Page 58 - Practical Control Engineering a Guide for Engineers, Managers, and Practitioners

Introduction to Developing Control Algorithms 33

the reader's head about now.) With few exceptions, the linking never
went well and we would frequently destroy an expensive blank.
To address this, I hooked up a PC to the analog inputs and outputs
of the minicomputer on the research facility's floor and used the PC to
crudely simulate the process. I think I did this with an early version of
Quick BASIC. With this "process simulator" in place, we could test the
linking without worrying about destroying a blank. Again, the process
model in the PC was extremely crude but sufficient.
So, any algorithm that I cooked up went through a test simulation on
my desktop, a test simulation using the on-the-floor process simulator,
and finally, a test on the real process. We ruined no more blanks. It is
important to note that the simulation test systems, either on my desk or
on the floor, were not particularly sophisticated. I simply wanted to test
the control algorithm on a rough approximation of the process to see if,
at least, the algorithm was functionally correct. The process (and only the
process) itself would tell us if the algorithm was any good at control.
We ended up taking the debugged control system to a manufac-
turing plant where it was installed without any problems-no SMILH
required.
It's Too Complicated-Use the Process for Debugging
This topic is a sore point with many engineers, especially moi. Some
of my best friends, people who I greatly respect, are staunch support-
ers of the idea that in some cases one just has to use the process to test
out the control system. The decision to use the process to test a con-
trol system has taken place repeatedly during my career. Each time
I've seen the engineering online debugging costs (never anticipated)
overwhelmingly outstrip any cost that would be required to put together
a test system that simulates in some way the process to which the con-
trol system is to be applied.
Whether to simulate or not is a question that will not be settled
here. However, I will present my point of view, unpopular though it
is. Start with the big picture; the process and the control system with
all the interconnections (Fig. 2-10). Let's assume that there are many

FIGURE 2·1.0 Interactions between process and control system.