A Joint Distribution-Based Testability Metric Estimation Model for Unreliable Tests

The selection of tests required to make complex systems testable is a fundamental of system-level fault diagnosis. To evaluate the test selection, testability metric estimation (TME) is required. The influence of unreliable (imperfect) tests, whose outcomes are non-deterministic due to unstable environmental conditions, test equipment errors, and component tolerances, should be considered for accurate TME. Previously, researchers considered a TME model using a Bernoulli distribution with the assumption that the variations of different test outcomes are independent. However, this assumption is not always true. To address the issue, a joint distribution-based TME model was developed derived from the copula function to quantify the influence of dependent outcomes of unreliable tests. The efficacy of the developed TME model was verified with a linear voltage divider and a negative feedback circuit.