An MILP framework for batch reactive scheduling with limited discrete resources

This work introduces an efficient MILP mathematical framework for the reactive scheduling of resource-constrained multistage batch facilities. The approach is based on a continuous time domain representation that takes into account the schedule currently in progress, the updated information on old production batches still to be processed and new order arrivals, the present plant status and the actual availability of renewable discrete resources like processing units and manpower. Batch due dates and sequence-dependent changeovers can also be handled. The proposed technique is able to update the current schedule when unforeseen events like deviations in processing times, equipment breakdown or batch reprocessing occur. To avoid full-scale rescheduling, the approach allows partial modifications to the schedule in progress consisting of starting time shifting, limited resource reallocation and local batch reordering at any utilized resource item. The rescheduling algorithm is iteratively performed to restore feasibility at minimum increase of the selected performance measure. The make-span or the average order tardiness can be used as alternative objective functions. The methodology can also be applied to improve a non-optimal production schedule even if discrete resource constraints are to be considered. A large-scale case study involving unexpected equipment failures and manpower changes was successfully tackled. An efficient updated production schedule for the remaining time horizon was found at a reasonable CPU time.