[Fall 1998
Colloquiums]
[Department Homepage]
DEPARTMENT OF MATHEMATICS AND STATISTICS
OAKLAND UNIVERSITY
ROCHESTER, MICHIGAN 48309
Ananda Sen
Department of Mathematics and Statistics
Oakland University
A Journey Through The Discrete Reliability
Growth Methodologies
Abstract
As a system undergoes development, its reliability generally improves
as testing exposes failure modes and appropriate design changes or corrective
actions are subsequently implemented. Reliability growth models are pursued
to utilize the available test data from all stages in an integrated way
in order to obtain a precise estimate of system reliability. Since its inception,
reliability growth modeling has been viewed to be primarily applicable to
"continuous" time-to-failure data appearing in the context of
repairable systems. Concurrently, in regards top analyzing "success-failure"
data from single-shot operating systems, discrete reliability growth models
and methodologies underwent a substantial development, although the literature
is widely scattered and the surveys have been quite limited in this area.
The intent of this talk is to present a brief historical overview of the
models and methods employed in the past three decades for discrete failure
data exhibiting a growth (or decay) in reliability. Models will be discussed
in the classical framework as well as in the Bayesian paradigm, with occasional
analogies drawn from some of the popular continuous-time models. Special
emphasis will be provided to a class of regression models evolving from
the popular Duane hypothesis of linearity between cumulative number of failures
and cumulative operating time on a log-log scale. The talk will conclude
with an indication of the research avenues in this area that have thus far
remained unexplored.
372 Science and Engineering Building
Tuesday, November 24, 1998
3:004:00 P.M.
Refreshments at 2:303:00 P.M. in Room 368, Science and Engineering Building