Recent Control & Safety Systems Discussions
Similar Posts
Setting up separate shelved and out-of-service alarm lists and making out-of-service alarm list to automatically populate
Need help with diluent trim control
In need of help
Student in need of help
Enabling smooth transitioning of slave controllers to cascade when part of split range or ratio control scheme and master cannot be initialised
Share

Help with a separator oil box level control - oscillatory out flow, need smooth outflow

Hi everyone,

I am trying to control a horizontal separator oil box by maintaining steady outflow, and am not too concerned with the level set point.  The separator is the second vessel after the slug catcher, which is essentially a surge vessel.  The problem we are having is that the way these controllers were tuned originally is to maintain set point.

As an example, the separator oil box level controller 2-LIC-107, has a set point of 2600 mm.  There is a flow meter on this oil outlet line that I have converted to an FIC, so we can send a CAS set point.  From what I have been reading, the LIC should be proportional only (no integral), so that the response to inflow changes is proportional, and not exaggerated by the integral action trying to maintain set point.  Can someone give me some advice as to how we should be controlling this process?  Is there other DeltaV functions that could be used in this situation to maintain steady outflow? (I deadband, PIDplus, nonlinear gain etc.)

Thank you!

Jayme

Controls specialist

controls_wl

29 Replies

  • James Beall

    In reply to controls_wl:

    I recommend tight control on the interface. The "arrest time" is the same as Lambda in the ISA paper I sent to you. The arrest time on the IF level should be at least 3 x the dead time identified by the tuner. I would not use gap control on the IF level.

    I would use Lambda on the oil level but you need a long arrest time (lambda) to absorb variability. For integrating processes, I don't recommend NL_MINMOD = 0 since the level will oscillate between the gap limits. You might use NL_MINMOD =0.25, then you must make Reset=Reset normal/NL_MINMOD. The paper I suggested talked about how to see what Lambda (aka arrest time) you need to keep the level within +-ALV% for a given max load disturbance MLD (use nominal %out of LIC OUT (percent of scale).
  • controls_wl

    In reply to James Beall:

    Thank you for the great explanation.

    Just to make sure I am understanding....if I have NL_MINMOD of 0.25, and my reset value from insight is 65.4 sec normally, I need to make it 261.6? (regular reset/nl_minmod)?

    Thank you

    Jayme
  • James Beall

    In reply to controls_wl:

    Jayme,
    Yes, for an integrating process like most level. Note that you do not have to do this for self-regulating processes.
    The tuning you sent looked like it was for the IF level, it seems too fast for the oil level. If you can send me a sketch with normal I/F and Oil levels, separator dimensions and Level Transmitter location and length, I'll look it over. You can email them to me at james.beall@emerson.com if you prefer.
  • LaurentB
    I’ m little bit disappointed on last replies about KNL..... my suggestion was to use it only to reject process noise, not to improve robustness to get tight level. For me (may I’m wrong ????) nl_minmod =0 and do not touch value proposed by Insight.
    For Insight go to the menu and select option Expert, to get more tuning access. You can also on simulation TAB to get an overview ( not absolutely accurate) of the expected response on sp change or disturbance.
    Now if you use KNL not only for process noise rejection but to get tight level, Why to reduce integrale action outside the deadband, this action will be also reduce inside the dead deadband . Multiplication factor is not applied on Ti but on 1/Ti
  • James Beall

    In reply to LaurentB:

