Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Management (Business Administration)

First Advisor

Dan B. Rinks


Globalization of manufacturing along with increased competition has made effective planning and control more important than ever. At the same time, it is more difficult than ever to achieve effective planning and control due to larger leadtimes and shorter product life cycles. The objective of this research is to explore the importance of control strategy on materials management in global manufacturing networks. Control strategies in common use and others that have recently been proposed in the literature are reviewed and classified along a push/pull gradient. It is shown that one of them, the restoration control strategy, can be used to represent a wide range of pull systems as well as certain elements of push systems. Using concepts underlying the restoration strategy, two models are developed for aggregate planning in a global manufacturing network. One model requires that all demands be met whereas the other allows some sales to be lost. Application of either of the models to a specific network results in values for decision variables, including target inventories and restoration coefficients. Target inventories are aggregate values that can be disaggregated to finer levels of detail. Values for restoration coefficients help identify the best control strategy. Both models apply to multi-echelon networks of any design and under known demand. Both formulations are nonlinear, mixed-integer programming models that have proven to be difficult to solve for the general case. Relaxing the integrality constraints allows the models to be solved using commercially available software although optimality cannot be guaranteed due to nonconvexity of constraints. The models were applied to a specific network. The restoration model with no lost sales was found to have severe limitations; however, the restoration model that allows lost sales provided results that were stable. The relationships between the decision variables and holding costs, labor costs, and demand variation were explored using the simulation technique of batch means. Among other things, results indicated that a control strategy very similar to base stock was most appropriate for the specific network studied.