José Sánchez-Alarcos Ballesteros

 

Now is time for deep thought about Air France 447 crash

There are still so many questions that it is hard to say which lessons we are going to learn from Air France 447. However, we should consider the thoroughness of investigation that this accident requires.

In more than a century of aviation history, we have only two pieces of air accident research acclaimed worldwide as outstanding: Air Ontario 1363 and Mount Erebus. These events have a common point: their investigators went beyond visible causes and, even when their professional status could have been comfortable with limited conclusions, they both decided to seek the truth.

The research of Los Rodeos was very good but the difference between the outstanding quality of the work to determine what exactly happened and the trivial conclusions drawn from this work make this investigation far from masterly. The same can be said of the research into the mid-air explosions of the Comets. It was a very good job but strictly limited to technical variables.

There are some other accidents where it is easy to think that the investigation should have gone deeper. For instance, would the accident of Swissair-111 have happened if the flight had been performed by a DC-10 instead of the newer MD-11? An analysis of the workload for two or three people in the situation they confronted could suggest that the DC-10 – the older model – could have been better equipped to deal with the problem. If that seems reasonable…could not we reach conclusions about secondary effects of technological evolution? What about AeroPeru 603? In an older plane, would the pilots have been trapped by tunnel-vision centered in as trivial a variable as “rudder-ratio”? What about AA-965 and the automation of navigation?

The analysis of Air France 447, if the present hypothesis is confirmed, could be made in two different ways:

  • A minor bug in the design, under external and exceptional environment, triggers the sequence of facts leading to the accident. Improve the design and compensate those affected.
  • A minor bug in the design triggers some automatic reactions from the plane starting a snowball effect, so that the pilots are either unable to make the right decision or prevented from carrying it out. There is a systemic problem and the conclusions could be far-reaching.

Systemic problems make investigators nervous because they mean questioning rules and practices already accepted by the market and regulators. That is why many of them could take the easy way out: the investigation is limited to the most immediate evidence.

Furthermore, systemic problems do not have a clear and easy-to-identify guilty party. Everyone, from the manufacturers to the operators and regulators tries to conduct their business in the right way but this “right way” is included in a hard to handle global dynamic. If any of the stakeholders decides to act by himself, he is probably going to lose most or all of his market share.

Even if the theory of a bad design proves accurate it would be unfair to point to Airbus or any other manufacturer as careless. Manufacturers are in a permanent war against inefficiency because efficiency means cost-reduction and operators convert cost-reduction into lower prices. Since the competition is fierce, operators evaluate very carefully the efficiency they can get from every plane and this is a critical variable when making decisions about fleet composition. Since passengers are convinced that aviation is safe per se, the one who rejects the efficiency war, loses it and, hence, the market.

All of this is not new. Charles Perrow, in his book Normal Accidents, advised against something that he called ‘tightly-coupled organizations’ due to the risk linked to a search for efficiency. Jens Rasmussen established a rule as the base-line for safety maintenance: The operator has to be able to run cognitively, the programme that the machine is running. More recently, James Reason in The Human Contribution pointed to different situations where only people could save a situation. To make this possible requires the condition established by Rasmussen: a deep knowledge of the system or plane. The situation is exactly opposite and is known as ‘automation paradox’: As complexity of a plane increases, required training to fly it decreases. Furthermore, this could be one of the main incentives for automation.

Against this ‘alarmist’ position, actions of manufacturers and operators, accepted by regulators and passengers, have shown an opposite view:

  • Perhaps, not having a radio-operator in a plane is profitable but accidents like Avianca 052 or the requirement from ICAO about English say that something has been lost with the radio-operators.
  • Perhaps, not having a flight-engineer is profitable but accidents like Swissair 111 show situations where they were missed and in some others as Los Rodeos a more assertive position from the flight engineer could have avoided the accident.
  • Probably, the cross-crew rating is economically justified but there are many situations of negative-transfer where pilots have produced behaviours adequate for planes different from the one they were flying at that moment.
  • Probably, long-range flights with twin planes are justified but it is hard to get enough statistical information about the ability, and reliability, of ONE engine working solo to get the plane flying with asymmetric thrust.
  • Probably, the reduction of training time in pilots is justified but the increasing technological complexity and labels as ‘need-to-know’ and ‘nice-to-know’ can lead us to situations of ‘If-only-I-had-known’.

We cannot ignore that, despite this, safety indicators have been improving for a long time but this is only a part of the reality and could drive us to get the wrong message.

