Bayesian Decomposition Method for Computing the Reliability of an Oriented Network

For a nonoriented network, Bayesian decomposition is straightforward and well known. A keystone element is assumed perfect (shorted), then failed (open) and the reliabilities of the two subnetworks are calculated. But for an oriented network, when the keystone element is assumed perfect, it cannot always be shorted (the 2 nodes brought together), because even a perfect element still retains its orientation. A technique for choosing keystone elements is proposed.