Tuning Interactive (series) Pressure Loops

Hi,

I have a process of 3 closed-loop pressure loops that slowly (30 min or so) oscillate 4% around setpoint 100% of the time. I have been trying to tune these loops for the past week using ExperTune software my company already has but since I lack experience with the software and tuning in general, I have found it to be less than helpful. I have been reading about different types of loops and how they interact and was hoping to get some thoughts from the group.

For my specific case, the series of loops in order are: Boiler->Header->Steambox. The boiler and header loops are always running and have fixed set points of 125 and 30 PSIG respectively. The header feeds 5 parallel processes including the steambox but only the steambox valve is oscillating with the header. The other 4 processes are 3 temperature loops whose valves are typically full open/closed the majority of the time (bad control there) and a steam line with flow indication only which is using up any excess steam. The steambox setpoint changes depending on the product we run but stays fixed throughout the run (for reference, say it's around 15 PSIG).

At this point, I have confused myself on how I should be tuning these loops. If I've interpreted what I've read correctly around the internet (This being the best website I've found so far), I need to detune the loops so that they are not controlling at the same period. I believe I should be tuning the boiler loop to be the fastest and the steambox the slowest but my intuition is telling me that it should be the other way around. That is to say, I'd want the steambox to be the fastest because that is what I'm controlling for my end process and therein lies my confusion.

I had actually gone the way of trying to make the header and steambox feedforward loops and feeding the upstream of each into the downstream but I think I've misunderstood the purpose of feedforward control and I havent been able to make it work as I expected. The documentation on the blocks is poor at best which hasn't helped.

Any advice to my tuning situation would be helpful and greatly appreciated. Thank you.

I am assuming there is a control valve between the 3 sections and one on the output of the steam box.

1. I agree that you should try to use feed forwards. Estimating the control output is always better than responding to errors.
2. The gains change as a function of the pressure drops across the valves so NO ONE set of tuning parameters will work under all conditions.
3. In think your intuition is right.

This is going to be one b!tch of a system to tune but you already know that. As and engineer I would take a step back and look at some numbers and obvious facts.

1 The changes in the pressure drops will change the gains.
2. If the pressure is going to be controlled at steady state the flow in must equal the flow out. That means that if the down stream valve must increase flow to maintain pressure the other valves upstream should instantly increase flow too so the in-flow matches the out-flow.
3. There are probably dead times so feed forwards would help here too.

Think about this. If the up stream valves were always opened or closed so that the in-flow matches the out-flow down stream would you even need a PID? Yeah, you probably would but the PID wouldn't need to do much because the feed forwards are doing most of the work.

I have a simulation that is similar but it involves cascaded tanks. In the steady state condition the in-flow must equal the outflow.
http://deltamotion.com/peter/Mathcad/TwoTanks/Mathcad - t0p2 2pi tc - Alin's two tanks Cascade.pdf
You can see the math gets messy with only two tanks and your system is the equivalent of having 3 tanks.
Notice that in my simulation the gain changes because the flow through the valves will change as a function of the tank level or the pressure drop across the valve.

Notice that I start out with differential equations. I was helping a student on LinkedIn. I told him I wouldn't consider giving him a passing grade if he couldn't write the differential equations for the system.

I did not use feed forwards in my two tank simulation because it would make the control almost perfect and I wanted to show some interaction between the PIDs.

I said this was going to be a b!tch.

I would first try to determine if this is a matter of interacting control loops that are setting up the oscillation or some disturbance with a 30 minute period that’s causing this. I’d want to rule out the disturbance as cause before investing too much more time in tuning.

Here are some rhetorical questions that you might think about…

Is the oscillation smoothly sinusoidal or have more of a notched/boxy shape (can you provide trend screenshots)? Does the oscillation persist when the steambox is out of service? Is one of the other 4 steam users on a 30 minute on/off cycle (if that’s the case, there’s the source of your oscillation, and that’s where I’d seek further understanding of what’s happening in that subprocess).

One way to determine if there’s a disturbance causing the oscillation is to operate the system in manual for a time period on par with the period of oscillation. If an oscillation persists at the same period while the system is in manual, you’ve definitely some disturbance at play. If you can’t operate it all in manual, maybe a loop at a time while the plant is otherwise operating steadily. If, when one loop goes to manual, the oscillations stop, that’s where you look.

Perhaps there’s a valve somewhere that’s the culprit: a control valve with no positioner is in a limit cycle, a non-return valve that’s acting funny, wouldn’t be a faulty steam trap (not for a large system anyway)?

Does the oscillation go away or change period with changes in boiler firing rate? Does the oscillation go away or change period while operating at a different boiler or header setpoint?

Is the oscillation adversely affecting operations?

Good luck.