调用fL，fR等函数的多级欧米茄方案。
调用fL，fR等函数的多级欧米茄方案。
裴来凡
发布于 2022-05-28 15:31:52
发布于 2022-05-28 15:31:52
stability_domain.m
omega = 1 - sqrt(0.5); % Parameters in multistage omega scheme
alpha = (1 - 2*omega)/(1 - omega);
alphap = 1 - alpha;
omegap = 1 - 2*omega;
x = -0.03:0.01:0.5; y = 0:0.5:9; [X,Y] = meshgrid(x,y); Z = X + i*Y;
% Absolute value of g(z)
One = ones(size(Z));
F = abs(((One + alphap*omega*Z).^2).*(One + alpha*omegap*Z)...
.*((One - alpha*omega*Z).^(-2)).*((One - alphap*omegap*Z).^(-1)));
figure(1), clf, hold on
v = [ 0.9 0.95 1.0 ]; c = contour(x,y,F,v); clabel(c,'fontsize',18);
line([0 0],[0 9]) % Draw imaginary axis
title('Contour plot of |g(z|','fontsize',18)
% Value of abs[g(z)] on imaginary axis
z = i*3; ff = abs(((1 + alphap*omega*z)^2)*(1 + alpha*omegap*z)...
*((1 - alpha*omega*z)^(-2))*((1 - alphap*omegap*z)^(-1)))
exact_solution.m
function ye = exact_solution(t,x,D,nalpha,nbeta,u,h)
alpha = nalpha*pi; beta = nbeta*pi;
ye = h*cos(beta*(x - u*t)) + exp(-D*alpha*alpha*t)*cos(alpha*(x - u*t));
fL.m
function ye = fL(t,D,nalpha,nbeta,u,h) % Left boundary value
alpha = nalpha*pi; beta = nbeta*pi;
ye = h*cos(beta*u*t) + exp(-D*alpha*alpha*t)*cos(alpha*u*t);
fR.m
function ye = fR(t,D,nalpha,nbeta,u,bc,h) % Right boundary value
alpha = nalpha*pi; beta = nbeta*pi;
if bc == 1
ye = - h*beta*sin(beta*(1 - u*t))...
- alpha*exp(-D*alpha*alpha*t)*sin(alpha*(1 - u*t));
else
ye = 0.0;
end
multistage_omega_scheme.m
u = 1.1; % Velocity
D = 0.0005; % Diffusion coefficient
dt = 1/10; % Time step
n = 100; % Number of time steps
J = 30; % Number of grid cells
central = 1; % Enter 1 for central scheme or something else for upwind scheme
bc = 2; % Enter 1 for inhomogeneous Neumann or 2 for homogeneous Neumann
hom = 1; % Enter 1 for homogeneous right-hand side
% or 2 for inhomogeneous right-hand side
defcor = 1; % Enter 1 for defect correction or else something else
nalpha = 3; % Parameter in exact solution: alpha = nalpha*pi
nbeta = 2; % Parameter in exact solution: beta = nbeta*pi
%...............End of input...............................
omega = 1 - sqrt(0.5); % Parameters in multistage omega scheme
palpha = (1 - 2*omega)/(1 - omega);
palphap = 1 - palpha;
omegap = 1 - 2*omega;
T = n*dt; % Final time
% Definition of indexing
% x=0
% grid |---o---|---o---|---o---|---o---|
% 1 1 2 2 3 J
h = 1.0/J; % Cell size
d = D*dt/(h*h); % Diffusion number
sigma = u*dt/h; % Courant number
meshpeclet = sigma/d; % Mesh Peclet number
% Preallocation of J*J tridiagonal matrix A
Ac = spdiags([ ones(J,1) ones(J,1) ones(J,1)], [-1 0 1]',J,J); % Central
Au = spdiags([ ones(J,1) ones(J,1) ones(J,1)], [-1 0 1]',J,J); % Upwind
% Central scheme
Ac(1,1) = sigma/2 + 3*d; Ac(1,2) = sigma/2 - d;
for j = 2:J-1,
Ac(j,j-1) = -sigma/2 - d; Ac(j,j+1) = sigma/2 - d;
Ac(j,j) = - Ac(j,j-1) - Ac(j,j+1);
end
j = J; Ac(j,j-1) = -sigma/2 - d; Ac(j,j) = sigma/2 + d;
% Upwind scheme
Au(1,1) = sigma + 3*d; Au(1,2) = - d;
for j = 2:J-1,
Au(j,j-1) = -sigma - d; Au(j,j+1) = - d;
Au(j,j) = - Au(j,j-1) - Au(j,j+1);
end
j = J; Au(j,j-1) = -sigma - d; Au(j,j) = sigma + d;
% System to be solved is E*ynew = C*yold + rhs
