A neural network approach for helicopter airspeed prediction

Conventional pitot-static airspeed measurement systems do not yield accurate measurements when aircraft speed is below 40 knots. Recent studies have demonstrated that neural network approaches for predicting airspeed are quite promising. In this thesis, a back-propagation neural network is used to predict the airspeed of UH-60A and OH-6A helicopters in the low speed environment. The input data to the neural networks were obtained using the FLIGHTLAB flight simulator. The results obtained by flight simulation were validated by comparison to results of a previous study of the UH-60A helicopter based on actual flight data. The results of the work performed for this thesis show that at sea level the UH-60A low airspeed can be predicted with an accuracy of ﾱ 0.71 knots and ﾱ 0.88 knots for out of ground effect and in ground effect conditions respectively. OH-6A analyses were performed at two pressure altitudes. At sea level the OH-6A airspeed can be predicted with an accuracy ﾱ 0.75 knots when the aircraft is out of ground effect and ﾱ 0.88 knots when the helicopter is in ground effect. At a pressure altitude of 6000 feet OH-6A airspeed can be predicted with an accuracy of ﾱ 0.64 knots for both flight conditions.http://archive.org/details/aneuralnetworkpp109451101