Thruster fault feature extraction method for underwater vehicle

Abstract

A novel thruster fault feature extraction method for autonomous underwater vehicle(AUV) is presented in this article. When using fractal dimension method to extract the thrust fault feature from state quantity, the noise feature value may be greater than the fault feature value, which lead to failure of fault feature extraction. To solve this problem, this paper proposes an improved method based on empirical mode decomposition, fractal dimension and the short-time higher frequency component positioning algorithm to extract fault features from state quantitiy. In this paper, the empirical mode decomposition(EMD) is used to replace the filtering method of the fractal dimension. The rolling time windows are introduced into the high frequency part of EMD, and the fractal dimension mutation at the time of fault occurrence is captured by calculating the fractal dimension of small samples in each time window. Fault feature extraction is completed by extracting the maximum value of the fractal dimension mutation. When using the fractal dimension method to extract the thruster fault feature from the control quantity, the calculation time is too long. To reduce the calculation time,this paper proposes a fault feature extraction and identification method based on fractal dimension and the support vector regression algorithm(SVR). In this paper, the ideal SVR model between control quantity and fractal dimension for different thruster fault is established, and the fault feature is extracted from the original data by SVR model.The effectiveness of the improved method is verified by the pool experimental data of an underwater vehicle.