The existing fault diagnosis methods of rotating machinery constructed with both shallow learning and deep learning models are mostly based on vibration analysis under steady rotating speed. However, the rotating speed frequently changes to meet practical engineering needs. The shallow learning models largely depend on domain experience of feature extraction, and training a deep learning model requires large samples and a long time. In addition, vibration monitoring has the shortcomings of contact measurement, small coverage, and noise interference. To address these problems, this article proposes a new fault diagnosis method with the least square interactive support matrix machine (LSISMM) and infrared thermal images. In this method, a novel matrix-form classifier called LSISMM is constructed under the concept of nonparallel interactive hyperplanes to fully leverage the structure information of infrared thermal images. To improve the computation efficiency, a new least square loss constraint is designed for LSISMM. Besides, we derive an effective solution framework based on the alternating direction method of the multiplier (ADMM) framework. The constructed LSISMM is directly used to analyze the collected thermal images of rotating machinery under time-varying speeds. Experiment results demonstrate that the proposed method is superior to state-of-the-art methods in terms of diagnosis accuracy and efficiency, especially under small thermal image samples.

Vibration signals and infrared images have different advantages and characteristics. Although a few recent researches have explored their information fusion in rotating machinery fault diagnosis, they show limited performance when facing strong interference and imbalanced cases. Therefore, a fusion framework based on confidence weight support matrix machine (CWSMM) is proposed. In this framework, CWSMM can not only fully leverage the structure information of infrared thermography images and vibration time–frequency images, but also has the following novelties. First, CWSMM uses dynamic penalty factors for different class samples to address the class imbalance problem. Second, by using the prior knowledge of matrix samples, a confidence weight assignment strategy is designed for CWSMM to improve the robustness. Last, the Dempster–Shafer (D-S) evidence theory is applied to fuse the posterior probability outputs of CWSMMs using different measurements. Experiment results demonstrate that the proposed method has promising fault diagnosis performance, specifically under strong interference and imbalanced datasets.

The vast majority of the existing diagnostic studies using deep learning techniques for rotating machinery focus on the vibration analysis under steady rotating speed. Nevertheless, the collected vibration signals are sensitive to time-varying speeds and the vibration sensors may cause structure damage of equipment after long-term close contact. Aiming at these aforementioned problems, a modified Gaussian convolutional deep belief network driven by infrared thermal imaging is proposed to automatically diagnose different faults of rotor-bearing system under time-varying speeds. First, infrared thermal images are measured to characterize the working states of rotor-bearing system to reduce the impact of changeable speeds. Second, Gaussian units are used to construct Gaussian convolutional deep belief network to well deal with infrared thermal images. Finally, trackable learning rate is designed to modify the training algorithm to enhance the performance. The comparison results verify the feasibility of the proposed method, which outperforms the other methods.