A Fault Detection Method with Ensemble Empirical Mode Decomposition and Support Vector Data Description

In order for the fault detection of processes with noise and nonlinearity, a method based on Ensemble Empirical Mode Decomposition (EEMD) and Support Vector Data Description (SVDD) is proposed. In this work, EEMD-based denoising method is utilized to remove the noise from the original dataset. The SVDD model is then developed to handle the nonlinear data for fault detection. The proposed method contains three steps. Firstly, the original dataset is decomposed into a series of Intrinsic Mode Functions (IMFs) by the EEMD method. Each IMF characterizes the corresponding scale information of the data. Secondly, the original data is reconstructed using the partial reconstruction denoising method. Only the relevant IMFs which mostly contain useful information are retained, and the IMFs that primarily carry noise are discarded. The optimal number of relevant IMFs is selected based on the Signal-to-Noise Ratio (SNR). Finally, the SVDD model is constructed on the reconstructed data to detect faults. The effectiveness of the proposed method is demonstrated by a numerical example. The results show the proposed method performs better compared with other existing methods.