Scholarlink

<p>Precisely determining the dynamic viscosity of pure fatty acid methyl esters (FAMEs) is a fundamental requirement for various energy and chemical sectors, particularly in the design and processing of biofuels. The present research leverages advanced machine learning frameworks to estimate FAME viscosity by incorporating crucial input features, specifically temperature, pressure, molar mass, and elemental composition including carbon (wt. %), hydrogen (wt. %), and oxygen (wt. %). To accomplish this, Gradient Boosting Decision Tree (GBDT) networks were structured and tested against a robust dataset containing <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(488\)</EquationSource> </InlineEquation> experimental observation points gathered from seven different types of FAMEs. The modeling phase utilized four distinct tuning strategies: Batch Bayesian Optimization (BBO), Evolution Strategies (ES), Bayesian Probability Improvement (BPI), and Gaussian Processes Optimization (GPO). To quantify the precision of the developed models, statistical metrics such as the coefficient of determination (<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\({R}^{2}\)</EquationSource> </InlineEquation>), Mean Squared Error (<InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(MSE\)</EquationSource> </InlineEquation>), and Average Absolute Relative Error (<InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(AARE\text{\%}\)</EquationSource> </InlineEquation>) were utilized. Through a rigorous comparative analysis, the GBDT-BPI approach was identified as the most highly accurate formulation, delivering an outstanding overall <InlineEquation ID="IEq5"> <EquationSource Format="TEX">\({R}^{2}\)</EquationSource> </InlineEquation> of <InlineEquation ID="IEq6"> <EquationSource Format="TEX">\(0.998716\)</EquationSource> </InlineEquation> coupled with an <InlineEquation ID="IEq7"> <EquationSource Format="TEX">\(MSE\)</EquationSource> </InlineEquation> of <InlineEquation ID="IEq8"> <EquationSource Format="TEX">\(0.003416\)</EquationSource> </InlineEquation>. Additionally, feature importance evaluations using the SHAP methodology indicated that system pressure and temperature act as the dominant driving forces dictating viscosity behavior, followed subsequently by hydrogen content, oxygen content, molar mass, and carbon content. The investigation also juxtaposed the computational demands of the applied optimization protocols. While BPI yielded superior accuracy, the GPO algorithm proved to be the most computationally economical, completing its execution in merely <InlineEquation ID="IEq9"> <EquationSource Format="TEX">\(182.9\)</EquationSource> </InlineEquation> seconds. Ultimately, these outcomes underline the profound capability of data-driven computational tools to rapidly and accurately evaluate the viscometric properties of FAMEs, offering an indispensable theoretical foundation for industrial operations seeking to maximize process efficiency and cost-effectiveness.</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

Rigorous data-driven scrutinization of viscosity for pure fatty acid methyl esters

  • 期刊论文

  • 发表日期：

    2026年5月22日

查看全文
  • Ahmad Adel Abu-Shareha,
  • Magdi E. A. Zaki,
  • Raed Alfilh,
  • Gadug Sudhamsu,
  • Prabhat Kumar Sahu,
  • Murari Devakannan Kamalesh,
  • Sumit Sharma,
  • Sobhi M. Gomha,
  • Soraya Hussaini

摘要

Precisely determining the dynamic viscosity of pure fatty acid methyl esters (FAMEs) is a fundamental requirement for various energy and chemical sectors, particularly in the design and processing of biofuels. The present research leverages advanced machine learning frameworks to estimate FAME viscosity by incorporating crucial input features, specifically temperature, pressure, molar mass, and elemental composition including carbon (wt. %), hydrogen (wt. %), and oxygen (wt. %). To accomplish this, Gradient Boosting Decision Tree (GBDT) networks were structured and tested against a robust dataset containing \(488\) experimental observation points gathered from seven different types of FAMEs. The modeling phase utilized four distinct tuning strategies: Batch Bayesian Optimization (BBO), Evolution Strategies (ES), Bayesian Probability Improvement (BPI), and Gaussian Processes Optimization (GPO). To quantify the precision of the developed models, statistical metrics such as the coefficient of determination ( \({R}^{2}\) ), Mean Squared Error ( \(MSE\) ), and Average Absolute Relative Error ( \(AARE\text{\%}\) ) were utilized. Through a rigorous comparative analysis, the GBDT-BPI approach was identified as the most highly accurate formulation, delivering an outstanding overall \({R}^{2}\) of \(0.998716\) coupled with an \(MSE\) of \(0.003416\) . Additionally, feature importance evaluations using the SHAP methodology indicated that system pressure and temperature act as the dominant driving forces dictating viscosity behavior, followed subsequently by hydrogen content, oxygen content, molar mass, and carbon content. The investigation also juxtaposed the computational demands of the applied optimization protocols. While BPI yielded superior accuracy, the GPO algorithm proved to be the most computationally economical, completing its execution in merely \(182.9\) seconds. Ultimately, these outcomes underline the profound capability of data-driven computational tools to rapidly and accurately evaluate the viscometric properties of FAMEs, offering an indispensable theoretical foundation for industrial operations seeking to maximize process efficiency and cost-effectiveness.