off.

    Re: new MH370 debris
    Anything new to report? Mainly I was expecting a short write-up.

  157. Victor Iannello says:

    @TBill: The probability that two blades fracture from fatigue at the same time is very low. In this case, one fractured from fatigue and the second one was likely damaged by impact from the first. The design requirements should apply to any collateral damage caused by the thrown blade.

  158. airlandseaman says:

    The “new debris” report is in process. There will be some new info in the report, but the bottom line is that the debris is from the #8 or #9 spoiler (99%).

  159. Victor Iannello says:

    @airlandseaman: What interests me is the possibility that the spoiler separated in flight before impact, which then raises the question of whether the spoiler was raised during an overspeed condition. The possibility of a raised spoiler has other implications about the flight conditions and whether there were pilot and/or autopilot inputs.

    It will be interesting to see if you and Don believe the spoiler was raised or lowered at separation.

  160. Tim says:

    Re UA328,
    The incident is so similar to UA1175, the NTSB could just reprint the same report and just add a paragraph to explain the engine oil leaked and caught fire on the reverser cascades!

    In the UA1175 report it shows the gear box mountings fractured and the casing cracked.

  161. David says:

    @Geoge G. Measured and methodical work you are doing there George.

    An observation. Your major polynomial has the A/P authority being exceeded before 39:40 and a course rate hump before that, though to me its running out of authority would produce an instantaneous rate change, a discontinuity, so would not suit a polynomial’s smoothing.

  162. Andrew says:


    Sorry, I missed your earlier comment (30768).

    RE: ‘…”In practice, no attempt should be made to counter the asymmetric yawing moment by banking towards the live engine and leaving the rudder central or free.” Yet that is what the two axis autopilot does.’

    I suspect that’s one of the reasons autopilots have limited roll authority. Another reason is to make it easier for the pilot to recover if the autopilot suffers a hard-over failure. Two-axis autopilots have been the ‘gold standard’ in aircraft design for many years, although large aircraft use 3-axis control for runway alignment during autoland operations. The two-axis design obviously requires the pilot to make the correct rudder input where necessary.

    RE: ‘…the question arises as to whether, this situation in SJ182 was within the bounds of the foreseeable, if pilots are inattentive (eg on long haul) or get distracted, or are confused in IMC.

    There have been several accidents where asymmetric thrust under autopilot control was a factor, including China Airlines 006 over the Pacific in 1985, China Southern 3943 at Dalian in 1992, and Tarom 371 at Bucharest in 1995. Of those, China Airlines 006 is perhaps the best known and best documented. The NTSB accident report is available here.

    The certification rules that are applicable today do foresee circumstances where pilots might not recognise a sustained autopilot control input or an asymmetric thrust condition. However, the emphasis is on alerting to ensure the pilot is aware of the condition, and can provide manual inputs where necessary. Paragraph 45.d.(3)(a) of AC 25.1329-1C (CHG 1) – Approval of Flight Guidance Systems states:

    “If the autopilot is holding a sustained lateral control command, it could be indicative of an unusual operating condition for which the autopilot is compensating. Examples of such unusual operating conditions are asymmetric lift and/or drag due to asymmetric icing, fuel imbalance, or asymmetric thrust. In the worst case, the autopilot may be operating at or near its full authority in one direction. If the autopilot were to disengage while holding this lateral trim, the result would be that the airplane could undergo a rolling moment that could possibly take the pilot by surprise. Therefore, a timely alert should be considered to permit the crew to manually disengage the autopilot and take control prior to any automatic disengagement that might result from the condition.”

    Paragraph 88.d. of the same document states:

    “During autothrust operation, it is possible that a failure (for example, engine failure, throttle lever jam, or thrust control cable jam) could result in a significant asymmetric thrust failure condition that may be aggravated by the continued use of the autothrust system. Because the FGS could potentially compensate for the asymmetric condition with roll (and possibly yaw) control, the pilot may not immediately be aware of the developing situation. Therefore, an alert should be considered as a means of mitigation to draw the pilot’s attention to an asymmetric thrust condition during FGS operation.”

    The B737 “Classic” (ie -300/-400/-500) was certified over 30 years ago and much of that was grandfathered from the original B737 type certificate that was granted in 1967. The alerting requirements were not incorporated in AC 25.1329-1 until 2006, when AC 25.1329-1B was issued. Consequently, they were not applicable when the B737-500 was certified.

  163. Don Thompson says:


    The fire, evident in the t/r cascades, is a concern. Or fires considering the ‘cold-stream’ fan duct is split left-right

    I geolocated the out-of-the-window view during the Twitter/@michaelagiulia video clip. During the clip, N777UA is near directly south of Bennett, CO. Correlating with FR24, that position places N777UA approx 10mins into the event.

    The engine lube oil tank remained attached to the fan case. The engine oil capacity is approx 20l, not a large quantity. However, even the short period necessary to assess the event, before pulling the engine cut off lever, may have resulted in significant fuel leakage should a fuel line have been fractured (per G-VIIO, where leakage was estimated at 365l of fuel in 30sec). The hydraulic EDP is not isolated until engine fire suppression is activated, a broken hydraulic feed line presents another accelerant source.

    The t/r cascades on both sides exhibit fire damage. A misting/spray leak of any engine related flammable liquid may have been the initiator.

  164. George G says:


    SJ182, re our previous comments.

    I have been persuaded that:
    My agreement to the use of the word “discontinuity” was misguided.
    Naturally, there is no discontinuity in actual turnrate.
    A more appropriate description may have been to say that the data indicates there is a “distinct change in turnrate” at about 14:39:40.
    This is in contrast to relatively smooth progressive change in direction during much of the rest of the flight, once the data scatter has been smoothed out.

  165. George G says:

    @Don T

    Is there an emergency valve in the hydraulic pressure supply line to the thrust reversers ?

    If not, then … as Don discusses.

  166. Don Thompson says:

    @George G

    The T/R Isolation Valve assembly includes, among other things, an Isolation Shut Off Valve. The ISOV is open only when the T/Rs are in use.

    I’m now confident that the initial event didn’t include fire. That 9-10 minutes after the fan blade separation a hydraulic line in the EDP supply or pressure side then fractured as a consequence of the continous vibration thus initiating the fire.

    During the 32s roll of the video clip the fire appears to subside. The Non-Normal Checklist for Engine Fire instructs that the fire handle should be operated, if the FIRE ENG message persists after 30 seconds, the fire handle should then be turned to its opposite stop so as to discharge the second bottle. The video clip captures the fire as it established and as it was extinguished.

  167. George G says:

    @Don Thompson

  168. Andrew says:

    @Don Thompson
    @George G

    I doubt the fire subsided as a result of the fire extinguishing system. The fire detection loops monitor the ‘hot’ section of the engine inside the nacelle and I doubt they would have detected a fire in the cold section around the T/R cascades. Similarly, the fire extinguisher bottles discharge halon into the nacelle around the hot section and would likely have had no effect on a fire burning in the cold section.

    I think it’s more likely the crew initiated the Engine Severe Damage checklist, which also requires the Engine Fire Switch to be pulled, but without activating the extinguishers. The crew might have subsequently operated the fire extinguishers in response to reports of fire from the cabin, but as I said above, it’s likely they would not have had any effect on the fire.

    A former Boeing propulsion engineer suggested on PPRuNe that the fire might have been caused by the ignition of leaking fluid of some description (fuel, oil, hydraulic – take your pick!) that caused resin within the composite components of the T/R cascades to ignite. The fire then self-extinguished once all the residual fluid and resin had burnt away.

