The above graph is taken from the DSTG book “Bayesian Methods in the Search for MH370, ” page 90. It shows the probability distribution of MH370’s endpoint in the southern Indian Ocean based on analysis of the different autopilot modes available to whoever was in charge of the plane during its final six hours. It was published earlier this year and so represents contemporary understanding of these issues. As you can see, the DSTG estimated that the probability that the plane hit the 7th arc north of 34 degrees south longitude is effectively zero.
I interviewed Neil Gordon, lead author of the paper, on August 11. At that time, he told me that experts within the official search had already determined that the BFO values at 0:19 indicated that the plane was in a steep descent, on the order of 15,000 feet per minute.
Such a rate of descent would necessarily indicate that the plane could not have hit the ocean very far from the 7th arc. Nevertheless, Fugro Equator, which was still conducting its broad towfish scan of the search area at the time, spent most of its time searching the area on the inside edge of the search zone in the main area, between 37.5 and 35 degrees south latitude, about 25 nautical miles inside the 7th arc. At no point between the time of our interview and the end of the towfish scan in October did Equator scan anywhere north of 34 degrees south.
Shortly thereafter, the ATSB hosted a meeting of the experts it had consulted in the course of the investigation, and the result of their discussion was published on December 20 of this year as “MH370 – First Principles Review.” This document confirms what Gordon told me, that the group believed that the BFO data meant that the plane had to have been in a steep dive at the time of the final ping. What’s more, the report specified that this implied that the plane could not have flown more than 25 nautical miles from the 7th arc, and indeed most likely impacted the sea within 15 nautical miles.
By the analysis presented above, a conclusion is fairly obvious: the plane must have come to rest somewhere south of 34 degrees south, within 25 nautical miles of the seventh arc. Since this area has already been thoroughly scanned, then the implication is that the plane did not come to rest on the Indian Ocean seabed where the Inmarsat signals indicate it should have.
I would suggest that at this point the search should have been considered completed.
Nevertheless, the “First Principles Review” states on page 15 that the experts’ renewed analysis of the 777 autopilot dynamics indicates that the plane could have crossed the 7th arc “up to 33°S in latitude along the 7th arc.”
Then in the Conclusions section on page 23 the authors describe “a remaining area of high probability between latitudes 32.5°S and 36°S along the 7th arc,” while the accompanying illustration depicts a northern limit at 32.25 degrees south.
In other words, without any explanation, the northern limit of the aircraft’s possible impact point has moved from 34 degrees south in the Bayesian Methods paper in early 2016 to 33 degrees south on page 15 in the “First Principles Review” released at the end of the year. Then eight pages later within the same report the northern limit has moved, again without explanation, a half a degree further north. And half a page later it has moved a quarter of a degree further still.
Is the ATSB sincere in moving the northern limit in this way? If so, I wonder why they did not further search out this area when they had the chance, instead of continuing to scan an area that they apparently had already concluded the plane could not plausibly have reached.
I should point out at this point that the area between 34 south and 35.5 south has been scanned to a total widtch of 37 nautical miles, and the area between 32.5 and 34 has been searched to a total width 23 nautical miles. Thus even if the ATSB’s new northern limits are correct, they still should have found the plane.
As a result of the above I would suggest that:
a) Even though most recent report describes “the need to search an additional area representing approximately 25,000 km²,” the conduct of the ATSB’s search does not suggest that they earnestly believe that the plane could lie in this area. If they did, they could have searched out the highest-probability portions of this area with the time and resources at their disposal. Indeed, they could be searching it right now, as I write this. Obviously they are not.
b) The ATSB knew, in issuing the report, that Malaysia and China would not agree to search the newly suggested area, because it fails to meet the agreed-upon criteria for an extension (“credible new information… that can be used to identify the specific location of the aircraft”). Thus mooting this area would allow them to claim that there remained areas of significant probability that they had been forced to leave unsearched. This, in effect, would allow them to claim that their analysis had been correct but that they had fallen victim to bad luck.
c) The ATSB’s sophisticated mathematical analysis of the Inmarsat data, combined with debris drift analysis and other factors, allowed them to define an area of the southern Indian Ocean in which the plane could plausibly have come to rest. A long, exhaustive and expensive search has determined that it is not there.
d) The ATSB did not fall victim to bad luck. On the contrary, they have demonstrated with great robustness that the Inmarsat data is not compatible with the physical facts of the case.
e) Something is wrong with the Inmarsat data.
@Gysbreght, I fear you may be ahead of me on the this one. What do you feel this graph tells us?
BTW I am going to my in-laws tomorrow; my father-in-law is a onetime PC hobbyist and has retrieved one of his old XP machines and loaded it with FS9 and the 777-200LR add-on. I married well…
@Gloria, Your comments are pretty insane. From now on I’m going to manually approve your contributions, so you may notice a delay before they appear, or they may not appear at all.
@jeffwise: Positive Z is north. Positive X is east.
@Matt Moriarty
Thank you for the sim study results. I am confused if the results contradict the PMDG 777 model behavior that VictorI is using.
@Susie C
I have seen that interview and it just strikes me that guy H is being defensive, evasive, and political. He is trying to say the military handled the MH370 situation correctly (by ignoring it), but a better example of correct behavior was this year when the UK RAF intercepted and escorted a Mexican passenger plane which had lost “communications”.
