[ Follow Ups ] [ Post Followup ] [ Control Consulting ] [ FAQ ]
Posted by Mike McCarty on September 28, 2000 at 12:05:52:
In Reply to: Gain Consistency posted by Graeme Taylor on September 19, 2000 at 17:18:23:
Don has some good comments on this situation. This is fairly common, and
most engineers
don't pick up on it like you have. Although, this is a good application to
understand
the implications of your models since it is a small controller and you can
study the
various 2x2 response matrixes.
Another situation you should investigate with the relative gains of these
2x2 matrix responses,
is with your top reflux and btms temp vs C5's and C4's. When the controller is trying
to maintain
both compositions with the reflux and btms temp, it will either push the reflux
and temp higher or
push them both lower depending on the gains. Since each MV will help one spec
and hurt the other,
the controller will tend to continue to load up the column with these MVs, or
will unload the
column, when controlling both compositions. Which way it goes can't be
controlled by LP costs,
the relative gains of the models must be set properly. And obviously you want
the controller to
maximize column load rather than minimize it. The analysis of these models is
done the same way
you did for looking at the pressure and btms temperature. In this case, if you
are holding the top
C5's constant using reflux, you would expect the responses to give a decrease
in btms C4's when
firing up the reboiler.
As Don said, you may need to "tweek" your controller gains to get
the responses that you need. But
it should not require too much modifications, and I think this is justifiable
because there is always
some uncertainty in these models (due to instrumentation issues,
non-linearities of plant operations,
data analysis techniques, quality of plant test moves, etc.).
Hope this helps.
Mike