Department of Engineering Sciences
Numerical Solution for PDEs
Assignment
MATLAB Programming for Numerical Solution for PDEs
Name Reg No
Maaz Khan CV-2463
Submitted To:
Dr. Muhammad Siraj
Programming for FTCS Scheme
n +1
Um = λU nm −1 + (1 - 2 λ)U nm + λU nm +1
a=0;
b=2;
L=b-a;
ti=0;
tf=0.1;
m=4;
T=tf-ti;
n=2;
h=(b-a)/m;
k=T/n;
alpha=1;
lambda=alpha*k/h^2;
% Check stability condition
if lambda > 0.5
warning('The explicit scheme is unstable (lambda = %.2f > 0.5). Reduce time step k for
stability.', lambda);
end
Nx=L/h+1;
Nt=T/k+1;
x=linspace(a,b,Nx);
t=linspace(ti,tf,Nt);
U=zeros(Nt,Nx);
% Bounday Conditions
U(:,1)=0;
U(:,end)=0;
% Initial Condition
U(1,:)= sin((pi/2)*x);
%FTCS
for n=1:Nt-1
for m=2:Nx-1
U(n+1,m)=lambda*(U(n,m+1)+U(n,m-1))+(1-2*lambda)*(U(n,m));
end
end
U_exact=zeros(Nt,Nx);
for n=1:Nt
U_exact(n,:)=exp(-((pi^2)/4)*t(n))*sin((pi/2)*x);
end
error=abs(U_exact-U);
max_error=max(error(:));
mean_error=mean(error(:));
% Line graph
figure;
plot(x, U(end,:), 'b-', 'DisplayName', 'FTCS Numerical');
hold on;
plot(x, U_exact(end, :), 'o-', 'DisplayName', 'Exact Solution');
xlabel('x');
ylabel('U(x, t)');
title(sprintf('Heat Equation Solution at t=%.2f', T));
legend({'FTCS Numerical', 'Exact Solution'}, 'Location', 'northeast');
grid on;
hold off;
% Print errors
fprintf('Maximum Error: %.6f\n', max_error);
fprintf('Average Error: %.6f\n', mean_error);
Exact Solution
Approximate Solution
Error
Maximum Error: 0.001932
Average Error: 0.000487
Plot
MATLAB Programming for BTCS Method
n−1
Um = -λU nm +1 + (1 + 2 λ)U nm + λU nm −1
a=0;
b=2;
L=b-a;
ti=0;
tf=0.1;
T=tf-ti;
alpha=1;
m=4;
N=2;
k=T/N;
h=L/m;
lambda=alpha*k/h^2;
Nt=round(T/k)+1;
Nx=round(L/h)+1;
x=linspace(a,b,Nx);
t=linspace(ti,tf,Nt);
U=zeros(Nt,Nx);
U(:,1)=0;
U(:,end)=0;
U(1,:)=sin((pi/2)*x);
%BTCS
A=diag((1+2*lambda)*ones(Nx-2,1))+...
diag(-lambda*ones(Nx-3,1),1)+...
diag(-lambda*ones(Nx-3,1),-1);
for n=1:Nt-1
B=U(n,2:Nx-1)';
W=A\B;
U(n+1,2:Nx-1)=W;
end
U_exact=zeros(Nt,Nx);
for n=1:Nt
U_exact(n,:)=exp(-((pi^2)/4)*t(n))*sin((pi/2)*x);
end
error=abs(U_exact-U);
max_error=max(error(:));
mean_error=mean(error(:));
% Line graph
figure;
plot(x, U(end,:), 'g-', 'DisplayName', 'BTCS Numerical');
hold on;
plot(x, U_exact(end, :), 'y-', 'DisplayName', 'Exact Solution');
xlabel('x');
ylabel('U(x, t)');
title(sprintf('Heat Equation Solution at t=%.2f', T));
legend({'BTCS Numerical', 'Exact Solution'}, 'Location', 'northeast');
grid on;
hold off;
% Print errors
fprintf('Maximum Error: %.6f\n', max_error);
fprintf('Average Error: %.6f\n', mean_error);
Exact Solution
Approximate Solution
Plot
Error
Maximum Error: 0.019912
Average Error: 0.005006
Crank Nicolson Method
n
−λ n+1 λ n+1 λ λ n
U m−1 + (1 + λ)U nm+1 - U m+1 = U + (1 - λ)U nm + U m+1
2 2 2 m−1 2
a=0;
b=1;
L=b-a;
ti=0;
tf=0.4;
T=tf-ti;
h=0.25;
k=0.1;
alpha=1/16;
lambda=alpha*k/h^2;
Nx=round(L/h)+1;
Nt=round(T/k)+1;
x=linspace(a,b,Nx);
t=linspace(ti,tf,Nt);
U=zeros(Nt,Nx);
U(:,1)=0;
U(:,end)=0;
U(1,:)=sin(pi*x);
% Constant Coefficient Matrix A
A=diag(2*(1+lambda)*ones(Nx-2,1))+...
