Performance of the Finite Strip Method for Structural Analysis on a Parallel Computer

. In this paper, we address the efficiency and parallelizability of the finite strip method for the analysis of certain structural problems. This method, first developed in the context of thin plate bending analysis, is a semi-analytical finite element process. It takes advantage of orthogonal properties of harmonic functions in the stiffness matrix formulation to yield a block diagonal stiffness matrix. The method not only simplifies the data manipulation in forming the matrix but also reduces the bandwidth of the assembled matrix, resulting in great savings in the solution process. Furthermore, the block diagonal structure of the matrix offers obvious advantages for solving the problem on parallel computers. Numerical experiments and the performance of this method on an Alliant FX/8 minisupercomputer are reported. Some parallel implementations of this method on the University of Illinois Cedar multiprocessor are presented. 1. Introduction. Recent developments in supercomputing techno..