Field Measurements-Based Probabilistic Stability Analysis of an Earth Dam

Soil property uncertainties are unavoidable for the design/evaluation of earth dams, which are large structures made of naturally variable materials (i.e., soils/rocks). This paper evaluates the stability of an existing earth dam under the consideration of such uncertainties by means of probabilistic analysis. The measurements collected during the dam construction or from laboratory tests are exploited to model the relevant soil properties by random variables or random fields. One objective of the current study lies in showing a probabilistic analysis procedure for practical engineering including soil variability modelling and reliability analysis. The considered dam problem is high dimensional with over 2000 input variables due to the discretization of random fields and the low autocorrelations distances. Five reliability methods, including sampling methods and surrogate-based techniques, are compared, aiming to investigate their accuracy and efficiency when dealing with such challenging problems. The presented results are expected to be useful for engineers to probabilistically analyze earth dams; some recommendations about method selection and soil spatial modelling for the reliability analysis using random fields are also available.

Keywords: Earth dam, Reliability analysis, Spatial variability modelling, Monte Carlo Simulation