I am trying to solve the 2D time dependent heat equation using finite difference method in Matlab. The code is below:
%Spatial variable on x direction
Lx=1;
delta=0.1;
xmin=-Lx/2;
xmax=Lx/2;
Nx=(xmax-xmin)/delta;
x=linspace(xmin,xmax,Nx);
%Spatial variable on y direction
Ly=1;
delta=0.1;
ymin=-Ly/2;
ymax=Ly/2;
Ny=(ymax-ymin)/delta;
y=linspace(ymin,ymax,Ny);
%Total matrix size
N = (Nx * Ny);
%Time variable
dt=0.002;
tmin=0;
tmax=1;
nt=(tmax-tmin)/dt;
tspan=linspace(tmin,tmax,nt);
%Create a meshgrid
[X,Y] = meshgrid(x,y);
% Defining initial state:
T0=exp(-(X.^2+Y.^2));
%reshape the initial condition to a vector
T_reshape = reshape(T0,N,1);
% Constructing the 1D spatial matrix
A=zeros(N,N);
I = eye(N);
%the diagonal elements
for m=1:N %the number of rows
for n=1:N %the number of columns
if (m==n)
A(m,n)=-2/delta^2;
end
%Boundary conditions: A(1,N)==A(N,1)
if(n==N)&&(m==1)
A(m,n)=1;
end
if(n==1)&&(m==N)
A(m,n)=1;
end
end
end
%the off-diagonal elements
for n=1:N-1
A(n+1,n)=1/delta^2; %the value of each lower off-diagonal elements
end
for n=2:N
A(n-1,n)=1/delta^2; %the value of each upper off-diagonal element
end
%create the 2D matrix
B = kron(A,I)+kron(I,A);
% Solve the equation
[Time,Tem]=ode45('dTDistribution',tspan,T_reshape,[],B,delta);
The function that is being called here is as following:
%Define the function
function dT=dTDistribution(tspan,T_reshape,dummy,B,delta)
dT = B.*T_reshape;
end
My problem is that the dimension of my matrix B is different than the dimensions of the initial condition T_reshape, therefore, the multiplication of B.*T_reshape won't be possible. I'm wondering how can I change the dimension of T_reshape to make the multiplication valid. Hope anyone could help.
Thank you.
Thank you for looking at my problem, but I have figured out the mistake in the code. Since A is the 1D matrix, then its size should be either (Nx,Nx) or (Ny,Ny) and then when taking the tensor product to get B the 2D matrix its size will be (N,N). However in the code, A has the size of (N,N) and as a result B is blowing up making the multiplication afterwards not possible.
Thanks.
Related
I am trying to solve a system of differential equations by the Euler's method. First, I have encoded my system into a vector. Then I pass the initial conditions to the function ode_Euler.
However, there's somehting wrong about my attempt. I am getting this error:
>> nm06p03a
Unable to perform assignment because the size of the left side is 1-by-2 and the size of the right side is 2-by-2.
Error in ode_Euler (line 11)
y(k+1,:)= y(k,:) +h*feval(f,t(k),y(k,:));
Error in nm06p03a (line 12)
tic, [tE,xE]=ode_Euler(f,tspan,x0,N); t_Euler=toc;
This is my code so far:
clear, clf
f=#(t,x)[-x(2)+x(1)*x(2); x(1)-(0.5.*x(1).^2)+(0.5.*x(2).^2)]; %Encoding system of differential equations into a vector
t0=0; tf=10;
tspan=[t0 tf];
N=100;
x0s=[0.2 0]; % A matrix consisting of initial points
for iter=1:size(x0s,1)
x0=x0s(iter,:);
tic, [tE,xE]=ode_Euler(f,tspan,x0,N); t_Euler=toc;
subplot(220+iter),
plot(tE,xE,'r:')
legend('ode_ Euler')
end
Here is the Euler's method:
function [t,y]=ode_Euler(f,tspan,y0,N)
if nargin<4|N<=0, N=100; end
if nargin<3, y0=0; end
h=(tspan(2)-tspan(1))/N;
t=tspan(1)+[0:N]'*h;
y(1,:)=y0(:)'; %make it a row vector
for k=1:N
y(k+1,:)= y(k,:) +h*feval(f,t(k),y(k,:));
end
When I use this other method ode_Heun, I get the same error:
function [t,y]=ode_Heun(f,tspan,y0,N)
if nargin<4|N<=0, N=100; end
if nargin<3, y0=0; end
h=(tspan(2)-tspan(1))/N; % Step-size
t=tspan(1)+[0:N]'*h; % Time vector
y(1,:)=y0(:)'; % make the initial value a row vector
for k=1:N
fk= feval(f,t(k),y(k,:)); y(k+1,:)= y(k,:)+h*fk; % Eq.(6.2.3)
y(k+1,:)= y(k,:) +h/2*(fk +feval(f,t(k+1),y(k+1,:))); % Eq.(6.2.4)
end
Can I get some help to understand the problem with my code?
