Multi-sensor heterogeneous data fusion method for rotor system diagnosis based on multi-mode residual network and discriminant correlation analysis

Deep learning algorithms have become a research hotpot in the field of fault diagnosis due to their powerful feature adaptive extraction capabilities. However, the monitoring data used as the input of deep learning typically includes only single-sensor data (e.g. vibration signals), not the multi-sensor unstructured data (e.g. infrared images) that can provide fault information from different angles, resulting in low diagnostic accuracy and weak generalization ability of the built model. To apply multi-sensor heterogeneous monitoring data fully, a novel fusion diagnosis method based on multi-mode residual network (M-ResNet) and discriminant correlation analysis (DCA) is proposed in this paper. Firstly, M-ResNet composed of one-dimensional ResNet (1D-ResNet) and two-dimensional ResNet (2D-ResNet) is constructed to extract richer and more comprehensive fault features. 1D-ResNet and 2D-ResNet are used to process structured vibration signals and unstructured infrared images respectively, so as to obtain more comprehensive fault information. Secondly, DCA is used to fuse the vibration signal features and infrared image features extracted by M-ResNet to maximize the intra-class correlation and eliminate inter-class correlation of features, thereby improving the accuracy of fault diagnosis. Finally, the fusion features are input into the Softmax classifier to achieve fault classification. The effectiveness of the proposed method is verified by experiments on the rotor system, where it achieves a remarkable classification rate of 99.79%. Compared with the similar methods, the proposed method exhibits outstanding performance, indicating the feasibility of using multi-sensor heterogeneous data for rotor system fault diagnosis.