The application of simulated annealing to an industrial scheduling problem

Construction of factory schedules using reverse simulation

“…The first is the case study used in the papers of both Reynolds et al [4,5] and Charalambous et al [10,9]. The three stage plant consists of two first stage processing units, five storage units and five final stage processing units.…”

“…Tasks in the earlier stages of the plant may make full use of any gaps that are present in the partially constructed schedule. This contrasts with the approach of Reynolds et al [4,5], where the plant simulation progresses strictly forward in time, eliminating the possibility of utilizing such gaps.…”

“…Techniques applied to this problem include mixed integer linear and non-linear programming [6][7][8] and metaheuristics including simulated annealing [4,5,9] and genetic algorithms [10]. In all of the approaches that use metaheuristics, the metaheuristics create a sequence or sequences of jobs.…”

“…Reynolds et al [4,5] do not split the demand for each product in this way. The total amount of product assigned to a final stage unit is considered to be a job.…”

“…Storage units must be modelled explicitly. Earlier research led to the production of an approach combining simulated annealing with a simulation of the plant [4,5] that produced high quality results for some factories. In this paper we improve on these results by replacing the plant simulation with a schedule constructor that uses reverse simulations to determine how to schedule individual jobs.…”