I feel a little remiss—I forgot to dig through French investigative propaganda as I did with their NTSB counterparts concerning the 2009 air carrier stall-induced crashes. I had every intention to read through the BEA report on AF 447 thoroughly and report on my findings on the five year anniversary of the A330’s crash this past June 1st, but I got wrapped up with MH 370 and MH 17. Naturally, the resident historian beats me to it…
While reading the NTSB’s CO 3407 report for the five-year anniversary of the February 12, 2009 Colgan crash, I was deeply disturbed that there was no proximate source I could find for the August 2012 revamp the FAA made of air carrier stall recovery procedures:
This FAA Advisory Circular predates the Transport Canada document by a year. Yet AC 120-109 is from August 2012—three and a half years after the Colgan crash. I’m going to go out on a limb, and postulate the complete revamping of air carrier stall recovery procedures (more of a back to basics—after all, “lower the nose” is not news to pilots) wasn’t related to the investigation of CO 3407.
I needn’t have looked very far; the date on the BEA report says it all—“published July 2012.” Prior to the French report, the industry-standard in stalls was “minimize altitude loss.” The damned Colgan stall recovery procedure required their pilots to maintain altitude and heading. In essence, air carriers were permitted (even directed) to treat stalls as a non-event. One month after the BEA released their findings, the FAA quietly releases an Advisory Circular that finally gives stalls their due:
(D) Reducing AOA is the proper way to recover from a stall event. Pilots must accept that reducing the aeroplane’s AOA may often result in altitude loss. The amount of altitude loss will be affected by the aeroplane’s operational environment (e.g., entry altitude, aeroplane weight, density altitude, bank angle, aeroplane configuration, etc.). At high altitudes, stall recovery may require thousands of feet,
(E) Differences between high and low altitude stalls; pitch rate and sensitivity of flight controls, thrust available for recovery, and altitude loss,
(F) The need to apply nose down elevator inputs to reduce AOA when stalled at excessively low pitch attitudes and/or at large bank angles, including bank angles exceeding ninety degrees,
(G) Noises associated with stick shakers and autopilot disconnect aural alerts or alarms can cause confusion in the cockpit,
(H) Understanding that early recognition and return of the aeroplane to a controlled and safe state are the most important factors in recovering from stall events. Only after recovering to a safe manoeuvring speed and AOA should the pilot focus on establishing an assigned heading, altitude, and airspeed
AC-120-109 throws heading and altitude control during stalls out the window; there are now three rules: 1. AOA—less, 2. Airspeed—more, and 3. DON’T CRASH. This very sensible return to basics (every private pilot learns to break stalls with “lower the nose, add power”) is thanks to the smearing of an unfortunate (and very dead) French pilot.
…to Stall Smears…
To this very day, Junior First Officer Bonin is accused of pitching the A330’s nose skyward, inducing AF 447’s stall:
Under the nose-up inputs the flight attitude increases progressively beyond ten degrees and the airplane starts an ascending trajectory. The PF at this moment makes some nose-down inputs and alternately from left to right to counter the roll. The airplane climb speed which had reached more than 7,000 feet/minute goes down progressively to 700 feet/minute and the roll is still equivalent from left to right to around 10 degrees.
7,000 foot/min climb. In the past I flew the EMB-170, a fly-by-wire aircraft (albeit of Brazilian, not European manufacture), but have never seen a turbine-powered transport climb at 7,000 fpm at sea level, let alone FL350. I think I’m missing something:
Before disconnection, the autopilot maintained the aeroplane’s flight path by countering light to moderate turbulence; the autothrust had a slight reduction to adjust the cruise mach towards the value selected on the FCU of 0.80.
“Moderate turbulence,” which is defined as:
Turbulence that is similar to Light Turbulence but of greater intensity. Changes in altitude and/or attitude occur but the aircraft remains in positive control at all times. It usually causes variation in indicated speed. Report as Moderate Turbulence.”
