부산대학교 도서관

We demonstrate how model-based diagnosis techniques can be applied directly in the frequency domain. Frequency domain models are used in reformulations of three diagnosis approaches: optimization-based parameter estimation; parameter estimation based on a particle filter formulated in frequency-domain; and analytic redundancy relations. Working in the frequency domain enables the parallelization of the diagnosis algorithms since the model's algebraic equations can be evaluated independently. Hence we can reduce the time for reaching a diagnosis. This approach is particularly useful for rotating machinery, where the fault modes exhibit rich dynamics better seen in the frequency domain. As an illustrative example we demonstrate our approach by diagnosing parametric faults in a DC motor.