A fault detection algorithm for turbopump based on lifting wavelet and LMS

A fault detection algorithm for turbopump based on lifting wavelet and LMS is proposed in this paper, to deal with the health monitoring of turbopump which is of high failure rate. Lifting wavelet transform is used for signal decomposition and single-scale reconstruction. And fault feature is extracted from the weighted average energy of approximation signals and detail signals, and its sequence is filtered by LMS (Least mean square error adaptive algorithm). After calculating the ratio between the fault feature point and the mean value of its neighboring local feature points, the failure of turbopump can be identified according to the change of the ratio. This algorithm is verified by using simulation vibration acceleration signals of a certain type of turbopump to simulate the process of hot commissioning. The results indicate that the algorithm presented in this paper can effectively detect the failure of turbopump with good performance of real-time and accuracy.