Predictive control of flexible manufacturing systems

Consideration is given to the dynamic scheduling problem in flexible manufacturing systems. It is assumed that a medium-scale flexible manufacturing system consisting of three to eight machines with overlapping functionality but possibly different effectivity is to be run under the JIT (just-in-time) philosophy. This means that all jobs have to meet their due dates but the starting times of the processing (due dates for delivered parts or previous stages of the production) should be as late as possible. Dynamic online scheduling means that the scheduling algorithms must react to the distances in the production process within short periods of time, typically some seconds to a few minutes. Under this constrain, an exact optimization cannot be achieved because of the enormous complexity of the problem. A mathematical description of the system is presented, and possible approaches to the generation and improvement of suboptimal schedules based on priority rules, and the prediction of the behavior of the system under the regime of such rules, are discussed.<>