On reducing piping uncertainties: A Bayesian decision approach

On reducing piping uncertainties: A Bayesian decision approach

Schweckendiek, T.

Vrouwenvelder, A.C.W.M. (promotor)
Vrijling, J.K. (promotor)

Civil Engineering and Geosciences

Hydraulic Engineering

2014-07-04

Dikes and levees play a crucial role in flood protection in deltaic areas such as the Netherlands. Internal erosion piping or under-seepage is a major cause of levee failures and a main contributor to the probability of failure of river levees due to the large (mostly geotechnical) uncertainties. The present thesis investigates how geotechnical uncertainties can be reduced and how we can provide input for rational investment decisions for uncertainty reduction measures such as monitoring or site investigation. The general trade-off is between investing in uncertain reduction and realizing cost reductions of retrofitting measures necessary to achieve the required reliability target. The key ingredients of the approach are Bayesian posterior and decision analysis. Posterior analysis allows us to update the piping reliability with new information; Bayesian decision analysis enables us to estimate the consequences and costs of the considered decision options. The goal of the decision analysis is to identify the strategy with the least expected cost which meets the reliability target as set by the safety standard. The essential strategy options are (a) investing in uncertainty reduction and (b) retrofitting (i.e., taking physical measures to increase the structural reliability). Within each strategy we optimize the "design parameters" such as the site investigation density or the width of piping berms. Several sources of information are investigated in this thesis, the first being field performance observations made during substantial loading conditions such as seepage or sand boils. Whereas earlier studies only considered survival information (in Dutch: "bewezen sterkte"), the proposed approach allows to incorporate much more detailed performance observations indicating good or poor performance of the levee. The case study results suggest that the probability of piping failure can decrease or increase roughly one order of magnitude depending on the prior uncertainties and the observation made. Another source of information is monitoring the response of the hydraulic head in the aquifer, which can have a considerable effect on the piping reliability, because it provides information on the geo-hydrological properties in the foundation of the levee. The same holds for site investigation such as soundings, which allow us to "map" the stratification including the thickness of the blanket layer, which is very important for the sub-mechanisms uplift and heave. Pre-posterior decision analysis enables us to determine the optimal monitoring configuration or site investigation density such that the sum of investigation cost and expected retrofitting cost is minimized. The application examples elaborated in this thesis suggest that investments in back-analysis of historical observations, monitoring and site investigation can be very cost-effective. The results also show that a framework which does not consider the benefits of risk reduction beyond meeting the reliability target is sub-optimal in an overall risk sense.

flood defenses
reliability
internal erosion
uplift
heave
piping
risk-based decision support
Bayesian updating
field observations
pore pressures
monitoring
site investigation
random fields
kriging
conditional simulation
decision analysis

https://doi.org/10.4233/uuid:f9be2f7e-7009-4c73-afe5-8b4bb16e956f

9789053358801

Institutional Repository

doctoral thesis

(c) 2014 Schweckendiek, T.