Research Abstract |
This research is to present an acceleration method in the finite element analysis arising in fluid flow and structual engineerings. For this purpose, special computational machine has been originated, of which main processors are 8086/8087 microprocessors. In the finite element computations, the computations original to the method has to be carried out. To accelerate, the whole computation, it is necessary to present the algorithm to accelerate the original computation. In this research, multi-microprocessors are attached to the presonal computers. The computation with the microprocessors has been accelerated 5.7 times faster than that by the computation without microprocessors. Those results are almost the same as in the computations of fluid flow and structural analysis. In the fluid flow analysis, the shallow water flow equationis used. In the structual analysis, the linear elastic equationis solved by the conjugate gradient method. Contrary to this, the computation cannot be accelerated even though over four microprocessors. The reason seems as follows, i.e., the computational time of the real computation is shorter than the computational time to distribute the data to each microprocessore. To improve this defect, the attached type of the microprocessor cannot be replaced to allin one type. Using a large scale computer, the algorithm for the acceleration method of the finite element analysis has been conducted. The algorithm presented in this research is referred to a the mapping procedure, which performs the computation element by element. Those algorithm has been applied to the nonlinear analysis of the fluid flow, which include the velockity correction method, boundary type method, moving boundary problem, Lagrangian analysis, Arbitrary Lagrangina-Eulerian analysis, Adaptive finite element method, etc.
|