Coverage from Flight Global

New search for missing AF447 to start in March

Investigators reveal extent of latest AF447 search effort

Metron report
 

Hi there

The Metron report give all the details about how the posterior distribution maps (posterior to the Phase III, to be used for the Phase IV, see fig 32 & 33 p 35) are estimated. It combines several prior distributions (linked to flight dynamics, the reverse drift analysis, the bathymetry, etc...) with posteriors (the several search phases) to build the Phase III posterior distribution.

I am a bit puzzled by several underlying assumptions:

-the FD (flight dynamics) distribution is built over 9 cases (quite low !) which are not all departing from FL 350, and this FD distribution aggregates cases with very different vertical speeds (from 12 kft/mn to 32 kft/mn): is the relatively good state of the recovered Galley G2 compatible with such high vertical speeds (20, 25, 30 kft/mn) ? are all the 9 cases compatible with the BEA findings about the attitude of the plane when it impacted the surface ("en ligne de vol") ?

-some prior distributions are combined in a fixed 30% versus 70% fashion, with little/subjective explanations:
"The prior distribution before surface search by aircraft and ships is taken to be a mixture of 70% of the FD Prior given in section 3.1 and 30% of the RD Prior given in section 3.2.2." (page 9, FD=Flight Dynamics, RD=Reverse Drift).
"Underwater Search Prior. For the purpose of evaluating the underwater search, we formed a prior which is a mixture of 30% of the Surface Search Posterior in Figure 14 and 70% of the FD Prior in Figure 2." (page 17)

This seems to explain why the final distribution used for Phase IV is mainly located on the LKP (in a small radius < 20 NM).
Jeff

ChristianJ wrote:

I would agree. Like any feedback servo, hydraulics are naturally rate limited in update terms by the inherent phase lags in the system. That is, the lag or delay between demand input and controlled entity reaching demand position. In the case of, for example, an aircraft rudder, this might be in the seconds range for full scale input. Because this is a rate of change limit, it means that increasing oscillatory demand input above the slew rate capability will result in ever decreasing amplitude at the control surface. I would think it unbelievable that any modern aircraft servo could have corners in the response plot that would permit oscillation at all. It's such fundamental part of servo design.

With a hydraulic failure, say a valve stuck so that the surface slews to one or other limit, all bets are off as the system is then running open loop. Istr, this sort of failure has been implicated a/c losses in the past, but forget the details. As for hydraulic capacity, I would expect any modern aircraft system to have a flow rate at least double that of worse case expected, but may be wrong there.

Have an old book on servo design that has a section on hydraulic servos:

Ahrendt, Servomechanism Practice, McGraw Hill, 1955. P 191. Old, but worth a read. The basic theory doesn't change...

Regards,

Chris

And based on the Metron report, the earlier searches may not have been entirely unsuccessful.

Position 3.628N / 30.504 W

Not so far from the normal route from ORARO to TASIL, and from the first debris (VS incl.) and bodies to be recovered after 6 days.
No dimension given for these manmade objects ?
Jeff

If I got it right, they are basically planning to do a search of 20nm around LKP.

That sounds like a plan !
From my point of view a very reasonable approach.

Seems to take into consideration that any travel much beyond 20nm after upset is not very likely, given the timeframe of < 5mins and the typical aircraft attitude / speed / flight path after a major upset.
Plus the fact that the Northern half of the circle has already been searched rather intensely and now extending to the South.

Somehow I'm pretty confident that they will find something....

If not there must be a story after 2:15, which is not really supported by the last ACARS message missing.

Let's cross fingers !!!

Hyperveloce;

The Prior using 70% FD and based on MAK/IAC data does seem rather misplaced. The descents used were, in the majority, of high rate and short period, and do not equate well to the facts associated with AF447 as we know them.

The Metron Report contains a good number of "assumptions" (IMO) and some I find rather difficult to swallow. Firstly, I noted that areas initially grid searched from the air by the FAB have been 'de-rated' for the reason that Metron were unable to obtain details on the manner in which the searches were conducted.

Bearing in mind that the BEA had commissioned Metron to undertake their analysis, it seems problematic that data essential to the work could not be accessed. I likewise have reservations about the initial air searches, and have posited my opinion to that effect in previous posts along the lines that the FAB's media arm often portrayed operations that didn't take place.

