A Diagnostics Framework Based On Abductive Description Logic Reasoning

This paper presents a flexible approach to automated diagnostics for complex technical systems, built on a firm theoretical and methodological basis. To this end we devise a conceptual logic-based model for diagnostics inspired by ISO standards, and subsequently investigate different mappings of this formal model into the framework of relaxed abduction, a novel non-standard description logic (DL) reasoning task introduced lately. This framework allows for the robust interpretation of potentially incomplete information with respect to imperfect diagnostic models. We investigate the use of both causal and anti-causal models representing the diagnostic knowledge, and evaluate a prototypical application of the approach to steam and gas turbine diagnostics with encouraging results.