Fault diagnosis approach based on approximate entropy feature extraction with EMD and support vector machines

A fault diagnosis approach based on approximate entropy feature selection with empirical mode decomposition and support vector machines is proposed. Firstly, the EMD method is used to decompose the vibration signals into a number of intrinsic mode functions. Secondly, approximate entropy of these intrinsic mode functions which contains main fault information are computed and obtained. Finally, the approximate entropy serve as feature vectors to be input to the multi-class support vector machines and the work conditions and fault patterns are identified by the output of the classifier. This fault diagnosis method is used to recognize the three common faults of ball bearings. The results show that this approach can effectively classify the working conditions and fault patterns of ball bearings accurately even when the number of samples is small.