Discrete Reliability Growth
Abstract
Three methods for weighting an exponential regression model to estimate discrete reliability growth were derived. The first method systematically applies more weight to less variable estimates. The second method similarly applies heavier weight to the most recent estimate. The third method allows the user to choose the weighting scheme. These methods were evaluated against eight patterns of actual reliability by altering a previously developed Monte-Carlo simulation. Their performance was then compared to the unweighted exponential regression and Maximum Likelihood Estimate with Discounting (MLEWD) models. A comparison is made of the least squares estimates for reliability growth using two different unbiased estimates for the negative of the natural logarithm of the failure rate.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1988
- Accession Number
- ADA202027
Entities
People
- Pamela A. Markiewicz
Organizations
- Naval Postgraduate School