Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Development of a GPU-accelerated implicit material point method for geotechnical engineering
A graphic processing unit (GPU)-accelerated implicit material point method (IMPM) is proposed in this paper, aiming at solving large-scale geotechnical engineering problems efficiently. The Cholesky decomposition direct solution method and the preconditioned conjugate gradient (PCG) iteration method are implemented to solve the governing equation implicitly. In order to build an efficient parallel computation framework, the sequential processes in these solution methods are optimized by adopting advancing parallel computational algorithms. The risk of data race during parallel computation is avoided using atomic operation. The GPU-accelerated IMPM is firstly tested by a 1-D compress column and cantilever beam simulation to validate the accuracy of the proposed IMPM. Then, the computational efficiency is tested using the sand column collapse simulation. The solution of the governing equation is the most time-consuming process, occupying more than 95% of the computational time. The PCG iteration method shows higher efficiency compared to Cholesky decomposition direct solution method. By analysing the memory usage, it is found that memory occupation is the primary limitation on the simulation scale of IMPM, especially using the Cholesky decomposition direct solution method. Finally, the GPU-accelerated IMPM is implemented in the simulation of the Xinmo landslide, showing high accuracy and computational efficiency.
Development of a GPU-accelerated implicit material point method for geotechnical engineering
A graphic processing unit (GPU)-accelerated implicit material point method (IMPM) is proposed in this paper, aiming at solving large-scale geotechnical engineering problems efficiently. The Cholesky decomposition direct solution method and the preconditioned conjugate gradient (PCG) iteration method are implemented to solve the governing equation implicitly. In order to build an efficient parallel computation framework, the sequential processes in these solution methods are optimized by adopting advancing parallel computational algorithms. The risk of data race during parallel computation is avoided using atomic operation. The GPU-accelerated IMPM is firstly tested by a 1-D compress column and cantilever beam simulation to validate the accuracy of the proposed IMPM. Then, the computational efficiency is tested using the sand column collapse simulation. The solution of the governing equation is the most time-consuming process, occupying more than 95% of the computational time. The PCG iteration method shows higher efficiency compared to Cholesky decomposition direct solution method. By analysing the memory usage, it is found that memory occupation is the primary limitation on the simulation scale of IMPM, especially using the Cholesky decomposition direct solution method. Finally, the GPU-accelerated IMPM is implemented in the simulation of the Xinmo landslide, showing high accuracy and computational efficiency.
Development of a GPU-accelerated implicit material point method for geotechnical engineering
Acta Geotech.
Wang, Bin (Autor:in) / Chen, PengLin (Autor:in) / Wang, Di (Autor:in) / Liu, Lei-Lei (Autor:in) / Zhang, Wei (Autor:in)
Acta Geotechnica ; 19 ; 3729-3749
01.06.2024
21 pages
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
GPU-accelerated method , Implicit material point method , Large-scale landslides , Sparse matrix Engineering , Geoengineering, Foundations, Hydraulics , Solid Mechanics , Geotechnical Engineering & Applied Earth Sciences , Soil Science & Conservation , Soft and Granular Matter, Complex Fluids and Microfluidics
Development of a GPU-accelerated implicit material point method for geotechnical engineering
| Springer Verlag | 2024
Development of an implicit material point method for geotechnical applications
| British Library Online Contents | 2016
Development of an implicit material point method for geotechnical applications
| Online Contents | 2016
Improved double-point material point method for dynamic geotechnical problems
| DataCite | 2024