Scholarlink

<p>The growing demand for cleaner energy and the reduction of greenhouse gas emissions have highlighted anhydrous ethanol as a key renewable fuel due to its applications as a fuel, gasoline additive, and biodiesel reactant. Traditional dehydration via azeotropic distillation with cyclohexane is energy-intensive and environmentally harmful, prompting the search for safer alternatives. This study modeled and simulated the ethanol dehydration process using monoethylene glycol in Aspen Plus v12.1, implementing conventional control strategies optimized by Luyben’s method, which identified tray 31 as the system’s most sensitive point. Dynamic simulations demonstrated process robustness, with rapid recovery from disturbances in feed and heat duty, maintaining 99.999 wt% ethanol purity and solvent recovery. Additionally, artificial intelligence (AI) models—specifically decision tree, random forest, and LightGBM—were developed to control top product composition using easily measurable variables. These models significantly outperformed linear regression, with the decision tree achieving R² = 0.9970, MAE = 4.19 × 10⁻⁷, and RMSE = 2.20 × 10⁻⁶, maintaining ethanol molar fractions above 0.9996 even under dynamic disturbances. Despite their strong performance, the industrial adoption of AI-based controllers is still limited. However, with the implementation of real-time validation, significant advancements in computational processing capacity, and improvements in techniques to prevent overfitting, artificial intelligence models can become a promising tool for industrial process control.</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

Implementation of advanced control with artificial intelligence in anhydrous ethanol production

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月11日

查看全文
  • Marcos Gabriel Lopes da Silva,
  • Davi Soares de Lima,
  • Andréa Selene Embirassú Xavier,
  • Allan Almeida Albuquerque

摘要

The growing demand for cleaner energy and the reduction of greenhouse gas emissions have highlighted anhydrous ethanol as a key renewable fuel due to its applications as a fuel, gasoline additive, and biodiesel reactant. Traditional dehydration via azeotropic distillation with cyclohexane is energy-intensive and environmentally harmful, prompting the search for safer alternatives. This study modeled and simulated the ethanol dehydration process using monoethylene glycol in Aspen Plus v12.1, implementing conventional control strategies optimized by Luyben’s method, which identified tray 31 as the system’s most sensitive point. Dynamic simulations demonstrated process robustness, with rapid recovery from disturbances in feed and heat duty, maintaining 99.999 wt% ethanol purity and solvent recovery. Additionally, artificial intelligence (AI) models—specifically decision tree, random forest, and LightGBM—were developed to control top product composition using easily measurable variables. These models significantly outperformed linear regression, with the decision tree achieving R² = 0.9970, MAE = 4.19 × 10⁻⁷, and RMSE = 2.20 × 10⁻⁶, maintaining ethanol molar fractions above 0.9996 even under dynamic disturbances. Despite their strong performance, the industrial adoption of AI-based controllers is still limited. However, with the implementation of real-time validation, significant advancements in computational processing capacity, and improvements in techniques to prevent overfitting, artificial intelligence models can become a promising tool for industrial process control.

Graphical abstract