Binomial reliability demonstration tests with dependent data

摘要

A reliability demonstration test (RDT) is a useful method of demonstrating the capability of a product or process. With continuous data are available, demonstration of capability is often done with a capability index and is preferable. However, in case of discrete data, a binomial reliability demonstration test (BRDT) is appropriate. The BRDT for binary data has two components associated with its quality statement, which can both be independently set by the user. The first is the confidence associated with the test and the second is the reliability of the process or product. An example is that there is 95 percent confidence that the reliability is at least 0.99. For BRDT a sequence of samples are taken and tested and the result is either a pass or failure and the probability of failure is p. The samples in sequence take the values either 0 or 1 and the results need not have to be always independent. In addition to the confidence and the reliability, the maximum allowable number of failures and the sample size are associated with BRDT. The sample size n can be computed based on the specified values of the confidence parameter a, the reliability parameter p, and the maximum number of failures allowed k. The sample size can be calculated using the inequality derived by Jaech (Ref. 1) for acceptance sampling. This can be solved by iterating for values of n until the inequality is satisfied by the smallest value of n. There are many methods suggested by various studies for the derivation of the sample size for continuous data. (23 refs.)