I have a project which utilizes neural network as the controller for the project. Anyone who has done this before?, pls. tell me a site where i can find some examples about this topic. because it is quite hard.

 

Dear George,

I have been working with MPC and Neural Networks from the last 3 years. I have been using PROCESS PERFECT from PAVILION as the controler. You can find some information about this software in "www.pavtechc.com":http://www.pavtechc.com . I have been using MATLAB neural network toolbox to develop neural network models. You can find a lot of informations about nnets in the help of this toolbox (pdf).
If you give some more information about your process I can give some tips about how to use nnets for control.

Regards
Leandro Dariva

 

Dear Leandro,
Thanks for your information.Well my research is actually about the pH control of a neutralization process between a wastewater(considered to be strongly basic) and a strong acid. The study will be done using the neural network toolbox, matlab and simulink. Hope you can help me with this one.
thanks in advance.pls email me ( "[email protected]", mailto:[email protected] )if you have any informations.

Thank you so much,
George

 

Dear Sir
I am a rookie and delighted to hear someone is really working with MPC and NN. Can you please list few control applications where you found those control methods to be very good (compared to PID for example),

Thanks
Tim

 

Dear George,

A pH control is really an excellent example of non-linear process. To keep you process in the region of pH~7 is very difficult because the process gain changes very strongly in this point. I can help on building the stady-state model. Unfortunatelly I don't have a implemetation of a MPC able to use NN in MATLAB. What I have been doing is to tranfer the MATLAB model to PAVILON which actually made the control. I have some functions to construct and to analyse the NN models.

Regards
Dariva

 

Dear Tim,

We have Nonlinear MPC using NN controling Polypropilene and Polyetilene Process. The main variables controled are Reactor Density, Production Rate, Flow Index, Concentrations ans polymer density (for PE). What is you relation with MPC and NN?

Regards
Dariva

 

You might check out

"www.brainwave.com":http://www.brainwave.com

which describes a model based, adaptive, predictive controller. We have some projects where customers have employed this product in conjunction with a neural net and achieved very good results

 

Please refer to these articles as to why neural networks should NEVER be used in process control schemes.

[5] Turner, P, Guiver, J, “Neural Networks – A comprehensively unsuitable technology for control”, Proceedings of Aspenworld 2002, Washington DC, (Oct 2002)
[6] Turner, P., Guiver, J., “Neural Network APC: Fact or fantasy?”, Control Solutions Magazine, “Optimizing Your Process” supplement, pp. 16-20, (June 2001)
[7] Turner, P., Guiver, J., Lines, B., Keenan, M., “On the commercial application of next generation nonlinear model based predictive control”, Session Applications of Model Predictive Control, paper 144f, AIChE Spring Meeting, New Orleans, LA, Mar 30th – Apr 3rd 2003.
[8] Turner, P., Guiver, J., Lines, B., “Introducing The State Space Bounded Derivative Network for Commercial Transition Control”, Session Transition Control, paper FP18-2, 2003, Americal Automatic Control Conference, Denver, CO, June 4-6, 2003.
[9] Turner, P., “Automated Product Grade Transitions: Exposing the inherent and latent dangers of neural networks in manufacturing Process Control”, EUnite 2003, Safety Critical Task Force Meeting, Session AC3, Oulu, Finland, 10th-11th July, 2003