On Fault Diagnosis using Bayesian Networks ; A Case Study of Combinational Adders.

In this paper, we use Bayesian networks to reduce the set of vectors for the test and the diagnosis of combinational circuits. We are able to integrate any fault model (such as bit-flip and stuck-at models) and consider either single or multiple faults. We apply our method to adders and obtain a minimum set of vectors for a complete diagnosis in the case of the bit-flip model. A very good diagnosis coverage for the stuck-at fault model is found with a minimum set of test vectors and a complete diagnosis by adding few vectors.