Fault diagnosis of induction motor using decision tree with an optimal feature selection

Time vibration signals are measured to extract a feature set for fault diagnostics of induction motor. Feature selection by decision tree and genetic algorithm (GA) is presented in this paper to remove irrelevant information in the feature set. New data with the selected features is used to train a decision tree, which is an expert system for classification. Testing results show that systems with selected features can reliably diagnose different conditions of induction motor, which has better performance compared to original one without feature selection.