基于非参数回归的土层参数反演分析算法

Back Analysis Algorithm of Soil Parameter Based on Non-parametric Regression

在岩土工程设计过程中,由于存在试验条件、土体性质变异性、土层取样方法以及试样扰动等不确定性因素,使得现场试验得到的土层参数结果具有离散性,而室内结果往往与原位土体的性质有一定差异。为此,提出一种土层参数反演分析算法,选取实测目标,结合均匀试验和非参数回归,建立土层参数与实测目标之间的高效响应面,对每个工况在一定范围内枚举出的大量土层参数样本进行计算得到目标计算值。考虑到目标实测值测量的不确定性,结合可变容差法,按照工况逐次对目标计算值进行筛分,从而得到符合条件的计算值对应的土层参数取值区间。基坑工程算例表明了该算法的可行性。

Because of the uncertainty factors such as test condition,the variability of soil properties,soil sampling methods,sample disturbance and so on,test results of soil parameters obtained by field test have the discreteness. And results obtained by laboratory test have some difference with in-situ soil properties. This paper puts forw ard a kind of back analysis algorithm of soil parameters. The measured targets are selected. Combining with uniform test and non-parametric regression,the efficient response surface betw een soil parameters and the measured targets is established. The calculated values of targets are obtained by calculating soil parameters samples enumerated in the definite range for each working condition. Considering the measurement uncertainty of the target values,the calculated values are screened by combining with the variable tolerance method according to the measured target values and the order of working conditions. The ranges of soil parameters that correspond to the qualified calculated values are obtained. A foundation pit engineering example is used to show the feasibility of the algorithm.