diag(-lambda*ones(Nx-3,1),1)+...
diag(-lambda*ones(Nx-3,1),-1);
% Constant Coefficient Matrix B
B=diag(2*(1-lambda)*ones(Nx-2,1))+...
diag(lambda*ones(Nx-3,1),1)+...
diag(lambda*ones(Nx-3,1),-1);
% Time-stepping Loop using Crank-Nicolson
for n=1:Nt-1
% Right Hand Side of Scheme B*U(n,2:Nx-1)
Y=B*U(n,2:Nx-1)';
% Solve System: A * U(n+1, 2:Nx-1)' = Y
U(n+1,2:Nx-1)=A\Y;
end
U_exact=zeros(Nt,Nx);
for n=1:Nt
U_exact(n,:) = sin(pi * x) * exp(- (pi^2 / 16) * t(n));
end
error=abs(U_exact-U);
max_error=max(error(:));
mean_error=mean(error(:));
% Line graph
figure;
plot(x, U(end,:), 'black-', 'DisplayName', 'Crank Nicolson Numerical Method');
hold on;
plot(x, U_exact(end, :), 'o-', 'DisplayName', 'Exact Solution');
xlabel('x');
ylabel('U(x, t)');
title(sprintf('Heat Equation Solution at t=%.2f', T));
legend({'Crank Nicolson Method', 'Exact Solution'}, 'Location', 'northeast');
grid on;
hold off;
% Print errors
fprintf('Maximum Error: %.6f\n', max_error);
fprintf('Average Error: %.6f\n', mean_error);
Exact Solution
Approximate Solution
Error
Maximum Error: 0.009557
Average Error: 0.002406
Plot
Richardson Method
n
n +1 λ
Um = U + 2λ[U nm −1- 2U nm + U nm +1]
2 m−1
a=0;
b=2;
L=b-a;
ti=0;
tf=0.1;
T=tf-ti;
m=4;
N=2;
h=L/m;
k=T/N;
alpha=1;
lambda=alpha*k/h^2;
Nx=round(L/h)+1;
Nt=round(T/k)+1;
x=linspace(a,b,Nx);
t=linspace(ti,tf,Nt);
U=zeros(Nt,Nx);
U(:,1)=0;
U(:,end)=0;
U(1,:)=sin(pi*x/2);
% Compute U_j^{-1} using backward extrapolation
U_minus1=zeros(1,Nx);
for m=2:Nx-1
U_minus1(m)=U(1,m)-lambda*(U(1,m+1)-2*U(1,m)+U(1,m-1));
end
% Apply Richardson Scheme
for n=1:Nt-1
for m=2:Nx-1
U(n+1,m)=U_minus1(m)+2*lambda*(U(n,m+1)-2*(U(n,m))+U(n,m-1));
end
U_minus1=U(n,:);
end
U_exact=zeros(Nt,Nx);
for n=1:Nt
U_exact(n,:)=exp(-((pi^2)/4)*t(n))*sin((pi/2)*x);
end
error=abs(U_exact-U);
max_error=max(error(:));
mean_error=mean(error(:));
% Line graph
figure;
plot(x, U(end,:), 'g-', 'DisplayName', 'Richardson Numerical Method');
hold on;
plot(x, U_exact(end, :), 'o-', 'DisplayName', 'Exact Solution');
xlabel('x');
ylabel('U(x, t)');
title(sprintf('Heat Equation Solution at t=%.2f', T));
legend({'Richardson Numerical Method', 'Exact Solution'}, 'Location', 'northeast');
grid on;
hold off;
% Print errors
fprintf('Maximum Error: %.6f\n', max_error);
fprintf('Average Error: %.6f\n', mean_error);
Exact Solution
Approximate Solution
Error:
Maximum Error: 0.011793
Average Error: 0.002074
Plot:
Comparison of Numerical Methods
Error (Max /
Method Stability Accuracy Complexity Remarks
Avg)
Easiest to implement but can
Conditionally stable (λ ≤ Low to Simple, explicit 0.001932 / become unstable and
FTCS
0.5) Moderate scheme 0.000487 inaccurate for large time
steps.
Implicit scheme; Stable for any timetable, but
Moderate to 0.019912 /
BTCS Unconditionally stable requires solving computationally heavier than
High 0.005006
linear systems FTCS.
Best balance between
High (second Semi-implicit;
Crank- 0.009557 / stability and accuracy; very
Unconditionally stable order in time moderate
Nicolson 0.002406 suitable for time-dependent
and space) complexity
PDEs.
Conditionally stable More accurate than FTCS,
Explicit, but more 0.011793 /
Richardson (depends on Moderate but sensitive to initialization
error-prone 0.002074
implementation) and parameter choice.