y(k,:) is a row vector, while the return value of f is a column vector. Per broadcasting rules the sum of a row and a column vector is a matrix as the sum of the matrices of repeated row and column vectors.
This is not very logical in the context of vector and matrix operations, but can make sense in the processing of (finite) sequences of vectors. Unfortunately that distinction is not realized and enforced in the type system.
I tried everything and looked everywhere but can't find any solution for my question.
clc
clear all
%% Solving the Ordinary Differential Equation
G = 6.67408e-11; %Gravitational constant
M = 10; %Mass of the fixed object
r = 1; %Distance between the objects
tspan = [0 100000]; %Time Progression from 0 to 100000s
conditions = [1;0]; %y0= 1m apart, v0=0 m/s
F=#(t,y)var_r(y,G,M,r);
[t,y]=ode45(F,tspan,conditions); %ODE solver algorithm
%%part1: Plotting the Graph
% plot(t,y(:,1)); %Plotting the Graph
% xlabel('time (s)')
% ylabel('distance (m)')
%% part2: Animation of Results
plot(0,0,'b.','MarkerSize', 40);
hold on %to keep the first graph
for i=1:length(t)
k = plot(y(i,1),0,'r.','MarkerSize', 12);
pause(0.05);
axis([-1 2 -2 2]) %Defining the Axis
xlabel('X-axis') %X-Axis Label
ylabel('Y-axis') %Y-Axis Label
delete(k)
end
function yd=var_r(y,G,M,r) %function of variable r
g = (G*M)/(r + y(1))^2;
yd = [y(2); -g];
end
this is the code where I'm trying to replace the ode45 with the runge kutta method but its giving me errors. my runge kutta function:
function y = Runge_Kutta(f,x0,xf,y0,h)
n= (xf-x0)/h;
y=zeros(n+1,1);
x=(x0:h:xf);
y(1) = y0;
for i=1:n
k1 = f(x(i),y(i));
k2= f(x(i)+ h/2 , y(i) +h*(k1)/2);
y(i+1) = y(i)+(h*k2);
end
plot(x,y,'-.M')
legend('RKM')
title ('solution of y(x)');
xlabel('x');
ylabel('y(x)')
hold on
end
Before converting your ode45( ) solution to manually written RK scheme, it doesn't even look like your ode45( ) solution is correct. It appears you have a gravitational problem set up where the initial velocity is 0 so a small object will simply fall into a large mass M on a line (rectilinear motion), and that is why you have scalar position and velocity.
Going with this assumption, r is something you should be calculating on the fly, not using as a fixed input to the derivative function. E.g., I would have expected something like this:
F=#(t,y)var_r(y,G,M); % get rid of r
:
function yd=var_r(y,G,M) % function of current position y(1) and velocity y(2)
g = (G*M)/y(1)^2; % gravity accel based on current position
yd = [y(2); -g]; % assumes y(1) is positive, so acceleration is negative
end
The small object must start with a positive initial position for the derivative code to be valid as you have it written. As the small object falls into the large mass M, the above will only hold until it hits the surface or atmosphere of M. Or if you model M as a point mass, then this scheme will become increasingly difficult to integrate correctly because the acceleration becomes large without bound as the small mass gets very close to the point mass M. You would definitely need a variable step size approach in this case. The solution becomes invalid if it goes "through" mass M. In fact, once the speed gets too large the whole setup becomes invalid because of relativistic effects.
