Probabilistic Expert Systems for Gas Turbine Diagnostics

Knowledge Industries (KI) will explore the technical feasibillity of applying probabilistic expert systems to development of T800 gas turbine diagnostics. KI is a leading developer of probabilistic expert systems and has developed other gas turbine diagnostic systems. In this project, KI will develop a prototype probabilistic expert system for the T800, investigate use of thermodynamic modeling as a component of a diagnostic system, and develop a software architecture and development plan for fielding a Phase II developed expert system. The basis of a probabilistic expert system is a probabilistic model of the faults and observables for a given diagnostic problem. We will use probabilistic causal networks called belief networks to model the relationships among important variables for the T800 gas turbine. Belief networks are a flexible and maintainable methodology for construction of large diagnostic systems. KI has developed state-of-the-art software tools and techniques for constructing and evaluating belief networks and we will evaluate the feasibility of probabilistic modeling of the T800 by constructing a prototype belief network and expert system, including a thermodynamic analysis of the engine. A major component of the program will focus on development of a software development and delivery plan for shop-based, line-based, and on-board diagnostics.