Parallel Evolution Finite Element Method on PC Cluster
A parallel evolution FEM algorithm based on PC cluster was designed by combining Genetic Algorithms (GAs) and domain decomposed parallel finite element method. This algorithms flow was described as follow: firstly GAs generates a random population, then the parallel FEM exerts its virtues huge computing capacity and fast computing speed, after the parallel FEM positive analysis, the master PC computes the fitness which was the square sum of the difference between FEM analysis and in-site monitor of the key points displacement, then the GAs generates a group of new parameters again by the evolution operations, the operations would done until the best parameters have been founded. This algorithm conquers the long time consuming disadvantage of traditional back analysis algorithm. Based on this algorithm the paper developed a parallel computing program GAPFEM3D using C++ and MPI library on PC cluster. GAPFEM3D exerts both GAs attribute: large search space, quickly converge and parallel FEMs attribute: largescale capability, high efficiency. Finally it was applied to a large-scale geoengineering successfully. The back analysis results validate the robust and efficiency of this algorithm.
PC Cluster Domain Decomposed Method Genetic Algorithms
Zhongliang Ru Hongbo Zhao Mingtian Li
College of Civil Engineering, Henan Polytechnic University Jiaozuo, Henan 454000, China Department of Civil engineering, Shandong Jiaotong University Jinan, Shandong 250023, China
国际会议
杭州
英文
77-81
2006-10-12（万方平台首次上网日期，不代表论文的发表时间）