Page 54 - Practical Control Engineering a Guide for Engineers, Managers, and Practitioners
P. 54

Introduction  to  Developing  Control  Algorithms   29




                                                          ~
                                           r----• Success          ,.
                                             Tune
                                                                  ,-
              Develop
               process



             F1caURE 2·9  The benefits of separation: interaction, evolution, synergism, and
             problem isolation.



        2-4  Some General Comments about
              Debugging Control Algorithms
             This is a sore subject with a lot of engineers, yours truly included.
             Perhaps it's best to simply tell a couple of war stories.
             Rookie Fright
             I joined a large manufacturing company with only a couple of years
             of experience after leaving graduate school with a Ph.D. in chemical
             engineering. Although I had been into a  plethora of hobbies and
             projects before college, my life as a professional student had been
             reclusive and narrow and I had no hands-on engineering experi-
             ence, nor much interest in gaining any-hey, I was an applied math-
             ematician (I thought)!
             When in Doubt, Simulate-Not!
             Given the comments of the above paragraph, I really was good for
             little other than generating sophisticated mathematical simulations.
             At my previous job I had been adept at making mathematical models
             of  complicated  processes,  cooking  up  complicated  algorithms  for
             controlling the model,  and  using the process model to  show how
             wonderfully the control algorithm would work.
                The rationale for using simulation to develop a control algorithm
             is simple and, in my opinion, quite incorrect. This approach is flawed
             because the model basically contains the knowledge of the modeler
             and little else. When it is put through its paces, it will surprise no one.
             Furthermore, the model will likely not contain any of the subtle idio-
             syncrasies of the real process-idiosyncrasies that might defeat the
             control algorithm developed by using the model.
                The motivation for using a mathematical model often comes from
             a manager who has little actual knowledge of mathematical model-
             ing. He has, however, observed that suggesting mathematical model-
             ing as a solution to some difficult process problem often comes across
             as an enlightened commandment.
   49   50   51   52   53   54   55   56   57   58   59