A statistical model for the unconfined compressive strength of deep-mixed columns

Abstract

The strength of soil improved by deep mixing processes often exhibits significant spatial variability. This paper presents a statistical model for strength distribution in deep-mixed columns, based on the assumption that spatial variation in strength arises from spatial variation in binder mass fraction. The statistical inputs of the model are the probability density function and coefficient of variation of the binder distribution, both of which were obtained from centrifuge model data and were assumed to be independent of the amount of the binder introduced into the ground. The theoretical framework of the model is first developed, followed by an assessment of the probability density function for the binder distribution. The variation of binder mass fraction within a deep mixing column is fitted by a truncated normal distribution. The model is benchmarked against core strength data from three phases of deep mixing operation in a construction project. The simplicity of the model precludes consideration of the effects of detailed mixing blade configuration, in situ variation in soil properties as well as workmanship. In spite of this, the computed histograms of strength distribution gave a reasonable reflection of the measured strength distribution. This suggests that despite the variety and complexity of deep mixing processes, one may still be able to develop a simple and workable model of the strength distribution taking into account just a few important factors.