  169. George G says:

    @Don Thompson,

    Just as a point of clarification/query.
    I initially thought the fire level was decaying due to lessening of the amount of flame holder, i.e. the normally cold duct components having somewhat already burnt off.
    Reviewing the video, I still think that.
    Do you think the discharge of the second fire bottle would really have much effect when the was fire was so exposed ?

  170. George G says:

    Informative reply. Didn’t refresh before I sent last comment.

  171. Don Thompson says:

    @Andrew – thank you for those insights. The ‘Eng Svr Damage/Sep’ checklist, including pull/lift of the fire handle, would be more appropriate than ‘Eng Lim/Surge/Stall’.

    @George G – having reviewed more images captured at earlier time points during the incident, a smaller outbreak of flames was evident. I agree that release of engine extinguishant would not likely have an effect on the flames, considering the location where they initially appear or where flames spread.

    Following the fan blade off, the T/R Sleeves were initially intact. ‘Hayden Smith, Instagram:@speedbird5280’ captured a number of images from the ground in which a small outbreak of flame is evident on the nacelle ‘keel’, the low point of the TR assembly. Later images from the cabin show the sleeves intact.

    That suggests that some fire accelerant (fuel, hyd, or lube oil) had pooled and ignited in the lower TR assembly, specifically the void between the inner and outer sleeve of the T/R sleeve (illustration).

    As Andrew describes the resin used to impregnate the CF in forming the CFRP skin of the structures is flammable but the nomex honeycomb core is a fire retardant material. The post event views of through the fan duct show that the inner face of the inner sleeve exhibits little, if any, fire damage.

    It’s possible that Twitter:@MichaelaGiulia captured the video clip soon after the outer sleeve had finally weakened and detached to expose what lies within the TR sleeve, i.e. the cascades.

    Somewhere under the flight path, west of the RD129-RD30 junction south of Bennett, the inboard TR sleeve may be found, charred on its inner face, less so on the outer face. Following the loss of those outer skins the flames likely burnt out.

    I was initially drawn to this by the possibility to locate the out-of-the-window view in Twitter:@MichaelaGiulia’s video clip, which I achieved, and have been working back from that point.

  172. Barry Carlson says:

    @Don Thompson,

    The following is a HiRes 4032 x 3024 px JPG image of a LUCAS Control Unit, Boeing Part Number S315W363-2 photographed on the ground in the Broomfield area.

    Appears that it and associated lines were cut out by departing blade.

  173. Don Thompson says:

    @Barry Carlson

    I understand that the local police have been acting as collection agents.

    That part is the control switch unit for thrust reverser PDOS (powered door opening system), the ‘down’ button remains, the ‘up’ button gone. It would normally be found on the aft face of the inlet cowl.

  174. David says:

    @Andrew. Returning to SJ182 loss of control. Thank you for your comprehensive comments.

    “The alerting requirements were not incorporated in AC 25.1329-1 until 2006, when AC 25.1329-1B was issued. Consequently, they were not applicable when the B737-500 was certified.”

    I note that the AC’s provide guidance and are not mandatory, as your quotes convey (“…should be considered”).

    §25.1329(j) requires, “Following disengagement of the autopilot, a warning (visual and auditory) must be provided to each pilot and be timely and distinct from all other cockpit warnings.” I note the ‘and’.

    From the below, not unexpectedly, that also was adopted in 2006:

    I see that not all think an auditory alert is worthwhile, some research being in the ‘inattentional deafness’ report mentioned.

    But also I see that §25.1309 could be construed as relevant. 1309(c) is apparently his source.

    Depending on what the inquiry finds as the likely chain of events that led to this accident, it might be that it raises the question as to whether it would be feasible to fit, retrospectively, visual and auditory alerts in the legacy 737’s, at least.

    However presumably that would be impractical, since if it was found necessary for safety reasons to make that change for aircraft designs certified post 2006, even the date of .1309, why otherwise would that not have been retrospective (assuming retrospectivity consideration is a normal requirement)?

    Also though there is another §25.1329 requirement at (d) and (e) of which I assume (e) to be the more likely to be relevant, “Under rare normal and non-normal conditions, disengagement of any automatic control function of a flight guidance system may not result in a transient any greater than a significant transient, as defined in paragraph (n)(2) of this section.”

    That says, “(2) A significant transient may lead to a significant reduction in safety margins, an increase in flightcrew workload, discomfort to the flightcrew, or physical distress to the passengers or cabin crew, possibly including non-fatal injuries. Significant transients do not require, in order to remain within or recover to the normal flight envelope, any of the following:
    …….(iii) Accelerations or attitudes in the airplane that might result in further hazard to secured or non-secured occupants.

    Mind, I do not find that easy to interpret!

    I wonder if that is relevant here and if so when that was included in §25.1329?

    If relevant that leaves a selection of §25.1329(e) and (j), and .1309(c).

    Possibly I delve too deeply at this stage but am mindful of the 737-8 Max outcome.

  175. David says:

    @George G. Thanks for 4Dd and the explanation under.
    From your 4Da I see the FDR to FR24 difference to highest altitude is a bit over a second, consistent with that earlier Report’s FDR timing of the left turn.

  176. George G says:


    For the record, the time for pilot intervention preceded the Autopilot dis-engagement by (say) half a minute, is my guess.

    Unless at least one pilot was proficient in unusual manoeuvring and with aerobatic like responses, when the autopilot did dis-engage and the aircraft subsequently rolled towards inverted, or very close to that, then that was it.

    Now, if there had been a dog in the cockpit trained to monitor for unusual SUSTAINED conditions whilst under autopilot control, and trained to BARK three times to alert the pilots.

    A well trained dog could have then yelled: “LOOK ! .. THRUST !”
    or “LOOK ! … ADI !”

    In this “Modern Day” a mini non-executive Electronic Flight Engineer could have done the same thing, so long as it was fed input from the existing computers.

    Probably cheaper than dog food.

  177. David says:

    @George G. I have inferred from that there was no alert system though to be fair, conceivably that had been retro-fitted and was ineffective in this instance.

    Yes ‘they’ did away with the flight engineer and gave pilots that job.

    Then we have aircraft flying that would not meet current airworthiness requirements yet quite possibly with pilot training standards based on the modern.

    I suppose motor vehicles are similar. All the same I would prefer to drive my family in a modern car (crash worthiness and even (come in @DennisW) anti-skid braking.

    Likewise as a passenger I prefer my aircraft modern.

    Maybe EICAS etc aim to be the Electronic Flight Engineer.

  178. Victor Iannello says:

    @David: Has the A320 had a better safety record than the B737?

  179. vodkaferret says:


    “@David: Has the A320 had a better safety record than the B737?”

    Erm… Boeings’ own document would seem to suggest so:

  180. Victor Iannello says:

    @vodkaferret: That’s an interesting reference. Looking at the statistics on page 13, which includes data through 2019, the hull loss accident rates per million departures are an identical 0.18 for the 737-600/700/800/900 versus the A320/321/318/319. On the other hand, the rate for the older 737-300/400/500 is 0.76, and the rate for the even older 737-100/200 is 1.8. So, the differentiation is not really B737 versus A320, but rather old versus new, as David asserted.

  181. vodkaferret says:


    Maybe I misunderstood your point – I thought you were implying that 737 was at least as safe / safer than an A320. And your figures are correct.. however the figures for the 320/321/318/319 should probably be more appropriately compared to the 737 3/4/500 series, not the later -600 and up… in which case the 0.18 vs .076 difference seems significant.

