Kai Wang

Traditional domain adaptation methods often assume balanced data distributions. However, this assumption is frequently violated in real-world industrial scenarios, where normal samples predominate while fault samples are inherently scarce. Under severe class imbalance, conventional decision boundaries tend to shift toward minority fault regions. This shift leads to persistently high misclassification rates for rare fault samples. To overcome this limitation, we propose the Dynamic Maximum Triple-View Classifier Discrepancy (DMTVCD) network, which integrates a Triple-View Classifier (TVC) Architecture and a Primary–Auxiliary Fused Cooperative Loss (PAFL). Specifically, the TVC employs auxiliary binary classifiers to aggregate fine-grained fault sub-classes into a unified “Fault Super-class.” This constructs a robust “normal-fault” binary boundary that effectively counteracts class imbalance. Driven by the PAFL, this boundary acts as a hierarchical geometric constraint to suppress the primary classifier’s tendency to misclassify faults as normal samples, thereby enhancing feature discriminability. Furthermore, a dynamic weighting strategy is introduced to assign large initial weights. This forces the model to bypass simple decision logic dominated by the majority class, ensuring a smooth transition from global exploration to fine-grained alignment. Extensive evaluations on the CWRU and JNU datasets demonstrate that DMTVCD consistently outperforms state-of-the-art approaches under high imbalance ratios (e.g., 20:1).

The scarcity of a large amount of labeled data for adequately training of deep learning models, along with their restricted generalization capabilities, persistently hinders the realworld practical application of data-driven deep learning in fewshot fault diagnosis and transfer task fault diagnosis. This paper proposes a self-supervised Wide Kernel Time-Frequency Fusion (WTFF) contrastive learning method that leverages extensive unlabeled signals to extract discriminative time-frequency fusion features, thereby enhancing fault diagnosis performance even with a limited number of labeled samples. Moreover, the WTFF integrates a multi-layer time-frequency wide convolutional neural network (TFCNN) encoder with a novel local and global timefrequency contrastive loss (LGTFCL) to capture time frequency consistency by facilitating the alignment of time-domain and frequency-domain feature embeddings across the shallow and deep network layers. In the fine-tuning phase, time frequency features across various levels learned from transferred pretrained model are fused to extract signal characteristics that exhibit both time and frequency discrimination. The proposed method demonstrates superior diagnostic accuracy and robustness in experiments involving few-shot and transfer learning-based fault diagnosis. The source code of the proposed method will be available at: https://github.com/KWflyer/WTFF.

In real industrial scenarios, the limited available labeled fault data and existed significant data distribution differences between the source domain and the target domain emplace challenges and obstacles for cross-domain fault diagnosis. In this paper, we proposed a novel cluster contrastive learning method (CLCO) for few-shot learning (even one-shot learning) and cross-domain fault diagnosis under complex fault modes, fault severities, and variable working conditions. The proposed CLCO combines clustering strategy and contrastive learning to learn discriminate domain invariant feature representation. Moreover, the consistency in the characteristic distribution of data is viewed as the pseudo label information captured by k-means clustering, which are subsequently embedded into strengthen contrastive loss (SCL) function of proposed CLCO model. Then, the pseudo label information is exploited to guide self-supervised contrastive pre-training for learning domain invariant features based on feature similarity comparison of positive and negative pairs. Extensive experimental results demonstrate that our proposed CLCO outperforms existing state-of-the-art methods by a large margin.

Big-data mining brings new challenges and opportunities for engineering design, such as customer-needs mining, sentiment analysis, knowledge discovery, etc. At the early phase of conceptual design, designers urgently need to synthesize their own internal knowledge and wide external knowledge to solve design problems. However, on the one hand, it is time-consuming and laborious for designers to manually browse massive volumes of web documents and scientific literature to acquire external knowledge. On the other hand, how to extract concepts and discover meaningful concept associations automatically and accurately from these textual data to inspire designers’ idea generation? To address the above problems, we propose a novel data-driven concept network based on machine learning to capture design concepts and meaningful concept combinations as useful knowledge by mining the web documents and literature, which is further exploited to inspire designers to generate creative ideas. Moreover, the proposed approach contains three key steps: concept vector representation based on machine learning, semantic distance quantification based on concept clustering, and possible concept combinations based on natural language processing technologies, which is expected to provide designers with inspirational stimuli to solve design problems. A demonstration of conceptual design for detecting the fault location in transmission lines has been taken to validate the practicability and effectiveness of this approach.

Although representation learning has been proved to be a promising and effective solution for intelligent fault diagnosis, existing methods still encounter classification performance degradation due to large intra-class variations of real-world applications. In this work, we present a novel representation learning method, namely class level autoencoder (CLAE) for fault diagnosis which aims to learn representative and discriminative features from vibration signals. Specifically, we formulate a novel loss function for representation learning by jointly minimizing the basic and class level reconstruction errors to restrain intra-class variations in the feature space. In addition, we propose a new and simple feature pooling strategy to effectively fuse various meaningful local features to capture efficient and inherent fault pattern information about the input. Extensive experimental results on rolling bearing dataset demonstrate that our proposed method achieves very competitive results compared to state-of-the-art methods. The code of the proposed method is available at https://github.com/KWflyer/CLAE.

Existing intelligent fault diagnosis methods depend mostly on single-scale vibration signals, which not only ignores the latent useful information of other different scales, but also underestimates the complementary benefits across scales. In this work, we show the advantage of learning combination of multi-scale information by aiming to automatically capture complementary and discriminative feature representations from different scales of vibration signals. Specifically, we combine the merit of different activation functions in a joint learning fashion and propose a novel joint multiple reconstructions autoencoder (JMRAE), whose training objective is to jointly optimise multiple reconstruction losses. JMRAE aims to jointly learn discriminative and robust scale-specific feature representations. In addition, we design a new multi-scale representations fusion network (MRFN) model to effectively fuse multi-scale feature representations learned concurrently by per-scale JMRAE model for maximising the discriminative capability of the scale-fused features. The objective of MRFN is to benefit all different scales from each other for improving the classification performance. Extensive experimental results demonstrate the effectiveness of the proposed method on rolling bearing dataset.

Affective analysis of images in social networks has drawn much attention, and the texts surrounding images are proven to provide valuable semantic meanings about image content, which can hardly be represented by low-level visual features. In this paper, we propose a novel approach for visual affective classification (VAC) task. This approach combines visual representations along with novel text features through a fusion scheme based on Dempster-Shafer (D-S) Evidence Theory. Specifically, we not only investigate different types of visual features and fusion methods for VAC, but also propose textual features to effectively capture emotional semantics from the short text associated to images based on word similarity. Experiments are conducted on three public available databases: the International Affective Picture System (IAPS), the Artistic Photos and the MirFlickr Affect set. The results demonstrate that the proposed approach combining visual and textual features provides promising results for VAC task.