Re: Levels

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

Posted by Dan O'Connor on April 10, 2001 at 19:25:46:

In Reply to: Levels posted by Graeme Taylor on April 02, 2001 at 16:41:21:

: Can anyone give me their thoughts/experiences on the following?

: I am currently working on a crude unit DMC where we have the atmospheric level as a dependent variable. The data shows a clear model for the lowest pump back reflux to the atmos level, as would be expected (lowest sidestream is total draw). Also the data shows good relationships to dependent variables in the vacuum overheads for moves in this reflux. Should all these relationships be included in the final model?

: I think that including all these relationships will lead to the controller over predicting the impact of moving this pump back reflux, as the controller will see a requirement to move the vacuum feed as well to balance the level. The controller will predict the move in the vacuum feed to have impacts throughout the entire tower (due to the model being developed for the average feed quality in the test), where as in reality the increase in vacuum feed will be solely vacuum overheads material.

: What is the “accepted” solution here, include all the models or not?

Graeme,

I think you need to include them, but you also need to check the consistency of the model you build.
If you zero all of the vacuum curves (represented by the change in pumpback), then when you increase 1 bbl (per time) of pumpback, the controller sees the level
in the bottom change by 1bbl, and the predicted change in the vacuum tower conditions will be the same as
what is represented by the average vacuum tower feed. However, moving the pumpback is basically putting VLGO
into the vacuum tower, with basically no change in VHGO or bottoms (unless the lighter material has a significant
effect on the vacuum). My goal for the DMC models I develop is to make them represent reality as closely as possible. I would
leave them in, but make sure that your response curves (gains) are consistent with the actual change in material balance. One way
to do this is with simulate. Another is to load the gains into a spreadsheet.

Follow Ups:

Re: Levels Graeme Taylor 16:28:29 4/22/01 (0)

Post a Followup

Name    :

E-Mail  :

Subject :

Comments:

: : Can anyone give me their thoughts/experiences on the following?
: : I am currently working on a crude unit DMC where we have the atmospheric level as a dependent variable. The data shows a clear model for the lowest pump back reflux to the atmos level, as would be expected (lowest sidestream is total draw). Also the data shows good relationships to dependent variables in the vacuum overheads for moves in this reflux. Should all these relationships be included in the final model?
: : I think that including all these relationships will lead to the controller over predicting the impact of moving this pump back reflux, as the controller will see a requirement to move the vacuum feed as well to balance the level. The controller will predict the move in the vacuum feed to have impacts throughout the entire tower (due to the model being developed for the average feed quality in the test), where as in reality the increase in vacuum feed will be solely vacuum overheads material.
: : What is the “accepted” solution here, include all the models or not?
: Graeme,
: I think you need to include them, but you also need to check the consistency of the model you build.
: If you zero all of the vacuum curves (represented by the change in pumpback), then when you increase 1 bbl (per time) of pumpback, the controller sees the level
: in the bottom change by 1bbl, and the predicted change in the vacuum tower conditions will be the same as
: what is represented by the average vacuum tower feed. However, moving the pumpback is basically putting VLGO
: into the vacuum tower, with basically no change in VHGO or bottoms (unless the lighter material has a significant
: effect on the vacuum). My goal for the DMC models I develop is to make them represent reality as closely as possible. I would
: leave them in, but make sure that your response curves (gains) are consistent with the actual change in material balance. One way
: to do this is with simulate. Another is to load the gains into a spreadsheet.

Optional Link URL:

Link Title:

Optional Image URL:

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