Scholarlink

<p>The issue of mitigating the high computational cost associated with high-fidelity simulation models has garnered significant attention in research, and the sequence approximation optimization method has emerged as the principal approach for tackling this challenge. This paper introduces an adaptive radial basis function neural network (RBFNN) approximation model, known as adaptive RBFNN based on multi-point sequences (ARMS-MIGA), which integrates the minimization prediction (MP) criterion, the maximum-minimum distance (MD) criterion, and the multi-island genetic algorithm (MIGA) for optimization. Two numerical examples of single-objective optimization are employed to validate the faster convergence capability and improved optimization performance of ARMS-MIGA. Furthermore, a weighted integrated improvement function construction method is proposed to extend ARMS-MIGA to multi-objective optimization design by integrating the multi-criteria decision-making method. Then, the feasibility of the proposed method is demonstrated through two specific engineering optimization case studies. Finally, a comparison of the ARMS-MIGA-based multi-objective optimization method with classical local and global multi-objective optimization methods highlights the ARMS-MIGA method's superior capability in achieving optimality and minimizing computation time. These results indicate that ARMS-MIGA provides a robust and efficient approach for multi-objective optimization in engineering applications, while providing critical insights into addressing complex design challenges across similar engineering applications.</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

A novel multi-objective optimization method based on adaptive RBFNN-MIGA with multi-point sequences

  • 期刊论文

  • 发表日期：

    2025年5月1日

查看全文
  • Xuan Zhou,
  • Ping Yu,
  • Jiangqi Long

摘要

The issue of mitigating the high computational cost associated with high-fidelity simulation models has garnered significant attention in research, and the sequence approximation optimization method has emerged as the principal approach for tackling this challenge. This paper introduces an adaptive radial basis function neural network (RBFNN) approximation model, known as adaptive RBFNN based on multi-point sequences (ARMS-MIGA), which integrates the minimization prediction (MP) criterion, the maximum-minimum distance (MD) criterion, and the multi-island genetic algorithm (MIGA) for optimization. Two numerical examples of single-objective optimization are employed to validate the faster convergence capability and improved optimization performance of ARMS-MIGA. Furthermore, a weighted integrated improvement function construction method is proposed to extend ARMS-MIGA to multi-objective optimization design by integrating the multi-criteria decision-making method. Then, the feasibility of the proposed method is demonstrated through two specific engineering optimization case studies. Finally, a comparison of the ARMS-MIGA-based multi-objective optimization method with classical local and global multi-objective optimization methods highlights the ARMS-MIGA method's superior capability in achieving optimality and minimizing computation time. These results indicate that ARMS-MIGA provides a robust and efficient approach for multi-objective optimization in engineering applications, while providing critical insights into addressing complex design challenges across similar engineering applications.