    Let’s not even get into neo vs max 😉

    I certainly agree that newer the aeroplane the safer (on average) it is, and if that was the point you were making I’m more than happy to concede it 🙂

  182. vodkaferret says:

    as a side note I have no idea where those smileys appeared from on my last reply as I certainly didn’t type them!!

  183. Cui Shineng says:


    Distinguished gentlemans, I want to know that have any conclusions about whether the new debris belong to MH370, or not?

  184. Victor Iannello says:

    @vodkaferret: I wasn’t trying to make a point because I was not aware of the data. I had read in less technical articles that the B737 and A320 had similar safety records. I now better understand the context of that statement. The B737NG series (600-900) was introduced to compete with the A320, and the safety records are almost identical.

  185. vodkaferret says:


    As I said, I think I misunderstood your original point

    As the data in that Boeing paper shows : the worst of the A320/1/etc series has a hull loss rate comparable with the best of the 737 series. So it would be hard to argue (using Boeings’ own report) that the two airframe have similar safety levels.

    But I’m really not trying to get into a fight. If you are saying newer airframe are safer than old, I wholeheartedly agree. If you are saying both a320 and b737 are incredibly safe – I wholeheartedly agree. And hull loss rates – while in some ways they are the ultimate measure of safety, in other ways the small sample size iss a killer. So we both know the limitations of that data.

    Only if you argue that the 737 has a better safety record than the a320 will we see the statistics differently. and even if thats thats the case, we can agree to differ!

  186. TBill says:

    @Cui Shineng
    I am not the expert, but above ALSM says almost definitely yes it is part of MH370. Not only is it a part of MH370, it is potentially one of the more important parts as far as determining nature of end of flight scenario.

  187. airlandseaman says:

    Cui Shineng: The new debris is either the #8 or #9 spoiler (one of 2 inboard spoilers on the right wing) from a B777-200. Almost certainly from 9M-MRO/MH370.

  188. vodkaferret says:

    or to put it another way, in answer to your question:

    “@David: Has the A320 had a better safety record than the B737?”

    The answer is yes, it does.

  189. Cui Shineng says:


    Expect the news report.

  190. Andrew says:


    RE: I note that the AC’s provide guidance and are not mandatory, as your quotes convey (“…should be considered”).”

    Yes, the guidance in the AC provides an acceptable means for demonstrating compliance with the regulations, but it is not mandatory. The introduction states:

    “The guidance in this AC is neither mandatory nor regulatory in nature and does not constitute a regulation. It describes acceptable means, but not the only means, for demonstrating compliance with the applicable regulations. We will consider other methods of showing compliance that an applicant may elect to present. While these guidelines are not mandatory, they are derived from extensive FAA and industry experience in determining compliance with the relevant regulations. If, however, we become aware of circumstances that convince us that following this AC would not result in compliance with the applicable regulations, we will not be bound by the terms of this AC, and we may require additional substantiation or design changes as a basis for finding compliance.”

    RE: “I see that not all think an auditory alert is worthwhile, some research being in the ‘inattentional deafness’ report mentioned.”

    Aural alerts have been mandated for some time, but research (and experience) has shown that acute stress tends to cause human hearing to become dulled or even shut down as part of the ‘flight or fight’ response. That process is known as auditory exclusion, or tunnel hearing (see: Understanding Stress – Part 6: Auditory Exclusion). Consequently, pilots can miss aural cues in highly stressful situations. It’s also possible that aural cues might be drowned out in the cacophony of alerts that occur in more complex failure scenarios. These phenomena are thought to have been factors in the AF447 accident, when the pilots failed to respond to the aural stall warning. The regulators have responded by mandating visual alerts in addition to aural alerts. On newer Boeing and Airbus aircraft, visual alerts have been implemented by red warnings on the pilots’ PFDs (eg ‘STALL STALL’, ‘WINDSHEAR’, ‘AP OFF’, ‘ENGINE FAIL’, ‘PULL UP’), where, hopefully, they won’t be missed.

    RE: “Depending on what the inquiry finds as the likely chain of events that led to this accident, it might be that it raises the question as to whether it would be feasible to fit, retrospectively, visual and auditory alerts in the legacy 737’s, at least.”

    Even if it were possible, I think the regulators and airlines would push back against such a recommendation, given the likely cost. I suspect they would focus on improvements to training as a way of mitigating the deficiency.

  191. David says:

    @Andrew. Thanks.

    “I suspect they would focus on improvements to training as a way of mitigating the deficiency.”

    That does seem to be the flavour of the Preliminary Report.

  192. David says:

    For those interested in the cause of the UA 328 fan blade failure the below gives a clue at question 7.
    Also some good shots of debris etc.elsewhere

  193. CanisMagnusRufus says:

    …. very quiet here…

  194. Victor Iannello says:

    While Bobby Ulich continues his work on the statistical analysis of the drift data, I spent a small amount of time and looked more closely at Drifter 101703, and compared that to the modeled results for paths ending near the time and position of the recovered flaperon. I assumed that the reporting delay for the flaperon was short.

    You might recall, Drifter 101703 was already undrogued as it passed near the 7th arc on March 8, 2014, at (-23.416,103.006). It last reported its position on July 23, 2015, near Saint Brandon at (-16.723,59.861), which is about 642 km to the northeast from where the flaperon was found on Reunion at (-20.916,55.649) on July 29, 2015. I asked myself the question: If we only knew the drifter’s last position, could we use CSIRO’s drift model results to estimate the position along the 7th arc on March 8, 2014? We can then compare this with the actual known position.

    I first identified which of the 86,400 trials passed near the drifter’s last reported time and position within a window of 16 km and 16 days. There were 70 trials that met the window criteria, and represent a range of starting positions between 10S and 40S, using the drift model results for the generic trials, for which a windage of 1.2% is applied as a proxy for Stokes drift. Of these 70 trials, the trial that starts closest to the actual position of the drifter was Trial 12682, which started less than 10 km from the drifter’s position on March 8, 2014. So we have a trial that starts and ends very close to the actual drifter. I would declare this a modeling success.

    The top pane of this figure shows these results. The drifter (yellow) meanders around the 7th arc before it is transported west across the Indian Ocean. Trial 12682 (green) shows similar behavior to the actual drifter. Of course, the start and end positions are almost identical due to the selection criteria.

    The bottom pane of the same figure shows analogous results for the flaperon, except CSIRO’s drift results are computed using the measured hydrodynamic parameters from CSIRO’s flaperon experiments, and of course we don’t know the location of the flaperon on March 8, 2014. There are 51 trials that satisfy the time and position window criteria. The start positions of these 51 trials range from around 41S to 23S, with only a single trial north of 31S, and a cluster around 34S. These results suggest a northern impact is unlikely, although a more complete statistical analysis would require more trials (with a larger time/distance window) similar to what Bobby is now calculating for multiple recovered debris.

    I’ve asked Bobby to use his methodology to predict the position of the drifter on March 8, 2014, assuming we only knew the position on July 23, 2015. Based on the results shown in the top pane of the figure, I doubt that there will be a high degree of latitude discrimination, but certainly there will be latitudes that can be rejected. It will be interesting to compare these results with the known position on March 8, 2014. It should be an unbiased test of the methodology.

  195. David says:

    @Victor. Of 70 trial generics that arrived at time and place, from a wide range of latitudes but all from the 7th arc, I think it unsurprising that one should be from near where the undrogued drifter had been.