z = zeros(J,1); e = ones(J,1); ID = spdiags([z e z],[-1 0 1],J,J);
Ec1 = ID + palpha*omega*Ac; Cc1 = ID - palphap*omega*Ac;
Ec2 = ID + palphap*omegap*Ac; Cc2 = ID - palpha*omegap*Ac;
Eu1 = ID + palpha*omega*Au; Cu1 = ID - palphap*omega*Au;
Eu2 = ID + palphap*omegap*Au; Cu2 = ID - palpha*omegap*Au;
if defcor ~= 1 % No defect correction
if central == 1 % Central scheme
E1 = Ec1; C1 = Cc1; E2 = Ec2; C2 = Cc2;
else % Upwind scheme
E1 = Eu1; C1 = Cu1; E2 = Eu2; C2 = Cu2;
end
end
% Preallocations
yold = zeros(J,1); % Numerical solution at previous time
ynew = zeros(J,1); % Numerical solution at new time
x = [1:J]'*h - h/2; % Coordinates of cell centers
%for j = 1:J, x(j)=(j-0.5)*h; end
t = 0; yold = exact_solution(t,x,D,nalpha,nbeta,u,hom-1);
rhs = zeros(J,1); % Right-hand side
beta = nbeta*pi;
if defcor ~= 1 % No defect correction
for i = 1:n
% Stage 1
rhs = (hom-1)*beta*beta*D*dt*omega*cos(beta*(x-u*t));
rhs(1) = rhs(1) + palphap*omega*(sigma + 2*d)*fL(t,D,nalpha,nbeta,u,hom-1)...
+ palpha*omega*(sigma + 2*d)*fL(t+omega*dt,D,nalpha,nbeta,u,hom-1);
if central == 1 % Central scheme
rhs(J) = rhs(J) -...
palphap*omega*(sigma/2 - d)*h*fR(t,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*(sigma/2 - d)*fR(t+omega*dt,D,nalpha,nbeta,u,bc,hom-1);
else % Upwind scheme
rhs(J) = rhs(J) - palphap*omega*( - d)*h*fR(t,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*( - d)*fR(t+ omega*dt,D,nalpha,nbeta,u,bc,hom-1);
end
rhs = rhs + C1*yold; ynew = E1\rhs;
% Stage 2
yold = ynew;
rhs = (hom-1)*beta*beta*D*dt*omegap*cos(beta*(x-u*(t + dt - omega*dt)));
rhs(1) = rhs(1) + ...
palpha*omegap*(sigma + 2*d)*fL(t + omega*dt,D,nalpha,nbeta,u,hom-1)...
+ palphap*omegap*(sigma + 2*d)*fL(t + dt - omega*dt,D,nalpha,nbeta,u,hom-1);
if central == 1 % Central scheme
rhs(J) = rhs(J) -...
palpha*omegap*(sigma/2 - d)*h*fR(t + omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palphap*omegap*(sigma/2 - d)*fR(t+dt-omega*dt,D,nalpha,nbeta,u,bc,hom-1);
else % Upwind scheme
rhs(J) = rhs(J) -...
palpha*omegap*( - d)*h*fR(t + omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palphap*omegap*( - d)*fR(t + dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1);
end
rhs = rhs + C2*yold; ynew = E2\rhs;
% Stage 3
yold = ynew;
rhs = (hom-1)*beta*beta*D*dt*omega*cos(beta*(x - u*(t + dt)));
rhs(1) = rhs(1) +...
palphap*omega*(sigma + 2*d)*fL(t+ dt- omega*dt,D,nalpha,nbeta,u,hom-1)...
+ palpha*omega*(sigma + 2*d)*fL(t+dt,D,nalpha,nbeta,u,hom-1);
if central == 1 % Central scheme
rhs(J) = rhs(J) -...
palphap*omega*(sigma/2-d)*h*fR(t+ dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*(sigma/2 - d)*fR(t+dt,D,nalpha,nbeta,u,bc,hom-1);
else % Upwind scheme
rhs(J) = rhs(J) -...
palphap*omega*( - d)*h*fR(t+ dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*( - d)*fR(t+ dt,D,nalpha,nbeta,u,bc,hom-1);
end
rhs = rhs + C1*yold; ynew = E1\rhs;
t = t + dt; yold = ynew;
end
else % One step of defect correction
dy = zeros(J,1); % Correction
drhs = zeros(J,1); % Correction of right-hand side
for i = 1:n
% Stage 1: upwind step
rhs = (hom-1)*beta*beta*D*dt*omega*cos(beta*(x-u*t));
rhs(1) = rhs(1) + palphap*omega*(sigma + 2*d)*fL(t,D,nalpha,nbeta,u,hom-1)...
