Re: [foxboro] Feedforward control question
- From: "Rick Rys" <rys@xxxxxxxxxxxxxx>
- To: <foxboro@xxxxxxxxxxxxx>
- Date: Tue, 24 Jul 2007 00:04:53 -0400
Greg,
Pat has some good suggestions - I like to use the BIAS input of the PIDA
block, as it is more proven, but still can get exciting with some unexpected
LIMOPT behavior. I look at it as the Feedforward variable directly connected
to the PID.OUT, and the PID part just biasing the Feedforward.
Of all the upsets hitting this distillation column is feedflow really the
most significant upset. What about feed composition, feed temperature,
upsets in reboiler, or condenser, or perhaps upsets caused by poor
regulation of column pressure or maybe changing pressure setpoints. Even if
it is the most severe of all your upsets, Feedforward on feedflow (or feed
composition for that matter) can be quite difficult, and can result in
amplifying the upset rather than improving control. For example, if the
Feedforward dynamics are not correctly matched to the real column the
temperature controller may have already handled the upset, only to have the
Feedforward computation arrive as a second upset, or maybe the Feedforward
corrections arrive before the upset. There is a significant deadtime
component in the dynamics and it needs to accurate.
The first point to remember is that the only reason you are really using the
tray temperature as a control point is that you don't have a measurement
that is fast, accurate, and reliable that directly measures the bottom
composition. The relationship between the tray temperature and the
composition variable you are really trying to control needs to be
understood, because even when the temperature is controlling to the setpoint
you might not really have the desired bottoms composition. That said,
temperature control can often be a very reliable and effective method to
improve regulation of the bottoms (or overhead) composition.
Multiplicative vs. Additive may not be such a big difference, but getting
the dynamics right is the hard problem. The FFTUNE block just might have
the horsepower to solve this problem, but it will be a lot of work to figure
out how to make it work and handle the problem of changing dynamics. I
don't have enough experience to give much advice on FFTUNE. But, think of
the dynamics, i.e. cascading liquid down trays? or packing? That is just the
hydraulic lag, and then there is the more dominant time constant for the
composition of the liquid on the trays, that must dynamically reach a new
steady state.
Now suppose that the feed temperature changes, or the column pressure
changes, Now the percent of the feed that is vapor changes (A subcooled feed
will condense internal vapor). Vapor feed goes up to the condenser and
cascades back down through the whole column. A liquid feed only goes down
the section below the feed tray. Roughly a 2:1 difference in the time delay
for vapor feed vs. liquid feed, if the feed is near the middle.
If this is a new column, then you don't yet have the experience to know if
the bottom temperature control loop is fast enough to handle the most
typical feedflow upsets with the simpler feedback only arrangement. If you
are trying to simultaneously control the OVHD composition, perhaps with a
second temperature controller it is likely that this interacting control
loop will be the main source of your upsets to the bottoms temperature
controller. Sometimes you can adjust the controls and operation of the
upstream process, easier than trying to compensate for it.
The FBTUNE block can be fabulous when you just use the PRETUNE function to
tune up the Steam Flow and especially the Tray Temperature controller.
Basically put the PIDA in Manual, and just set the THRESH above the noise
band and the BUMP (BMP) large enough to exceed the THRESH. It typically
gets great tuning settings in 1 attempt with only a small upset from the
symmetrical bipolar pulse. Of course you should tune the Steam flow first
and make sure it is solid, before focusing on the Temperature loop. And be
sure to use derivative in the temperature loop (MODOPT=5). The Self-tune can
work, but requires more time to get results and needs to be watched or
severely limited to make sure it does not go unstable. The SELFTUNE works
with the controller in AUTO, but it can go to sleep if there is not loop
activity. Generally the PRETUNE is sufficient and very fast. But don't
bother with any of this until column pressure is rock steady.
Rick Rys
www.R2controls.com
508-339-6633 Home Office
508-369-5186 Cellular
_______________________________________________________________________
This mailing list is neither sponsored nor endorsed by Invensys Process
Systems (formerly The Foxboro Company). Use the info you obtain here at
your own risks. Read http://www.thecassandraproject.org/disclaimer.html
foxboro mailing list: //www.freelists.org/list/foxboro
to subscribe: mailto:foxboro-request@xxxxxxxxxxxxx?subject=join
to unsubscribe: mailto:foxboro-request@xxxxxxxxxxxxx?subject=leave
Other related posts:
