Scholarlink

<p>Process intensification can improve esterification productivity with lower energy demand. This work develops a machine learning assisted Barnacles Mating Optimization (BMO) framework for solvent free isoamyl acetate synthesis in a miniaturized intensified reactor, targeting ester maximization and energy minimization. Decision tree, SVM, and ANN models were trained to predict ester concentration from temperature, acid to alcohol ratio, flowrate, and retention time, and bootstrap resampling improved all models. The bootstrap ANN achieved the highest accuracy (R<sup>2</sup> = 0.989) and was selected as the surrogate for optimization. Using ANN BMO, ester maximization converged to 45.343&#xa0;°C, ratio 0.6116, 51.092&#xa0;µL min<sup>−1</sup>, and 42.701&#xa0;min, yielding 1.240&#xa0;mol L<sup>−1</sup> ester with 47.056&#xa0;mW energy. Energy minimization produced a boundary solution at 30.0&#xa0;°C, ratio 0.60, 40.0&#xa0;µL min<sup>−1</sup>, and 33.0&#xa0;min, giving 9.22&#xa0;mW but low ester (0.2559&#xa0;mol L<sup>−1</sup>). Multi objective optimization generated a Pareto trade off where higher ester requires higher energy, with temperature identified as the primary decision variable driving the compromise.</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 Assisted-Barnacles Mating Optimization of non-enzymatic isoamyl acetate in a solvent-free system using miniaturized intensified reactor

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月21日

查看全文
  • Fakhrony Sholahudin Rohman,
  • Dinie Muhammad,
  • Nurhazwani Yusoff Azudin,
  • Muhamad Nazri Murat,
  • Cheong Sheng Lee,
  • Siti Nor Azreen Ahmad Termizi,
  • Ashraf Azmi,
  • Syamsul Rizal Abd Shukor

摘要

Process intensification can improve esterification productivity with lower energy demand. This work develops a machine learning assisted Barnacles Mating Optimization (BMO) framework for solvent free isoamyl acetate synthesis in a miniaturized intensified reactor, targeting ester maximization and energy minimization. Decision tree, SVM, and ANN models were trained to predict ester concentration from temperature, acid to alcohol ratio, flowrate, and retention time, and bootstrap resampling improved all models. The bootstrap ANN achieved the highest accuracy (R2 = 0.989) and was selected as the surrogate for optimization. Using ANN BMO, ester maximization converged to 45.343 °C, ratio 0.6116, 51.092 µL min−1, and 42.701 min, yielding 1.240 mol L−1 ester with 47.056 mW energy. Energy minimization produced a boundary solution at 30.0 °C, ratio 0.60, 40.0 µL min−1, and 33.0 min, giving 9.22 mW but low ester (0.2559 mol L−1). Multi objective optimization generated a Pareto trade off where higher ester requires higher energy, with temperature identified as the primary decision variable driving the compromise.