A cyclic approach to large-scale short-term planning in chemical batch production

In the chemical industry, final products arise from chemical and physical transformations of materials in processing units. We consider the case of batch production mode, where the total requirements for intermediate and final products are divided into individual batches. To produce a batch, the inputs are first loaded into a processing unit, then a transformation process is executed, and finally the output is unloaded from the processing unit. In general, storage facilities of limited capacity are available for stocking raw materials, intermediates, and final products. We present a novel cyclic approach to solving large-scale instances of the minimum-makespan production scheduling problem. This problem can be decomposed into a batching and a batch scheduling problem. The basic idea of the cyclic approach consists in reducing the size of the batch scheduling problem by computing a cyclic sub-schedule, which needs to be executed several times. Using a mixed-integer nonlinear programming formulation of the batching problem one can compute the set of batches of one cycle and the number of cycles needed to satisfy the primary requirements. The sub-schedule is then obtained by scheduling the batches on the processing units subject to material-availability and storage-capacity constraints. In an experimental performance analysis, we applied this cyclic approach to a set of 70 test instances. For each instance, we obtained a better feasible solution within much less CPU time than a state-of-the-art method from the literature.