Fault detection of non-linear systems based on multi-form quasi-armax modeling and its application to the ship benchmark

Sie sind hier: Homepage > Suche

Fault detection of non-linear systems based on multi-form quasi-armax modeling and its application to the ship benchmark

Kumamaru, K. / Inoue, K. / Hosoyamada, Y. / Söderström, T.

Model-based fault detection method is applied to a ship propulsion system. When modelling the object system, quasi-ARMAX (Auto-Regressive Moving Average with eXogenous signal) model with multi-model form is used. In this model, the system non-linearity is incorporated into model parameters by using NNMs (Non-linear Non-parametric Models). Thus the model has a linear structure and can effectively be used for control design and system analysis of general non-linear systems in the frame work of linear system theory. However, the KDI can no longer be directly applied to evaluate the distortion of identified Quasi-ARMAX model, since the model is essentially non-linear, and the corresponding likelihood function becomes non-gaussian. To solve this problem, a fault detection method based on multi-model has been proposed for general non linear systems. The basic idea of the method is as follows: when unified NNMs are used for expression of all parameters in Quasi-ARMAX model, it can be trans formed to a multi-model form that consists of several local linear ARMAX models. If the identification is performed based on the Quasi-ARMAX model with the multi-model form, then the model distortion due to a fault will be reflected in the local ARMAX model. Therefore, the model distortion in the local model can be evaluated by the KDI and distortion of the identified model is estimated by considering an appropriate weight for each KDI. The effectiveness of the method is verified through simulation studies on the ship propulsion system.