Scholarlink

<p>Optimization algorithms are essential across science and engineering, and because problem scales grow, constraints get tighter, and the no-free-lunch theorem still applies, new methods keep appearing to trade exploration and exploitation in better ways while staying simple and fast. We present Attrition Mill Optimization (AMO), a physics-inspired metaheuristic modeled on attrition mills. AMO follows a compact three-stage structure—Initialization, Operation, and a Rejuvenation Cycle—to maintain a strong exploration–exploitation balance with very few controls. Built-in Opposition-Based Learning (OBL) delays premature convergence and preserves diversity in both low- and high-dimensional settings. Across 19 benchmark functions and three real engineering problems, AMO attains competitive or better fitness with short runtimes against 16 diverse metaheuristics. We also introduce Multi-Objective AMO (MOAMO), which adds an external archive, crowding-distance selection, and dynamic reference guidance. On the Zitzler–Deb–Thiele (ZDT) and Deb–Thiele–Laumanns–Zitzler (DTLZ) suites, MOAMO shows stable convergence toward the Pareto front with good coverage.</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

Attrition mill optimization algorithm: a novel approach for solving engineering optimization problems

  • 期刊论文

  • 发表日期：

    2025年11月13日

查看全文
  • Amir Mohammad Sharafaddini,
  • Behnam Mohammad Hasani Zade,
  • Najme Mansouri

摘要

Optimization algorithms are essential across science and engineering, and because problem scales grow, constraints get tighter, and the no-free-lunch theorem still applies, new methods keep appearing to trade exploration and exploitation in better ways while staying simple and fast. We present Attrition Mill Optimization (AMO), a physics-inspired metaheuristic modeled on attrition mills. AMO follows a compact three-stage structure—Initialization, Operation, and a Rejuvenation Cycle—to maintain a strong exploration–exploitation balance with very few controls. Built-in Opposition-Based Learning (OBL) delays premature convergence and preserves diversity in both low- and high-dimensional settings. Across 19 benchmark functions and three real engineering problems, AMO attains competitive or better fitness with short runtimes against 16 diverse metaheuristics. We also introduce Multi-Objective AMO (MOAMO), which adds an external archive, crowding-distance selection, and dynamic reference guidance. On the Zitzler–Deb–Thiele (ZDT) and Deb–Thiele–Laumanns–Zitzler (DTLZ) suites, MOAMO shows stable convergence toward the Pareto front with good coverage.