Author(s)

JinYuan Hu

Affiliation(s)

School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan 430065, Hubei, China.

Corresponding Author

JinYuan Hu

ABSTRACT

The rapid development of deep learning has brought transformative advances to intelligent fault diagnosis, providing powerful end-to-end feature learning capabilities that enable more effective analysis of rolling bearing vibration signals. However, conventional convolutional neural network (CNN), with their fixed architectures, have difficulty capturing the dynamically changing time-frequency features of vibration signals. In addition, most existing models lack effective mechanisms to suppress noise and vibration interference during monitoring, leading to a marked drop in diagnostic accuracy under non-stationary and noisy conditions.To improve the model’s ability to process non-stationary signals, this study introduces a multi-module diagnostic framework, VMD-CNN-Transformer, which integrates Variational Mode Decomposition (VMD), CNN, and Transformer architectures. The framework first applies VMD to decompose the vibration signals into representative intrinsic mode functions, enhancing the multi-scale representation of the original signals. The CNN module then extracts key local features and integrates multi-scale information. Finally, the Transformer captures long-range dependencies, allowing detailed characterization of complex fault patterns.Comparative experiments on benchmark datasets, including CWRU, XJTU, and DIRG, show that the proposed method achieves superior robustness and generalization under challenging conditions with noise and varying operating states. The framework outperforms mainstream methods and provides a novel technical solution for intelligent industrial equipment monitoring, demonstrating strong potential for practical engineering applications.

KEYWORDS

Rolling bearing; Variational mode decomposition; Convolutional neural network; Transformer; Fault diagnosis

CITE THIS PAPER

JinYuan Hu. Intelligent fault diagnosis of rolling bearings based on VMD-CNN-Transformer. World Journal of Engineering Research. 2025, 3(2): 50-56. DOI: https://doi.org/10.61784/wjer3029.

REFERENCES

[1] Yang Y, Zhai J, Wang H, et al. An Improved Fault Diagnosis Method for Rolling Bearing Based on Relief-F and Optimized Random Forests Algorithm. Machines, 2025, 13(3): 183-183.

[2] Yan H, Shang L, Chen W, et al. An adaptive hierarchical hybrid kernel ELM optimized by aquila optimizer algorithm for bearing fault diagnosis. Scientific Reports, 2025, 15(1): 11990-11990.

[3] Liao W, Fu W,Yang K, et al. Multi-scale residual neural network with enhanced gated recurrent unit for fault diagnosis of rolling bearing. Measurement Science and Technology, 2024, 35(5).

[4] Feisa T T, Gebremedhen S H, Kibrete F, et al. One‐Dimensional Deep Convolutional Neural Network‐Based Intelligent Fault Diagnosis Method for Bearings Under Unbalanced Health and High‐Class Health States. Structural Control and Health Monitoring, 2025, 2025(1): 6498371-6498371.

[5] Bharatheedasan K, Maity T, Kumaraswamidhas L, et al. Enhanced fault diagnosis and remaining useful life prediction of rolling bearings using a hybrid multilayer perceptron and LSTM network model. Alexandria Engineering Journal, 2025: 115355-369.

[6] Li X, Ma J,Wu J, et al. Transformer-based conditional generative transfer learning network for cross domain fault diagnosis under limited data. Scientific Reports, 2025, 15(1): 6836-6836.

[7] Zhilin D, Dezun Z, Lingli C. An intelligent bearing fault diagnosis framework: one-dimensional improved self-attention-enhanced CNN and empirical wavelet transform. Nonlinear Dynamics, 2024, 112(8): 6439-6459.

[8] Sahu D, Dewangan K R, Matharu S P S. Hybrid CNN-LSTM model for fault diagnosis of rolling element bearings with operational defects. International Journal on Interactive Design and Manufacturing (IJIDeM), 2024: 1-12.

[9] Sarunyoo B, Pradit F, Chitchai S, et al. Adaptive meta-learning extreme learning machine with golden eagle optimization and logistic map for forecasting the incomplete data of solar irradiance. Energy and AI, 2023: 13.

[10] XiaX, WangX, ChenW. A Hybrid Fault Diagnosis Model for Rolling Bearing With Optimized VMD and Fuzzy Dispersion Entropy. International Journal of Rotating Machinery, 2025, 2025(1): 7990867-7990867.

[11] Chen H,Yu Y . Acoustic Emission Diagnosis of Rolling Bearing Faults based on Optimized VMD-Transformer. Frontiers in Computing and Intelligent Systems, 2025, 11(3): 19-24.

[12] Wang Y, Zhu K, Wang X, et al. An extended iterative filtering and composite multiscale fractional-order Boltzmann-Shannon interaction entropy for rolling bearing fault diagnosis. Applied Acoustics, 2025: 236110699-110699.

[13] Dragomiretskiy K, Zosso D. Variational Mode Decomposition. IEEE Transactions on Signal Processing, 2014, 62(3): 531-544.

[14] Liu T, Diao F, Yao W, et al. Study on Motion Response Prediction of Offshore Platform Based on Multi-Sea State Samples and EMD Algorithm. Water, 2024, 16(23): 3441-3441.

[15] Chen Q, Zhang F, Wang Y, et al. Bearing fault diagnosis based on efficient cross space multiscale CNN transformer parallelism. Scientific Reports, 2025, 15(1): 12344-12344.

[16] Layton W O, Peng S, Steinmetz T S. ReLU, Sparseness, and the Encoding of Optic Flow in Neural Networks. Sensors, 2024, 24(23): 7453-7453.

[17] Vaswani A, Shazeer N, Parmar N, et al. Attention Is All You Need. arXiv, 2017.

[18] Zhang F, Yin J, Wu N, et al. A dual-path model merging CNN and RNN with attention mechanism for crop classification. European Journal of Agronomy, 2024: 159127273-127273.