Kicking Tin: FDX 80
It has been just over a year since FDX 80 crashed at Japan’s Narita IAP while landing (I commented about it here and here). The Japanese Transportation Safety Board recently released a progress report containing factual information (reported here, but no conclusions.
Relying upon time lapse images and digital flight data recorder (DFDR) plots, I am going to take a whack at being a crash investigator and attempt to determine the precise cause for the crash.
Warning: Acres of dry specialist writing follow.
The mishap aircraft was an MD11. The weather was clear, but windy: 320 degrees at 23 knots gusting to 34 (25 to 38 mph). Since the aircraft was landing on RWY 34L, the wind components were 22 − 33 head and 8 − 12 left cross, well within aircraft limits.
There were no equipment problems. The aircrew had complied with all standard operating procedures. The First Officer (FO) was the pilot flying (PF).
The PF disconnected the autopilot, which was using the instrument landing system for vertical and horizontal guidance at 212’ above ground level (AGL). The autothrottle system remained engaged throughout the mishap sequence.
The approach met stabilized approach criteria, and was normal down to 50’ AGL. The composite below is a sequence of five images taken over the two seconds preceding touchdown. Times are local. The first in the sequence shows the aircraft at 20’ AGL.
This is the last image of that sequence, to make its touchdown attitude visible:
Approximately one second later, the aircraft has all landing gear on the runway, and the spoilers are starting to deploy:
Within two seconds after touchdown, the aircraft becomes airborne, reaching 16’ AGL:
As it comes back down, the aircraft pitches over, impacting first on the NLG, then on the left MLG
This impact overloaded the left wing, which fractured just inboard of the left MLG.
Since the aircraft still has flying airspeed, the lift differential resulting from the still intact right wing flips the aircraft on its back, and off the left side of the runway.
There is the what. The problems remaining, increasing in difficulty, are: how, within eight seconds, how did this approach go from normal to disaster; and why did the how happen, so as to avoid further repetition?
To answer the “how” question, the DFDR is essential. (In the following analysis, for clarity I have clipped portions of the DFDR depiction I linked to above.)
First, some decipheratory notes:
From the top, this shows:
The essentially constant slope of the RA line up to one second before touchdown shows several things. It was consistent with a normal descent rate for the 3 degree ILS glideslope and a ground speed of roughly 130 knots (approach airspeed minus headwind component). It also shows that the airplane did not follow the prescribed landing profile: At 50’ AGL, shift the aimpoint down the runway to break the descent rate, then starting at 30’, establish the landing pitch attitude of 7.5 degrees.
As a consequence, when the MLG touched down (at 2 below), the airplane’s descent rate was essentially unchanged. The impact was 1.67G — roughly what an airplane experiences landing on an aircraft carrier. Further, because of the vertical speed, and the center of gravity being ahead of the MLG, the remaining vertical component of kinetic energy de-rotated the airplane, causing the spike loading of 2.25G (at 4 below) when the NLG hit the pavement. Between 4 and 7, the entire airframe is “ringing”, which is causing the roughly 2.5 cycle per second oscillations in the trace. In this interval, the airplane has bounced back into the air.
If the film stopped here, the end of the story would have been an embarrassing landing, but not quite so abrupt as to require maintenance to perform a hard-landing inspection.
Unfortunately, the film continues to roll, and the aircraft hits the runway again, at 7, this time as if dropped from 16 feet.
Airplanes are not designed for this.
That is the how, but it doesn’t get us to why.
Warning: speculation follows.
At about 20’ AGL the pilot initiates the pitch attitude increase (directly above the “P” in “[Yoke] Pos (deg)” below) that should have happened at 50 feet. However, since the aircraft cannot respond instantaneously, the pitch attitude does not start to change until about a quarter second later: the action and reaction are slightly out of phase. That is inevitable; bad things happen when the rate of pitch inputs starts to pump the natural oscillatory frequency of the combined control-fuselage-wing system. Think of a spring whose combination of resilience and mass will cause it to bounce at one cycle every two and a half seconds (My memory banks are dragging that up for a typical airliner; flight manuals tend to be silent on the subject.)
At the first impact, the back pressure is relaxed, but as the nose is rapidly falling through, the yoke comes back again.
