Posted by Mike McCarty on October 07, 1999 at 15:05:14:
This is the first of what we expect to be a regular feature on the board. "Tech tips" posted by various CCI engineers to provide control project tips and caveats, and encourage more discussion. Please post comments or questions you may have on our tips. We hope this will prove useful.
On complex distillation columns (like atmospheric crude towers), do you keep (or install) Q controllers at the DCS level on the intermediate P/A streams when implementing an MPC application?
A lot of DCS systems already have Q controllers configured on these pump-arounds, but usually it doesn't buy you very much to keep these in an MPC application. The main problem is that these are essentially temperature controllers, and instead of stabilizing the operation of your column they can easily destabilize it. If you have a top temperature controller, and two or three intermediate Q controllers, you have a situation where these controllers will fight each other, or at least significantly increase the settling time of the column.
As the top TC adds reflux to cool down the temperature, this will affect the first Q calculation, which will change this P/A flow. This affects the column, which the top TC has to respond to. The same thing happens going down the column with the other Q's. These tend not to stabilize the column like you would expect them to. Also the MPC responses can take into account these temperature changes in the column, and the MPC controller usually is more stable with fixed P/A rates.
One problem with intermediate Q controllers is the typical operating LMTD of the heat exchanger in this loop. If you have very close approach temperatures, or are at the capacity of the heat exchanger, the Q controller will end up making big moves in the P/A flow rate to control the heat duty. Even with a constant calculated heat duty, a big change in P/A flow rate will tend to upset the operation of that section of the column rather than stabilize it.
Another thing to watch for on any Q controller is problems with the dynamics of the system. Q is calculated as proportional to Flow*(Tin-Tout), but those three inputs do not have the same dynamic responses. Flow (as an MV) and Tin (as a FFWD) can change immediately or in a step, but Tout will only change based on the heat exchanger dynamics. Therefore, with no dynamic compensations, the calculated Q will not be stable with a Flow or Tin change - it will exhibit a lead type spike and then a slow decay to the final value. In order to configure a good Q controller, the Flow and Tin components should be dynamically compensated to match the Tout response.
One place where an intermediate Q controller can provide benefits is on a loop where some other controller is changing the heat duty. For instance, a light ends column reboiler is moving the Crude column P/A flow to control it's temperatures. Also, the P/A loop has an additional exchanger which can be used to maintain the total P/A loop heat removal. This is an ideal application for a Q controller. A proper configuration here would decouple the effects of the light ends column from the Crude column, and would allow simultaneous testing on the light ends and Crude columns.