EE - Forum Styles

Tuning Issue

Were using deltav tune / Deltav adapt for the first time to tune a temperature loop into a gas fire heater and the results are very disappointing.

the process gain is around 1.2, time constant about 800 seconds, dead-time around 200 seconds.

We have approved a adapt model and running the PID in partial mode.

using a tuning method of PID, and Lambda values of between 1.25 and 0.4, the response is just too slow. 

When the heater starts up, the temp set-point is ramped from ambient (20°C) to 270°C, the temperature gets to around 250 then gets to set-point over a few hours which is too slow and the heater only runs for a couple of hours each batch. we need to get to setpoint ASAP with minimum overshoot.

Any ideas what im doing wrong?

7 Replies

  • Hey Stuart!
    The dead time and time constant seem a bit long for the "typical" fired heater. What is the model quality? Does the temperature loop output go straight to the valve or is it cascade to a fuel flow or pressure loop? If it goes straight to the valve, and the valve has significant dead band (amount of signal reversal required to reverse the valve closure member), the process gain identified by Process Learning may off.

    Also, it would be helpful to see the output of the controller in your plot.

  • In reply to James Beall:

    Hi James, thanks for your continued support.

    model quality is between 50 - 60. the temp does go to a ratio block, then cascades to fuel flow controller and combustion air flow controllers. these do have good control.

    dont currently have copies of trends but could get copies tomorrow if they will be useful. the pv is following the output signal

    this loop was controlling good pre our deltav upgrade. It looks like we had different tuning values in our controller then what was in the upgraded database.

    the slides you sent were great, many thanks. would be great if i had a recording your of presentation to go with the slides, to hear the detail which goes with the slides.

  • In reply to Stuart Jolley:

    Thanks for the OUT plot.

    Now that I've read your note in more detail, I see that the heater is operated in a batch mode. Thus, the State Variable should probably be PV with models at various temperatures from cold to hot. What State Variable is being used?

    Also, the "Lambda" tuning method in the Adaptive Tuning is the most simple of the 11 different process models and tuning provided in the full Lambda Tuning suite provided with the InSight "Entech Toolkit" add on. The Entech Toolkit would likely identify this as a second order self-regulating process, suitable for Derivative action based on the measure 2nd order lag. In Adaptive Tuning, the IMC-PID tuning is similar to Lambda tuning but will bas the derivative on rule of thumb without identifying the second order lag. So, you will get derivative but if the model is not accurate, this tuning will likely not be satisfactory either.

    We can arrange WebEx tutorial on InSight! Please have your Emerson representative reach out to me to set it up with you and your team.
  • In reply to James Beall:

    state variable is PV.

    Were not using the Entech Toolkit.

    I will ask our Emerson rep to contact you.

    Im not sure where i go from here in terms of tuning. I cant leave the process as it is but dont seem to have many other options for tuning.
  • In reply to Stuart Jolley:

    Although I think ratio control is a powerful and underused control scheme, it can have drawbacks if the independent variable can go to (or start as) as a small value. The initial heat requirement at low temperature and low feed rate is that required to heat the heater mass. Thus, the gain can actually change significantly. We would need to know more about the system but I believe our team could help with this application. I think we could help remotely to lower cost of support and provide a quicker response! We can talk more about this!
  • In reply to James Beall:

    I am curious what the feed flow through the heater is doing during the heat up. If you can, send another (sorry!) plot with all of the above plus the feed flow plus the SP for the TIC.