Losing Competence Part V: Asiana 214 and the Loss of Control Accidents

document.write(" serif">Automation and Hubris

Bernard Ziegler designed the Airbus to be pilot-proof. He is a good pilot, and he noticed that many pilots are less skilled than himself. In the interest of safety, he designed an airplane that could not be stalled. But it has been known for thousands of years that hubris is followed by nemesis, that Pride goeth before destruction, and an haughty spirit before a fall. (Proverbs, 16:18)

Hubris is arrogance before the gods. The goddess Nemesis alone can see the fine line between doing the best work you can and believing that your work is somehow superior. Cross the line and she is ruthless, finding your fatal flaw and using it to bring you down.

AF 447 was the fall of the hero. Pilot carelessness led the airplane into a line of thunderstorms. Supercooled water drops overwhelmed the pitot heaters, temporarily removing all three sources of airspeed information. The autopilot dropped off. The flight control computers switched from Normal to Alternate Law. The airplane can be stalled in Alternate Law.

Human or robot, there is always a fatal flaw.

How can we work with imperfection?

Don't Bow Down

Mankind, when confronted with the complicated or the divine, tends to bow down in worship. This can be hazardous in aviation.

The new automation – glass cockpit, fly by wire, IRS and GPS – together bring a change at least as momentous as going from props to jets in the 1960's. The aircraft is now such a capable pilot on her own that she almost seems real. We called the Airbus Fifi. Rather than bowing down, we found it was much better to treat her like a person. Dare to know her and maintain a relationship.

In her early days, frustrated pilots would exclaim, “What the #$%* is it doing now?” On a go-around at KLGA the map display would disappear, the airplane sailing off the edge of the world because it had passed the last waypoint in the flight plan. Or on a miss from a visual approach at KMIA the power would suddenly go to idle. Finger trouble with the Autothrust. She was trying to maintain go-around speed.

But the answers are right in front of you on the FMA. (Flight Mode Annunciator, at the top of the Primary Flight Display) We began speaking for her, calling out any change in the FMA, so we all knew what she was doing, or thought she was doing.

And yes, most of the time she was a damn good pilot. Just as we are. Exactly the same, including the occasional lapse. Which is why there is more than one pilot aboard. And which is why the human pilot should never bow down and never step aside. Know her (the automation, Fifi, the airplane) as well as you can. Always monitor her as you would a human pilot and call out anything unusual. And if she's not doing what she is supposed to, take over. For those interested in pursuing the subject, there is an excellent video, Children of Magenta, of a lecture by an American Airlines training captain. The take-away is the same: if she's not doing what you want, take over and fly by hand. You don't have time to figure out what you did wrong with the automation.

Crew Concept – and Not Just Humans

Moving from props to jets, pilots were introduced to many new concepts: mach tuck, dutch roll, deep stall, etc. Perhaps the most important were the long, shallow drag curve and the slow spool-up time of the engines.

Moving into the fly-by-wire era, we have to accept that the airplane (her automation) is part of the crew. Philosophically, it is perhaps a stretch, but in the real world of the cockpit it is a game changer and a life saver. As soon as you accept that the airplane is part of the crew – not a superior or inferior, but an equal – everything starts to make sense. She sounds the cricket as the autopilot drops off. In Alternate Law she says Stall, Stall as the panicked pilot pulls back on the sidestick.

But if you're on approach below 1000 feet (critical phase of flight) and the descent rate is 1300 feet per minute and the airspeed is below Vapp then someone isn't doing what needs to be done. (Without a glideslope the airplane will not understand that something is wrong.) The software doesn't care if the airplane crashes. She is a good pilot but she has absolutely no self-preservation instinct, no will to live. Human pilots have, or they have no place on the flight deck.

Losing Competence Part IV: Asiana 214 and the Loss of Control Accidents

document.write(" serif">The Drag Curve

Why did I say, yesterday, that saying pull back was exactly the wrong thing to do?


This is a drag curve. Every airplane has one. This looks like the drag curve of a small airplane, say a C-172. The curve for the B-777 is the same, except stretched out in the speed axis, like this:

During descent a jet's engines are at idle. Descending at perhaps 320 knots, it is near the right-hand end of the curve. Because of the 250-knot speed limit below 10,000 feet, at about 12,000 feet the pilot will pull back on the control column and hold the nose a little higher, slowing the rate of descent and trading kinetic energy (speed) into potential energy (reducing the rate of descent). As he does so he is moving left on the speed axis, toward the low point of the drag curve. Notice that drag is reducing as he slows. This means the rate of descent will decrease.

