The purpose of this study is to improve the accuracy and generalization ability of intelligent fault diagnosis models for rolling bearings under varying operating conditions. By integrating multidimensional features through multi-view learning (MVL) and utilizing Mamba feature fusion, the method aims to address the challenge of data distribution differences that reduce diagnostic accuracy when working conditions change. The approach also incorporates domain adaptation techniques to align source and target domain data, ensuring robust and accurate fault detection. This work seeks to enhance fault diagnosis performance, reduce maintenance costs and ensure operational continuity in industrial environments.
This paper proposes an integrating multidimensional feature method based on multi-view learning (IMDF-MVL) for intelligent fault diagnosis of rolling bearings. MVL is used to capture multidimensional fault features, while Mamba feature fusion combines features from different views to enhance the model’s generalization ability. Domain adaptation is applied to align data distributions between source and target domains. Experimental validation is conducted by comparing IMDF-MVL with state-of-the-art methods, demonstrating its superior diagnostic accuracy and robustness under varying conditions. The proposed approach aims to provide an effective solution for real-world industrial fault detection applications.
The findings of this study demonstrate that the proposed IMDF-MVL method significantly outperforms existing fault diagnosis models, such as DCTLN, NCNN, InDo-DDM, GMVTDA and RTDGN, in both source and target domain datasets. On the source domain, IMDF-MVL achieves an average diagnostic accuracy of 99.98 and 99.89%, highlighting its high efficiency and stability. In target domain transfer experiments, even without target domain fine-tuning, the method achieves diagnostic accuracies of 93.71 and 63.40%, indicating its robustness under changing operating conditions. These results confirm the method’s ability to maintain diagnostic performance and improve generalization across diverse scenarios.
The originality of this study lies in the integration of multidimensional feature extraction through multi-view learning (MVL) and Mamba feature fusion, addressing the challenge of fault diagnosis under varying operating conditions. By leveraging domain adaptation techniques, the proposed IMDF-MVL method aligns data distributions between source and target domains, enhancing model generalization. This work contributes to the advancement of intelligent fault diagnosis by providing a robust and effective approach for rolling bearings, with potential applications in other rotating machinery. The method’s ability to maintain high diagnostic accuracy across diverse conditions offers significant value in industrial operation and maintenance.
