Fault-diagnosis method for INS/GPS integrated navigation system based on PNN and genetic algorithm

Aiming at the characteristics of no redundant observation information and possible faults of INS/GPS integrated navigation, an improved fault diagnosis method for integrated navigation based on state chi-square test and adaptive probabilistic neural network (APNN) is proposed. Through monitoring the state vectors of the integrated navigation system, the normalized results of state chi-square test can be extracted as the input for APNN to identify the pattern of the fault, and then implement fault locating and isolating. The APNN is improved with GA algorithm to achieve the optimum number of the pattern layer neurons and match the smoothing parameters for each transfer function of neurons, which improve the generalization ability and diagnostic accuracy of the neural network. Simulation results show that the new method can not only identify the faults efficiently, but also improve the reliability and safety of the system.