Error using transfer functions

Please help me sort out the issue. I am not understanding why the gain block dimension is shown as 2, where as i am expecting only 1. Please see the attached screenshot of simulink blocks.

[Birdman]

Please share the simulink model(export to r2017a)?

[Shyam]

I am using R2016b. Please let me know how to export. I have included screenshot of full model and the error message.

[Birdman]

You do not have to export then. Just send.

[Shyam]

I have uploaded.

[Birdman]

I am not getting any error in simulation, but it is running very slow.

[Shyam]

I get these errors, included in attached screenshot.

[Shyam]

Hi, errors are no longer there. I just closed and opened the model. But, as you said, it is running very slow. Any idea, why...? It runs normally if the compensator blocks are not there (2 transfer functions and 1 gain block), which adds to the summer

Answer 1

[Birdman]

It is normal for simulation to run faster if you substract Transfer functions from the model but the main reason it is slow is that there are too many mathematical calculations to be done and too many states coming up from transfer functions and integrators. The variable-step solver tries to automatically adjust the step size for solving the states but since this model is an electrical system and since naturally electrical systems tends to behave quickly, the solver reduces the step size to almost 1e-4 and this results in simulation to slow. I have some suggestions:

-First, try to find the most suitable sampling time for your model. This means discretizing the continuous blocks in a suitable manner(Transfer Functions and Integrators). Because the variable step solver automatically increases the step size(narrows) in order to give you a solution but this results in slow simulation. I suggest you to look for Bode Plots of Transfer functions and controllers in your model.

-Second, be careful with the potential algebraic loops while discretizing the system.

[Shyam]

Thank you, let me try.

[Birdman]

Ok, let me know the results.

[Shyam]

Could you please help me run this simulation.

[Birdman]

What are your controller parameters, J, B, Jem and Bem?

[Shyam]

I have uploaded the init file.

[Birdman]

Firstly, how did you select your Kp and Ki parameters? They were not realistic and eventually I had to decrease them. The reason that it gives error for their initial values is that your control signal rises so quickly that the solver step size can not adapt to its quickness no matter how it tightens the step size. Therefore, one solution would be to decrease the controller parameters. I dropped Ki to 70 and Kp to 5. Still, there is an unstable response, which you need to check out. You may resdesign your controller. Please do not pick the controller parameters randomly. Hope this helps.

[Shyam]

Thank you very much for your time and effort. kp and ki were not put randomly, but the design could be wrong. I will check and reply. Once again, thank you.

[Birdman]

Ok, let me know the results.

[Shyam]

There was an error in my earlier PI design I am trying to design the PI for this system (attached). Could you please suggest some method to do it. I have derived the transfer function, act/input. I am unable to see what is the open loop transfer function so that i can design the PI using bode

[Shyam]

Thanks, i figured out the issue.

[Birdman]

I was about to solve it, too. Glad you figured that out.

Answer 2

[Shyam]

Another problem involving simple transfer function. I have attached the simulink model and init mfile. I am getting the error regarding singularity in the solution.

[Birdman]

I solved it by changing your solver to Fixed-Step and adding some delays to prevent algebraic loop. But please make sure that your model is reasonably defined.

[Shyam]

how to change it to .slx Reasonably defined meaning that values of the parameters are reasonable? Thanks, i will check that.

[Birdman]

Values that will not cause algebraic loop. You do not need to change to slx.

[Shyam]

ok. When i click on the link, i am seeing some code instead of simulink blocks.

[Birdman]

No, you have to download it first to your computer. Then open it at Simulink :)

[Shyam]

There is no download option. As i click on the link, the code appears. If i right-click also, there is no download option. Sorry to disturb with these silly things.

[Birdman]

Right click and click on Save link as

[Shyam]

I am getting the following error message (attached)

[Birdman]

What is your MATLAB version?

[Shyam]

R2016b..Today license is getting expired. I am installing the latest version now.

[Birdman]

Then you can easily open this one in the latest version. Therefore I am not sending the exported version to R2016b.

[Shyam]

ok..

[Ramu Nair R]

I installed latest version and opened the model. I am trying to use memory blocks instead of delay block. Also, is it necessary to use z domain for transfer function. The block is a part of a bigger system, which are in continuous time, fixed step and solver is Runge-kutta

[Ramu Nair R]

It worked. The problem was with initial condition. Thank you very much for your time and effort.