Linearized model and discrete model?

[Vicky Vicky]

I already have a linearized model in state-space form for my MPC implementation,

How do I convert to a discrete form to be used in the MPC implementation since it requires

a discrete state-space form? 

[Barp]

Vicky Vicky,

Under the Control Design Palette, you can find the Model Conversion VIs that convert Continuous Models into Discrete and, at the same time, you have convert between type of models and you can convert any Linear Time Invariant Model into State-Space Model. Notice that you have to create a model in a way the VIs accept and you can do that on the "Model Construction" palette.

[altenbach]

It would help if you could write a paragraph explaining in a bit more details what you are trying to do.

Not everybody is familiar with your possibly personal definition of MPC. 

[Vicky Vicky · Accepted solution]

Hi Barp, altenbach and all....

I write to give details of my interest in implementing

an MPC controller for a four-tank process.

   Attached is my VI and a note with details, please how is the

discretization implemented...I hope to get the VI back after discretizing.

 regards, Vicky . 

Message Edited by Vicky Vicky on 05-11-2010 08:31 AM

[Barp]

You have all the tools available in Control Design Palette to do the discretization. Just for fun, I tried to code your formulas and create the model directly. The Function "CD Continuous to Discrete.vi" does the conversion of 0.1 s using the "Zero-Order Hold" method. For more information, please consult the manual and online help.

Altenbach, the Model Predictive Control (MPC) is a control algorithm used by the process control industry that run an optimization algorithm that tries to keep the process variable away from the constraints and try to preview how the system behaves and "predicts" which is the best controller for this process.

Message Edited by Barp on 05-12-2010 08:29 AM

[Barp]

ah... the snippet did not keep the variable values. Please use the VI in attachment to look at this VI working.