20,000 PSI System Odd PID Loop Behavior
20,000 PSI System Odd PID Loop Behavior
(OP)
We have a 20K pressure washing system installed in our plant. The system is run via a high pressure pump that draws from a water reservoir. On startup, the system starts by recirculating water through a dump valve back to the reservoir. The system then turns on a PI control loop with the output of the dump valve clamped at 21% (~4,000 psi) for ~30 seconds to drive air from the system and allow operators to check for leaks at low pressure. After 30 seconds, operators enable "High Pressure", which removes the clamp and allows the PI loop to control without the clamp. Setpoint is typically 19,000 psi.
Initial settings were P: 0.3 %/%, and I: 0.05 rep/sec. Bias is 0%. The controller is a standard ISA position PID loop.
Valve Position = Kc * (Error + Error * dt/Ti + Td * Derivative) + CV Bias.
The tuning parameters are: P = Kc, I = Kc/Ti,
I then tuned it, eventually reaching P: 0.2 %/% and I: 0.3 rep/sec. Reducing P below 0.2 %/% completely eliminated that cycling, and the new controller values have very little (but still some!) cycling. Reducing even further completely eliminated cycling, with a smooth ramp, but resulted in a much longer time to achieve SP.
Upon releasing the clamp, the below behavior was observed. The second graph is the final tuning parameters.
Initial:

Final:

With initial settings, the system would cycle, then eventually smooth out. Given the loop is PI only (no derivative control), I don't see any reason that anything would drive the valve back down to make it cycle like it did. The only reason I see as a possibility is this: The sample rate for recording and display is only 2 seconds. It is possible that system pressure went above setpoint due to integral windup during the clamp, causing the valve to re-close quickly, with all of this occurring before the screen or historian updated. It is odd, though, that the historian recorded fairly repeatable values - if the cycling frequency were not the exact same as the sampling frequency, it should have recorded some very high values. The controller also does not have any output rate of change limiter enabled that would only allow the valve to open X%/second
What is causing this cycling phenomenon?
Edit: New information. The current bit as described below is currently set a "0". Since we are starting at 100% error on startup (pressure is near 0 and setpoint is 19,000), the valve is calling for >21% open, leading to BIT 4 setting a negative initial integral value. Does that make sense?
Bit 4: Anti-reset windup action. When this bit is 0, the anti-reset-windup action uses a reset (integral term) back-calculation. When the output is clamped, the accumulated integral term is replaced with whatever value is necessary to produce the clamped output exactly.
When the bit is 1, the accumulated integral term is replaced with the value of the integral term at the start of the calculation. In this way, the pre-clamp integral value is retained as long as the output is clamped. This option is not recommended for new applications. Refer to CV Amplitude and Rate Limits below.
Initial settings were P: 0.3 %/%, and I: 0.05 rep/sec. Bias is 0%. The controller is a standard ISA position PID loop.
Valve Position = Kc * (Error + Error * dt/Ti + Td * Derivative) + CV Bias.
The tuning parameters are: P = Kc, I = Kc/Ti,
I then tuned it, eventually reaching P: 0.2 %/% and I: 0.3 rep/sec. Reducing P below 0.2 %/% completely eliminated that cycling, and the new controller values have very little (but still some!) cycling. Reducing even further completely eliminated cycling, with a smooth ramp, but resulted in a much longer time to achieve SP.
Upon releasing the clamp, the below behavior was observed. The second graph is the final tuning parameters.
Initial:

Final:

With initial settings, the system would cycle, then eventually smooth out. Given the loop is PI only (no derivative control), I don't see any reason that anything would drive the valve back down to make it cycle like it did. The only reason I see as a possibility is this: The sample rate for recording and display is only 2 seconds. It is possible that system pressure went above setpoint due to integral windup during the clamp, causing the valve to re-close quickly, with all of this occurring before the screen or historian updated. It is odd, though, that the historian recorded fairly repeatable values - if the cycling frequency were not the exact same as the sampling frequency, it should have recorded some very high values. The controller also does not have any output rate of change limiter enabled that would only allow the valve to open X%/second
What is causing this cycling phenomenon?
Edit: New information. The current bit as described below is currently set a "0". Since we are starting at 100% error on startup (pressure is near 0 and setpoint is 19,000), the valve is calling for >21% open, leading to BIT 4 setting a negative initial integral value. Does that make sense?
Bit 4: Anti-reset windup action. When this bit is 0, the anti-reset-windup action uses a reset (integral term) back-calculation. When the output is clamped, the accumulated integral term is replaced with whatever value is necessary to produce the clamped output exactly.
When the bit is 1, the accumulated integral term is replaced with the value of the integral term at the start of the calculation. In this way, the pre-clamp integral value is retained as long as the output is clamped. This option is not recommended for new applications. Refer to CV Amplitude and Rate Limits below.
RE: 20,000 PSI System Odd PID Loop Behavior
you might compare the piston cycle rate to you pulsations,
RE: 20,000 PSI System Odd PID Loop Behavior
RE: 20,000 PSI System Odd PID Loop Behavior
TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm
RE: 20,000 PSI System Odd PID Loop Behavior
Brad Waybright
It's all okay as long as it's okay.
RE: 20,000 PSI System Odd PID Loop Behavior
RE: 20,000 PSI System Odd PID Loop Behavior
PID is most adequate to obtain both.
Not sure I can help, just my two cents...
Life is not about waiting for the storm to pass. It's about learning dance in the rain.
RE: 20,000 PSI System Odd PID Loop Behavior
I cannot help either?just update for learning
More than 500.000 Parts available ship worldwide in fast delivery time (PLC, HMI, Inverters, Servo Drives, CNC, Motors - Encoders, Software, Low Voltage...
https://panasonicservomotor.com/HG-KR23/
RE: 20,000 PSI System Odd PID Loop Behavior
I have not made any headway on this. The cycling phenomenon would exist if I increased the gain again to a higher level. As it is, I've reduced the gain until the cycling was no longer evident. I'm still completely perplexed as to why the system cycles in the first place at higher gains; it is a PI controller, and it certainly isn't going much above setpoint (because doing so would blow the rupture disc on the pump).
If PV is below setpoint for a PI controller when the loop is initiated, there should not be anything that drives the valve lower - both the P and the I will be opening the valve. The only thing I can think to check is have our controls check the built-in formula to make sure the P and the I have the same sign (both should be positive wrt error).
RE: 20,000 PSI System Odd PID Loop Behavior
RE: 20,000 PSI System Odd PID Loop Behavior
TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm
RE: 20,000 PSI System Odd PID Loop Behavior
It would also be good to have the valve position.
Peter Nachtwey
Delta Computer Systems
http://www.deltamotion.com
http://forum.deltamotion.com/
RE: 20,000 PSI System Odd PID Loop Behavior
Just a further note on the system design - the control valve is a 3-way valve that dumps excess water back to the reservoir tank. Change in system pressure is nearly instantaneous with valve position changed due to the incompressibility of water.
IRStuff, I'm not sure I understand you. The change in pressure is near instant with change in valve position, so process response is extremely quick. If I were to change valve position manually from 15% to 80%, pressure would immediately jump from 4,500 psi to ~18,000 psi. The largest lag in the system is likely valve travel speed and/or sensing element response time. If you are talking about making loop changes before the system has a chance to feed back the response (valve travel time + sensing element change time + processing time vs loop update rate), then I get where you are coming from. The loop update rate needs to be a good bit slower than the sensed response for the loop to be stable. FYI, the loop update rate is 2 seconds.
I'm inclined to agree with Compositepro (and maybe IRStuff, too, if that's the point you are making?). It's likely that something physically in the loop (valve travel time, pressure sensor element update time?) is too slow for the loop udpate rate, and is causing instability.
RE: 20,000 PSI System Odd PID Loop Behavior
PID control has changed a lot from it's earliest implementations employed using electro-mechanical and pneumatic control. The PIDE instruction in Allen-Bradley Logix controllers applies the gain parameter to the process velocity (or rate-of-change if you will) instead of the error or PV signal by itself. It's very likely that any newer PID device employs a much more sophisticated and complex algorithm than what we learned years ago. That makes predicting the effect of a single tuning parameter more difficult. Also, my experience shows that the final tuning constants normally end up being unusually small, at least as far as the gain component is concerned.
Brad Waybright
It's all okay as long as it's okay.
RE: 20,000 PSI System Odd PID Loop Behavior
TTFN (ta ta for now)
I can do absolutely anything. I'm an expert! https://www.youtube.com/watch?v=BKorP55Aqvg
FAQ731-376: Eng-Tips.com Forum Policies forum1529: Translation Assistance for Engineers Entire Forum list http://www.eng-tips.com/forumlist.cfm
RE: 20,000 PSI System Odd PID Loop Behavior
Also, at that pressure, the pipe is stretching.
That makes two springs that are probably not accounted for in your system model.
Mike Halloran
Corinth, NY, USA
RE: 20,000 PSI System Odd PID Loop Behavior