MPC

[YD17]

Hello,

I have implemented MPC for simulation and it works fine bt when i try to implement on real system with some modifications i am nt getting any results. I dnt have much idea too how it works, read alot of papers bt its all simulation based. I would appriciate if anyone would guide me through this.

Thanking u 

[GerdW]

Hi YD,

I have implemented MPC for simulation

When you use abbrevations you should explain them.

What is "MPC"?

Your keyboard is missing some vocals ("bt", "nt", dnt") - maybe you need to replace it?

[Barp]

Assuming the MPC is Model Predictive Control, you need provide more information about what is not 'working'. What are the modification you did? How did you define your model? What is the setup that you have?

Basically, MPC is tricky to use if you don't have complete knowledge of your system (you can't do model predictive with having a good knowledge of the system). In addition, any model-based control needs validation to be sure if the model matches the system.

[YD17]

Hello SIr,

I am working on temperature system with time delay and for that i want to implement Model predictive control. I have model of my system, by using the example of MPC with delay in labview i have done simulation of my model, but now i want to implement on my system, bt i dont know much about the parameters and block used to implement MPC on system.I have tried some modifications in simulation VI but it doesnt seem to work.

Thanking You

[Barp]

Again, the key point of using model predictive control is for you to define the model of the system. You need to specify the following:

a) the temperature system, in general, has at least one time-constant (pole). But it can also have other poles, although in general, they are all slow. You need to define the order of your system, i.e. how many poles does it have.

b) the time delay most likely has a value. You need to measure this value, which needs to be as correct as possible.

c) notice that the algorithm for Model Predictive Control does not take into account delays. You have to convert the 'continuous delay' into a discrete delay and 'append' your model. This example demonstrate that:

C:\Program Files (x86)\National Instruments\LabVIEW 2015\examples\Control and Simulation\Control Design\MPC\CDEx MPC Transfer Function Model With Delay.vi

Also, keep in mind that your simulation could not really take into account the delay information. If you create a first-order model with delay using a "Construct Transfer Function" model, add a delay and wire to the Transfer Function, then that function will NOT take into account the information. You have to explicitly use the 'delay' function from the Control and Simulation Loop.

Due to all the issues above, maybe your best bet is to use a Smith Predictor. This is a bit better to use then MPC for this kind of problem. Here is a shipping example of it:

C:\Program Files (x86)\National Instruments\LabVIEW 2015\examples\Control and Simulation\Case Studies\Process Control\Smith Predictor\Smith Predictor.vi

And here is a discussion about using it:

http://forums.ni.com/t5/LabVIEW/smith-predictor/td-p/3276061

Notice that without specific information (model) and how you are coding it, it is impossible for us to help you. You must put some more efffort to gives a better idea what you have and how you are implementing it.

And finally, you have to show that your model and your real system has similar features. A graph of a step response of the simulated system and the real system would help to demonstrate you know both cases.

[supriyad]

I have developed inverted pendulum control using MPC tool box on cRIO. But its taking large delay and after that it runs very fast. For same inverted pendulum i have used pole-placement method and LQR on cRIO, both are working fine. How to resolve the issue? Shall i upload image of interconnection?.if anyone has implemented plz give me suggestions about interconnections?