Class imbalance issue has been a major problem in mechanical fault detection, which exists when the number of instances presenting in a class is significantly fewer than that in another class. This article focuses on the problem of zero-shot fault detection of rolling bearing, which is the extreme case of class imbalance. Aiming at this problem, a two-stage zero-shot fault recognition method is proposed. First, inspired by the conditional generative adversarial network, a novel feature generating network which is composed of a feature extractor, a discriminator, and a generator is designed to capture the potential distribution of normal samples. Then, the generator will generate abundant pseudofault features by adding an additional sequence to the condition. Second, an improved deep neural network is trained with these synthetic pseudofault features as the classifier. Specially, a condition index is designed to represent different fault classes so that it can recognize the unseen fault samples. Finally, the effectiveness of the proposed method is verified by three datasets and a comparison method is also given to show the superiority. Results show that the feature generation network can effectively detect the typical faults even though the fault data are unavailable during training, which is practical for industrial application.

Deep learning has demonstrated splendid performance in mechanical fault diagnosis on condition that source and target data are identically distributed. In engineering practice, however, the domain shift between source and target domains significantly limits the further application of intelligent algorithms. Despite various transfer techniques proposed, either they focus on single-source domain adaptation (SDA) or they utilize multisource domain globally or locally, which both cannot address the cross-domain diagnosis effectively, especially with category shift. To this end, we propose globally localized multisource DA for cross-domain fault diagnosis with category shift. Specifically, we construct a GlocalNet to fuse multisource information comprehensively, which consists of a feature generator and three classifiers. By optimizing the Wasserstein discrepancy of classifiers locally and accumulative higher order multisource moment globally, multisource DA is achieved from domain and class levels thus to reduce the shift on domain and category. To refine the classifier at sample level, a distilling strategy is presented. Finally, an adaptive weighting policy is employed for reliable result. To evaluate the effectiveness, the proposed method is compared with multiple methods on four bearing vibration datasets. Experimental results indicate the superiority and practicability of the proposed method for cross-domain fault diagnosis.

Intelligent fault detection has been widely used for feature extraction and fault classification. However, various complex signal processing methods are adopted in many researches. The paper presents a novel deep learning network via shunt-wound restricted Boltzmann machines with layerwise feature learning to learn the fault features from big raw vibration signals directly. The network consists of split layer, predict layer, update layer, dephasing layer and softmax layer. The shunt-wound restricted Boltzmann machines are used in both predict layer and update layer to improve feature extraction ability and to ensure effective training of the network. The operation of predicting and updating is considered to be a layerwise feature learning which makes the network effective. The proposed method is verified by three bearing datasets and a few related researches are introduced for comparison. Results prove that the proposed method is capable of extracting sensitive features for fault detection.

The implementation of condition monitoring and fault diagnosis are of special importance for ensuring wind turbine (WT) operation safely and stably. In practice, however, the fault data of WT is limited, which makes it hard to identify faults of WT accurately using existing intelligent diagnosis methods. To address this, signals augmented self-taught learning networks (SASLN) is proposed for the fault diagnosis of the generator, which is one of the most important parts in WT. In SASLN, fault signal samples are generated by the Wasserstein distance guided generative adversarial networks to expand the limited training dataset. The sufficient generated signal samples are used to pre-train the self-taught learning networks (SLN) to enhance the generalization ability of SLN. Then, the weights of SLN are fine-tuned using a small number of real signal samples for accurate fault classification. The effectiveness of SASLN is verified by two bearing vibration datasets. The results show that SASLN can achieve fairly high fault classification accuracy using small training samples. Besides, SASLN has good robustness in the noisy working environment and can also identify faults even in variable loads and variable rotating speeds cases, which makes it meaningful for decreasing the running costs and improving the maintenance management of WT.