Turbine Cavitation State Recognition Based on BP Neural Network

摘要

Aiming at the difficulties in identifying the cavitation state of hydraulic turbines, a method of identification the cavitation state of hydraulic turbines based on BP neural network is proposed. Firstly, the acoustic emission(AE) signals under different cavitation states are collected by AE sensors. After noise reduction pretreatment, the characteristics of the cavitation signals are extracted by parameter analysis and lifting wavelet energy analysis, and the feature vectors under different cavitation states are constructed. In order to verify the accuracy of feature vector selection, the feature vectors in different cavitation states are input into BP neural network for state recognition. The results show that the accuracy of state recognition is as high as 88%. The proposed method can recognize different cavitation states.