Roller Bearing Fault Diagnosis Based on Integrated Fault Feature and SVM

Purpose The purpose of this paper is to provide high accuracy and rapid fault detection simultaneously using integrated fault features and support vector machine. Methods This paper first proposes a new fault feature extraction approach that separates the signals of integrated fault features (IFF) rapidly. The singular values are obtained by singular value decomposition (SVD) of Hilbert spectrum which is attained by intrinsic mode functions (IMFs) through empirical mode decomposition (EMD), and then combined with the permutation entropy (PE) of signal to form the IFF vector. Next, the support vector machine (SVM) is proposed as the classifier to further enhance the fault diagnosis performance. Particle swarm optimization (PSO) is employed in this paper to optimally tune the parameters of SVM. Results On two public data platforms, the classification accuracy of IFF with SVM can reach 98.1% and 99.43%, which is 19.7% and 9.4% higher than that of single feature value with SVM at most Conclusion In this paper, a novel IFF extraction method has been proposed to improve the computational efficiency and accuracy of fault diagnosis for roller bearings. At the same time, the proposed method has good classification capability for various types of roller bearings and different sample number. This result is helpful to provide a new way of feature vector selection.