Fault Diagnosis of Turbine Generator Vibration Based on Supervision of Data-Driven

Vibration detection system of turbine generator can obtain large amounts of data resources; however, there are no effective methods to excavate useful knowledge from these massive data. In this paper, a new approach for fault diagnosis of turbine generator based on supervision of data-driven is proposed. This algorithm begin with the given classification data, using the representative points on behalf of class mean values, using the weighted distances in place of Euclidean distances. And establishing the iterative algorithm to search the optimal representative points, what´s more, the algorithm steps are given. Finally, employing the method to identify 3 kinds of common fault states for turbine generator, the experiment results shows that this algorithm can solve the problem of fault classification, it provide us an effective way to diagnosis the fault for turbine generator.