关键词: 不确定性分析, 可靠性设计, 统计方法, 参数变异性, 贝叶斯理论, 数据优化

Abstract: Based on interpreting the importance of statistical work of soil parameters, uncertainties, the theoretical foundations and superiorities of many statistical methods are analyzed and contrasted. The research results about soil variability are summarized systematically via reviews and collections of the involved literatures at home and abroad, including index parameters, deformation parameters, strength parameters, consolidation parameters of lab tests and the parameters from CPT, VST and SPT. It is shown that the traditional method and the three-sigma rule, which are both based on the random-variable theory, will never overcome defects of model itself in spite of improvements in any case. The random field-based approaches model the soil profiles by superior random field theory and show strong vitality. It will replace the traditional method along with gradually mature of in situ tests finally. The graphical three-sigma rule, converging theoretical advance and applicability, is a worthy recommended method. The results of soil variability show that the variations of laboratory index parameters are small, the laboratory consolidation parameters are big and the other parameters from lab tests and in situ tests are in the middle. The results also show significance for selecting and optimizing parameters and reliability analysis of some important projects.

Key words: uncertainty analysis, reliability design, statistical method, variability of parameter, Bayesian theory, data optimization