VictorI said; “I have shown how the plane was flown from 45S1 to 45S2”.
You have shown how a sim plane could be flown from 45S1 to 45S2.
This does not explain, for instance, the precise 3999.99 feet altitude
value found in the sim data, a value that in FSX may have application as
the immediate value prior to 4000 at which background colur changes may
occur in certain add-on softwares to MSFS (not proven to apply to FS9,
but Shah may not have known that – Shah was a safety-conscious,
professional pilot, but no-one ever suggested he was a particularly savy
computer user).
David said; “experimentation/proving with/of the program.”
David is suggesting that he was experimenting with the program to test
the flight sim program – not to obtain any data to apply in his
real world flying. There was a certain interest in the experience of the
flight sim, comparative to how it approached his experience of real
flying, and to ‘fly’ to places familiar, rarely visited, or those that he
would never fly and land at – but are you asserting that the computer
problems his kin/friends & colleagues say he told them about, for many
months before March 2014, was all an act?
You apparently are suggesting that Shah found it so enjoyable to
comtemplate the act of flying to suicide, that he then eventually went
& actioned a suicide flight, with some happenstance murders to boot.
OK, that’s your theory.
Why not relate to us which files in the list of dated files in Table 7 of
Data-from-Flight-Simulator-Computer.pdf, match which particular files in
the FLT files recovered from MK25? It is not as if that information were
Shahs bank account details or passwords – rather, it is info which is
only specifically relevant to the flight sim data.
@JeffWise @all
Jeff, in an earlier post I challenged your conclusion that the Inmarsat data was wrong as did several others. You then asked those that did to prove that it was not wrong.
It is certainly possible that your statement could be true, but your cannot absolutely make the statement.
What is true is that the search zone predicted by the DSTG Bayesian model using the Inmarsat BTO/BFO data is not correct because MH370 has not been found in the predicted zone.
To blame this exclusively on the BTO/BFO data set means that you accept the entire DSTG model as 100% correct.
Even DSTG no longer believes in the assumptions that form the basis of the prediction of the 120K km2 zone. But this is an embarrassing admission.
There are many assumptions that could clearly be challenged. The AP modes were raised, but there are others that could be challenged.
But I doubt DSTG has any interest in revisiting the model after spending $180 million on the search centered on Longitude 88.1 E, Latitude 38.0 S.
Any comments on the maneuver model assumptions used in the DSTG book?
@JeffW
FYI you can put FS9 on a Win10 computer. There is a so-called “no CD” patch, which allows FS9 or FS9.1 to run without CD#4 in the CD drive. Win10 needs this patch. Not sure yet about the PSS model maybe I finally try that install soon.
@Jeff Wise
You said: ”
@Michael Helms, Whatever mechanism by which the data was spoofed, assuming it was spoofed, would have returned BFO values that indicated a flight to the south no matter how many incoming calls the plane received.
”
I am aware that there would be BFO, and if spoofed they could be used to show “somehow it flew south”. But i think without the BTOs it would not be sufficient to convince the perps that the authorities would set all energy in this faint information and ignore other leads like from the not reliable witnesses, which may appear , e.g. that the authorities would eventually judge else than you about Kate Tee (“Kate Tee’s alleged sighting is a rogue wave of horseshit”) or the Maldives sighting.
So i would think the perps would come to the conclusion, such a spoof is not worth the energy they would need to invest.
As for my first point: Since You did not answer to this, may i ask once more in other words:
If “e) Something is wrong with the Isat data” is valid, why would/should we take serious only that version, that is mirroring the then true path by some 180 degree?
Just an example. If a live person was at the controls, the aircraft could have executed a 360 degree clearing turn between pings. The time between arcs might still remain within the minimum Mach considerations. The turn model did not allow for this.
You simulator people are consider fuel flow which also affects weight and this impacts aerodynamics – major impact on duration. The DSTG model assumed infinite fuel and had no fuel consumption model. They did some post analysis to toss profiles that exceeded Boeing range estimates.
All of the validation flights were typical airline long range cruise profiles. Even here some figures showed weird branches that were not well explained in the book. May a embedded fuel model would have eliminated these “false” paths.
All that we know is that for the data provided by Inmarsat and the model described in the DSTG book, the model did not make a good prediction.
@TBill: I run FS9/PSS 777 on Windows 10 with the “No CD” patch without a problem. I have FSX/PMDG 777 installed on the same machine. They happily co-exist.
@buyerninety said, “This does not explain, for instance, the precise 3999.99 feet altitude value found in the sim data…”
No, actually we explained this. After the simulation was paused at 45S2, the altitude changed to 4,000 ft, and the flight file saved, the value appearing in the flight file is “Altitude=+003999.99”, exactly as it appears in the recovered file.
Regarding what you represent my theory to be, it is not. There is no sense on my commenting on your theory of my theory.
How has the FSx stability been on replaying the data recovered from ZS’s computer for those who are replicating the installation on their own computers?
‘Is the new ATSB search area sound?’
I actualy wonder how ‘new’ this area is.
