SCHEDULING OF MULTIPURPOSE BATCH PLANTS WITH DIFFERENT STORAGE POLICIES: A COMPARATIVE STUDY BETWEEN S-GRAPH AND MILP METHODS

Batch plants are the best option to manufacture relatively small quantities of speciality chemicals, pharmaceuticals and agrochemicals high-value products with variable demand patterns. The most general case corresponds to the so-called multipurpose batch plants where a wide variety of products with different processing recipes can be produced by sharing flexible processing equipment capable of performing different batch operations. Furthermore, intermediate storage is usually employed to increase the productivity and expand the plant availability (Graells et al., 1996). In order to take full advantage of this inherent operational flexibility, highly efficient optimization methods are needed to deal rigorously with industrial-scale short-term production scheduling problems. Different MILP-based scheduling models have been proposed to manage the complexity of multipurpose batch processes. MILP formulations provide a rigorous mathematical representation of the problem features that can subsequently be solved through different solvers, commonly LP-based branch and bound methods. Méndez and Cerdá (2003) developed a continuous-time MILP formulation that relies on the notion of general precedence. This model is able to define a single binary variable to sequence processing tasks as well as related storage operations without losing optimality. On the other hand, alternative solution methods have emerged over the last years. This is the case of the S-graph framewor