Scholarlink

<p>To guarantee the desired performance of magnetorheological (MR) damper-controlled structural systems under uncertainties, a new reliability-based design optimization (RBDO) method is developed to efficiently optimize both controller parameters and damper deployment. First, the RBDO problem is tactfully addressed from the perspective of active learning-based quantile estimation. On this basis, a parallel multi-point enrichment is developed in the active learning workflow by integrating the stepwise uncertainty reduction with the weighted <i>K</i>-means clustering strategy. Further, the NSGA-II multi-objective optimization algorithm is employed to generate multiple design solutions per iteration, strategically selecting the training samples in regions around the design solutions. The proposed method is preliminarily verified on an analytical benchmark example, showcasing its favorable efficacy. Then, the optimization results of MR damper-controlled structures reveal consistent patterns across all Pareto-optimal solutions in control algorithms and damper deployment. These patterns indicate a strong underlying framework for optimal configurations, providing valuable insights into the systematic design of MR damper-controlled structures. A comparative analysis with an existing RBDO method further validates the efficacy of the proposed method in terms of addressing complex optimization challenges.</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

Fast parallel active learning for reliability-based design optimization of magnetorheological damper-controlled structures

  • 期刊论文

  • 发表日期：

    2025年7月21日

查看全文
  • Pei Pei,
  • Ser Tong Quek,
  • Tong Zhou,
  • Zhihong Pan,
  • Yongbo Peng

摘要

To guarantee the desired performance of magnetorheological (MR) damper-controlled structural systems under uncertainties, a new reliability-based design optimization (RBDO) method is developed to efficiently optimize both controller parameters and damper deployment. First, the RBDO problem is tactfully addressed from the perspective of active learning-based quantile estimation. On this basis, a parallel multi-point enrichment is developed in the active learning workflow by integrating the stepwise uncertainty reduction with the weighted K-means clustering strategy. Further, the NSGA-II multi-objective optimization algorithm is employed to generate multiple design solutions per iteration, strategically selecting the training samples in regions around the design solutions. The proposed method is preliminarily verified on an analytical benchmark example, showcasing its favorable efficacy. Then, the optimization results of MR damper-controlled structures reveal consistent patterns across all Pareto-optimal solutions in control algorithms and damper deployment. These patterns indicate a strong underlying framework for optimal configurations, providing valuable insights into the systematic design of MR damper-controlled structures. A comparative analysis with an existing RBDO method further validates the efficacy of the proposed method in terms of addressing complex optimization challenges.