Fault Diagnosis Method For Nonlinear System Based On PFCM Algorithm With Mahalanobis Distance

With the development of industrial technology, there is a trend that the industrial process is becoming more and more complex. Moreover, even an error may lead to a huge economic loss or even casualties. The fault type can be diagnosed by fault diagnosis technology which can prevent equipment damage and casualties caused by a sudden fault. The modern industrial process is a complex nonlinear system in which precise mathematical models are hard to be established so that the fault diagnosis technology has been worldwide concern due to the fact that it’s does not depend on mathematical models.Fuzzy clustering is a method that does not depend on the mathematical model. In the process of diagnosis, the system operating mode can be obtained from the monitoring data. Based on the research of existing clustering algorithms, it is found that these algorithms can’t meet the needs of fault diagnosis of nonlinear systems. This paper proposes an algorithm of probabilistic-fuzzy C-mean clustering based on Mahalanobis distance(PFCM-M), and the simulation results prove its effectiveness. The main contents of this paper are as follows:1. This paper first discusses HCM FCM PCM and PFCM algorithms. Experiments prove that PFCM algorithm can solve the noise sensitivity defect of FCM and the coincident clusters of PCM. And because it is based on Euclidean distance, so it is not good for linear and elliptical distribution of data clustering.2. In this paper, the PFCM-M algorithm is proposed to solve the problem that the PFCM algorithm is difficult to deal with the linear and elliptic distribution data. For complex nonlinear system, this paper proposes a method that combines principal component analysis(PCA) with PFCM-M algorithm, and utilizes the classic TE process to prove PFCM-M algorithm not only better than PFCM algorithm in the clustering effect, but also solve PFCM algorithm difficult to identify the unknown fault.