(A) study on design algorithm for optimal supervisor based on complete min-cut

Abstract

In this thesis we deal with a design method of the optimal supervisor based on complete min-cut. The objective is minimization of the net cost of the supervisor under the condition that the controlled plant must be a trim generator. The cost of control incurred when an event is disabled by the controller, and the penalty of control is incurred whenever states are reachable in the controlled plant. To attain our objective, we propose a design algorithm for the optimal supervisor. In addition we propose a procedure to design optimal supervisor for multiple model case. Moreover when robust supervisor exists for given multiple model, a cost and penalty assignment scheme with which we can design minimally restricted robust supervisor is proposed. As a case study, a resistance spot welding process is considered. It is multiple modelled under the assumption that it is not known whether the shunt effect exists or not. Then design the optimal supervisor with cost and penalty assignment.