An intelligent pattern recognition method for machine fault diagnosis

This paper proposes a novel pattern recognition method for rotating machine fault diagnosis. In this work, the proposed method firstly employs the local mean decomposition (LMD) algorithm to decompose the raw vibration signals into a small number of product functions (PFs), and then, the energy of each useful PF is computed and normalized to form an original feature vector, so an original data table about machine faults can be constructed via these feature vectors; subsequently, the table are processed using kernel principal component analysis (KPCA) to extract the principal feature and compute the corresponding feature values; lastly, the low-dimensional features and their values are input into least squares support vector machine (LS-SVM) for fault identification. The experimental results show that the proposed method can effectively extract the fault features and can accurately identify the different machine faults.