Scholarlink

<p>This study addresses the challenges of efficient modeling and machine learning (ML) hyperparameter optimization (HPO) for active magnetic bearing rotors (AMBs-rotor) by proposing an Improved Gorilla Troops Optimizer (IGTO). The IGTO integrates Circle chaotic mapping, Lévy–Cauchy flights, and a generalized opposition-based learning (GOBL) strategy to substantially enhance global search capabilities and effectively avoid local optima. Benchmark experiments on twelve classical test functions demonstrate that IGTO outperforms the standard GTO and other state-of-the-art optimizers in both exploration and exploitation. Furthermore, IGTO autonomously and efficiently completes HPO for three representative ensemble learning models, surpassing both GTO and random search (RNG) methods. A surrogate model for the AMBs-rotor, incorporating both control gains and bias current, constructed via IGTO shows significant improvements in predictive accuracy and robustness compared to conventional HPO-tuned ML models. Simultaneously, an ablation removing Circle, Lévy–Cauchy, and GOBL in turn shows each is beneficial; the complete IGTO attains the lowest MSE and highest <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^{2}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation>.When applied to critical speed optimization, the IGTO-based surrogate not only greatly reduces computational cost but also shortens the testing cycle, offering an efficient and cost-effective solution for AMBs-rotor critical speed design.</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

Machine learning surrogate model optimized by improved gorilla troops optimizer for rotor dynamic optimization of active magnetic bearings

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Jiahang Cui,
  • Jiahong Li,
  • Feichao Cai,
  • Zhenmin Zhao

摘要

This study addresses the challenges of efficient modeling and machine learning (ML) hyperparameter optimization (HPO) for active magnetic bearing rotors (AMBs-rotor) by proposing an Improved Gorilla Troops Optimizer (IGTO). The IGTO integrates Circle chaotic mapping, Lévy–Cauchy flights, and a generalized opposition-based learning (GOBL) strategy to substantially enhance global search capabilities and effectively avoid local optima. Benchmark experiments on twelve classical test functions demonstrate that IGTO outperforms the standard GTO and other state-of-the-art optimizers in both exploration and exploitation. Furthermore, IGTO autonomously and efficiently completes HPO for three representative ensemble learning models, surpassing both GTO and random search (RNG) methods. A surrogate model for the AMBs-rotor, incorporating both control gains and bias current, constructed via IGTO shows significant improvements in predictive accuracy and robustness compared to conventional HPO-tuned ML models. Simultaneously, an ablation removing Circle, Lévy–Cauchy, and GOBL in turn shows each is beneficial; the complete IGTO attains the lowest MSE and highest \(R^{2}\) R 2 .When applied to critical speed optimization, the IGTO-based surrogate not only greatly reduces computational cost but also shortens the testing cycle, offering an efficient and cost-effective solution for AMBs-rotor critical speed design.