Use of Weibull Failure Rates
Often one of the largest costs of an operational phase, estimating reparable costs remains more an art than
science. A number of different probability distributions are commonly assumed to underlie failure rates for predicting future reparable requirements. In the case where a limited amount of data is available on demonstrated failure rates, the Weibull can be a versatile and useful estimating tool. This paper demonstrates how observed failure rate data from a time-limited test can be adapted to a Weibull distribution for predicting future failure rates. Since the Weibull can be adjusted to represent an exponential, a binomial, or a Gaussian distribution, this paper discusses how to determine which shape is appropriate, and when to assume the shape parameter rather than estimate it. Real program data is used in examples of how the Weibull can be applied.
Virginia Stouffer is a SCEA-certified cost analyst with Masters’ degrees in Economics and Electrical Engineering. She has worked for LMI since 1997 and consults in areas of feasibility, cost, analysis, and modeling. She works in the areas of aviation, defense, electronics, and environment.