Scholarlink

<p>This paper introduces an enhanced version of the whale optimization algorithm (WOA), which utilizes machine learning (ML) algorithms, including neural network (NN) and support vector regression (SVR), to optimize manufacturing problems. The WOA mimics the social behavior of humpback whales, particularly their bubble-net feeding strategy, and is widely used to solve various optimization problems due to its superiority over other evolutionary and swarm-based techniques. However, WOA exhibits some limitations, such as the trapping of local minima, a slower convergence rate, and a lack of tradeoff between exploration and exploitation. To address this challenge and improve performance, we optimized the WOA parameters using NN and SVR techniques. A dataset has been prepared to train the models. The proposed method has been compared with several WOA variants and established optimization algorithms across 23 well-known benchmark functions, four constrained engineering design problems, three machining optimization problems, and four unconstrained engineering design problems, demonstrating its robustness and broad applicability. Finally, with minor modifications, we applied to a discrete manufacturing problem, such as the machining operation sequence optimization. We demonstrated their effectiveness through various case studies from the literature. The experimental results show that the proposed algorithms yield comparable and improved results to other well-established techniques while solving different manufacturing problems. Specifically, for the machining operation sequence optimization problem, comparative results show that in Case Study 1, NN-WOA achieves improvements in the fitness function ranging from 4 to 46%, while SVR-WOA achieves improvements between 2 and 36% compared to other methods. In Case Study 2, NN-WOA shows improvements in the range of 10–17%, whereas SVR-WOA achieves improvements between 6 and 15%. Similarly, in Case Study 3, NN–WOA demonstrates improvements ranging from 2 to 27%, while SVR-WOA shows improvements between 2 and 25% relative to the other approaches considered.</p> Graphical abstract <p></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

Enhanced whale optimization algorithm for optimizing manufacturing problems through machine learning algorithms

  • 期刊论文

  • 发表日期：

    2026年7月31日

查看全文
  • Gobinda Chandra Behera,
  • Sankha Deb

摘要

This paper introduces an enhanced version of the whale optimization algorithm (WOA), which utilizes machine learning (ML) algorithms, including neural network (NN) and support vector regression (SVR), to optimize manufacturing problems. The WOA mimics the social behavior of humpback whales, particularly their bubble-net feeding strategy, and is widely used to solve various optimization problems due to its superiority over other evolutionary and swarm-based techniques. However, WOA exhibits some limitations, such as the trapping of local minima, a slower convergence rate, and a lack of tradeoff between exploration and exploitation. To address this challenge and improve performance, we optimized the WOA parameters using NN and SVR techniques. A dataset has been prepared to train the models. The proposed method has been compared with several WOA variants and established optimization algorithms across 23 well-known benchmark functions, four constrained engineering design problems, three machining optimization problems, and four unconstrained engineering design problems, demonstrating its robustness and broad applicability. Finally, with minor modifications, we applied to a discrete manufacturing problem, such as the machining operation sequence optimization. We demonstrated their effectiveness through various case studies from the literature. The experimental results show that the proposed algorithms yield comparable and improved results to other well-established techniques while solving different manufacturing problems. Specifically, for the machining operation sequence optimization problem, comparative results show that in Case Study 1, NN-WOA achieves improvements in the fitness function ranging from 4 to 46%, while SVR-WOA achieves improvements between 2 and 36% compared to other methods. In Case Study 2, NN-WOA shows improvements in the range of 10–17%, whereas SVR-WOA achieves improvements between 6 and 15%. Similarly, in Case Study 3, NN–WOA demonstrates improvements ranging from 2 to 27%, while SVR-WOA shows improvements between 2 and 25% relative to the other approaches considered.

Graphical abstract