A robust hierarchical production planning for a capacitated two-stage production system El-Houssaine Aghezzaf a, , Carles Sitompul a, , Frank Van den Broecke b, a Department of Industrial Management, Faculty of Engineering, Ghent University Technologiepark Zwijnaarde 903, B-9052 Zwijnaarde, Belgium b Production Planning Division, Agfa Gevaert NV, Septestraat 27, B-2640 Mortsel, Belgium article info Article history: Received 25 June 2010 Received in revised form 5 December 2010 Accepted 6 December 2010 Available online 13 December 2010 Keywords: Robust planning Hierarchical production Cyclical planning Periodic review abstract In this paper, we propose a robust hierarchical production planning approach for a two-stage real world capacitated production system operating in an uncertain environment. The first stage of the system pro- duces a set of semi-finished products having relatively stable annual demands, and the second finishing stage produces finished products having highly variable weekly demands. The fixed production setup costs incurred at the first stage are considerably high. Fixed production setup costs incurred at the second stage are fairly small compared to those of the first stage. We propose an integrated hierarchical planning model, where semi-finished products from the first stage (i.e. the aggregate level) are disaggregated into finished products to be produced in the second stage (i.e. the operational level). As a result of the rela- tively stable demands and the high setup costs experienced at the first stage, a cyclical aggregate plan- ning model is proposed for production planning at the upper level of the hierarchical plan. Based on this aggregate plan, a modified periodic review policy is then proposed for production planning at the lower level. Finally, a coupling plan, linking the two planning levels, is proposed to ensure the feasibility of the disaggregation process at every period. Ó 2010 Elsevier Ltd. All rights reserved. 1. Introduction The two-stage production system discussed in this paper is a real world production process having some particular features which, if taken into account in the planning process, may unequiv- ocally lead to effective production plans. The first stage of the pro- duction system produces some 200 different semi-finished products which are either shipped directly in this current status to other manufacturers or processed further in the second stage to produce some 10,000 different finished products. Therefore, pro- duction plans developed for the first stage can naturally be consid- ered as aggregate production plans (i.e. the aggregate level), which can subsequently be disaggregated to provide plans for the second stage (i.e. the operational level). The first special feature of the pro- duction system is that fixed production setup costs experienced at the first stage are significantly higher than those experienced at the second stage. This fact differs from the usual hierarchical pro- duction planning processes which consider that fixed costs at the aggregate level are negligible and take these costs into account only at the detailed level (i.e. the operational level). In addition, as it is often the case in almost all manufacturing systems, it is dif- ficult to accurately forecast finished product demands. This is what usually makes aggregate plans less effective in practice. The second special feature is related to the fact that demands of the semi-fin- ished products are relatively stable, due to the strong position of the manufacturers in the semi-finished product market. Moreover, the variability of the finished product demands can be approxi- mated by looking back into the past demand realizations. The main objective of this research is to develop a robust hier- archical production planning approach for the two-stage produc- tion system, which explicitly considers those special features during the planning process. The production planning approach also takes the variability of finished product demands into account, which consequently generates plans for the whole system that are robust (i.e. less sensitive to the demand variability). The proposed planning approach is based on the hierarchical production planning approach presented by Bitran and Hax (1977), which benefits are well-established in the literature (see also Bitran, Haas, & Hax, 1981, 1982). In a typical hierarchical production planning model, the objective is mainly to decompose a large and complex planning problem into less complex planning sub-problems resulting in consistent aggregate and master pro- duction schedules. In the production system at hand, the fact that demands of semi-finished products are relatively stable suggests that, even though demands of the finished products are random, planning at the level of semi-finished products may have some stabilizing effect on the aggregate production planning of the whole system. 0360-8352/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.cie.2010.12.005 Corresponding authors. E-mail address: elhoussaine.aghezzaf@ugent.be (E.-H. Aghezzaf). Computers & Industrial Engineering 60 (2011) 361–372 Contents lists available at ScienceDirect Computers & Industrial Engineering journal homepage: www.elsevier.com/locate/caie