Some discussion takes place as to the survival of one or both ULBs based on the relative closeness of their individual locations. Metron then goes on to state,
Surely, that information was available to them and could have been provided by the BEA. There is nothing I have seen, through published information or otherwise, that leads me to believe that the ULBs suffered some gross 'g' forces the remainder of the aircraft didn't.

The comment regarding the SNA Emeraude's sonar is not helpful, and after other performances by the French Navy, is not unexpected.

So based simply on the fact that nothing was positively identified in all the areas searched around the LKP, the probability of the aircraft being closer to the LKP is the likely outcome of this analysis. You will note there is no bias in any direction, but just the red blob within 20NM of LKP.

Finally, its not my intention to rubbish the Metron Report, but rather to highlight some debatable points.

SaturnV;

I've got the distinct impression that images of possible debris will turn out to be nothing more than items jettisoned from a passing ship. The one on the left looks like it could be a heavy wire-rope strop containing spliced eyes at each end.

I would expect the sidescan sonar to reveal traces of the major structural items, hull, wings etc.. and jettisoned debris would then be mapped in the vicinity.

The following is a readable English translation of the article appearing in Le Figaro yesterday morning:-

After three previous search failures, this attempt is the last chance. Thierry Mariani, the Secretary of State for Transport, on Friday morning met with the families of victims of the Flight AF447 disaster and announced the launch of a Phase 4 search. Reportedly, a former Ifremer vessel, "Alucia" will soon depart from Seattle to Recife via Panama with two Remus Autonomous Underwater Vehicles (AUVs) onboard, leased to Woods Hole Oceanographic Institute. The ship should arrive in Recife, Brazil in mid-March and will uplift a third Remus AUV, leased from the German Geomar Institute, and members of the crew and investigators. The vessel will leave Recife at the end of March on the first of three voyages, each of a months duration.

This new search phase, at a cost of US$7.1 million, is funded by Air France and Airbus. This is less costly than the previous which had been US$11.5 million. If one believes the members of the investigation, this downgrading does not diminish the chances of finding the wreckage and black boxes. Quite the contrary. The earlier Phase 3 search led by the Office of Investigations and Analysis (BEA), had leased expensive equipment, in order to pick up items of debris, that have proved unsuitable and unusable in the conditions encountered. These devices were never out of the holds of the vessel due to the failure of the search. This time, the sole mission is to locate the debris field of the aircraft. If successful, a ROV such as Ifremer's Nautilus will be dispatched to undertake the underwater operations needed to recover pieces of wreckage and possibly the flight recorders of the aircraft.

For this last chance operation, the investigation methods have also changed. The Phase 3 search (spring 2010) had focused on an area defined by backtracking studies, i.e the analysis of drifting debris and recovered bodies to define the position of the aircraft. But once on site, investigators had found the real currents exceeded 1 knot, while the backtracking analysis was on the basis of theoretical currents of 0.1 to 0.2 knots.

This time, investigators have opted for the systematic combing of an area enclosed by a 40km radius of the last known position of the aircraft, offshore Brazil. The search should begin with the sweep of areas that have not been explored with certainty last year and then move further south. Many experts believe indeed that the search had focused on an area too far north.

When interviewed by Le Figaro, those close to the investigation said that deep water images taken in the previous search gave them confidence. The only possible cause of failure is that the aircraft is not in this area, e.g. it has turned around, or has continued its flight longer than expected.

Phase 4 - Search Schedule
 

Dates for the Phase 4 Search

2011-02-09 M/V ALUCIA DEPARTS SEATTLE
2011-03-14 M/V ALUCIA/WHOI TEAM ARRIVE BRAZIL
2011-03-18 DEPARTURE LEG 1
2011-04-23 RETURN LEG 1
2011-04-25 DEPARTURE LEG 2
2011-05-31 RETURN LEG 2
2011-06-02 DEPARTURE LEG 3
2011-07-08 RETURN LEG 3

Hi,

From the association AF447 victims site:
Google english translation
Google Vertaling

Original link:
Une 4ème phase de recherche en mer va débuter en février... | Technique aéronautique

syseng68k, I believe this is a non-starter unless you posit the pilot was trying to get a little more speed by really flapping the control surfaces far harder than they are designed to take. There's a fairly simple reason for this beyond the fact that the design engineers were very probably quite competent with servo design. Think about the hydraulic loads during landing and take off when the landing gear and flaps are being extended and retracted. Can anything in flight even in a storm equal that load on the hydraulic pumps?

Unless my impression that the landing gear and flaps are hydraulic motors is wrong the entire issue is not worth commenting upon it's so far fetched. Even if the motors are electric for these two functions the design for the servo system would account for worst case operations as you noted.

