Cutting human error in the air

It's often said that flying is the safest way to travel, and in the last 20 years there has been a steady reduction in the accident rate of commercial flights. Incidences of mechanical failure now account for a small percentage of total accidents, thanks mainly to advances in reliability and structural integrity. But around 75% of accidents are caused by the same factor: human error. “Air crashes due to mechanical failure have gone right down, but we've seen no decrease in those caused by human error,” says Paul Salmon of Brunel University. Salmon, who is Human Factors research fellow at Brunel's System Engineering and Design Department, has helped to devise a way of identifying and eliminating elements of cockpit design that could increase the chance of mistakes – dubbed 'design-induced' pilot error. The methodology, called HET (Human Error Template), is adapted from similar techniques used in the nuclear and petrochemical industries, but is specific to the needs of aviation. “If you can understand the characteristics of the task and the technology, you can predict the errors that will occur,” he says. “You describe the tasks within a system, then apply these techniques.” Within aviation, these methods had been used in air traffic control. But the continuing number of design-induced pilot errors led the Federal Aviation Authority to insist on a similar system for planes themselves. A historical example was seen in 1940s planes, in which the toggle switch for lowering the landing gear looked identical to the one that lowered the wing flaps. To make things worse, the two switches were situated side-by-side in the cockpit. “Lots of pilots mistakenly changed the flaps, then crash-landed without the landing gear down,” says Salmon. Fast-forward by 50 years and similar problems still happen. The crash of an A320 plane at Strasbourg Airport – in which the crew mistakenly set the rate of descent too high – was blamed on a dual-function control: pushed one way, it controlled rate of descent; pushed the other way, it controlled the angle of descent. Using HET, says Salmon, would have flagged up the inherent danger of using a dual-function control. The method is a checklist-type approach that has 12 'error modes' – such as 'failure to execute', 'wrong task executed' and 'task performed too early/late'. The researchers were able to predict what errors might occur for a specific flight task and cockpit – and was able to check this against data for actual errors, which was gathered from pilots. HET would be used to 'walk through' the design in a methodical fashion, predicting potential errors that could occur based on the way in which components are designed. Common 'low level' pilot errors include setting airspeed or altitude incorrectly – so any design improvements would aim to make these tasks for straightforward to perform. The project to develop HET – whose other partners include Marshall Ergonomics and the Universities of Cranfield, Lund and Limerick – is now finished. The next step, says Salmon, would be to do more rigorous testing on it. “We gathered data for a specific flight task and cockpit,” he says. “We hope people will take it up and conduct some more validation studies – on a wider range of cockpits and tasks.” Pointers The 'Human Error Template' (HET) helps designers to 'walk through' a cockpit design in order to identify potential problems The method is a 12-step checklist that is applied sequentially to flight tasks – such as setting airspeed or altitude The project is finished, but could be refined by more validation studies