ICF10B

Where are multinomial expression including space coordinates j p ( ) x y , , are coefficients undefined. j c ( )    x y x y p p p T 1 , 1 2 3 = = p   (4-a)  1 2 3 ( , ) x y c c c T = c (4-b) In the MLSM, the coefficients c are obtained by minimizing the following weighted square expression . j J ( ) ( ) [ ] 2 2 0 ) ( i i i i i n i i w u u u J w = ∆ − − =∑ x x x x (5) ) ( i iw w x x = − (6-a) ( ) ( ) ( , ) ( , ) ( , ) 0 i i i j i i ji i i i i p x y c x y u x y u u u − = − ∆ = x x (6-b) Where  i i i x y = x i x x = is the coordinate of a node i in the neighborhood of evaluation point and is the nodal value at . is the number of nodes in the neighborhood of x . x iu n ) i x x− (w is a weight function defined in the neighborhood of x . We used four-dimensional spline function in Eqn. 7 as ) ( i x− w . x i x x− is norm of vector i x x− and stands for the distance from the evaluation point to node . x i x ) 3.0 (0 8.0 ( ) 1.0 6.0 4 3 2 ρ ρ ρ ρ  ≤ ≤       −      +         = − w w w w w r r r r w r (7) Where i wr x x = − and ρ is a radius of circular domain of influence in two dimensions. The stationary of in Eqn. 5 with respect to leads to the following equation. J )x(c ( ) B u A c x y x y x y , ( , ) ( , ) = (8-a) or ( ) B u A c x y x y x y , ( , ) ( , ) 1− = (8-b) Where A , B and u in Eqn. 8 are defined by following equations. [ ] k ik jk ij ij A A w p p = = , A j , 1 i k n ~ ~ 3 1= = (9-a) [ ] i ji ij ij B B wP = = , B ,i j 1 3~= (9-b) u (9-c)  n T u u u ⋅⋅⋅ = 1 2  Eqn. 8-a must be solved accurately to retain the accuracy of the MLSM interpolant. When the matrix is not well conditioned, Eqn. 8-b cannot be solved with desired accuracy. However the necessity for solving Eqn. 8-a can be eliminated by diagonalizing the matrix . To diagonalize the matrix A , we obtained orthogonal basis function q by using Schmidt orthogonalization procedure. The coefficients c can be expressed by using Eqn. 8-a with . Thus, by substituting the coefficients c for Eqn. 3, the MLSM interpolant approximation can be expressed as follows: A A ( ) x x, i j ( ) x x, iq j ( ) ( ) ( ) ( ) i j k j i k k i u q w q q x y w x x x x x x x x x x , ) ( , , ) ( , 2 0 − = − u (10) By using a method developed by T. Belytschko et al, we divide the analytical domain into regular and

RkJQdWJsaXNoZXIy MjM0NDE=