Bearing fault diagnosis method based on ICEEMDAN-PCA-CNN

In scenarios involving the heavy-load operation of large machinery and high-frequency vibration environments, the raw bearing fault signals collected during equipment operation are characterized by high levels of noise. Existing data-based bearing fault diagnosis models have weak noise resistance. To address this issue, this paper proposes a bearing fault diagnosis method based on Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN), Principal Component Analysis (PCA), and Convolutional Neural Network (CNN). First, the ICEEMDAN algorithm is used to decompose the bearing acoustic signal and extract the intrinsic mode functions. Then, PCA is applied to reduce the dimensionality of the high-dimensional feature data. Finally, a CNN model is constructed to classify the processed data and achieve fault diagnosis. The method is validated using experimentally obtained bearing fault signals, and the results confirm the effectiveness of the proposed method for early fault prediction of fan bearings.