MH370 Flaperon Failure Analysis

MH370 Flaperon Failure Analysis

Tom Kenyon prepared a preliminary analysis based on published photography of the right-side flaperon from MH370 found in Réunion which was made available last October in a post here. He later (in early December 2015) updated and expanded his analysis (version 2.0), and that is now available by clicking here (2.6 MB PDF).



Comparison of MH370 Flight Paths Based on Waypoints

Comparison of MH370 Flight Paths
Based on Waypoints

Richard Godfrey
1st February 2016

Many different flight models have been developed to try and fit the Inmarsat satellite data to plausible flight paths that may have been flown by MH370 under an assumption of autopilot control.

Here are a couple of typical examples:

  1. An assumed true track (TT) mode usually results in the choice of a late final major turn (FMT), fitting a course of around 186.2°T (i.e. azimuth 186.2 degrees from true north); this results in a RMS BTO error of 9.2 km and a RMS BFO error of 3.7 Hz. (RMS = root mean square; BTO = burst timing offset; BFO = burst frequency offset).
  1. A magnetic heading (MH) mode usually results in the choice of an early FMT, fitting a course of around 193.8°M (i.e. degrees from magnetic north), which results in a RMS BTO error of 18.3 km and a RMS BFO error of 3.9 Hz.

There are, of course, many other conceivable flight paths depending on assumed values for the location of the FMT, the speed mode, the altitude, step climbs, lateral offsets, winds, magnetic declination, etc.

There are two problems in particular with these typical flight model examples:

  1. The BFO error obtained at 19:41 UTC is much higher than at other times later in the flight: around 8 Hz compared with a typical BFO error of between 1 Hz and 4 Hz.
  1. The BFO data for the first attempted phone call shows a fairly steady path for 61 seconds (between 18:39:55 and 18:40:56 UTC) near either 161°T or 200°T (i.e. at azimuths close to equal angles each side of due south).

In order to see if these two problems could be resolved, I have examined alternative flight paths using a single waypoint entered into the Flight Management System (FMS) as opposed to a course selected using the heading bug; Mike Exner has argued that, if waypoint navigation were used after the FMT, then it is more likely that a single waypoint was entered.

First, I entered each possible single waypoint from both an ‘Early’ and a ‘Late’ FMT (at times and locations shown in Table 1 below) into my flight model. I kept the altitude constant at 34,000 feet and speed constant at Mach 0.817 in all cases.

Next, I determined the RMS BTO and BFO errors in each case from 19:41 to 00:11 UTC (Table 1 below); I also noted the BTO and BFO errors at 19:41 UTC (Table 2 below).

The results are given in the tables below. Green is used to highlight the best result, orange for the second best, and yellow for BTO errors below 50 km or BFO errors below 4 Hz.

Waypoint Table 1

Table 1: Derived RMS BTO and BFO errors for assumed paths to the waypoints as indicated in the far-left column for Early and Late FMTs at the locations shown. 

Waypoint Table 2

Table 2: BTO and BFO errors at 19:41 UTC for paths to waypoints as shown from both Early and Late FMTs as defined in Table 1. 



  • The BTO error is clearly a better differentiator of the results than the BFO error.
  • The result with the smallest BTO error is not the same as the result with the smallest BFO error (the plot of BTO error against BFO error gives a smooth curve for both early and late FMTs).
  • It is easier to find results with a small BTO error with a Late FMT.
  • The result with the smallest RMS BTO error was the waypoint 3090S with a Late FMT (as Don Thompson has previously suggested).
  • The result that gave the smallest BTO error at 19:41 UTC was the waypoint ISBIX with a Late FMT.
  • The smallest BFO error at 19:41 UTC was the waypoint SELSU for an Early FMT (course 215.6°T) and the waypoint NIXUL for a Late FMT (also course 215.6°T).
  • Because these smallest BFO errors at 19:41 UTC result from a course of around 215°T and this was also not far off the possible path at 18:39:55 to 18:40:56 UTC (as mentioned above), further investigations were undertaken, as below.