{^_^}

mm43
Agreed !
Most of the scenarios were more or less instant drop or spiral dive sometimes with mid air disintegration. AF447 took significantly longer than most of those cited.

However, assuming a major upset of AF447 it is still very unlikely that it traveled an effective distance of much more than 20nm IF it was in the water at 2:15.

quite a number of pages back I made a very rough calculation / guestimate / however you want to call it, assuming a rather 'benign' scenario to get the plane into an 'arrival' attitude matching the description of BEA.
That smooth scenario would cater for ~25nm assuming no change of flight path (which you btw. would not expect in a LoC situation).
That's why I still assume anything more than 20nm from LKP rather unlikely.

Hyperveloce, no dimensions. One of the objects to my eye looks like a piece of large, hard-sided luggage.

The position of these particular targets is just at the left edge of the search grid done by BR-6 on June 6. While the report does not say, I assume it was this plane flying this grid that was involved in the initial sightings. The search grids are given in more detail than the previous BEA reports, and this sector (with the targets) was not overflown until June 6. (If I have erred in this, I'm sure mm43, who is far more knowledgeable, will correct me on that.)

These targets 18, 19, 20, are on the left edge of a grid scanned by Orion (sonar) and Remus (photography on the last day of the search, and a month after Orion). To the left and north of these targets is a void that was not scanned (or scanned with any confidence), or photographed.

If I recall the debris field for AI 182 correctly, smaller items first, with items getting larger and more numerous as one proceeded along the track.

All the prior/posterior estimations based on the reverse drift analysis (including the initial aerial/naval searches, no matter the number of days from the accident) are uniformly/subjectively underweighted (versus the flight dynamics priors [*]) throughout the whole estimation process. This is why we get a distribution mainly located 20 NM around the LKP with a circular symetry (like the FD prior distribution), and with "holes" in it (linked to the absence of success of the Phases I if we assume the ULBs ere functional, II and III). The BEA seems on the same stance: too much uncertainties on the backdrifting studies: the searches could start from LKP and systematically probe the whole unsearched 40 NM disk region. The several targets of interest are really up north of LKP meaning far beyond the FD distribution for the targets 18, 19, 20, and if we have to assume that these targets could have been floating/drifting several days, I don't see how they could be major parts of the airframe.
JF
[*] FD distribution: in the Metron report it is a 50%-50% mixture between a uniform 40 NM radius disk and the 9 cases distribution.
- there are also large uncertainties about the AF 447 terminal trajectory & scenario after the loss of AP/ATHR: which pich & thrust mismatch or loss of control ? large roll unbalance ? Mach buffet ? stall ? how deep ? spin ? spiral ? adverse crew reaction ? late recovery ? secondary stall ?... this would militate for a uniform or circular symetric distribution.
- but there are some BEA initial findings about the AF 447 impact conditions that should be taken into account for the FD distribution:
flight duration starting from the AP/ATHR OFF between 3 min 45 sec and 4 min 15 hence outlying the entire 9 cases distribution (the 20 NM radius),
Galley G2 putting a limit on the vertical speed probably not compatible with spiral dives,
A/C attitude en "ligne de vol" and deformations of the rear frames (suggesting a significant horizontal speed) probably not compatible with a spiral dives or spin
- a limited number of compatible past cases to estimate a FD prior 2D-distribution (bad for the statistical significance of the estimated distribution), which is assumed with a circular symetry (meaning that a uniform/equal probability is allowed to any route deviation within [-180°,180°], -180° or 180° = U-turn)

kijangnim, I presume you are an authority on the attenuation of sound underwater? Would you care to tell us some more about hearing it from WELL below that 1000' level? What about interference patterns? What about simple square law signal attenuation with distance? What about non-linear properties of water? What about its dissipative characteristics for sound waves at the pinger's frequency? Please, do tell us more about what supports your bald assertion. I'm ready to read.

{^_^}

Hyperveloce, you exercised your human prerogative for handwaving there. How do you get the plane outside the 20nm radius without it going faster than anybody thought it was going or was capable of going and still have it go off the air in a fashion indicative of hitting the water within 4.25 minutes?

Actually a 20nm circle is a little off beam. Something more akin to a cardioid might be the correct figure to search within. The initial point for the cardioid would be last known position. Extrapolate what a plane could do in fully controlled flight to push the outside of the rest of the envelope. Then search everything inside that envelope. They are taking the "easy way out" by simply searching the full circle when there's lower probability at 20 nm backwards along their projected flight path. But there's rather low probability they could have pushed the plane past the 20nm forward point.

