Integrating Multi-Source Information from Site Investigation for Probabilistic Characterization of Undrained Shear Strength
Integrating Multi-Source Information from Site Investigation for Probabilistic Characterization of Undrained Shear Strength
复制标题
DOI：
10.1061/9780784480724.005
复制
发表时间：
2017-06
期刊：
影响因子：
--
通讯作者：
Meng-Yao Shen;Z. Cao;Dianqing Li;Yu Wang
中科院分区：
文献类型：
--
作者：
Meng-Yao Shen;Z. Cao;Dianqing Li;Yu Wang
本文提出了一种贝叶斯序贯更新（BSU）方法，集成了从岩土工程现场调查获得的多源信息的概率表征粘土的不排水抗剪强度。在此，多源信息包括在项目之前可获得的知识（即先验知识）和来自各种测试程序的测试结果，例如超固结比（OCR）、标准贯入试验（SPT）和圆锥贯入试验（CPT）数据。在这项研究中，OCR，SPT和CPT数据顺序纳入BSU框架更新的知识onsufor确定其特定网站的统计和概率分布。BSU框架允许使用来自不同位置的不同测试程序的多种类型的测试结果。所提出的方法进行了说明和验证，使用OCR，SPT和CPT数据模拟从虚拟粘土网站，其中真正的统计和概率分布的suare已知。结果表明，所提出的BSU方法结合先验知识与多种类型的测试结果在一个一致的和系统的方式，它提供了合理的统计估计的岩土参数的组合信息的基础上。此外，还利用模拟数据探讨了数据质量和数量的影响。
This paper presents a Bayesian sequential updating (BSU) approach that integrates multi-source information obtained from geotechnical site investigation for probabilistic characterization of undrained shear strengthsuof clay. Herein, the multi-source information includes the knowledge available prior to the project (namely prior knowledge) and test results from various testing procedures, such as overconsolidation ratio (OCR), standard penetration test (SPT), and cone penetration test (CPT) data. In this study, the OCR, SPT, and CPT data are sequentially incorporated into a BSU framework to update the knowledge onsufor determination of its site-specific statistics and probability distributions. The BSU framework allows using multiple types of test results from different test procedures at different locations. The proposed approach is illustrated and validated using OCR, SPT and CPT data simulated from a virtual clay site, where true statistics and probability distributions ofsuare known. Results showed that the proposed BSU approach combines prior knowledge with multiple types of test results in a consistent and systematic manner, and it provides reasonable statistical estimates of geotechnical parameters based on the combined information. In addition, effects of data quality and quantity are also explored using simulated data.