Further Investigations
Using pairs of waypoints, the best result obtained was for ISBIX and 3090S; but this still did not solve the BFO problems at the time of the first phone call (near 18:40) or at 19:41 UTC.

However, using triplets of waypoints, the trio NISOK, ISBIX and 3090S led to a solution to both problems:

  1. The path obtained at 18:39:55 to 18:40:56 UTC was 208.9°T, which gives a BFO error of -2.6 Hz and a constant course during the timeframe of the first call.
  1. The BFO error obtained at 19:41 UTC is just 0.5 Hz.
  1. The overall RMS BFO error is 1.7 Hz.

NISOK is a waypoint between waypoint SELSU and waypoint NIXUL, and on the same longitude (see Table 2).

The resultant flight path is shown below. It could be argued that such a path might be that followed by someone wanting to avoid Indonesia to the extent possible, whilst confusing any observer as to the ultimate intention*.

Waypoint Analysis Flight Path


It is also noted that this path fits the location and directions of the suggested sightings by Kate Tee, although the timing is out by 30 minutes from her evaluation of the time in question (which is acknowledged to be a possibility).


*Note added by Duncan Steel: A contrary argument would be that the ultimate disappearance of MH370, and in particular its apparent non-detection by Indonesian radars, is conditional upon such a path having been taken. That is, such a path having been taken either accidentally or automatically (i.e. no pilot intent) results in the aircraft’s disappearance, and if such a path had not been taken then we would be more likely to know where it went. The path that Richard has suggested, based on detailed analysis, might be considered to have a low probability a priori; but a posteriori it might be that the disappearance of the aircraft is contingent upon that path (or one very much like it) having been followed, whether by chance or by intent.


Is the search for MH370 being conducted in the right area?

Is the search for MH370 being conducted in the right area?

Richard Godfrey
(with vital contributions from Henrik Rydberg, Victor Iannello, Brian Anderson, and other IG members)
6th January 2016
(updated 7th & 8th January)

Prefatory remarks by Duncan Steel

As time goes by and the wreckage of MH370 has not been found in the priority search area, the Independent Group (IG) has continued to work on exploring other possibilities, being determined that no proverbial stone should be left unturned. As one IG member (Sid Bennett) put it rather aptly,
There is nothing new (as far as I know) in the data that we have available to work with, but there is a concern as the search goes on that we (everyone)  may have made some fundamental conceptual error.

That possibility − some fundamental conceptual error − remains. However, the immediate thing that we question is our assumptions. IG members have already indicated (as noted in the text that follows below) that those contributing to the Australian Transport Safety Bureau’s assessment of most likely crash area have made various dubious assumptions, some of which might not even have been recognised by those in question. In particular the IG has identified various errors in the book prepared by the (Australian) Defence Science and Technology Group (as linked below), only a few of which have yet been described and discussed on this website; future posts here will say more in that regard.

Turning back to the IG’s own assumptions, to date we have been fundamentally assuming that in deriving a best estimation for the route taken by MH370 down into the southern Indian Ocean after its turn southwards at the northwest of the Malacca Strait we should use only the Inmarsat BTO and BFO data, coupled with other plausible indicators (the fuel-defined limits; the apparent aircraft speed prior to the southward turn; and the aircraft altitude as limited by the radar detections through to 18:22 UTC). In essence, as I noted early in the piece in 2014, we allow Occam’s Razor to be our guiding principle: do not assume anything that you do not need to assume.

However, we now have a new piece of information. Simply: the aircraft has not been found within the priority search zone. If that continues to be the case then we must consider other possibilities which might conform to the known data (Inmarsat BTO and BFO values, and the fuel limits which can work either way, either setting a range limit or else requiring fuel to be burnt more quickly per unit distance) and lead to a revised end-point for MH370 that is outside of the search zone, and north of it (given that the fuel limitation prohibits end points further south).

