Process flexibility design in heterogeneous and unbalanced networks: A stochastic programming approach

Most studies of process flexibility design have focused on homogeneous networks, whereas production systems in practice usually differ in many aspects, such as plant efficiency and product profitability. This research investigates the impacts of two dimensions of production system heterogeneity, plant uniformity and product similarity, on process flexibility design in unbalanced networks, where the numbers of plants and products are not equal. We model the design of flexible process structures under uncertain market demand as a two-stage stochastic programming problem and solve it by applying Benders decomposition with a set of acceleration techniques. To overcome slow convergence of the exact algorithm, we also develop an efficient optimization-based heuristic capable of obtaining solutions with optimality gaps less than 6% on average for realistic-scale production systems (e.g., with five plants and 10 types of products). Numerical results using the proposed heuristic show that flexibility designs are influenced by both dimensions of system heterogeneity, though the desired level of flexibility is more sensitive to the effect of plant uniformity than that of product similarity.