Quantifying Site-specific Geotechnical Variability by Markov Chain Monte Carlo Simulation

Project: Research

View graph of relations


Soils and rocks are natural geo-materials, whose properties are affected by manyspatially-varying factors during the geological process, such as the properties of theirparent materials, weathering and erosion processes, transportation agents, andsedimentation conditions. In-situ geotechnical properties therefore vary spatially andsuch variability is site-specific/project-specific in nature. During geotechnical siteinvestigation, however, only a small portion of geo-materials is examined. The amount ofgeotechnical data is generally too sparse to sufficiently characterize geotechnicalvariability of the site. This can lead to a difficult situation commonly referred to as“unexpected site condition” in Hong Kong. Unexpected site condition often causes bothsignificant project delays and additional project costs. For example, in 2014, constructionof the high-speed railway that connects Hong Kong and mainland China and has abudget of HK$62.4 billion suffered from at least a two-year project delay and requiredan additional several billion HK$. Unexpected site condition has been listed as a mainreason for the delay and over-budget problems.This project aims to develop methodology and computer software to quantify site-specificgeotechnical variability. The proposed method will integrate sparse project-specifictest data with pre-existing site knowledge (e.g., geological information,geotechnical property statistics, and local engineering experience available prior to theproject) under a Bayesian framework to provide updated knowledge of geotechnicalproperties. It will express the updated knowledge numerically and quantify theirvariability through Markov Chain Monte Carlo (MCMC) simulation. The major researchcomponents include (1) theoretical development of the MCMC-based Bayesian method,(2) calibration of the developed method by field data and observations, and (3)development of computer software for the developed method. Supported by a donationfor research project from industry, some preliminary studies were performed and showedpromising results.This project will effectively solve the problem of how to quantify site-specificgeotechnical variability and obtain meaningful statistics of geotechnical properties fromsparse project-specific test data. It will offer computer software to facilitate suchquantification in engineering practices. The research outcome will be beneficial to notonly Hong Kong but also many other countries where probability-based design codes(e.g., Eurocodes or Load and Resistance Factor Design codes) have been implemented orprobabilistic analyses are performed in geotechnical engineering (e.g., quantitative riskassessments of geohazards such as landslides and earthquakes), because quantificationof site-specific geotechnical variability is a prerequisite for the implementation of thesedesign codes or probabilistic analyses.


Project number9042172
Grant typeGRF
Effective start/end date1/07/1526/03/19

    Research areas

  • Geotechnical variability,Geotechnical uncertainty,Reliability-based design,Subset Simulation,