Research

Extreme events happen in a variety of natural, engineering and financial systems. They are away from their normal state which can cause severe consequences. The physical understanding, quantification and prediction of extreme events thus are of vital importance. Currently, I am working on quantifying the extreme response of ocean structures to environments, particularly waves. 

The typical context of the Bayesian experimental design is to learn the quantity of interest (QoI) associated with a black-box (usually deterministic) function with known random input.  Two key components in Bayesian experimental design is the modeling of the black-box function and the utility function of selecting design points.  For function learning, I am particularly interested in Gaussian process, which is a useful probabilistic machine learning tool providing point-wise prediction as well as the uncertainty of the prediction. The prediction uncertainty is the corner stone in building the utility function, which straightforwardly aims to reduce the uncertainty of the estimation of QoI. 

Tidal current energy has attracted more and more attention with the increasing demand for renewable energy. To improve the tidal current turbine’s competitiveness in the energy area, it is recommended to rely on tidal turbine arrays as they are more cost-effective than single turbines. Therefore,how to distribute turbines in an array becomes the key point, and a number of important investigations were reported in the past decades. We extend the two scale array model describing the dynamics of an array which horizontally blocks a constant-velocity channel in two ways, firstly, a three scale array model allowing vertical deployment of turbines for additional benefits, and secondly, a combined array-channel model to include the effects of array on channel (reduce flow velocity).