Re: DMC vs RMPCT

[ Follow Ups ] [ Post Followup ] [ Control Consulting ] [ FAQ ]

Posted by Anonymous on August 28, 2000 at 14:32:40:

In Reply to: Re: DMC vs RMPCT posted by Dan O'Connor on August 17, 2000 at 14:30:11:

I didn’t realize that Honeywell derives their Laplace transforms from the step weight models – it’s no wonder that you would get similar results. Perhaps the question should be whether there is better technology than what is contained in **both** the Aspen Tech package and the HW package? There are several software packages currently available (e.g., the Matlab System Identification Toolbox, the Control Arts Controller Model Identification software, MIDSA, Shell's ID package, etc.) that directly fit low-order transform models, which can then be directly used to generate impulse weight models for the controllers.

It’s not much of a problem to determine the form of the transfer function. You can use step weights for an initial guess (some of the programs above do this) or you could simply look at some step responses to estimate the model order. Also, there are several statistical tests in these packages to let you know if you have the right order. In contrast, there’s very little statistical analysis in step-weight identification to ensure that the parameters are correct.

As for the problem with odd shapes, you can use compound transfer functions (MacGregor uses one in his paper). Basically, these are two transfer functions for a single input added together, one with a longer deadtime. Some of the packages mentioned previously have this capability as well.

Follow Ups:

Re: DMC (1)

Post a Followup

Name :
E-Mail :
Subject :

Comments:
: I didn’t realize that Honeywell derives their Laplace transforms from the step weight models – it’s no wonder that you would get similar results. Perhaps the question should be whether there is better technology than what is contained in **both** the Aspen Tech package and the HW package? There are several software packages currently available (e.g., the Matlab System Identification Toolbox, the Control Arts Controller Model Identification software, MIDSA, Shell's ID package, etc.) that directly fit low-order transform models, which can then be directly used to generate impulse weight models for the controllers. : It’s not much of a problem to determine the form of the transfer function. You can use step weights for an initial guess (some of the programs above do this) or you could simply look at some step responses to estimate the model order. Also, there are several statistical tests in these packages to let you know if you have the right order. In contrast, there’s very little statistical analysis in step-weight identification to ensure that the parameters are correct. : As for the problem with odd shapes, you can use compound transfer functions (MacGregor uses one in his paper). Basically, these are two transfer functions for a single input added together, one with a longer deadtime. Some of the packages mentioned previously have this capability as well.
Optional Link URL:
Link Title:
Optional Image URL:

[ Follow Ups ] [ Post Followup ] [ Control Consulting ] [ FAQ ]