    While their drifts west were similar, though the tracks converge about 9˚, before that they were quite different.

    I get the impression that the trial’s track was longer and if so its speed was higher.

    That would be a shame since had the two followed near-identical tracks that would have supported the premise that the undrogued drifters’ speed was current plus 1.2% of wind. As it is, if the trial’s track is indeed longer, we do not know whether local eddies or Stoke’s drift, or both, would cause that. Similarly if of the same length but separated, a different Stoke’s drift could balance different current.

    About the flaperon modelling, I suggest we can have less confidence that the course and speed assumed were accurate. I have belaboured this before but if useful here I can repeat that.

  196. 370Location says:

    @Victor – Your top diagram and methodology is fine, because we know the exact time and location of 101703 before it went quiet. Understandably you added a window of 16 km and 16 days to acquire the precomputed model paths.

    The second chart does not apply, because you applied the same 16 day window of arrival at Reunion, but we don’t know when the debris arrived. There were reports of seeing similar debris much earlier at Reunion. We know that debris from farther north arrives earlier, because it gets into the Southern Equatorial Current without wandering around first. Saying, “These results suggest a northern impact is unlikely…”, or that with a similar approach, “certainly there will be latitudes that can be rejected.” shows a selection bias for your candidate area. Your group’s previous reports have penalized or excluded early arrivals in other ways.

    If the recent Port Elizabeth debris find turns out to match MH370, it will be quite clear that debris can be found even years after it was expected to arrive.

  197. Victor Iannello says:

    @370Location: I clearly stated the assumption of no report delay for the flaperon. The presence of barnacles is consistent with little time on the beach. Sure, it’s possible the flaperon beached and was carried back out to sea several times before it was found. For this analysis, I assumed the simplest of assumptions, which is the flaperon first beached when it was found, consistent with the presence of barnacles. That’s not applying selection bias.

    So what do you believe is the upper limit for the delay between when the flaperon was reported and when it first beached?

    And nobody has said that no debris will arrive later than a certain date. The problem with analyzing late arriving debris is that you lose all latitude discrimination for the impact location.

  198. Victor Iannello says:

    @David: We don’t know that undrogued drifters are transported with essentially no windage but with Stokes Drift, although I don’t know why an undrogued drifter would behave differently than other generic debris. We don’t know whether the actual flaperon behaved like the flaperon mock-ups that were used to experimentally determine the hydrodynamic parameters. We don’t know how well CSIRO’s BRAN2015 model simulates the historical current and wind. Also, the number of trials falling within the time/distance window was too low to apply meaningful statistics.

    The purpose of this exercise was only to visualize the trend of starting points along the arc that are consistent with the position of the drifter and the flaperon in July 2015. On the other hand, Bobby is working on more rigorous methods.

  199. Victor Iannello says:

    @370Location said: Saying, “These results suggest a northern impact is unlikely…”, or that with a similar approach, “certainly there will be latitudes that can be rejected.” shows a selection bias for your candidate area. Your group’s previous reports have penalized or excluded early arrivals in other ways.

    To eliminate any concerns about selection bias introduced by assuming no reporting delay (which I maintain is consistent with the presence of barnacles), I ran two other flaperon cases, shown in this figure. For the first case, a reporting delay of 60 days is assumed. The number of trials falling in the 16-km/16-day window is reduced to 15, and there are no trials starting north of 24S latitude. For the second case, I used only a 16-km window with no time constraints. The number of trials is now 379, with only a single trial north of 24S latitude at around 20S latitude.

    This should eliminate any concerns that I have selected a time window that is biased against an impact at northern latitudes. The possibility of an impact at northern latitudes is only allowed if the drift model is not representative of the historical winds, currents, and the hydrodynamic behavior of the flaperon.

  200. paul smithson says:


    Thanks for that little write up with your usual clarity and conciseness. This is exactly the same conclusion that David Griffiths reached – that the debris finds in the Western Indian ocean tell you that the debris came from somewhere between 20S and 40S, with things getting increasingly unlikely at beyond each end. And this is why I remain convinced that the debris modelling is a very poor discriminator of the likely latitude of origin if WIO debris finds is the analytic endpoint.

    That is before you also consider uncertainties embedded in the modelling – both the ocean surface current modelling (which I believe is very good) and the overlay of leeway/stokes drift vector (which I think is open to significant uncertainty). These sensitivities presumably blur discriminatory power even further.

    Modelled latitude discrimination appears to improve if we take non-appearance of debris in W Australia as our outcome variable. However, to accept that you also need to believe:
    a) that the leeway/stokes drift vector estimates in the model are good
    b) that modelled convergence with the coast results in beaching (ignoring tides or how “sticky” the coastline is as an accumulator of flotsam)
    c) that stuff arriving on the coast would have been found and reported [including the majority length of coastline not covered by regular clean-ups]

    For all of the reasons above, I am deeply sceptical of using drift-models as an indicator of likely origin, whether used on its own or in combination with other lines of evidence. I suppose it is better than nothing, but I would attach a very low weighting to the contribution of drift models to predicted origin/origin range.

  201. paul smithson says:

    In relation to (b) above
    “Stranding is determined either by comparing with the current grid or by reading a file containing a coastline contour based on the Global Self-consistent Hierarchical High-resolution Shoreline database (GSHHS). The latter method is highly recommended (and the default) as it allows much more precise stranding of particles.”

  202. Victor Iannello says:

    @Paul Smithson: In the CSIRO model, I believe the stranding of trials includes interactions of the virtual particles with an accurate shoreline database. For some trials, after beaching, the particle re-enters the ocean and is further transported. For these cases, we arrest the particle from further transport after the first beaching, as evidenced by a small incremental displacement over several days.

    I think it is more straightforward to use the drift model to eliminate potential impact points than to discriminate or prioritize among the possible impact points. But this is exactly what Bobby is attempting to do. Bobby has spent much of his career in teasing out the signal from low signal-to-noise data sets, and I remain hopeful that we learn something from his work.

  203. David says:

    @Paul Smithson. “For all of the reasons above, I am deeply sceptical of using drift-models as an indicator of likely origin, whether used on its own or in combination with other lines of evidence.”

    Agree with “on its own” though I think including “in combination” may be a little harsh.

    Regarding “on its own”, the CSIRO’s approach has been to characterise its findings as ‘consistent with’; a broad brush. However that was refined to specifics by the First Principles Review of November 2016, in identifying future search prospects. The drift work pertinent was described later in the CSIRO’s, The Search for MH370 and Ocean Drift paper dated December 2016. It included reliance on the flaperon modelling that was based on field work with the replica; as evinced by the Review’s page 21, where it describes the influence of that flaperon drift assessment on the SATCOM Probability Density Function.

    Now about “in combination” though, that paper included also a comparison of the areas of the aerial search with the CSIRO’s modelling of where wreckage should have been at that search timing. That allowed for the crash to have been up to 45 NM from the 7th arc and utilised updated drift modelling when compared to that for the actual search.

    Because of the relatively short time lapse, errors in drift direction that might exercise tolerances in long range drifts should be much the less, in principle anyway. The outcome, at the above paper’s page 23 was, “Our analysis (Fig. 4.1) shows that the surface search strongly contra-indicates regions north of 32°S as being the location of the accident. It also mildly contra-indicates the region between 34°S and 32°S. The region between 34°S and 38°S emerges as the area where debris from an aircraft impact was highly likely to have escaped detection.” I suggest that that work might carry more weight than the analysis of drift to great distances, particularly of the flaperon.

