TABLE OF CONTENTS The real world presents many problems to the control system engineer. For example:
Processes may not be well behaved; that is, they may have an inverse response or a long dead time. Hence, they may not be amenable to approximation as first-order plus dead time.
The processes may consist of multiple-input, multiple-output (MIMO) interacting control loops.
Rigorous dynamic models suitable for control purposes may not be tractable.
The process may be subject to random disturbances.
Several forms of constraints may be present:
on process variables,
on manipulated variables,
on auxiliary variables,
on rate-of-change of variables,
on combinations of variables.
The constraints may vary with time or with operating conditions.
No feasible solution may be available that will satisfy all the constraints simultaneously.
Processes are not necessarily "square"; they may be "fat" or "thin," where
Square means same number of outputs as inputs (zero degrees of freedom),
Fat means more inputs than outputs (excess degrees of freedom; this usually leads to an optimization opportunity),
Thin means more outputs than inputs (insufficient degrees of freedom; all of the outputs cannot be controlled simultaneously),
A valve going into saturation, a controller being placed in manual, or an analyzer failure may transpose a square or fat system into a thin system.
Many of these problems were discussed in earlier chapters. For example, feedforward control was applied to handle measurable disturbances. Override control was applied to handle constraints, though we did not note that situations may arise for which there is no feasible...
