(316e) Cyberinfrastructure for Supporting the Development of Computational Models for Enterprise-Wide Optimization

A multidisciplinary group at Carnegie Mellon University (Biegler, Hooker, Grossmann), Lehigh University (Linderoth) and the University of Pittsburgh (Schaeffer) composed of chemical engineers and operations researchers, have undertaken a new project in the area of Enterprise-wide Optimization (EWO). This project provides an example of a research collaboration that lies at the interface of Cyberinfrastructure with Process Systems Engineering and Operations Research. The major elements of the cyberinfrastructure include capabilities for integration of computational and IT tools, communication among geographically distributed groups, and access to grid computing. The goal of this project is to develop novel computational models for improving the operation of the petroleum and chemical industries. Due to increasing pressure for reducing costs and inventories, Enterprise-wide Optimization (EWO) has become the "holy grail" in these industries in order to remain competitive in the global marketplace. This view was for instance reinforced at the conference, ?Foundations of Computer-Aided Process Operations? that took place in Coral Springs in January 2003. The theme was ?A View to the Future Integration of R&D, Manufacturing and the Global Supply Chain.? It became clear that there is great interest among the process industries such as petroleum, chemical, pharmaceutical, consumer products, to achieve the goal of EWO (see http://www.cheme.cmu.edu/focapo).

EWO involves optimizing the operations of supply, manufacturing (batch or continuous) and distribution activities of a company. The major operational activities include planning, scheduling, real-time optimization and inventory control. A major focus in EWO is the scheduling of the manufacturing facilities, as well as their modeling at the proper level of detail, often with nonlinear process models. One of the key features in EWO is integration of the information and decision making among the various functions that comprise the supply chain of the company. Integration of information is being achieved with modern IT tools such as SAP and Oracle that allow the sharing and instantaneous flow of information along the various organizations in a company. These tools, which are largely transactional in nature, do not generally provide comprehensive decision making capabilities that account for complex trade-offs and interactions across the various functions, subsystems and levels of decision making. This means that companies are faced with the problem of deciding as to whether to develop their own in-house tools for integration, often with inadequate resources, or else make use of commercial software from vendors that often cannot be easily tailored to the application at hand.

In order to achieve EWO throughout the process industry our research effort is aimed at developing a new generation of computational tools that allow the full integration and large-scale solution of the optimization models, as well as the incorporation of accurate models for the manufacturing facilities. These tools that are to be deployed through a cyberinfrastructure and complemented with modern IT tools, will address the following items:

a) Modeling. Development of production planning and scheduling models for the various components of the supply chain, including nonlinear manufacturing processes that through integration can ultimately achieve enterprise-wide optimization. Major emphasis here is the development of novel mathematical programming and logic-based models that can be effectively integrated to capture the complexity of the various operations.

b) Multi-scale optimization. Strategies for coordinating the optimization of models over geographically distributed locations and for a given time horizon spanning from weeks to years. This includes long-term strategic decisions (years) related to sourcing and investment, medium-term decisions (months) related to tactical decisions on production planning and material flow, and short-term operational decisions (weeks, days) related to scheduling and control. The major emphasis is on novel decomposition procedures for large-scale computation that can effectively work across large spatial and temporal scales.

c) Optimization under Uncertainty. The goal here is to account for stochastic variations in order to effectively handle the effect of uncertainties (e.g. demands, yields, breakdowns). The major emphasis here is the development of novel and effective stochastic programming models and tools.

d) Algorithms and advanced computer architectures. In order to support the three points above, which require extensive computational capabilities, efficient algorithms will be developed that effectively exploit modern computer architectures. A major emphasis here will be the use grid computing for mixed-integer and stochastic optimization in collaboration with Jeff Linderoth's group at Lehigh University, which should provide a unique testbed for cyberinfrastructure.

We present some preliminary results to demonstrate how new analytical IT tools deployed in a cyberinfrastructure should help to realize the full potential of Enterprise-wide Optimization.