News:

Precision Simulator update 10.180 (14 October 2024) is now available.
Navburo update 13 (23 November 2022) is now available.
NG FMC and More is released.

Main Menu

Air France jet missing over the Atlantic

Started by Phil Bunch, Mon, 1 Jun 2009 14:41

Peter Lang

I read in some german news, that Airbus adviced Air France to change the pitot tubes. But AF only did change the change those from A320 series.

Another interesting article In found here (also in german):
http://www.welt.de/vermischtes/article3853525/Experte-beklagt-absurde-Zustaende-im-Cockpit.html

Here it is said, that 2001 an A 330 ran out of fuel, due to a detached fuel pipe. They lost their complete fuel in 13 minutes. After this both engines failed and the A 330 glided within 19 minutes into the azores and could land safely.

Experts have been very frustrated that they could not find any records for 21 minutes on the flight data recorder. The electrical power also failed.

What would have happenend, if the azores did not come across incidentally?

Peter

Phil Bunch

Quote from: Peter LangI read in some german news, that Airbus adviced Air France to change the pitot tubes. But AF only did change the change those from A320 series.

Another interesting article In found here (also in german):
http://www.welt.de/vermischtes/article3853525/Experte-beklagt-absurde-Zustaende-im-Cockpit.html

Here it is said, that 2001 an A 330 ran out of fuel, due to a detached fuel pipe. They lost their complete fuel in 13 minutes. After this both engines failed and the A 330 glided within 19 minutes into the azores and could land safely.

Experts have been very frustrated that they could not find any records for 21 minutes on the flight data recorder. The electrical power also failed.

What would have happenend, if the azores did not come across incidentally?

Peter

Another update, supporting your post:

http://online.wsj.com/article/SB124428319732091447.html#printMode

Excerpt:

"Lost Plane Was Due for Change of Speed Sensors"



"Air France said in a statement Saturday that starting in May 2008, some of its A330s and similar A340 models "experienced incidents involving a loss of airspeed data" while cruising at high altitude. Air France and Airbus concluded the problems arose because the probes briefly iced-over, the airline said."

"The airline said it had begun replacing the units on the Airbus models in April, following tests indicating that a newer-model sensors would give fewer faulty airspeed readings at high altitude due to icing. Air France said it has now accelerated the replacement program and also reminded its pilots how to handle a loss of airspeed data."
Best wishes,

Phil Bunch

Jeroen Hoppenbrouwers

Somebody somewhere will now be thinking about a system that detects gross changes in indicated airspeed that do not match up with what the IRU says, and tries to work out a method to detect faulty pitot tube indications without endangering the stall warning system too much.

Peter Lang

Quote from: Phil BunchAir France said it has now accelerated the replacement program and also reminded its pilots how to handle a loss of airspeed data."

A question is in my mind since some days. We live in a time of INS and GPS. It should be possible to get altitude and groundspeed data from this. (Even the GPS in the cars can show groundspeed) The winds are also known. So why do all airspeed issues depend only from the pitot tubes? Ok, the pitot tubes are rather reliable, have haeting systems and are common use. But for safety or backup reasons it should be possible to get rather reliable data in case of pitot tube failure, at least to bring an airplane safely to an alternate. Especially at night time, where there is no visual reference.

It is not acceptable that computers get scared, when they have to handle different airspeed indications.

Just some thoughts
Peter

Qavion

QuoteThe winds are also known.

How?

Dominic Manzer

#65
Quote from: Peter LangIt is not acceptable that computers get scared, when they have to handle different airspeed indications.

I agree, It is very difficult to get programmers to screen input data for errors such as impossibly rapid changes or to resolve conflicting data without producing undesirable outputs. Programers often miss screening out data format problems such as text data where numeric should be. Format screening should be mandatory in any software project but I've found it missing in even high reliability code!

I have 2 questions.

1. Dose anybody have the complete ACARS data steam annotated with time and plane text descriptors? Some of the media seam to have it.

2. The cabin pressure alarm was described as "Cabin Rate" out of spec and could be triggered by fast pressure decreases or increases. Is it possible to trigger the cabin rate warning by a steep dive with the pressurization working normally?


[JH: edited the word "pressure" for funny spelling error]

Jeroen Hoppenbrouwers

It is not just programmers that often don't screen everything.

