Decisions regarding capital investments are among the most important that managers of a firm must make. An analysis must be made first of markets, pricing, costs, and other factors which influence the expected outcome of any proposed investment. Capacity planning decisions are affected by investment decisions. These decisions, large and small, directly influence the success of a company and require the appropriate commitment of substantial resources over long periods of time. The purpose of capacity planning models is to help determine the amounts and kinds of capacity appropriate under varying types of demand and cost restrictions. Recently the maximization of present worth subject to physical constraints is the common criterion used in evaluating the proper capacity policy. Failure of industry to recognize the economic life of its productive facilities and to replace them on time can have regrettable consequences. This research is dedicated to the development of economic decision rules for the medium to long run capacity planning of new manufacturing technology with different maintenance strategies under physical deterioration and obsolescence.

In this dissertation, I developed a unified complex network framework for environmental decision-making. I focused on complex network systems arising in the context of green logistics, including global supply chains, and electronic waste recycling. The framework that I developed is able to handle many decision-makers at the tiers of the networks, and enables the prediction of the flows of the materials between tiers as well as the prices at the tiers, along with the emissions generated, and the incurred costs and profits. I first developed a theoretical framework for supply chain networks with environmental concerns in the context of decision-making in the Information Age today. I allowed different decision-makers to weight the criteria (including the environmental ones) in distinct fashion. Subsequently, I generalized the original model to a global supply chain network model which included environmental criteria, electronic commerce, and risk management. I then developed an integrated reverse supply chain management framework that allows for the modeling, analysis, and computation of the material flows as well as the prices associated with the different tiers in the multitiered electronic recycling network. I also extended this model in order to deal with the environmental risk caused by hazardous material generated from the recycling processes. I assumed that the environmental risk due to hazardous material depends on the amount of residual hazardous material that is not extracted from electronic wastes by the processor as well as on the storage amount of hazardous waste that is not disposed yet by the hazardous material disposer. The models and computational methods were based on the methodologies of variational inequality theory for the study of the statics (cf. Nagurney (1999)) and projected dynamical systems for the dynamics (cf. Nagurney and Zhang (1996)).

This dissertation addresses three important problems with the shared objective of increasing the efficiency of supply chains. The first one considers a forward distribution network. The other two deal with the reverse flow of product in remanufacturing environments. The dissertation is divided accordingly into three well differentiated parts. The first part considers a multistage inventory system composed of a single warehouse that receives a single product from a single supplier and replenishes the inventory of n retailers through direct shipments in Full-Truckload transportation mode. Based on the structural properties of optimal solutions derived, we develop an exact algorithm that runs in polynomial time for a given number of retailers. To overcome the computational burden when the number of retailers is large, we propose two additional algorithms based on Lagrangian decomposition. Computational experiments show the effectiveness of the algorithms and the gains associated with coordinated versus decentralized systems. The second part of the dissertation considers the joint production, procurement and inventory control problem for an assembly system where one of the two components in the end product is produced through remanufacturing. By acquiring additional information on the quality of returned products, the firm can adopt a more selective recovery process to reduce the uncertainty in the remanufacturing yield. We show that the optimal recovery lot size first increases, then decreases as more information is acquired in a special case and the convexity of the expected total cost in the recovery lot size in a slightly more general model. The third part of the dissertation investigates the profitability of remanufactured products for a monopoly firm in a single period setting. The firm has to decide (1) whether to offer both new and remanufactured products or to offer only new products to the market, (2) whether to offer some new products as substitutes to remanufactured products. We characterize a threshold for the remanufacturing cost under which offering remanufacturing products is profitable, and a threshold for the production cost of new products which is a necessary condition for the substitution option to be profitable.

Effective decision-making is a key prerequisite for a successful organization. Today’s organizations are large and continue to grow in size and scope. This development leads to higher complexity in managing and controlling organizations. Consequentially, selecting the right course of action has become more difficult for the individual decision-maker. Considering only immediate and local effects of actions reduces decision complexity but also decision quality. Effective decision-makers need to take into consideration the consequences of their actions on different time and organizational scales. In our research we develop a framework for multiscale decision-making, which gives decision-makers the ability to make opportune decisions in the face of multiscale system properties. Also, we provide organizations with a tool with which they can gauge the consequences of various organizational parameters on hierarchically interacting agents over multiple organizational and temporal scales. We begin our investigation with a model of two hierarchically interacting agents in a superior-subordinate relationship. The agents influence each other's rewards and chances of success with their decisions. This bi-directional influence creates a game-theoretic situation. Using the concept of Nash equilibria, we determine the agents' optimal strategies for different organizational parameters. Results are presented through phase diagrams, which graphically capture how variations in organizational parameters affect agent behavior. We extend this model of hierarchical interaction between two agents to a generalized tree-structured interaction network of many agents. This network resembles the typical organizational form of an enterprise. We visualize the hierarchical agent interaction with dependency graphs, which provide a compact representation of the organization and the associated parameters. In a final step, we extend the one-period model to allow for multiple time periods. We use Markov decision processes to model the multi-time-scale interactions. We take into consideration that decisions on lower organizational scales are made at a higher frequency than decisions at higher organizational scales. By considering the interdependence of decision frequency and hierarchal level, we fuse the temporal scale with the organizational scale which results in one comprehensive multiscale decision-making model.

