Statistical Analysis of Core Strength of Deep Cement Mixing Ground

Abstract

Cement-treated ground by deep cement mixing has been used for several geotechnical structures widely. In quality assurance processes of this method, statistical parameters, mean and standard deviation, of unconfined compressive strength of cement-treated soils are used. The mean and standard deviation are normally used to assure the quality of the improved ground. These parameters evaluated from core strength data are the sample statistical parameters, indicating these parameters involve the statistical uncertainty. Thus the evaluation of the statistical uncertainty is needed when assuring the quality of the improved ground precisely. Moreover, the spatial correlation exists in core strength data. The statistical uncertainty emerging in the evaluation of the population statistical parameters is possibly affected by the spatial correlation. This paper presents the statistical analysis of core strength data observed in several deep cement mixing projects. The mean, standard deviation, and autocorrelation distance, were adopted as the statistical parameters of the strength. The type of the probability distribution of the core strength was investigated by the Kolmogoronv-Smirnov (K-S) test. The goodness fit of the normal and log-normal distributions was examined against the core strength data. The autocorrelation distance, which is the parameter representing the characteristic of the spatial correlation, was calculated from the distribution of the core strength using the maximum-likelihood method. The statistical uncertainty of the statistical parameters was evaluated using a Bayesian inference approach. In the Bayesian inference approach, a Markov chain Monte Carlo method was adopted to calculate the realizations of the population statistical parameters. The analysis results indicated the statistical uncertainty included in the statistical parameters is significantly affected by the spatial correlation.