Adaptive probabilistic neural network fault-diagnosis algorithm for SINS/GPS integrated navigation system

In order to insure the SINS/GSP integrated navigation system work stably and reliably, a new method based on adaptive probabilistic neural network (APNN) was put forward to detect and isolate the fault of integrated navigation system. State chi-square test (SCST) was used to real-time monitor the navigation system. The components of STSC were used as the inputs of APNN to determine the type of fault. Gauss function was chosen as the excitation function of APNN. Since the width of Gauss kernel function significantly affects the generality of network, the cross-validation method was used to estimate the smooth factor, and the particle swarm optimization (PSO) algorithm was adopted to optimize them. The simulation results show that this algorithm can accurately detect the fault of navigation system, and can improve the reliability and safety of the system.