Thursday, April 01, 2010

Fighting PID's

brinkmann27 (Electrical)
12 Mar 10 10:29
We have the ML1400, 1762-IF4 and 1762-OF4 modules. The system consists of one pump supplying 2 water systems. Each system has a flowmeter and a modulating valve. The flow is to be split 50/50 thru each system. The PID's are set up to take half the desired flow and adjust the modulating valve to achieve this. One issue is they are butterfly valves and at the desired flow being 10-20 % open, rather than the desired mid range at desired flow. I'm not sure of the valve resolution at the low end of the valve. (I'm not a valve guy)

I haven't been to the site yet, the customer reports the PID's the unit will "hunt" with a large range. Never stabilizing near the set point. I understand PID tuning is more of an art than science. To me this is a "P" system not needing the I or D. Question: What would be good starting parameters for the gain, and loop update? Do we need a larger (than 0) deadband to allow the system to stabilize?

How do we make these guys quit fighting?

Thanks in advance!  

djs (Electrical)
12 Mar 10 12:51
You might never be able to make the system work. The issue is your actuator gain is too high. In other words the valve at 10% open when at operating setpoint. This means your valve gain is nearly 10. Most valves have a hysteresis of 1 to 2% which is 10 to 10% of your setpoint. In other words the closest you will be able to come to matching flows is +- 20%!
itsmoked (Electrical)
12 Mar 10 15:34
You might also ask this question in forum378: Pipelines, Piping and Fluid Mechanics engineering.

Keith Cress
kcress -

brinkmann27 (Electrical)
12 Mar 10 15:46
Thanks Keith
cjw32 (Nuclear)
17 Mar 10 5:11

Sounds like you have the system responding far too quickly because your proportional gain is far too high for the characteristics of the valve.

Have you tried ziegler-nichols loop tuning? It gives a good starting point from which you can then fine tune the loop.

We have valves that operate at around 25% to 30% for flow control and it works perfectly.  We deliberately chose an over-damped response as our system is chemically very sensititive to sudden flow changes.  

You'll need to check the flow characteristics of the exact valve you've installed but for a generic butterfly valve you're flow is relative to the position so you're operating at 10-20% of flow so sounds like you've oversized the valves.

To be honest, I would start with a low proportional gain and a little integral to get an overdamped response.  You'll need to check the speicific set-up for the PID controller as a lower I setting usually means a higher frequency of intgrating the error and therefore a faster response, so you might want to set the I setting higher.

Once you have an overdamped response you can start increasing the P to speed up the initial response (but if you're operating at the lower end of the valve then I'd probably keep the P value low).  I'd then start decreasing the I to increase the valve response.   But again there are so many factors to take into account, such as the dead-time response, that it's hard to give specifics.

I would also look at the equation for the PID controller you have as well as that will give you an idea of how it will respond.
FrancisL (Computer)
22 Mar 10 8:14
As DJS said, it sounds like the valve sizing is wrong, so the process gain is too high. No matter how well you tune it the result will still have limit cycles due to valve stiction



Post a Comment

Subscribe to Post Comments [Atom]

<< Home