The issue with software development is that if you do not trust your input data, you have to write programs that are 90% paranoid checking of everything, and 10% getting work done. If your company or manager is under pressure to deliver a complex system in time and on budget, and you have to struggle to meet the specs of 10% of the code, why spending nine times more on checking that "should never be necessary"?

Of course it should happen, but the question is whether it is affordable. In aviation and other safety-critical software, there already is much more paranoia going on than elsewhere, but there is always a statistical and economical limit.


Jeroen

Peter Lang

Quote from: QavionHow?

Roughly from the weather charts and the data provided from other aircrafts flying along the ariways.
Enroute winds are also needed for fuel calculation. This is a part of flight preparation.

I admit that these data may vary ± 10 or perhaps ± 20 kts. But it should be eonough to get an airplane to the alternate in worst case scenaro.

Peter

PS:
It should also be possible to calculate the wind data roughly with these parameters:
- ground speed by GPS vs. INS / IRS data.
- engine data vs. AOA data.

Dominic Manzer

Quote from: Jeroen HoppenbrouwersIt is not just programmers that often don't screen everything.

Of course it should happen, but the question is whether it is affordable. In aviation and other safety-critical software, there already is much more paranoia going on than elsewhere, but there is always a statistical and economical limit.

Jeroen, I agree, lots of things don't get checked during design, but the trick is to limit the system's reactions to failures and unanticipated events too something that can be worked with. Back in the days of analog electronics, data inputs were bandwidth and range restricted to limit the effect up stream failures had even-though this significantly increased the size and weight of the circuitry.

Affordability is an interesting and difficult problem. Most of my experience is with spacecraft, the combined cost of the spacecraft and launch vehicle often are in the 100 to 200 million dollar range. Operating cost for 10 or 20 years can double overall program cost. I'm not sure of the cost of a new A330 but it's probably close to 100 million but this pails in-comparison to the liability that a crash can incur, approaching 1 Billion, ($1,000,000.)

In a well organized software development program, basic input and consistency checking is very cheep and can be built into the software development environment software. As proof: the ACARS system had no problem detecting the inconsistent airspeed data. The very procedure that Airbus recommends the pilots use to correct for a failed airspeed indication can be directly implemented in software to at least keep the aircraft stable until a definitive resolution is made by the pilots. As designers, maintainers and users of high-tech systems we should not except that exhaustive checking of simple routine things is too expensive, the computers we use to design them are very good at ensuring that routine checks are performed as long as you implement a well thought-out policy at the beginning of a program.


It's beginning to look like they entered and were unable to recover from a Hi altitude stall. If the pilots were not able to determine correct airspeed stall recovery could be impossible. Imagine having an overspeed stall when you think you are flying too slow! You keep the nose down and apply power, the situation rapidly gets worse. Or an underspeed stall when airspeed looks OK. With the turbulence of the storm, mild disorientation from the storm or the aircraft CG off centerline, the aircraft could easily end up on it's back witch is nearly unrecoverable with even good wether and a good aircraft. (No Failures) This is why I'm looking for the timeline and details of the pressure warning, these will quickly narrow the range of possibilities.

John Davis PC

A Bit more information about Black Boxes as discussed in the first part of the thread.

http://www.boreme.com/boreme/funny-2009/flight-box-recorders-p1.php

Cheers  PC

Qavion

Thanks, John.
Interesting.... On this type of DFDR, instead(?) of a water jacket, they use bicarbonate of soda!
They also show a magnetic tape reel type of CVR, but there are also solid state types with longer recording times.

QuoteI admit that these data may vary ± 10 or perhaps ± 20 kts. But it should be eonough to get an airplane to the alternate in worst case scenaro.

Peter

PS:
It should also be possible to calculate the wind data roughly with these parameters:
- ground speed by GPS vs. INS / IRS data.
- engine data vs. AOA data.


Not sure what you mean by "vs.", Peter. For example, GPS and IRS data is essentially the same kind of data: Groundspeed and position. No indicated airspeed data, which is essential to flight.

Don't know about the Airbus, but the engines also, normally, use the airplane's pitot/static system.
 AOA sensors are also susceptible to icing (that's why they are heated).

