Scholarlink

<p>Thermal diffusivity (TD) is one of significantly important parameters in the design of ground heat exchangers for hydrothermal coupling as well as heat–mass transfer, which depends on the soil type, density and water content. Due to the dependence on several soil properties, estimating TD in the laboratory is quite laborious. As a result, machine learning (ML) algorithms have gained significant attention in geotechnical engineering due to their ability to predict complex nonlinear relationships. Since the prediction of soil’s TD remains an unexplored area, this paper presents an estimation of TD of non-plastic soils using four machine learning approaches: ANN, SVM, ANN-PSO and XGB. The results indicate that the XGB is best-performing ML model followed by ANN, SVM and ANN-PSO. Further, XGB-SHAP analysis indicates that the water content (<i>w</i>), sand (<i>S</i>) and dry density (<i>γ</i><sub><i>d</i></sub>) are the majorly affecting input parameters for predicting TD of non-plastic soil. Moreover, it can be stated that the use of ML models will overcome the cost and labor-intensive drawbacks of experimental measurement of TD of non-plastic soils and thereby enabling sustainable applications in energy geotechnical projects.</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

Prediction of Thermal Diffusivity of Non-Plastic Soil for the Design of Ground Heat Exchanger Using Machine Learning Approach

  • 期刊论文

  • 发表日期：

    2025年5月1日

查看全文
  • Namit Jaiswal,
  • Pawan Kishor Sah,
  • Shiv Shankar Kumar,
  • Bhabani Shankar Das,
  • Anubhav Baranwal

摘要

Thermal diffusivity (TD) is one of significantly important parameters in the design of ground heat exchangers for hydrothermal coupling as well as heat–mass transfer, which depends on the soil type, density and water content. Due to the dependence on several soil properties, estimating TD in the laboratory is quite laborious. As a result, machine learning (ML) algorithms have gained significant attention in geotechnical engineering due to their ability to predict complex nonlinear relationships. Since the prediction of soil’s TD remains an unexplored area, this paper presents an estimation of TD of non-plastic soils using four machine learning approaches: ANN, SVM, ANN-PSO and XGB. The results indicate that the XGB is best-performing ML model followed by ANN, SVM and ANN-PSO. Further, XGB-SHAP analysis indicates that the water content (w), sand (S) and dry density (γd) are the majorly affecting input parameters for predicting TD of non-plastic soil. Moreover, it can be stated that the use of ML models will overcome the cost and labor-intensive drawbacks of experimental measurement of TD of non-plastic soils and thereby enabling sustainable applications in energy geotechnical projects.