Fault Diagnosis Of Gas Turbine Based On Fuzzy Clustering

As gas turbine’s high efficiency, easy to assemble and maintain, and other advantages, it’s widely used in industry, energy and transport fields. Diagnose and solve fault timely and correctly, develop related technique, both would bring huge benefit to social economy. With modern technique, especially the information technique and computer science, many experts and scholars did lots of work in this field, and made a big progress.However, the traditional fault diagnosis method of gas turbine is mainly focus on tremble signal, and processes it with wavelet or nerve network technique. With this method, can we only solve certain faults that would cause abnormal tremble signal, but can do nothing to deal with other faults related with temperature or pressure.Research shows that the thermal parameters of gas turbine such as temperature, pressure, flow, have very deep relationship with the run-time state. Also, the classical fault, including cautery, begrime and damage, will lead to thermal parameter change. This paper directly use the thermal parameter sampled in run-time, classify it into different clusters, which can avoid the complex mathematical modeling.Considering the complexity of the gas turbine, faults’influence with each other, and the noise inside and outside as well, it is necessary to use the fuzzy technique the describe the faults which has a ambiguous boundary. This paper introduces fuzzy cluster into fault diagnosis of gas turbine, which use the fuzzy method to cluster the run-time data, and get one standard normal pattern and several standard fault patterns. Then recognize the data, judge pattern it belongs to, and can get the result if any issue occurs, then decide which fault it is and gives its possibility.In fuzzy clustering, considering the characteristic thermal parameters of gas turbine, this paper applies the EFCM algorithm in fault diagnosis of gas turbine, and also improve it via adding weight value of minimum and area width, which makes it more accurate, and makes the iteration compute of cluster center not necessary, and reduces the time complex. Result shows that it can correctly cluster different faults. And also compare the EFCM and FCM, result indicates that EFCM is better than FCM, and is feasible and suitable in online fault diagnosis.