Browse All Digital Items

The problem to which this thesis is addressed is known as the n by m scheduling problem. It can be briefly described as follows: n commodities are, each in a prescribed order, to be processed on one or more of m facilities which are initially idle, ready to provide service, and uncommited in the future. All ofthe n commodities are ready to be processed initially, and each must be processed for a prescribed period of time on each facility to which it is to be routed.

Implemented to provide the users with as much power as possible, LETEXT conforms to design criteria which include service to the users and security and privacy of the users' information.

First-order methods for convex and nonconvex optimization have been an important research topic in the past few years. This talk studies and develops efficient algorithms of first-order type, to solve a variety of problems. We first focus on the widely studied gradient-based methods in composite convex optimization problems that arise extensively in compressed sensing and machine learning.

This study considers the effect of priority dispatching in a labor and machine limited production system. A simulation model of an actual shop is developed using GASP II. The shop is characterized as a flow shop with cycling in which lots are subject to stochastic yields and reworks. Five labor sections, each operating on three shifts, exist in the shop.

Forward Osmosis is a natural phenomenon that takes places across a semi-permeable membrane when there is a concentration difference across the membrane. Pure water permeates to the highly concentrated channel until the concentration across the membrane equilibrates. In water desalination applications, the same principle is applied. Spiral-wound membrane, flat sheet, or hollow fiber module are typical configurations in forward osmosis desalination modules.

Methods used by electric power companies for estimating future loads as a basis for timing generating capacity additions have been examined. It has been found that these estimates deviate widely from time to time from the loads subsequently experienced. A method is developed for interrelating the total annual energy requirements, annual energy sales by classes, and the annual peak loads as a possible basis for improving the timing of generating capacity additions.

The conversion of a decision table into an optimal decision tree is an integral part of most decision table compilers. This thesis establishes some fundamental concepts of a heuristic approach to the conversion of limited entry decision tables into decision trees and proposes algorithms for the latter two objectives.