And why we all should
Turbine-powered airplanes don’t crash very often these days, so when they do we should all take notice. Upon reviewing such accidents from the past few years, it’s clear there is a disturbing trend in modern cockpits: pilots struggle to control the airplane after the autopilot suddenly quits flying. Now before you start bemoaning the current state of stick and rudder skills and urging all pilots to start flight training in a Cub, let’s consider another (more nuanced) option.
Can you recover if George walks off the job?
Air France 447 is probably the best-known accident, since it garnered non-stop news coverage and the airplane wasn’t found for almost two years after it went down in the Atlantic Ocean. While the accident was complicated, the short version is that the autopilot kicked off (at night, in IMC, over the ocean) and the pilots proceeded to stall the airplane and fly it stalled all the way to the ocean. There is plenty of blame to go around, from system design to crew training, but the important lesson is that the pilots were not in the loop. When suddenly forced to fly the airplane, they did not have the situational awareness or the training to handle it.
Another accident, and one that received less attention, involved a Pilatus PC-12 in Florida. The final report is not complete yet, so all the details aren’t known, but it’s clear that the pilot lost control of the airplane in IMC and it crashed nearly straight down. The question is why. One plausible theory is similar to the Air France accident: the airplane, at near gross weight after making a fuel stop, was climbing through 25,000 ft. in the clouds when the autopilot disconnected due to turbulence. The Pilatus has a stick shaker that will disconnect the autopilot at high angles of attack, and the slow speed required to climb at FL250 plus the convective weather in the area could have caused exactly that. This is not unique to the PC-12 – most autopilots will disconnect if the turbulence gets bad enough.
If he did lose the autopilot, the single pilot, who was relatively new to the airplane, may have simply lost control in the clouds. I’ve been in a similar situation and it’s quite unnerving, as you go from monitoring the autopilot to flying (often with no flight director) in an instant. You do not have long to get the airplane under control.
None of this is to suggest that autopilots are dangerous or that they should be avoided (as some Luddites argue). Modern autopilots are nothing short of amazing, and I firmly believe they increase safety. For example, single pilot IFR is certainly possible without an autopilot, but the safety margins are thinner. If you have an autopilot, you should use it. If you don’t have an autopilot, you should add one.
But even the most ardent autopilot supporter has to admit they have limitations. And in some critical situations, like the accidents described above, autopilots essentially throw up their hands and say to the pilots, “your airplane!” There is no warning that the autopilot may be about to shut off and no in-between state: it’s there one second and gone the next.
One response to this could be to design autopilots that react better, and some pilots have advocated this approach. Theoretically, autopilots could have some type of fall back mode, where they don’t disconnect completely, but revert to basic attitude hold. Better annunciators are another option, where the autopilot explains why it disconnected and whether it can be turned on again.
Some of this may be possible, but a far more practical (and more affordable) option is for pilots to regularly practice this failure scenario. Call it unusual attitudes for modern airplanes – we need to experience what it’s like to be unexpectedly thrust on stage. I had never done it myself until very recently, and in talking to a number of other pilots, I haven’t met one yet who does this on a regular basis. We all practice emergency scenarios that are exceedingly unlikely to happen, like engine failures, and yet we ignore a scenario that has proven to be both possible and fatal.
The best way to practice autopilot unusual attitudes is to go up with an instructor or safety pilot and fly like you would on a real trip. Put on the hood and engage the autopilot, then have your co-pilot randomly disconnect the autopilot, both straight and level and in turns. Be realistic – if you spend a lot of time with your head down, looking at your iPad, practice recovering from this position. Staying in the loop is a critical part of the exercise, since many of us may relax in cruise flight and may not be spring-loaded to fly.
In particular, it’s the transition that counts. Many instrument pilots feel a little uneasy when they first transition from VMC to IMC, especially if they haven’t flown lately. The same is true with autopilots: the key is those first few seconds after you lose the autopilot. It’s basic attitude flying, and the focus should be on flying the airplane and nothing else. Also consider that the autopilot may have disconnected due to an AHRS or air data issue (as in the case of Air France), so it’s a time to rely on your profiles. What power setting and what pitch attitude lead to straight and level? You should know that cold.
Curmudgeons may scoff at this as another sign that modern avionics have made us all bad pilots. But I think that’s an overreaction. Even the best new technologies still have quirks to learn and new procedures to practice. Nobody argues that celestial navigation is better than GPS simply because a GPS receiver can fail and stars don’t. The tradeoffs are worth it.
Practicing autopilot failures is just part of being a safe, modern pilot who embraces all the tools available. Don’t just read about it in the manual – get out there and fly it!
- Weather flying means learning to read clouds - November 11, 2024
- What’s wrong with the teardrop pattern entry - September 25, 2024
- What matters for VFR proficiency: better landings - August 26, 2024
You make great points here. What stands out to me is in the last paragraph, the word “tools.” Autopilot is a tool, not a crutch or replacement for the pilot. Autopilot must be monitored just like a human pilot sitting next to you to ensure everything is as expected.
