LAUSR

Abstract This paper investigates fault feature extraction for autonomous underwater vehicles (AUVs) with weak thruster fault. When the conventional feature extraction method based on time-frequency domain decomposition is used to extract the weak fault feature of the thruster, the frequency bands of the fault feature and disturbance feature are overlapped, such that it is difficult to extract the fault feature accurately. To solve this problem, a novel extraction method for fault feature is developed based on an optimized sparse decomposition algorithm. Two problems are encountered when directly using the existing sparse decomposition algorithm to diagnose weak thruster fault. The first problem is that during the decomposition of time-domain signals, the accuracy is relatively low. A time-shift operator-based decomposition algorithm is proposed in this study to address this problem. The second problem is that during the extraction of weak fault feature of the thruster, the difference between the fault feature and disturbance feature is small. To address this problem, a feature extraction method based on fault weight matrix is proposed. Finally, pool-experimental verifications are presented.