If they are convinced it must be within 25Nm of the 7th arc they already scanned most of this ‘new area’ up to and even somewhat beyond 33S according to @Richard Cole’s latest graphic as I read it well:
https://twitter.com/richard_e_cole
If this is what it looks like Mh370 did not crash very close to the 7th arc in this area and this still gives no proof or indication the 0:19 BFO conclusions of a steep high speed crash are correct. On the contrary I would suggest.
Combined with the latest big shunk of trailing-edge debris with all the other wing/control surface related debris finds, a high speed dive crash becomes more and more unlikely IMO.
The debris trail shows obviously overwhelming mostly pieces of debris from an airplane that would seperate during a relatively low speed horizontal entering of the water.
So is the new area sound? Probably not enough IMO.
But it has to be searched to make sure it’s not there either, for perhaps it is.
@Jeff Wise: “What do you feel this graph tells us?”
Look upon it as lift coefficient versus angle of attack. To obtain the lift coefficient for a weight of 180 tons and 37650 ft altitude the value on the vertical scale must be multiplied by a factor of 0.281.
VictorI has told us that in the B777 model used in MSFS the zero-lift AoA is -(minus)4.2 degrees. The graph illustrates that the values of loadfactor and AoA that can be observed in Victor’s simulation of 27 seconds of flight correlate reasonably well(*) with that zero-lift AoA and a linear cL-alpha relationship. I honestly don’t understand Victor when he says that the FCOM description of the B777 PFD is irrelevant for the MSFS simulation of that PFD.
(*)Bearing in mind that the loadfactor shown is rounded to only two significant digits.
HAVE LOTS OF FUN WITH FS9 TOMORROW!
Since there seems to be some confusion regarding the simulation variables stored in the flight files, I am including below all the SimVar values that existed after 27s of my simulation, before and after the change in altitude to 4,000 ft.
It is important to note that the vertical speed before the change in altitude is YVelWorld = -12.3 fps = -738 fpm (descending). Yet, after the change in altitude, YVelWorld = 24.5 fps = 1,470 fpm (climbing). This once again demonstrates why in the recovered data set for 45S2, the plane was at 4,000 ft and climbing. It is an artefact of how the altitude was changed, and it is repeatable.
45S2 before changing altitude to 4,000 ft
[SimVars.0]
Latitude=S045° 07′ 47.7081″
Longitude=E104° 08′ 38.4512″
Altitude=+037653.90
Pitch=-4.3087744937088219
Bank=-0.13368120603826217
Heading=-173.91712176294601
PVelBodyAxis=0.00018614309386330585
BVelBodyAxis=-0.0020271815217738072
HVelBodyAxis=-0.0022713792578914439
XVelBodyAxis=-41.58377886262317
YVelBodyAxis=-54.603579041882462
ZVelBodyAxis=560.31061062745562
DynPres=106.56777021305984
AGL=37654.545100655138
SimOnGround=False
PAccBodyAxis=0.0036237087108499772
BAccBodyAxis=-0.040136805729038924
HAccBodyAxis=-0.044906914081202023
XAccBodyAxis=0.52229238740584405
YAccBodyAxis=-1.4274932715008966
ZAccBodyAxis=-1.1198455434698196
PVelWorld=0.00018084306173458725
BVelWorld=-0.0018560135825955791
HVelWorld=-0.0022782465263876672
XVelWorld=-18.164120121706603
YVelWorld=-12.255331923815755
ZVelWorld=-564.0732245518218
IceAccum=0
AutoFeatherSwitch=False
MaximumGForce=2.2000000000000002
MinimumGForce=0.14499999999999999
45S2 after changing altitude to 4,000 ft
[SimVars.0]
Latitude=S045° 07′ 47.9993″
Longitude=E104° 08′ 38.4000″
Altitude=+003999.99
Pitch=-4.3087744394544405
Bank=-0.13368115784497478
Heading=-174.0096132162906
PVelBodyAxis=0
BVelBodyAxis=0
HVelBodyAxis=0
XVelBodyAxis=0
YVelBodyAxis=0
ZVelBodyAxis=325.58327269368777
DynPres=114.38446786895769
AGL=37654.545100655138
SimOnGround=False
PAccBodyAxis=0.0036237087108499772
BAccBodyAxis=-0.040136805729038924
HAccBodyAxis=-0.044906914081202023
XAccBodyAxis=0.52229238740584405
YAccBodyAxis=-1.4274932715008966
ZAccBodyAxis=-1.1198455434698196
PVelWorld=0.00018084306173458725
BVelWorld=-0.0018560135825955791
HVelWorld=-0.0022782465263876672
XVelWorld=-33.882354848785759
YVelWorld=24.461539685191507
ZVelWorld=-322.89020564957741
IceAccum=0
AutoFeatherSwitch=False
MaximumGForce=2.2000000000000002
MinimumGForce=0.14499999999999999
@Michael Helms, No one outside of Inmarsat knew that they had started to log BTO values, so any perp who took the plane and spoofed the data would not have known that there would be any point in attempting to spoof it (if it is even possible.) From an abuction perspective, BFO values alone would be more desirable because they are too vague to allow a search zone. The BTO value has permitted a search whose failure have called the whole shenanigans into question.
