Fault Diagnosis of Rotating Machinery System Based on Multi-source Fusion Using RP-RGB-HLCNN Model

Aiming at the problem that the signal collected by a single sensor may cause the loss of fault information, a rotating mechanical system diagnosis method based on multi-source information fusion based on RGB color image fusion technology is proposed. Convert the multi-channel data collected from different position sensors into a recursive grayscale image (RGI). RGB image fusion technology is used to fuse the RGIs of different channels into new color images, and then the dataset composed of color images is sent to the improved deep convolutional network (CNN) for feature fusion extraction and fault diagnosis. The method is tested on the dataset of CWRU with multi-channel synchronous acquisition process. The universality and effectiveness of the information fusion method are verified through the comparative analysis before and after the fusion and the analysis of the average accuracy rate under the division of datasets with different proportions.