Gas turbine fault detection and identification by using fuzzy clustering methods

In this paper fault detection of gas turbine using signal based methods has been studied. First gas turbine data from a precise, validated and existent simulator of gas turbine acquired. This simulator is capable of simulating four common faults of gas turbine. There is one normal class and four different fault classes. Data from sixteen variables had stored. Each class has 780 samples of observations. Feature selection methods have been applied on the data set. Principal component analysis (PCA) and linear discriminant analysis (LDA) used as feature selection methods. Fuzzy C-means and Gustafson-Kessel clustering algorithm used to fault detection and identification of the faults of the gas turbine. Confidence matrix and correct rate methods used to evaluate the clustering performance. The simulation results show that clustering methods has acceptable performance for fault detection and identification of the gas turbine faults. The combination of LDA and Gustafson-kessel clustering algorithm had the best performance as 96.96% correct rate.