Characterization of spatial variability with observed responses: application of displacement back estimation

Abstract

Soil spatial variability is difficult to evaluate due to insufficient test data. An alternative option is estimation by indirect methods such as inverse analysis. In this paper, two examples are presented to demonstrate the capability and accuracy of the probabilistic estimation method to characterize soil spatial variability with displacement responses. The first example is a soil slope subject to a surcharge load, in which the spatially varied field of the elastic modulus is estimated with displacements. The results show that estimations based on horizontal displacements were more accurate than those based on vertical displacements. The accuracy of the estimated field was substantially reduced by increasing variance of elastic modulus. However, the estimation was generally acceptable as the error was not more than 10%, even for the high variance case (COV_E=1.5). The accuracy of estimation was also affected by the type of covariance function and the correlation length. When the correlation length decreased, the accuracy of estimation was reduced. The second example is a validation of laboratory model tests where a horizontal load was applied on a layered ground. The estimated thicknesses of soil layers were close to those in the real situation, which demonstrates the capacity of the estimation method.

概要

目的

由于现场勘察和室内土工试验数据的不足,因此 土体空间变异性难以估计。通过间接方法如反演 分析方法进行估算是一个有效的途径,而土体参 数空间变异性概率反演估计的准确性受变异特 性自身影响。本文旨在通过算例研究和模型试验 验证,明确影响土体空间变异性反演准确性的关 键因素,以期为岩土勘察测试工程实践提供 参考。

创新点

1. 通过土坡空间变异性反演分析,揭示数据类型、 变异系数、相关长度和协方差函数类型等对反演 的影响;2. 室内分层土模型试验验证表明,概率 反演分析方法可有效地识别土体层厚和内摩擦 角变异性。

方法

1. 通过边坡数值算例,研究位移监测数据类型、 土体相关长度、弹性模量变异系数以及协方差函 数对弹性模量空间变异性的位移反分析的影响 (图5、6、9、11 和12)。2. 开展室内模型试验, 利用粒子图像测试技术获取位移监测数据,对分 层土体内摩擦角的变异性进行识别,并研究软弱 夹层位置与厚度对反分析的影响(图14)。

结论

1. 水平位移比竖直位移更适合用于位移反分析。 2. 反分析精度在可接受范围内,且对于高变异性 的情况(COVE=1.5),误差不超过10%;此外, 反分析精度还受协方差函数类型和相关长度的 影响。3. 反分析可识别出模型试验的土体分层, 并且对内摩擦角的估计误差小于10%。