Use of uncertainty methodology in identification and classification of soils based upon CPT

摘要

The current Cone Penetration Test (CPT) soil engineering classifications have two kinds of uncertainties: randomness and fuzziness. Research indicates that possible solutions to these uncertainties can only be worked out through modeling them. Therefore, a systematic investigation is performed and some preparative tasks are done in advance. First, an efficient soil classification index, U, is defined and several CPT soil classification charts are simplified accordingly. Second, a moving window approach based upon an Intraclass Correlation Coefficient (ICC) to determine normal soil behavior units is adopted so that a correlation between soil types and soil behavior units can be established. Based upon these approaches, a preliminary data reduction is performed on the raw CPT data from eight sites, and the characteristics and distributions of the soil behavior units are determined and discussed for seven soil types.;Two statistical criteria, Region Estimation and Point Estimation, based upon distributions of soil behavior units are then developed to predict soil type using CPT data. Also, a fuzzy subset approach is suggested to handle the fuzziness and randomness. In this CPT fuzzy soil engineering classification, a new naming system is used. The randomness of CPT soil engineering classification is put into the conceptual framework of three new soil types. The fuzziness is then described by fuzzy membership functions. These functions are derived from the modification of the density functions of corresponding compositional soil groups.;Finally, a new package of CPT soil engineering classification is suggested. It consists of following procedures: (1) Transform a CPT sounding profile of parameters (tip resistance, q$sb{rm c},$ and friction ratio, FR) by conformal mapping to a corresponding profile of soil classification index, U; (2) Layer the U profile by ICC moving window method and calculate the mean of U values for each layer to determine the soil behavior unit; (3) Predict the soil type of each layer by matching the soil behavior unit of that layer with the classification criteria suggested in this study.;Several sets of CPT soil engineering classification criteria are recommended in this dissertation. They are the indicators of an evolution process from the purely empirical to the purely theoretical.