In industrial scenarios, noise interference and feature overlap often result in blurred classification boundaries, compromising the reliability of rolling bearing fault diagnosis. An adaptive distribution-aware discriminative loss (ADADL) is introduced, through which intraclass thresholds are dynamically adjusted and interclass boundaries are optimized, thereby enhancing compactness and separability in the feature space. By integrating it with the cross-entropy loss, ADADL yields marked gains in diagnostic accuracy on both the Case Western Reserve University benchmark and real-world datasets, particularly under conditions of class imbalance and high noise. Visualization analyses further confirm its ability to sharpen clustering boundaries, suppress feature overlap, and effectively mitigate blurred decision regions.

Accurate and timely foam grade prediction plays an important role in the flotation foam industry process. However, the information between foam characteristic series and foam grade series at different sampling times often does not match, making the prediction result lagging behind. A multistep fusion transformer (MSFT) model is designed in this article. First, we extract multiple froth time series as input to correlate feature information and grade information under multiple time series, then, a self-attention structure is designed to fuse at multiple scales, which enhances the degree of information correlation under different time series, finally, the information matrix is passed through the fully connected layer to obtain the final prediction result. Compared with the existing froth grade network recurrent neural networks (RNN), long short-term memory (LSTM), gated recurrent unit, Transformer, Enc–Dec (RNN), feature reconstruction–regression, Siamese time series and difference (LSTM), and FlotationNet models, the MSFT model has reduced the baseline by 30.3%, 30.3%, 30%, 66.9%, 30%, 45.8%, 55.2%, and 52.5%, respectively, among all indicators.

To prevent industrial fault propagation, it is important to clarify the relationship between industrial system components and identify the fault propagation path efficiently and timely, aiming at the problems in fault propagation path identification, this article presents an equipment fault propagation analysis approach based on unstable points identification to solve such issues. First, the fault propagation diagram is created by analyzing the industrial complex system components. Furthermore, to address the issues of traditional interpretative structural modeling (ISM), the bilateral rotation interpretative structural modeling (BRISM) method is proposed to stratify the components and detect the unstable points. Finally, the fault propagation graph is analyzed using the PageRank algorithm to update the unstable points and edge weights between different nodes to identify the fault propagation paths. The results indicate that the proposed method can effectively identify fault propagation paths between components and can be applied to various industrial systems.

In contemporary industries, diagnosing bearing faults is crucial, yet the complexity and diversity of these faults pose challenges to traditional methods. Existing algorithms typically treat compound faults as independent events, overlooking the interrelations among different faults, which constrains the performance in diagnosing the faults with diverse semantic complexities. Also, the research on leveraging multimodal data to enhance fault diagnosis accuracy is limited. To overcome the weakness mentioned above, a multimodal knowledge graph (MKG) construction method based on multimodel data, including time series vibration signals, spectrum, and description text of datasets, is proposed. Subsequently, a fault diagnosis method utilizing a MKG completion model based on a relation cascade graph attention network is designed to capture the relationship between various faults. Experimental results on an MKG constructed from seven bearing datasets demonstrate the robustness of the proposed method.

Mobile edge computing effectively enhances service quality and decreases system cost by processing resource-intensive tasks at the network edge. Today, unmanned aerial vehicles (UAVs) are increasingly being utilized for task offloading services in remote areas due to their convenient deployment and flexible mobility. However, the complex task environment when using UAVs brings great challenges to the optimization strategy’s capacity to solve and converge in a stable manner. To solve this issue, a differential rate rule (DRR) is proposed in this work with the goal of improving the update stability of the agent in the actor–critic reinforcement learning (RL). Second, a UAV-assisted task offloading algorithm called DSUTO is designed based on DRR and maximum entropy RL. Finally, a UAV-assisted mobile device-edge-cloud collaborative computing model is constructed with time-varying channel obstacles and user movement, thus solving a multi-objective joint optimization problem on the task completion cost (including delay and energy consumption) and UAV endurance under resource constraints. The experiment results demonstrate that DSUTO not only has excellent performance in terms of convergence and stability, but also significantly reduces the total system cost by 21.38% compared with the latest benchmark algorithms under complex environment conditions.

Concerning the problems of harrowing extraction and poor classification accuracy of fault features in rolling bearing vibration signals, a fault feature selection and classification method based on multi-indicator fusion is proposed. First, the original signal is decomposed through the improved complementary ensemble local mean decomposition method into several physically meaningful product functions (PF) and single residual components; then, the three indicators of kurtosis, correlation coefficient, and Kulback–Leibler divergence are combined to extract the most suitable PF components for signal reconstruction. Ultimately, the reconstructed signal's multidomain characteristics and entropy value features are retrieved and fed into the LightGBM classifier for classification in order to achieve an intelligent diagnosis of rolling bearing problems. The statistical results demonstrate that the proposed method can efficiently identify the functional PF components and has notable benefits in extracting features from diverse experimental datasets and detecting faults.