Time-dependent reliability analysis of flood defences

• Main Author: Buijs, Foekje Akke
• Published: University of Newcastle Upon Tyne 2008
• Subjects: 627
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.489304

The aim of this thesis is to investigate how the time-dependent behaviour of flood defence properties can be appropriately characterised and incorporated in a reliabilitybased approach. Such an approach is required in a maintenance optimisation framework for flood defence management. The first objective shows that existing structural reliability methods are suitable for the analysis and incorporation of asset time-dependent processes in flood defence (system) reliability. Recent progress on quantitative maintenance optimisation frameworks for flood defence management is drawn together and complemented by theory from other engineering disciplines. The second objective develops three importance measure types to indicate the relevance of the time-dependent processes in the context of a rational maintenance optimisation approach. These importance measures support practical operational management as well as maintenance optimisation model design. The third objective develops a modelling methodology to describe asset time-dependent processes of flood defences by a statistical model. The first phase in the modelling methodology is problem formulation. The second conceptualisation phase is a five-step analysis of the asset time-dependent process. Firstly, existing field observations and scientific understanding are assembled. Secondly, the excitation, ancillary and affected features and uncertainty types of the asset time-dependent process are analysed. The third step describes the character of the process conditional on the excitation. The fourth step analyses the dependencies between different asset time-dependent processes. The fifth step formulates alternative statistical models for the asset time-dependent process. The last phase in the modelling methodology is parameter estimation, calibration and model corroboration. Historical observations on asset time-dependent processes are scarce and can either be used for further extension of this phase or Bayesian posterior updating. The fourth objective demonstrates the methods developed in this thesis in a (system) reliability model of the Dartford Creek to Swanscombe Marshes flood defence system along the Thames Estuary.