A Lightweight Model for Bearing Fault Diagnosis Based on Gramian Angular Field and Coordinate Attention

The key to ensuring rotating machinery’s safe and reliable operation is efficient accurate faults diagnosis. Intelligent fault diagnosis technology based on deep learning (DL) has gained increasing attention. A critical challenge how embed the characteristics of time series into DL obtain stable features that correlate with equipment conditions. This study proposes a lightweight rolling bearing method Gramian angular field (GAF) coordinated attention (CA) improve recognition performance efficiency. Firstly, domain signal encoded GAF images after downsampling segmentation. retains temporal relation provides valuable for DL. Secondly, convolution neural network (CNN) model constructed through depthwise separable convolution, inverse residual block, linear bottleneck layer learn advanced features. After that, CA employed capture long-range dependencies identify precise position information nearly no additional computational overhead. proposed tested evaluated by CWRU dataset experimental dataset. results demonstrate CNN (GAF-CA-CNN) can effectively reduce calculation overhead achieve high diagnostic accuracy.