中文摘要：

实际工程中岩土参数的分布区间均为有限区间，左端不可能为负值，右端不可能无限大．利用现有的经典分布函数（包括正态分布、对数正态分布、贝塔分布、威布尔分布、极值分布等）来拟合岩土参数的实际分布时，存在区间不匹配的困难，必须进行截尾处理，进而导致截尾区间内概率累积值永远小于1．为此，选取了5种截尾区间，根据5组代表性岩土参数的实测样本，基于正态信息扩散原理对其相应的概率密度函数进行推断，利用K—S检验法对所得的概率分布函数进行检验．通过区间取值匹配、拟合检验值、累积概率值和概率函数曲线的比较，提出了以3口统计原理为基础，同时考虑偏度进行调整的截尾区间确定标准．研究结果表明：在确定的截尾区间内，正态信息扩散推断方法所得概率分布的检验值均优于传统分布，而且在误差为10^（-5）量级水平上满足概率累积值等于1的要求．此外，正态信息扩散分布能够反映岩土参数实际分布的随机波动，也克服了单峰型分布的不足．

英文摘要：

In practical engineering, the distribution interval of geotechnical parameters is limited. The difficulties of interval mismatch will be faced when the actual distribution of geotechnical parameters are fitting with the classical distribution including normal distribution, lognormal distribution, Beta distribution, Weibull distribution and extreme value distribution; and truncation processing is required; however, which will lead to the probability cumulative value in truncation interval is always less than 1. In this paper, five truncated intervals are selected; and based on 5 groups measured sample of representative geotechnical parameters, the corresponding probability＇density functions are deduced with the normal information diffusion method （NID method） ; and the Kolmogorov-Smirnov method is used to test the goodness-of-fitting. Through the comparison of the interval value matching, fitting test values, cumulative probability values and probability function curves, a new integral interval standard combining 3σ statistical principle and the effect of skewness is put forward. The results show that in the determined truncated interval, the random fluctuation of the actual sample distribution of geotechnical parameters can be reflected well with the normal information diffusion distribution. The test value of the probability distribution with NID method will less than that of the probability distribution with the classical distribution fitting method, and the requirement that the cumulative probability value is equal to 1 with the fitting error in the 10^-5 magnitude level is also satisfied. In addition, the random fluctuation of the actual sample distribution of geotechnical parame- ters can be reflected well with the normal information diffusion distribution, which overcomes the short- comings of single peak distribution.