A hierarchical methodology is developed for the overall design of manufacturing systems. The methodology consists of solutions to four levels of problems, namely, (1) manufacturing system selection, (2) shop loading, (3) machine loading and tool allocation, and (4) testing the feasibility of a schedule and determining strategies for the operational control of the system. Although these problem levels are developed in a hierarchical sense, they can be applied independently by assuming appropriate inputs to the problem level under consideration. The third and the fourth level problems are addressed in this research for the flexible manufacturing system. The first level of the hierarchical methodology addresses the problem of manufacturing system selection. The mathematical 4 model formulated for this problem captures the basic and integrated relationships among the systems and system components. This model provides a practical approach and a precise tool to determine an optimal mix of systems, to assign appropriate machines to each system, and to select the best material handling system for each system to best suit long-term production requirements at minimum costs. The second level of the hierarchical methodology addresses the shop loading problem. A mathematical model is developed for partitioning parts among the manufacturing systems selected at the first level to minimize total operating costs. For the third level problem, a mathematical model is formulated to obtain routings of parts through an FMS and to assign appropriate cutting tools to each machine in the system to minimize total machining cost. For the fourth level problem, a simulation model is developed for testing the feasibility of the solution obtained at the third level. It also helps to determine strategies for the operational control of the system. The computational experience with the mathematical models is presented using the MPSX-MIP/370 package. Sensitivity analysis is also performed to further understand system behavior under various operating conditions. Several new findings of the research are reported. Because of the special structure of the mathematical models, a computational refinement for their solution is also proposed based on Lagrangian relaxation. / Ph. D.