Scholarlink

<p>Understanding the permeability of granular systems is essential for flow-related studies, but it is complex due to the influence of grain morphology. This paper proposes new explicit interpretable permeability and pore models for granular materials using the multivariate adaptive regression splines (MARS) algorithm. The models use a dataset generated in a previous study and establish connections between particle morphology descriptors and permeability parameters, as well as pore indexes. Within this framework, correlation of permeability and pore characteristics is also investigated. Sensitivity analyses are then performed on the developed hydraulic conductivity model. The proposed explicit models are simple in form and have fewer parameters. The models were trained using fivefold cross-validation on 80% of the randomly selected data from the database, and then tested on the remaining 20% of the data. The effectiveness of the models is quantitatively evaluated through statistical indicators. Results show high <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(R^{2}\)</EquationSource> <EquationSource Format="MATHML"><math> <msup> <mi>R</mi> <mn>2</mn> </msup> </math></EquationSource> </InlineEquation> values (&gt; 0.73) and low RMSE (close to zero), demonstrating the effectiveness of the proposed models. This work provides valuable insights into the impact of particle morphology on permeability and contributes to the development of explicit permeability models for granular materials.</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

Machine Learning-Based Interpretable Formalization of Permeability Using Particle Morphology Descriptors

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Hadi Fathipour-Azar

摘要

Understanding the permeability of granular systems is essential for flow-related studies, but it is complex due to the influence of grain morphology. This paper proposes new explicit interpretable permeability and pore models for granular materials using the multivariate adaptive regression splines (MARS) algorithm. The models use a dataset generated in a previous study and establish connections between particle morphology descriptors and permeability parameters, as well as pore indexes. Within this framework, correlation of permeability and pore characteristics is also investigated. Sensitivity analyses are then performed on the developed hydraulic conductivity model. The proposed explicit models are simple in form and have fewer parameters. The models were trained using fivefold cross-validation on 80% of the randomly selected data from the database, and then tested on the remaining 20% of the data. The effectiveness of the models is quantitatively evaluated through statistical indicators. Results show high \(R^{2}\) R 2 values (> 0.73) and low RMSE (close to zero), demonstrating the effectiveness of the proposed models. This work provides valuable insights into the impact of particle morphology on permeability and contributes to the development of explicit permeability models for granular materials.