• 全国中文核心期刊
  • 中国科技核心期刊
  • 美国工程索引(EI)收录期刊
  • Scopus数据库收录期刊
王荣鑫, 普建明, 张建民, 王睿. 高性能岩土工程流固耦合仿真方法研究[J]. 岩土工程学报. DOI: 10.11779/CJGE20240382
引用本文: 王荣鑫, 普建明, 张建民, 王睿. 高性能岩土工程流固耦合仿真方法研究[J]. 岩土工程学报. DOI: 10.11779/CJGE20240382
High-performance Solid-fluid Coupled Simulation Method for Geotechnical Engineering[J]. Chinese Journal of Geotechnical Engineering. DOI: 10.11779/CJGE20240382
Citation: High-performance Solid-fluid Coupled Simulation Method for Geotechnical Engineering[J]. Chinese Journal of Geotechnical Engineering. DOI: 10.11779/CJGE20240382

高性能岩土工程流固耦合仿真方法研究

High-performance Solid-fluid Coupled Simulation Method for Geotechnical Engineering

  • 摘要: 超大型岩土工程的数值仿真需求对于计算能力提出了新的挑战。为了研究饱和土层中复杂结构的静动力响应,需要发展基于高性能计算的岩土体流固耦合求解方法。本文提出一种有限元-有限体积(FEM-FVM)流固耦合仿真方法,并发展了适应于并行求解的全隐式流固耦合算法(FICM)、全显式流固耦合算法(FECM)、固显流隐耦合算法(SEFIM)。文章通过对比分析不同方法的求解精度、单核计算效率和并行计算性能,为算法选用提供了理论支持。在求解精度方面,三种方法均能较为精确的求解太沙基一维固结问题。在单核计算效率方面,对比发现FICM对于求解小规模静力流固耦合具有一定优势,而FECM和SEFIM更适合求解动力流固耦合问题。在并行计算性能方面,对比分析了三种不同方法的并行可拓展性与加速比,并针对FECM进行了不同有效单核负载下的可扩展性研究和不同总负载下的加速比研究。结果表明,对于千万自由度量级大规模岩土工程流固耦合仿真问题,FECM和SEFIM均具有较大的优势。其中,显式的FECM方法对于上亿自由度的超大规模问题仍能保持较高的并行加速比。

     

    Abstract: Research on mega scale geotechnical engineering projects has raised new challenges for computational capability. In order to study the static and dynamic response of complex structures in saturated soil, it is essential to develop high-performance computing methods for porous media solid-fluid coupled analysis. This study proposes a finite element method-finite volume method (FEM-FVM) coupled framework for parallel solution of solid-fluid coupling problems, and develops three solution methods, including: Fully Implicit Coupled Method (FICM), Fully Explicit Coupled Method (FECM), and Solid Explicit Fluid Implicit Method (SEFIM). A comparative analysis of these methods in terms of accuracy, single core efficiency, and parallel performance is conducted, providing theoretical support and practical guidance for appropriate method selection. Regarding accuracy, the study analyzes the excess pore water pressure time-history obtained by different methods for a Terzaghi one-dimensional consolidation problem, validating all three methods. In terms of single core efficiency, the study shows that for small scale problems, FICM is more suited for static problems, whereas FECM and SEFIM are more suitable for dynamic problems. Regarding parallel computing performance, the analysis of scalability and speedup indicates that FECM and SEFIM have significant advantage over the fully implicit method when solving large scale problems over 10 million degrees-of-freedom. Especially, the explicit FECM can achieve high scalability for problems with over 100 million degrees-of-freedom.

     

/

返回文章
返回