PID For Flow Control

[technics]

We are writing code for a two channel flow control that will direct two valves in response to input from their respective Ultrasonic Flow Meters. The control is to proportion each channel in ratios from 10/90 to 90/10. The actual overall flowrate is determined by a downstream flow control valve that we do not control, so all we are doing is maintaining the proportion.

I belive that the best method is to use a single PID loop that will increase the valve on channel #1 at the same time decrease the valve on channel #2 by an equal amount. So should the proportion be 51% channel #1 and 49% channel %2, we would set channel #1 to full open and #2 to full closed waiting for the down stream valve to open. When the downstream valve is open the flowrate would immediately be measured on the open valve and proportioning would begin. 

I expect that the downstream valve would eventually reach steady state, so that subsequent changes in position would be slight.

My question is - what do you think about that? Will it work? with a follow up question of - How might this work for three channels?

Looking forward to your response!

[AndyClegg]

Why do you think the valves would be fully open or fully closed ? Are you working with solenoid valves that can only be open or closed ?

If the measured proportion is different to the desired proportion (setpoint of the PID) then the controller will just adjust the valve openings (more open or more closed) according to how big the error is.

For the type of control action you are looking for look up the term "split range". However, I think you need to think about the control objectives and define them more clearly - since if all you are bothered about is the proportion, then you don't need to have two valves being controlled (look up the term "ratio control"). I think implicit in your description is an secondary objective to keep the total flow constant. If that is the case then the method you described won't reliably do that since you are relying on the flow through each valve to have an identical characteristic and pressure conditions - which are unlikley.

If you want to keep the total flow constant, then you should have one PID that does that by adjusting one of the valves, and have a ratio controller on the other valve to keep the proportion constant.

Hope this helps.

[technics]

Thank you for your opinion but I think you are missing something. (1) the valves are described full open and full closed because they have to start at some position, not that they are on/off valves. (2) Further, for typical ratio controls, one of the streams is wild while the other controls to its proper proportion. The trouble here is that  the pump pressure for the wild stream may be higher than the capability of the second stream. Therfore it won't matter how open the second FCV is, there will be no flow through that channel. (3) We do not have any control over the overall flowrate. That would be too easy.

[AndyClegg]

I think you'd need to provide a simple P&ID to show the layout of the three valves, and measurements and the proposed control loops (overall flow and composition control) before I can understand what it is you are proposing. Also it now seems your question is more around start-up rather than steady-state operation. Normally I consider the normal running / steady-state control actions as the main objective, as there is always someway to get the loop start-up (e.g. by operating procedures, initial conditions for controller outputs, bias terms) as start-up operation is very specific to the application (e.g. how pipes are pressurised, safety implications, etc).

It sounds like the valves are in addition to controlling the composition are used to reduce the supply pressure, so are right that if they start fully open fully closed they will eventually settle out. It would be better to have the set to roughly the right position at start-up (via the PID controller's initial output setting) so you don't have to wait so long, but it depends how critical that is - maybe not at all.

By trying to have the two valves that give the composition and a third valve for the overal flow - then you could get some interaction between the loops - but you will be minimising that interaction by making the first two valves maintain a roughly constant combined flow (it sound like split range is what you need for these). However, I would have the two loops tuned differently to help prevent any problems from interaction - overall flow control loop tuned to have a fast response, and the composition controller to be 5x slower.

If the setpoint for the overall flow controller is changed a lot and the composition can't be maintained with sufficient accuracy, you may need to consider feedforward control from the overall flow controller to the composition controller.