Don't miss our weekly PhD newsletter | Sign up now Don't miss our weekly PhD newsletter | Sign up now

Geomechanical Inverse Modelling And Uncertainty Quantification

   Department of Civil and Structural Engineering

This project is no longer listed on and may not be available.

Click here to search for PhD studentship opportunities
  Dr Domenico Bau  Applications accepted all year round  Self-Funded PhD Students Only

About the Project

Detailed predictive modelling of coupled geomechanical processes at the continuum-scale is vital for addressing decision making in areas such as geological carbon sequestration, subsurface waste disposal, land subsidence, groundwater and geothermal resources development, and reservoir engineering. In recent years, novel types of near surface and subsurface observations have offered valuable insight into the processes that govern the migration of fluids in the subsurface, the deformation of geological layers, and the parameters necessary to model these processes. The information retained by these data can thus be extracted by developing inverse models that merge field observations into the response of coupled geomechanical models.
Since these models typically involve large numbers of inputs and outputs, they result computationally very expensive. This limits significantly the quantitative assessment model uncertainties, which is generally problematic if only a limited number of simulations can be performed. There is thus a great interest in building surrogate models, that is, fast approximations of full-scale simulators, to carry out tasks such as parameter estimation and history-matching, and consequently make predictions that support an informed decision making under conditions of uncertainty.
The aim of this project is is to develop novel model surrogate techniques within data assimilation frameworks for estimation of the parameters of geomechanical simulators. The focus will be on data–driven surrogates, such as Gaussian process estimators, where the dimension of the parameter space is reduced by projecting it into a smaller subspace using principal component methods, such as the Karhunen-Loeve expansion. Different techniques to inform the selection of initial surrogate training sets will be studied, combined with dynamic machine-learning sampling methods that progressively inform where to run the full scale geomechanical simulator as convergence of the inverse model is approached.
Engineering (12) Mathematics (25)

Funding Notes

Pre-requisite qualification
The ideal candidate is a student with a passion for the computational aspects of engineering applications, and propensity to understand in-depth the physics, mathematics, statistics, and numerics behind the algorithms used to develop geomechanical deformation models.
Search Suggestions
Search suggestions

Based on your current searches we recommend the following search filters.