So let's discuss what the plane is doing to get it out past that 20nm and how it still managed to apparently meet the water or be in such bad condition it ceased transmissions in only about 4.25 minutes.

Please play some time line scenarios by us.

{^_^}

JD-EE
That would probably be the ideal approach, assuming that a 180° turn at cruise speed would already consume 5-10nm, thus reducing the effective range in the opposite direction to some extent.
So a cardioid ranging 25nm in flight direction and roughly 15nm in opposite direction from LKP would proably have the highest likelyhood of a hit.
But the concentric circle of 20nm seems to me a pragmatic approach with a very good probability compared to the -from my very humble point of view- somewhat odd search areas of the first three phases.
From my perspective thzey ignored the general Flight Dynamics to much and relied too much on the very unreliable (due to unpredictable relatively high speed currents) backdrift analysis.
So I remain confident this time.

I quite like this approach to explaining AF447....
 

a Reasonable Interim Conclusion: see link below=

Significant Regulatory & Related Activity|Aviation Safety Journal
.
.

Re: The link.
Hmmm, I read it and must say:
I'm not impressed.
At all.

OK, nice collection of the AD's but in the text so much gets mixed up, is unprecisely described (e.g. mixing up QF72 and the Perpignan A320).
I would say: skip it.

There have been MUCH better scenarios and descriptions been done in this Forum / Thread.

Hi there,
Sorry for the handwaving.
I am just a bit puzzled about the methodology to estimate the probability maps, but the good thing is that the entire unexplored regions will be systematically probed (this limits the impact of the Phase IV priors), whether an airliner can exit the 20 NM radius or not, whether the maximum probability to find the wreckage is on the LKP or not. Just waiting the effective search now.
JF

PJ2

Hi. Sorry for the delay, and thank you for the usual brilliant post. One thing, you mention a "disconnect" of AP does not degrade flight Law. I was of the impression an involuntary (auto) drop commanded Alternate Law? If Aero and accel considerations caused the A/P to quit, that does not trigger Law Change?? I have long believed the AP dropout was caused by its inability (programmed) to keep up?

bear

Bearfoil

Which-ever way you look at this "chicken & egg" situation at 0210z, I am reminded of the the favorable CRM analysis made of the THY crew involved in the B738 accident at AMS. The investigation confirmed that they in fact did everything correctly in a tight ATC controlled environment, but were badly let down by an unexpected automation condition not documented nor trained for. Unfortunately their epitaph acknowledges that, "The combination of training standards and experience is not always enough to protect crews from the subtle effects of automation failures during automated, human-monitored flight", and there may well be some relevance to the events we are dealing with here.

mm43

The quoted conclusion is a glaring example of the incongruity (and potentially deadly results) of a human crew flying with an automatic or fbw system they do not "understand". Inherent in this unfortunate statement by a regulator (of all things) is the mirage of safety due automation.

Pilots have been caught out with a bird's nest in air intake, a mud dauber's nest in pitot, or dual failures since the "day". There is and can be no excuse for introducing advanced control systems and then making an effort to "isolate" the crew from its operation.

What is the goal of automation? Pilotless flight? Fine. Say so. With 447, it would appear that a cascade of warnings and alerts of varying criticality deluged the dark and lumpy cockpit with distracts of almost sinister veil.

THY had alot of experience to hand, but the state of the "Art" caused a disconnect at the most critical of times, (Always the case??) It would seem that the most inopportune times are populated more heavily with the breakdown in "Awareness".

Will we find out if Captain DuBois' flight was besieged with a clumsy "reversion" in Flight Laws and Manual command?? One hopes the evidence will be found. That it is easy to entertain such a thought is worrisome. Were the PF's lulled into reliance on the AP for too long? Did they allow a dangerous drift into complete trust of the systems?? Would we use computers comparable to those aboard in our most mundane business tasks? Computers cannot think, they Act. Or Not. It is the Not part that gets people in trouble when "Novelties" are not trained. The ebb and flow of flight data and control command needs a new 'seamstress'.

Wasn't THY more a flight data reads issue? Unreliable airspeed syndrome is just that. There is no real Airspeed issue at all, only (sic) the air data displayed by the sensors. In 447's case, Pitch and Power should have saved the day. Duff airspeed read has been recovered thousands of times.