Regarding improvements to autopilot systems, an attitude hold default would make sense. I was always taught to maintain attitude rather than fight for altitude if ever caught in significant turbulence. Is this the same for you?
That’s the way I learned as well. Chasing airspeed or altitude will not do you much good in that situation. When in doubt, fly that attitude.
John – you make a very good suggestion on training with autopilots! Kinda strange that autopilot failure recovery training has never been adopted as part of standard advanced pilot training and flight reviews … and given the fact that autopilots have been widely available on light aircraft for many decades.
As to your mention that the pilots in several high profile aviation accidents were not maintaining situational awareness (most recently with the Asiana 214) with auto-controls engaged, it’s simply incredible that professional pilots do that, regardless of any amount of cockpit automation available and in use.
I am just a lowly private pilot who, considering how much cross-country flying I do, installed in my simple fixed gear airplane a modern S-Tec A/P and Garmin IFR approach-certified panel-mount GPS coupled to the A/P. I use the Garmin and the A/P for at least a portion of nearly every cross country flight. Whether the A/P is engaged or not, my instrument scan goes on continuously, to make sure the airplane is going where I intend it to go, and performing as expected. It just never occurs to me that as PIC that I can ever afford to take on the mentality of a mere passenger. In fact, even when I am flying right seat with someone else as a mere passenger, I cannot help but scan the panel pretty much continuously!
I would think that autopilot failure training should rate right up there with attitude indicator failure training.
On two recent ferry flights I have seen unannounced autopilot failures where the autopilot just rolled the airplane into a spiral (in visual conditions, fortunately for me). During autopilot certification, these failure modes are tested and it is verified that a pilot can recover after a short period of recognition; but that assumes close monitoring. I could see where a pilot with his head in the iPad could have a real problem.
The problem with training as you suggest John, in the airplane, is that the instructor directed autopilot disconnect will be announced (unless he pulls the CB, which still only simulates a total failure with no rapid rolloff or pitch). I think random quiet failures of different modes including turbulence disconnects, in a simulator, would be much more effective.
“Be realistic – if you spend a lot of time with your head down, looking at your iPad, practice recovering from this position.”
Sounds like an article needs to be written about “See-And-Avoid” procedures and techniques. And what’s wrong with wanting to train in a cub? I have to say the smug remarks about “stick and rudder” ” are pretty disconcerting. I’ll fly with those “Curmudgeons” any day! You bring the valid points of training for systems failure I completely agree. However, I feel it would’ve been best accepted in a slightly better tone.
Adam, I fly tailwheel airplanes all the time and I’ve written before about my love for Cubs – heck, I’ve even been called a curmudgeon myself. I’m not picking on them/us, just pointing out that oftentimes we fall back on our pet causes. In the aftermath of the Air France crash, there was a lot of self-righteous talk about how the new generation of pilots aren’t “real aviators” or whatever. I think expecting every pilot to train in a tailwheel airplane (especially if all they’ll ever fly is a Boeing) is unrealistic. Let’s train the systems and train the potential failures.
What is different in control of tailwheelers when inflight??? Maybe none!
Celestial navigation. Can. Fail. It is called clouds.
I would love to learn celestial nav just as a novelty. But yes, no sense in looking to the stars in an obscured sky. I’m glad to be living in the 21st century. Lots of enhancements toward safety in this age. Unfortunately, the ultimate price of lives lost had to occur to know what we know now. I believe a perfect blend of proficiency in new age technology as well as basic fundamentals makes the idea pilot. None of this is possible however without a pilot displaying safe ADM. I like these articles. New reader here!
I had a autopilot “failure” (actually, it was a pilot failure) recently. We had been flying below a convective cloud, trying to get out from under it. We had powered back to minimize airspeed and possible turbulence damage. ATC called for an immediate climb (into the cloud) because of obstacle conflicts. I activated climb mode. Did you see what I missed?
Yup, on-autopilot, climbing, low power setting eventually equals VERY high angle of attack. As my tardy scan finally came across the Aspen AI all I saw was red chevrons pointing down. It took a second or two before I understood what was happening.
Power up, ease the nose over (don’t want to scare the little woman, or look uncool, don’t ya know).
Blessedly, we didn’t stall. That was WAY to close for comfort.
How does that old saying go? “Your mistakes will make you stronger if they don’t kill you.”
Joe
“I activated climb mode.”
Unless your AP includes autothrottles, you did no such thing. This goes back to probably your second lesson as a student pilot. Unless you absolutely intend to alter your airspeed for some purpose, a climb ALWAYS requires an increase in power; a descent ALWAYS requires a decrease. You can cheat with alterations to gear and flaps settings, but there’s very limited cheat authority available.
This is what brought down that 777 in SFO.
Two old jokes: What makes an airplane fly? Money. What makes an airplane climb? Power. Both are true. Only one is funny.
