Scheduling is a critical decision-making procedure in chemical process industries. Its vital role in the performance of the whole supply chain network is strongly emphasized by a large number of authors. Chemical process industries are dynamic in nature and, therefore, different kinds of unexpected events, such as equipments breakdowns, changes in processing and/or setup times, inadequacy of raw materials, due date modifications, orders cancellations and prices changes, occur quite frequently. Underestimating uncertainty and its impact can lead to decisions that neither safeguard a company against threats, nor take advantage of eventual opportunities that higher levels of uncertainty may provide [1]. Therefore, deterministic studies considering nominal plant operating conditions and production capacity during the whole time horizon are not the most appropriate to deal with the real scenario.
Nowadays, two approaches are mainly used to deal with uncertainty in process industry environments: proactive and reactive scheduling. Proactive scheduling allows to fully exploit the flexibility of the process and, consequently to meet production goals to a higher degree. However, it requires information to characterize the uncertainty and incorporate it into the scheduling process. Additionally, it disregards the possibility to react to new information in the future, thus reducing the optimization capabilities [2]. As a result, if the revealed uncertainty is not taken into account in the initial representation then significant disturbances and infeasibilities may appear in scheduling process thus affecting the performance and the effectiveness of the supply chain entity. Otherwise, reactive scheduling modifies the predictive schedule to adjust it to changes to production environment. Uncertainty is dealt on-line and scheduling decision-making is based on updated plant information; however, this is not always possible. Moreover, it entails an additional computing cost, and the use of sophisticated optimization methods which are still limited.
Most of the literature considers only one of the aforementioned approaches at a time, and little research has focused on their integration. The aim of this work is to encompass both methods and compare their effectiveness to deal with uncertainty thus suggesting an hybrid proactive-reactive approach.
A reactive scheduling model is adopted based on a MILP continuous-time formulation from previous contributions [3]. Only limited changes into the proactive schedule are allowed during the rescheduling process rather than a time-consuming model involving full-scale rescheduling. Afterwards, the former model was appropriately modified into an equivalent proactive scheduling model. The two models are compared through several case studies in order to shed light on the specific features of each one of them. An evaluation criterion based on economical issues has been established for comparing the effectiveness of both approaches. As far as it concerns the reactive approach, a rescheduling cost is also considered.
In summary, taking into account the above considerations and comparison results it seems challenging to study the development of a new scheduling approach which contains the advantages of these two different scheduling approaches to deal with uncertainty. The integration results into a new scheduling model that tackles unforeseen events more efficiently. Moreover, this model is compared with the aforementioned scheduling approaches in order to highlight its particular characteristics. This novel approach is proposed as an efficient method for dealing with the uncertainty in order to improve the decision-making process in the scheduling of chemical multiproduct batch industries.
Acknowledgements
Financial support received from "Ministerio de Educación y Ciencia" (FPU grants), Universitat Politécnica de Catalunya (UPC research grant) and European Community (project PRISM-MRTN-CT-2004-512233) is fully appreciated. Besides, financial support from Generalitat de Catalunya (project no. I0898), and MEC (DPI2006-05673) projects is gratefully acknowledged.
References
[1] Gupta A. and Maranas C., (2003) Computers and Chemical Engineering, 27, 1219-1227. [2] Engell S., Markert A., Sand G., Schultz R., Schulz C., (2001), Eds Springer, Berlin, Ch. Online optimization of large scale systems, 649-676. [3] Méndez C., Cerdá J., (2004) Computers and Chemical Engineering, 28, 1059-1068.