Scholarlink

<p>This study presents a parametric design and optimization approach for bucket drum lunar regolith collector. Using discrete element method (DEM) simulations, the operational performance of the collector was analyzed, focusing on filling efficiency, collection rate, and evacuation rate. Three surrogate models—radial basis function (RBF), Gaussian process regression (GPR), and support vector regression (SVR)—were constructed to form a composite surrogate model. The performance of four multi-objective optimization algorithms (MOPSO, NSGA-II, SPEA-II, PESA-II) was compared, with MOPSO demonstrating the best results. An adaptive surrogate model invocation mechanism based on absolute error of leave-one-out cross-validation (AELOOCV) further enhanced optimization accuracy. The entropy weight method and TOPSIS were employed to select the optimal solution from the Pareto set, leading to improvements of 10.353% in filling efficiency, 13.275% in collection rate, and 12.070% in evacuation rate. The study highlights the effectiveness of combining surrogate models with advanced optimization algorithms in lunar soil collection design.</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

Multi-objective optimization of bucket drum for lunar regolith collectors with multi-surrogate model based on adaptive invocation mechanism

  • 期刊论文

  • 发表日期：

    2025年4月16日

查看全文
  • Haoran Li,
  • Yuyue Gao,
  • Lieyun Ding,
  • Cheng Zhou,
  • Shifeng Wen,
  • Yan Zhou

摘要

This study presents a parametric design and optimization approach for bucket drum lunar regolith collector. Using discrete element method (DEM) simulations, the operational performance of the collector was analyzed, focusing on filling efficiency, collection rate, and evacuation rate. Three surrogate models—radial basis function (RBF), Gaussian process regression (GPR), and support vector regression (SVR)—were constructed to form a composite surrogate model. The performance of four multi-objective optimization algorithms (MOPSO, NSGA-II, SPEA-II, PESA-II) was compared, with MOPSO demonstrating the best results. An adaptive surrogate model invocation mechanism based on absolute error of leave-one-out cross-validation (AELOOCV) further enhanced optimization accuracy. The entropy weight method and TOPSIS were employed to select the optimal solution from the Pareto set, leading to improvements of 10.353% in filling efficiency, 13.275% in collection rate, and 12.070% in evacuation rate. The study highlights the effectiveness of combining surrogate models with advanced optimization algorithms in lunar soil collection design.