A possibility − that is, one might say an assumption to be trialled − is that the turn to the south at the northwest of the Malacca Strait was made by a conscious or semi-conscious pilot after some severe malady on board, the intent being an emergency landing at Banda Aceh. That is not a new suggestion, and in later posts the IG will pay some attention to detailing what is involved, and giving due credit to whoever first suggested it in terms of a fit to the BTO and BFO data. The point here is that a gradual descent starting from the turn south might explain the BFO values; the BTO values might be fitted; and the additional fuel consumption from flying at a lower altitude (indeed at a high speed and therefore with high drag) might exhaust the fuel at the right time (after 00:11 UTC) but with a much-reduced distance flown.

This scenario would involve the pilot being again incapacitated after making that southerly turn and starting that approach from afar, the autopilot then continuing to fly the aircraft as it had been instructed until the end of the flight. As I wrote above, this complete scenario will be described in more detail in later posts.

Immediately, however, the relaxation of the constraints (use only the BTO and BFO data) previously imposed, due to the non-discovery of the wreckage of MH370, has led to a variety of other steps forward in terms of the IG re-assessing where MH370 might have ended up. For instance, how curved paths might have occurred, and related matters. The post that follows below stems from this. In collaboration with several other IG members Richard Godfrey has found that, dependent upon the autopilot roll mode that is assumed, it is feasible that the aircraft actually ended up rather further north than the current priority search zone. A strong recommendation therefore follows, and the ATSB is urged to repeat the analysis described below in order to confirm its veracity, and then alter the search zone accordingly.

The IG published a report dated 26th September 2014 in which the most-likely end-point of MH370 was identified to be near 37.71S, 88.75E. This conclusion was based on a best fit of the BTO and BFO satellite data; it was also in accord with the calculated range given the available fuel. Both considerations (satellite data, fuel limitations) made use of an assumption of near-level high-altitude flight at a speed in accord with long-range cruise (LRC) under autopilot control.

The ATSB report update and DSTG book preprint made publicly available on 3rd December 2015 indicate a search area centred on 38.0S, 88.3E and therefore aligns closely to the previous IG publication.

However, as no aircraft debris has been found around this location, it seems reasonable to question the assumptions necessarily made in performing the calculations.

IG members Richard Godfrey and Victor Iannello published an initial response to the latest ATSB report and DSTG book on 9th December 2015, showing that the Bayesian approach employed by the DSTG team has a pronounced bias favouring straight flight paths (and thus militating against curved flight paths) as well as a bias toward higher speeds/against lower speeds. Indeed by allowing only speeds between 0.73 Mach and 0.84 Mach the DSTG analysis precludes many curved paths.

In a follow-up post dated 14th December 2015 another IG member, Brian Anderson, explained in more detail various misinterpretations of the Aircraft Control Modes that appear in the DSTG book (and thus detrimentally affect the analysis). Anderson explained in that post the four autopilot roll modes, referenced to either True North or Magnetic North, and following either a Heading or a Track:
1. True Heading.
2. Magnetic Heading.
3. True Track.
4. Magnetic Track.

If the autopilot roll mode was True Track then the aircraft followed a straight path or loxodrome. In that mode the aircraft compensates for the effect of winds and does not need to adjust for the effect of magnetic declination.

In the other three options (1,2 and 4 above) the aircraft follows a curved path: there is no compensation in either of the heading modes (1,2) for wind field, and there are adjustments in the magnetic modes (2,4) for the magnetic declination (which varies with location i.e. latitude and longitude).

In the present post I show that an assumption that the autopilot roll mode is True Track and the aircraft therefore follows a straight path (a loxodrome) may lead to an incorrect estimation of the place where MH370 crashed into the southern Indian Ocean (SIO).

