Scholarlink

<p>This study investigates the physical properties of the ionic liquid amide ([P222(1O1)][TFSA]), focusing on densities, molality (<i>m</i><sub><i>1</i></sub>), and CO₂ solubility expressed in terms of mole fraction (<i>x₁</i>) as functions of equilibrium pressure (<i>p</i>). Utilizing response surface methodology (RSM) and artificial neural networks, the research spans high-pressure conditions (up to 50&#xa0;MPa) and temperatures from 273.15 to 353.15&#xa0;K for density measurements, and temperatures from 303.15 to 333.15&#xa0;K with pressures up to 6&#xa0;MPa for CO<sub>2</sub> solubility and molality. The strong statistical significance of the model is underscored by high F values and p values less than 0.05, indicating that the independent variables significantly influence the outcomes. The close alignment of predicted R² and Adjusted R² values demonstrates the accuracy of the RSM models in estimating the studied properties. Additionally, a multilayer perceptron architecture was optimized, revealing that configurations with 15 and 10 neurons yielded optimal results for CO<sub>2</sub> solubility and density, respectively. A comparative analysis of training algorithms identified the Levenberg–Marquardt method as the most effective, obtaining the minimal MSE for CO<sub>2</sub> solubility (0.00012921) and molality (0.00048875).</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

RSM and ANN approach for modeling of TFSA amide ionic liquid: insights into CO2 solubility and equilibrium pressure

  • 期刊论文

  • 发表日期：

    2026年5月21日

查看全文
  • Zohreh Khoshraftar

摘要

This study investigates the physical properties of the ionic liquid amide ([P222(1O1)][TFSA]), focusing on densities, molality (m1), and CO₂ solubility expressed in terms of mole fraction (x₁) as functions of equilibrium pressure (p). Utilizing response surface methodology (RSM) and artificial neural networks, the research spans high-pressure conditions (up to 50 MPa) and temperatures from 273.15 to 353.15 K for density measurements, and temperatures from 303.15 to 333.15 K with pressures up to 6 MPa for CO2 solubility and molality. The strong statistical significance of the model is underscored by high F values and p values less than 0.05, indicating that the independent variables significantly influence the outcomes. The close alignment of predicted R² and Adjusted R² values demonstrates the accuracy of the RSM models in estimating the studied properties. Additionally, a multilayer perceptron architecture was optimized, revealing that configurations with 15 and 10 neurons yielded optimal results for CO2 solubility and density, respectively. A comparative analysis of training algorithms identified the Levenberg–Marquardt method as the most effective, obtaining the minimal MSE for CO2 solubility (0.00012921) and molality (0.00048875).