A Novel Bearing Intelligent Fault Diagnosis Method Based on Modified VMD and 1-D CNN

Abstract

Bearings are one of the most critical components of rotating machinery. Timely and accurate bearing fault detection and diagnosis are very important for the reliability and safe operation of industrial systems. Due to the non-stationary, non-linear, and high-dimensional characteristics of the rolling bearing vibration signal, the signal pre-processing and fault feature extraction are particularly important, because they can directly affect the efficiency and accuracy of fault diagnosis. This paper proposes an improved intelligent diagnosis method combining the modified Variational Mode Decomposition (MVMD) and the 1-D Convolutional Neural Network (1-D CNN). This method first uses the traditional Empirical Mode Decomposition (EMD) to adaptively extract the Intrinsic Mode Function (IMF), and then proposes a sensitive IMFs evaluation index to choose sensitive IMFs and then reconstruct the vibration signal. The VMD iterates over the reconstructed signal and searches for the optimal solution of the variational model, determines the frequency center and bandwidth of each component, and realizes the frequency domain splitting of the reconstructed signal and the separation of the modes. Finally, the processed data is divided into train set and test set, and feature extraction training is performed on the training set through 1-D CNN. The trained CNN network uses the test set to evaluate the accuracy of fault diagnosis. The experimental results show that the method can not only accurately decompose the bearing vibration signal and mine the available fault features, but also classify different fault types.