MICHAEL PHELAN / LONDON
A University of Cambridge project aimed at enabling mechanical systems to "learn" to adapt to damage in real time could improve air safety by helping to prevent accidents such as the 1992 loss of an El Al Boeing 747-200 Freighter in Amsterdam, say UK researchers.
Model Predictive Control, a fault-tolerant control system being developed by the university's engineering department, aims to restore as far as possible a system's normal behaviour by using the remaining functioning controls, by comparing actual behaviour to that of a reference model.
Speaking to the UK Institute of Measurement and Control, Jan Maciejowski, head of the control group at Cambridge, said the controller was tested on models representing El Al flight 1862, which crashed after the number three and four engines detached from the aircraft.
The crew managed to keep the 747 flying for about 8min, but lost control at low speed after deploying high-lift devices, and crashed into an apartment block.
"Delft University had very detailed models of the aircraft's behaviour before and after the engine separation," says Maciejowski. "Using simulated pilot control inputs, and by comparing the reference aircraft model's expected behaviour to that observed in the damaged model, the controller attempted to track that behaviour, to some extent hiding the effects of the damage from the pilot," says Maciejowski.
The key to adapting the system - which is already used in the petrochemical industry - to aircraft applications is to speed up the controller's learning phase.
"An adaptation like that required for the flight 1862 scenario would need to be made within a couple of minutes," Maciejowski says, "but this should be possible with today's computing power."
While primary uses for the technique are still in other industries, the group is hopeful that the aviation safety community will show interest, leading to real applications. Maciejowski says the technology may be of particular interest in autonomous unmanned air vehicles, where controllers often struggle to cope with failures.
Source: Flight International