+ palpha*omega*(sigma + 2*d)*fL(t+omega*dt,D,nalpha,nbeta,u,hom-1);
drhs = rhs; % Preparation of rhs for correction step
drhs(J) = drhs(J) -...
palphap*omega*(sigma/2 - d)*h*fR(t,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*(sigma/2 - d)*fR(t+omega*dt,D,nalpha,nbeta,u,bc,hom-1);
rhs(J) = rhs(J) - palphap*omega*( - d)*h*fR(t,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*( - d)*fR(t+ omega*dt,D,nalpha,nbeta,u,bc,hom-1);
rhs = rhs + Cu1*yold; ynew = Eu1\rhs;
% Stage 1: correction step
drhs = drhs - Ec1*ynew + Cc1*yold; dy = Eu1\drhs; ynew = ynew + dy;
% Stage 2: upwind step
yold = ynew;
rhs = (hom-1)*beta*beta*D*dt*omegap*cos(beta*(x-u*(t + dt - omega*dt)));
rhs(1) = rhs(1) + ...
palpha*omegap*(sigma + 2*d)*fL(t + omega*dt,D,nalpha,nbeta,u,hom-1)...
+ palphap*omegap*(sigma + 2*d)*fL(t + dt - omega*dt,D,nalpha,nbeta,u,hom-1);
drhs = rhs; % Preparation of rhs for correction step
drhs(J) = drhs(J) -...
palpha*omegap*(sigma/2 - d)*h*fR(t + omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palphap*omegap*(sigma/2 - d)*fR(t+dt-omega*dt,D,nalpha,nbeta,u,bc,hom-1);
rhs(J) = rhs(J) -...
palpha*omegap*( - d)*h*fR(t + omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palphap*omegap*( - d)*fR(t + dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1);
rhs = rhs + Cu2*yold; ynew = Eu2\rhs;
% Stage 2: correction step
drhs = drhs - Ec2*ynew + Cc2*yold; dy = Eu2\drhs; ynew = ynew + dy;
% Stage 3: upwind step
yold = ynew;
rhs = (hom-1)*beta*beta*D*dt*omega*cos(beta*(x - u*(t + dt)));
rhs(1) = rhs(1) +...
palphap*omega*(sigma + 2*d)*fL(t+ dt- omega*dt,D,nalpha,nbeta,u,hom-1)...
+ palpha*omega*(sigma + 2*d)*fL(t+dt,D,nalpha,nbeta,u,hom-1);
drhs = rhs; % Preparation of rhs for correction step
drhs(J) =drhs(J) -...
palphap*omega*(sigma/2-d)*h*fR(t+ dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*(sigma/2 - d)*fR(t+dt,D,nalpha,nbeta,u,bc,hom-1);
rhs(J) = rhs(J) -...
palphap*omega*( - d)*h*fR(t+ dt- omega*dt,D,nalpha,nbeta,u,bc,hom-1)...
- palpha*omega*( - d)*fR(t+ dt,D,nalpha,nbeta,u,bc,hom-1);
rhs = rhs + Cu1*yold;
ynew = Eu1\rhs;
% Stage 3: correction step
drhs = drhs - Ec1*ynew + Cc1*yold; dy = Eu1\drhs; ynew = ynew + dy;
t = t + dt; yold = ynew;
end
end
solex = exact_solution(t,x,D,nalpha,nbeta,u,hom-1);
error = ynew - solex; % Compute error norms
norm1 = norm(error,1)/J
norm2 = norm(error,2)/sqrt(J)
norminf = norm(error,inf)
figure(1), clf, hold on
xx = 0:0.02:1; plot (xx,exact_solution(t,xx,D,nalpha,nbeta,u,hom-1),'-');
axis([0 1 -1 1]); plot (x,ynew,'o');
if central == 1, s1 = ['Central fractional step omega-scheme'];
else, s1 = ['Upwind fractional step omega-scheme'];
end
if defcor == 1
s1 = ['Fractional step omega-scheme with defect correction'];
end
if bc == 1, s1 = [s1, ', inhomogeneous Neumann'];
else s1 = [s1, ', homogeneous Neumann'];
end
title(s1,'fontsize',12);
if hom ==1, s2 = ['Homogeneous differential equation']
else, s2 = ['Inhomogeneous differential equation']
end
s3 = ['omega=',num2str(omega),' D=',num2str(D),' u=',num2str(u),...
' h=',num2str(h),' dt=',num2str(dt),' t=',num2str(T),' n=',num2str(n)]
s4 = ['meshpeclet=',num2str(meshpeclet),' Courant=',num2str(sigma),...
' d=',num2str(2*d) ]
本文参与 腾讯云自媒体分享计划，分享自微信公众号。
原始发表：2020-05-24，如有侵权请联系 cloudcommunity@tencent.com 删除
评论
登录后参与评论
推荐阅读