A novel method of intelligent diagnosis is proposed for structural faults in low-speed rotating machinery. It uses a hybrid scheme to automatically identify health states in a complex mechanical system by combining an improved mode decomposition approach, a Gramian angular summation field and a convolutional neural network. The proposed method is tested with a dataset affected by noise to evaluate its performance and generalizability, which are both essential in fault diagnosis. Experimental results show that the proposed scheme is superior to traditional machine learning methods. It not only provides a more efficient and widely applicable way to learn intrinsic fault signatures, it also has enhanced feature learning capability to improve precision and generalizability. Finally, the reasons for the method’s high performance are analysed to determine the best general features for an adaptive classifier that is generalizable to diverse operating conditions.

machinery fault diagnosis is an attractive but challenging task, especially for low-speed conditions. Therefore, a new discriminative approach that introduces robust principle component analysis (RPCA) and multi-kernel to deep neural networks is proposed to perform intelligent fault diagnosis. Firstly, RPCA is applied to extract fault signals from extreme background noise based on its sensitivity to grossly corrupted data. Secondly, two cascaded MKPCA stages with additional robustness to distortions in feature extraction are used to enhance the energy of spectrum symptom and overcome the tricky issues of low-speed machinery. Especially, the multi-kernel is introduced into the basic PCA filters to learn the data-adapting convolution filter and gain additional robustness to nonlinearity in the signal. Finally, the proposed method is demonstrated on signals from laboratory tests (with a slightly damaged defect in a bearing) and structural fault data, outperforming those of traditional machine learning and classical deep learning methods. Moreover, hidden information of the network is visualized to analyse the reasons for its high performance.