RESEARCH TOPICS
Soil-foundation interaction. We primarily study how structures, foundations and soils interact, and how they are affected by vibratory forces and ground vibration.
Our models mainly use finite element discretizations for the structure and foundation parts, and some boundary element formulation for the soil part. These models are typically computationally expensive, which is why some of our work is dedicated to improving numerical integration efficiency and high-performance computing strategies.
High-performance computing. This branch of work of our laboratory is dedicated to computer efficiency. Our first strategy is to improve the efficiency of numerical integration methods, by using advanced series extrapolation and adaptive quadrature schemes. The second strategy is to use massively parallel computing devices such as GPUs. But more recently, we have also been looking into the potential of machine learning schemes to solve computationally expensive engineering problems such as numerical integrals.
Machine learning. These trendy neural networks are great to deal with problems with poor noise-to-signal ratios. In our lab, we have been using them to identify buried landslide victims from soil spectrograms, to recognize keystrokes from computer keyboards, as well as pattern recognition schemes for puzzle solution and efficient numerical integration.
Topology optimization. We have been looking into how the properties of the soil affect the optimal shapes of the structures that they support. As the soil deforms under the weight of the structure and various external forces, it demands the structure to have intricate features to compensate for this flexibility.
Bio-inspired foundations. In this line of work, we take inspiration from tree roots to understand how to build better foundations for buildings. Roots are slick, material-saving solutions for foundations that evolved optimized shapes over millions of years, and have shown to outperform our bulky, unsustainable manmade foundations in almost every way.