    Hi Laurent,
    You idea for KNL (=0) in the gap to reduce noise is perfect, and very ingenious, for self-regulation processes! And, as I said, there is not need to modify the Reset for in or out of the gap. However, for an integrating process like most levels, it is rare that the PV holds at a steady state value in open loop control (KNL=0), it will either be increasing or decreasing, thus the Gap control with 0 gain in the gap will create an oscillation of the level and controller output. Reducing the KNL in the gap stops the oscillation and reduces the transfer of PV noise to the Output (which could be a valve or SP to a secondary loop). Secondly, with an integrating response, reducing the controller gain and leaving the Reset at the same value (assuming it was calculated to be critically damped as Lambda tuning does) will result in oscillations of the loop. Thus, I recommend slowing the Reset down (larger number) so that when the lower gain is being used in the Gap, the loop won't oscillate. If you indeed take KNL to 0, then even for the integrating case, you can leave the Reset as calculated by InSight (or other method) for the case of outside the gap. However, since NL Gain option does not allow a different Reset time outside of the Gap, it has to stay slower. That is why for these cases I use the 3-region "Gain Scheduler" template in the Library to have a separate set of Gain and Reset when the PV is close to the SP (in the Gap) and a different set of Gain and Reset when "outside the Gap", thus having non-oscillatory tuning in both areas.
    I think in this case, the I/F tight level controller and the Oil level "averaging, aka slow" control apps have been comingled. I was speaking more to the Oil level case and was a bit surprised by the recommended tuning...maybe it was for the I/F level controller with tight control. Good questions/discussion! Let's get some more details and we'll talk more!
    James
  • LaurentB
    Hi James
    Thx for your clear explanations. In fact we were not speaking exactly of the same things regarding Lambda tuning. As you said lambda tuning can be to much agressives for these application. Reducing integrale action is not the direct consequence of KNL but and overall rules.
  • LaurentB

    In reply to LaurentB:

    it's also very ingenious to use Gain scheduler template. It's a very usefull template but used too much rarely !!!
    And you are so right that level is not steady at SP when the loop is open !!! And having KNL 0.25 is very ingenious comparing Zero which introduce a long period oscillation around the deadband. It's too late for me to go back on FPSO to modify it ;-)
  • Ejiogu Evans Nyagazie
    Please,  I need help on how to resolve this issue 

  • controls_wl

    In reply to James Beall:

    Thank you again James, and everyone who has helped on this issue thus far!

    I will be applying the gap tuning to the separator A oil box on Monday.  I was able to do some testing with it on Thursday, but operations did not want to leave it over night, so will get back to it next week.

    On another note, I applied the gap and lambda tuning to our interface.  (I know you said it is not recommend, but I will tell you why I did it, and why it appears to be working in our situation)

    The interface probe is an Agar (probe length is very short, around 20") inserted at a 45 degree angle in the interface zone at the back end of the separator.  We have always had issues with interface on the separators, as it is the second vessel after the surge vessel (slug catcher or free water KO),  and with our thermal oil sands production, we do get a rag layer some times.  With that being said, I always felt something was wrong with the way the interface and water dump controlled.  The interface probe PV and the water dump output would constantly oscillate, which made me realize that the probe length being so short (and on a 45 degree angle) was causing the controller to constantly oscillate.  With the gap tuning, it seems to have calmed down the water dump valve output substantially, and has propagated the affects to the oil box and the next vessel (treater).  When the operators run the vessel in manual, they are able to basically flat line everything.  to do that, they make smaller moves periodically on the water dump, and then check physically oil/water taps in the vessel to make sure the water level is where they want it.  With what you suggested, the controller is now doing this to some extent.  We are trusting the direction of the inteface PV, but not trying to maintain a strict set point.

    Here are some trends of the progress so far (and we have not even implemented the gap tuning on the oil box yet!!

    Before we applied gap and re tuned with lambda (notice the water dump valve output)

    After gap tuning and re tune with lambda (output is a lot more steady instead of up down oscillations in phase opposition to the interface PV)

    So far, operations has said that oil treating is good, and Separator A water taps etc. seem to be good.

