A Novel Hybrid Model for the Prediction and Classification of Rolling Bearing Condition

Rotating machinery is a key piece of equipment for tremendous engineering operations. Vibration analysis is a powerful tool for monitoring the condition of rotating machinery. Furthermore, vibration signals have the characteristics of time series. Hence, it is necessary to monitor the condition of vibration signal series to avoid any catastrophic failure. To this end, this paper proposes an effective condition monitoring strategy under a hybrid method framework. First, we add variational mode decomposition (VMD) to preprocess the data points listed in a time order into a subseries, namely intrinsic mode functions (IMFs). Then the framework of the hybrid prediction model, namely the autoregressive moving average (ARMA)-artificial neural network (ANN), is adopted to forecast the IMF series. Next, we select the sensitive modes that contain the prime information of the original signal and that can imply the condition of the machinery. Subsequently, we apply the support vector machine (SVM) classification model to identify the multiple condition patterns based on the multi-domain features extracted from sensitive modes. Finally, the vibration signals from the Case Western Reserve University (CWRU) laboratory are utilized to verify the effectiveness of our proposed method. The comparison results demonstrate advantages in prediction and condition monitoring.