An INS Fault Diagnosis Method Based on State Chi-Square Test and Optimized GA-BP Neural Network

The failure of the INS will directly lead to navigation errors in an INS/GNSS integrated system. This paper presents an INS fault detection and diagnosis method using state Chi-square test (SCST)and optimized genetic algorithm and Back-Propagation neural network(GA-BP). Firstly, SCST is used to detect fault and the results of SCST is extracted as the feature data of fault diagnosis. Secondly, the selection strategy and calculation method of crossover and mutation probability are optimized to improve the performance of genetic algorithm. Then the optimized genetic algorithm is used for global search to optimally acquire the initial weight and the threshold of the BP neural network. Finally, two optimized GA-BP neural networks are applied to locate the fault and obtain the magnitude and type. The results of the simulation demonstrate that the method is able to diagnose the fault location, magnitude and type with high accuracy.