    Now on to the Separator A oil box

    Before:

    After:

    Treater oil out (downstream vessel after separator A, at which point goes to spec sales oil)

    Before

    After:

    Here are all of the tuning values:

    2-LIC-106 IF control

    Range: 0-100%

    Typically SP: 62 % currently

    Gain: 0.35

    Reset: 436

    NL_GAP: 5

    NL_HYST: 5

    NL_MINMOD: 0.15

    NL_TBAND: 9

    2-LIC-107 Oil box level control

    Range: 1676 to 3048 mm

    Typical SP: 2600 mm

    Max/min for surge capacity: unknown, but think it is around 2400 (anything lower will push gas to next vessel) to around 2900 mm (may be able to go higher as the capillary pad is not at the top of the vessel, so there is some room)

    Values when we did the gap trial yesterday

    Gain: 3.33

    Reset: 553

    NL_GAP: 80

    NL_HYST: 80

    NL_MINMOD: 0.25

    NL_TBAND: 230

    I will post some trends of the oil box after we leave the gap tuning on for a day or so.  If you have any other suggestions on how to smooth it out more, please do share!  I was wondering if I should have 2-LIC-106 interface cascade  flow set point to        2-FIC-104A (newly created flow controller on interface zone water dump).  It is a vortex flow meter and is pretty reliable.  We also have a wedge meter on the oil box dump, but it is not very reliable (entrained gas etc.), would it still be worth while making it into an FIC for master/slave control?

    Also, could you explain how dynamic reset limit works (does it go on the master or slave loop?)

    And also "Use PV for BKCAL"  Is the what the slave controller should have?

    Thank you again! Appreciate the help so far, I am learning so much.

  • Lou Heavner

    In reply to controls_wl:

    I have not chimed in because you have gotten some pretty good advice. But I will alert you that when a level loop has been operating with an aggressively moved outlet valve, the valve can be physically worn out. When you go to tune the loop using the lambda method, you may find that the response of the valve is poor. Hopefully it will be good enough to go until the next time you can repair the valve. But in my experience, this is usually the limiting factor of performance when decide to lambda tune a level with or without a cascaded flow controller or with or without gap control.
  • controls_wl

    In reply to Lou Heavner:

    Thank you for the great tip! At this facility, it is most often run to fail. I have just got into APC and will definitely make sure the instrumentation is serviced or replaced before we tune or apply a new strategy. Would you recommend full valve rebuild (packing and trim/plug inspection) along with valve calibration and response testing?

    Jayme
  • Lou Heavner

    In reply to controls_wl:

    I'm not surprised your facility runs to fail... a lot of oil and gas producers run the same way. I wouldn't recommend anything without seeing more data. What you have may already be good enough. I was just warning you that control valves that are worn can be a limit on control capability. Make a 1/4% change in output. Then make a 1/2% change. Then a 1 % change. See how much change you have to make for the flow to change indicating the valve actually moved. Then reverse and go in the other direction.. See how much change is required to make the valve move. You can trend the controller output and flow on the same trend. If you have position feedback (readback), you can trend it, too. For me, usually a 1% valve is good enough and 1/2% or better is a pretty good valve. If it is much more than 1%, you will see the characteristic wave pattern when the loop is in auto. The output is a sawtooth wave and the flow (or actual valve position) is a square wave, both in a limit cycle. Ultimately, if you can stabilize the flow and maintain the level, you have satisfied the control objective.

    BTW, there may be economic incentive not to run to fail, and do some proactive maintenance. We have a group at Emerson that can provide consulting in that area. If you are interested, ask about Operational Certainty Consulting. They can help with a lot more than instrumentation and controls. They have expertise and FMEAs for almost every kind of equipment in almost every kind of process.

    Lou
  • MPHymel

    In reply to Lou Heavner:

    And while we are on the subject, stiction is often misdiagnosed as a tuning issue. Have you positioners on the valve, or just an ItoP transducer?
  • controls_wl

    In reply to MPHymel:

    It is a DVC, but we have a work order in to verify its accuracy. Thank you for the tip!

This is the official online community site of the Emerson Global Users Exchange, a forum for the free exchange of non-proprietary information among the global user community of all Emerson Automation Solution's products and services. Our goal is to improve the efficiency and use of automation systems and solutions employed at members’ facilities by sharing our knowledge, experiences, and application information.

User Groups | World Areas | Community Guidelines | Legal Information | Privacy Policy | Contact Community Manager

Website translation provided by Website Translator

© 2015-2022 Emerson Global Users Exchange. All rights reserved.