DVER: A tool chain for cross-validation and perfection of discrete model-based diagnostic systems

The safe and reliable operation of complex aerospace vehicles requires rapid and accurate detection and identification of system faults. Discrete model-based fault diagnostic systems (FDS) are often used when rapid real-time fault detection is needed. Given a vector of current discretized sensor readings, such a diagnostic algorithm produces fault hypotheses as its output. In this paper, we present DVER (Diagnostic VERification), a tool set to support the automation of validation and improvement of the central FDS components - the discrete model and associated sensor data processing capabilities. We present the results of applying the DVER tool chain to a modelbased FDS for the NASA Advanced Diagnostic and Prognostic Testbed (ADAPT) system, a testbed for a redundant electrical power distribution system.