FLOWSHOP SEQUENCING WITH NON-PERMUTATION SCHEDULES

The general belief is that near-optimal solutions can be obtained for the scheduling of serial multiproduct batch processes under the assumption of a permutation schedule (PS), which limits the search space. In this paper, we present a study of the nature of the serial multiproduct batch scheduling problem and provide justifications for considering serial flowshop scheduling under a non-permutation schedule (NPS). The problem addressed involves minimization of makespan for a given time matrix with unlimited intermediate storage between units. A comparison is made between PS and NPS solutions. NPS solutions are obtained by enumerative search for small problems and by simulated annealing (SA) for large problems. It is observed that the quality of an NPS solution compared to a PS solution is dependent on the range of processing times and also on the size of the problem. For most problems, the use of an NPS is necessary to obtain optimal or near-optimal solutions. An extension of the rapid access extensive search (RAES) heuristic, which we call generalized rapid access extensive search (GRAES), is also developed for obtaining an NPS solution. GRAES is shown to be a good technique for fast computation of NPS solutions, although the quality of solutions obtained by SA is better.