Testability demonstration test planning based on sequential posterior odds test method

Abstract

Testability is a new critical characteristic of equipment, and testability demonstration test plays an important role in assuring the designed testability levels. As a result of the complex structure and higher cost of the current equipment, a reasonable testability demonstration test plan can both guarantee the desired testability level and effectively decrease the test cost. This article considers the testability demonstration test planning with the binomial distribution. The sequential posterior odds test method that combines Bayesian method and sequential probability ratio test method is used to form a sequential decision process by calculating the Bayes factor and making judgment according to the test result sequentially. With a sequential decision criterion, the test result space is divided into three subspaces, based on which a Bayes factor threshold calculation method is presented according to the restriction parameters and prior probability density function of the testability index. Operation characteristic and average sampling number are compared among the classical method, sequential probability ratio test method and the sequential posterior odds test method, which shows that with nearly the same operation characteristic, sequential posterior odds test method has smaller average sampling number than sequential probability ratio test method and classical method. The impact of prior parameters on the sequential posterior odds test method is analyzed, and the results indicate that the sequential posterior odds test method can perform well with a range of prior parameters. Physical testability demonstration tests are carried out to verify the features of sequential posterior odds test and sequential probability ratio test methods. The result indicates that with the same restriction parameters and test condition, the sample size of sequential posterior odds test is smaller than the sample size of sequential probability ratio test as a result of the introduction of prior information.