Scholarlink

<p>Despite the large number of metaheuristic algorithms developed for solving optimization problems, nature-inspired algorithms still face challenges such as premature convergence, getting trapped in local optima, and failing to maintain a proper balance between exploration and exploitation. This study aims to develop a robust optimization algorithm that addresses these limitations by maintaining an effective balance between global exploration and local exploitation, thereby improving convergence toward optimal solutions. In this paper, the tree pruning optimization (TPO) algorithm is proposed, inspired by arborists' tree pruning methods. TPO begins by removing dead or diseased branches and then selects the best leader by eliminating competing leaders. The removal of weaker and suckering branches represents the exploration phase, while selecting a dominant leader and removing other competing branches represents the exploitation phase. These two phases are controlled using a random variable. Elitism and a crossover operator are employed to select the best branches for the next iteration. To evaluate the proposed TPO algorithm, benchmark suites from CEC2019 (10 functions) and CEC2022 (12 functions) were used, comparing TPO against nine well-known and competitive optimization algorithms. TPO outperformed all compared algorithms individually in at least 7 out of 10 functions in CEC2019 and 8 out of 12 functions in CEC2022. It achieved strong average ranking scores of 1.8 and 1.4, respectively. Statistical analyses confirmed the superior performance of TPO in most cases, supported by <i>p</i> values (<i>p</i> &lt; 0.05) and boxplot visualizations. Furthermore, TPO demonstrated its effectiveness on real-world engineering problems. It achieved the best results and ranked first in solving the speed reducer design, tension/compression spring design, and welded beam design problems, outperforming all competitors in terms of minimizing cost and satisfying constraints.</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 metaheuristic algorithm inspired by tree pruning for solving benchmark optimization and engineering design problems

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Hardi M. Mohammed

摘要

Despite the large number of metaheuristic algorithms developed for solving optimization problems, nature-inspired algorithms still face challenges such as premature convergence, getting trapped in local optima, and failing to maintain a proper balance between exploration and exploitation. This study aims to develop a robust optimization algorithm that addresses these limitations by maintaining an effective balance between global exploration and local exploitation, thereby improving convergence toward optimal solutions. In this paper, the tree pruning optimization (TPO) algorithm is proposed, inspired by arborists' tree pruning methods. TPO begins by removing dead or diseased branches and then selects the best leader by eliminating competing leaders. The removal of weaker and suckering branches represents the exploration phase, while selecting a dominant leader and removing other competing branches represents the exploitation phase. These two phases are controlled using a random variable. Elitism and a crossover operator are employed to select the best branches for the next iteration. To evaluate the proposed TPO algorithm, benchmark suites from CEC2019 (10 functions) and CEC2022 (12 functions) were used, comparing TPO against nine well-known and competitive optimization algorithms. TPO outperformed all compared algorithms individually in at least 7 out of 10 functions in CEC2019 and 8 out of 12 functions in CEC2022. It achieved strong average ranking scores of 1.8 and 1.4, respectively. Statistical analyses confirmed the superior performance of TPO in most cases, supported by p values (p < 0.05) and boxplot visualizations. Furthermore, TPO demonstrated its effectiveness on real-world engineering problems. It achieved the best results and ranked first in solving the speed reducer design, tension/compression spring design, and welded beam design problems, outperforming all competitors in terms of minimizing cost and satisfying constraints.