A Novel Abnormal Vibration State Recognition Method for Rolling Mill

The vibration of rolling mill is nonlinear and non-stationary. To effectively extract rolling mill vibration signal features for fault diagnosis, a method of feature vector extraction based on empirical wavelet transform (EWT) and fuzzy entropy (FE) is proposed. The method extracts the midpoint between the adjacent maximum values in the frequency domain of vibration signals, adaptively divides the Fourier spectrum, and constructs a suitable wavelet filter group to calculate different intrinsic mode functions (IMFs). Subsequently, calculate the fuzzy entropy of IMFs to be used as the feature vector. In addition, probabilistic neural networks (PNN) classifier is used to identify the types of vibration faults. Experimental results indicate that the accuracy of the method is higher than those based on Empirical mode decomposition (EMD) and multi-scale fuzzy entropy (MFE), under the limited sample condition. Thus, a referential diagnostic method is provided for practical applications.