1. DATA-CENTRIC QUASI-SITE-SPECIFIC PREDICTION FOR SOIL MODULUS.
- Author
-
Jianye Ching and Ming-Chieh Kuo
- Subjects
DATABASES ,SOILS ,CONFIDENCE intervals - Abstract
This paper compiles a new soil property database called SOIL-DMT/8/7186 that contains 8 parameters of 7,186 soil records from 701 sites worldwide, including the modulus parameter obtained by the dilatometer test (DMT). With this database, the paper demonstrates how to construct a quasi-site-specific model that can predict the soil modulus for a target site (the Bothkennar test site, UK). First, the hierarchical Bayesian model (HBM) developed by the first author is used to learn the site-specific characteristics of the 701 sites in the database. The learned HBM can produce a prior model for the target site. Then, this prior model is further updated by sparse target-site data into a (posterior) quasi-site-specific model. The quasi-site-specific model can then be used to predict the soil modulus of the target site based on inexpensive site investigation data such as SPT blow count and CPT cone tip resistance. A remarkable observation is that the effectiveness of the quasi-site-specific model generally depends on the soil property database: the quasi-site-specific model based on a database more relevant to the target site can produce a narrower 95% confidence interval for the soil modulus. Moreover, it is found that a small database that contains only a few sites relevant to the target site may still perform well if these sites are properly chosen. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF