Thursday, 24 January 2013

Gethomeitis

Hello everyone, sorry it's been a while but endless revision has been driving me round the bend and I haven't felt like I had much positive to say.

The phase one school finals are finally out the way and I managed a very respectable 97% average in the exams (thanks mainly to two very short, very easy comms papers). Perhaps once I have finally caught up on sleep and given my brain a rest I will be able to feel positive about that.

We now have a welcome week at home to prepare for the 'real' exams which start on 4th. The format and difficulty should be identical to the school finals so barring any nasty surprises I can't see me or any of my colleagues on the BA programme coming unstuck. We will then be more than half way through ground school and and after that... two weeks off! Happy days.

The subject of this post is gethomeitis (Get-Home-Itis). This is not as it sounds a painful disease of the big toe. It does not mean that I've been desperate to get home to my wife (though I have). It is a rather ugly aviation term meaning simply the tendency of pilots to want to complete the flight as planned and land at the destination no matter what.

Gethomeitis? In gliding, a low approach over unlandable
terrain might get you home, but would a landing in a field
five miles back have been the safer option?
Gethomeitis has been the cause of an awful lot of aviation accidents; not just gliders and private planes but airliners too. One study in France reckons that gethomeitis caused 42% of all fatal accidents to aircraft on visual flights between '91 and '96.

In 2001, a regional airliner Crossair 3597 crashed in Zurich killing 24 of the 28 people on board. The highly experienced Captain Lutz attempted to land in poor conditions on a runway that was not equipped with an instrument landing system, just as the airport was closing.

There is nothing intrinsically wrong with attempting an approach in touch-and-go conditions provided the correct procedures are followed. Lutz however put the plane into a fast, steep descent and — crucially — flew well below the minimum safe altitude for the approach without visual contact with the runway. He then claimed he could see the airport when he could not possibly have done and continued the approach. The relatively inexperienced first officer did not intervene, with the tragic result that the plane crashed into hills some miles short of the airport.

How could a highly trained and experienced airline captain to make a string of such poor judgements? To answer this, we need to look at some psychology.

It's quite fashionable among trainee pilots to treat the subject of Human Performance rather dismissively, and to be fair there is a lot of meaningless verbiage in the textbook. Or as my wife called it when helping me to revise, "mumbojumbo". She is a master of unintentional puns.

But buried in the management speak there are some gems, including this insight into human decision making (bear with me here).

It turns out that people are hopeless at assessing risk. Absolutely terrible. After all, would we spend £6 billion per year on gambling in UK if we weren't? Would we ever drink and drive? Would anyone smoke?

There are two major biases in the way people assess risk and benefits.

1. Given a choice between two positive outcomes, people will choose the one that is more certain. For example, if I offered you the choice of £10 cash now, no strings attached, or £100 in August 2014 "probably, if I remember", you'd no doubt take the tenner. I'm not known for my good memory, and it's very hard to turn down a dead-cert.

2. Given a choice between two negatives, people tend to avoid the more certain one. Gambling is the typical example; if you have lost money the temptation to up the stakes is incredibly strong, even though the chances of a big win are remote. When our outcomes are negative, suddenly we favour the long-shot.

And the closer we are to our goal, the more reluctant we are to give it up. There is even a name for this — the Concorde fallacy.

Despite endless problems, the British and French governments repeatedly ploughed incredible amounts of money into developing Concorde. Costs spiralled utterly out of control, but so much had already been invested that failure simply could not be countenanced. Eventually six times more than the original budget was spent. They did finally succeed in getting it into service, and it was an engineering marvel, but financially it was a disaster.

You've probably already realised that this phenomenon explains gethomeitis. In the position of Captain Lutz of Crossair flight 3597, you basically have two options:

1. Play it safe and divert to an alternative airport. This will really annoy the passengers, it will cost the company a lot of money in hotels, food, lost customers and another flight the next day, it will mess up all the rosters because the aircraft and crew are in the wrong place. In short, it is a pain in the arse for all concerned; a certain negative situation.

2. Continue and try to land in poor conditions. If successful, which is highly likely, none of the above will happen.

Of course there is the tiny risk of total disaster, but as we have seen people routinely underestimate this (or we probably wouldn't drive cars) and cling onto the much larger chances of avoiding the known problems of diverting.

Put this way, you can start to understand — though perhaps not forgive — his behaviour. On top of this, he apparently was right at the end of his maximum duty hours and therefore tired, and he wanted to go to his grand-daughter's birthday party the next day. These considerations should not be important but hey, we are all human.

So how can we avoid making similarly bad decisions, not just in flying, but in life generally?

The answer is to flip around the options to make them both positive. As we saw above, when options are positive, people tend to favour the dead-cert over the long-shot. We could rewrite Captain Lutz' options like this:

1. Avoid the certain inconvenience of diverting and all the associated hassle and costs by simply continuing on and hoping the landing goes well. Or

2. Avoid the slight possibility of total catastrophe, loss of the aircraft and multiple fatalities by diverting to a safer airport.

It's obvious, isn't it? Put that way, how could you not take the safe option?

Captain Lutz was probably going through his ground school training in about 1950, most likely in the military. I very much doubt that he had the benefit of modern psychology to assist his decision making. If he had, perhaps he would have looked at the situation differently, and him and 26 others would be alive today.

The modern airline pilot must continuously make decisions and judgement calls based on experience and knowledge, and despite their best efforts and training they are subject to the same human foibles as the rest of us. So this sort of knowledge is not incidental, it is absolutely central to the job. Human Performance is a compulsory part of our training for a reason — it may turn out to be the most important subject we will ever study.

Thursday, 3 January 2013

What is the dog there for?

Hello and happy new year to everyone!

