Error using * Incorrect dimensions for matrix multiplication. Error in hbg_implict (line 23) out = A*y + B*u - E*yp;

Hey guys, Whenever i try to run cal my function I keep getting this error, and i am not able to find the error, because i think my matrix dimensions are right. would really appreciate if someone can fix it. Thank you in advance.

function out = hbg_implict(t, y, yp)
index = t*100000; 
ts = round(index);
n=0;
even = n(rem(ts,2)==0); 
if even == 0 %Change to 100kHz 
    il1=0; 
    il2=1;
    il3=1; 
    hsm = hss(il1, il2, il3); 
else
    il1=1; 
    il2=0;
    il3=1;
    hsm = hss(il1, il2, il3); 
end
u = [28;0]; % Input V 
d = length(hsm(:,1));
w = length(u(:,1)); 
A = hsm(:,1:d);
B = hsm(:,d+1:d+w);
E = hsm(:,d+w+1:d+w+d); 
out = A*y + B*u - E*yp;
end


function hsm = hss(il1, il2, il3)
L1 = 10e-3;
L2 = 2.25e-7; 
L3 = 1e-5;
R1 = 1.73;
R2 = 2e-5;
a = 0.00902;

A = [-1*xor(il1,il2)/(L1*R1) -a*xor(il1,il2)/L2 0 0 0 
    a*xor(il1,il2)/L1 -il3/(L2*R2) 0 0 0
    0 0 ~il3/(L3*R2) 0 0 
    0 0 0 ~xor(il1,il2)/L1 0
    0 il3/L2 0 0 -il3/L3];

B = [il1 il2 
    0 0 
    0 0 
    0 0
    0 0]; 

E = [xor(il1,il2) 0 0 0 0 
    0 1 0 0 il3 
    0 0 ~il3 0 0 
    0 0 0 0 0 
    0 0 0 0 0];

% Reduce size of model 
order = length(A(1,:));
z = zeros(1,order);
i = 1; 
for n = (1:order) 
    
% Delete rows of zeros 
if sum(A(n,:)) == 0 
else
    Ared1(i,:) = A(n,:); 
    Bred1(i,:) = B(n,:); 
    Ered1(i,:) = E(n,:);
    i = i+1; 
end; 
end;
i = 1; 
for n = (1:order) % Delete Columns of zeros 
    if sum(Ared1(:,n)) == 0 
    else
        Ared2(:,i) = Ared1(:,n); 
        Bred2 = Bred1; 
        Ered2(:,i) = Ered1(:,n); 
        i = i+1; 
    end; 
end;
hsm = [Ared2 Bred2 Ered2]; 
end

The solver that i am using it, is shown below:

tspan = [0:(1/1000000):9.5];
y0 = zeros(1,3)'; 
yp0 = zeros(1,3); 
[t_imp, y_imp] = ode15i(@hbg_implict, tspan, y0, yp0);

% Run simulation until event 
tspan1 = [0:1e-3:(5-(1e-3))];
y01=[0 0 0]';
yp01 = zeros(1,3)';
[t1,y1] = ode15i(@hbg_implict, tspan1, y01, yp01);

% Concatenate results vectors 
t = [t1; t2]; 
y = [y1; y2];

Answer 1

[Walter Roberson]

I know what is happening with your code, but at the moment I do not know why.

The second time your function is called, yp is a row vector instead of a column vector.

ode15i() documents that you should be receiving a column vector for that parameter, so I suspect a MATLAB bug.

The workaround is not difficult:

out = A*y + B*u - E*yp(:);

[rohit singh]

I tried it, but it is still not working.

[Walter Roberson]

yp0 = zeros(1,3); 

That line there: you need to change it to make yp0 a column vector. That is not what the documentation says, but it is necessary.

% Concatenate results vectors 
t = [t1; t2]; 
y = [y1; y2];

Problem: you do not have a t2 or y2. You created t_imp and y_imp and t1 and y1.

It is not obvious to me what would be wanted there. Your second run is a subset of the first run, just every 1000 samples out of it for a while.

[Walter Roberson]

I submitted a bug report about ode15i not treating the column orientation properly.

[rohit singh]

I have changed the yp0 zeros (1,3) to a vector. There were asign missing. And t_imp and y_imp should be defined as t and y.

But still getting the same error.

[Walter Roberson]

I made the change to yp0 and ran your code, and it worked until it went to try to do the [t1; t2]

If you post the repaired code for

% Concatenate results vectors 

then I will try again.

