Neural Network Based Low Airspeed Sensor

A neural network based low airspeed sensor is developed using flight test data from an SH-60 helicopter specifically collected for this purpose. Inputs to the network are parameters that are typically monitored by a flight data recorder or a health and usage monitoring system. Approximately five percent of the data is utilized for network training. Prediction results indicate that accurate prediction of low airspeed magnitude and direction are possible with a root-mean-square error just under three knots. The neural network based low airspeed sensor can be utilized in real time to improve pilot situational awareness or implemented as a non-real time algorithm to improve assignment of fatigue damage through enhanced structural usage monitoring in the low airspeed flight regime. The neural network based low airspeed sensor provides a viable and affordable alternative to advanced mechanical low airspeed sensor systems.