Probabilistic-Based Analysis of MSE Walls Using the Latin Hypercube Sampling Method

Traditionally, factors of safety have been used to design mechanically stabilized earth (MSE) walls; however, the uncertainties associated with input parameters have not been considered. The probabilistic analysis of MSE walls was an appropriate approach to consider all uncertainties. The authors of this paper investigated the performance of MSE walls by using reliability and sensitivity analyses. A finite-element model of an MSE wall was developed and verified on the basis of field measurements of a reference wall to approximate performance functions. The failure probability of the MSE wall was determined using the first-order second-moment (FOSM), first-order reliability method (FORM), and improved Latin hypercube sampling (iLHS) method, considering external and internal stability in addition to the horizontal wall displacement limit state. Uncertainties associated with soil and reinforcement properties, soil–geosynthetic interface friction angle, and the magnitude of surcharge load were considered for the reliability analysis. Probabilistic characteristics of soil unit weight, soil friction angle, soil–geosynthetic friction angle, geosynthetic tensile strength, and modulus of elasticity of the geosynthetic were experimentally determined, and over 500 tests were performed. A sensitivity analysis was then performed through the FOSM and FORM, and the most effective random variables were determined according to each limit state function. The reliability analysis revealed that, although the FOSM and FORM were appropriate methods to determine the initial approximation of the failure probability of a MSE wall, the iLHS method was more practical and accurate for this analysis because of the nonlinearity of the limit states. Furthermore, the sensitivity analysis results indicated that the most effective parameter in sliding and pullout failure was the soil–geosynthetic interface friction angle, and in overturning, it was the soil friction angle. Moreover, the most effective parameter, second only to the tensile strength of the geosynthetic, was the surcharge load in the rupture limit state. In conclusion, the most significant random variables for horizontal wall displacement in the sequence were the modulus of elasticity of the geosynthetic, magnitude of the surcharge load, and soil friction angle.