Represent Simulink Integrator block as Matlab Function

Hi.

I need to implement the following behavior :

The Integrator and my_Integrator blocks have to be equivalent I/O.

How should I write the Matlab Function ?

Thanks for any reply.

[Azzi Abdelmalek]

What is your goal?

[River Rock]

Please see http://www.mathworks.com/matlabcentral/answers/55705#answer_67551

Answer 1

[Ryan G]

As this looks like a homework problem, I can't answer directly. However I will point you in the direction of persistent variables.

Answer 2

[Azzi Abdelmalek]

I don't know why do you need this, maybe if you explain exactly what you need, there is better way

[River Rock]

The model diagram should be as light as possible, as shown in the first attached picture: one input and one output, no additional simulink blocks.

[Azzi Abdelmalek]

You can't do it

[River Rock]

Sure we can. Ryan's idea works.

Would it be easier to use a S-function instead of a MATLAB Function ?

[Azzi Abdelmalek]

What about time t?

[Azzi Abdelmalek]

Can you post the solution?

[Azzi Abdelmalek]

It works with Ryan's hint, but you still need a clock as an input.

[River Rock]

The solution it's already posted above. But since you and Guy disapprove this, I am waiting for the better solution.

I didn't need any clock input , please see http://www.mathworks.com/matlabcentral/answers/55705-represent-simulink-integrator-block-as-matlab-function#comment_115977

And even if I did, I guess it could have been fetched using the get_param command (never tried this though)

[Azzi Abdelmalek]

Ok, I see, If T is constant, you must then set, in model configuration parameters your fixed step time to T, and also your step block sample time to T. In this case you don't need a clock.

function y = fcn(u)
persistent   uold yold
T=0.01;
if isempty(uold)
 uold=0;yold=0;
end
y = u*T+yold-(u-uold)*T/2
yold=y;uold=u;

[River Rock]

Changing the sample time of the Step block to 0.01 removed the previous offset. Thanks

Answer 3

[Guy Rouleau]

This is not a good idea. The MATLAB function is not designed for this purpose.

[Edward Rodriguez]

Excuse me, so, What would be a good idea to implement numerical integration methods in blocks in Simulink?

Answer 4

[River Rock]

My main goal is to implement the differential equations of a physical system using a single Matlab Function. As the sums and gains were easy to represent, I couldn't find any alternative for the integration.

[Ryan G]

Why do you need to use MATLAB over SL blocks? Look into the persistent variables, it will get the job done, although as Guy mentioned, it's not the best way to do it.

[River Rock]

Your idea would be to declare a persistent variable for the numeric integration ?

Y(s) = U(s) / s => y(z) = yOld + u(z)

so you declare persistent x = y to stand for yOld ?

Can I avoid this low-level arithmetic and call a predefined method instead (ode45 for example) ?

[Ryan G]

What you have written is close it would be more like:

y(z) = yOld+u(z)/SampleTime

You cannot use the ODE solver in the MATLAB function block.

[River Rock]

Any idea on how to get rid of this offset ?

http://i.imgur.com/cxr5H.png

Code looks like:

function y = fcn(u)
%#codegen
T=0.01;

persistent yOld;
persistent uOld;

if (isempty(yOld))
    yOld = 0;
end

if (isempty(uOld))
    uOld = 0;
end

y = yOld + (T/2)* (u + uOld);

%y=yOld + u*T;
yOld = y;
uOld = u;

Answer 5

[River Rock]

Can anybody suggest a better way of implementing the numerical integration ? The code has to be written inside the Matlab Function Block though.

[Kiran Mahmood]

I'm working on a similar problem. Did u find the solution ? Need help.