Scholarlink

<p>The cable circuit design is required to optimize the complex interdependencies of parameters like power losses, signal quality, and consumed materials, and such types of multicriteria optimization approaches are not valid anymore in many cases under traditional optimization methods. The paper discusses the Knowledge Graph-Based Optimization Framework (KGNOF) that solves this problem by integrating relational graph convolutional networks with the non-dominated sorting genetic algorithm II. The new approach based on knowledge graphs understood the hierarchical relations between circuit parameters and NSGA-II works towards finding the diverse set of Pareto-optimal solutions. This new approach improves the accuracy of predictions considerably: RMSE is 0.0123 and MAE is 0.0105 versus traditional approaches. Accordingly, it improves 12% of enhanced diversity along the Pareto front concerning classical NSGA-II types, thereby improving the trade-off between conflicting optimization objectives. Sensitivity analyses identify the set of critical parameters that affect performance, and cluster analysis provides further validation of the framework’s performance in correctly mapping circuit topologies. The model is scalable and computationally efficient, making it fit for applications at industrial scales. This study partially filling a gap in the literature and directly providing the task of cable circuit design a more intuitive and practical solution.</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

Advancing Multi-Objective Optimization in Cable Circuit Design Using Knowledge Graph-Based Framework

  • 期刊论文

  • 发表日期：

    2025年9月16日

查看全文
  • Fengyi He,
  • Xiaoming Su,
  • Ruihua Li,
  • Binyan Zheng,
  • Yan Yang

摘要

The cable circuit design is required to optimize the complex interdependencies of parameters like power losses, signal quality, and consumed materials, and such types of multicriteria optimization approaches are not valid anymore in many cases under traditional optimization methods. The paper discusses the Knowledge Graph-Based Optimization Framework (KGNOF) that solves this problem by integrating relational graph convolutional networks with the non-dominated sorting genetic algorithm II. The new approach based on knowledge graphs understood the hierarchical relations between circuit parameters and NSGA-II works towards finding the diverse set of Pareto-optimal solutions. This new approach improves the accuracy of predictions considerably: RMSE is 0.0123 and MAE is 0.0105 versus traditional approaches. Accordingly, it improves 12% of enhanced diversity along the Pareto front concerning classical NSGA-II types, thereby improving the trade-off between conflicting optimization objectives. Sensitivity analyses identify the set of critical parameters that affect performance, and cluster analysis provides further validation of the framework’s performance in correctly mapping circuit topologies. The model is scalable and computationally efficient, making it fit for applications at industrial scales. This study partially filling a gap in the literature and directly providing the task of cable circuit design a more intuitive and practical solution.