Reliability-based performance of embankments improved with deep mixing considering spatial variability of material properties

Geotechnical structures are generally designed using the deterministic approach, where uniform properties are used over the material space. This approach is not appropriate for deep cement mixed (DCM) ground due to the highly variable nature of material properties within the cement mixed soil. Although a large overdesign factor can reduce the risk of unsatisfactory performance, this method does not provide a systematic approach. Proper understanding of the variability-dependent performance of DCM ground is important to specify the design strength as well as to specify the quality controlling criteria. Therefore, in this paper, the reliability-based performance of an embankment stabilized using DCM wall panels underneath the side slopes was investigated. The spatial variability of elastic modulus and shear strength properties was considered. The strength variation was specified using the coefficient of variation (COV), mean strength, and the spatial correlation length. Material properties for different cases were randomly generated from a lognormal distribution with a specific mean strength and a COV using a computer program written in MATLAB. The ABAQUS finite element analysis was executed for each case within this MATLAB program and the reliability was determined using 1,500 Monte Carlo realizations. Results demonstrate that the spatial correlation length of strength properties has a great influence on the reliability-based performance of the embankment. The most critical probability of unsatisfactory performance was observed when the spatial correlation length is 0.18 times the width of DCM wall panels, at all mean strengths and COVs. The lateral deformation data obtained from Monte Carlo simulations showed increasing upper bound with increasing spatial correlation length and COV. The partial factor of safety (PFOS) has significantly contributed to the skewness of the deformation distribution without any change to the upper and lower bounds of the distribution.