This is the onset of a pilot induced oscillation (PIO), where the control inputs and aircraft responses (bottom two traces) quickly, and increasingly diverge. Eight seconds from the initiating pitch input, control and response are completely out of phase with each other.I had a student do this to me in a previous life, under similar circumstances. It is amazing how quickly things can get completely out of hand.
While I am fairly confident that the mishap board will conclude there was a PIO, there is a tiger lurking in the shadows. The MD11 has nearly cornered the market on this kind of mishap. Googling [md11 “hard landing”] confirms this plane has a reputation unlike any other.
I think it boils down to a single design issue. The picture below (taken from the Wikipedia MD11 article) shows the MD11, and the DC10 from which it derives.
Even to the uninitiated, the horizontal stabilizer is conspicuously smaller than on the DC10. Additionally, since the MD11 fuselage is longer, the mass is distributed further from the center of gravity, meaning it has a higher polar moment of inertia; that is, more force is required to move the airplane in the pitch axis. Yet because the wing — which is what the horizontal stabilizer effectively is — is smaller, it cannot generate as much lift, which is what the pilots are using to move the nose up and down.
Smoothly controlled, that is unnoticeable. However, it seems certain that a combination of design characteristics makes the MD11 less well behaved when pushed, and more inclined to PIOs. All designs are compromises. The MD11 high altitude cruise, which, it seems, came at some expense to low speed handling.
Yes, but. The crash didn’t happen until 06:28:25, five seconds after initial touchdown, and two seconds before the final impact. Until that moment, application of go around thrust gets the crew another approach and landing, as well as greatly extending the rest of their lives. The same applies to nearly all the other MD11 landing mishaps.
Over the last year, there have been significant changes in MD11 landing training in the simulator that are, so far as I know, unique to the type.
We still have to figure out how to get the wetware to make a timely decision to take it around.
Relying upon time lapse images and digital flight data recorder (DFDR) plots, I am going to take a whack at being a crash investigator and attempt to determine the precise cause for the crash.
Warning: Acres of dry specialist writing follow.
The mishap aircraft was an MD11. The weather was clear, but windy: 320 degrees at 23 knots gusting to 34 (25 to 38 mph). Since the aircraft was landing on RWY 34L, the wind components were 22 − 33 head and 8 − 12 left cross, well within aircraft limits.
There were no equipment problems. The aircrew had complied with all standard operating procedures. The First Officer (FO) was the pilot flying (PF).
The PF disconnected the autopilot, which was using the instrument landing system for vertical and horizontal guidance at 212’ above ground level (AGL). The autothrottle system remained engaged throughout the mishap sequence.
The approach met stabilized approach criteria, and was normal down to 50’ AGL. The composite below is a sequence of five images taken over the two seconds preceding touchdown. Times are local. The first in the sequence shows the aircraft at 20’ AGL.
This is the last image of that sequence, to make its touchdown attitude visible:
Approximately one second later, the aircraft has all landing gear on the runway, and the spoilers are starting to deploy:
Within two seconds after touchdown, the aircraft becomes airborne, reaching 16’ AGL:
As it comes back down, the aircraft pitches over, impacting first on the NLG, then on the left MLG
This impact overloaded the left wing, which fractured just inboard of the left MLG.
Since the aircraft still has flying airspeed, the lift differential resulting from the still intact right wing flips the aircraft on its back, and off the left side of the runway.
There is the what. The problems remaining, increasing in difficulty, are: how, within eight seconds, how did this approach go from normal to disaster; and why did the how happen, so as to avoid further repetition?
To answer the “how” question, the DFDR is essential. (In the following analysis, for clarity I have clipped portions of the DFDR depiction I linked to above.)
First, some decipheratory notes:
- The vertical grid lines are time in two second intervals; the left edge is 10 seconds before touchdown.
- As is standard for DFDR depictions, the traces are jammed together in order to make the time relationship between different data types easier to see; however, it does come at the expense of readability.
- The encircled numbers visible in some DFDR extracts refer to image numbers.
