We Need 3-D Safety

The new automatic flight control systems in airplanes ranging from fixed-gear singles to business jets are just amazing. I never imagined I would see such capability except in multimillion dollar jets, but it's there in a Skyhawk. The big flat-glass displays get all of the attention, but it is the flight guidance computer at the heart of the integrated system that does the real work.

Using GPS guidance with accuracy improved by the wide area augmentation system (WAAS), these advanced systems from Garmin, Collins, Bendix/King, Honeywell, Avidyne, L-3 and others can automatically fly any published IFR procedure. The systems are smart enough to consider groundspeed and calculate when to start a turn so there is a smooth capture of the next leg of a procedure without overshooting. Vertical profiles are also stored in the database and the system can automatically descend and level off at the proper altitude for each segment of the procedure. The precision is simply fantastic.

But there is an unintended pilot trap in these wonderful systems, and that is that only a few can also handle the third dimension of flight -- airspeed control. Managing power and perhaps using drag devices such as spoilers and flaps is left entirely to the human pilot while the integrated automatic flight control system flies the airplane perfectly along the preprogrammed 2-D path in space. And the humans are not holding up their end of the deal every time.

The most recent example of a human failure to manage airspeed -- the third dimension of flight -- while the autopilot flew the other two Ds was the Dash 8 crash at Buffalo. According to preliminary reports the autopilot was engaged and flying the approach with the proper control over lateral and vertical paths. But when the automatic system leveled the airplane, as it should, the crew didn't add power. Airspeed quickly decayed and the crew clearly botched what should have been a routine recovery from an approach to a stall.

But the terribly wrong reaction to the stick shaker stall warning, and then stall barrier stick pusher, were simply tragic outcomes of the fundamental problem, which was failure of the human pilot to handle the one out of three dimensions of flight that the automatic system didn't control.

We already know more about the Dash 8 crash than we do about general aviation accidents because the turboprop airliner was required to have a very sophisticated flight data recorder. Investigators knew almost immediately all the details of the flight up to the fatal stall, including power settings and control positions. None of those facts are recorded in a typical general aviation accident.

However, there have been a steady stream of GA crashes with a similar profile to the Dash 8 accident. ATC radar shows the airplane flying the proper flight path, leveling off, losing airspeed, followed by the inevitable loss of control. There is no recorder to positively show that the autopilot system was engaged and controlling the lateral and vertical flight paths of those accident airplanes, but I'd bet my house that it was.

The potential for a pilot to fail to correctly manage airspeed control with the autopilot flying is particularly worrisome in turboprops more so than in piston airplanes or jets. The reason is that at low power settings the propeller on a turboprop engine goes to a very flat pitch to maintain the necessary rpm of the turbine section without producing too much thrust.

A propeller in flat pitch is essentially a huge air brake. The flat pitch of the blades gives the propeller arc enough solidity to block air flow and add huge amounts of drag. Anybody who doesn't understand that should try yanking the throttles to idle on a turboprop such as the King Air 90 while still several feet above the runway. That will make for a spot landing you won't forget.

That also means that when the autopilot system is flying a procedure, the pilot of a turboprop needs to be very prompt to add power when the system levels the airplane off before airspeed can disappear due to propeller drag. The Dash 8 at Buffalo lost 25 percent of its airspeed within seconds of leveling off.

The traditional solutions -- not really solutions but only reactions -- to this type of accident is to blame the pilots involved, demand more training, and damn the airline system that allows relatively low-time pilots to get jobs that pay next to nothing. But that reaction does nothing to further aviation safety.

Real advances in safety are made by expecting human errors to occur and then devising ways to avoid dying because of those mistakes. The safety record in airline and business jet flying is so good because there are multiple layers of systems that can detect a human error and warn of, or automatically correct, the potential problem. One of the most important of those systems is the autothrottle, which controls the third dimension of flight with the same precision the autopilot system delivers in the other two dimensions.

An autothrottle is necessary to advance safety because the autopilots are now so good in lateral and vertical control. It doesn't take long to develop so much confidence in a fully capable autopilot system that you find yourself looking away from the primary flight instruments to perform other chores, such as dial radios, look at charts, read checklists and so on, while the autopilot maneuvers on course. The system makes its course and altitude changes with such precision and dependability that we all soon take it for granted. But in all but a few business jets we, the human pilots, must still be paying very close attention to the airspeed as the airplane turns, descends and levels off. And airspeed control is at least as important, actually even more so, than being exactly on course and altitude.

The airlines recognized the critical nature of autothrottles decades ago and all jet airliners have them. It took longer in business jets, but at the top end Gulfstreams offered autothrottles more than 25 years ago and now they are available in Globals, Challengers and other higher-end models. Southwest has been an outlier in not using the autothrottles that Boeing installs in the 737s it buys under what I think is the bizarre notion that not using an autothrottle "keeps pilots in the loop." If you really believe that then don't use the autopilot for the other two dimensions of control and really keep the pilots in the loop. But nobody does that.

There are two reasons autothrottles have not proliferated as fast as the phenomenal 2-D capability of the autopilot system in general aviation. One is that pilots -- such as chief pilots at Southwest -- have not understood the important safety advantage they deliver. The other reason is that the systems so far have been very expensive.

I think the reason so few pilots understand the value of an autothrottle is that so few have had the chance to fly airplanes with a system installed. Manipulating the power to maintain a target airspeed seems easier to most of us than controlling bank and pitch angles to maintain a perfect course and altitude. But that's not true, and the only way you can appreciate that is to fly with an autothrottle system. When I am hand-flying with the autothrottle engaged it seems like more than half the workload has disappeared. It's hard to realize how every change in bank and pitch require an adjustment in power until you see the autothrottle moving in response to every flight control movement. It's like magic.

Autothrottles are expensive because they require precision servos to move the throttles, must have high-level monitoring to detect any failures, and must be smart enough to abide by all engine limitations. Autothrottles in airliners are essential to the automatic landing capability used in Category III approaches so you can understand how the FAA has demanded an extremely high level of reliability.

But I think the FAA should change the standards to make autothrottles available in all airplanes at an acceptable cost, just as the marvelous automatic flight control systems have become. The key is to lower the demands for reliability and monitoring so the autothrottle is only a little more complicated than the cruise speed control in your car. The only really critical capability is the absolute necessity to be able to shut it off if it fails. If and when it does fail the pilot is no worse off than he is now, having to monitor and adjust power by hand.

Like in the car, an airspeed hold autothrottle would deliver a huge percentage of the value for a small fraction of the cost of a full capability system. Of course it's great to dial in the target airspeed, or in recent systems have the autothrottle look up the published target airspeed from the flight management database. But that adds cost and complexity. If there was an economical system where you simply engaged the autothrottle when the airplane was on the desired airspeed, and the system held that speed, safety would be greatly enhanced and pilot workload reduced.

The avionics manufacturers can make such a system but they need to hear a demand from airplane manufacturers and pilots. It's starting to happen. I know of a few situations where customers have refused to buy a business jet until an autothrottle was available. Embraer is offering cruise airspeed control in its Phenom 100 light jet, and is working with Garmin to hold airspeed on approach. But so far nothing is going on for turboprops and pistons because not enough pilots know how valuable an automatic airspeed control system is and are demanding it.

It is true that aviation makes progress toward safer flying after serious accidents, and I think the Dash 8 crash at Buffalo with its detailed documentation and high profile brings the value of autothrottles to everyone's attention. We have to know that pilots, no matter how experienced or well trained, can and will make mistakes and we must have a system that backs up the human. An autothrottle would have changed the outcome at Buffalo, and who knows in how many general aviation accidents.