MATLAB Answers

System Identification Toolbox - Continous or discrete model?

11 views (last 30 days)
Nils B
Nils B on 20 Jul 2019
Commented: Star Strider on 20 Jul 2019
Hello,
I want to identify a control process and have gathered samples of a thermocouple over time with a sample time of 1s.
My question is, which kind of model (continous or discrete) would I ideally use to estimate the process with the Identification Toolbox GUI?
From my understanding both could make sense. Discrete because my data is sampled and continous because the discrete data is derived from a
continous signal(falling and rising temperature).
Maybe you can help me to understand, what kind of model should be used in which situation?
Best regards

  0 Comments

Sign in to comment.

Accepted Answer

Star Strider
Star Strider on 20 Jul 2019
It would seem to depend on your eventual application of the identified model. If you’re going to construct a completely analogue (continuous) system in hardware, identify it as such. If you are going to use sampled signals and implement a discrete controller, then identify it as a discrete system.
Also, I have nothing against the GUI, however it is likely easier to use the necessary functions themselves to generate the final identified model information. It depends on what you want to do with the results.

  0 Comments

Sign in to comment.

More Answers (1)

Nils B
Nils B on 20 Jul 2019
Thank you, I think I might get it now. Actually I want to construct a control loop in MATLAB - I just want to derive a model from the experimental data to test the control loop, before I put it into the hardware.
Well I though about using the functions as well however it seems rather complicated if you want to test things out. For example:
I have 3 heat generators and 3 thermocouples to measure temperature. For the sample data i turn on heat generator 1 and measure all thermocouples for a while. I do the same with the two other heat generators. I then have a 3x3 matrice, one field for each control loop.
I have not found a function yet, which can merge these 3 experiments with different y1,y2,y3 and u1,u2,u3 into a single model. Thats why I prefer to identify each single model in the GUI and try to combine them later. If you know a faster way let me now.
Best regards

  1 Comment

Star Strider
Star Strider on 20 Jul 2019
As always, my pleasure.
I agree that identifying different models is preferable, and you will later have to determine if it is possible to combine them.
The GUI is convenient at the outset, as you describe, to explore the various options. However I have always used the functions to get the final results, since that is simply easier.

Sign in to comment.

Sign in to answer this question.


Translated by