The New Jersey Institute of Technology's
Electronic Theses & Dissertations Project

Title:	Improving the reliability of electronic systems by minimization of the variances and stabilization of the mean values of the system performance criteria
Author:	Neu, Emil Carl
View Online:	njit-etd1966-004 (xiv, 165 pages ~ 4.7 MB pdf)
Department:	Department of Electrical Engineering
Degree:	Doctor of Engineering Science
Program:	Electrical Engineering
Document Type:	Dissertation
Advisory Committee:	Misra, Raj Pratap (Committee chair) Zambuto, Mauro (Committee member) Russell, Frederick A. (Committee member) Anderson, L. Bryce (Committee member)
Date:	1966-06
Keywords:	Electronics. Reliability (Engineering)
Availability:	Unrestricted
Abstract:	The parameters of an electronic system are not deterministic variables; but rather because of variations in manufacturing processes, they are randomly distributed variables. In addition, aging mechanisms will cause these parameters to drift with time. As a result of both of these factors, the system performance criteria deviate from their initial design center values. This research studied the effect of component parameter variations upon these electronic system performance criteria and then presented a thesis to minimize the effect of these variations. The distributions of the performance criteria were obtained as functions of the parameter distributions, by first approximating the performance criteria by the linear terms of a Taylor's series. This technique made it possible to obtain relatively simple expressions for the mean values and variances of the performance criteria. Then by assuming that the performance criteria were normally distributed, the probabilities of the performance criteria being within the required limits were determined. This analysis showed that the probabilities of the system performance criteria being out of tolerance could be minimized, if the mean values of the performance criteria were held at their design center levels and if the variances of these criteria were kept as small as possible. It was seen that the mean values of the performance criteria could be set at their design center levels by proper choice of the parameter mean values. The variances of the criteria were kept as small as possible by first minimizing these variances with respect to the system parameter mean values with the constraint that the initial performance criteria means assume their design center levels. Increases in the variances of the performance criteria with time, as well as drift of the mean values of these criteria, were prevented by satisfying certain relationships among the parameter drift rates. The results of this research include the extension and refinement of techniques for determining the distributions of system performance criteria as functions of system parameter distributions. In addition, methods were developed for the selection of the mean values of the parameters and the parameter drift rates, so as to minimize the variances of the system performance criteria, while at the same time preventing drift of the mean values of these criteria. Expressions for computing the coefficient of linear correlation between two variables whose values where selected in a non-random manner were also obtained.