Neural Network Limit Avoidance System for Rotorcraft

Rotorcraft limit envelopes are complex and difficult for pilot compliance. Exceedances are unsafe and shorten useful life of the aircraft, while piloting that isn't close to limits cannot fully exploit performance capabilities. Accurate, automatic limit-onset prediction/detection would alert the pilot and trigger automatic limit-avoidance control to prevent exceedances. Neural network (NN) techniques offer the enabling technology for real-time prediction and detection. Self-designing controller (SDC) technology offers a versatile means for limit avoidance. The proposed NN synthesis algorithm will select the most relevant sensed variables in a database of flight recordings and evolve a constrained-complexity, nonlinear polynomial neural network. This NN will recognize the proximity of envelope limits. The NN limit-onset predictor/detector will be demonstrated in Phase I and combined with the SDC limit-avoidance controller in Phase II. Barron Associates, Inc. is teamed with Stewart Hughes Ltd., the worldwide leader in systems for rotorcraft health and usage monitoring, and with Analytic Services Inc., who will provide expertise of an experience helicopter test pilot/engineer. Commercialization of the work will proceed concurrently with Phase II and III.