    Surprising though that included no assessment as to the likelihood that the wreckage would have been spotted if there, though their later Part IV paper went into that. Using one area as an example, where its analysis indicated that the prospects of spotting were assessed as high, it concluded that that positive conclusion could apply to the rest.

    Of course its west Indian Ocean work had been updated in April 2017 (by its Part II) by the outcome of its field work on the genuine flaperon, that confirming the broad-brush earlier ‘consistent with’ conclusion while attributing a higher confidence to that. The modelling included two flaperon course offsets in view of field results’ scatter and to test for sensitivity.

    As you know, all that has been supplemented since to include most of the wreckage which now has been recovered, update to the CSIRO’s BRAN2015 model and now the BRAN2020. Even so, David Griffin of the CSIRO has iterated after these refinements that the outcome should not be relied on except in the broad.

    Other long range drift studies have contributed too though, so far as I know, they offer no more precision.

    Thanks for the reference on stranding, a key aspect. So many unknowns and, I suspect, generalisations, plus the local knowledge required. Here may be the latest version of the shore line/coastline data base. Because of size, or I cannot, I have not accessed its parts.
    The related links are relatively accessible I notice.

    @Victor. “We don’t know that undrogued drifters are transported with essentially no windage but with Stokes Drift,….” There is the CSIRO’s discussion of that at para 2.4 of their ocean search drift report of December 2016, but as per the above this does not allow for their bio-fouling, and most of the CSIRO experiments were with GDP buoy replicas. Also wind measurements were at about 3 metres above sea level vice the NOAA’s 10.

    “…although I don’t know why an undrogued drifter would behave differently than other generic debris.” The offshore generic experiments were with replicas of a flap fairing and ‘Roy’, neither with biofouling (though as first discovered beached, ‘Roy’ was heavily fouled). As it was, ‘Roy’ did not drift close to the undrogued drifters at all. The flap fairing ended up close but had diverged on the way.

    Besides the outer right flap part that ended in Tanzania is likely to have drifted like the flaperon, off-downwind but quite possibly with less Stoke’s drift so less speed than that but more than the generic’s. Probably these add to not much difference under “consistent with” tolerances but could affect the more precise.

    “Bobby is working on more rigorous methods.” Excellent.

  204. Arto L says:


    It could be a good idea to re-examine also the probability of debris NOT ending up in Australia. CSIRO already did that in 2018, and the results were sort of conflicting with other data. CSIRO’s analysis showed that if the crash site was south of latitude 33S, there should have been significant beaching of debris in Australia, where none was found. However, at the same time it was estimated that the probability of non-detection of debris during aerial search for latitudes 26S-33S is near zero.

  205. 370Location says:

    @VictorI: Thank you for running the additional plots. You asked what the upper limit should be for the delay before discovery. Even 60 days is not enough. Oleksandr Nesterov ran a very nice simulation from an 8.36S origin based on his research:

    It shows the debris field heading into the SEC and directly to Madagascar and Mauritius/Reunion, arriving from late Aug to early Sep 2014. That’s 10 months earlier than debris was found. Your time window selects for debris that meanders early on, then goes direct to the destination. A northern origin does the opposite, being carried directly to the area of the finds, then meandering there. One big difference is that no paths from the northern arc go to WA, while paths S of 30S clearly start out east toward WA beaches.

    It’s surprising to see none of your simulated paths N of 24S but one at 20S make it to within 16 km of the Reunion find (though the island itself spans 75 km). The first GEOMAR drift report on the flaperon pointed right at Java. Their second one with more detailed Stokes drift focused on 20S:

    The 2019 GEOMAR report looked at variations in the Stokes drift component, finding that the correct component was uncertain, with values for origins along the arc from 50% (Java) to 150% (40S). Not surprisingly, they chose a value of 100% which matched the area around 30S

    I recently posted an animation of the same CSIRO drift paths that you’re using. It seems clear that almost any origin along the arc will eventually have debris appear where it was found. It may not be possible to discriminate or reject origins, except those farthest south.

    You are using barnacles to determine the date of arrival, assuming it wasn’t caught in a gyre before beaching. The temperature range for barnacle attachment may be an important clue, as detailed in my report above. The De Deckker report (FIR appendix F) report on barnacle growth indicated that they attached at the warm end of the reproductive range, progressed to colder water in the austral summer, and then ended at the expected temperature for Reunion. The conclusion says that for a crash site west of Australia, growth should have commenced immediately near the cold end of the range. Many investigators have considered that report inconclusive, because it fits with a tropical crash site, not cold waters. Some pieces of debris found on the African coast had no attached gooseneck barnacles, but were covered in snails and other fouling. Those items may have never left tropical waters. While not good for discriminating a particular latitude, it’s a strong indicator for N vs S on the arc. I’m not claiming it rules out a southern location, as you appear to be doing with northern arc origins and drift analysis. I’m asking you to not rule out other evidence for a viable site near Java based on a narrow analysis of single drift model using in this case only one piece of debris.

  206. TBill says:

    Re: the CSIRO results (above 33S), that sounds consistent with Prof Chari who likes 32S with allowance for some distance 65 nm from Arc7.

    Re: Lack of Confirmed Visual Debris (26-33S)- that’s a hard one – some debris was seen but not confirmed. Also depends on distance from Arc7 searched and nature of crash.

  207. Victor Iannello says:

    @370Location: You claimed that I chose a time window based on selection bias favoring 34S. I showed that even with no time window, the CSIRO drift model for the flaperon shows that a northern impact is still unlikely. You point to Oleksander’s model, which shows debris arriving at Reunion about 10 months before the flaperon was found. (Oleksander’s model includes particles with a range of assumed leeway.) Are you really suggesting that the flaperon might have first beached 10 months before it was reported in July 2015?

  208. paul smithson says:

    “Agree with “on its own” though I think including “in combination” may be a little harsh.” Perhaps so – hence my 10% weighting comment at the end. However, I do think it is important to note that we could be significantly misled by drift model inferences if the parameters are seriously off. And the leeway factor + directional offset to true wind could easily be seriously off – as the flaperon results version 1 and version 2 showed. I would be more comfortable with including the drift-inferred likely origin if it included a sensitivity analysis of variation in key input parameters. That would clearly make a fuzzy prediction fuzzier still. Which is why using it to rule out search areas is a risky thing to do. In fact, I regard the use of the “drift evidence” to rule out 40S at the first principles review to have been an unforgivable misinterpretation of evidence.

    See fig 11, p20 of

    “Figure 11 shows the result of the SATCOM PDF updated with the search results. The areas searched were removed from the PDF. The residual probability is located in areas yet to be searched, with two clear areas of interest: north of the current search area and south of the current search area.”

    Between these, the bottom end was clearly the better prospect on basis of fit to the sat evidence (DSTG Figure 10.3, sans BFO). Yet First Principles went on (Fig 13) to conclude that “The section south of the indicative search area ….becomes much less favourable once the drift results are incorporated.”

    In other words, they used drift models to exclude searching the southern end. And the rest is history.

  209. Victor Iannello says:

    Arto said: CSIRO’s analysis showed that if the crash site was south of latitude 33S, there should have been significant beaching of debris in Australia, where none was found.

    Interestingly, Chari Pattiaratchi believes that offshore currents preclude any debris from reaching the shores of Western Australia, and therefore the absence of debris should not be used to discriminate latitude. He bases this on his study of drifters. David Griffin disagrees.

  210. George Tilton says:


    I would side with Dr. Pattiaratchi.

