This thesis examines hybrid estimators lor trusted health assessment of dynamic systems, focussing on applications from the field of aviation. Reliable fault detection and assessment of the level of degradation of a system are mandatory to achieve a cost-benefit when using a condition-based maintenance strategy. The chosen method can be verified to achieve consistent estimates. The integrated approach and the systematic verification of consistency is the main contribution of this thesis, analysed theore tie ally and demonstrated practically ill an aerospace application. Recursive Bayesian estimation techniques are a stochastic approach of stale estimation; this enahles derived methods to determine the probability of a failure being currently present in a system. This distinguishes these methods from other deterministic approaches, where the level of uncertainty is in general added ad hoc, for example by fuzzy thresholds. The assessment of consistency of a recursive Bayesian estimator in a nonlinear application is, however, often neglected. In this work, consistency testing is studied intensively. The developed library of estimators enables large-scale Monte Carlo simulations. Thus, the objective comparison of different estimators with common boundary conditions becomes feasible. The normalised estimation error squared is thu central criteria fur this analysis. Recursive Bayesian estimators with Gaussian approximation of the posterior density are considered in this work. Addressed are variants of the Kalman filter (KF), namely the extended KF (EKF), the iterated KKF, the unscented KF (UKF) and the iterated UKF. Extending the recursive Bayesian estimators for handling of hybrid systems allows to treat the fault detection problem. The autonomous multiple-model (AMM) algorithm, the generalised pseudo-Bayesian (GPB) algorithm of order 1 and the interacting multiple-model (IMM) algorithm are used in this work. For assessing the integrity of the discrete mode estimate a new criterion, the mode probability error squared is proposed. This allows lo assess different hybrid estimators regarding the confidence in the estimated mode. As the mode estimate is used to detect a failure in the system, this criterion allows to assess the quality of fault detection. In the considered applications, the IMM proved superior to the AMM and GPB1 regarding consistency of the underlying state and mode estimates.
Trusted health assessment of dynamic systems based on hybrid joint estimation
Sichere Zustandserfassung dynamischer Systeme basierend auf hybriden Abschätzungsverfahren
2010
168 Seiten, 55 Bilder, 16 Tabellen, 84 Quellen
Hochschulschrift
Englisch
Zustandsschätzung , Zustandsüberwachung , dynamisches System , Luftfahrt , Fehlernachweis , Zuverlässigkeit , zuverlässigkeitsorientierte Instandhaltung , Bayes-Theorem , statistische Qualitätskontrolle , rekursive Schätzung , stochastisches Verfahren , Unsicherheitenbehandlung , Fuzzy-Logik , Konsistenz , Monte-Carlo-Simulation , Kalman-Filter , Algorithmus
Trusted Autonomy for Spaceflight Systems
AIAA | 2005
|Blockchain Methods for Trusted Avionics Systems
IEEE | 2019
|British Library Online Contents | 2006
NAECON tutorials: Trusted systems and electronics
IEEE | 2017
|