土体性质互相关性的估计以及不确定性的量化分析

Quantification of Statistical Uncertainty for Cross-correlation between Multiple Soil Properties

土体参数如土体黏聚力、内摩擦角之间存在互相关性,这一特性对岩土设计分析具有重要意义。互相关性可以用互相关系数表征。准确的计算这一系数需要较多试验数据,而实际工程中岩土的试验数据较少。这就导致所得互相关系数的不确定性较大。为此,介绍了如何用自助抽样法去量化这一不确定性,并以工程实例说明。

Soil properties such as cohesion(c') and internal friction angle(φ') often exhibit cross-correlation, which has been shown to greatly affect geotechnical design. The cross-correlation between c' and φ' is characterized by a cross-correlation coefficient ρ, reliable estimation of which, however, generally requires extensive data. In practice, the number of measurement data is usually sparse, leading to great statistical uncertainty in the estimated ρ. This paper adopts the bootstrap method, using which the statistical uncertainty in estimated ρ is objectively and explicitly quantified, c' and φ' data from a real site are adopted for illustration.