A Heuristic for the Short-Term Planning of Multi- Product Continuous Plants

We consider the short-term production planning of multi-product continuous plants. In the literature this problem is generally modeled as a large-size monolithic mixed-integer linear or nonlinear program. In this paper we follow a decomposition approach which partitions the problem into an operations planning and an operations scheduling problem. The operations planning problem consists in assigning the production tasks to the processing units, fixing the operating conditions of the tasks, and computing the number of operations executed for each task. This problem can be formulated as a nonlinear program of moderate size containing only one integer variable. The solution to the operations planning problem provides a set of operations, which have to be scheduled on the processing units. For this operations scheduling problem we present a novel mixedinteger linear programming formulation as well as a fast priority-rule based scheduling method. In contrast to the classical monolithic approaches, our new method is able to find good feasible schedules for large-scale instances within less than five seconds.