A novel fault diagnosis method for rolling bearings based on dual branch architecture ECVmamba-1DCNN model

To address the issues of large parameter count, high computational complexity, and poor noise resistance in deep learning based rolling bearing fault diagnosis models, a lightweight intelligent fault diagnosis model based on ECVmamba-1DCNN is proposed. The original vibration signal of the bearing is converted into a time-frequency image through continuous wavelet transform, and the image data is combined with time series data to form a composite dataset to fully characterize the time-frequency domain characteristics of the signal. Constructing a dual branch architecture ECVmamba-1DCNN model, utilizing ECVmamba branches to process time-frequency images and capture global fault dependencies; Using 1DCNN branches to process temporal signals, extracting local instantaneous features, and achieving complementary features through adaptive weighted decision fusion. Verified in the 5-class task of the dataset, the analysis results show that the diagnostic accuracy of the constructed model surpasses mainstream CNN and Transformer diagnostic models, with an average accuracy of 98.42%; Moreover, due to the integration of ES2D and ECA modules, it has significant advantages in reducing computational complexity and enhancing the ability to distinguish fault features from noise.