In current research works, a number of intelligent fault diagnosis methods have been proposed with the assistance of domain adaptation approach, which attempt to distinguish the health modes for target domain data using the diagnostic knowledge learned from source domain data. An important assumption for these methods is that the label information for the source domain data should be known in advance. However, the high-quality condition monitoring data with sufficient label information is difficult to be acquired in the actual field, which can greatly hinder the effectiveness of domain adaptation–based fault diagnosis methods. The simulation model of the rotating machine is an effective approach to provide an insight into the characteristics of the mechanical equipment, which can also easily carry the sufficient label information for the mechanical equipment under various operating conditions. In this study, a simulation data–driven domain adaptation approach is proposed for the intelligent fault diagnosis of mechanical equipment. The simulation data from a rotor-bearing system are used to build the source domain data set, and the diagnostic knowledge learned from the simulation data is used to realize the healthy mode identification of mechanical equipment in the actual field. The proposed domain adaption approach consists of two parts. The first part is to achieve the conditional distribution alignment between source domain data and target domain supervised data in an alternative way. The second part is to achieve the marginal distribution alignment between source domain data and target domain unsupervised data in an adversarial training process. The proposed domain adaptation method is evaluated on two case studies, the diagnostic results on two case studies indicate that the proposed domain adaptation method is capable of realizing the fault diagnosis of mechanical equipment using the diagnostic knowledge learned from simulation data.
Rolling bearings are the vital components of rotary machines. The collected data of rolling bearing have strong noise interference, massive unlabeled samples, and different fault features. Thus, a deep transfer learning method is proposed for rolling bearings fault diagnosis under variable operating conditions. To obtain robust feature representation, the denoising autoencoder is used to denoise and reduce dimension of unlabeled rolling bearing signals. For those unlabeled target domain signals, a feature matching method based on multi-kernel maximum mean discrepancies between source domain and target domain is adopted to get enough labeled target domain samples. Then, these rolling bearing signals are converted to multi-dimensional graph samples and fed into a convolutional neural network model for fault diagnosis. To improve the generalization of convolutional neural network under variable operating conditions, we combine modelbased transfer learning with feature-based transfer learning to initialize and optimize the convolutional neural network parameters. The effectiveness of the proposed method is validated through several comparative experiments of Case Western Reserve University data. The results demonstrate that the proposed method can learn features adaptively from noisy data and increase the accuracy rate by 2%-8% comparing with other models.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.