QuoteI admit that these data may vary ± 10 or perhaps ± 20 kts. But it should be eonough to get an airplane to the alternate in worst case scenaro.

I'm sure I can think of some much worse case scenarios... e.g. localised windshear, wind gusts of a hundred knots or more, etc, not experienced by aircraft flying along the same route a few minutes earlier (and I doubt that there would be aircraft flying all routes every few minutes).
+/-20kts could stall or overspeed the aircraft, especially at high altitudes.

Rgds.
Q>

Phil Bunch

http://online.wsj.com/article/SB124428319732091447.html#printMode

The above URL is the latest updated WSJ news story on the crash with some new details, etc.

A quote from the article follows:

"Modern jetliners have backup speed-measuring systems and pilots train for such situations, so damaged probes, by themselves, should not cause a plane to crash, industry officials say."

Is this true as a general statement for trans-oceanic airliners?  I thought there were no backup air speed subsystems except for multiple pitot tubes and their associated components.  Also,I thought that one's only indication of an air speed accuracy problem in the dark over an ocean at cruise altitude would be something like "air speed indicators disagree", or perhaps a grossly impossible speed indication, along with some alarms and error condition messages on the cockpit displays.

Finally, my aging memory tells me that the speed difference between nice stable cruise and a stall at cruise altitude is relatively small.  Is this qualitatively correct?  If so, then accurate air speed at cruise takes on special importance, of course.  It wouldn't be enough to be "sort of working" or "working but with substantially reduced accuracy".

Along this line of thought, I recall an incident in which an Air China crew more or less dozed off during cruise and failed to notice that an engine of their 747 had stopped making much thrust (perhaps it completely stopped - can't recall).  As a result, the other 3 engines and the autopilot tried to hold the requested cruise altitude as long as possible, but the airliner ultimately stalled out and went into a high-speed dive, finally waking up the crew.  They managed to rescue the diving, possibly supersonic aircraft before it hit the ocean, but considerable damage was done to the aircraft by excessive airspeed and mechanical stresses as I vaguely recall.  Fortunately, it was able to fly to the San Francisco airport destination.   I think "Captain Tarmack" described this incident in one of his columns or online forum posts.

With their vulnerability to everything from icing to insects to protective tape covers being left on the pitot tube ports, one would like to have a backup system such as GPS-based speed available to the crew.  This wishful thinking of course would only make sense if it is airworthy in a formal, strict sense through rigorous engineering analysis and testing, none of which I can contribute to even for informal discussion purposes.
Best wishes,

Phil Bunch

Jeroen Hoppenbrouwers

#72
If you max out your cruise altitude, you indeed are left with about 10 knots between overspeed buffet and stall speed. On the 744, the red/yellow bands would nearly touch.

This is one of the reasons that in case of expected or experienced turbulence, you should not be near your maximum altitude. One wind gust and off you go.

Air speed or Mach here is the absolute key. GPS or IRS speed does not tell you anything as your wings don't care at all about anything but (indicated) air speed/Mach. In stable conditions away from the Coffin Corner you may be able to fall back onto ground speed, but at the edge of the envelope it's useless.

Qavion

Quote"Modern jetliners have backup speed-measuring systems and pilots train for such situations, so damaged probes, by themselves, should not cause a plane to crash, industry officials say."

They are probably talking about Standby Instruments (although these sometimes use the same pitot/static sensors). AOA & thrust coupling is another possibility for maintaining airspeed, but I get the impression that pilots don't often get the opportunity to practise this kind of thing.
There have been lots of instances where pilots haven't been able to identify the faulty component/s, so they use the wrong instruments.

Most aircraft these days have groundspeed displays based on IRS or GPS, but as Jeroen says, they don't tell you what your airspeed is.

Dominic Manzer

To understand how air data can be so disastrously wrong, one has to know why it is important and how the instruments work. I'll try to give a brief discussion of each, but brief is not easy.

Air speed is critical to most phases of flight. Take off rotation occurs just before stall speed. Best performance climb out occurs just above stall speed. G forces wile turning at lower speeds require nearly stall lift from the wings. High altitude flight is conducted in what is called the coffin corner of the flight envelope. Here stall speed increases due to the thinning air to nearly the speed of sound. Turns at high altitude increase the wing loading when there is little margin for increase. The only 2 phases of flight you don't need to worry about air speed are (1) mid altitude moderate speed straight and level flight and (2) parked on the tarmac.

