Reference: Srinivas, S. Building Diagnostic Models from Functional Schematics. Knowledge Systems Laboratory, February, 1994.
Abstract: Diagnosis in Engineering systems is a very important problem. Though some progress has been made in providing software aids to diagnosis the current technologies suffer from many drawbacks. Bayesian networks are a very attractive modeling scheme for diagnosis which address many of these drawbacks. However, hand construction of diagnosis models in the form of Bayesian networks is a very difficult task. We develop a methodology for automatically constructing diagnosis models in the form of Bayesian networks from a functional schematic describing the Engineering system. Other than the basic diagnosis capability we show how our methodology is extended to become a unifying framework encompassing concepts such as hierarchical modeling, dynamic systems, optimal test generation and repair, maintenence scheduling and reliability theory.
Full paper available as ps.