Search

In recent years, deep learning methods, especially convolutional neural network (CNN), have received extensive attentions and applications in fault diagnosis. However, recent studies have shown that the shift-invariance of CNN is not good enough, resulting in fragile feature extraction and sharp reduction in model performance when the shift occurs in the input. To improve the shift-invariance of CNN, considering the periodic characteristics of vibration signals, a multiscale CNN with feature alignment (MSCNN-FA) is proposed for bearing fault diagnosis under different working conditions. First, by analyzing the operating principles of the convolutional layer and pooling layer, a feature alignment module including single-stride convolution layer, adaptive max-pooling layer, and global average pooling layer is designed to obtain aligned features. Next, to extract shift-invariant robust features from vibration signals, a multiscale convolution strategy is utilized, and a feature-aligned multiscale feature extractor is constructed. Finally, a classifier composed of fully connected (FC) layers is constructed for bearing fault diagnosis. The effectiveness of the method is verified by a rolling bearing experiment, which outperforms other related existing CNN-based methods in terms of diagnosis accuracy and feature robustness. [ABSTRACT FROM AUTHOR]

Recently, deep transfer learning based intelligent fault diagnosis has been widely investigated, and the tasks that source and target domains share the same fault categories have been well addressed. However, due to complexity and uncertainty of mechanical equipment, unknown new faults may occur unexpectedly. This problem has received less attention in the current research, which seriously limited the application of deep transfer learning. In this article, a two-stage transfer adversarial network is proposed for multiple new faults detection of rotating machinery. First, a novel deep transfer learning model is constructed based on an adversarial learning strategy, which can effectively separate multiple unlabeled new fault types from labeled known ones. Second, an unsupervised convolutional autoencoders model with silhouette coefficient is built to recognize the number of new fault types. Extensive experiments on a gearbox dataset validate the practicability of the proposed scheme. The results suggest that it is promising to address fault diagnosis transfer tasks in which the multiple new faults occur in the target domain, which greatly expand the application of deep transfer learning. [ABSTRACT FROM AUTHOR]

In recent years, deep learning has become a promising tool for rotary machinery fault diagnosis, but it works well only when testing samples and training samples are independent and identically distributed. In practice, rotary machinery usually works under variable speed. The change of speed leads to the variation of samples’ distribution, which can significantly decrease the performance of the deep learning model. Scholars try to utilize transfer learning techniques for solving this problem. However, most exiting methods can just work well under target speed instead of all speed, while the target samples are always required in model training. In this article, a deep semisupervised domain generalization network (DSDGN) is proposed for rotary machinery fault diagnosis under variable speed, which can generalize the model to the fault diagnosis task under unseen speed. Under the setting of semisupervised domain generalization, only one fully labeled source (LS) domain data set and one totally unlabeled source (US) domain data set are available during training. To make full use of these data, the proposed method simultaneously utilizes Wasserstein generative adversarial network with gradient penalty (WGAN-GP)-based adversarial learning and pseudolabel-based semisupervised learning for training. The transmission and bearing fault diagnosis cases are utilized for evaluation. The comparative experiments indicate that the proposed method has a better performance than other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In practical industrial applications, the working conditions of machinery are changing with long-term operation, and the health status is declining with the degradation of crucial components. When the working condition changes, prior diagnosis models cannot be generalized from one condition to another. To solve this challenging issue, in this article a robust weight-shared capsule network (WSCN) is introduced for intelligent fault diagnosis of machinery under varying working conditions. First, taking raw accelerometer signals as inputs, one-dimensional convolutional neural network is constructed to extract discriminative characteristics. Second, various capsule layers based on multistacked weight-shared capsules are developed to enhance the generalization performance for further fault classification. Finally, margin loss function as well as agreement-based dynamic routing algorithm are employed to optimize the WSCN. In this article, two diagnosis cases are carried out to demonstrate the generalization performance of the WSCN which obtains higher accuracy under varying working conditions than that of other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Deep transfer learning has attracted many attentions in machine intelligent fault diagnosis. However, most existed deep transfer learning algorithms encounter difficulties to detect a new emerging fault in target domain because these methods assume that the source and target domains have the same fault categories. Unfortunately, in real-world applications, new fault may emerge during machine running, which is not the same as those faults for training diagnosis models. To solve this problem, a novel fault diagnosis method named deep adversarial transfer learning network (DATLN) is proposed for new emerging fault detection. First, a one-dimension convolutional neural network is constructed to learn invariant features from the raw vibration signals of the source and target domains. Then, a multiple label classifier is trained to recognize known fault classes of the source and target domains. Finally, a decision boundary is built for the new emerging fault detection by training a classifier to recognize some target samples as new ones. Experiments on rolling bearing and gearbox demonstrate that the DATLN can implement the faults recognition with high accuracy and outperform other transfer learning methods when a new fault emerging in the target domain. [ABSTRACT FROM AUTHOR]

With the manufacturing industry stepping into the emerging new era of big data and intelligence, the amount of data collected from perception and monitoring systems with multiple smart sensors has increased tremendously. Such huge amount of multisensory data may not only power many aspects of fault diagnosis, but also bring great opportunities and challenges in modern manufacturing industry. In addition, with respect to intelligent fault diagnosis for machinery, few researches have been focused on the compound fault diagnosis under big-data circumstance. Therefore, a novel, intelligent, compound, fault decoupling method based on deep capsule network (CN) and ensemble learning is developed for compound fault decoupling and diagnosis using multisensory data. First, a decoupling CN (DCN) is constructed as the basic model. Second, taking the full advantage of multisensory data, the DCN model can be pretrained with multiple sensor data, which can obtain various pretrained DCN models. Finally, combining with ensemble learning skill, the pretrained DCN models are integrated by a combination strategy to obtain the deep ensemble CN (DECN) model for intelligent compound fault decoupling and diagnosis. The performance of the DECN model is validated on an automobile transmission (AT) data set with two compound faults, and the experimental results illustrate that the DECN model obtains higher diagnosis accuracy and decouples the compound fault correctly. [ABSTRACT FROM AUTHOR]