In my case, we have HP/IP feedwater pumps with 9 user valves. I have not used a cascade control scheme like you have shown in your example, but in the past we have used valve position as the process variable for the speed controller with a fixed 90% setpoint.
As you've shown, we looked at the valve with the greatest % open and that was the process variable for the speed controller. With this configuration, the feedback we had from the commissioning group at site was that the loop did not control adequately due to the varying process dynamics associated with each of the user demands. For example, one of the valves was for HP feedwater and was part of the three element feedwater control loop and one of the valves was for attemperation and was part of a temperature control loop with a much smaller flow.
I never received any actual data back from the field to support their claims, however, I don't want to make the same mistake on the Becancour job. Does the problem make sense to you? I have modified the logic so that the two primary users (HP and IP feedwater) are controlling the speed based on the greater valve position (SP fixed at 90%), and I have added override controllers for each of the 7 users. This way, the valve position setpoints can be independently set, and each controller can be tuned differently. (I did consider using adaptive tuning, but opted for this solution instead).
I would greatly appreciate your comments with respect to this dilemma, and thanks again for your help, and the article.
Leslie Aitchison, PE, SNC Lavalin Constructors Inc., Redmond Wash.
ANSWER:
THE PROBLEM you have described in connection with your past project makes good sense, because every time a new valve becomes the measured variable (the most open one), the gain of the final control element in the loop (valve size) and therefore the gain of the total loop also changes. This necessitates the re-tuning of the controller (adjusting its gain) so that the gain product of the total loop will be returned to 0.5.
The cascade configuration I have shown in my column in the November issue on page 16 (Rick, please insert the figure here. Please increase its scale, because the published version was not sufficiently legible.) is one that I have used a lot and it seems to work well. Note that the cascade slave VPC has only an integral control mode and is provided with external feedback from the supply pressure transmitter, while the PC is provided with external feedback from the pump speed.
I also found that in case of 3-element feedwater valves one should not use the control signal as the measurement of valve opening, (because of the characterizing positioners used to linearize these valves), but should detect the valve opening by directly detecting the stem position.
As to your own control configuration, I do not believe that it is safe to leave the pump discharge pressure uncontrolled. From your description, it is not clear to me how the active control signal is selected from among the group VPCs and override VPCs?
Also, you have to be careful to prevent reset wind-up in the blocked (unselected) 6 VPCs, so that when control is switched from the operating 7th VPC and one of the 6 is called upon to take control, the transition will be bumpless.
As the 3rd part of my column on pump controls in the March 2005 issue points out, one really needs to have a full understanding of the pump characteristics, process dynamics, valve and positioner characteristics, to be sure that the system will be stable and efficient for sure.
RESPONSE:
THANK YOU so much for your response. I think we will likely modify our configuration to include the discharge pressure control as you've described.On past projects we have not had any problems with not directly controlling the discharge pressure (at least, none that were reported back from site!), but we normally configure a minimum speed override controller that will override the active control signal in order to maintain a minimum speed on the pump, and we have configured a high discharge pressure override controller, but this was primarily to compensate for pressure limitations on some piping!
The active control signal between the group VPC (HP and IP feedwater valve position) and the override VPCs is simply the greatest of the outputs. Under normal conditions the primary demands, the HP and IP feedwater will control the pump speed, however if one of the other user control valves opens more than it's setpoint, it's PID output will began to increase, once this output value is greater than the output from the group VPC, it will become the active control signal. For the HP and IP feedwater valve positions, we are using position transmitters and reading 4-20 mA position signals in the DCS.
We are using an Emerson Delta V system for this project, and they have developed a special output block for this type of override control. The PID loops that are not actively in control of the output have their integral action "turned off" so reset wind-up is not a problem. The transition from one PID in control to another PID is bumpless with this block.
Thanks again and I look forward to your next column!
Leslie Aitchison, PE, SNC Lavalin Constructors Inc., Redmond Wash.
| Need an Expert Opinion? |
Send your knotty control and automation questions to Béla Lipták at [email protected]. Lipták will either answer the questions himself or bring in specialists from his cadre of Instrument Engineers Handbook, Fourth Edition co-authors. Questions and answers will be published online on www.ControlGlobal.com. Each month, Lipták will pick the best question, and publish it, along with the answer in CONTROL. If your question is chosen, a prize will be awarded.