    I plotted the ATSB search area on a 1943 US Military Map of SIO currents and there appears to be a very low chance of debris reaching Western Australia.

  211. Andrew says:

    The NTSB has released an investigation update for UA328.

    The update states:
    “FDR data indicate the engine made an uncommanded shutdown and the engine fire warning activated shortly after. The flight crew declared an emergency with ATC and stated their intention to return to DEN for an emergency landing. The flight crew began to complete checklists, including the engine fire checklist. As part of the checklist, the flight crew discharged both fire bottles into the engine, but the engine fire warning did not extinguish until the airplane was on an extended downwind for landing.”

    It seems I was wrong about the fire indication, but even so, I’m not surprised the extinguishers didn’t quell the fire.

  212. Andrew says:

    Boeing AERO:
    Fire Protection: Engines and Auxiliary Power Units

    If the PW4000 fire detection system is similar to that of the typical engine in the article, I assume the fire was detected by the loops in the lower bifurcation:
    Figure 1: Compartmentation in a typical engine pod
    Figure4: Typical engine fire and overheat detector locations

  213. David says:

    @Paul Smithson. Looking at the diagrams and argument in CSIRO’s 8th December 2016 paper, which I presume mirrors the drift information put to the Review, these generally support its selection of the northern PDF search area. (Fig 4.1, right, from 38˚S to below 39˚and-a-bit, estimates a 18-20% chance of flotsam being missed in the aerial searches there but the drift outcome renders that academic).

    The pretty scattered data and shortage of supporting analyses surely reflect the timeframe available. The major offshore data-gathering field trial was concluded on 25th October, 2016, just a little over a month before the 2-4 November Review – in which to model, interpret, prepare a presentation, deliver at least a summary to meeting participants beforehand and travel.

    Whether the Review’s decision to search north was in the circumstances “unforgivable” would depend on the extent to which its participants felt free to seek supplementary work. I think it quite likely that the real-world urgency for the meeting’s outcome would have inhibited that.

    As to whether, now, with more information on the flaperon and further flotsam recoveries and after more analysis there should be more sensitivity analysis, well even if so, two oceanographers who have conducted informed studies are still not supporting a search south, even though the merits of searching an unsearched area are obvious.

    Also though, another piece of new information has been the finding that there would be insufficient fuel to reach the southern unsearched zone anyway. While that prevails it renders as hypothetical that question of whether drift excludes the southern area.

    For my part I would have liked to have seen more extensive field testing to be able to refine crash latitude (and longitude). However Dr David Griffin has said that while he would have liked to conduct more field work, there has been no tasking.

  214. 370Location says:

    @VictorI asks: “Are you really suggesting that the flaperon might have first beached 10 months before it was reported in July 2015?”

    Not at all. In the next sentences I pointed out that unlike southern arc origins, a site nearer the Southern Equatorial Current gets taken directly toward Madagascar, and has time to meander there (before beaching). I further on suggested that the flaperon may have been caught in a gyre.

    My best estimate for the path of the flaperon from Java based on barnacle growth is that the winter waters around Madagascar/Reunion/Mauritius got cold enough for larval attachment just as the debris arrived there. It may have meandered farther south or even back into warmer summer waters before beaching, as the attached barnacles can continue to grow over a broader range of temperatures.

    I’d like to point out that I have not done my own drift modeling for a practical reason beyond the difficulty of getting it right. I’m basing my candidate site on additional new acoustic evidence. It doesn’t matter that there are higher probability drift origins, because I have a very specific location found by seismic epicenter.

    (My current thinking on the seismic source is debris hitting the seafloor. Autocorrelations with the strongest 1:15:18 impulse appear to show several weaker matches from just before and up to an hour after.)

    In terms of drift, it only matters that analysis shows a reasonable match with where debris was found. This is the same reasoning for the flight path. It only matters that there is a flyable path that reasonably matches all the factual evidence.

    In terms of barnacle growth, I’m saying that a Java origin is a better match with the evidence than a cold water crash site.

    I have no need to make certain assumptions like the intentions of the pilot, when he stopped piloting, how hard the plane hit, or even much detail about the early portion of the flight. I’m certainly supportive of all research for other scenarios than mine – whatever it takes to resume the search and find the plane. That of course includes reality checks for improbable theories and speculation, but I don’t think my candidate site should be pushed into that category.

  215. 370Location says:

    @Paul Smithson:

    My apologies if this has been discussed long before my participation here, or if I’m oversimplifying, but it would seem that the accuracy of the drift modeling is crucial to your candidate site.

    I assume that your 44.5S 89E site like others in that area far south is based on satellite imagery like Tomnod. It’s my understanding that aerial and surface searches were conducted there because that’s where a 7th Arc debris field was estimated to have drifted by that time. (Well documented).

    Those early drift estimates are now moot, but suppose that a particular detail in one of those images actually was drifting debris from the plane. That imaged location would not be the impact site. It would take reverse drift modeling over the time from crash to image date, using the best available methods, to determine the drift origin and crash site.

    My expectation is that the origin would be almost as close to the 7th arc as the difference in the accuracy between the latest drift models and the earliest estimates.

  216. Victor Iannello says:

    @David: There is no way that anybody will resume the search effort along the 7th arc only based on drift results. The drift results have to be part of a larger body of evidence.

  217. Victor Iannello says:

    @370Location: I look at candidate sites as having strengths and weaknesses. The strengths increase the probability of the site being the true POI, and the weaknesses reduce the probability. The location precision of your site near the 7th arc is a strength. Some of the weaknesses (off the top of my head) are the complexity of the path after 19:41, the final BFOs, the timing, the difficulty assigning the event with a scenario for a large implosion, and the poor agreement with drift models. None of those are show-stoppers, but we have to at least acknowledge the weaknesses.

    The site 34.2S has its own set of weaknesses, including the complex path between 18:28 and 19:41, and the previous null search result. However, it is supported by a simple path after 19:41, excellent match to the BFO, fuel models, and drift models.

    The southerly site around 40S proposed by Paul Smithson is attractive because of the simplicity of the path after 18:28 and because some of the proposed area has not been searched. It’s main weakness seems to be from insufficient fuel. If we can somehow reconcile the discrepancy between the modeled and required fuel consumption, I think this location would be a strong contender.

  218. paul smithson says:

    @370Location. I abandoned the notion of 44.5S (and systematic BTO distortion) some years ago. My favoured scenario now is an early FMT completed soon after 18:33 and great circle path at M0.84, FL340. This path is predicted to cross 7th arc between 39.6S and 39.9S and my expectation would be that the wreckage lies very close to the 7th arc (BTO=18390 at 10,000ft or less).

    As Victor says, the chief difficulty is fuel endurance. I have previously suggested extending fuel endurance by: a) idle descent near end of flight – buys you 25 mins in the air for ~600kg of fuel; b) unpowered final descent ~25 mins with electrical power supplied by APU – and final APU re-start for the 7th arc logon attempt; c) the fuel models are out; (d) a final step climb to a max regular cruise altitude (?FL420) with lower headwinds – which also increases duration of idle/unpowered descent.

    (a) and (b) have a problem that the 00:11 isn’t consistent with descending at typical ROD of -1250 to -1400 fpm. And you loose quite a lot of speed/distance if you start idle/unpowered descent early. So to get to 7th arc you need the TAS reduction to be offset by a combination of lower headwind and the BTO gained from lower altitude (~+50). For (c) to be viable, the fuel models need to be out by ~7%

    I am quite sure that some combination of a or b plus c can get you there – but you then stray into the realms of requiring an unlikely combination of factors and you can’t be descending at 00:11. Out of these, I reckon (d) is the simplest but I haven’t modelled it as I don’t have wind fields for FL420. You might even complete the step climb shortly before 22:40.

