Multiobjective Evolutionary Optimization of Batch Process Scheduling Under Environmental and Economic Concerns Elisabet Capo ´n-Garcı ´a Dept. of Chemistry and Applied Biosciences, ETH Zu ¨rich, Zu ¨rich 8093, Switzerland Aaro ´n D. Bojarski, Antonio Espun ˜ a, and Luis Puigjaner Dept. of Chemical Engineering, CEPIMA, Universitat Polite `cnica de Catalunya, ETSEIB, Barcelona 08028, Spain DOI 10.1002/aic.13841 Published online in Wiley Online Library (wileyonlinelibrary.com). The simultaneous consideration of economic and environmental objectives in batch production scheduling is today a subject of major concern. However, it constitutes a complex problem whose solution necessarily entails production trade-offs. Unfortunately, a rigorous multiobjective optimization approach to solve this kind of problem often implies high computational effort and time, which seriously undermine its applicability to day-to-day operation in industrial practice. Hence, this work presents a hybrid optimization strategy based on rigorous local search and genetic algorithm to efficiently deal with industrial scale batch scheduling problems. Thus, a deeper insight into the combined environmental and economic issues when considering the trade-offs of adopting a particular schedule is provided. The proposed methodology is applied to a case study concerning a multiproduct acrylic fiber production plant, where product changeovers influence the problem results. The proposed strategy stands for a marked improvement in effectively incorporating multiobjective optimization in short-term plant operation. V V C 2012 American Institute of Chemical Engineers AIChE J, 00: 000–000, 2012 Keywords: LCA, multiobjective optimization, evolutionary algorithms, multiproduct batch scheduling, product changeover, sequence dependency Introduction Tighter economic demands and increasingly stringent safety and environmental policies in process industries require more sophisticated decision-making tools. Thus, shorter periods of time for decision-making are imposed by market and process uncertainty affecting all enterprise levels. Specifically, the scheduling function, which is related to the organization of the human and other resources used in a company and directly linked to the satisfaction of customer demands, significantly affects the company results. Environmental, safety, and economic concerns are the ba- sis for most of the criteria that can be used to analyze the scheduling problem. In practice, the decision maker often faces the situation of choosing a given solution under differ- ent and conflicting objectives such as cost, performance, reli- ability, safety and productivity among others. 1 The absence of a universal and unique criteria to establish the goodness of a given schedule increases the complexity of the decision- making process, because a large number of solutions may be suitable. Therefore, the goal of management is frequently multifold and prefers to be guided through a set of good alternatives available rather than a single ‘‘best one’’ pro- posal. The analysis of the decision maker’s alternatives under conflicting objectives is performed by means of multi- criteria decision analysis. Therefore, efficient multiobjective optimization strategies, which can generate different alterna- tives for decision analysis in production scheduling, stand a great chance of adding competitive advantage to business. 2 In the literature, several authors present alternative meth- odologies to account for engineering, economic, and environ- mental trade-offs arising in batch process scheduling. How- ever, most of existing contributions are problem oriented. Two different approaches are usually adopted, namely, (1) techniques based on mathematical programming and (2) techniques based on heuristics and metaheuristics. Mathematical programming approaches are extensively used for solving scheduling problems; remarkable examples are reviewed by Mendez et al. 3 The most important feature provided by these approaches is that optimal solutions can be proved global under certain problem circumstances (e.g., linearity, convexity, casting the problem into a MILP). How- ever, the combinatorial nature of the scheduling problem poses serious computational difficulties when using mathe- matical rigorous approaches. Therefore, it is widely recog- nized that full-space mathematical programming approaches cannot tackle long scheduling time horizons or with many batches to be assigned to many resources, that is, problems of important size. 4 Even more, the integration of multiobjec- tive issues at the operational level increases the problem complexity. Therefore, the generation of multiobjective opti- mal solutions for scheduling problems, which entail almost immediate results for day-to-day decisions, can only be Additional Supporting Information may be found in the online version of this article. Correspondence concerning this article should be addressed to L. Puigjaner at luis.puigjaner@upc.edu. V V C 2012 American Institute of Chemical Engineers AIChE Journal 1 2012 Vol. 00, No. 0