This also answere BTW your question that you reiterate at the end: how does this allow us to determine the true path? I suggest that you read “The Plane That Wasn’t There,” but the short answer is that the presumably unspoofed BTO values by themselves indicate the track that would have been followed.
@VictorI, Thank you, that is extremely enlightening.
@Gysbreght, Thank you!
@Hank, You wrote, “All that we know is that for the data provided by Inmarsat and the model described in the DSTG book, the model did not make a good prediction.” That’s one way of looking at it, and it is entirely correct if one assumes that the Inmarsat data could not possibly have been spoofed.
For the last two years or so, we have been playing a sort of game. Those of us who think that the SDU was rebooted so that hijackers could spoof the BFO data (call us “Spoofs”) thought that the seabed search would turn up no wreckage. Those who felt that it would be impossible to spoof the BTO data (call them the “NoSpoofs”) believed that the plane lay within a region along the seventh arc. At the beginning, the NoSpoofs were extremely confident that the plane would be found. (NoSpoofs today might deny this but you can go back and Google what they said.) Then, as time went by, they said (or rather, they didn’t say, but this was their position), “we’ve been unlucky,” and then “we’ve been really unlucky,” and then “we’ve been really, really, really unlucky.”
Whereas from the Spoofs perspective, the game started out with us saying to the UnSpoofs, “you may be right,” to then, as more of the probability heat map was searched out, “you’re probably not right,” to finally, “you’re almost certainly not right.”
It’s possible to cook up explanations of how the plane might have wound up in the southern Indian Ocean outside the area already scanned (one example being the loiter scenario you describe) but from an a priori probabilistic perspective the odds for this are very low.
So this is how the game ends: with the official search being put into indefinite suspension, and the NoSpoofs (include the ATSB and yourself) saying, “Oops, we were unlucky,” and Spoofs like myself saying: “You screwed up by failing to take into consideration the possibility that the SDU reboot was evidence of tampering. You weren’t unlucky, you were bamboozled.”
Well, that’s not exactly the end of the story, as we will no doubt continue to pick over other data for a while, including the debris. I personally feel that the flaperon trailing-edge barnacles and Mg/Ca analysis are a smoking gun confirming the spoof hypothesis, while others take the same tack that they do with the SDU reboot: “I’m sure there’s an innocent explanation.”
@Jeff
There is a third category – competent nospoofs. I never endorsed a spoof, and I am have been firmly of the opinion (from before an underwater search was started) that the initiation of an underwater search was foolish based on the available data. You can Google that as well.
The ATSB has not been unlucky. They have been stupid.
@jeffwise said, “It’s possible to cook up explanations of how the plane might have wound up in the southern Indian Ocean outside the area already scanned (one example being the loiter scenario you describe) but from an a priori probabilistic perspective the odds for this are very low.”
Do you realize what you just said? That reasoning is completely circular.
It is the a priori probability distributions that were ASSIGNED (hence a priori) for number of manoeuvers and vertical speed at BFO times that led to the current search area. I have been saying for some time, including just after the DSTG report was issued, that the DSTG results, i.e., the current search area, are only as good as these assumptions. It was a place to start the search, but not finding the plane there does NOT invalidate the satellite data.
@JeffW
Lot’s of folks feel the pilot(s) were faking a north or westerly flight before SIO. That being the case, if I am the pilot, no way am I going to just hang a simple left turn near Banda Aceh. VictorI shows a holding pattern at VOVX, but I’d be inclined to think MH370 went northwest on waypoints, possibly with an IGARI-type double-back turn around. You point to loiter as some stretch of the imagination, rather I see “loiter” as the logical path.
@Jeff,
Until it was determined by some really smart people in the IG how the BFO data was able to determine how the path south worked, quite a lot of path analysis was done to try to identify a northern path based on the BTO data alone. I think that all those attempts failed. The primary reason being that the pattern of winds at the time, particularly the very strong westerly winds across the Bay of Bengal, northern India and Pakistan etc made it impossible to get any reasonable fit.
Here is a piece I co-authored with Richard Godfrey on Dec 9, 2015, just days after the DSTG report was issued, that explains why the DSTG results favored straight and high speed paths. Paths with either a loiter or a large number of turns, even if they occurred before 19:41, would be ranked low.
I am an unSpoof, but I never claimed that the aircraft would be found in the current search area. In fact, here was the conclusion from that paper:
***
In summary, the recent DSTG report presents a generalized, Bayesian method to prioritize the search zone. Although the ATSB is using this report as justification of its current search area, the alignment between the new results and previous results are a result of an inherent bias towards straight, high speed paths. More generalized methods would allow a larger range of end points. This bias should be acknowledged.
***
https://web.archive.org/web/20161025150724/http://www.duncansteel.com/archives/2090
@TBill: The proposed loiter at Car Nicobar was based on an observation that the straight path after 19:41 that aligns with McMurdo Station AND satisfies the satellite data also aligns with Car Nicobar. In fact, we really don’t know what happened between 18:28 and 19:41. The failure of the current search area demonstrates to me that the flight was not fast and straight between 18:28 and 19:41. For many reasons, the ATSB is unlikely to reach this conclusion.