Reliance of the entire Airframe on flight controls that are operated by a system that doesn't have any idea what it is doing is regrettable, and here may be a case in point. A perfectly serviceable Aircraft and its people destroyed by Human Pride, Stubborn adherence to antiquated ideas, etc.

Ice? Upset? Mechanical Malfunction/Failure ?? Let us all hope there is a more thoughtful conclusion to this accident than the laughable and pathetic conclusion made by the authority in the case of THY Amsterdam.

regards,

bear

First, bearfoil, I see a lot of emotional assertions with no facts in evidence.

Second, put your money where your mouth is and design a flight control system that works your way for that airframe. It ain't that simple, McGee.

{^_^}

Bearfoil;
An involuntary autopilot disconnect does not in itself result in any degrading to alternate 1 or 2, or direct law.

However, some system abnormalities or failure conditions which do result in a degradation to alternate or direct law also result in an involuntary disconnection of the autopilot and autothrust systems.

The difference is illustrated below - the following summarizes both crew-actioned autopilot disconnections and involuntary disconnections which would not be associated with any degradation of flight control laws:

AP DISENGAGEMENT
AP1 or 2 disengages when :
– The pilot presses the takeover pushbutton on the sidestick.
– The pilot presses the corresponding AP pushbutton on the FCU.
– The pilot pushes on the sidestick harder than a certain threshold. (Disengagement through rudder
pedals is only active on ground).
– The other AP is engaged, except when localizer/glideslope modes are armed or engaged, or rollout or go-around mode is engaged.
– Both thrust levers are set above the MCT detent and the aircraft is on ground.
– In a non-precision approach, with FINAL APP mode engaged, the aircraft reaches :
• The MDA/MDH minus 50 feet (if entered), or 400 feet AGL (if no MDA/MDH is entered), or
• The Missed Approach Point (MAP), depending on which one comes first.
– One of the engagement conditions is lost.
– Furthermore, in normal law with all protections available, the AP will disconnect if :
• High speed protection is active.
• Angle-of-attack protection is active (α prot + 1° is reached).
• Pitch attitude exceeds 25° up, or 13° down, or bank angle exceeds 45°.
The standard way for the flight crew to disengage the AP is to press the takeover pushbutton on the sidestick.
When AP is OFF, the associated FCU pushbutton goes off, and the “AP1” (or AP2) is removed from the PFD's FMA.

The following is a table of results/outcomes from various system degradations which result in alternate laws 1 or 2 or direct law, and an involuntary disconnection of the autopilot & autothrust systems:

Depending on the type of failures affecting the flight control system, or peripherals, there are 3 possible reconfiguration levels :
– Alternate law (ALT 1 or ALT 2)
– Direct law, or
– Mechanical.

http://i277.photobucket.com/albums/k...06-07_1518.jpg

Thank you PJ2 for these tables.

One can notice that ADR DISAGREE does not entail A/P LOST condition.

Which means that under ADR DISAGREE conditions, A/P can be reengaged, presumably when the problem is over... Or almost over... See again the AD previously mentioned...

AP Reconnect and Prim And Sec Faults
 

I am having a hard time reconciling the AD scenario of an AP reconnect with the Prim1 and Sec1 faults.

1. It appears that the Prim1 and Sec1 faults would not cause a second disconnect?

2. Would it be a reasonable action to attempt a AP reconnect and command a Prim 1 and Sec 1 Shutdown?

3. If commanded, in what order would the FCPC and FCSC's be shut down, eg. Prim 1 then Sec 1 or Prim 1 then Prim 2, etc.?

Search Update
 

The dive support vessel "Alucia" departed Harbor Island, Seattle, WA at 10/1430z for Port Suape, Brazil via the Panama Canal. The Puget Sound pilot disembarked off Port Angeles at 10/2010z and the transit time to Balboa, Panama is 13 days 19 hours at 12 knots (3,975NM).
Edit :: Vessel anchored / slow maneuvers in Freshwater Bay, west of Port Angeles at 10/2112z. Why? Possibly testing dynamic positioning (DP) mode. Your guess is as good as mine!

Alucia is currently en route
 

mm43

Its seems to me that tests have been conducted previously between North Bay/Blue Ridge and Richmond Beach.

But now, what's the matter? What is succeeding in Freshwater Bay?
The vessel seems to be currently (Speed 0.3 kn, heading 130˚) coming back to Eliot Bay.

