UBC Theses and Dissertations
Statistical characterization of soil profiles using in situ tests Wickremesinghe, Damika Sampath
Several statistical procedures that would enhance the site characterization capabilities of insitu test data with special emphasis on the cone penetrometer test have been proposed and presented. Two methods to identify different soil layers from a profile have been described. One of these procedures is based on the effects of the individual parameters, namely, cone bearing, sleeve friction and pore pressure, while the other method employs a multivariate scheme of analysis, which has the capability of handling all three or any two parameters, simultaneously. The advantages of these statistical methods over the conventional methods of soil layer identification, have also been highlighted. Critical levels of the values of the Intraclass Correlation coefficient and the D statistic have been proposed for the identification of layer boundaries as primary or secondary for both sand and clay type soils. Methods of trend analysis have been proposed while the complications arising from the presence of correlations have been discussed. The role played by methods of statistical filtering and smoothing, in the identification of trends, have also been illustrated. Statistical procedures have been proposed, for the purpose of verification of non-stationarity or stationarity, in the event it cannot be determined from a visual inspection. The need for the consideration of geotechnical data as random has been emphasized, together with applications of random field theory in the determination of exceedance probabilities of given threshold values over spatial averages of a soil layer. A computationally more convenient method for the determination of the scale of fluctuation has been proposed while emphasizing its importance in several areas of applications, with respect to the cone penetration test. Time Series methods have been employed in order to model the stationary component of soil profiles and also have been extended to obtain the measurement noise of different test methods. A comparison of the measurement noise of different insitu testing devices, obtained by the time series method has been compared to a procedure based solely on the autocorrelation function of the data, resulting in a good agreement. The relatively low value of measurement noise obtained for the cone penetration test confirms its superiority over other insitu testing methods like the field vane test which gave fairly high estimates of the measurement noise. A two dimensional interpolation procedure considering the correlation between data points has been recommended. This procedure which uses the autocorrelation function, has been applied to a set of cone penetrometer test data and the results of which have been compared with the actual profile at that location. The reasonable comparison of the predicted with the actual, clearly indicate the need for the consideration of correlations if they do exist, in interpolating geotechnical data in two or three dimensions. IBM - PC compatible interactive micro computer programs have been developed in order to perform most of the techniques proposed in the thesis. These programs cater to any type of data format and have several inbuilt options available to the user. Detailed user manuals for these programs are also available.
Item Citations and Data