Re: DMC vs RMPCT

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

Posted by Don Snowden on August 17, 2000 at 15:08:32:

In Reply to: Re: DMC vs RMPCT posted by Anonymous on August 15, 2000 at 20:01:45:

CCI associate David Hoffman is very familiar with MacGregor’s work. There is a lot of good work here but there are many issues surrounding transfer function identification that I would like to defer to another day.

There are a couple of points that you may not realize. As I understand it, Honeywell’s identification package FIRST uses the finite impulse form to analyze the plant data, then it simply uses the Laplace transform models to fit to the finite impulse responses! I personally think this is not all that bad (i.e. verses identifying the plant data directly with the Laplace transforms for the reasons you mentioned: not knowing what the shape of the response looks like).

As far as the “odd” shaped responses, this can be caused by numerous issues but can be very real. In most cases, it will have to do with the interaction of a PID loop (e.g. TIC in the bottom of a column). I think one would get much better behave response (e.g. 1st and 2nd order curves) if you just opened up many of the slower PID loops. But this is not what we recommend. We like to keep, for example, one end of the column “tied-down” with a faster acting PID to reject unmeasured disturbances.

As you mentioned, the bottom line is this: we can get “similar” response models with either package. The issue right now is the length-of-time it takes to implement the analysis. I don’t want to elaborate too much more on this topic as I will start to sound bias.

Follow Ups:

Post a Followup

Name    :

E-Mail  :

Subject :

Comments:

: CCI associate David Hoffman is very familiar with MacGregor’s work.  There is a lot of good work here but there are many issues surrounding transfer function identification that I would like to defer to another day.

: There are a couple of points that you may not realize.  As I understand it, Honeywell’s identification package FIRST uses the finite impulse form to analyze the plant data, then it simply uses the Laplace transform models to fit to the finite impulse responses!  I personally think this is not all that bad (i.e. verses identifying the plant data directly with the Laplace transforms for the reasons you mentioned: not knowing what the shape of the response looks like).  

: As far as the “odd” shaped responses, this can be caused by numerous issues but can be very real.  In most cases, it will have to do with the interaction of a PID loop (e.g. TIC in the bottom of a column).  I think one would get much better behave response (e.g. 1st and 2nd order curves) if you just opened up many of the slower PID loops.  But this is not what we recommend.  We like to keep, for example, one end of the column “tied-down” with a faster acting PID to reject unmeasured disturbances.

: As you mentioned, the bottom line is this: we can get “similar” response models with either package.  The issue right now is the length-of-time it takes to implement the analysis.  I don’t want to elaborate too much more on this topic as I will start to sound bias.

Optional Link URL:

Link Title:

Optional Image URL:

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