Air speed can only be computed from measurements of the air being flown through. From memory, air data systems can provide 4 measurements of use to the pilot, barometric altitude, air speed, stall warning and angle of attack. Please let me know of I missed any. Most air data measurements are dependent on an accurate determination of the static (non-moving) pressure of the air the plane is flying through and is supplied by a static source.

The simplest static source is holes on the sides of the pitot tube that are perpendicular to the airflow. Clearly a pitot tube static source can be corrupted by the same problems as the pitot dynamic pressure (air speed), such as ice, blocked lines, insects and being covered. On most aircraft, If the static sources are corrupted, all the air data information is wrong. A few types of aircraft, use different forms of static sources and air data ports, such as carefully placed holes in the aircraft skin. These systems have to be carefully calibrated in flight test and are even more sensitive to contamination than pitot tubes. They are used mostly on military planes that have problems with pitot tubes, such as supersonic flight and stealth.

Barometric Altitude is the most fundamental air data presented to the pilot in the form of the altimeter. Barometric altitude is calculated by subtracting the static pressure from a reference pressure and multiplying by a constant to convert the pressure difference to feet of altitude. At low altitude, the reference pressure is the barometric pressure at the airport of interest. Taking off, the reference is set by adjusting the altimeter to the actual elevation of the runway. Landing, the pilot has to obtain a barometric pressure measured on the ground and manually input that pressure into the altimeter.

The Blue Angels once crashed four or five planes at once because they violated their air-show barometric altitude procedure. The acrobatic procedures are written in altitude above ground level (AGL) so the procedure dose not need to be rewritten (and relearned) for each airport. But this requires that the altimeter be set to sea level before takeoff. Most of their practices and air-shows are at airports close to sea level making altitude above sea level (ASL) and AGL nearly identical. Before takeoff, they set the altimeters to the altitude of the runway as you would for a normal flight instead of sea-level. Apparently this group of pilots had been making this mistake for sometime and got away with it because the error was less than the built in safety margin. They did a loop, and toped out 1000 feet to low because the ground was 1000 feet higher normal.

For high altitude flight the altimeter is set to "standard pressure" so that all aircraft will fly at the same altitude when commanded to a specific flight level. The actual distance above the ground of a specific flight level varies by several hundred feet with the wether from day to day but also is different at different places at the same time.

Air speed is calculated by subtracting the static pleasure from the pitot dynamic pressure then converting to velocity. Air speed is fundamental for nearly all phases of flight and most critical at landing and high altitudes. Before GPS and other radio navigation adds it was necessary to guess what the winds were to estimate ground speed for navigation. A friend was taking flying lessons and the instructor asked him what he thought their ground speed was. He knew the air speed was 75 knots and they had a 15 knot head wind at takeoff. Thinking the wind would be higher at 3000 feet he said 50 guessing the wind was 25. The instructor said, "Look down out the side window." They were over the Chesapeake bay bridge going backwards! The head wind was about 100 knots. The plane could not fly faster so they had to fly at low level to find lower winds to get back to the airport.

Stall warning is determined by a stall warning pitot that faces backwards on the upper rear wing surface. It looks for air recirculating (moving forward) over the wing as the stall vortex begins. It may use the primary static source or use it's own static source.  At low speeds it measures the velocity the air is moving over the wing and produces the stall warning when the air flow reverses (moves forward) impinging on the dynamic port of the probe changing the dynamic pressure from lower than to higher than the static pressure. On most small planes air moves from the dynamic port through a reed buzzer and out the static port. The sound gets louder and higher pitched as the stall deepens. This systems requires a dedicated static source as the volume of airflow through the buzzer would corrupt the static reference for other uses. At high altitudes the calibration of the stall warning may need to be changed and could be different for both overspeed and underspeed. In high altitude stalls, the air flow still separates form the wing causing reduced lift, but the flow direction may not change. I don't know how stall warning is implemented on large jets and is probably dependent on the aerodynamics of the particular wing design. More information is welcome.

