Solve Poission equation using Conjugate gradient solver.
8 views (last 30 days)
Show older comments
I am trying to solve Poisson equation using Conjugate gradient solve.
But the numerical solution does not converge to exact solution....
close all;
N = 100; % points in the each direction
x0 = 0;
xN = 2;
y0 = 0;
yN = 1;
dx = (xN-x0)/N;
dy = (yN-y0)/N;
dx2_dy2 = dx^2 / dy^2;
x = linspace(x0,xN,N+1);
y = linspace(y0,yN,N+1);
[X Y] = meshgrid(x,y);
Q = 5 * pi^2 * sin(pi .*X) .* sin(2*pi.*Y);
Qint = Q(2:N,2:N); % obtain interior
q = Qint(:); % transform into a vector
% True solution
Pd = sin(pi .* X) .* sin(2*pi .*Y);
tol = 1e-5;
Pcg = zeros(N+1, N+1);
x = zeros((N-1)^2, 1);
x_old = x;
err = 1;
iter = 0;
TN = gallery('tridiag', N-1, -1, 2, -1);
A = kron(speye(N-1), TN) + kron(TN, eye(N-1));
A = A / dx^2;
b = reshape(Q(2:N, 2:N), [], 1);
r = b - A * x;
p = r;
tic;
while (err > tol)
iter = iter + 1;
Ap = A * p;
alpha = (r' * r) / (p' * Ap);
x_new = x + alpha * p;
r_new = r - alpha * Ap;
residual_norm = norm(r_new);
err = norm(x_new - x)
if err < tol
break;
end
beta = (r_new' * r_new) / (r' * r);
p = r_new + beta * p;
r = r_new;
x = x_new;
end
toc;
Pcg(2:N, 2:N) = reshape(x, N-1, N-1);
cg_iter = iter
% Filled Contour Plots
figure(1);
% Numerical Solution Pj
contourf(X, Y, Pcg, 20, 'LineStyle', 'none');
colorbar;
title('Numerical Solution (Pcg)');
xlabel('x');
ylabel('y');
% error Solution Pj - Pd
figure(2);
contourf(X, Y, Pcg - Pd, 20, 'LineStyle', 'none');
title('Error (Pcg)');
xlabel('x');
ylabel('y');
colorbar
1 Comment
Torsten
on 10 Dec 2024
Your matrix A is only suited for an equidistant grid in both directions (x and y).
In your case, it has to be modified because dx = 2*dy.
Answers (0)
See Also
Categories
Find more on Eigenvalue Problems in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
