|Thesis abstract: |
This work proposes an adaptive diagnosis technique for complex digital device and system called Automatic Fault Detective. This technique is based on Bayesian Networks, it allows the description of the system at an high level of abstraction to simplify the task of test engineer teams. In this thesis, different directions of research are covered, aiming at an optimization of the cost of a diagnostic process: the identification of minimum cost initial test set for fault detection, an adaptive step-by-step execution sequencing of tests, a robustness analysis of the obtained diagnostic conclusions. The proposed methods are verified on both simulated systems (to prove correctness of results) and implemented on some real industrial case studies.