ON REGRESSION TECHNIQUES FOR ESTIMATING THE PARAMETERS OF WEIBULL CUMULATIVE DISTRIBUTIONS.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OHIO SCHOOL OF ENGINEERING
Pagination or Media Count:
A regression technique for estimating Weibull scale and shape parameters is introduced and experimentally investigated. The derivation involves maximizing the likelihood function of a set of normal density functions which give the probabilities of sub fi items failing until time ti. The estimation technique was evaluated using simulated failure data and the estimators were found to be optimal according to Kolmogorov-Smirnov tests. When compared with the Harter-Moore MLE iteration technique the regression method was 145 times faster and yielded Kolmogorov-Smirnov values 26 percent lower. Of 1100 estimations only one could not be accepted at the .20 confidence level as having come from the failure data. The method is well suited for small computers such as the IBM 1620 and may be employed to analyze incomplete or censored samples of failure data. Author
- Operations Research
- Manufacturing and Industrial Engineering and Control of Production Systems