As he slows further for approach, he moves toward the low point of the curve. As you can see, in a jet the curve at this point is pretty flat, so the transition from one side of the curve to the other is a subtle one, covering, let's say, from 170 knots to 220 knots. In this range the pilot can pull back, raise the nose, and slow down without much affecting the rate of descent. Subtle though it is, this low point on the curve, the minimum drag speed, is extremely important.

Why? Because as he slows further for final approach, say to 137 knots, he is on what pilots call the back side of the drag curve. (You will more often hear back side of the power curve, but it is the same thing.) Everything changes on the back side of the drag curve. Slower speed requires more power, not less. If you don't add power, pulling back increases the descent rate. Pilots know that to stretch a glide, you (counterintuitively) have to push, not pull.

In a jet Vapp (final approach speed) is quite a way up the left side of the curve. You can see that as you slow, the drag increases, causing you to slow further. The airplane is “speed unstable”, which means that a small change in speed from turbulence or control input becomes a larger change if left to itself. What the pilot has to do is “catch” the speed as he approaches Vapp by adding power. The pilots of Asiana 214 forgot this step, and continued to slide up the left side of the curve.

There is an old pilot saying: Pull back, you go up. Pull back more, you go down. Now you are closer to understanding why, but you may ask, What happens when you get to the end, where the curve stops? To answer, we have to look at the lift curve.

Instead of drag and speed, we are now graphing lift against Angle Of Attack (AOA, or alpha). (In 1G flight there is an equivalence between AOA and speed, so it's OK to think of the AOA axis as speed, except in this graph slow is on the right, with larger angles of attack.)

AOA is the angle at which the air meets the wing. What pulling back on the stick or the control column really does is increase the angle of attack. So if the pilot needs to increase lift, he pulls back, moving along the curve to the right. But the wing only “flies” in a narrow range of AOA, from 1 degree or less up to about 16 degrees. Above that the airflow starts to detach from the upper surface of the wing and the wing rapidly loses efficiency. That is called the stall. Now the wing is more like a board. It is still pushing air around, but not nearly as much air. The wing is stalled.

You can see in the graph that as you move to the right, increasing angle of attack, lift increases in a straight line and then suddenly goes over the top like a roller coaster track. Indeed, that's just what it feels like when an aircraft stalls. Now as you pull back more lift decreases, and the bottom drops out. That's what happened to Air France 447. When the autopilot kicked off the pilot panicked and pulled. For a brief period they went up. The pilot kept pulling, and for four minutes the aircraft fell, at angles of attack well above the stall, until it hit the sea.

So that's why the drag curve ends on the left-hand side. The end of the curve is the stall. Asiana 214 was almost there, and that's why the stick shaker went off. They weren't stalled yet, but they were within a whisker. They were sitting on top of the roller coaster. And when they pitched up even further in that last second before impact (you can see it clearly the YouTube Video and in this re-enactment) it did nothing to change their flight path. All it did was lower the tail and landing gear.

Now you know why.

Next: Why aren't pilots better trained and more experienced?

Losing Competence Part III: Asiana 214 and the Loss of Control Accidents

document.write(" serif">Today's News

NTSB Chairwoman Deborah Hersman continues to impress. Quoted today in the San Francisco Chronicle, she says: “What I'm telling you is that from 500 feet to 100 feet, there is no mention of speed.” That's on page A10. On page A12 there are two articles, Do pilots have adequate skills? by airline pilot James F. Atkinson, and When will FAA require alerts? by lawyer Robert A. Clifford. (I am not including links to these articles because you would have to be a Chronicle subscriber to read them.)

Atkinson rightly addresses basic flying skills and airmanship, pointing out that today's automated systems actually undermine skills. Clifford calls for mandatory low airspeed alerts, missing the point that this would make the pilots even dumber. (It is worth analyzing the terrific save by the captain of the Quantas A380 that had the uncontained engine failure. There was so much damage and so many (hundreds) of ECAM alerts that he finally said, Stop ECAM. Lets go backwards and just see what we've got left. That critical decision was the key to saving the airplane.)

Analysis of the Last Minute

At 1000 feet, 54 seconds before impact, someone says, Sink Rate. The throttles are at idle. The training captain tells the trainee, who is flying the airplane, to pull back. This is exactly the wrong thing to do. We will explore why that is so in greater detail in another post, but for now we'll say that they were at idle, on the back side of the drag curve, so total drag is increasing with angle of attack. As any pilot knows (see Stick and Rudder, 1944), instead of correcting the sink rate, pulling back on the control column actually increases the rate of descent and causes the speed to decay faster and faster.

At 200 feet, 18 seconds before impact, the training captain realizes they are too slow and moves to engage the autothrottle. After saying pull back he does nothing for 36 seconds while the airplane descends at over 1300 feet per minute. The target for any approach is 700 feet per minute. The engines are still at idle. They are well below approach speed.