Maybe you could explain in more detail if your system is supposed to be set up this way, and what the purpose of the integration is. If it is really supposed to be a 2D or 3D problem, then more states need to be added.
For your manual Runge-Kutta code, you completely forgot to integrate the velocity so this is going to fail miserably. You need to carry a 2-element state from step to step, not a scalar as you are currently doing. E.g., something like this:
y=zeros(2,n+1); % 2-element state as columns of the y variable
x=(x0:h:xf);
y(:,1) = y0; % initial state is the first 2-element column
% change all the scalar y(i) to column y(:,i)
for i=1:n
k1 = f(x(i),y(:,i));
k2= f(x(i)+ h/2 , y(:,i) +h*(k1)/2);
y(:,i+1) = y(:,i)+(h*k2);
end
plot(x,y(1,:),'-.M') % plot the position part of the solution
This is all assuming the f that gets passed in is the same F you have in your original code.
y(1) is the first scalar element in the data structure of y (this counts in column-first order). You want to generate in y a list of column vectors, as your ODE is a system with state dimension 2. Thus you need to generate y with that format, y=zeros(length(x0),n+1); and then address the list entries as matrix columns y(:,1)=x0 and the same modification in every place where you extract or assign a list entry.
Matlab introduce various short-cuts that, if used consequently, lead to contradictions (I think the script-hater rant (german) is still valid in large parts). Essentially, unlike in other systems, Matlab gives direct access to the underlying data structure of matrices. y(k) is the element of the underlying flat array (that is interpreted column-first in Matlab like in Fortran, unlike, e.g., Numpy where it is row-first).
Only the two-index access is to the matrix with its dimensions. So y(:,k) is the k-th matrix column and y(k,:) the k-th matrix row. The single-index access is nice for row or column vectors, but leads immediately to problems when collecting such vectors in lists, as these lists are automatically matrices.
I have an nx2 matrix r in Matlab reporting n draws from a bivariate normal distribution
n=1000;
m1=0.3;
m2=-m1;
v1=0.2;
n=10000;
v2=2;
rho=0.5;
mu = [m1, m2];
sigma = [v1,rho*sqrt(v1)*sqrt(v2);rho*sqrt(v1)*sqrt(v2),v2];
r = mvnrnd(mu,sigma,n);
I want to normalise these draws to the unit square [0,1]^2
First option
rmax1=max(r(:,1));
rmin1=min(r(:,1));
rmax2=max(r(:,2));
rmin2=min(r(:,2));
rnew=zeros(n,2);
for i=1:n
rnew(i,1)=(r(i,1)-rmin1)/(rmax1-rmin1);
rnew(i,2)=(r(i,2)-rmin2)/(rmax2-rmin2);
end
Second option
rmin1, rmax1, rmin2, rmax2 may be quite variable due to the sampling process. An alternative is applying the 68–95–99.7 rule (here) and I am asking for some help on how to generalise it to a bivariate normal (in particular Step 1 below). Here's my idea
%Step 1: transform the draws in r into draws from a bivariate normal
%with variance-covariance matrix equal to the 2x2 identity matrix
%and mean equal to mu
%How?
%Let t be the transformed vector
%Step 2: apply the 68–95–99.7 rule to each column of t
tmax1=mu(1)+3*1;
tmin1=mu(1)-3*1;
tmax2=mu(2)+3*1;
tmin2=mu(2)-3*1;
tnew=zeros(n,2);
for i=1:n
tnew(i,1)=(t(i,1)-tmin1)/(tmax1-tmin1);
tnew(i,2)=(t(i,1)-tmin2)/(tmax2-tmin2);
end
%Step 3: discard potential values (very few) outside [0,1]
In your case the x and y coordinates of the random vector are correlated, so it's not just a transformation in x and in y independently. You first need to rotate your samples so that x and y become uncorrelated (then the covariance matrix will be diagonal. You don't need it to be the identity, since anywya you normalize later). Then you can apply the transformation you call "2nd option" to the new x and y independently. Shortly, you need to diagonalize the covariance matrix.
