Re: Applications for Small DMC controllers

Name    : 
E-Mail  : 
Subject : 

Comments:
: Jim, I'm also interested in how other people in the industry have used multivariable control (i.e. not just DMC) for smaller problems.

: I have applied multivariable control to a problem as small as a simple 1MV, 1 FFwd, and 2 CVs (one of these as a constraint variable).  The application was a simple heat exchanger/distillation/flash of phenol from BPA.  Although it was a simple application and I'm sure someone, more creative than me, could have easily installed a DCS based control scheme.  However, DMC was easily installed without much difficulty or retuning.  At the end-of-the-day, the controller worked well and plant rates were significantly increased.

: I guess their are several key points to warrant the application of multivariable control: 1) do you already possess a license for the software?, 2) Are the operation's & engineering folks already "comfortable" with the technology?, 3) Is the hardware in place to "run" the technology?.  In otherwords, if you have the technology in place why not use it! If the application passes these first three points, then I think that you then look for any application that is "interactive", potentially more constraints than degrees of freedom, and anything with significant dynamics.  

: I know we usually say that we don't want to "waste" our time applying multivariable control to simple problems, but in reality, the application may be warranted simply based on ease-of-maintenance.  I seen some of the "sophisticated PID" algorithms that can be quite difficult to maintain given the complexity of the task at hand.  Case in point, Charlie Cutler will tell you the reason he developed DMC was to get out of the "headache" of all the logic blocks.  DMC has the embedded LP that can do most of those checks for you.

: I think your question is a timely one in the sense that much of the focus of multivariable control has been on the "big" applications (e.g. FCCU , crude units, etc..).  Now that many of the bigger companies have "completed" many of the big guys, its time to go after the lesser players.  Even though the benefits may be less, hopefully the effort would be less as well (hopefully, sometimes the small ones can be just as difficult as the big ones?!?)

: My only word of warning would be to watch out for applications that involve "questionable" instrumentation (e.g. furnace louvers or dampers).  One thing PID does better than model predictive control is handling setpoints, or outputs, that don't respond repeatably.  That's why we spend so much time trying to get the regulatory system in the best configuration as possible (e.g. leaving flow controllers and temperature controllers in PID control).


Optional Link URL:  
Link Title:         
Optional Image URL: