Probabilistic assessment of an earth dam stability design using the adaptive polynomial chaos expansion

Abstract

Assessment of the stability for an earth dam under seismic loadings is presented in a probabilistic framework. The geotechnical parameters measured at the dam site permit a realistic description of the soil variability for the dam and the superficial foundation. An active learning sparse polynomial chaos expansions (A-bSPCE) combined with Monte Carlo simulation is employed to quantify the uncertainty propagation. The proposed dam design is assessed within two seismic scenarios: the first one considers a constant pseudo-static acceleration while the second one accounts for the uncertainties in seismic loadings by modeling the pseudo-static acceleration as a random variable. The probabilistic assessments for variation of the amplitude and standard deviation of the seismic loading are performed. In addition, the effects of the correlation coefficient between the soil’s shear strength parameters and the soil spatial variability are discussed. It is found that these effects are more significant in the first scenario. The obtained results permit an evaluation of the preliminary dam design and to provide guidance on seismic safety enhancement for the final design phase. The present study also highlights the accuracy and efficiency of the employed reliability method. It is shown that the total computational time of a probabilistic analysis can be reduced from several days to only 20 min by using the A-bSPCE.