Empirical Determination of WRSK Component Failure Distributions.

Current methodology for determining components of a War Readiness Spares Kit (WRSK) is based on mean times to failure for the entire U.S. Air Force inventory. Failures on a specific aircraft type during a calendar period are divided by flying hours logged on that aircraft type during the same period. Kit configuration is then optimized using the Poisson distribution to approximate kit demand behavior. The Poisson process requires a single statistic, mean value of distribution, for which world-wide mean time to failures is used. There are several alternatives to the Poisson distribution. Negative binomial distribution has been suggested as a more appropriate model, and it has been demonstrated that the WRSK configuration would change under this assumption. To test the suitability of both distributions, two analyses were undertaken of mean time to verified failures of the fleet and assumed constant failure rate of WRSK items among the fleet. In both analyses, the hypothesis that data is Poisson distributed could not be rejected for the majority of the cases. In several cases, however, the Poisson distribution was rejected as an adequate fit for the data and the negative binomial was a viable alternative. Negative binomial also appeared to be a better model in some of the cases where the Poisson distribution could not be rejected. (Author)