    For the final 90 mins of flight, an altitude increase doesn’t lose as much TAS as it would earlier in the flight because the ISA deltas are so much lower and you have nearly constant temperature from FL360 and up. I believe fuel efficiency (kg/hr) at FL420 is something like 12% better than FL340 for same M number.

    The other possible contributory factor is that the Wx models were a bit out. Models often disagree on where transitions lie, even if they are largely in agreement on what the weather looks like either side of the transition. A more southerly transition means your strong headwinds are encountered a little later.

  219. David says:

    @Paul Smithson. A footnote to my above is that there has been no check that I am aware of that BRAN2020 modelling would not alter the conclusions drawn from the aerial search results comparison using the then BRAN2015.

  220. paul smithson says:

    @David. Yes, that’s another fly in the ointment. Though I would expect it to have smaller impact than leeway vector assumptions.

  221. Hi from CAPTIO, long time …
    I did some backlog readings 🙂

    “There are virtually endless scenarios for MH370’s End of Flight (EOF) although it seems that some are more probable than others. Several EOF scenarios have been presented and analyzed by contributors.”

    A. Kamoulakos from CAPTIO just published some early elements on a breakage analysis with the possible conclusion of a missed ditching event (asymetric first impact). Here it is :

    ” P.S. Your feedback on CAPTIO integrity is fully accepted based on my respect for your impeccable credibility. FWIW I was mistaken, my work was credited. However incorrectly credited as (highlighted by CAPTIO). Additional works published appear to be used in CAPTIO presentations and videos found on the web. I guess I would request that if our work is used by CAPTIO the source should be stated, certainly not presented to the public as original work by CAPTIO, that is very misleading.”

    We ow you apologies for not crediting one picture of yours that we used in CAPTIO flaperon breakage documentation. It skipped through our quality review and checks. We did not claim it was our original work as we tended to believe it was one from the French DGA report as your editing of the picture with the French policemen carrying the flaperon looks so “official”. We forgot it was actually coming from your report 🙂
    Again sorry for this. We will make sure to correct this in future publication or updates.

  222. Victor Iannello says:

    @David: My casual observation is that BRAN2020 drift results show the same general trends as BRAN2015. The problem with using BRAN2020 is it has not been as thoroughly evaluated by David Griffin as BRAN2015.

  223. airlandseaman says:

    Don and I will be making a presentation on the results of our spoiler analysis to the MH370 Press Conference tomorrow at 0800 UTC. The paper will be released shortly after. I’ll post a link here at that time.

  224. David says:

    @Victor. ‘Same general trends’. Yes, in January I drew up a comparison of the outcomes of BRAN2020 runs he had made with his earlier and asked for his comment. The comparisons included recoveries with time vs model predictions depicted in his “north”, “south” and “east” figures, and his MP4s animating the drifts of both BRANs.

    He responded that BRAN2015 and BRAN2020 are basically equivalent, in other words as I understand it modelling using BRAN2020 does not change his conclusions.

    That subject came up on Richard Godfrey’s blog. I can post the comparison here should that be of interest.*

    That was all about the drift to Africa and its islands. I imagine the same would apply to the short drifts to the surface search areas though I do not know.

    *There has also been work on Richard’s site which questions what conclusions can be drawn from lack of recoveries in WA.

  225. paul smithson says:

    @Victor, @Andrew. If you want to program a step climb in advance – no further action required, does that change in altitude need to be associated with a waypoint (whether named/charted or virtual)? Or can it be assigned to occur at any designated time, aircraft gross weight, fuel status or whatever?

  226. Kenyon says:

    Please find a link below to my consolidated Flaperon studies. This is about as far as I can take the analysis for now. If there are further developments the Flaperon Analysis can be revisited for update.

    Many have you have provided challenge and positive critique over the many years, thank you.

  227. Kenyon says:

    @Jean-Luc Marchand

    No problem and thank you. Based on Victor’s feedback to me it was assumed to be a minor oversight in comparison to your body of work.

  228. Andrew says:

    @paul smithson

    RE: “If you want to program a step climb in advance – no further action required, does that change in altitude need to be associated with a waypoint (whether named/charted or virtual)?”

    There are two types of step climb – planned and calculated. A planned step climb is associated with a waypoint. A calculated step climb is calculated by the FMC at the optimum point, given the aircraft weight, entered winds/temperatures, step size and cost index (ECON/RTA speed modes only). HOWEVER, in the cruise, it is not possible to program a step climb to occur without further action by the pilot. The pilot must set a higher altitude on the MCP (which can be done in advance) and push the altitude selector to initiate the climb (which cannot be done in advance). Alternatively, the pilot can set a higher altitude on the MCP, enter the new cruise altitude on the CRZ ALT line of the FMC VNAV page and press EXEC to execute the change and initiate the climb. Either way, the aircraft will not climb unless the pilot takes action to initiate the climb at the appropriate point. Any step climbs entered into or calculated by the FMC only affect the FMC predictions (ie time & fuel); they do not control the aircraft.

  229. Mick Gilbert says:


    A thorough body of work, Tom. Very well done.

    More’s the pity a comparable effort was not applied to each element of recovered wreckage by the relevant investigatory bodies.

  230. Andrew says:

    @paul smithson

    My apologies, there is a work-around that would allow the pilot to set up a step climb that would be automatically executed. The pilot would first need to force the FMC back into the climb phase. He could then enter the current altitude as a constraint at the waypoint where he wanted to climb, with the higher step altitude entered as the CRZ ALT. The aircraft would then climb automatically upon reaching the waypoint where the constraint was entered, provided the MCP had been pre-set to the higher altitude and VNAV engaged. That procedure could only be completed at an entered waypoint.

  231. airlandseaman says:

    Here is the Jeffreys Bay Debris analysis:

  232. Victor Iannello says:

    @Andrew, @paul smithson:

    I think there are two ways to perform automated climbs:

    1) As Andrew said, use waypoint altitude constraints in climb mode with the MCP altitude set to the final altitude. Custom waypoints along an LNAV leg could be programmed using distance relative to standard waypoints without the need to enter coordinates.

    2) Set the MCP speed and a very high altitude, and do an FLCH climb. The available thrust should constrain the highest altitude. As fuel is burned, the plane should continuously climb. (I believe some years ago I tested this in the PMDG777 simulator.)

  233. Tim says:

    Thanks Mike & Don for your spoiler analysis.
    As I see it, evidence is possibly now pointing towards a right wing failure before sea impact.
    Can this be used to determine the direction of the final spiral dive? To me, happy to be corrected, it might imply a right hand spiral, as I feel the right wing would experience higher aerodynamic loading in a right hand spiral.

    Over to the aerodynamicists for discussion!

  234. Victor Iannello says:

    @Tom Kenyon: That report is excellent, including the section in which you answer questions posed by fellow investigators. Thank you for your years of hard work. It’ll take a bit of time for all of us to thoroughly review it.

    @airlandseaman: You and Don make a strong case for both the identity of the part as well as a plausible failure mechanism.

    @all: Today we have two reports released discussing possible failure mechanisms for recovered wing parts (flaperon and spoiler). Both reports suggest the parts separated during a high speed dive rather than from the impact or from a controlled ditching. That’s significant.

