Fault Diagnosis of Turbine Based on Data-Driven

A large number of real-time data and fault history data of turbine could be got through DCS, but the ability of data processing is lagging, a new method of fault diagnosis based on supervision of data-driven for turbine is introduced which is. The method of classification replace given data with points, using the weighted distance in place of Euclidean distance, establishing the iterative algorithm to search optimal representative point. The algorithm steps are given. According to the number of inconsistent samples points in different types of faults, the complexity relations of fault classification data is divided into the simple data, complex data.This paper points out a new algorithm of fault diagnosis which is based on representative points clustering; we could use the algorithm to analyze the turbine fault.