A novel methodology of weighted residual for nonlinear computations

W. Chen

Published 1999-05-02Version 1

One of strengths in the finite element (FE) and Galerkin methods is their capability to apply weak formulations via integration by parts, which leads to elements matching at lower degree of continuity and relaxes requirements of choosing basis functions. However, when applied to nonlinear problems, the methods of this type require a great amount of computing effort of repeated numerical integration. It is well known that the method of weighted residual is the very basis of various popular numerical techniques including the FE and Galerkin methods. This paper presents a novel methodology of weighted residual for nonlinear computation with objectives to avoid the above-mentioned shortcomings. It is shown that the presented nonlinear formulations of the FE and Galerkin methods can be expressed in the Hadamard product form as in the collocation and finite difference methods. Therefore, the recently developed SJT product approach can be applied in the evaluation of the Jacobian matrix of these nonlinear formulations. This also provides possibility to introduce the nonlinear uncoupling technique to the FE and Galerkin nonlinear computing. Furthermore, the present scheme of weighted residuals also greatly eases the use of the least square and boundary element methods to nonlinear problems