Parallel computing using semianalytical finite element method
Jeong Rae Kim, Wie Dae Kim, Seung Jo Kim · AIAA Journal · 1994
In this paper, a parallel computing technique using a semianalytical finite element method and its implementation method are studied, and the speedup in the elapsed computing time is discussed. The structural analysis of an axisymmetric body with nonaxisymmetric boundary conditions is considered by using a semianalytical finite element method. The displacements and the boundary conditions are expressed by a Fourier series. By the orthogonality properties of the Fourier series, the equilibrium equations are fully decoupled for each mode. Therefore, these decoupled equations can be well adapted to the individual processors of a parallel computer. Two example problems are considered for the verification of the developed code and for the application of the parallel processing to practical problems. The computing time by parallel processing is compared with that by sequential processing, and a large amount of improvement in the elapsed computing time is achieved by parallel processing. From the computational experiments, a speedup of 4.6 times in elapsed time is observed with six processors of Alliant FX2812 at Seoul National University. 14 refs.