Abstract - Pekny - 9402948 Industrial chemical processing success requires efficient management of capacity and the ability to quickly adapt to changes in the marketplace. A large number of constraints, the increasing scope of operations, time and efficiency pressures, and the combinatorial complexity implied by discrete decisions make scheduling and planning problems difficult to solve. This research will lay the foundation for developing process scheduling tools that are based on rigorous optimization methods, are robust to failure, easy to use in an industrial environment, can be quickly integrated with corporate databases, and can address very large and highly constrained scheduling and planning problems. Rigorous optimization methods provide an estimate of the best achievable performance available and, in the best case, they produce a schedule that is provably optimal for a given industrial scenario. The focus for this project is to provide the methodology and information necessary for understanding and controlling the critical components of process performance in a system that is sufficiently flexible to incorporate technological advances and that is designed to account for the fact that the scheduling and planning function will continue to evolve over time. The prototype system to be used is based on the premise that all such systems can be viewed in a modular fashion. The objectives of the research are: (1) development of a formulation specifically tuned to sequencing considerations that will complement existing formulations and be able to address a range of processes in considerable detail, (2) development of an efficient means to strengthen the formulation of scheduling problems through the introduction of strong inequalities derived from problem physics, (3) development of large scale integer programming solutions capability specific to scheduling and planning problems and sufficient to address long time scales, and (4) integration of symbolic reasoning and mathematical programming methods to reduce the training required to use a sophisticated system and mitigate the robustness problems inherent with exclusive use of optimization methods on combinatorial optimization problems.