@JeffWise @DennisW @Victori, @TBill
Victori, TBill, and DennisW – You seem to share my view that flawed assumptions and reasoning by ATSB could result in a serach area that does not contain MH370.
Jeff, thanks for the description of the different perspectives of blog participants. I am new to the blog but have been following the MH370 story since it began. I am clearly a NoSpoof.
I cannot rule out this is another EgyptAir 990 or SilkAir Flight 185. The responsible crew member (most likely Z) did not want the FDR ever found and decided to plant MH370 in the SIO. The Malaysian government would rather not find the aircraft than have it confirmed to be a suicide – just like Egypt and Indonesia for those suicide flights. Z would have known about Air France 447 and the role that ACARS played – and turned it off. He could have glided the aircraft many miles from Arc 7 intercept.
The alternative fire and hypoxia, diversion to Penang, followed by a flight attendant avoiding land by going northwest from Penang and staying over water could also be true. As TBill notes a holding pattern could be used and then maybe off the the SIO by some final decision to not harm anyone on the ground.
As Victori notes the a priori distributions and maneuver models decided by DSTG were questionable from the start.
I am certain that many of the folks with flight simulators could construct a route that would cross each of the seven arcs at the correct time and run out of fuel near arc 7 and not require the aircraft to be within the designated search zone. They could use S-turns and clearing circles that Z might have done just for something to do.
ISIL hijacking MH370 to Kazakhstan, planting aircraft parts to drift towards Africa, and hacking the SDU in flight or the Perth ground station as a cover and the other Spoof theories too far for me to go. A well executed suicide by Z or a fire followed by a do no harm on the ground evasion are as afar as I can get.
@VictorI
You use the word “ASSIGNED” for “the a priori probability distributions” and “use the word “assumptions” for DTSG interpretation.
Is that another way of saying both are arbitrary, but the assignations cannot be “invalidated”, only the DSTG’s “assumption” of them can?
@Matt Moriarty,
It seems that your SIM test actually confirms what the manuals say. Depending on which manual one reads there are two statements: (i) the aircraft maintains the same heading, or (ii) the aircraft maintains the same track.
Care is necessary with the interpretation of these statements. They need to be read as a pilot, in an operational sense. In doing that both statements are true, at least over the length of time that one might encounter in an operational sense, while seemingly saying different things.
The statements do not say anything about reverting to anything previously selected, nor whether NORM or TRUE is selected as the heading reference.
Victor’s observations of his FSim operation also confirm this.
If the heading is maintained then that heading will be the True heading of the aircraft at the point of the discontinuity [or end of route]. The fact that the heading display on the MFD is MAG is incidental.
The aircraft, at that instant is also flying a track that is wind corrected. If the wind stayed constant then both the track and the heading would remain constant.
Good luck getting a definitive statement from Honeywell. Many months ago I tried, and I copy here the response to my question.
“If LNAV is engaged and the aircraft encounters a route discontinuity (either within the route or at the end of the route), the operation is LNAV will remain engaged and take a snapshot of the current magnetic heading and maintain that heading until another roll mode is selected or the discontinuity is resolved. This is referred in the design as the Heading Hold Control mode. There is no “Track Hold” method associated with this type of scenario.”
Note that even this requires some interpretation, because while the reply says “snapshot” of the current magnetic heading, we have to appreciate that this heading is actually referenced to the True heading at that instant and it is actually the True heading that the aircraft uses to manage its track.
Hence my view: the aircraft will maintain a constant True heading, and the track will be subject to changes in the wind field.
I note that the ATSB came to a different conclusion after talking to a number of operational B777 pilots. That is not surprising. I have also talked about this question to a number of B777 pilots and they all gave the same answer, But, when questioned further about what happens to the aircraft track over time they did not know, because the situation never arises in a normal operational sense. Perhaps the ATSB have incorrectly interpreted the answers they received.
@Susie Crowe @Victori
The DSTG model creates a probability distribution for parameters at the last radar fix. The exact Mach number, altitude, heading and other parameters are not known so when the create a random trajectory to evaluate these distributions are used.
They also crated a model to create the test trajectories based on speed, altitude, and heading changes at random intervals and magnitudes. The simulation creates a random flight and the BTO/BFO measurements are used to weight the probability for that flight given the measurements.
The unweighted flights are like a starburst from the initial position. All of the ones going northeast, for example, are eliminated. But the model does create flights going everywhere from the start location.
The assigned distributions are not arbitrary but there are judgements made between the last ACARS report and the state of the aircraft at the last radar fix. The start distributions and modelling used for the random flights are all subject to DSTG judgement.
@VictorI
In 2016 hindsight, that was a good paper.
Re: sim studies, have you given any thought to the G-forces changes after 10N90E? Gysbreght has shown an analysis of the radar near IGARI showing a quick climb to FL400+ and then a quick descend. I did a test, I was able to get to about FL440 at 3000 ft/min and then stalled, saved case, turned the nose down to descend and it was actually the nose down/descend step that generated the 2+ G force. Obviously I am suggesting a potential parallel between IGARI manuever in the sim studies it occurs at 10N90E.
@Hank
You said, “As Victori notes the a priori distributions and maneuver models decided by DSTG were questionable from the start.”
