Improved EMD Local Energy with SVM for Fault Diagnosis in Air Compressor

The data generated from the machines are generally nonlinear and non-stationary in nature. Extracting relevant information from the data plays a major role in the fault diagnosis. This paper proposes local energy based feature extraction technique derived from improved empirical mode decomposition. Relevancy of feature is examined by correlation method. Support vector machine is used for classification of features. The proposed approach is compared with full signal energy using Hilbert transform on EMD. Improved empirical mode decomposition is used for decomposing acoustic signal into intrinsic mode function in lesser time compared to conventional EMD. Acoustic signals are acquired from most sensitive position of air compressor. Although acoustic signal-based machine health monitoring has not been applied to same extent as vibration signal. In this paper, acoustic signal has been used because of its advantage. The experimental results show the acceptable levels of average accuracy.