Queueing networks have been used as modeling tools for a variety of applications including manufacturing, telecommunications, and general service systems. Open, finite queueing networks take into account the fact that most real life systems have limited storage space. Analysis of open, finite queueing networks is extremely difficult and very few exact results are reported in the literature. The main purposes of this dissertation are: (1) To develop analytically tractable and computationally efficient bounds on the throughput of finite queueing networks. (2) To use the bounds in developing an approximation to the throughput function from such networks. (3) To study the routing optimization problem with the bounds and the approximations. The general optimization problems that exist in queueing networks are discussed and the optimal routing problem is studied. In particular, this dissertation deals with the static routing problem, wherein customers arriving at split junctions in series-parallel queueing networks are assigned routing probabilities to be routed along different arcs and nodes so as to maximize total throughput. An approximation of throughput in split topology networks is developed and this is used to specify sub-optimal ranges of routing probabilities. This concept is then extended to larger series-parallel networks without blocking across split and merge junctions. A linear time algorithm is then developed to route customers in these series-parallel networks which is shown to be quite efficient and robust. In the case of series-parallel networks with blocking across split junctions another approximation is developed for the throughput function to account for the interdependencies between different service streams. A non-linear programming algorithm is developed to solve the optimal routing problem based on this approximation. Several experimental situations are studied and the results of the algorithm are verified. Finally, some avenues for future research, including multiple classes of customers and their optimal routing in queueing networks, are explored along the lines developed in the dissertation.

No description available.

This dissertation is concerned with a facility layout and traffic flow network design problem where the focus is on assigning activities to locations in order to minimize the congestion within a circulation system. This dissertation first presents a series of mathematical models of the facility layout problem in which stochastic congestion within the movement system is modelled. The models can be used to assist the design and placement of departments or activities within a facility or surrounding environment where the congestion of the pedestrian or customer traffic flow is of major concern. Numerical examples are provided to illustrate the difference between the various formulations. Second, a new heuristic algorithm called STEP, Sample Test pair-wise Exchange Procedure, to solve these complex problems is developed. With its straightforward approach, the algorithm can solve large-scale Quadratic and Stochastic Quadratic Assignment Problems with efficient computing times and excellent solution performance. Computational experience for solving many test examples is presented. Extensive work has been carried out for selecting and evaluating appropriate facility network design configurations. This is done by evaluating the performance of the alternative network designs such as star, grid, and ring topologies in terms of customers' sojourn time in the system. Finally, the dissertation concludes with a discussion of open problems and directions for future research.

The purpose of this research is to explore aspects of statistical quality control methodology that may be usefulIy applied to monitor component reliability both in life test experiments and in service. The object of such monitoring would be to detect alterations in the expected life of the component which may be caused by changes in materials, in manufacturing practices, or in other factors that did not appear to be significant in the sense of quality of conformance to specifications. The work is divided into two major sections. The first section considers the monitoring of results from life tests when the distributional form and parameter values have been established during pre-production testing. It is assumed that the underlying life of the product can be described by a two-parameter Weibull distribution. In an attempt to monitor the scale and shape parameters quasi-independently, two control charts are developed using the first and the nth order statistics from a sample of size m. The control chart to be used to monitor the shape parameter was found to be disappointingly ineffective. The other control chart, however, was found to give two distinctly different signals depending upon the affected parameter, and thus can be used alone to accomplish the intended purpose. The second section considers the problem of monitoring component reliability after the component has been placed in service. Under the assumptions that the component is in continuous service and that censoring due to unrelated causes is not significant, control charts are developed that monitor the aging profile of the component based upon a quantal response model for an arbitrary life distribution. A simple method for estimating the control limits for a two-parameter Weibull life distribution is examined. The method could be applied to many other life models. The work concludes with consideration of how such quality control methodologies could be practically employed, and makes recommendations for additional areas of research.

Flexibility in operational decision making can be used to great advantage in many manufacturing and assembly environments. Mixed technology printed circuit board assembly is offered as one such environment. A heuristic-based dynamic control methodology is developed taking advantage of a natural control hierarchy at three levels. The methodology makes use of system level product mix determination, intercell routing, and machine level buffer sequencing to obtain improved system performance. It is demonstrated that a concentration on key areas of integration between the three control levels is of vital importance in such a strategy. Integrated, multi-level control strategies are shown to be superior to stand-alone flexible controls, producing synergistic results which are better than those of the individual control elements acting separately. The integrated solution approach is further shown to provide robust system improvement over a wide range of system conditions.

The discrete optimization technique called local search yields impressive results on many important combinatorial optimization problems. Its tendency to return solutions that are sub-optimal, however, remains a serious limitation of the approach. In this dissertation we propose and investigate a general technique that seeks to overcome this difficulty while maintaining the speed that makes local search an attractive approach. We focus on constrained problems which have an induced non-uniform neighborhood structure, and we examine three such problems experimentally. Our example set is: (1) a variant of Stock Cutting; (2) a precedence constrained routing problem; and (3) a weighted version of the latter. Using our techniques we have been able to improve significantly the quality of solutions found in each of these domains over that found by the traditional local search algorithm, with only modest increases in running time. Our technique is independent of problem domain and, therefore, is applicable to a wide variety of problems.