- The leftmost dashed vertical line is when the MLG touches down, second is NLG, and the solid vertical line is the terminal NLG impact
From the top, this shows:
- Calibrated airspeed, which varied from 156 to 163 knots (180 − 188 mph) over the approach’s last ten seconds. Under the mishap flight conditions, calibrated and indicated (what the pilots see) airspeeds are essentially identical.
- Engine power. The auto throttle system is increasing power from ten to seven seconds prior to touchdown in order to capture the final approach airspeed of 160 knots, then starts pulling power to maintain airspeed until 5 seconds prior to touchdown, where it initiates a pre-programmed retard to starting at 50’ AGL.
- Radar Altitude (RA), the distance between the airplane and the ground directly beneath it. It decreases nearly linearly from 105’ AGL ten seconds before touchdown, for an average descent rate of 630 fpm. Note a DFDR limitation: it is recording RA at one second intervals, which means RA alone cannot be used to calculate vertical speed at touchdown. This is the sort of thing 9/11 Truthers, for reasons of mental disease or defect, can never take on board.
- The first image in the composite photo above corresponds to the right parens in “(ft)” in this DFDR extract.
The essentially constant slope of the RA line up to one second before touchdown shows several things. It was consistent with a normal descent rate for the 3 degree ILS glideslope and a ground speed of roughly 130 knots (approach airspeed minus headwind component). It also shows that the airplane did not follow the prescribed landing profile: At 50’ AGL, shift the aimpoint down the runway to break the descent rate, then starting at 30’, establish the landing pitch attitude of 7.5 degrees.
As a consequence, when the MLG touched down (at 2 below), the airplane’s descent rate was essentially unchanged. The impact was 1.67G — roughly what an airplane experiences landing on an aircraft carrier. Further, because of the vertical speed, and the center of gravity being ahead of the MLG, the remaining vertical component of kinetic energy de-rotated the airplane, causing the spike loading of 2.25G (at 4 below) when the NLG hit the pavement. Between 4 and 7, the entire airframe is “ringing”, which is causing the roughly 2.5 cycle per second oscillations in the trace. In this interval, the airplane has bounced back into the air.
If the film stopped here, the end of the story would have been an embarrassing landing, but not quite so abrupt as to require maintenance to perform a hard-landing inspection.
Unfortunately, the film continues to roll, and the aircraft hits the runway again, at 7, this time as if dropped from 16 feet.
Airplanes are not designed for this.
That is the how, but it doesn’t get us to why.
Warning: speculation follows.
At about 20’ AGL the pilot initiates the pitch attitude increase (directly above the “P” in “[Yoke] Pos (deg)” below) that should have happened at 50 feet. However, since the aircraft cannot respond instantaneously, the pitch attitude does not start to change until about a quarter second later: the action and reaction are slightly out of phase. That is inevitable; bad things happen when the rate of pitch inputs starts to pump the natural oscillatory frequency of the combined control-fuselage-wing system. Think of a spring whose combination of resilience and mass will cause it to bounce at one cycle every two and a half seconds (My memory banks are dragging that up for a typical airliner; flight manuals tend to be silent on the subject.)
At the first impact, the back pressure is relaxed, but as the nose is rapidly falling through, the yoke comes back again.
This is the onset of a pilot induced oscillation (PIO), where the control inputs and aircraft responses (bottom two traces) quickly, and increasingly diverge. Eight seconds from the initiating pitch input, control and response are completely out of phase with each other.
While I am fairly confident that the mishap board will conclude there was a PIO, there is a tiger lurking in the shadows. The MD11 has nearly cornered the market on this kind of mishap. Googling [md11 “hard landing”] confirms this plane has a reputation unlike any other.
I think it boils down to a single design issue. The picture below (taken from the Wikipedia MD11 article) shows the MD11, and the DC10 from which it derives.
Even to the uninitiated, the horizontal stabilizer is conspicuously smaller than on the DC10. Additionally, since the MD11 fuselage is longer, the mass is distributed further from the center of gravity, meaning it has a higher polar moment of inertia; that is, more force is required to move the airplane in the pitch axis. Yet because the wing — which is what the horizontal stabilizer effectively is — is smaller, it cannot generate as much lift, which is what the pilots are using to move the nose up and down.