Of course, it is also possible that a more deterministic route was followed based on waypoint(s) entered in LNAV (Lateral NAVigation) mode. Then Great Circle paths would have been followed between the designated waypoint(s), and after overflying the last waypoint the plane would have continued on its final bearing.

Current ATSB Search Area
The current definition of the ATSB search area extends only from 35.5S to 39.5S, with a width 40 NM either side of the 7th Arc.

Previously the ATSB defined the search area to extend from 33.0S to 39.5S (i.e. starting at a more-northerly latitude), but along a narrower band adjacent to the 7th Arc.

The diagram below dated 15th December 2015 is courtesy of Richard Cole.



Latitude Distribution at 00:11 UTC by Autopilot Roll Mode
Henrik Rydberg (IG) has run a comparison of the autopilot roll modes to show the latitude distribution at 00:11 UTC (i.e. at the 6th Arc) based on a large Monte Carlo simulation and employing uniform BFO scatter between ±25 Hz. Henrik’s results are shown in his graph as below.


At 00:11 UTC the range of latitudes from 29.0S to 40.0S along the 6th Arc covers over 99% of Henrik’s four distributions (one for each autopilot roll mode).

As can be seen in the plot above, the different autopilot roll modes give different latitude peaks and distributions. In the following table, I show there is a good fit to the satellite BTO and BFO data at 00:11 UTC. That is, any of the four autopilot roll modes is feasible in terms of what we know from the Inmarsat BTO and BFO data.

Update 8th January: The following table from RG includes the indicated positions at 00:19 (UTC) [i.e. 7th Arc]. 


There is also a general fit to the fuel range in each case, when one allows for the additional distance flown due to the curvature of some of the flight paths and the fuel used manoeuvring in the Strait of Malacca before the Final Major Turn (FMT) southward.

The area previously searched by the ATSB from 33.0S to 39.5S (and near the 7th Arc) corresponds in general to the range of latitudes in the True Track autopilot roll mode. The peak at 36.5S in the True Track mode at 00:11 UTC on the 6th Arc translates to an end point at 37.43S on the 7th Arc at 00:19 UTC. This aligns with the original IG estimate of 37.71S, and the recent ATSB estimate of 38.0S.

The vital point being made here is this: if the autopilot roll mode in operation on board MH370 were any of the three possibilities other than True Track, then the aircraft would have arrived in the SIO at a latitude north of the current search zone (i.e. north of 35.5S, and possibly as far north as 29S).

The ATSB should also include a search 20 NM either side of the 7th Arc from 29.0S to 33.0S (and also those areas down to 35.5S which have not yet been searched out to ±20 NM), because we do not know which autopilot roll mode was in use at the time of the disappearance of MH370.

Together with revisiting the small percentage of areas within the original priority search zone already examined, where it was not possible to discount potential aircraft debris findings, the extension of the search area further north should be a higher priority than extending the width of the search area to 40 NM either side of the 7th Arc, that being the current search strategy.

Addendum from Mike Exner: January 7th. 

Readers should be aware that not all four autopilot roll modes are equally probable to have been in use. The two ‘Normal’ (or Magnetic) modes are far more likely than either of the True modes.

Quantifying such relative likelihoods as formal mathematical probabilities is perhaps intractable, but that does not mean we should totally ignore the fact that some modes are more likely than others due to the standard operating procedures (SOPs) that are generally employed.

As explained by Brian Anderson it is SOP for all navigation between latitudes 60S and 60N (i.e. far from the magnetic poles) to use the magnetic reference. If the flight was controlled by the autopilot (i.e. Flight Management System, FMS) after overflying a final waypoint (WP), then it was more likely to have followed the Magnetic (Normal) heading last used at that final WP, making an end point on the 7th arc somewhere further northeast most likely (i.e. at a latitude north of about 35S, as indicated in Richard Godfrey’s post and Henrik Rydberg’s graph, above).


Space Scientist, Author & Broadcaster