Why? Was it a lack of appropriate intellect, an unwillingness to admit full comprehension? Why would a team assigned to an aviation disaster of this magnitude, be “questionable”
@Susie Crow: I am being a little sloppy with my terminology to make things simple. The a priori distribution for the number of manoeuvers was probably based on statistics from previous commercial flights, where aircraft generally fly straight and fast. To apply the same distribution to MH370 in my view is erroneous. There might be a mathematical justification to do this based on sampling prior flights, but not a logical justification in that the sample does not represent MH370.
Possibly I should have said, •allowed to• be questionable
@TBill: I have studied the radar data, and I have flip-flopped many times on the interpretation of the speed data after IGARI. My current thinking is the plane flew at a nearly constant M0.84/FL340 until 18:22. I think the speed distortions were produced by missing data points and inaccurate position data, which produced speed artefacts when the estimator used this data to produce the trajectories.
We really need the raw radar data to settle this. I have been pushing for this for a long time.
Here is a paper I wrote (before the DSTG report) that discusses the radar data:
https://www.dropbox.com/s/zh9rfqa6rxy582m/2015-08-18%20Radar%20Data%20for%20MH370.pdf?dl=0
@matt
“Your resistance, which you predicate on ‘he seems like a nice guy, nothing in his psych background’ FLIES in the face of what appears to have taken place, and most significantly, of where the plane ended up….Hint (and for StevenG): Not in the shallow waters off Cocos, or any other nearby island.”
so where do you think the plane is and why it ended up right there?! Do you really think anyone would go through such a complicated evasion procedure just to kill himself?! It doesn’t add up whatever suicide scenario you imagine that ends around the possible non-searched areas of the 7th arc.
@Susie Crowe
ATSB selected the Australian military R&D organization (DSTG) to lead the effort to define the search area with help from many other organizations including Inmarsat and Boeing.
DSTG is a competent group with experience in target tracking algorithms. The leader of the group Dr. Gordon was one of the founders of the concept known as a particle filter. The PF concept has been used for missile tracking problems. Inmarsat used a different approach to define the surface search zone in March 2014 – not a PF.
I have no insight into whether Dr. Gordon promoted the PF and why the earlier Inmarsat approach was 100% dropped.
As Victori points out the PF was used to rank sample trajectories based on the BTO/BFO measurement but much judgement was used for the rules used to create the samples. Assumptions were made such as turns could only be made from -180 to +180 degrees. This would prevent a 360 degree turn. The intervals between random turns had rates based on prior flights so it would also not be possible to have a sample of two sequential 180 degree turns. So these rules for sample flights constrained MH370 to fly long airline routes.
One constraint that bugs me is that they assumed infinite fuel for each sample route. They intended to deal with this at the end by dropping some routes – but this is not specified how it is actually done. It would have been possible to constrain the fuel to 43,800 kg if fuel burn was modeled, but it was not. This is separate from the BTO/BFO data, but the fuel was not considered and should have been.
It would be too embarrassing for ATSB to look too closely at the assumptions now that the many millions have been spent.
You can download the DSTG book and read about the model. It contains very complex math, but it is possible to follow the concept of the model and its assumptions.
This is a very complex problem and the DSTG people did outstanding work, but I question how much push back anyone provided at the time the zone was first defined.
I sometimes think that government “lab rats” have huge egos and don’t take criticism well, but I really do not know anything about Dr. Gordon and the deliberations of the larger working group.
@Brian
“Hence my view: the aircraft will maintain a constant True heading, and the track will be subject to changes in the wind field.”
What you say certainly makes sense especially in the context of the reply you got from Honeywell. However, what Matt said resonates with me if I were leading the design on the AP. When LNAV is engaged a unique great circle route in generated between the start and end points. What I would do is simply maintain this route around the planet until instructed to do otherwise. That seems like the simplest possible implementation, and it too makes sense.
I took our host’s advice and used Google to find some other statements of mine, being what he calls an unSpoof. Many of us had severe reservations about the methodology in the DSTG report, and expressed it. In fact, in these posts on Reddit from a year ago, I predicted that the plane would not be found in the current search area, but not due to tampered satellite data.
***
The fact that “generic curved paths” would only be allowed if they were consistent with constant magnetic track/heading means there was a bias towards straight paths, which is what I said. In fact, even those magnetic track/heading paths were not allowed because to match the BTO would require a speed slower than the lower limit of M0.73 at FL350. The lower limit of M0.73 is completely indefensible because permissible fuel flow rates are possible at much lower true air speeds.
And yes, there are MANY solutions possible if you allow lower speeds and the possibility of curved paths, especially with a higher variation in the BFO. For instance, a circling pattern above Sumatra is possible, as is a descent at 18:40 while flying north with a later turn south. This would push the northern limits to 20S latitude or even further north, depending on the variations of track and speed that are allowed in the solution set.
The current search zone was considered the highest probability because at the time the most likely scenario was a straight, ghost flight. This also kept the search area to a practical size. But now that most of that zone has been searched with nothing found, the assumptions that led to this zone need to be re-visited, and constraints relaxed.
My point in criticizing the DSTG report was to show that we should have no increased confidence that the prior analyses were correct based on these new analytical results. So unless there is additional evidence available to the ATSB and not us (such as a positive but unreported scan), it is unlikely the plane will be found in the current search area.
