Scholarlink

<p>The experimental data were used to predict equations for the dynamic viscosity and thermal conductivity of six nanofluids employing various machine learning methods. Nanoparticles were oxidized multi-walled carbon nanotubes (MWCNT) and surfactant-assisted MWCNT, dispersed in water and ethylene glycol (EG) ratios as 0:100, 10:90, and 20:80. The data were collected for concentrations of 0.125, 0.25, and 0.5 mass% and temperatures of 50–70 °C. This study applies seven machine learning methods: the Ordinary Least Squares (OLS), the Huber regression, the Decision tree regressor, the AdaBoost regressor, the Theil-Sen regressor, Random forest (RF), Support vector machine (SVM) in the linear and the polynomial formation independently reach the best <i>R</i>-squared (<i>R</i><sup>2</sup>). Using the EG temperature and the EG thermal conductivity, the models and the created formulas for the thermal conductivity of the EG/H<sub>2</sub>O (90:10) and thermal conductivity of the EG/H<sub>2</sub>O (80:20). The best <i>R</i><sup>2</sup> for making the formula and model to the EG/H<sub>2</sub>O (90:10) thermal conductivity, respectively, are 0.995, 0.999 reached by the Theil-Sen regressor and the RF and SVM (kernel = RBF) regressors in linear formation. For the EG/H<sub>2</sub>O (80:20), thermal conductivity is 0.992 and 0.998, respectively, reached by the OLS (polynomial formation) and the RF (linear formation) regressor. The EG + 0.5% oxidized MWCNTs temperature and thermal conductivity are the base for creating the formula and model for the EG/H<sub>2</sub>O (90:10) + 0.5% oxidized MWCNTs thermal conductivity, and the EG/H<sub>2</sub>O (80:20) + 0.5% oxidized MWCNTs thermal conductivity, the <i>R</i><sup>2</sup> are, respectively, 0.996 and 0.999 in polynomial and the linear formations reach by the OLS and the RF regressor, and 0.962 and 0.998 in the polynomial and the linear formations by the OLS (for formula) and the AdaBoost and the RF (for model). For reaching the dynamic viscosity of the EG/H<sub>2</sub>O (90:10) and the EG/H<sub>2</sub>O (80:20) using the EG temperature and dynamic viscosity, and the best <i>R</i><sup>2</sup> results for building formula and model, respectively, are 0.998 and 0.999 in polynomial and linear formation by the OLS and the RF, and 0.999 and 0.999 by the Theil-Sen regressor and the RF regressor.</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

Ethylene glycol-based nanofluids: machine learning predictions for improved solar thermal performance

  • 期刊论文

  • 发表日期：

    2025年5月22日

查看全文
  • Nazim Hasan,
  • Hakim S. Sultan Aljibori,
  • Othman Hakami,
  • Abdullah Ali Alamri,
  • Shadma Tasneem,
  • Mohammad Ehtisham Khan,
  • Abdul Amir H. Kadhum,
  • Mahidzal Dahari,
  • M. R. Safaei

摘要

The experimental data were used to predict equations for the dynamic viscosity and thermal conductivity of six nanofluids employing various machine learning methods. Nanoparticles were oxidized multi-walled carbon nanotubes (MWCNT) and surfactant-assisted MWCNT, dispersed in water and ethylene glycol (EG) ratios as 0:100, 10:90, and 20:80. The data were collected for concentrations of 0.125, 0.25, and 0.5 mass% and temperatures of 50–70 °C. This study applies seven machine learning methods: the Ordinary Least Squares (OLS), the Huber regression, the Decision tree regressor, the AdaBoost regressor, the Theil-Sen regressor, Random forest (RF), Support vector machine (SVM) in the linear and the polynomial formation independently reach the best R-squared (R2). Using the EG temperature and the EG thermal conductivity, the models and the created formulas for the thermal conductivity of the EG/H2O (90:10) and thermal conductivity of the EG/H2O (80:20). The best R2 for making the formula and model to the EG/H2O (90:10) thermal conductivity, respectively, are 0.995, 0.999 reached by the Theil-Sen regressor and the RF and SVM (kernel = RBF) regressors in linear formation. For the EG/H2O (80:20), thermal conductivity is 0.992 and 0.998, respectively, reached by the OLS (polynomial formation) and the RF (linear formation) regressor. The EG + 0.5% oxidized MWCNTs temperature and thermal conductivity are the base for creating the formula and model for the EG/H2O (90:10) + 0.5% oxidized MWCNTs thermal conductivity, and the EG/H2O (80:20) + 0.5% oxidized MWCNTs thermal conductivity, the R2 are, respectively, 0.996 and 0.999 in polynomial and the linear formations reach by the OLS and the RF regressor, and 0.962 and 0.998 in the polynomial and the linear formations by the OLS (for formula) and the AdaBoost and the RF (for model). For reaching the dynamic viscosity of the EG/H2O (90:10) and the EG/H2O (80:20) using the EG temperature and dynamic viscosity, and the best R2 results for building formula and model, respectively, are 0.998 and 0.999 in polynomial and linear formation by the OLS and the RF, and 0.999 and 0.999 by the Theil-Sen regressor and the RF regressor.