Dealing with structural constraints in self-optimizing control engineering

This paper deals with new design methods for self-optimizing control structures, i.e., the selection of controlled variables which guarantee almost optimal operation. Two types of structural solutions are considered. The first type has been previously addressed by other authors [1], [2], [3] and [4] and refers to structures where all controlled variables are linear combinations of the same process variable set. In the practically more relevant second type, each controlled variable consists of a linear combination of an individual process variable set. Design methods for both types are developed. They are compared with respect to optimality and computational performance by means of repeated random tests. For further illustration, the newly developed methods are applied to a previously considered practical example from the academic field [5] and [3].