Neural network-based online fault-tolerance design for flight control systems

To enhance the reliability of flight control system, an approach of sensors online fault-tolerance based Radio Basis Function neural network (RBFNN) was proposed. The RBFNN was designed by using improved gradient optimization algorithm to make progress in NN learning speed and mapping capability. Based on network approximation, online adaptive neural network identification was built. Considering the characteristic of real-time and closed loop feedback control system, a set of identification models were applied for sensors fault isolation and signal reconfiguration to ensure stability of the closed loop feedback system. The scheme was demonstrated through simulations applying the flight control system of a fighter. Results show that sensor online fault isolation and reconfiguration are achieved.