- Title
- Integrating multiple sources of data for inhomogeneous soil profile
- Creator
- Xie, Jiawei
- Relation
- University of Newcastle Research Higher Degree Thesis
- Resource Type
- thesis
- Date
- 2023
- Description
- Research Doctorate - Doctor of Philosophy (PhD)
- Description
- Due to the complexity of the geological process, natural soils can be highly variable in space. However, in most cases, only sparse measurement data can be achieved for characterizing their geotechnical properties. Hence, it is needed to use limited data to estimate the properties of a large area. It is therefore a challenging task to better site characterization by considering the spatial correlation in the geotechnical properties and making the efficient use of multiple sources of testing data to enhance the reliability of the interpreted results. To address these challenges, various methods have been developed in this thesis to make efficient use of sparse site investigation data. For single-source sparse measurement data, the Gaussian process regression method has been investigated in which spatial auto-correlation can be explicitly considered. The kernel selection and parameter estimation issues have been extensively studied. Then a novel geotechnical distance field based extra trees method is proposed to interpolate the sparse measurement data based on implicitly considering the spatial correlation. This method can provide quick and accurate estimates of soil properties and includes a way to quantify associated prediction uncertainty, allowing for objective evaluation of the reliability of subsurface modeling results.Several solutions have been proposed to integrate multiple sources of measurement data. A Bayesian updating method is utilized to integrate laboratory testing data and geophysical data considering the measurement errors. Then the intrinsic collocated cokriging method is utilized to integrate geotechnical data (cone penetration test (CPT)) and geophysical data (Multi-channel analysis surface wave test (MASW))). The geophysical data is utilized to estimate the horizontal scale of fluctuation which is very difficult to be estimated based on geotechnical data only. A multi-level machine learning framework is proposed to upgrade the resolution of the subsurface well log data based on core data. Combining these two sources of data offers a cost-effective way to achieve higher-resolution test results. Finally, a new solution to integrate the borehole and CPT data with the tree-based method is proposed. A probabilistic mapping matrix between the unified soil classification system and soil behavior type classification system (USCS-SBT) is built based on a collected municipal database with collocated borehole and CPT data. A boundary dictionary method is proposed to improve the performance of tree-based models on complex soil layer boundary conditions. The optimal distribution of soil layers can be selected based on combining multiple borehole information and pruning the structure of trees.
- Subject
- geotechnical site investigation; data integration; sparse data interpolation; spatial correlation; machine learning; thesis by publication
- Identifier
- http://hdl.handle.net/1959.13/1508409
- Identifier
- uon:56123
- Rights
- Copyright 2023 Jiawei Xie
- Language
- eng
- Full Text
- Hits: 331
- Visitors: 370
- Downloads: 48
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | ATTACHMENT01 | Thesis | 24 MB | Adobe Acrobat PDF | View Details Download | ||
| View Details Download | ATTACHMENT02 | Abstract | 125 KB | Adobe Acrobat PDF | View Details Download |