The main objectives of this research are twofold: (1) a synthesis of descriptive, normative, and prescriptive aspects of decision making and (2) a synthesis of decision theoretical approaches for decision aiding and artificial intelligence techniques for problem solving. The result of this research is expected to be a computational theory of decision making for developing automated decision tools. The application domain under consideration is the scheduling problem in manufacturing.

The traditional approach to manufacturing control systems is the so-called push strategy, such as material requirement planning (MRP). The Japanese alternative is the ``pull'' strategy in which the material supply is just-in-time (JIT). It is becoming evident that neither one is consistently better than the other. In fact, in many instances, a hybrid approach is more superior, depending upon the manufacturing system. However, it is not known which characteristics of the manufacturing system affect the performance of different strategies. The main objectives of this research are to identify the factors affecting the performance of the manufacturing control strategy and to design an optimal hybrid strategy parameterized by the characteristics of the manufacturing system.

Empirical data indicate that microlevel engineering and management by themselves are not adequate for improving the rate of increase of productivity in manufacturing. This research identifies technology management and production techniques to be the major contributors to an increase in productivity. Under investigation is development of models for organizing macro- (e.g., government, university) and micro- (e.g., firm, union) agents around technology development and adoption, as well as innovative production processes.

A first-of-its-kind emulator for an FMS has been constructed to test algorithms for the distributed real-time scheduling and control of large-scale, discrete-event systems. A network of 13 computers defines the control hierarchy. An automated guided vehicle and an automated storage and retrieval system are included for material handling. Four generic processing and one fixturing station will have virtual processing capability. Eventually we will construct a second FMS emulator and a coordinator to synchronize production in both FMSs. The emulator will also be employed as an educational laboratory in several manufacturing engineering courses.

A hierarchical, object-oriented programmable logic simulator (HOOPLS) has been developed for modeling large-scale, discrete-event systems controlled by a hierarchy of computers. For each controller, a coordinated object is defined within which multiple entity types can be assigned to various subordinate resources to execute assigned tasks. HOOPLS focuses on the modeling of interactions among the controllers that coordinate the flow of the entities. HOOPLS is currently being applied to modeling several manufacturing systems and is also being used as a computer-aided design tool for the construction of a physical emulator for a flexible manufacturing system.

A new hierarchical, object-oriented framework for process planning is being developed. In an experimental implementation, we first developed specialized software for detailing the elements of the product structure diagram and the processing plan for the product, and then provided specialized windows where production data collected during the manufacturing of the product could be entered. The plans and collected data will be stored in a object-oriented database and will eventually provide an immediate input in the construction and validation of simulation models for the manufacturing facility.

A new object-oriented simulation approach is being developed for the production planning problem. The approach will permit the product structure diagram and specialized policies for managing the inventory for each product to be considered. Finally, the simulation approach defines a generalized queue for all staged and dispatched orders so that a generic interface to any investigated decision algorithm for managing the production can be provided. Eventually, the modeling will be expanded to consider the purchasing of materials from external vendors, material and capacity requirements planning, and real-time production scheduling.

This research deals with integrating three stages of the quality control and assurance process in a manufacturing or assembly environment: design, manufacturing, and fault-detection stages. The principal objectives of this research are twofold. First, a probabilistic network model (PNM) is to be developed as a diagnostic tool. For this purpose, the existing PNMs, such as influence diagrams, belief networks, and path analysis, are investigated. The second objective is the integration of the PNM with the data from SPC charts and with the robust design techniques. The results are to be applied to the Italtel's printed circuit board production process in Milan, Italy.

The objective of this work is to develop analytical and computational means for a comprehensive geometric analysis of a single lamina in a composite shell. The analysis is necessary for assuring both the strength and the geometric feasibility (the absense of tow jamming in each lamina) of the layup at any location in the shell. The analysis involves an analytical construction of fiber trajectories on a given doubly curved surface, starting with two intersecting fibers chosen to represent the initial conditions of the problem. The output will include the fiber lengths in each of the two fiber arrays (determining the cutout pattern for the lamina) and the fiber angle (which must be within the admissible range). A fast and computationally efficient geometric analysis is a prerequisite for a more challenging problem the geometric optimization of the reinforcement layout.

When designing a laminated composite shell of double curvature, the reinforcement lay-up specified (and, perhaps, optimized) at one given location in the shell is not always statically adequate or even geometrically feasible elsewhere. This limitation underlies the need for the geometric optimization of individual laminae. A Chebyshev net serving as an analytical model of a lamina is uniquely determined by specifying the trajectories of two intersecting fibers. Accordingly, the geometric optimization problem is stated as the following minimax problem. Given the directions of two intersecting fibers at one point on the surface, determine the two reference fiber trajectories such as to minimize the maximum skewing in the net. The search algorithm employs automated ordered modifications of the reference trajectories.