So, the altitude and attitude were changing, and speed was decreasing. Upon autopilot (AP) disconnect, the A330 wasn’t likely to be trimmed out. Moreover, at a slower speed the aircraft must maintain a greater AOA, so the pitch trim was likely increasing. This was followed by “prolonged drops” in speed:
The first disturbances in speeds 1 and 2 occurred at about 2 h 10 min 04, causing the autopilot to disconnect, which was signalled by a visual and an aural (cavalry charge) warning. The crew did not necessarily perceive these transient losses of speed information and the associated losses of altitude.
The first prolonged drop (at least 5 seconds) in speed on the right-side PFD began not later than 2 h 10 min 07. It caused a drop in the altitude displayed on this PFD of approximately 330 ft. From 2 h 10 min 08, the speed became abnormal on the left.
When the autopilot disconnected, the roll angle increased in two seconds from 0 to
+8.4 degrees without any inputs on the sidesticks.
Wait, why is only roll mentioned? Why is there no mention of trim positions or other AP-commanded pitch and/or yaw? The NTSB listed such values for the BUF crash that occurred less than four months previously:
[B]ecause the autopilot altitude hold mode was engaged when the airplane leveled off at 2,300 feet, the autopilot continued to add nose-up pitch trim to maintain altitude as the airspeed slowed. During the time that the low-speed cue was in view, the airplane’s pitch trim increased from 1° to 7° nose up, and the pitch attitude of the airplane increased from 3° to 9° nose up.
The above aircraft experienced a major pitch upset seconds later, which American investigators claim could only be triggered by pilot input. I beg to differ. Excessive pitch trim coupled with a high power setting will send the nose climbing for the heavens.
But at least the NTSB explores the physics associated with the crash of CO 3407. The BEA report is so devoid of such analysis the 7,000 foot/min climb is not mentioned in the narrative—the initial press release was more informative.
…to Sweeping Stalls Under the Rug
An Air France A330 experienced a massive pitch upset on the night of June 1, 2009. Something caused that pitch upset. To French investigators, a human being’s hand naturally holds all the blame…it’s not like losing airspeed data can create immense confusion which would flabbergast the most seasoned of aviators:
1.18.6 Previous Accidents and Recommendations
Accidents with a relation to airspeed problems
Accident on 1st December 1974 to the Boeing 727 operated by Northwest Airlines
The aeroplane was scheduled to undertake flight 6231 between New York JFK, NY (United States) and Buffalo, NY. About 10 minutes after take-off, the crew noticed that the speed and the rate of climb were very high, respectively 405 kt and 6,500 ft/min. A little later the overspeed warning triggered, quickly followed by the stall warning (stickshaker). The crew attributed the stickshaker to the appearance of « Mach buffet » and tried to reduce the indicated speed. The aeroplane levelled off towards 24,800 ft and then stalled. It went into an uncontrolled spiral spin during which the stabilizer separated from the aeroplane. It struck the ground about 1 minute 20 after beginning its descent.
Accident on 6 February 1996 to the Boeing 757 operated by Birgenair
The aeroplane was scheduled to undertake flight 301 from Puerto Plata (Dominican Republic) to Frankfurt. During the takeoff run the Captain noticed that his speed display was not working. The copilot’s was working so he decided to continue the takeoff. During climb towards 4,700 ft the Captain’s speed display indicated 350 kt, which led the autopilot to increase the pitch attitude and the autothrottle to reduce thrust. The crew received « Mach airspeed » and « rudder ratio » warnings. The different speed displays and the simultaneous triggering of the overspeed and stall warnings (stickshaker) led to confusion in the cockpit. Noticing finally that the aeroplane was losing speed and altitude, the crew disconnected the autopilot and applied maximum thrust. A short time later, a GPWS warning sounded and the aircraft struck the sea a few seconds later.
Accident on 2 October 1996 to Boeing 757 operated by Aeroperu
The aeroplane was scheduled to undertake flight 603 from Lima (Peru) to Santiago (Chile). Immediately after takeoff the crew noticed that the altitude and speed displays were changing in an abnormal manner. They received a windshear warning, despite very calm weather and declared an emergency with the intention of returning to land at Lima. The aeroplane climbed up to a maximum of 13,000 ft and then began to descend. During the descent, the speed displayed to the Captain was so high that it triggered the overspeed warning even though the stall warning (stickshaker) was also active. The total confusion that ensued in the cockpit led the pilots to depend on the altitude indications given by the controller without realising that it was information supplied by the aeroplane itself in response to a radar signal which was thus false. After about 30 minutes of flight the aeroplane finally struck the sea off the coast of Lima.
