Posted by Naidu on November 30, 1999 at 13:53:36:
In Reply to: Re: High Purity Tower DMC control - effect of operating state variation on controller performance posted by Don Snowden on November 29, 1999 at 13:19:36:
The parameters & description of the system are as follows:
This is a high purity tower based in a refinery.
It processes feed from the gas plant as well as from outside vendors.
Feed contains roughly 70% Propylene and 30 % propane with ethane i
n ppm range.
Feed enters the tower under flow control.
40 lb steam provides reboil heat.
Steam rate is used to manage / control the bottoms level.
Bottom product therefore leaves under flow control.
Over head rate is used to control condenser accumulator level.
Reflux rate can be independently controlled.
There is no top or bottoms temperature control. Pressure floats although
vent rate flow controllers are available.
At the time of original plant test,
approx. feed - 250 BPH
approx. bottoms - 63 BPH
approx. reflux - 2627 BPH
approx. ovhd product - 171 BPH
approx. steam rate - 60 MLBHR
Overhead propylene spec is set between 100.0 % & 99.5 %.
Bottoms propylene spec is set between 10.0 % & 5 %.
This is a typical propylene / propane splitter with over
100 stages and high reflux ratio.
Immediately after the original implementation, the controller ran staisfactorily,
with the LP
1)driving overhead spec towards the low limit - 99.5% mostly with
bottoms,
2)driving bottoms spec towards the high limit - 10% mostly with
feed,
3)driving flood indicator just below the high limit mostly with reflux.
No transformations were implemented for overhead and bottoms spec.
Due to the change in the operating philosphy, this tower is now used to
process the swings in the Propane / Propylene inventory while a newer
splitter is base loaded with max feed.
Maximum variations are as floows:
30% - feed, 30% - bottoms, 7% - reflux, 19% - overhead product
and about 12% for steam. While DMC was on feed changes were made by
simply adjusting feed MV high & low limits as and when needed.
Before the operating philosophy chnage - controller performance was
typical of a good DMC installation. Smooth, stable always maximizing profit
with steady accum. pred error.
Now under variable conditions, when big feed changes are made along with correlated
& coordinated changes in bottoms & reflux, accumulated prediction errors
are not steady - they swing widely indicating severe model mismatch.
Ampltidute of prediction errors are much higher than normal.
Besides analyzing the splitter performance, I have been reading
publications on high purity distillation dynamics and control.
These are inherently nonlinear with the top & bottoms ends of
both stripping & frationation sections pinched in Both composition
and temperature. The desired separation takes placein the middle of
the stripping section.
I believe that I have to use multiple models befitting operating
range variation dictated by feed change.
I do appreciate your comments and apologize for the lengthy posting.
Any comments / suggestions welcome. I will pass along how I am
going to solve this problem & the results.