Abstract: Inaccuracies in the present proportionate stratified sampling algorithm based on fuzzy failure rate data is one main cause which makes the testability demonstration results untrustworthy. In order to solve the problem, a new allocation plan is formed which takes into consideration the empirical prior information combined with Monte Carlo simulation. First, the Gamma distribution is applied to fit failure rate data by considering the randomness and order of failure rate, and then, the parameters in Gamma distribution is calculated by limit fractiles. Quantities of random numbers following Gamma distribution are gained through limiting samples, and the mean value of the random numbers is applied to allocate the failure sample. Finally, a conclusion is drawn that the allocation plan based on the method proposed in this article is rational and the result of corresponding testability demonstration test is much more trustworthy.

Key words: testability demonstration tests, allocation of failure sample, stratified sampling, failure rate, Gamma distribution, Monte Carlo methods