10 replies to this topic

[NoobMi]

Hi I am a new process engineer and is tasked to do some PID tunings for my plant cascade controllers

 

Cascade Loop is a level controller (master) linked to the steam flow rate (slave) of a reboiler in distillation tower.

 

Currently, the cascade loop is not controlling effectively, where it keep oscillating at avg of 50% +- 5%. We were hoping to control it within +-2%.

 

So how do i actually perform an online tuning (i.e. on a 'live' distillation tower)? Is it possible to do it without modelling the process? (i feel it might be too difficult for me to model it at the moment)

Do i have to stabilize the level first before any adjustment to the P & I settings? or during the oscillations i can start to adjust and observe for any effects on the overshoot?

[breizh]

Hi NoobMi

Why don't you ask the E&I team to support you ? This should be the job of an  Instrument Engineer /technician

My 2 cents

 

Breizh

Edited by breizh, 07 June 2013 - 05:45 PM.

[thorium90]

I have some possible suggestions. Take it off cascade first. See if the slave controller is stable or not. If the control is reasonable, you can then try on cascade mode, some trial and error, small changes at a time. Open loop tests can help you make a reasonable local approximation of the process dynamics.

It could be possible the instability is caused by something other than this controller... Eg: other controllers.

Edited by thorium90, 07 June 2013 - 10:20 AM.

[Bobby Strain]

Is the level oscillating 50%, or 5%, or neither of these? It would help to know more about the system installation, size, configuration, etc.

 

Bobby

[NoobMi]

Hi NoobMi
Why don't you ask the E&I team to support you ? This should be the job of an  Instrument Engineer /technician
My 2 cents
 
Breizh

normally we will ask outside vendor for help but for this time my boss want us to learn ourself to be less dependent on outside help.
our E&I engineer doesnt do DCS related stuff.

[NoobMi]

I have some possible suggestions. Take it off cascade first. See if the slave controller is stable or not. If the control is reasonable, you can then try on cascade mode, some trial and error, small changes at a time. Open loop tests can help you make a reasonable local approximation of the process dynamics.
It could be possible the instability is caused by something other than this controller... Eg: other controllers.

by taking off cascade meant setting the slave to auto mode and change its SV to see if i can stabilize the level at a certain steam flow rate?
if i can do it means it is stable?

i have also thought abt open loop test. Just to conform my approach is correct:
1) Break cascade and stabilized level
2) increase steam flow by 10% and observe changes to level
3) get dead time and slope of change to find PI settings.
will that be correct?

Edited by NoobMi, 07 June 2013 - 07:19 PM.

[NoobMi]

Is the level oscillating 50%, or 5%, or neither of these? It would help to know more about the system installation, size, configuration, etc.
 
Bobby

the level SV target is at 50% and the cascade controller is controlling it at 45-55% with an oscillating period of 20mins.

so my target is to reduce the oscillation to ~48-52%.

will tuning of the slave be sufficient and how do i know when to tune the master controller and how to tune the master controller?

[Bobby Strain]

You need help. You can find it at McMasters and at Emerson. A short read will get you on your way. It sounds as if there is no real problem. Sometimes it is good to have a fluctuating level for good reason. You need to step back and look at the whole system. And establish your control objectives. Saving energy, maintaining specification product, and achieve production rate objectives should direct your focus. Here is a superb treatise.

 

    http://openbookproje...ok/liii_0v2.pdf

 

Bobby

Edited by Bobby Strain, 07 June 2013 - 09:56 PM.

[curious_cat]

One idea may be to forget the notion of a set point and think in terms of a set band. 

[Subject_Name_Here]

Hello, fellow process engineer! Improving the process is in our job description, so anything that impacts the process, we are responsible for. The best way to tune anything is by trial and error. There are plenty of ways to calculate the PID settings, but typically, as even literature admits, they often fail to produce the best results.

 

I'm going to assume your operators start the controls in manual, hunt for the settings, and throw it in auto and that's when you're seeing the oscillations? It's always easiest to tune a loop by watching the process variable (PV) and controlled output (CO) on a graph. A good loop will have an overshoot decay of about 25% or less on each swing.

 

Adjust the PID settings on the master loop first while online after reaching steady state. Follow these guidlines:

 

GAIN (P) - The gain controls the magnitude of change in the CO in response to error. If you see drastic changes in the CO then reduce the gain. I usually change the gain by 0.25 increments, but of course this depends on your controller and process (every PLC/DCS has their own slightly modified PID equation). If you see not enough change in the CO then increase gain. If the controller operates better in manual mode, then this term needs to be reduced.

 

RESET (I) - The integral action will adjust for error over time. Say your PV is continuously above or below the setpoint... In this case you need to decrease the integral term. Remember, the equation is 1/Integral gain, so when you increase it, the term drives the CO less and when you decrease it, it drives it more!

 

RATE (D) - This term is usually not used and set to zero, however, it can be extremely helpful in some cases, especially if a lot of lag time is involved. By lag time, I mean the time it takes to see a change in the PV from when you changed the CO. If you want to minimize overshoot, this may be your solution. The derivative term allows the loop to sense when the rate of change in the process varaible is too much and it changes the CO accordingly.

[thorium90]

I would agree with Bobby's analysis. +-5% oscillation on a column level does not seem like a real problem. It probably is designed for some amount of oscillation which is likely caused by the controller taking larger responses, probably for disturbance rejection related to trip scenarios. ie: this controller of yours was likely tuned for disturbance rejection rather than smooth setpoint tracking. Its not uncommon for column levels to have large periods of oscillation due to the integrating nature of level control loops. Further more, yours is also coupled with a heating device (steam), which has longer response times compared to flow and pressure control loops. Therefore, my preliminary observation is same as Bobby's; you don't have a real problem.

 

However, a P&ID would shed some light on what are some other possibilities for the oscillation. A list of the protections for your column would help analyse whether it would be advisable to tweak the PID settings. Some screenshot of the trends showing the PV, SP and OP of the controllers in question as well as other relevant parameters might help reveal some underlying characteristics of the process that you did not realise initially.

 

I've attached an article from the web in case you decide to go ahead anyway.

Attached Files

•  CascadeTuning.pdf   340.69KB   27 downloads

Edited by thorium90, 09 June 2013 - 07:55 AM.