The accident rate is lower but it is not possible to ignore that the typology of accidents seems to be different and that requires a specific analysis.

The label ‘situational awareness’ in the past was usually applied to navigation and ‘situation’ was usually a synonym for ‘spatial situation’. Nowadays, the same label points to a wider reality that sometimes means confusion not related with navigation but with working-modes. There is improvement because things that designers and pilots already knew are done in better and more efficient ways. At the same time, strange situations can be harder to manage and that is why new accidents arise.

It would be nonsense to accuse the manufacturers because they play a game where they do not control all the pieces. It is enough to observe the two biggest manufacturers to see that they have imitated each other to get the most efficient solution:

  • The A-310 solution, without a flight engineer was followed by Boeing 757 and 767, even when the wide cabins able to have a flight engineer were not in the final design.
  • When Boeing introduced twin planes for long-range flights, the sincerity of the complaints by Airbus was tested when they started to sell the 330s and now the 350s.
  • When Airbus introduced FBW (fly-by-wire) technology in the 320s, Boeing showed a more conservative position, but only for a short time as the 777 and the new-brand 787 easily prove.

The path is clear: efficiency wins and, with it, automation, complexity and training requirements arriving to extremes as Multicrew Pilot Licenses.

Some years ago, when automated planes started to appear in the market, an old pilot said something that became notorious: 'I’m ready to fly a plane as automated as you want…as far it has a red button that converts it instantly into a DC-9.'

It seems commonsense to have this possibility that is, in colloquial terms, something quite similar to the Rasmussen rule that requires two conditions to be met:

  • The pilot has to be able to fly a traditional plane like the DC-9.
  • The plane, once all the software is out, needs to have a design making it possible to be flown by hand without big stability problems or lack of basic information.

Are these two conditions met in the most modern planes and with the present training practices?

It would be good if the investigation of a big accident like the AF-447 could go beyond the change of a component. It would be good if someone decided to use this opportunity to make an overhaul of the air transportation system, including the role of the passenger who lives in a kind of Matrix believing that aviation is a non-risk activity instead of a controlled-risk activity.

If the passenger were conscious of this fact, perhaps he would be start to be interested about who controls the risks, and how.

The present path of growing automation and complexity and decreasing training is not good but there is not an individual actor prepared stop it because that would mean losing position in the market.

Errors in efficient organizations and efficient machines are also efficient because they use the same channels to produce their effects; those of the normal operation. Now, the accident of AF-447 appears and, if the expected cause is confirmed, while it won’t be the first time that this dynamic produces a bad outcome, it could be the worst to date.

All the stakeholders have a big decision in their hands: to change a little piece of a plane and wait for any other little piece that, under an unforeseen environment, shows a defective design and starts an snowball effect, or to challenge the whole system.

Efficiency dynamics could be evil because nobody has the resources to stop the war. Many of the readers can remember the cold war where it was impossible to stop the armaments race because the one doing it could lose. The one deciding to fix the situation by himself loses. The defective piece could be today from Airbus and tomorrow from Boeing but the problem is far beyond Airbus or Boeing. There is a race that somebody – everybody – has to stop.

The Rasmussen rule keeps its validity. Applied to aviation, it means that the pilot has to know every single moment what the plane is doing, how it is doing it and why, and if required, he has to be able, without limitations to get a full control of the plane. To do that, ‘user knowledge’ is not enough.

We have to know how to differentiate false from authentic decreases in complexity. Many pilots fall in love with the screens because of the quantity of information they can obtain from them. Boeing pointed out that the model 747-400 had decreased controls and indicators by two thirds when compared with earlier models. That is the false complexity reduction and it is very easy to show with an example: is it easier to handle a radio with a single button and a lot of modes or one that, for the same functions, has 20 single-mode buttons? Probably, the second one is going to be easier.

The advances have been important and they have been made following a path with big secondary effects. To go back is impossible and, perhaps, not even convenient. However, the course can be changed and this change needs to be by everyone, including the passenger.

José Sánchez-Alarcos Ballesteros

 

 

◄ This article was originally published by SEPLA, the Spanish Airline Pilots Association.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

description

Improving Air Safety through Orgnaizational Learning

by José Sánchez-Alarcos Ballesteros

Author's blog

 

 

 

 

 

 

For more on Normal Accidents and Normal Accident Theory read The ETTO Principle: Efficiency-Thoroughness Trade-off, by Erik Hollnagel

 

 

description

The Human Contribution, by James Reason. The author takes a selection of heroic recoveries and uncovers what is special about them and the heroic acts involved.