Konferensartikel

We compare a number of different methods for estimating model parameters based on external stimuli. We examine the trade-offs between the different methodologies with respect to the modelling effort necessary to implement them and the granularity of the estimation obtained. In implementing these methods, we utilize Modelica and FMI. As an application we show how these methods can be combined with component fault modes to provide effective real-time estimates of the health of a physical asset based on thermal sensor data. In particular, we contrast the effectiveness of the different estimators in predicting the degree and location of fault.

Parameter estimation, parameter tuning, fault diagnosis, Modelica, FMI