After an all-too-short break we are back at school with a bump, racing to fit in the remaining phase one material before exams start on 21st January. One of the subject we are working on is autoflight.

I'm sure it is no surprise to anyone reading this that passenger jets are flown, most of the time, by autopilots and not by the pilots pulling on control columns and shoving on pedals like you see in the movies. In fact, many jets do not even have control columns, but that's another story.

What you may not know is just how far this automation has gone. Our instructors enjoy winding us up about being button-pushers or sys-admins rather than real pilots, or about how little work we will actually do, or how boring the job is. This is partly because most of our instructors are frustrated ex-flight engineers who were made redundant by increasing automation on the flight deck.

The flight engineers are not the only ones to be consigned to history, the navigators and the second officers have gone too, all replaced by computers and sophisticated control systems.

Just the two pilots remain — but for how long? Are single pilot or even pilot-less cockpits the future?  To answer this we need to look at what the automatics can and cannot do.

Computers everywhere


It is computers that have made the high degree of automation we see today possible. Large planes are stuffed full of them. They have been responsible for a huge increase in safety, reliability and efficiency of operation.

There are computers running laser gyroscope-based inertial reference systems and computers that analyse the data from the probes and sensors outside the aircraft and still more computers communicating with radio navigations systems. These all feed into computers that work out the statistically most likely position of the aircraft and feed this information into many other sources.

A horizontal situation indicator (navigation display)
from an Airbus. Note the weather radar display (large
blobs) and the purple line showing the planned route
that the aircraft will take - with no help from you!
The main ones are the flight management computers that provide navigation both laterally and vertically, flight director computers that indicate to the pilot the most efficient way to fly this course, and of course autopilot and autothrottle computers that mean they do not have to.

When a control is actually moved, it may no longer be connected to anything physical. Instead it produces an electrical signal to, you guessed it, a computer that decides exactly how far and fast the control should move, and prevents the pilot or autopilot doing any manoeuvre that could overstress the aircraft. Other computers prevent stalling via stick shakers and pushers and even simulate the 'feel' of old fashioned mechanical controls.

Information on the aircraft's attitude, speed, altitude, vertical speed, heading and so on is displayed on a computerised readout called a Attitude Director Indicator, and another computer-controlled display called a Horizontal Situation Indicator gives navigation information, weather radar and so on (see picture). Between the pilots, a further pair of displays (fed by various computers) give information on engine performance and aircraft systems together with any warnings.

Still more computers contribute to safety by providing ground proximity warnings and traffic avoidance, fault warning systems, voice and data recorders and devices to allow the aircraft's position, height and identity to be constantly sent to other aircraft and controllers via radar and data link systems (you can see this information live on websites such as FlightRadar24.com)

The avionics bay of a Boeing 737 - where many of its computers live
Some of the most advanced computers are control and monitor the engines. With dozens of sensors they will spot impending problems before they become serious, and both the airline and the engine manufacturer will know about it before the plane even lands.

These are just the main computer systems that I can recall right now, there are more. Braking computers. Weather radar computers. Pressurisation controllers. I could go on.

All of the vital ones will be at least duplicated. In total there will typically be several hundred separate computers whirring away on a current passenger jet. This is not perhaps the most elegant way to organise things, and party explains why a new Airbus A380 will set you back around $403 millon. I'm sure that as the designs evolve the systems will become more and more integrated.

What they can do


Put simply, the computers can do almost all of the flying. They can do it more safely, comfortably and accurately than the human pilots 99% of the time, and with less fuel used.

A few pieces of key data and the route to fly (usually simply a company route number) are entered into the flight management computers before takeoff, and once airbourne the systems will be switched on and will often fly the entire journey. With some help, many can even land the plane.

When there is a change of plan, air traffic control will contact the pilots with a new flight level or waypoint for example, and the pilots have to check and tap this information into the flight management computer. The information is often sent electronically via a data link, so it's not hard to imagine closing the loop and depriving the pilots of this small task too.

What they can't do

Most landings are still flown manually despite autoland often being available, and this is a skillful and workload-intensive task. The computers do help out enormously by guiding the pilots on the correct glide slope and approach path and warning of various errors or deviations but ultimately the pilots usually land the aeroplane.

Currently, I don't know of any autopilot in use that can take off, so this still remains an important job for the pilots. The take-off is not a difficult manoeuvre, but if things go wrong a quick decision of whether to abort or continue is essential to avoid either running out of tarmac or trying to launch an unflyable plane. Even so, auto take-off cannot be far away.

So as it stands, the pilots take-off and land, and the rest of time the computers fly. The flight crew programs and monitors the systems and handles the communications — and of course they are in command not the machines — but it is the computers actually flying the plane.

So what are the two rather well paid guys in the smart uniforms really doing to earn their keep? The answer lies in that 0.1% of the time when things are not going to plan. The unexpected. The unpredicted. The unimaginable. This is where pilots must step in and do the one thing that computers cannot —  make critical judgements in difficult or unforeseen situations.

Some of these judgements are fairly routine, for example the best way to cross a line of storms or a decision about where or when to divert based on what you expect the weather (and other traffic) to do. Others are completely unique, such as a medical emergency or lunatic passenger or (rarely) major equipment failure.

As our textbook puts it; "The ability of the pilot to evaluate evidence and come to conclusions will, in future, be the only reason for keeping him on the flight deck."

In other words, computers are stupid. They cannot, yet, operate outside of the conditions that the programmer intended and they cannot come up with create solutions to problems.

What about the dog?


And the dog? What is the dog there for? Well, so the joke goes, we will soon be moving to a new flight crew; one pilot and a dog.

The pilot is there to feed the dog, and the dog is there to bite the pilot if he touches anything.

Let's hope not, at least not for a good while yet!