Equipments regularly change working speeds during real-time production owing to process requirements. Applying deep learning models trained in a single speed domain straightforwardly to other unknown speed domains is extremely challenging single-domain generalization problem. Therefore, this article proposes a histogram matching mixup based sequential embedding network (HmmSeNet) for single-domain generalization of intelligent fault diagnosis under unknown speeds. HmmSeNet consists of four components: histogram matching mixup (HMM); sequential embedding (SE); separable convolution; and decision making. First, inspired by histogram matching and Mixup, the HMM data augmentation method is proposed. HMM is capable of synthesizing data with the same semantic information, but different distributions from a single source domain data during the training process, thus augmenting the source speed domain to the unknown speed domains. Then, SE utilizes trainable linear dimensional boosting to approximate the distribution between samples, which reduces the effect of sample amplitude distribution shifts caused by speed changes and allows the model to learn domain-invariant features. Finally, three layers of separable convolution and global average pooling are used to accomplish an accurate and robust recognition task. Experimental results on three datasets show that the proposed approach is only trained on a single speed domain, while it has good diagnostic performance on other unknown speed domains, even varying speed domains. The powerful generality and flexible deployment capability of HmmSeNet for speed changes are also demonstrated by ablation experimental analysis and dimensional analysis.

The transient features caused by a local fault are of vital importance for bearing fault diagnosis in an intelligent industry. Due to the uncertainty of fault forms and nonstationarity of operating conditions, the fault feature distribution influenced by the physical dynamic response of actual defect is always complex and irregular with morphological differences. This will bring embarrassments for an accurate fault diagnosis. Motivated by this, a convolution sparse self-learning (CSSL) is proposed in this article to accomplish an adaptive feature enhancement. In the view of image sparse processing, the representation for desired morphological structures is promoted by a two-dimensional optimizing approach with manifold sensing. From a randomly selected fragment, the time-frequency manifold learning is first applied to mine the latent structures. The image entropy is then introduced to adaptively output the optimal one as a sensing kernel. Therewith, this kernel is used to operate a shift-invariant sparse analysis on raw time-frequency image. Combining this rebuilt image with the raw phase, an enhanced signal is finally synthesized. In this manner, the desired transient morphology can be automatically mined, which is consistent with the physical dynamic response. Practical defective bearing data are analyzed to illustrate the effectiveness of the proposed method. Specifically, a comparison further illustrates that the proposed CSSL is superior in the morphological transient features enhancement.