Scholarlink

<p>Amidst the intertwined complexity and fragility of global supply chain networks, there is an urgent need to break through the limitations of traditional risk management’s static modeling. This study innovatively integrates dynamic Bayesian networks with lightweight convolutional neural networks to construct a full-chain intelligent management system encompassing risk identification, prediction and early warning, and decision-making response. The core contribution lies in the pioneering time-delay parameterized dynamic modeling method, which addresses the challenge of accurately characterizing the propagation of multi-level interruption events by quantifying the lag effects and cascading paths of risk conduction between nodes. A lightweight feature extraction architecture based on channel compression and integer quantization is designed, compressing the model size to the order of 0.5&#xa0;MB, enabling millisecond-level real-time response from edge devices. An event-driven cross-modal attention mechanism is developed, dynamically integrating multi-source heterogeneous information such as logistics monitoring images, inventory time-series data, and policy texts, strengthening the decision-making weight allocation of key risk signals. Empirical research shows that this framework exhibits significant advantages in the manufacturing, retail, and healthcare industries: it significantly improves the accuracy of interruption prediction compared to traditional methods, demonstrates robust performance under noise interference, and reduces the false alarm rate to industry-leading levels; through a dynamic strategy linkage mechanism, it effectively controls the loss caused by supply chain interruptions, and the edge deployment solution supports full-domain coverage from cloud servers to mobile terminals, achieving 28ms-level risk early warning response on devices such as Huawei Mate30. This achievement provides a technical paradigm for building resilient supply chain systems that combines theoretical rigor with engineering feasibility, significantly enhancing the digital risk prevention and control capabilities in the manufacturing and logistics industries.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial Intelligence for Intelligent, Resilient, and Sustainable Supply Chain Management

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月2日

查看全文
  • Zhonghuai Wang,
  • Yang Li

摘要

Amidst the intertwined complexity and fragility of global supply chain networks, there is an urgent need to break through the limitations of traditional risk management’s static modeling. This study innovatively integrates dynamic Bayesian networks with lightweight convolutional neural networks to construct a full-chain intelligent management system encompassing risk identification, prediction and early warning, and decision-making response. The core contribution lies in the pioneering time-delay parameterized dynamic modeling method, which addresses the challenge of accurately characterizing the propagation of multi-level interruption events by quantifying the lag effects and cascading paths of risk conduction between nodes. A lightweight feature extraction architecture based on channel compression and integer quantization is designed, compressing the model size to the order of 0.5 MB, enabling millisecond-level real-time response from edge devices. An event-driven cross-modal attention mechanism is developed, dynamically integrating multi-source heterogeneous information such as logistics monitoring images, inventory time-series data, and policy texts, strengthening the decision-making weight allocation of key risk signals. Empirical research shows that this framework exhibits significant advantages in the manufacturing, retail, and healthcare industries: it significantly improves the accuracy of interruption prediction compared to traditional methods, demonstrates robust performance under noise interference, and reduces the false alarm rate to industry-leading levels; through a dynamic strategy linkage mechanism, it effectively controls the loss caused by supply chain interruptions, and the edge deployment solution supports full-domain coverage from cloud servers to mobile terminals, achieving 28ms-level risk early warning response on devices such as Huawei Mate30. This achievement provides a technical paradigm for building resilient supply chain systems that combines theoretical rigor with engineering feasibility, significantly enhancing the digital risk prevention and control capabilities in the manufacturing and logistics industries.