[Walter Roberson]

Side note: please check the line

            Bred2 = Bred1;

That is copying a fixed size of information to Bred2, and that size will not necessarily be compatible with the Ared2 and Ered2 that are being created, as those are like

           Ared2(:,i) = Ared1(:,n);

[rohit singh]

function out = hbg_implict(t, y, yp)
index = t*100000; 
ts = round(index);
n=0;
even = n(rem(ts,2)==0); 
if even == 0 %Change to 100kHz 
    l1=0; 
    l2=1;
    l3=1; 
    hsm = hss(l1, l2, l3); 
else
    l1=1; 
    l2=0;
    l3=1;
    hsm = hss(l1, l2, l3); 
end
u = [28;0]; % Input V 
d = length(hsm(:,1));
w = length(u(:,1)); 
A = hsm(:,1:d);
B = hsm(:,d+1:d+w);
E = hsm(:,d+w+1:d+w+d); 
out = A*y + B*u - E*yp;
end


function hsm = hss(l1, l2, l3)
L1 = 10e-3;
L2 = 2.25e-7; 
L3 = 1e-5;
R1 = 1.73;
R2 = 2e-5;
a = 0.00902;

A = [-1*xor(l1,l2)/(L1*R1) -a*xor(l1,l2)/L2 0 0 0 
    a*xor(l1,l2)/L1 -l3/(L2*R2) 0 0 0
    0 0 0/(L3*R2) 0 0 
    0 0 0 1/L1 0
    0 l3/L2 0 0 -l3/L3];

B = [l1 l2 
    0 0 
    0 0 
    0 0  
    0 0]; 

E = [xor(l1,l2) 0 0 0 0 
    0 1 0 0 l3 
    0 0 0 0 0 
    0 0 0 0 0 
    0 0 0 0 0];


% Reduce size of model 
order = length(A(1,:));
z = zeros(1,order);
i = 1; 
tic
x = 0;

for n = (1:order) 
    
% Delete rows of zeros 
if sum(A(n,:)) == 0 
else
    Ared1(i,:) = A(n,:); 
    Bred1(i,:) = B(n,:); 
    Ered1(i,:) = E(n,:);
    i = i+1; 
end
end
i = 1; 
for n = (1:order) % Delete Columns of zeros 
    if sum(Ared1(:,n)) == 0 
    else
        Ared2(:,i) = Ared1(:,n); 
        Bred2 = Bred1; 
        Ered2(:,i) = Ered1(:,n); 
        i = i+1; 
    end 
end
hsm = [Ared2 Bred2 Ered2]; 
end

Solver:

tspan = [0:(1/1000000):9.5];
y0 = zeros(1,3)'; 
yp0 = zeros(1,3)'; 
[t, y] = ode15i(@hbg_implict, tspan, y0, yp0);

[rohit singh]

I will have a look

[rohit singh]

Walter try to use the new code that i have uploaded. Thanks.

[Walter Roberson]

Your new A matrix is rank 4, but your boundary conditions are still only length 3

out = A*y + B*u - E*yp;

A, B, E are all 4 x something, but your y and yp are only 3.

[rohit singh]

How do i solve this?

[Walter Roberson]

yp = zeros(1,4)';
yp0 = zeros(1,4)';

[rohit singh]

It worked!! thanks a lot.

Walter, do you know why matlab doesnt accept this symbol " ˜ " ?

It matlab accept this symbol then I dont have to put the values for the vairables in the matrix, it automatically takes it.

you can see in the code shown below where i am trying to use the symbol:

function out = hbg(t, y, yp)
index = t*100000; 
ts = round(index); 
n=0; even = n(rem(ts,2)==0);
if even == 0; %Change to 100kHz
    l1=0; l2=1; l3=1; 
    hsm = hss(l1, l2, l3);
else l1=1; l2=0; l3=1; 
    hsm = hss(l1, l2, l3);
end
u = [28;0]; % Input V 
d = length(hsm(:,1));
w = length(u(:,1)); 
A = hsm(:,1:d);
B = hsm(:,d+1:d+w); 
E = hsm(:,d+w+1:d+w+d); 
out = A*y + B*u - E*yp;
end



A = [-1*xor(l1,l2)/(L1*R1) -a*xor(l1,l2)/L2 0 0 0
    a*xor(l1,l2)/L1 -l3/(L2*R2) 0 0 0 
    0 0 ˜l3/(L3*R2) 0 0 
    0 0 0 ˜xor(l1,l2)/L1 0 
    0 l3/L2 0 0 -l3/L3];


B = [l1 l2 
    0 0 
    0 0
    0 0 
    0 0]; 

E = [xor(l1,l2) 0 0 0 0 
    0 1 0 0 l3 
    0 0 ˜l3 0 0 
    0 0 0 0 0 
    0 0 0 0 0];

[Walter Roberson]

That is the Unicode "small tilde" symbol, U+02DC . You need the normal tilde symbol, ~