Deterministic and probabilistic prediction of facing deformations of geosynthetic-reinforced MSE walls using a response surface approach

Abstract

This paper describes a strategy to compute deterministic and probabilistic estimates of maximum outward deformation of mechanically stabilized earth (MSE) modular block wall structures of particular type, materials and boundary conditions. The reference walls in this study were a group of MSE modular block walls constructed under carefully controlled laboratory conditions. A numerical FLAC model was first demonstrated to give reasonable estimates of measured maximum wall deformations. The model was then used to carry out a sensitivity analysis using a modest number of simulations with different values of key input parameters. Six key control parameters were selected based on the results of the sensitivity analysis and a review of the literature on wall performance. A response surface of full quadratic form was used to predict maximum normalized wall deformations as a function of the six key parameters. The results of a second set of 729 numerical simulations corresponding to the maximum number of possible permutations of three different values for each of the six control parameters were used to find the coefficient terms for the response surface equation using a conventional least squares method. The response surface method (RSM) equation was then used to carry out a large number of Monte Carlo simulations using random sampling of distributions for soil friction angle, surcharge pressure, block–block interface shear stiffness and reinforcement stiffness input parameters. The resulting statistics were used to construct a cumulative distribution function (CDF) for maximum normalized wall deformations. The CDF is then used to link reliability performance-based design to a reasonable range of serviceability criteria for wall deformations generated under operational conditions.