Causal Reasoning Under Uncertainty With Q/C-ENetworks: A Case Study On Preventive Diagnosis OfPower Transformers

The paper presents the formalism of quantified causal-evidential networks (Q/C-E networks) for causal reasoning under uncertainty in networks of propositions. First, the basic concept of C-E network is introduced. The issue of representing uncertainty about propositions and causal-evidential relations is then discussed and Q/C-E networks are defined. Methods for propagating and aggregating uncertainty in a Q/C-E network are proposed and their main properties are illustrated. The proposed approach has been successfully experimented in the design and development of ASTRA, a knowledge-based system for preventive diagnosis of power transformers.