CNN Bearing Fault Diagnosis Based on Symmetric Point Pattern Feature Fusion with Multi-Source Resonance Sparse Components.

To address the issue of low recognition accuracy caused by incomplete information, a CNN-based fault diagnosis method for rolling bearings using multi-source resonance sparse component feature fusion (RSSD-P) is proposed in this paper, which effectively resolves the problem of impact features being masked. In noise-contaminated environments, bearing vibration signals exhibit nonstationarity, obscuring fault characteristics. To overcome this, resonance sparse decomposition was employed to extract impact-related fault features. Furthermore, to fully utilize multi-sensor information and enhance fault representation, a symmetric dot pattern (SDP) method was introduced to fuse multi-source fault impact features, achieving effective integration of impact characteristics from multi-source vibration signals. A CNN-based approach incorporating multi-source resonance sparse component and SDP feature fusion was developed, and a bearing fault diagnosis model was established accordingly. Experimental results demonstrate that the proposed method achieves a fault recognition accuracy of 98.63% under varying operating conditions. Compared with other bearing fault diagnosis methods, the recognition precision is improved by 8.49%~17.8%, confirming its superior performance.