Theses and Dissertations

An investigation of the sample sizes required in order to find average queue lengths with specified accuracy is presented in this thesis. The queuing systems studied are single server models with infinite queues and no priorities.

Procedures are presented which provide management with a tool to make integrated decisions what are consistent with the interacting constraints of a multi-item, multi-source, warehouse constrained, stochastic environment.

Implicit in the classical techniques of smoothing and forecasting of discrete time series is the restriction that the data spans equal intervals of time. There exists a need to make forecasts at unequal time intervals for the transactio or even oriented business activity.

Integer programming is currently enjoying an increase in interest due in no small part to the potential use of high speed computer equipment to solve heretofore unmanageable prolem. The development and analysis of integer programming algorithms are presently restricted to small size problems. This limitation is also a concern of potential users of these algorithms. To use the algorithms successfully the user must know how these algorithms perform on the type and size of problems with which he is concerned.

The problem of balancing an assembly line has been facilitated by a number of analytic methods for which computer programs are available. The majority of such techniques accept only deterministic values as the element times of the operations.

A computational procedure for obtaining the minimum makespan solution for scheduling N jobs through M machines with job start constraints, due date constraints, and job passing allowed is investigated concerning the method of bringing jobs into solution.

An investigation of the properties of certain tandem queueing systems is presented in this study. A generalized tandem queueing simulator is developed and the results of the simulations for two systems are discussed.

The advantage of applying a learning curve to the server in a queueing system has not been obtainable dure to the lack of an easily computed solution for the queue where the mean service time is a function of the number of units previously served. Such a queue is solved in this thesis using the method of imbedded Markov chains.

The application of adaptive, steady-state control to batch type manufacturing processes is addressed. An empirical modeling technique based on steepest ascent is developed and the performance of an optimal control system is compared with that of an adaptive control system.

The one period, one dimension allocation model is described and developed mathematically. Two computational difficulties are investigated. If enumeration is used to progress from stage to stage through the model, then no difference could be found in the allocations due to the order of consideration of the projects. The projects may be numbered in an arbitrary manner and the allocations and the resulting total returns will be the same.

The electromagnetic properties of a live biological system are extremely important to many medical applications. While information about electromagnetic properties of tissues are available in the literature, little or no data are available for a single cell or subcellular structures. Microwave biological/cell detection has been demonstrated to be useful and promising in many medical applications due to its internal properties such as non-invasive, fast and label-free.

The policy-iteration technique is a method of selecting the optimal solution vector for a Markovian decision problem. This technique was analyzed to determine its sensitivity with respect to random errors in the stochastic elements in such problems.

A model for predicting product damage due to hurricanes in the Southern United States is presented.

An investigation of the properties of a class of bulk service queuing systems in which the server capacity and/or speed can be changed according to the number of customers in the system just after the end of each service. The model includes the possibility that some customers may stay with the server for more than one service interval.

The problem considered in this thesis is that of finding the probability distribution of waiting times in a queueing system consisting of: 1- a single server; 2- interarrival times are independent random variables each having the exponential distribution; and 3- service times are independent random variables having either the Erland or uniform distribution.