Abstract Observed reliability data from fielded systems is highly desirable because they implicitly account for all actual usage and environmental stresses. Many companies and large organizations have instituted automated field-failure reporting systems to organize and disseminate these data. Despite these advantages, field data must be used with caution because they often lack sufficient detail. Specifically, the precise times-to-failure are often not recorded and only cumulative failure quantities and operating times are available. When only data of this type are available, it is difficult to determine whether component or system hazard function varies with time or is constant (i.e., exponential distribution). Analysts often use the exponential distribution to model time-to-failure because the distribution parameter can be estimated with just the merged data. However, this can be dangerous if the exponential distribution is not appropriate. An approach is presented in this paper for Type II censored data, with and without replacement, to evaluate this assumption even when individual times-to-failure are not available. A hypothesis test is presented to test the suitability of the exponential distribution for a particular data set composed of multiple merged data records. Two examples are presented to demonstrate the approach. The hypothesis test readily rejects an exponential distribution assumption when the data originate from a Weibull distribution. This is a very important result because it has generally been assumed that time-to-failure data were always required to evaluate the suitability of specific time-to-failure distributions.
[1]
D. W. Coit,et al.
Practical reliability data and analysis
,
1986
.
[2]
Jerald F. Lawless,et al.
Statistical Models and Methods for Lifetime Data.
,
1983
.
[3]
M. Kendall,et al.
The advanced theory of statistics
,
1945
.
[4]
Frank M. Guess,et al.
Bayes estimation of component-reliability from masked system-life data
,
1996,
IEEE Trans. Reliab..
[5]
N. Singpurwalla,et al.
Methods for Statistical Analysis of Reliability and Life Data.
,
1975
.
[6]
C. J. Wang.
Graphical analysis of ill-collected interval data for a repairable system in vehicles
,
1991,
Annual Reliability and Maintainability Symposium. 1991 Proceedings.
[7]
J. S. Usher.
Weibull component reliability-prediction in the presence of masked data
,
1996,
IEEE Trans. Reliab..
[8]
Maurice G. Kendall,et al.
The advanced theory of statistics
,
1945
.
[9]
Marvin Zelen,et al.
Mathematical Theory of Reliability
,
1965
.