OVERVIEW

Self-tuning, or auto-tuning as it may also be called, is a technology that has made rapid strides in the past few years. Although the theoretical concepts underlying it have been around for some time, only recently has computing power become inexpensive enough for calculationintensive self-tuning techniques to be economically viable in commercial products. Today, practically all distributed control systems and microprocessor-based single-loop controllers, even those aimed at the low end of the marketplace, include some type of self-tuning.

There are several reasons to use self-tuning. If the process is non-linear that is, if it does not exhibit the same response at one operating point as another and if the process is operated under widely varying conditions, then the controller's tuning parameters should be changed to match operating conditions. This change could be accomplished by using an operator's log that tabulates various combinations of tuning parameters for different operating conditions. This method assumes that each operating condition can a priori be associated with a particular combination of tuning parameters. It also depends on an operator or instrument technician's diligence in entering the appropriate tuning parameters as operating condition changes. However, if process characteristics change rapidly, or cannot be categorized from simple measured data, it probably is unreasonable to expect frequent manual changes to the tuning parameters to produce satisfactory results.

An entirely different motivation for using self-tuning is the desire to have a procedure that will determine an acceptable set of tuning parameters automatically during startups. This would minimize the task of...