Search for: All records

Fault diagnosis of rolling bearings becomes an important research subject, where the data-driven deep learning-based techniques have been extensively exploited. While the state-of-the-art research has shown the substantial progresses in bearing fault diagnosis, they mostly were implemented upon the hypothesis that the location of bearing prone to failure already is known. Nevertheless, in actual practice many rolling bearings are installed in a complex machinery system, any of which is likely subject to fault. As such, fault diagnosis essentially is a process to achieve both fault localization and identification, which results in many fault scenarios to be handled. This will significantly degrade the fault diagnosis performance using conventional deep learning analysis. In this research, we aim to develop a new deep learning framework to address abovementioned challenge. We particularly design a hierarchical deep learning framework consisting of multiple sequentially deployed deep learning models built upon the transfer learning. This can improve the learning adequacy for a high-dimensional problem with many fault scenarios involved even under limited dataset, thereby enhancing the fault diagnosis performance. Without the prior knowledge regarding the fault location, this methodology is greatly favored by the sensor/data fusion which takes full advantage of the enriched pivot fault-related features in the measurements acquired from different accelerometers. Systematic case studies using the publicly accessible experimental rolling bearing dataset are carried out to validate this new methodology. 

Rolling bearing is a critical component of machinery that has been widely applied in manufacturing, transportation, aerospace, and power and energy industries. The timely and accurate bearing fault detection thus is of vital importance. Computational data-driven deep learning has recently become a prevailing approach for bearing fault detection. Despite the progress of the deep learning approach, the deep learning performance is hinged upon the size of labeled data, the acquisition of which is expensive in actual implementation. Unlabeled data, on the other hand, are inexpensive. In this research, we develop a new semi-supervised learning method built upon the autoencoder to fully utilize a large amount of unlabeled data together with limited labeled data to enhance fault detection performance. Compared with the state-of-the-art semi-supervised learning methods, this proposed method can be more conveniently implemented with fewer hyperparameters to be tuned. In this method, a joint loss is established to account for the effects of labeled and unlabeled data, which is subsequently used to direct the backpropagation training. Systematic case studies using the Case Western Reserve University (CWRU) rolling bearing dataset are carried out, in which the effectiveness of this new method is verified by comparing it with other well-established baseline methods. Specifically, nearly all emulation runs using the proposed methodology can lead to around 2%–5% accuracy increase, indicating its robustness in performance enhancement.