PROGRAMMING OF FINITE ELEMENT METHODS （有限元方法的编程）.pdf

PROGRAMMING OF FINITE ELEMENT METHODS IN MATLAB LONG CHEN We shall discuss how to implement the linear finite element method for solving the Pois- son equation. We begin with the data structure to represent the triangulation and boundary conditions, introduce the sparse matrix, and then discuss the assembling process. Since we use MATLAB as the programming language, we pay attention to an efficient programming style using sparse matrices in MATLAB. 1. DATA STRUCTURE OF TRIANGULATION We shall discuss the data structure to represent triangulations and boundary conditions. 1.1. Mesh data structure. The matrices node(1:N,1:d) and elem(1:NT,1:d+1) are used to represent a -dimensional triangulation embedded in , where N is the number of vertices and NT is the number of elements. These two matrices represent two different structure of a triangulation: elem for the topology and node for the geometric embedding. The matrix elem represents a set of abstract simplices. The index set is called the global index set of vertices. Here an vertex is thought as an abstract entity. For a simplex , is the local index set of . The matrix elem is the mapping (pointer) from the local index to the global one, i.e., elem(t,1:d+1) records the global indices of vertices which form the abstract -simplex t. Note that any permutation of vertices of a simplex will represent the same abstract simplex. The matrix node gives the geometric realization of the simplicial complex. For ex- ample, for a 2-D triangulation, node(k,1:2) contain - and -coordinates of the -th nodes. The geometric realization introduces an ordering of the simplex. For each elem(t,:), we shall always order the vertices of a simplex such that the signed area is positive. That is in 2-D, three vertices of a triangle is ordered counter-clockwise and in 3-D, the