Gas turbine shaft unbalance fault detection by using vibration data and neural networks

This study presents fault detection of a heavy duty V94.2 gas turbine which has 162.1 MW nominal power and 50 Hz nominal frequency and is located at Pareh Sar power plant, Gilan, Iran. For this purpose stored data include measurements of relative and absolute vibration of shaft bearings in both turbine and compressor sections. Signal processing techniques and mathematical transformations are used for feature extraction, as well as supervised and unsupervised methods for dimensionality reduction. Finally neural networks are employed for classification task and fault detection results for different methods are compared and discussed. Proposed techniques show zero FAR and MAR, when PNN is used with PCA or when MLP or RBF is used with LDA for dimensionality reduction.