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.

Bearing fault impulses are easily submerged by background noise, resulting in the inconspicuous fault feature and affecting the accuracy of the fault diagnosis. This paper presents a novel method for bearing the fault feature enhancement and diagnosis based on the statistical filtering and the 1.5-dimensional symmetric difference analytic energy operator (1.5D-SDAEO) to solve the problem. 1). The statistical filtering is used to filter the background noise under the standard distinction index. 2). The 1.5D-SDAEO is used to enhance the signal impulse, suppress the residual noise, and improve the SNR. 3). The dominant frequency in the energy spectrum is compared with the rolling bearing fault characteristic frequency to the fault diagnosis. The feasibility and the superiority of the presented method are verified by the simulation, engineering, and comparison experiments. All results show that the presented method can effectively enhance the fault feature and accurately diagnose the rolling bearing fault.

To precisely diagnose the rotating machinery structural faults, especially structural faults under low rotating speeds, a novel scheme based on combination of empirical mode decomposition (EMD), sample entropy, and deep belief network (DBN) is proposed in this paper. EMD can decompose a signal into several intrinsic mode functions (IMFs) with different signal-to-noise ratios (SNRs) and sample entropy is performed to extract the signals that carry fault information with high SNR. The extracted fault signal is reconstructed into a new vibration signal that will carry abundant fault information. DBN has strong feature extraction and classification performance. It is suitably performed to build the diagnosis model based on the reconstructed signal. The effectiveness of the proposed method is validated by structural faults signal and the comparative experiments (BPNN, CNN, time-domain signal only, frequency-domain signal only). The results show that the diagnosis accuracy of the proposed method is between 99% and 100%, the BPNN is less than 25%, and the CNN is between 70% and 95%, which means the verified, proposed method has a superior performance to diagnose the structural fault.

Low-speed bearing fault signal is weak and overwhelmed by background noise, and it is ineffective in diagnosing bearing faults. In this study, an improved statistical filtering and convolution neural network (CNN) low-speed bearing fault diagnosis method is presented. Statistical filtering is a convenient and effective signal-filtering method that can extract fault signals from strong background noise original signals. The improved method can deal with drawbacks in selecting the standard distinction index of statistical filtering algorithm. CNN is used to extract signal features automatically and build the diagnosis model for fault classification. An engineering experiment is performed with different condition-bearing fault signals to test this presented method, and the results show that the improved statistical filter and CNN are effective in diagnosing low-speed bearing signals. Comparison with traditional methods proves that this study has superior performance in low-speed bearing fault diagnosis.