Could they have done a missed approach that point? I will leave that for formal analysis, and point out only that these aircraft are designed to be able to do a baulked landing from any point before touchdown, but only if the engines are already spooled up and the speed is at approach speed, about 1.3 times the stall speed.

Ten seconds later, it was already game over. Perhaps the autothrottle had been armed, but most likely it had not actually been engaged, so thrust lever movement happened only now, at about 100 feet and 8 seconds before impact. And it will be another 5-7 seconds before the engines develop any useful power. So we see the slowest speed at 3 seconds before impact, at perhaps 40 feet above the water. This is where the passengers behind the wings see the plumes of water as the engines start to spool up. Meanwhile the stick shaker is going, indicating impending stall. Despite pulling back and belatedly adding power, they are still descending at 750 feet per minute, by my calculations. This is the first time anyone in the cockpit calls for a go-around. Of course it is too late. Way too late.

What the training captain should have done, back at 1000 feet and 54 seconds, is push the power up. Manually. With the thrust levers. The problem is that he would probably not been able to stabilize the approach from the idle thrust, slow (148 knots) and 1300 feet per minute sink rate descent. It would have been a neat parlor trick if he could have put on go-around thrust, pushed to counter the nose-up moment of the added thrust and bring the speed back up to bug (the approach speed), and then quickly brought the power back to approach power and held the speed. At 1000 feet he had room to fart around a bit, at least in theory. But airline Standard Operating Procedures (SOP's), his own airline's included, say that the approach must be stabilized by 1000 feet and remain stabilized, or else a go-around shall be performed.

So what the training captain really should have done is to say:

I have control.


That's it. That's the last time the training captain, the Pilot in Command, had any say in the matter. That's when the pilots, the crew, gave up having any influence over the outcome.

It is sad, but true. It must be said. The pilots were incompetent.

Next: aerodynamics they should have understood . . .

Losing Competence Part II: Asiana 214 and the Loss of Control Accidents

document.write(" serif">News and Public Relations

Deborah Hersman, The NTSB's Chairwoman, has taken some flak in the last few days. But from my perspective, she is one of the few in responsible positions who are looking good.

First a minor annoyance: on Saturday and Sunday news outlets kept repeating a young witness's observation that Asiana 214 came in “low and fast.” Many immediately available facts, including where the aircraft came to rest, made it a slam dunk that the aircraft was, instead, flying way too slowly.

Then on Monday the Korean Government announced they would be “inspecting all Korean B-777's”. On Tuesday and Wednesday various pilot unions called for Ms. Hersman's head, saying presumptions of pilot error were speculative and premature.

Please. I am used to the power players grandstanding their interests, but this is amateur hour. There is only so much ignorance out there.

The Last Thirty Seconds

Now let's get back to what we know. The cockpit cleared the breakwater nicely. The main landing gear and the tail did not. The speed at impact was 106 knots, within a knot or two of the stall speed. (The approach speed should have been 137 kt.) One and a half seconds before impact, engine power increased. Passengers in seats just behind the wing could see spumes of water being thrown up. At four seconds before impact the stick shaker operated, signifying an impending stall. At seven seconds someone is heard on the voice recorder calling for an increase in speed. In his interview the training captain said he went to push the throttles forward but the trainee already had. (Notice at least 5 1/2 seconds elapse between advancing the throttles and the increase in power. Seven seconds spool-up from idle is typical for a fanjet engine.) At 500 feet altitude the training captain became aware that they were too low (the PAPI lights were red over red) and he asks the trainee to pull back. The training captain also notices they are not aligned with the runway. Ms. Hersman says at that point they knew they were low and they were making lateral corrections to line up on the centerline of the runway.

These are the bare facts.

Flight Preparation in Seoul

Now let's go back twelve hours or so to the pilots' briefing. The dispatcher has already produced the flight plan with its route as close as possible to the minimum time track. The weather is good over the Pacific and at destination. Most likely they have fuel for a close alternate, such as Sacramento. It looks easy. But somewhere in the data available to the dispatcher and pilots are these two lines:

ISFO 06/005 SFO NAV ILS RWY 28L GP OTS WEF 1306011400-1308222359

ISFO 06/004 SFO NAV ILS RWY 28R GP OTS WEF 1306011400-1308222359

San Francisco airport (KSFO) always lands on runways 28L and 28R unless a winter storm blows through. With today's light winds and good visibility it is a near certainty that these runways will be in use. But decoding the two lines above (they are called NOTAMS, for Notices To AirMen) we find that the GlidePath (GP) is OuT of Service (OTS) for both runways. The When in EFfect (WEF) is from June 1, 2013 to August 22, 2013 at midnight. This is important because the aircraft cannot fly these approaches on autopilot in the way the pilots are used to.

