Approximate model-based diagnosis using greedy stochastic search

Most algorithms for computing diagnoses within a model-based diagnosis framework axe deterministic. Such algorithms guarantee soundness and completeness, but are NP-hard. To overcome this complexity problem, we propose a novel approximation approach for multiple-fault diagnosis, based on a greedy stochastic algorithm called SAFARI (StochAstic Fault diagnosis AlgoRIthm). SAFARI sacrifices guarantees of optimality, but for models in which component failure modes are defined solely in terms of a deviation from nominal behavior (known as weak fault models), it can compute 80-90% of all cardinality-minimal diagnoses, several orders of magnitude faster than state-of-the-art deterministic algorithms. We have applied this algorithm to the 74XXX and ISCAS-85 suites of benchmark combinatorial circuits, demonstrating order-of-magnitude speedup over a well-known deterministic algorithm, CDA*, for multiple-fault diagnoses.