Real-time linearization in Simulink for nonlinear MISO system

I am developing a model-predictive controller for a nonlinear multiple-input-single-output system and need to linearize the Simulink model in real time to get two transfer functions. Unless there is a better way to do this, I am trying to use the Time-based linearization block in Simulink. If I have two possible input-output pairs (after declaring two input points and one output point), how do I distinguish and access the transfer function I want using the Time-based linearization block?

Answer 1

[Kaustubha Govind]

If you have two inputs and one output, I'm not sure how you can get two distinct transfer functions for the same output. Is it possible to break the output into two components - one that depends only on the first input and one that depends only on the second input? If yes, you may want to create your model so that you have two output ports, one for each component.

[Victor]

Here is a clarification of what I would like to do. I would like to embed this into the Simulink model such that the model is linearized at each time step:

io(1) = linio('Model/Input 1',1,'in');

io(2) = linio('Model/Input 2',1,'in');

io(3) = linio('Model/Output',1,'out','on');

setlinio('Model',io)

op = operpoint('Model');

linsys1 = linearize('Model',op,io([1 3]));

linsys2 = linearize('Model',op,io([2 3]));

I tried putting this into a Simulink embedded function block and received an error saying that linio is not supported for stand alone code generation. Any other way to do this?

[Victor]

I got rid of the first error by typing this in the first line:

eml.extrinsic('linio','operpoint','linearize')

However, now I get a very long error message for the line:

op = operpoint('Model');

I'm still not sure this is the best way to linearize my Simulink model in real time because I have to introduce so many work-arounds. Does my question make more sense now?

[Kaustubha Govind]

But linearize is a static analysis tool. You cannot use it while the model is running. I don't know a lot about Controls, but you might also want to check out this toolbox: http://www.mathworks.com/help/releases/R2011b/toolbox/mpc/mpc_product_page.html

Answer 2

[Arkadiy Turevskiy]

You cannot use linearize in real time.

Your best bet is to linearize the model at several operating conditions offline, design MPC controller for each linear model, and then switch between these MPC controllers at run time as shown in this demo.

HTH.

Arkadiy

Answer 3

[maddikayala venkateswarlu]

Try this link, here you will get idea that how to model a MISO system. https://in.mathworks.com/help/ident/examples/dealing-with-multi-variable-systems-identification-and-analysis.html?searchHighlight=steamEng&s_tid=doc_srchtitle