Fault diagnosis of turbine generator vibration based on wavelet packet and data-driven

Vibration is one of the most common faults of turbine-generator; traditional signal processing method can not effectively extract the eigenvector of the fault signals. In this paper, using wavelet packet technology analysis the spectrum of turbine vibration signals, extracting the frequency eigenvectors of the signal and use them as the classification data; Using the representative points on behalf of classes, extracting the quantified values of the contribution which the indicators made to classifications, and using them as the weights to the definite the weighted distances between sample points and representing points, what's more, taking the weighted distances in place of Euclidean distances, establishing the iterative algorithm to search optimal representative points, moreover, the algorithm steps are given. Therefore, we can apply this algorithm to diagnosis the fault for turbine generator. The experiment results shows that the proposed method which diagnosis the vibration fault for the turbine generator fault is pretty effective, in addition, reducing the workload and the time of the diagnosis process.