These past accidents related to unreliable airspeed (I included all three that the BEA mention in their report–the above list is complete with no omissions) have several common elements. First, both stall and overspeed triggered in every case. Stall recovery requires dropping the nose and adding power. Overspeed avoidance pitches the nose up and reduces power. Well, which is it?
We know now, after hundreds have died, that the aircraft in question were stalling; but in the moment how were the (now dead) pilots supposed to ascertain which indication was unreliable? On three separate occasions (at least) the crews could not answer that question, and everyone onboard died as a result. It was a conundrum that probably could only be solved safe on the ground; not in the air with multiple, contradictory alarms blaring.
More importantly, unreliable data is…unsurprisingly, unreliable. The Aeroperu crew asking ATC for independent altitude data resulted in the erroneous onboard values being parroted back to them. Should have disabled their transponder and forced the controller to rely on primary radar, a solution the “experts” want to take away from pilots.
Overspeed receives just a cursory examination in the BEA report, but very well could have taken down AF 447 if the circumstances had matched the other three unreliable airspeed accidents. Recall Birgenair:
During climb towards 4,700 ft the Captain’s speed display indicated 350 kt, which led the autopilot to increase the pitch attitude and the autothrottle to reduce thrust.
Onboard automation pitches the nose up, and from personal experience in an overspeed situation the protective mode pitches up hard. The risk of supersonic airflow causing a subsonic airliner to pitch down or “tuck under” necessitates the overspeed mode’s abruptness. Mach tuck risk is great enough that overspeed protection usually will trigger when a plane is flown past VMO/MMO manually, just as stall protections don’t disengage when the AP deactivates (alternate law, however, is another matter). Come to think of it, an overspeed protection pitch-up would explain AF 447’s 7,000 foot/min climb, had the overspeed warning gone off simultaneously…
Wait, unreliable data stems from failures inside pilot tubes, static ports, and ADS probes; pressure-sensing devices common to all aircraft. How was this crashed Air France Airbus (an A330 that was manufactured in Toulouse, France) immune to the twin overspeed/stall conundrum that plagued the Northwest 727 and two Peruvian/Turkish 757s manufactured in Renton, Washington State…oh, right. The BEA is a French agency; I think I just answered my question.
The most important lesson to learn from the crash of AF 447 (and CO 3407) is the flight crew can only rely on their own instruments—they need at least one source of good data. Military aviation has had that source for decades, and that source might have been able to save 278 lives in 2009.
The Almighty Angle of Attack Indexer
Here’s how it works:
When the amber donut in the center of the indexer is the only indicator illuminated, you’re flying just the right speed for approach. This is called being on-speed. If you get too slow for the approach the green chevron will illuminate. The color and direction are important…it tells you exactly what you need to do to fix the situation. Green means go…add power. Adding power will flatten out your approach and reduce the angle of attack. Your angle of attack is high because you’re slow or your descent angle isn’t steep enough. The chevron points down…lower the nose. Lowering the nose will increase your rate of descent. Either action will cause the chevron to disappear in favor of the amber donut, but always remember that with a correction of pitch you’re going to need to adjust the power. The opposite is the red chevron. The red chevron points up. Pitch up. Red means stop…back off the power. In either case you’ll increase angle of attack, which is exactly what the indexer is telling you need to do. Keeping the amber donut illuminated will keep you on a nice slightly nose-high glide to the deck.
If Bonin and Renslow had an AOA chevron pointing downward, telling each pilot to mash the stick forward to the stops, neither man would likely be dead today. Modern turbine-powered transports already have AOA vanes; all pilots need is the information to be displayed in the cockpit. With today’s glass cockpits, adding an AOA indicator and indexer to the PFD should be child’s play.