And here lies the unspoken problem. If the plane is not found in the current search zone, it means the possible set of end points defines a region that is impractically large to search.
***
https://www.reddit.com/r/MH370/comments/3wjef2/response_to_the_atsb_report_and_dstg_book_on_mh370/cxxgz9d/?
***
I won’t advocate abandoning the search area, but I do agree that “more generalized methods would allow everyone including Iannello and Godfrey to admit no-one has a **** clue where it is or might be (to a practical searchable area size).” Now that the “single turn, straight path” solution has been searched unsuccessfully, anybody that claims to know the end point with any level of precision is a fool. That is my main criticism of the DSTG report.
https://www.reddit.com/r/MH370/comments/3wjef2/response_to_the_atsb_report_and_dstg_book_on_mh370/cxxh5fl/
@Dennis,
“maintain this route around the planet” . .
How does the aircraft know what the route is once the endpoint is not specified? On a Great Circle path [unless the path is a meridian] the [T] heading changes continuously.
Victori
Well said. I agree regarding the DSTG work.
It seemed that the approach used initially by Inmarsat did not depend on trajectory modeling, but this approach was dropped.
It seems that if you connect every “point” on arc 5 with every “point” on arc 6 that most of these line segments could not be possible. The transit time is fixed for each segment. So each segment has a velocity. Segments with a velocity greater than Mach 0.9 cruise could be killed. The shortest distance could be limited by some minimum Mach, but S-turns or a clearing turn could reduce effective speeds and allow for a direct perpendicular path. BFO allows for a velocity vector at each arc. I can see that possible paths could be constructed using this type of search logic. I do not know how Inmarsat did it but they did predict a likely splash point for the surface search. I do not know how their model compared with the DSTG zone.
@Brian Anderson: @DennisW is correct that an initial position and track will define the great circle path. The track will vary with position along the path. The question is whether this value is updated after a discontinuity or end of route. Nobody seems to really know the answer.
@Hank: The first efforts at path reconstruction did this very thing. Turns were only allowed at handshake times. A constant ground speed (and then later, constant Mach with estimated meteorological conditions) were used for each path. So depending on the speed range, a large variation in end points would result.
The IG and others argued that preference should be given to straight, fast paths because that is how pilots fly planes. I was in the camp that believed that this was a practical way to limit the search zone, and I supported the search in this area around 37.5S latitude. However, I also knew this search zone was based on an assumption, and therefore would be a crapshoot.
My main problem with the DSTG report was that I didn’t feel it was provided any better prediction than the straight, fast, deterministic predictions. I said it was written in a way to impress rather than illuminate. Once the search area in the immediate vicinity of 37.5S turned up nothing, I predicted the larger search area recommended in the DSTG report would fail.
In fairness, I also studied potential northern paths, and how the BFO might have been spoofed. (I won’t provide links unless you are curious.) And I was very suspicious about irregularities regarding the flaperon and the first find by Blaine Gibson. But after Blaine was able to repeat his ability to find debris and after the numerous other finds by others, I ranked the planting scenario as quite low. When the simulator data on the captain’s computer surfaced, in my mind, that was the final blow to any possibility of a spoof.
@Brian Anderson & @DennisW:
In my “path integrator” tool that I have used to reconstruct MH370 paths, I have a model to generate great circle paths. It is based on the following differential relationship, which is valid for ellipsoids:
d(alpha) = sin (phi) d(lambda)
where alpha is the track, phi is the latitude, and lambda is the longitude. So starting with a position (phi, lambda) and track (alpha), knowing the speed, you can create a small rhumb line segment at constant track and calculate the new position. You can then update the track based on the new phi and lambda. In the limit of infinitesimally small rhumb line segments, the great circle path generated in this way is exact.
That said, I don’t think the navigation system in a plane works this way because there is the potential for cumulative errors. A more likely way would be that a track is found and continuously updated based on the current position and the next waypoint.
For this reason, I don’t think a great circle path would be followed after a route discontinuity.
@VictorI
No, the big sims are booked back-to-back as far as the eye can see so I only had a few minutes to do each test (plus some takeoffs and landings, which I wasn’t about to deny myself in my fave big jet).
I’m impressed that MSFS modeled end of route correctly – so much so that I’m going to ask you to do another test. Since a mostly north-south flight isn’t a very good test of true vs great circle, when you have time, can you try this?
-Dial up ZERO WIND
-Load up a bunch of fuel and line up on RWY28R at KORD (bound for EGLL but the plane doesn’t know that)
-Leave the ND in MAG REF
-In the FMC input only
DPA
KORD
N45W83
(altitudes at your discretion, either VNAV or ALT, whatever you’re used to)
-After takeoff, DIRECT DPA in LNAV.
-Approaching DPA, go HDG SEL and make a manual left turn all the way around to a HDG of 052
-near the rollout to 052, go DIRECT KORD, then back to LNAV, overfly O’Hare and watch the next leg go magenta
(your MAG HDG after KORD should be about 052 [048T])
-speed up the sim until you’re almost at N45W83
-overfly it, triggering end-of-route / stay in LNAV.
