Application of Wavelet Packet and Data-Driven in Fault Diagnosis for Hydropower Units

Vibration is one of the most common faults for hydropower units; traditional signal processing methods can not effectively extract the eigenvectors of the fault signals. In this paper, using wavelet packet analyze the spectrum of the vibration signals for hydropower units, extracting the frequency eigenvectors of the signal and use them as the classification data, taking good advantage of the feature of the data and extracting the quantified values of the contribution which the indicators made to classifications, whatpsilas more, taking the weighted distances in place of Euclidean distances, establishing the iterative algorithm to search optimal representative points. Therefore, we can apply this algorithm to diagnosis the fault for hydropower units. The experiment results shows that it is very appropriate for us to use the proposed method to diagnosis the vibration fault for hydropower units.