Angle of Attack (AOA) is not normally displayed to the pilot even in aircraft that are equipped with it. Test pilots who have used AOA love it and want it an all aircraft. It provides a better indication of stall than traditional stall warning, and can be useful in other phases of flight. Again, the stall AOA may be different for high altitude stalls than low altitude. The beauty of AOA is that stall is always at the same AOL regardless of airspeed, weight, air density or other variables. Angle of attack can be used as a proxy for air speed as the relationship with airspeed and weight is constant. The simplest way to measure angle of attack is with a horizontal wind vain in the airflow. But the sensor must be held a considerable distance from the airframe to get clean air and is thus not practical for operational aircraft. All modern flight control systems compute angle of attack (because it is such a good measure of performance) form traditional air data sources. So, if the air speed is corrupted so is the AOA computed from the same air data sources.


Other methods of determining airspeed. I can can only think of 2 that will work, Doppler radar and acoustic distance measurement. There may be one or two other methods but not many more.

Acoustic distance measurement. It is not obvious why this works, but it does. By placing acoustic transducers across a horizontal space open to air flow continuous measurements of the distance can be made. (Say between two engines). When stopped all this tells you is the distance between the engines, but when the air is moving relative to the plane, the air pushes the sound waves rearward relative to the aircraft. To be picked up by the receiving transducer the sound has to travel diagonally relative to the still air. This diagonal motion increases both the time and distance the sound travels between the two transducers (the receiving transducer has moved forward during the transit time of the sound.) This method works well and is in use for measuring fluid flows in pipes and ducts.  A horizontal gap should work on aircraft up too about Mac 0.95. Reconfiguring to a longitudinal (length wise) gap will produce 2 different measurements, 1 for forward and 1 for rearward. Producing 2 different times for the same pair of transducers provides a nice consistency check. As speed increases, the forward measurement will take longer and signal strength will be reduced. But the rearward measurement will take less time and increase the signal strength as speed increases, taking 1/2 as long at the speed of sound.

Doppler radar. This method should already be available from Doppler wether radars all ready on the marked. Doppler radar not only measures the strength of reflected radar energy from rain and ice it also measures the velocity that the particles are moving relative to the aircraft by measuring the shift in frequency of the reflected energy. This is the same method used by police radar for traffic speed. In police moving radar (in car), the receiver uses the return energy from the stationary ground to calculate the speed the radar is moving and uses that speed to calculate the ground speed of other vehicles. Airborne Doppler wether radar has to do the same thing, measure the frequency of the ground return to determine the aircraft ground speed. The frequency shift of the wether return is the relative difference in wind speed and aircraft speed so the aircraft ground speed must be subtracted to produce a usable wind ground speed wether map. The raw wether return frequency shift is the airspeed of the aircraft when the return is from air immediately in-front of the plane. The raw relative speed map shows what the airspeed will be when the aircraft arrives at a particular point if current conditions persist. This is the only technology currently available that can give a true airspeed without using air pressure ports. The angle of attack can be read directly from the radar by looking for the angle of maximum airspeed.

A 3D Doppler Radar is capable of providing nearly all the information that aircrews use to fly. It can produce air speed, and AOA. The combination of air speed and AOA provides stall warning. The ground return provides altitude above ground level (AGL), rather than barometric altitude, also it is the AGL in-front of the aircraft, rather than under it. But when landing, this will produce the altitude above the runway, a more useable reading than above sea level.

Qavion

QuoteFrom memory, air data systems can provide 4 measurements of use to the pilot, barometric altitude, air speed, stall warning and angle of attack.

... vertical speed, total air temperature, static air temperature, mach, TAS, impact pressure for engine fuel control, etc....

AOA vanes are used on all Boeings and the 737NG even has an optional AOA indication on the primary flight display. AOA vanes are the primary reference for computing stall speeds on the 744.

QuoteA few types of aircraft, use different forms of static sources and air data ports, such as carefully placed holes in the aircraft skin.

I guess you could say Boeings also have different forms of static sources. Some are located on the pitot probes, others are on the fuselage.

Cheers.
Q>

Peter Lang

Quote from: QavionI'm sure I can think of some much worse case scenarios... e.g. localised windshear, wind gusts of a hundred knots or more, etc, not experienced by aircraft flying along the same route a few minutes earlier (and I doubt that there would be aircraft flying all routes every few minutes).
+/-20kts could stall or overspeed the aircraft, especially at high altitudes.