As a side note, your code adds/subtracts 3 times 1, instead of 3 times the standard deviation. Also, you can avoid the for loop, using (e.g) Matlab's bsxfun which applies an operation between matrix and vector:
t = bsxfun(#minus,r,mean(r,1)); % center the data
[v, d] = eig(sigma); % find the directions for projection
t = t * v; % the projected data is uncorrelated
sigma_new = sqrt(diag(d)); % that's the std in the new coordinates
% now transform each coordinate independently
tmax1 = 3*sigma_new(1);
tmin1 = -3*sigma_new(1);
tmax2 = 3*sigma_new(2);
tmin2 = -3*sigma_new(2);
tnew = bsxfun(#minus, t, [tmin1, tmin2]);
tnew = bsxfun(#rdivide, tnew, [tmax1-tmin1, tmax2-tmin2]);
You still need to discard the few samples which are out of [0,1], as you wrote.
Suppose we have this Hamiltonian:
n = 10;
H = ones(n,n);
The density matrix is:
Ro = sym('r',[n,n]);%Density matrix
The equation of motion is:
H*Ro-Ro*H
The above equation of motion is the right hand side of the equation, the left hand side is the time derivative of density matrix.
How can I solve the equation of motion in Matlab without the symbolic math toolbox? I need to change the value of n. It can be up to 100.
In your dynamics function, reshape between vectors and matrices in order to use MATLAB's standard ode functions, which (to my knowledge) require vector inputs. Note, the symbolic toolbox isn't used anywhere in this solution. R can be any size n-by-n, within the constraints of your machine's memory.
function dR = dynfun(R,H)
%// R = n^2-by-1 vector
%// H = n-by-n matrix
n = sqrt(length(R));
R = reshape(R,[n,n]); %// reshape R to n-by-n matrix
dR = H*R-R*H;
dR = dR(:); %// reshape dR to n^2-by-1 vector
end
Call the ODE solver:
[tout,Rout] = ode45(#(t,R) dynfun(R,H), [0,T], R0(:));
where T is final time, R0 is n-by-n initial condition, tout are the output time steps, and Rout is the solution trajectory. Note due to the reshaping, Rout will be k-by-n^2, where k is the number of time steps. You'll need to reshape each row of Rout if you want to have the actual matrices over time.
I have a problem when calculate discrete Fourier transform in MATLAB, apparently get the right result but when plot the amplitude of the frequencies obtained you can see values very close to zero which should be exactly zero. I use my own implementation:
function [y] = Discrete_Fourier_Transform(x)
N=length(x);
y=zeros(1,N);
for k = 1:N
for n = 1:N
y(k) = y(k) + x(n)*exp( -1j*2*pi*(n-1)*(k-1)/N );
end;
end;
end
I know it's better to use fft of MATLAB, but I need to use my own implementation as it is for college.
The code I used to generate the square wave:
x = [ones(1,8), -ones(1,8)];
for i=1:63
x = [x, ones(1,8), -ones(1,8)];
end
MATLAB version: R2013a(8.1.0.604) 64 bits
I have tried everything that has happened to me but I do not have much experience using MATLAB and I have not found information relevant to this issue in forums. I hope someone can help me.
Thanks in advance.
This will be a numerical problem. The values are in the range of 1e-15, while the DFT of your signal has values in the range of 1e+02. Most likely this won't lead to any errors when doing further processing. You can calculate the total squared error between your DFT and the MATLAB fft function by
y = fft(x);
yh = Discrete_Fourier_Transform(x);
sum(abs(yh - y).^2)
ans =
3.1327e-20
which is basically zero. I would therefore conclude: your DFT function works just fine.
Just one small remark: You can easily vectorize the DFT.
n = 0:1:N-1;
k = 0:1:N-1;
y = exp(-1j*2*pi/N * n'*k) * x(:);
With n'*k you create a matrix with all combinations of n and k. You then take the exp(...) of each of those matrix elements. With x(:) you make sure x is a column vector, so you can do the matrix multiplication (...)*x which automatically sums over all k's. Actually, I just notice, this is exactly the well-known matrix form of the DFT.