This work deals with the introduction of heuristics on the design of digital control systems, so as to combine rule-based techniques with conventional control methods. The heuristics is implemented by means of a rulebased supervisor designed from the expert knowledge about the process. From the expert knowledge about the process it is possible to divide the process operation into a set of so called ‘states’. These ‘states’ may include for instance different operating points in which the system can be considered linear or zones where the system is operating under some type of nonlinearity. For each particular state it is possible to select and tune an appropriate control algorithm (e.g. a linear control algorithm for the former states and fuzzy, neural or rule-based controller for the zone with nonlinear-specific behaviour). The goal of the supervisor, which is executed at each sampling period, is then to track the system operation, detect a system state change and switch to the control algorithm associated to the state reached. A methodology for the design of such type of supervisors is presented. It addresses how the information acquired from the process has to be organized in a simple manner in order to be used to track process transitions between states. In particular, the idea of detector modules, whose function is to establish the tendency of the relevant process signals, is proposed. This information is combined in what is called ‘propositions’. Particular combinations of such propositions control the transitions between states. Finally, the complete state transition diagram is implemented as a set of rules of type if-then-else.
Read full abstract