The Research On System Fault Diagnosis For Multifunction Underwater Vehicle

With the development of the activities in deep sea, the application of underwater vehicles is widespread. Underwater vehicle that works in very tough environment and needs to finish many tasks, therefore, it is very crucial to design a reliable control system to guarantee its safety. To realize the self fault diagnosis of the underwater vehicle's control system is the core symbol of the intelligent level and is the problem that the researchers need to deal with.Research on system fault diagnosis for multifunction underwater vehicle is undertaken in this thesis. The main task is studying the faults of the sensor system and thrusters in order to improve the performance of the control system, therefore, the underwater vehicle can enhance its survivability when it works in the unknown sea environment.Firstly, the wavelet transform is undertaken for the sensor information of the underwater vehicle, and the extreme points of the wavelet transform are used to detect the jumping faults of the signal. In order to decrease the noise's influence, the threshold method is brought in. The disturbance of the noise can be bucked by setting the threshold for the high frequency parameters of the wavelet transform. The experiment results tells that this method can enhance the detect accuracy of the jumping signal. As to the oscillation of the outputs of the positioning sonar, the linear smoothing method is adopted. The cubical curve fitting and kalman filter are designed and comparison experiments are conducted among them, and the experiment results say: underwater vehicle as the specific research plant, linear smoothing method is not only very simple but also direct and effective.To deal with the thruster fault diagnosis, a nonlinear sliding mode observer is designed. A new transfer function is constructed to replace the symbol function in sliding mode observer to overcome the jump which is caused by the symbol function transfers near the origin zone. Therefore, the aim to decrease the buffeting of the sliding mode observer can be realized. Because of the discontinuity and distort of the fault diagnosis method based on threshold, the fuzzy fault diagnosis is discussed to improve the accuracy of the fault diagnosis. A fuzzy residual evaluator is designed to analyze the residual signal and it is applied to thruster fault diagnosis. The simulation results indicate that fuzzy fault diagnosis method can heighten the accuracy of the fault diagnosis, and the robustness of the fault diagnosis can be strengthened.An improved wavelet neural network structure is adopted here. The learning algorithm is deduced. The wavelet neural network which trained well is applied to do the system identification for the underwater vehicle, and the fault diagnosis can be achieved by comparing the outputs between the wavelet neural network and real sensors. The thruster fault diagnosis based on wavelet neural network is finished and satisfied results are obtained. This method can decrease the inputs for the neural network. Therefore, the neural network structure can be simplified.The research results display that the fault diagnosis methods presented here can realize the fault diagnosis of the control system, and it is very useful in the field of underwater vehicle.