Scholarlink

<p>The paper presents a new model for the mass transport processes in fractured media of fractal structure. The proposed model considers the factors of time and space non-locality, particularly, of the moisture transport in the system of fractures. The finite-difference technique is presented for the solution of an initial-boundary value problem for the proposed model. The peculiarities of its implementation on GPU are analysed. The results of a series of numerical experiments prove the accuracy and convergence of the scheme and the efficiency of its parallel implementation showing up to <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\sim \)</EquationSource> <EquationSource Format="MATHML"><math> <mo>∼</mo> </math></EquationSource> </InlineEquation> 43-times acceleration compared with single-threaded execution on CPU. An additional 1.33-times acceleration can be achieved using a lower-precision data type with slight loss in the accuracy of solutions on meshes with less than 256 nodes. With the change of parameter values the new model can describe a wide range of transport processes’ behaviours, both accelerated and retarded compared with the classical model. Consideration of the proposed model as semi-empirical and fitting the values of its parameters to the measurements of moisture pressure and substance concentration can improve the accuracy of transport processes prediction.</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

Fractional-order simulation of mass transport in fractured saturated soils using GPU

  • 期刊论文

  • 发表日期：

    2025年9月8日

查看全文
  • Vsevolod Bohaienko,
  • Volodymyr Bulavatsky

摘要

The paper presents a new model for the mass transport processes in fractured media of fractal structure. The proposed model considers the factors of time and space non-locality, particularly, of the moisture transport in the system of fractures. The finite-difference technique is presented for the solution of an initial-boundary value problem for the proposed model. The peculiarities of its implementation on GPU are analysed. The results of a series of numerical experiments prove the accuracy and convergence of the scheme and the efficiency of its parallel implementation showing up to \(\sim \) ∼ 43-times acceleration compared with single-threaded execution on CPU. An additional 1.33-times acceleration can be achieved using a lower-precision data type with slight loss in the accuracy of solutions on meshes with less than 256 nodes. With the change of parameter values the new model can describe a wide range of transport processes’ behaviours, both accelerated and retarded compared with the classical model. Consideration of the proposed model as semi-empirical and fitting the values of its parameters to the measurements of moisture pressure and substance concentration can improve the accuracy of transport processes prediction.