In intelligent manufacturing systems, accurate and timely fault diagnosis is crucial for ensuring a safe and stable manufacturing process. While transfer learning (TL) can mitigate the need for extensive labeled data, not all historical datasets are applicable to specific fault diagnosis tasks, and the use of inappropriate datasets can deteriorate the accuracy of TL models. To address these issues, a TL fault diagnosis method based on cloud-edge collaboration is proposed. First, a variational autoencoder TL algorithm based on multiscale convolution and domain fusion (MSDF-VAE) is presented to effectively leverage extensive historical fault data, particularly in scenarios with limited labeled samples. Second, a lightweight autoencoder model (LAE) is employed to improve the reusability and specificity of historical data and fault diagnosis models by analyzing the correlation between historical data and the current data. Additionally, to reduce latency and meet real-time requirements for fault diagnosis tasks, a cloud-edge collaborative framework is proposed, within which MSDF-VAE and LAE are deployed. This approach enables real-time diagnosis using the MSDF-VAE model at the edge layer, while the cloud layer concurrently trains a high-precision model with the selected data by the LAE. The experiments verify the accuracy of the MSDF-VAE and confirm the effectiveness of the proposed cloud-edge collaboration framework.

Semantic expertise remains a reliable foundation for industrial decision-making, while Large Language Models (LLMs) can augment the often limited empirical knowledge by generating domain-specific insights, though the quality of this generative knowledge is uncertain. Integrating LLMs with the collective wisdom of multiple stakeholders could enhance the quality and scale of knowledge, yet this integration might inadvertently raise privacy concerns for stakeholders. In response to this challenge, Federated Learning (FL) is harnessed to improve the knowledge base quality by cryptically leveraging other stakeholders’ knowledge, where knowledge base is represented in Knowledge Graph (KG) form. Initially, a multi-field hyperbolic (MFH) graph embedding method vectorizes entities, furnishing mathematical representations in lieu of solely semantic meanings. The FL framework subsequently encrypted identifies and fuses common entities, whereby the updated entities’ embedding can refine other private entities’ embedding locally, thus enhancing the overall KG quality. Finally, the KG complement method refines and clarifies triplets to improve the overall quality of the KG. An experiment assesses the proposed approach across different industrial KGs, confirming its effectiveness as a viable solution for collaborative KG creation, all while maintaining data security.

Human–robot collaborative disassembly (HRCD) is a promising approach in remanufacturing, leveraging robot's efficiency and human's adaptability for disassembling end-of-life (EoL) products. However, HRCD often encounters numerous choices with uncertain outcomes, posing significant challenges. To address this issue, an HRCD sequence planning model is introduced, providing a quantitative analysis of various decisions with explanations. Initially, HRCD constraint graph is constructed for targeted EoL product based on semantic documents. Subsequently, a Dirichlet Bayesian network (DiBN) is employed to generate feasible sequences based on the HRCD constraint graph, effectively quantifying uncertainty. Then, a fine-tuned large language model (LLM) with tailored prompts is utilized to quantitatively analyze DiBN-based sequences. The DiBN is updated with high-performing sequences from LLM, mitigating the limited knowledge about specific EoL products. Furthermore, a generative adversarial network is proposed to integrate the aforementioned modules for effective training. The effectiveness of the proposed method is demonstrated through two HRCD case studies.

Fault isolation, or fault location, aims to identify anomalous components at the start of the maintenance process. However, fault isolation within complex equipment can be challenging due to constraints on the scarcity of labeled data and the intricate interaction among various substructures. To overcome this challenge, an embedding-based Bayesian Network (BN) probability inference is proposed to locate the fault components, where the embedding, derived from semantic meanings, can approximate the actual fault distribution within BN. First, a Fault Graph (FG) is established based on the equipment's mechanical structure and its mechanisms. Then, a Multifield hyperbolic embedding is employed to vectorize the nodes in the FG, thereby preserving the inherent logic maximally. Following this, the FG is transformed into the BN, which facilitates the prediction of the faulty component based on available evidence, using the well-trained graph embedding. An empirical study on oil drilling equipment showcases the graph embedding properties and inference performance of the proposed method by comparing it with other cutting-edge methods and traditional scenarios.

Serverless edge computing adopts an event-based paradigm that provides back-end services and dynamically provisions resources as needed, resulting in efficient resource utilization. To improve the end-to-end latency and revenue, service providers need to optimize the number and placement of serverless containers while considering the system cost (i.e., latency cost and container running cost) incurred by the provisioning. The particular reason for this circumstance is that frequently creating and destroying containers not only increases the system cost but also degrades the time responsiveness due to the cold-start process. Function caching is a common approach to mitigate the coldstart issue. However, function caching requires extra hardware resources and hence incurs extra system costs. Furthermore, the dynamic and bursty nature of serverless invocations remains an under-explored area. Hence, it is vitally important for service providers to conduct a context-aware request distribution and container caching policy for serverless edge computing. In this paper, we study the request distribution and container caching problem in serverless edge computing. We prove the proposed problem is NP-hard and hence difficult to find a global optimal solution. We jointly consider the distributed and resourceconstrained nature of edge computing and propose an optimized request distribution algorithm that adapts to the dynamics of serverless invocations with a theoretical performance guarantee. Also, we propose a context-aware probabilistic caching policy that incorporates a number of characteristics of serverless invocations. Via simulation and implementation results, we demonstrate the superiority of the proposed algorithm by outperforming existing caching policies in terms of the overall system cost and cold-start frequency by up to 62.1% and 69.1%, respectively.