A Bayesian approach to confidence in system testing

The authors present a purely Bayesian technique for computing the probability that the conclusion reached during a fault isolation is correct given only the results of the (potentially unreliable) tests that have been run and information about the reliability of those tests. This technique is then applied to both off-line and real-time system analysis as a means of determining the potential correctness of a fault isolation conclusion as well as to guarantee (if possible) an arbitrarily selected probability that a conclusion is correct.<>