Test set enrichment using a probabilistic fault model and the theory of output deviations