Casual [i.e., causal] nets for diagnosis

Abstract: "Cause-effect relations are widely used for fault diagnosis, but a perception gap seems to exist between the causal reasoning technology and the real world expectation. This is because the causal modeling of a power system falls into a cyclic loop if not treated properly. This paper shows how a Bayesian net model can be formulated without loop. Based on this model, a qualitative way is proposed to evaluate the net and the diagnosis problem is transformed into an optimization problem. A special organization called A-Teams is used to solve this transformed problem. Real examples are given to illustrate the process and demonstrate the results."