(reaching this fix your heading will be about 59M/51T if you came direct from KORD)
-speed up the sim again (as fast as it will go)
(Of course, you can do any version of the same idea. I just tried to do some of the figuring for you – two fixes that draw a great circle route toward a distant third fix.)
If it’s a constant true track, you should draw a rhumb line right into the north pole. If you end up anywhere near Heathrow, it has to be a great circle route.
Look forward to hearing what happens! I’m pretty sure you’ll get back to me sooner than Honeywell does.
@TBill
No contradiction. Victor and I got the same results in both a desktop sim and a $20 million one. The thing stays in LNAV at end of route but flies a “MAG TRK.” If Victor can do the test I suggest above – in the absence of a reply from Honeywell – it’ll be the closest thing we have to an answer about whether or not it’s a true course or a great circle after end-of-route.
What I will say is that I’m more confident that MH370’s final leg was flown in LNAV with a distant fix in the FMC.
@Gysbreght
Just stop already. It’s like watching a train wreck.
@Matt Moriaty wrote
“After overflying the final waypoint entered in the LEGS page, the autopilot STAYS IN LNAV but flies a MAG TRK. (I’m talking to Honeywell to see if it’s simply a wind-corrected MAG HDG or if it’s an actual great-circle-route like everything else in the FMC.”
I was intrigued by this experiment as it does align with the latest ATSB report:
“The remaining three lateral control modes generally resulted in the aircraft’s flight path ending north of the CTT and LNAV results, up to 33°S in latitude along the 7th arc.”
So, chances are MH370 followed a pre-programmed flight path, and after the final waypoint continued in the way described above (no manual entering of further direction).
I feel this is consistent with a ghost flight proper, afer some intentional diversion (probably for nefarious reasons), followed by an extended period of no pilot input (for example, the perpetrator passed out in the meantime).
What I find astonishing, however, is the way the First Principles reports set out its reviewed search area. All information were known in early 2016 (or mid 2016) at latest, but the search area was never revised. On the contrary,
“Go Phoenix was tasked to search from 32.8°S to 34 .5°S in the northern section of the search area while the two Fugro vessels were tasked to search from 34.5°S to 39 °S in the south. In April 2015, GO Phoenix completed the contract with the Malaysian Government and departed the search area. At this time, the tripartite Governments of Australia, Malaysia, and the People’s Republic of China agreed to expand the search area to 120,000 km² . Deep tow search operations continued throughout the winter months with the two Fugro vessels.”
So, the “correct” search area was abandoned right at the time when the search area was doubled down?
Why did it not make sense to give equal consideration to both the southern and northern areas?
@VictorI
The great circle geometry defined by two points could, in principle, use the starting point for the new waypoint when the terminal point is reached. In that way you could ping pong around the earth constantly circling on the very same great circle route. Unless, of course, your starting and ending points are antipodal in which case there are an infinite number of great circles to pick from.
@Brian Anderson
I just now saw your post. And, yeah, I’m not hoping for too much from Honeywell, thus my request of Victor. It will categorically clear up how MSFS interprets the Honeywell system. What level of that validity that has IRL can only be answered by Honeywell. I can tell you that three guys with 40,000 hrs between them (one a captain and another his instructor) couldn’t say definitively either!
But to clarify what I said and what Dennis chimed in on – what we’re arguing is that the distant waypoint used by the FMC to chart the TRK could actually the last one overflown, just drawn completely around the globe on whatever MAG HDG was instantaneously observed after overflight.
Every route between two points in the FMC is a great circle route and this has been the case since the dawn of ground-based RNAV. That being the case, it does seem logical that any TRK drawn after end-of-route would also be great-circle, even if the next “waypoint” was actually the one you just overflew.
A constant TRUE heading of anything OTHER than 090 or 270 will eventually carve a loxodrome over the planet. True is useful above 70N where the compass is meaningless. But only when you also have either INS or GPS or both to reorient you at each fix, wherein you (in the old days) would adjust your heading to make the next fix. But in a long-term overflight situation (especially in jets that tend to go more east-west than they do north-south) doesn’t seem to be a very intuitive design choice.
But the lack of public knowledge – even among high-time ATPs – shows you how uncharted these waters are. (Was that a pun? I can never tell.)
@Victori
You have obviously followed the search zone modelling activities very well. Thanks for your insights.
To me the hijack and hide in the north scenario was far less plausible than than the captain suicide. I also became convinced by the discovery of the simulation data.
Thanks again for your information.
@Matt M. & @Brian A. and @DennisW: I’ll run some experiments possibly tomorrow to determine how FSX behaves at the end of a route. Whether or not that is useful we can sort out later.
@Brian Anderson
Sorry, “on whatever MAG TRK was instantaneously observed.”
@Victor
Great!
In the meantime, do we all agree that constant true tracks (other than the cardinal points) over great distances all end up in one pole or another?
@Brian Anderson
Funny that the Honeywell guy told you something that contradicts the manual. From p 4-34:
“Whenever LNAV is engaged and the aircraft enters a route discontinuity, DISCONTINUITY is displayed in the scratchpad, and the aircraft maintains its existing track.”
Your guy said: “…There is no “Track Hold” method associated with this type of scenario.”
Hmm…