Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

GPU-accelerated Shell Simulation

: Junga, Sven
: Stork, André; Mueller-Roemer, Johannes

Darmstadt, 2020, 56 S.
Darmstadt, TU, Master Thesis, 2020
Master Thesis
Fraunhofer IGD ()
FEM simulation; GPU computing; GPU-based simulation; deformable models; Lead Topic: Digitized Work; Research Line: (Interactive) simulation (SIM)

In the context of finite element analysis of thin structures, the application of shell elements offers multiple advantages over volumetric elements. An abstraction of a three-dimensional solid towards a two-dimensional surface often allows a greatly reduced problem size and simplified meshing. Due to the nature of the finite element method of constructing and solving large sparse matrix systems, it is inherently suited to parallel implementations on modern graphics processors. In this work, shell element standards relevant in engineering practice are examined. Based on this examination, the triangular MITC3 (Mixed Interpolation of Tensorial Components) shell element is chosen and implemented for GPU optimization. The GPU implementation presented includes a simplex-based assembly approach and a preconditioned conjugate gradient solver adapted to the MITC3 shell element. Results show that an accurate double-precision finite element analysis is possible using the proposed implementation. Furthermore, comparing CPU and GPU execution times, a speed-up factor of nine is demonstrated for a problem size of one million degrees of freedom.