Fault localization for a series system when tests are unreliable

Official URL: http://risc01.sabanciuniv.edu/record=b1620283 (Table of Contents)

In this study, we consider a failed series system in which any of the components in the system can be the cause of the failure with different independent probabilities. We are allowed to sequentially test the components in the system to localize the faulty one. Prior probability that a component is the cause of the failure as well as the cost of testing a component are known. We consider unreliable tests in which a test can identify a component as working when in reality it is down, and vice versa. Therefore, there are costs corresponding to misclassi cation of the components in the system and the total expected cost of diagnosing becomes sum of inspection costs and misclassi cation costs. In this study we propose a model in which the repetition of tests are allowed at most once to minimize the total expected cost. Therefore, the aim here is to not only determine the best test sequence, but also the best repetition strategy with minimum expected cost. The complete mathematical model which covers all repetition strategies is introduced. Di erent characteristics of the model are discussed and numerical results are presented to demonstrate the possible cost reductions through repetition of tests.