Reasoning Framework for Diagnosis and Prognosis

Abstract

This paper describes a general-purpose probabilistic framework for reasoning in diagnosis and prognosis. The framework provides a mathematically rigorous way of handling uncertainty, which is often present in diagnosis and is inherent to prognosis. It is based on an extension of Bayesian network models and Bayesian inference. It coherently integrates multiple sources of evidence in diagnosis and prognosis, including component usage, environmental conditions of operation as well as component health and health trends. The framework has been applied to diagnosis of very complex transportation and aviation systems and to prognosis of electromechanical and electronic subsystems in aviation.