Software toolkit for aerospace systems diagnostics

This paper describes a component-based modeling approach for diagnosing continuous-valued systems. Our approach allows a user to build diagnostic models for complex continuous-valued systems based on a library of component models. Continuous-valued systems are modeled, and then dynamically transformed into discrete-valued models (in the form of causal networks), and diagnoses are generated using the causal network model-based diagnostic technology. The primary novel contributions of this work are (1) applying sophisticated and powerful model-based diagnostic techniques to hybrid systems, and (2) employing dynamic techniques for mapping hybrid models into the discrete models necessary for model-based diagnostic inference.