After reading this article and the comments, I can offer this: Hand flying skills, including unusual attitude recovery, can never be neglected. Never. If you, as a pilot, feel uncomfortable at all with your hand flying skills, take steps to fix the problem now. Find an instructor, tell him what you need and do it.
Simple rule of thumb – Learn everything possible about the aircraft, BEFORE flying it, then practice, practice, practice until a QUALIFIED instructor, in that aircraft, says you are good to go. Then go practice some more. Anything made by man or machine, is susceptible to failure, at the most inopportune time. Be ready to “break that chain” at the very first opportunity.
I have to agree; you’re dead on correct. As I read the original report on the Air France incident, I just couldn’t believe it. When I was doing my flight training for my private pilot certificate, my instructor did multiple simulations, having me with a hood bending my head down starring at my lap while he put the airplane in many puzzling attitudes, altitudes and engine speed configurations. Then he would shout, “you got it!” The task was to recover and return to the previous altitude, heading, speed, and straight & level flight. I thought it was excellent training and one of the best techniques I have practiced.
A few years later in some really bad weather between Grand Cayman & Jamaica, while flying right seat, that training made me recover and take over from the PIC who was suddenly overwhelmed (disoriented actually) when we hit a pocket in IMC conditions and were thrown off autopilot, attitude, altitude and crazy speeds with zero visibility. It took me right back to that training. I just imagined I had on the hood and my instructor was messing with me (except it was a very rough ride in that tropical weather).
I can’t stress enough the importance of that type of training, even for seasoned pilots. Thanks for raising awareness of this type of training.
READ THE MANUAL, you must be joking. Real pilots read the manual IF they have a checkride scheduled, otherwise it sits unopened and unread and with all the years previous updates still wrapped in cellophane.
Manuals and checklists are important, but so is awareness about when to skip ahead and do something NOW and then go back to complete the checklist.
An example, ditching in the Hudson River, once it was obvious that they were going into the water, hitting the FIRE buttons on each engine will shut the fuel OFF. And going to the last item, closing the cabin vents.
All else was nice to do, when the ditching starts at 10,000 feet. At less than 2,000 feet the pilots should know their emergency procedures well enough to do the critical steps as easily and quickly as possible.
Autopilots are a wonderful device, an idiot will happily fly the airplane into a corner, trimming all the way. When it disconnects the plane will be out of trim and it may take all the pilot’s strength to control the airplane and your free hand to spin the trim wheel. 99 times out of a hundred, the airplane will be in trim when a pilot takes over, the hundredth time it may be out of trim in every axis. It is really fun with an engine out.
So instructors need to teach into the dark corners and have their students look into those dark places. Simulators are wonderful, the multimillion dollar sims can do these things in complete safety if the sim instructor has a manual to tell him/her to set up the accident.
I owned a Cutlass RGII that had an autopilot that would cut out every so often and you couldn’t re-engage it until you shut down the aircraft. We could never figure it out but when it did let go would go into a pretty sharp roll to the left.
I can tell you I had plenty of practice with this situation and it became something as common as adjusting the altimeter every so often. Press the button to disengage always worked and the back up plan was pulling the CB, which I never had to do.
Had the unit into the shop multiple times replacing this and that.
i used to fly a lot with my dad in the midwest in a 310-had a terrific motorola(lear) autopilot but it would start riding the vertical speed indicator, and next thing you would notice is the plane go up and down. started to keep an eye on that puppy and never forgot the first rule of flying–fly the airplane. even if the autopilot was doing the work, you monitored everything that was going on. the second you saw the porpise start, you took the altitude hold off–re-trimmed, and reset the altitude hold. you always made sure you checked to see that it was working properly.
Most light aircraft nowadays have a glass cockpit with an associated flight director system. Pilots can be addicted to flying the flight director with the unintended consequence that they concentrate entirely on those two crossed needles or V-Bars and get out of practice at scanning all the flight instruments which make up the big picture as it were. Blind adherence to the FD can kill you. FD are meant to be an aid to instrument flying yet even in the airlines the FD is very rarely turned off even though it may be giving the pilot false information. This leads to FD addiction.
You should be able to turn off the FD and seamlessly switch to manual raw data instrument flying especially if the FD is not giving you the information you require. In an unusual attitude situation get rid of the FD straightaway as you need to instantly scan all flight instruments during recovery and not be seduced by dancing FD needles and find yourself losing control.
Older model Cessna 150 and 172 have an emergency procedure published in their POH.
Inadvertent IMC, release the control wheel, with rudder only steering set a standard rate turn on the turn and bank instrument for one minute then level the wings and fly out of the conditions.
The same procedure applies if losing visual reference at night. Adding power will cause climb to re-attain visual or reverse course.
I questioned over 100 Instructors and 5 Examiners flying these aircraft and none knew this “emergency” procedure!!!
Hands-off control should be practiced and be the normal method of control when hand flying.
An excellent reference to this is the 2014 Mar/Apr FAA Flight Safety-Brief page 13. All pilots should read this article.