Bayesian consideration of unknown sensor characteristics in fatigue-related structural health monitoring
Publication date
2019
Document type
Research article
Author
Berg, Thomas
Ende, Sven von
Lammering, Rolf
Organisational unit
ISSN
Series or journal
Probabilistic Engineering Mechanics
Periodical volume
56
First page
71
Last page
81
Peer-reviewed
✅
Part of the university bibliography
✅
Keyword
Fatigue Crack Propagation
Fatigue Testing
Probability Density Function
Abstract
In structural health monitoring, Bayesian updating is widely utilised in the analysis of noisy sequential data of dynamic systems with the objective of determining the state of damage of a structure or identifying its unknown dynamic characteristics or both. In the present work, this approach is enhanced to encompass the simultaneous handling of insufficient knowledge of sensor features – i.e. a non-applicable relation between state of damage and observations due to high uncertainty introduced by unknown measuring parameters – while given the nature of damage propagation as in fatigue-driven applications. Thereby, the statistical inversion problem of inferring unknown states of damage as well as unknown measuring and dynamic model parameters is addressed solely on the basis of observations and parameter-dependent functional expressions linking these quantities. As Bayesian updating provides the posterior belief on the unknown quantities in form of probability density functions, the question of state observability and parameter identifiability can be approached simultaneously. The methodology is applied to potential-drop measuring in a fatigue-loading scenario and its effectiveness is successfully demonstrated. © 2019 Elsevier Ltd
Cite as
Enthalten in: Probabilistic engineering mechanics. - Amsterdam [u.a.] : Elsevier Science, 1986. - Online-Ressource. - Bd. 56.2019
Version
Not applicable (or unknown)
Access right on openHSU
Metadata only access