Hi Qavion,

worst case scenario in this case was meant as no reliable airspeed indication due to loss of all pitot tubes. If you have gusts of 100 kts and more in FL 390 you will be in trouble if your pitot tubes are working or not.

For the GPS or INS backup: I think it should be possible to say that in a certain altitude with ISA conditions at a certain N1 a rather exact IAS, Mach, TAS and GS (with zero wind) is achieved. For the 747 e.g. cruise tables are in the manual in which you can see what IAS, MACH and TAS, N1 and FF parameters you have in a specified FL and grossweight range.

These figures of course differ due to weather conditions. A change of temperature will change IAS / Mach relation (and as I understand your comments also N1 and FF) and a change of wind changes TAS / GS relation.

Reversely it should be possible to calculate wind components from the existing parameters, at least for a short period of time which should be sufficient to descent to a safer more uncritical FL (perhaps the altitude in which an aircraft must descent in case of loss of cabin pressure)

The GPS knows the actual GS and TRK and FL. I'm not sure if the GPS also knows the HDG. But the INS/IRS should know. These data compared with the "should have" GS and HDG from the last known conditions gives you the wind component.

It is clear that this will not be an exact calculation. It only can be an approximate calculation. And its also not meant to conduct an entire flight this way. These thougts just came up to minimize the chance of accidents as we probably had on AF 447.

Peter

Hardy Heinlin

#77
Hi Dominic.

Quote from: Dominic ManzerThe Blue Angels once crashed four or five planes at once because they violated their air-show barometric altitude procedure. The acrobatic procedures are written in altitude above ground level (AGL) so the procedure dose not need to be rewritten (and relearned) for each airport. But this requires that the altimeter be set to sea level before takeoff.
I assume this is a typo and should read "airport elevation" instead of "sea level"?

Regards,

|-|ardy


Edit: Any my typo is: It shouldn't read airport elevation, but airport pressure. Pascal, not feet.

Dominic Manzer

Quote from: Hardy HeinlinHi Dominic.

Quote from: Dominic ManzerThe Blue Angels once crashed four or five planes at once because they violated their air-show barometric altitude procedure. The acrobatic procedures are written in altitude above ground level (AGL) so the procedure dose not need to be rewritten (and relearned) for each airport. But this requires that the altimeter be set to sea level before takeoff.
I assume this is a typo and should read "airport elevation" instead of "sea level"?

Regards,

|-|ardy
You made my point, by crashing. You made the same mistake they did! If you need say 1400 ft to complete a maneuver and start the maneuver when your altimeter says 1500 ASL, as the Blue Angles did, you will be OK as long as the ground elevation is less than 100 ft. If the ground elevation is 1000 ft, you started the maneuver at 500 ft AGL, have 900 ft too little altitude and are in big trouble.

Knowing how things work and why is essential when something unusual happens or is done. In the Blue Angels' case they memorized the procedure altitudes in AGL. The procedure is written in AGL so they hit the same altitude marks every-time, but the altimeter must be set to read AGL. Flying the routine in above Sea Level measurement at a new ground level airport every week is impossible, as the crash showed.

Hardy Heinlin

#79
I guess we're having a lingual rather than a technical misunderstanding here.

I'm saying:

If I want my baro altimeter to agree with the AGL-altitude printed on the chart (while flying at that AGL-altitude), I have to set my baro not to sea level (QNH), but so that it reads zero when I'm parking at the airport (QFE).

When I would be overflying La Paz at 1000 feet AGL and my baro would be set to sea level (QNH), my altimeter would read something around 14000! Not 1000.

And when I would land in La Paz with QNH set instead of QFE, my altimeter would read ca. 13000, not 0.

Don't say you disagree :-)

|-|ardy


That lingual misunderstanding may lie in in the word "this":
QuoteThe acrobatic procedures are written in altitude above ground level (AGL) so the procedure dose not need to be rewritten (and relearned) for each airport. But this requires that the altimeter be set to sea level before takeoff.
What is the word "this" refering to? To procedures written in AGL? Or to procedures rewritten in QNH?