Parameter Correlation Influence on Failure Probability Estimation for Multi-Functional Flood Defenses (PAP014365)

Código

PAP014365

Autores

Juan Pablo Aguilar Lopez, Ralph Schielen, Jord Warmink, Suzanne Hulscher

Tema

Flood risk management in mega cities

Resumo

Most relevant structural failure mechanisms of flood defenses that can occur during a flood event have been studied trough time by Dutch national authorities and practitioners. The addition of an extra structure to the embankments as a multifunctional approach has also been in the mind of water managers and architects for quite some time already. However, these additional structures can compromise the system safety as their own failure mechanisms could be triggered during a flood event as well. Most failure mechanisms of flood defense and non-flood defense structures, are described by different numerical models that recreate the physical processes experienced for certain load conditions. Geometrical, geotechnical and environmental data is required as input in these models to design and asses the reliability of these constructions. One of the biggest sources of uncertainty for estimating the probability of failure of any structure, comes when selecting the appropriate parameter values and statistical distributions of the input. This means, that the possible correlation of these input must also be taken into account when modeling a certain failure mechanism. At the moment the effects of spatial and temporal correlation of single variables is considered in the reliability analysis of conventional flood defenses. But for the case of multi-functional flood defenses, dependency between 2 different variables might be of great importance as well. The investigation of the possible correlation effects between these parameters is needed as it this dependency may change the outcome of the estimated probability of failure in unknown ways. Correlation is not only the degree of one variable to be able to describe another one but also the behavior in terms of linearity, monotonicity and robustness just to mention some of the most relevant. For example a variable can be more strongly correlated for extreme values than for lower ones even when describing an incremental behavior. Therefore the actual failure probability during an extreme event is bigger than the one estimated if two variables are assumed non correlated. In order to represent this possible relationships between parameters, the implementation of Copulas" or joint probability distributions implemented during the modeling processes for the different failure mechanisms. With this study it is intended to evaluate the risk of failure of a flood defense with an embedded structure, in a probabilistic manner while taking into account the correlation of variables that in practice are assumed independent. The outcome of this study will help designers and managers to understand which failure mechanisms might increase or reduce their probability of occurrence after embedding infrastructure into the actual flood defenses."