With chemical shipments reaching $506.5 billion in 2004, the chemical industry annually contributes 2.1% of the US GDP, being the largest manufacturing sector of the US economy on a value added basis. Over the last decade, however, global competition has increased resulting in a trade deficit of $5,064 million for the first time in 2002. At the same time, profit margins and capital spending of chemical companies have decreased, while regulatory, economic and trade policies are not always favorable. To remain healthy and viable in today's competitive environment chemical companies must maintain their technological edge (by increasing R&D investments) and achieve enterprise-wide optimal solutions using advanced optimization methods. At the tactical level, specifically, it is important to integrate medium-term planning decisions with short-term operational decisions. Despite the improvements in computer hardware and optimization software, production planning and scheduling problems remain notoriously difficult to solve and the state of the art is insufficient to meet industrial needs.

A general overview of the problems in process operations will be presented first, followed by a survey of existing methods for planning and scheduling problems. A novel continuous-time Mixed-Integer Linear Programming (MILP) model will be presented next. The proposed model accounts for resource (utility) constraints, variable batch sizes and processing times, batch mixing/splitting and sequence-dependent changeover times. A hybrid Mathematical/Constraint Programming method will also be presented. The main idea is to decompose the original scheduling problem into a MILP master problem and a Constraint Programming (CP) subproblem. The decisions about the type and number of tasks performed, as well as the assignment of units to tasks are made by the MILP master problem, while the CP subproblem checks the feasibility of a specific assignment. Numerical results show that the hybrid method is substantially faster than standalone MILP and CP models, enabling us to solve problems that were unsolvable with existing methods.

Finally, we will present an attainable region approach for the effective solution of production planning problems. In this approach, we analyze off-line a detailed scheduling model to obtain a convex approximation of the feasible production targets, i.e. the attainable region of each process network. This attainable region is expressed via simple inequalities that involve only planning variables, lending itself to effective integration with production planning formulations. The proposed description provides all the relevant scheduling information necessary to solve the planning problem with high quality, without involving the complicating (binary) scheduling variables.

If you'd like to meet with Prof. Maravelias during his visit, please contact Meredith Leach (x4-2655 or leach at egr.uri.edu).