A matheuristic approach for the design of multiproduct batch plants with parallel production lines

Batch processes are typically used to manufacture, among other, specialty and fine chemicals. As the construction of grassroot batch plants requires major investments, models for determining the optimal design of such plants have been developed over the past decades. These models are often formulated as Mixed Integer Linear Programming (MILP) models which are solved exactly. In a previous study, we introduced the concept of parallel production lines as a design option into existing mathematical plant design models. The design problem now also aims at optimising the number of production lines, their design and the allocation of products (and production quantities) to the installed lines. However, with this extension, the complexity increases significantly. To tackle this combinatorial divergence, we formulated a matheuristic solution approach which combines an iterated local search metaheuristic with exact MILP calculations. In this paper, the hybrid solution method is described and its performance, in comparison to an exact algorithm, is illustrated for several example problems. It was found that our matheuristic obtained very good solutions in significant lower computation time. As a consequence, this technique is suitable to solve more realistic instances and enables us to expand these design models with e.g. different objectives in the future.