Smoothly controlled, that is unnoticeable. However, it seems certain that a combination of design characteristics makes the MD11 less well behaved when pushed, and more inclined to PIOs. All designs are compromises. The MD11 high altitude cruise, which, it seems, came at some expense to low speed handling.
Yes, but. The crash didn’t happen until 06:28:25, five seconds after initial touchdown, and two seconds before the final impact. Until that moment, application of go around thrust gets the crew another approach and landing, as well as greatly extending the rest of their lives. The same applies to nearly all the other MD11 landing mishaps.
Over the last year, there have been significant changes in MD11 landing training in the simulator that are, so far as I know, unique to the type.
We still have to figure out how to get the wetware to make a timely decision to take it around.
posted by Hey Skipper at 8:42 PM
3 Comments:
Because Kahului Airport has strong winds, often across the runway, I have long idly wondered how pilots deal with that. Since your post last year, I have taken a lot more flights than I usually do in 12 months, and the curiosity is less and less idle.
Flew in today, as it happens, and it was so windy that paddleboarders were being blown out to sea left and right. Landing was kinda wobbly.
So much so that the passengers (me not included) applauded when we finally settled down.
Some flights are like a petal settling, others kinda like bumper cars. Is it the pilot or the conditions?
Do some pilots put it down like a petal every time?
Your post makes it sound like the MD11 is an accident just waiting to happen.
If so, how has it survived for so long?
Harry:
A crosswind means the airplane is not pointed where it is going. Ordinarily, that isn't a problem: the airplane points where it has to point in order to go where it needs to go.
However, on landing, it is a problem. Most airplanes are not designed to take significant side loads on the landing gear. So, at touchdown, the airplane has to be going in the direction it is pointed, which also has to be the same direction the runway wants to go, in the touchdown zone, and on the center line.
Starting somewhere between 200 and 50 feet above ground (depending upon individual technique and amount of crosswind), the pilot applies downwind rudder until the fuselage is aligned with the runway.
Alone, this yaw fixes heading, but not ground track, so the pilot also has to feed in opposite aileron. The result is to precisely balance downwind drift with upwind turn. Done correctly, the upwind MLG will touchdown first, then downwind MLG, then NLG.
Next time you are approaching an airfield with strong crosswinds, if you pay attention, you will feel the airplane rotate and bank. The effect is more pronounced the further forward or aft you are sitting.
Most aircraft have a 30-35 knot crosswind limit. The driving factor is typically how much bank angle the airplane can have at touchdown without dragging an engine or wingtip, then subtracting five degrees.
Some flights are like a petal settling, others kinda like bumper cars. Is it the pilot or the conditions?
Yes. I have flown with some guys who get a smooth touchdown without leaving a bunch of runway behind them every time. Others get the touchdown, but often at the expense of touching down long. Still others (like the last guy I flew with) waste runway and plant it.
Since I think one of the most useless things in the world is runway behind the airplane, I focus on a "concise" flare. On a 0-10 scale, I am at about a 9 for touchdown point, and 7 for smoothness.
The difference between "are we down yet" and a thunk is really just a matter of a couple, or fewer feet.
Conditions make a difference -- gusty winds mean the wings are producing changing amounts of lift and the airplane is experiencing varying drift angles. In other words, the difficulty factor quadruples.
Barry:
Your post makes it sound like the MD11 is an accident just waiting to happen.
If so, how has it survived for so long?
Accident waiting to happen. Hmmm.
I think it is clear that the airplane does not behave smoothly at its limits. It is sort of like a Porsche 911 in that regard: a perfectly fine car except for that drop throttle snap oversteer thing.
The biggest reason it has survived this long is that ETOPS regulations have made tri-jets uneconomical in passenger service, so virtually all of them are hauling freight. That means crashes like this kill only a couple people instead of a couple hundred.
Also, there are some training issues related to use of the auto throttles, which I don't have time to go into.
Finally, the penultimate reason for this mishap is the pilot mishandled the aircraft by flying it into a hard landing. The ultimate reason was, having hit the ground that hard and bounced, failing to advance the throttles to full power and take it around. It is that wetware thing I mentioned.
So, yes, it has its issues, but nothing that can't be managed.
Post a Comment
<< Home