From the Deep Ocean Quest website:
Alucia just got out of a overhaul/refit, didn't it? Maybe they had an equipment casualty. It can take some time to bring a ship up to fully seaworthy condition after major rework.

Alucia
 

The video attached to this report says that they planned tests in Puget Sound prior to departure, and then expected to head for Brazil at 10PM (local 2011/2/9).

Seattle-Based Ship Sets Out To Find Air France Flight 447 - News Story - KIRO Seattle

If they are still in the Seattle area, then either they found something to fix, or the reporter got the details wrong (has that ever happened? :O)

Alucia
 

They are anchored in Freshwater Bay, and I suspect they are waiting for something "essential".

Edit ::
11/1918z 48.1479N 123.5933W (anchored - Freshwater Bay)
12/1754z 48.1480N 123.5933W

Ship Tracker
 

Alucia seems to be not showing on the Seattle tracking site.
Try this link to see where it is.
Alucia ? HO3651 ? position and weather

Alucia
 

Machinbird;

Thanks for the alternative link. AIS positions were being received intermittently while the vessel was anchored in Freshwater Bay, and that was probably to do with terrain shadowing.

She got underway at 13/0300z and crossed to the northern side of the Juan De Fuca Strait to get into the outbound traffic lane. Is now proceeding west at 10.5KTS

Edit :: WX prognosis is not good - SW 10/15KT becoming S 20/30KTS. Mod to heavy W swell, and by Monday (PST) the wind will be S 30/40KT.

Hello,

She seems to be currently en route (11kts / 213°)
48.4188
-125.0979

oil slick and initiation event
 

Whatever happened to the oilsick ? Was it ever taken seriously in choosing the search area ? Not saying it should have been but just wondering.

On a separate point, whats the current view on the condition of the aircraft at the time the first ACARS message went out: was this the start of the event or some point down the chain ? Can't recall the opinions about the possibility that the systems had previously been misled by erroneous (but consistent) inputs and the aircraft was suddenly left in a poor trim condition at the time the AP gave up.

Mr Optimistic;
It vanished, i.e. evaporated or dispersed.http://images.ibsrv.net/ibsrv/res/sr...ilies/evil.gif
It was discounted due to it being well outside the area of probability calculated by the Drift Group.http://images.ibsrv.net/ibsrv/res/sr...lies/worry.gif Though it does fall within the statistical probability area defined in the Metron Report, and the area will be covered in the current Phase 4 search.
Ops not normal, and 7 thousand or more posts on this and the AF447 thread later, the only satisfactory answer will (possibly) be revealed on recovery of the CVR and DFDR.http://images.ibsrv.net/ibsrv/res/sr...lies/wink2.gif

MM43
 

Thanks as always.

mm43

Yes, of course... but CVR/DFR solid state memory cards data are granted only thirty days at 16,000 feet deep water.

More than tweny months after the wreckage, how many probability % to recover recorded data?

A/P disengagement
 

It seems to me that the distinction between "DOUBLE ADR FAULT" and "ADR DISAGREE" is rather subtle, as both conditions usually occur together. As I understand it, the former results from two ADR's internally faulting (each ADR faults itself when it sees a sudden change in its output), and the latter from the PRIMs comparison of one ADR against the other(s).

Air Caraibes recorded the following explanation for A/P OFF from 'Airbus engineers' (my translation):
<q>The warning "AUTO FLT AP OFF" appears on the first line of the ECAM. The AUTO FLIGHT system compares continuously the values it receives from the three ADR's. If this cross-check is no longer possible after loss of two ADR's, it results in disconnect of FD and automatic pilot. As the AUTO FLIGHT system is conceived, any sudden variation of CAS, Mach, or altitude implies the rejection of one ADR. Thus, due to icing of PITOT and TAT probes, one sees on the two PFD, i.e. the ADR's 1 & 2, a very rapid reduction of these parameters which change respectively from 273 kt to 85 kt, M0.80 to M0.26 and from 35000 ft to 34700 ft within 1m25s. NOTE: thresholds for reject of one ADR: CAS 20 kt during 0.45s, Mach 0.04 during 0.45s, altitude 400 ft during 0.45s. <u/q>

See also Paragraph 1.6.11.2 in BEA's 2nd Interim Report on AF447.

EDIT:: Both explanations don't quite line up with the diagram posted by PJ2. If ADR DISAGREE does not cause A/P disengagement, then it was probably caused by DOUBLE ADR FAULT due to a rapid change of CAS and Mach in at least two ADR's.

regards,
HN39