Bayesian Learning of Site-Specific Spatial Variability Using Sparse Geotechnical Data

Research output: Chapters, Conference Papers, Creative and Literary Works (RGC: 12, 32, 41, 45)32_Refereed conference paper (with ISBN/ISSN)peer-review

View graph of relations

Detail(s)

Original languageEnglish
Title of host publicationProceedings of the 7th International Symposium on Geotechnical Safety and Risk (ISGSR 2019)
EditorsJianye Ching, Dian-Qing Li, Jie Zhang
PublisherResearch Publishing (S) Pte. Ltd.
Pages553-558
Number of pages6
ISBN (Electronic)9789811127250
Publication statusPublished - Dec 2019

Conference

Title7th International Symposium on Geotechnical Safety and Risk (ISGSR 2019)
LocationNational Taiwan University of Science and Technology
PlaceTaiwan
CityTaipei
Period11 - 13 December 2019

Abstract

Geo-materials (e.g., soils and rocks) are natural materials, and their properties are affected by many spatially varying factors during their complex geological formation process, such as the textures of their parent materials, weathering processes, transportation agents, and sedimentation conditions. Geo-materials therefore may have different properties at different locations at a specific site. This spatial variability, to some extent, are unique and deterministic in every site as an outcome of the previous geological processes that the soils and rocks at a specific site have undergone. In recent years, random field model has been a popular tool to model spatial variability of geotechnical properties. In practice, measurement data at a specific site are used to estimate random field parameters, such as mean and standard deviation, as well as parameters (e.g., scale of fluctuation) of a pre-specified parametric form of correlation function. Estimation of these random field parameters generally require extensive measurements from a specific site, which are usually not available in geotechnical engineering practice. In addition, the underlying correlation function form is often unknown for a specific site, therefore it is also challenging to select the most suitable parametric function form of the correlation function when only limited measurements are available. Moreover, even if geotechnical properties are measured at every location, the conventional random field still produces stochastic sample in each realization rather than converging to the unique and deterministic spatial variation in the specific site. To address these issues, this paper presents a Bayesian learning method for random field modeling of site-specific spatial variability of geotechnical properties, which is data-driven and non-parametric. The method is developed based on supervised machine learning, and it bypasses the estimation of parametric correlation function. The learning results are demonstrated by generating random field samples of site-specific spatial variability. Simulated data is used to illustrate the performance of the method. Comparative study on conventional random field model is also provided. The results show that the Bayesian learning method performs well even when measurements are sparse.

Citation Format(s)

Bayesian Learning of Site-Specific Spatial Variability Using Sparse Geotechnical Data. / Hu, Yue; Zhao, Tengyuan; Wang, Yu.

Proceedings of the 7th International Symposium on Geotechnical Safety and Risk (ISGSR 2019). ed. / Jianye Ching; Dian-Qing Li; Jie Zhang. Research Publishing (S) Pte. Ltd., 2019. p. 553-558.

Research output: Chapters, Conference Papers, Creative and Literary Works (RGC: 12, 32, 41, 45)32_Refereed conference paper (with ISBN/ISSN)peer-review