Fault detection method of integrated navigation based on LVQ neural network

In the present study, a GPS/CNS/SINS federated filter model is proposed firstly to improve the low accuracy of fault detection in multi-sensor integrated navigation system. On this basis, an LVQ neural network assisted integrated navigation fault detection method is developed for LVQ (Learning Vector Quantization) networks with few design parameters, simple network structure and non-normalized input vectors during usage. The optimal number of neurons in the competitive layer is determined by K-CV (Cross Validation) verification method, and LVQ neural network is used to identify and classify the soft and hard faults added at different times. The simulation results indicate that compared with traditional neural network, LVQ neural network achieves higher detection accuracy (93%) with lower CPU usage. Thus, it is convinced that the study has great engineering significance and practical value.