Yuanjun Laili

Industrial standards are the basic specifications for the whole lifecycle management of complex products, including design, manufacturing, maintenance, remanufacturing, and recycling. As multidisciplinary knowledge is required for complex products, cross-standard reasoning and clause-level traceability become imperative for decision-support systems. However, existing retrieval-augmented generation (RAG) methods suffer from incomplete standard-scoped evidence retrieval, identifier and evidence-identity heterogeneity, and a mismatch between open-answer correctness and clause-level traceability. To overcome these drawbacks, this paper proposes a Source-Clause-Triple Anchored RAG (SCT-RAG) method specifically designed for traceable industrial standard retrieval. It establishes bidirectional bindings among source standards, clauses, and extracted knowledge triples using unique clause identifiers. Then, a standard-aware router classifies queries into single-standard, cross-standard, and correlation categories. Based on this classification, a standard-anchored retriever combines standard-number-directed lookup, within-standard dense enrichment, and clause-linked knowledge graph (KG) traversal to improve evidence coverage within the identified standard scope. Furthermore, an identity-preserving reranking and generation stage refines the evidence while preserving the unique clause identity of every candidate, and restricts the generator to cite only identifiers exposed in the retrieved context. Experiments on a 200-question test set over a corpus of 200 industrial standards show that the proposed method achieves a clause-level Hit@5 of 0.775, an answer accuracy of 0.755, and a Citation F1 of 0.616 under predicted routing, outperforming eight retrieval-augmented baselines and a closed-book control.

Model-Based Systems Engineering (MBSE) has gained significant attention from both industry and academia as an effective approach to managing product complexity. Despite its progress, current MBSE concepts, tools, languages, and methodologies face notable challenges in industrial applications, particularly in addressing design variability, ensuring model consistency, and enhancing operational efficiency. Based on the authors’ industry observations and literature analysis, this paper identifies the primary limitations of traditional MBSE, and introduces MBSE 2.0, a next-generation evolution characterized by comprehensive, integrated, and intelligent features. Key enabling technologies, such as model governance, integrated design methods, and AI-enhanced system design, are explored in detail. Additionally, several preliminary explorations were introduced under the guidance of the MBSE 2.0 philosophy. This study introduces the MBSE 2.0 concept to stimulate discussion and guide future efforts in academia and industry, emphasizing key advancements and highlighting several key and pressing perspectives to alleviate current limitations in industrial practice.

Complex industrial equipment plays a critical role in specific tasks within industrial edge scenarios. Predicting their health status accurately is essential to ensuring safety and reliability in the production process. However, real-world industrial edge scenarios often have limited resources and stringent real-time requirements, making it difficult to deploy high-precision deep learning models directly at the edge. To address this issue, this article proposes an efficient adaptive-gating binary neural network (ABNN). First, a trend-aware encoder (TAE) is proposed to optimize the binarization process of the input layer. Next, a learnable precision indicator (LPI) is proposed to adjust the inference precision level. Finally, an adaptive-gating convolution is proposed to improve the representational capabilities while maintaining the fitting ability without significantly increasing the computational cost. Additionally, a field-programmable gate array (FPGA) hardware accelerator is designed for the proposed network. ABNN achieves approximately a 7% improvement in accuracy and a 45% gain in efficiency compared to the baseline model.