Scribd
Upload
88 views

LU SIP Procedure

The document describes the Strongly Implicit Procedure (SIP) for solving partial differential equations. SIP replaces the sparse matrix A with a modified matrix M that can be decomposed into upper and lower triangular matrices L and U. The elements of L and U are determined by: 1) Approximating values at neighboring nodes in terms of central nodes 2) Solving the resulting equations sequentially from the southwest corner 3) Multiplying the residual vector by L^-1 to solve for intermediate vector R 4) Marching through R to solve for the update vector δn in reverse order

Uploaded by

kulov1592
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
88 views

LU SIP Procedure

The document describes the Strongly Implicit Procedure (SIP) for solving partial differential equations. SIP replaces the sparse matrix A with a modified matrix M that can be decomposed into upper and lower triangular matrices L and U. The elements of L and U are determined by: 1) Approximating values at neighboring nodes in terms of central nodes 2) Solving the resulting equations sequentially from the southwest corner 3) Multiplying the residual vector by L^-1 to solve for intermediate vector R 4) Marching through R to solve for the update vector δn in reverse order

Uploaded by

kulov1592
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

The algebraic equation for one CV is given by

l l l l l l AN N + AE E + AS S + AW W + AP P = qP

where P denotes the node at which the PDE is approximated and index l runs over the neighbor nodes involved in finite difference approximations. The node P and its neighbors form the so-called computational molecule. The system can be written in matrix notation as follows:

[ A] = Q

The object of strongly implicit procedure (SIP) proposed by Stone (1968) is to replace the sparse matrix [A] by a modified form [A+N] such that the modified matrix can be decomposed into upper and lower triangular sparse matrices denote by [U] and [L], respectively. HW3. Please determine all of the elements of [L] and [U] used in the Strongly Implicit Procedure (SIP). Hints: 1. [ M ] = [ L ][U ] = [ A] + [ N ]
= 1 1 1 1 1
l UN l UE

l MW

l M NW

l MS

l MP

l MN

l M SE

l ME

l W

l S

l P

1 1 1 1

l M W = LlW

l M NW = LlWU N

lN j

M Sl = LlS
l M P = LlWU E lN j l + LlSU N1 + LlP

l l M N = U N LlP l l M SE = LlSU E1 l l M E = U E LlP


* * 2. NW and SE can be approximated in terms of the values of at nodes corresponding to

the diagonals of [A].


* NW (W + N P ) * SE ( S + E P ) (<1)

3. The resulting equations are not only sufficient to determine all of the elements of [L] and [U], but they can be solved in sequential order beginning at the southwest corner of the grid:
l LlW = AW / (1 + U N lN j

)
l + LlSU E1 ) LlWU E lN j l LlSU N1

l LlS = ASl / (1 + U E1 ) l LlP = AP + ( LlWU N lN j

l l U N = ( AN LlWU N

lN j

)/ L

l P

l l l U E = ( AE LlSU E1 ) / LlP

The coefficients must be calculated in this order. For nodes next to boundaries, any matrix element that carries the index of a boundary node is understood to be zero. 4. We now turn to solving the system of equations with the aid of this approximate factorization. The equation relating the update to the residual is :

[ M ] n = [ L][U ] n = n

LlW

LlS

LlP

1 1 1 1 1
l UN 1 l UE

1 1 1

l N j l N j l 1 l 1 l l = l +1 l +1 lN lN j j 1

The equations are solved as in generic LU decomposition. Multiplication of the above equation by [L]-1 leads to:

[U ] n = [ L]1 n = [ L]1 {Q - [ A] n } = R n
RlN j lN j R l 1 l 1 l l R = R l +1 l +1 lN lN R j j

LlW

LlS

LlP

Rn is easily computed:
R l = ( l LlS R l 1 LlW R
l N j

)/ L

l P

This equation is to be solved by marching in the order of increasing l. When the computation of R is complete, we need to solve above equation:
1 l N j R l N j l 1 R l 1 l l = R l +1 R l +1 lN lN j R j 1

1 1 1 1 1
l UN 1 l UE

1 1 1

l l l = R l U N l +1 U E l + N

in order of decreasing index l. Then n+1= n+n

You might also like