Reliability Computing of Stochastic BEM Based on Grid Platform
Lei Zhang, Yi Liu, Jianfei Zhang · 2011
Stochastic BEM provides a powerful tool for the reliability computing of complex or large scale structure. Though simple and applicable, AFOSM Method has shortcomings when distribution forms of random variable are concerned. In some cases, the distribution types of variables can be worked out through statistics analysis. Thus it is necessary to adopt appropriate methods to work out possible impacts of variable distribution. Stochastic BEM, based on Monte-Carlo numerical simulation, is capable to simulate such impacts effectively as a method of statistical approximation. But the coefficient matrix is asymmetric full-matrix in the solution of BEM matrix equation, the computational complexity increase by N3 with the increase of boundary node number. On solving the complex large-scale three dimensions structure problem, the large computational scale has been a barrier in the application of boundary element method, especially for the Stochastic BEM computation. Grid computing is a new style of distributed computing technology after the cluster computing. Grid is a integrated environment of computing and resource, which realize sharing resource and cooperative teamwork .It can absorb various computing resources and translate the min to convenient and reliable computer power. On the campus grid platform, the single or multiple large-scale BEM computing could carry out expediently and efficiently. This paper took the case of a campus grid in Hohai university, and introduced the characteristic of grid computing and the three-layer frame of the campus grid such as resource layer, grid layer and application service layer, and the campus grid platform is a system based on two-level scheduler, respectively grid layer scheduler and resource layer scheduler, thus a two-level parallel strategy in the above two layers can be adopted for the computing of stochastic BEM which has two parallel specialties in both BEM computing and Monte-Carlo method. Then the realization of stochastic BEM based on grid platform was put forward. Finally, there is a test which shows the high performance and efficiency of the two-level parallel strategy for the stochastic BEM computing on the grid platform.