Research On Fault Diagnosis Of AUV Integrated Navigation System Based On Multi-scale Analysis

As one of the main equipment of underwater exploration and development for human beings,Autonomous Underwater Vehicle(AUV)has received increasing attention.The safety of AUV is especially important because of the long time working without cable,and fault diagnosis is the key technology for AUV to ensure that.The correctness of the navigation system is the primary prerequisite for the normal operation of the AUV Research on the navigation sensor fault diagnosis technology has important research significance and practical value for improving the safety of AUV and promoting the practical process.This paper takes the DR/BDS integrated navigation system of an AUV as the research object,analyses and gives the common fault types of navigation sensors,simulates the data of each fault type based on the actual experimental sensor data by using MATLAB,analyses the influence of fault on the integrated navigation system,and gives the multi-scale analysis method of data from the perspective of multi-scale analysis theory.Aiming at the multi-scale characteristic analysis of AUV navigation sensor fault signals,the performance characteristics of various types of faults on different scales are given,and a signal extraction method based on multi-scale entropy is proposed to realize the quantitative extraction of fault features.Secondly,the extracted multi-scale fault feature vectors are identified and classified by using the neural network to realize the fault diagnosis of AUV navigation sensors.The performance of BP neural network and wavelet neural network in AUV fault diagnosis is compared through simulation experiments.The experimental results show that the learning ability of wavelet neural network is stronger than that of BP neural network,but the performance can not accurately diagnose faults.Finally,in order to make the learning performance of the network meet the requirements of AUV navigation fault diagnosis,the wavelet neural network is improved.Momentum-driven adaptive learning rate algorithm is added to optimize the training speed of the network,and Levenberg-Marquardt algorithm is added to greatly improve the learning performance of the network.According to AUV navigation sensor fault situation,an improved LM wavelet neural network fault diagnosis method combined with multi-scale entropy fault feature extraction method is feasible and effective.