Derivation of a Statistical Framework for Strength Prediction in Deep Cement Mixing

Abstract

Deep cement mixing (DCM) is widely used all over the world. The variability of the properties of natural soils, combined with uncertainty and inaccuracy of construction operation of deep soil mixing, leads to nonuniformity of the binder distribution in the deep-mixed soil mass, therefore, the often highly variable strength. This has been a problem in construction practice. This paper introduces the derivation of a statistical framework for strength prediction in DCM, based on a statistical analysis of a large quantity of sample measurements in a series of centrifuge model tests on DCM reported in previous literatures. Through the derivation, beta distribution is verified and proposed to fit the variation of DCM binder concentration. Moreover, phase relationships within the DCM column are theoretically established. Based on an empirical relationship of unconfined compressive strength, soil cement ratio and water cement ratio, and the phase relationships, a statistical framework for the strength prediction of DCM columns has been derived from the distribution of binder concentration. This statistical framework allows engineers and researchers a first-cut estimation of the mean, standard deviation, and PDF of the strength of DCM columns, during design and construction of DCM, even before field trials are undertaken. Preliminary case study is partially presented.