Rolling Bearing Fault Diagnosis Based on Recurrence Plot

作者：Zheming Chen, Bu Xu, Zhong Zhang · 发表于：IEEE Access · 年份：2024 · DOI：10.1109/access.2024.3472454 · 被引用次数：9 · 研究领域：Gear and Bearing Dynamics Analysis、Advanced Measurement and Detection Methods、Advanced Decision-Making Techniques

A bearing fault diagnosis method based on recurrence plots and a fusion neural network is proposed to address the low recognition accuracy of noisy bearing vibration data. Compared to existing methods, this approach leverages the recurrence plot technique to convert vibration signals into color images, which carry more information than grayscale images. For the prediction model, the traditional convolutional neural network is enhanced by integrating bidirectional gated recurrent unit and multi-head attention mechanism, allowing it to capture temporal features alongside the spatial features typically extracted by convolutional neural network. The accuracy of the method exceeds 92% on two different bearing datasets, indicating its strong generalization performance. The results of ablation and comparison experiments demonstrate that the proposed model achieves high prediction accuracy even in the presence of strong noise, exhibiting robust noise immunity compared with other methods.