New Models for Reliability Growth.
PRINCETON UNIV NJ DEPT OF STATISTICS
Pagination or Media Count:
Reliability growth modeling is an important component in development testing of complex systems. It provides estimates of the current status of the process as well as forecasts of reliability at future time points. Several models have been investigated in the literature. In this report, we focus on comparing the performance of a model introduced by L. Crow and a family of models which are generalizations of those introduced in Braun and Paine. The novel feature of this latter family is that is models the logarithm of the intensity function of the failure process as a multiple linear regression of some function of the number of different failure modes uncovered and the number of recurrent failures. These models possess great flexibility in accommodating many different shapes of reliability growth curves including nonmonotonic ones. Three members of this family are compared with the Crow model over a range of simulated reliability growth patterns. One member, in particular, seems to provide short-and long term forecasts superior to those of Crows. These results suggest that these new models be in the modeling of field data. Three such data sets are considered here as well. Author
- Statistics and Probability