A real-time expert control system for dynamical processes

An approach for designing expert systems suitable for engineering process control is proposed. It is based on shallow-coupling systems in which the selection of a control routine is determined by observing the state variables of the process. Relevant technical issues for representing and processing the appropriate knowledge are addressed. The method has the advantage that it is real-time; that is, it is able to act almost instantaneously since it is based on observing the state variables. However, a disadvantage is that the information on which the action is based may not be sufficient. In general there should be a compromise between the amount of information necessary to make a decision and the time spent in collecting this information. It is found that the method is very suitable for nonlinear problems in which the process goes from linear to nonlinear, dead zone, or saturation through its trajectory, and a set of controllers must be appropriately chosen for each case. Implementation for a simulated continuous reactor has demonstrated the effectiveness of the method.<
>