Feature extraction and severity identification for autonomous underwater vehicle with weak thruster fault

Weak thruster fault feature extraction and fault severity identification methods for autonomous underwater vehicle (AUV) are studied in this paper. One of the traditional methods of fault feature extraction is based on wavelet transformation + modified Bayes (MB), then the grey relation analysis (GRA) method is used to identify the fault severity of the thruster. Above methods are efficient for strong fault of thruster, but for weak fault, problems exist in these methods are the ratio of fault eigenvalues to noise eigenvalues of the extracted feature is low and the identification accuracy of fault is not satisfactory. To overcome the above deficiencies, resonance-based sparse signal decomposition (RSSD) together with stochastic resonance (SR) + MB is proposed for thruster weak fault feature extraction. Euclidean distance together with grey relation (GR) method is proposed to promote the identification accuracy of weak thruster fault. Finally, the pool experiments are performed on Beaver II AUV, and the effectiveness of the proposed method is demonstrated in comparison.