Any changes / updates to MPC Pro in recent patches e.g. WS_21 / CTRL_06 (MQ)?

Our site recently deployed WS_21 patch to iFix as well as CTRL_06 patch to the controllers. Ops called last night befuddled by MPC-Pro controller on waste heat boiler - didn't seem to be behaving intuitively.

The "feedback" behavior (versus prediction) looks a little muted - actual temperature was low and the controller was still cutting fuel.

Also, the optimizer does not appear to be honoring the SP low limit or the ± SP range as configured.

Anyone ( ?) have any inside knowledge of possible fixes / enhancements that may have been applied? Just finished updating model so we're watching how it goes.

Thanks,

John

5 Replies

  • John, did you get any feedback on this issue? I recently upgraded to version 14LTS, applied the WS_21 and CTRL_06 patches and now my MPCs are "broken". They are absolutely doing the wrong thing. Have unrealistic target setpoints, minimizing MVs when they should be maximizing. These worked perfectly prior to the upgrade. We have gone through a version 12.3.1 to version 13.3.1 upgrade a few years ago with no issues. Currently have submitted a GSC call, but I would be curious if you are still having issues and if other users have experienced similar issues at version 14.
  • In reply to Jerry Harris:

    So far no feedback. My troubles are not as severe, but still can't explain why it doesn't obey SP limits etc. Did you try adjusting the optimizer targets? I should say I also repeated step tests and generated new models for the controller in question, which was performance related. It's the only one where we use the optimizer at the moment, which I'd have to say is following our objectives and priorities, but while violating e.g. setpoint high / low limits.

    Please share what you learn from the GSC if you can, thanks.

  • John,

    There may be a known problem, but I am not aware of it. There are a few things you might check, but I'll see if one of my in-house colleagues has seen this. Make sure your penalties weren't somehow changed. Make sure your constraint limits weren't changed and make sure some unexpected constraint isn't forcing the controller to act differently than you expect. If you are using MPCPro instead of MPCPlus, it is good practice to use shadow constraints. MV limits are always honored by the MPC, but the optimizer in MPCPro doesn't consider if the MV is limited or not. So if you configure a constraint variable that is the same as an MV (you should use SP, but you could use PV, of a flow loop PID, for example) and the same constraint limits as the MV and with a gain of 1 and DT and time constants of 0 for that MV pairing and a gain of 0 for all others, the optimizer will find a solution recognizing the constraint variable limits.

  • In reply to John Rezabek:

    I have tried adjusting some of the optimizer values, but it didn't help. Working with Todd Maras out of Austin right now. I have sent him my files and he is trying to duplicate the issue. I have two MPCs where we try to maximize feed rates against various constraints and they both "broke" after the upgrade. They are calculating "crazy" target setpoints for the constraint variables and thus driving my feed rates the wrong way. These worked perfectly for years (one since 2014 and one since 2016) until last week when we upgraded. I will share what I learn and if/when we fix it.
  • In reply to Lou Heavner:

    Lou, one of mine that started acting up is configured exactly as you mention using the shadow constraints (James Beall helped me with it). It has worked great since 2014, but it started acting up after my upgrade last week. Something happened, not sure what. Hopefully Todd Maras can recreate.