Here is where we have to move into sensitive territory. (There will be more of these before we're done.) At the end of the article Terror on Jet, in The New York Times on Monday, July 8, we find these lines:

Some experts said that pilots often have little opportunity to practice landings without the aid of such technology . . .

Still, given that the weather was ideal and the guide lights (PAPI, or Precision Approach Path Indicator) were on, making a visual landing should not have been difficult for most commercial pilots . . .

on a difficulty scale of 1 to 10, this was a 2 or 3 at the most . . .

Pilot Judgment

A pilot's most important skill is his judgment. (see my Developing Pilot Judgment) He must look at the tasks and maneuvers ahead and ask two questions: Can the airplane do it? and Can I do it?

The former is mostly hard data in the Aircraft Flight Manual Limitations section, but it is also practical knowledge of what the aircraft's systems can and can't do and an understanding of the feedback systems that tell the pilot if the job is being done. (A good example is the AutoThrottle).

The latter question is the more important of the two: Can I do it? The only way to answer is through experience, and it is not measured in flight hours.

Training: have I been trained in this maneuver?

Practice: have I practiced it on my own? What were the results?

Recency: have I done one in the last 30 days? 90 days?

When Asiana's pilots were preparing for the flight in Seoul, the two NOTAM lines about the glidepaths on 28L and 28R should have triggered a dialogue:

We're going to have to do an everything-off visual approach in KSFO. Has any of us been trained for this? Who has practiced one in the last year? Which of us has done one in the last 30 days? 90 days?”

I'd be willing to bet (I'll back this up in future installments) that none of the four pilots had flown a visual approach in the last 90 days. In that case, pilots with sound judgment would never have attempted the visual approach to 28L in KSFO. They would have diverted to Sacramento where there were long runways with functioning ILS systems.

Next: what else they did wrong . . .

Losing Competence: Asiana 214 and the “Loss Of Control” Accidents

document.write(" serif">Introduction

Asiana 214 is in my dreams. All day her last two minutes of manually controlled flight replay in my head. Searching for a cause does not distress me. The pilots were clearly incompetent. But how did we get to this pretty pass? My overwhelming sense that that's where we are distresses me greatly.

For the last few years disasters like this one have come to be known as “Loss Of Control” accidents. It's a catchy label, but it doesn't get to the heart of the matter. The pilots of these airplanes – I'm thinking of Colgan Air at Buffalo, Air France 447, and Asiana 214, but there are many more – these pilots fundamentally did not understand what was happening, so they were unable to do their jobs.

How has this come about? And how can it be fixed?

At this point I don't know if this is going to be a blog, a series of blogs, or a book. I know only that I must explain the technical issues, explore the commercial and financial forces as they interact with my trade, and try to map a path through this crisis.

Flying has grown into a huge industry. The era of limitless supplies of hugely keen, military-trained pilots is over. Worldwide, there will be a demand for over 600,000 pilots in the next decade. Where will they come from?

I love flying. Most of my working life has been in airplanes, flying and teaching. I see flying as a living link between the sailors of the past from Magellan to Jack Aubry and the space explorers of the future like Jim Kirk and Jean-Luc Picard.

Sailing, navigating, flying: these have always been apprenticeship trades. You learn the theory, but you also learn the practical, the hands on. You practice. You repeat. You get sharper. Gradually you come to understand what you have to do to stay sharp, to stay competent. (Or, in airline-speak, to maintain your compentency.) Then you pass your knowledge on.

Somehow this process has broken down. There is no single villain, no smoking gun. Instead there are many villains conspiring unwittingly, starting with you and me, airline customers, frequent flyers, looking for a painless flight and, most of all, for a deal.

Training a pilot, says Transport Canada (in the Flight Instructor's Guide), must be done right the first time. The pilot can't see his airplane moving through the air, because most of the time air is invisible. Instead he must imagine the air flowing over his wings and through his engines, and imagine the air pushing on his slipping or skidding fuselage and fin. Above all he must imagine the angle the airflow makes meeting his wings and how this critical parameter is related to lift and airspeed. He must viscerally feel the drag curve as he controls this angle, the Angle of Attack, as he slows his airplane for approach and landing. He must understand at all times where he is on that curve and what the consequences are. He must know how to fly.

It is not as easy as 1, 2, 3. It requires work and practice, and most of all it requires imagination.

Pilots don't call it that, of course. That sounds too much like an artist, an inventor, or a writer. Pilots refer to it as Situational Awareness. It's what was missing in all of these “Loss of Control” accidents. But why? And how can we fix it?

More to come . . .