The soil can be considered as a collection of grains or aggregates with microstructures, aggregates are composed of particles with interparticle voids, and grains or particles are made of minerals with atomic structures. Thus, the origin of mechanical behaviours/properties of the soil can be investigated by downscaling, and the mechanical modelling of the soil can be conducted by upscaling.

The constitutive modelling of soils is a fundamental pillar of advanced geotechnical analysis, aiming to mathematically represent the complex and unique stress-strain behaviour of soil under various loading paths and drainage conditions. Unlike simpler engineering materials, soil exhibits a host of intricate characteristics including nonlinearity, inelasticity, pressure-dependence, dilatancy, path-dependency, time-dependency, and temperature-dependency, which cannot be captured by elementary elastic models. Constitutive models are therefore developed as sets of equations that define these relationships, ranging from relatively simple frameworks like Mohr-Coulomb to more sophisticated critical state-based models such as Cam-Clay and its modern descendants. The primary challenge lies in creating a model that is both realistically representative of the soil's observed behaviour and computationally efficient for practical application in numerical simulations. A well-calibrated constitutive model is crucial for predicting real-world geotechnical performance, enabling engineers to more accurately forecast settlements, assess stability, and understand failure mechanisms in geotechnical structures.

Numerical methods have become an indispensable tool in modern geotechnical engineering, providing powerful capabilities to analyze complex soil-structure interaction problems that are often intractable using traditional closed-form analytical solutions. By discretizing a continuum into a finite set of elements, techniques like the Finite Element Method (FEM), Material Point Method (MPM), Hybrid Element Particle Method (HEPM), Finite Difference Method (FDM), Peridynamics method (PD), and Finite Discrete Element Method (FDEM) enable engineers to model the inherently nonlinear, inelastic, and heterogeneous behavior of soil and rock under various loading and environmental conditions.

We have extensively performed the application of AI in geotechnics due to the strong capacity of solving non-linear and high-dimensional problem of AI. For example, the optimization and Bayesian-based methods can bridge the gap between advanced constitutive theories and engineering practice; ML-based surrogate model can also be applied in engineering practice as the alternative to experimental and numerical methods, saving expenses for engineering design. Currently, we focuses on the following three topics.

We have developed various types of geotechnical tests: (1) scaled model testing, in which we aim to address the reclamation related geotechnical problems, such as sedimentation and consolidation of marine deposits, marine reclamation by super-fast consolidation methods, FRP pile behaviour filled with sand or cemented sand, and clay-FRP pile interaction under static and cyclic loadings; (2) advanced soil element testing, in which the soil element performance can be clarified by several advanced testing apparatuses, including multi-functional soil-structure interface testing apparatus, PolyU-patented true triaxial apparatus, dynamic hollow cylinder apparatus, and biaxial testing apparatus; and (3) soil microstructural testing, in which the soil microstructure can be well identified by X-ray computed micro-tomography, scanning electron microscope (SEM) and mercury intrusion porosimetry (MIP) test, with the relevant apparatus in PolyU and CEE.