The Atterberg limits are the index properties that most frequently used in the soil identification and classification. The soil liquid limit is the moisture content in which the soil changes from a plastic state to a liquid state. However, with the same physical definition, the liquid limit values obtained by various methods have shown to be not the same. In this study, three different testing methods were performed to determine the liquid limit values, namely the hard base Casagrande apparatus method, the soft base Casagrande apparatus method and the cone penetrometer method. The principle objective of this research is to study the uncertainty of the liquid limit values obtained by the three different testing methods.

The study is comprised of three different analyses of the li...[ Read more ]

The Atterberg limits are the index properties that most frequently used in the soil identification and classification. The soil liquid limit is the moisture content in which the soil changes from a plastic state to a liquid state. However, with the same physical definition, the liquid limit values obtained by various methods have shown to be not the same. In this study, three different testing methods were performed to determine the liquid limit values, namely the hard base Casagrande apparatus method, the soft base Casagrande apparatus method and the cone penetrometer method. The principle objective of this research is to study the uncertainty of the liquid limit values obtained by the three different testing methods.

The study is comprised of three different analyses of the liquid limit values. They are the comparison of the liquid limit values obtained by different testing methods, the application of single-factor analysis of variance (ANOVA) in the comparison of the liquid limit values obtained by the different testing methods and, lastly, the modeling and simulating of the liquid limit values obtained from the different testing methods with the use of Gaussian copula. The laboratory tests of liquid limit values from 17 soil samples were performed in order to obtain reliable experimental data for the study. Additional to that, a database has been developed which consists of over 200 data points from the experimental results obtained by different methods from various sources of soil samples from literature.

The comparison of the test result values reveals that the majority of the experimental data having higher liquid limit values which were obtained by the cone penetrometer method. Also, the % of error of the liquid limit values obtained by the hard base Casagrande apparatus method were shown to have a range between -26% and 70% with those obtained by the cone penetrometer method, while the % of error of the similar comparison between the liquid limit values obtained by the soft base Casagrande apparatus method and those obtained by the cone penetrometer method is ranging from -20% to 31.5%. Some regression models modeling the results obtained by the three different testing methods have also been proposed and discussed.

Subsequently, the single-factor ANOVA has been performed on the experimental results of liquid limit values for comparing the mean values of the liquid limits obtained by the three different testing methods. The null hypotheses that all of the mean values of liquid limit values obtained by the three different methods are the same have been proposed. And the analyses reveal that the hypotheses should be rejected in all of the 13 cases demonstrated in the study.

Furthermore, the modeling and simulating of the liquid limit values obtained by the three different testing methods has been demonstrated by adopting a relatively new approach, the copula approach. The empirical results of liquid limit values of commercially available soils and that of natural soils have been separated in two cases. The joint distributions of the liquid limit values from the two categories of soil samples have been modeled in order to investigate the uncertainty of liquid limit values obtained by different testing methods in these two categories. Moreover, the simulations of the liquid limit values with the joint distributions determined by the copula method have been revealed for further applications of the probability analyses.

Finally, the newly proposed image analysis program for the Casagrande apparatus test has been adopted in order to determine the closure of the groove more accurately. This is the first application of image analysis being implemented to the Casagrande apparatus method for the determination of the liquid limit values. This program aims at eliminating the testing error from the inaccurate eye judgement of the operators. The performance of the image analysis program was found to be satisfactory in the Casagrande apparatus test for soils in light colour.