Crash barrier

A sophisticated autopilot that could land a damaged plane safely is being developed in the UK.

When engines, ailerons or other systems fail, the pilot may be unaware of the extent of the problem and fly into disaster.

The fault-tolerant system could also be used in factories where failing components may cause production to be unnecessarily shut down, when subtle adjustments could keep operations going.

The technology reacts within certain guidelines to data picked up by sensors around the aircraft, said Dr Jan Maciejowski who is leading the research at Cambridge University.

He believes the technology could eventually run a model-predictive control system that in planes would prevent crashes like the 1992 Amsterdam air disaster. An Israeli El Al Boeing 747 crashed into a block of flats, killing the four crew and 39 others. Two of its engines had fallen off after take-off, but the pilot was unable to fly it back to safety.

‘This work is to create a system that can cope with the unexpected. For now we can measure an aircraft’s response by monitoring its roll and yaw rates and from that we can work out a new flight model based on the plane’s behaviour.’

The team developed the control system with the help of a simulation of the Amsterdam crash. Developed by the Delft University of Technology in the Netherlands, the simulation was based on information from the crashed plane’s flight data recorder. The team used this information to test its system.

At the time the pilot thought that the engines had merely shut down, not fallen off. This caused more damage to the wing that he would not have been aware of.

Maciejowski said that his system understood the full extent of the problem and flew the plane in a way that allowed it to reach the runway. ‘We were able to fly the damaged plane past the point where the pilot had lost control, and bring it down to ground level,’ said Maciejowski.

The system relies on optimisation algorithms, rather than a set of rules, to create a new flight model for the damaged aircraft. The algorithm is set to produce a result that will equate to safe flight.

As the computer receives flight data it makes aileron and tail plane adjustments at the rate of 10 times a second to try to match flight behaviour with the desired output of the algorithm.

The team is in discussions with Qinetiq and DSTL about possible applications for the system, initially on military aircraft. While crashes related to aerostructure damage are rare in the commercial world the military are interested in the technology for future unmanned aerial vehicle development.