Fault Diagnosis of Nonlinear Differential-Algebraic Systems Using Hybrid Estimation

Modern technical systems often contain components capable for extensive autonomous actions. Thus, an integrated system supervision is essential in enabling an adequate reaction for compensation of unpredictable substantial variations. This is addressed by fault detection, isolation and identification techniques discussed in this article. Therefore, an overview is given about modelling of systems subject to faults, continuous state estimation utilizing an unscented Kalman filter and hybrid state estimation by the interacting multiple model approach. These methods are generalized for application to nonlinear differential-algebraic equations, i.e. DAE systems. DAE systems arise in such fields as discretization of partial differential equations or optimization problems. However, the appearance of DAE systems most often results from an object-oriented modelling (OOM) approach. Since OOM is probably the most relevant approach for modelling complex systems, the generalization and adaptation of supervision methods to DAE is the principal subject of this contribution. Finally, the proposed fault identification approach is applied to a hydraulic system, and the related results are discussed in detail.