  235. paul smithson says:

    @Victor. Thanks for responding. “Set the MCP speed and a very high altitude, and do an FLCH climb. The available thrust should constrain the highest altitude. As fuel is burned, the plane should continuously climb. (I believe some years ago I tested this in the PMDG777 simulator.)” Yes, I recall you noodling around that scenario. I think it has no application in the case I am trying to resolve because it would continually push you above optimally fuel-efficient altitude – with lower fuel savings than optimally planned step(s).

  236. Don Thompson says:


    Thank you for your report, much kudos, it’s superb piece of work.


    I view the significance of a right wing failure, during descent, as important to help bookend the interval between 001937UTC and impact. I believe that additional Computer Aided Engineering work could help evaluate that hypothesis. Perhaps a structures effort rather than an aerodynamic one.

    @Jean-Luc Marchand,

    Is there any interim work from CAPTIO/Kamoulakos, published after the March>/a> and June papers of 2020 and this most recent PDF dated for the 8th March 2021?


    Thank you.

  237. paul smithson says:

    @Andrew. Many thanks for your earlier response. As my preferred path solution requires higher speeds than any ECON setting I think the automated action is excluded. Options are therefore planned or (your workaround) automated – in both cases associated with a waypoint, real or virtual. That’s very helpful.

    The path model aligns with IGEBO-RUNUT (and continuation of same great circle thereafter). I realise that this discontinuity behaviour is at odds with your level D sim experiment, but it is the solution that pops out of my methodology based on optimising single turn timing for BTO fit.

    You reach RUNUT (13.83S) at 2100 with fuel remaining 17.0 to 17.5T, GW=191. At that weight, indicated optimal altitude is FL400 (and you don’t get to optimal FL420 until GW=170T). So if you are choosing even number FLs, then best altitude at RUNUT/2100 and beyond would seem to be FL400. Does that sound about right?

  238. Viking says:


    Due to the special day today, I took a brief look at Victor’s Blog after a long period of absence. In doing so, I noticed, that I had unfortunately overlooked your detailed work from 2018 on the seismic detectors at Christmas Island. I am very sorry for overlooking your great work.

    After reading it, I would like to have a discussion with you on some important details. Initially, I think your signal at 23:51:50 did not come from MH370. Instead, I am practically sure it came from the airplane making the second contrail I found in the METEOSAT7 data. This airplane flew exactly over the airport at Christmas Island, and the timing is perfect with an uncertainty of less than plus/minus 15 seconds. In addition, you assigned the clearest feature with an inverse Doppler chirp to an A380. As some people may remember, I also assigned that particular contrail to an A380 based on its intensity, speed, and my guess that is was due to a flight between Melbourne and Dubai taking a detour to benefit from tail wind. At the time, nobody believed in my assignment, since it did not agree with the flight schedule, and since there was some chance it could instead have been a freight plane from New Zeeland. With your information form the seismic detectors, it may now be possible to prove the identity of this airplane from the special sound feature. Why is that important for MH370? The reason is that this airplane came quite close to the airplane making a contrail in the opposite direction in almost perfect agreement with my model for the MH370 flight. If we could identify the pilots on the westbound flight over CI, they might have noticed MH370.

    An added technical detail is that you say one of the polarization angles detected by the seismic detectors did not fit MH370. Could that be because the other flight is the source? I do not know the angle definitions in the seismic equipment, so I am unable to help you here. Is it possible for you to go through the data again?

    Most importantly, your seismic data strongly corroborate that what I found in the weather satellite data were in fact real contrails. This is strong evidence, since I looked up the two flight routes in statistical compilations from Flightradar24. The westbound route over the airport is used relatively often, while the other (MH370?) route was not flown at all during 2014 or 2015. This is of course only from a statistical point of view. It could have been flow a very small number of times without showing up in the general statistics. However, what is the probability of this happening with a timing precision better than 30 s and a position precision better than 6 km exactly at the time, angle and speed, where my model says MH370 should have flown this route? I am quite sure this is less than one in a million probability. In other words, the only viable way to dismiss the solution is to say the contrail is a false positive signal. However, with your independent corroboration of the other contrail this becomes increasingly difficult to defend.

    Finally, if MH370 does in fact show up in the data from Christmas Island, its signal should appear somewhere between 23:56:00 and 23:59:00. According to your general considerations, it should peak at exactly 23:58. I get the peak at 23:57:30, but that difference is insignificant. The reason is that I am uncertain how seismic energy is transmitted from an airplane passing 50 km from the island (air, water or combination?). However, the intensity will certainly be much smaller than from an airplane passing exactly over the airport, so I am afraid there is little chance of recognizing the signal during the seismic event.

  239. Victor Iannello says:

    @paul smithson said: As my preferred path solution requires higher speeds than any ECON setting I think the automated action is excluded.

    With CI=9999, the ECON speed is Mmo = 0.87. Surely, none of your paths require higher speeds.

  240. Victor Iannello says:

    @paul smithson: The other compromise you have to contend with is bleed air. With piloted flight, the altitude and speed can be optimized for range, but that requires a pressurized cabin. With no bleed air, fuel efficiency increases, but then there is no pilot to optimize the altitude and speed.

    Tough business.

  241. Victor Iannello says:

    From the New Strait Times:

    A source from the Transport Ministry told Bernama that the government would make an important announcement on the tragedy soon.


  242. Kenyon says:

    Mike and Don, excellent investigation, report, and link to EOF scenario. In addition, your report is most helpful to understand the debris photos as they relate to a fully functional B777 spoiler as well as provide insights to key dimensions.

    I find it curious that over the years much of the findings presented by various IG members and other VERY talented researchers that participate on Victor’s blog seem to offer debris discovery, concepts, ideas and, at times conclusive proof that contradicts the official efforts and narrative. All without the benefit of hands on access to the data and debris.

    Great report and segue to a better understanding of EOF scenarios.

  243. airlandseaman says:

    Tim: Re:

    “As I see it, evidence is possibly now pointing towards a right wing failure before sea impact. Can this be used to determine the direction of the final spiral dive? To me, happy to be corrected, it might imply a right hand spiral, as I feel the right wing would experience higher aerodynamic loading in a right hand spiral.”

    As you will recall from basic training, it will be the inside wing that stalls first turning base to final. For the same reason, it is more likely the outside wing in a high speed spiral that will fail first in a theoretical world. That said, there are too many things we still do not know to form an opinion on this now. The left and right wings will not have _exactly_ the same structural strength. I think it is also possible that some pieces from the left wing separated during the final seconds of the descent. That is something AAIB-MY should be looking at along with the analysis of the spoiler.

  244. Victor Iannello says:

    @airlandseaman, @Tim: Even with neutral ailerons, the rising wing (outer) will have higher angle of attack, airspeeds, and loads. But as Mike said, the wings will have some differences in strength. Also, possibly a component (such as the flaperon) on one wing failed first, and this failure increased the dynamic loading and resulted in a progressive failure of the entire wing.

  245. Arto L says:

    @Viking, @370Location:

    I’ve been looking at the METEOSAT7 images as well. I can easily spot the contrail you assign to the other plane (possible the flight between Melbourne and Dubai), but not the one you assign to MH370? Maybe its just me. Anyway, it would be interesting to see if there is any trace of the second aircraft about 10 minutes after Ed’s detection. For example in Ed’s and CAPTIO’s scenarios MH370 would approach closer to the Xmas Island and at lower altitude.

  246. Brian Anderson says:

    Mike and Don,

    Excellent detective work. Well done.

  247. Victor Iannello says:

    [Comments here are closed. Please continue the discussion under the new article.]