Research

Our proposed research initiative seeks to propel machine learning into the forefront of geotechnical engineering, with a vision to address critical challenges and revolutionise the field for the betterment of society. The overarching goals of our project align with the need to confront uncertainty, combat climate change through zero carbon emission strategies, address soil parameter heterogeneity, expedite finite element (FE) calculations e.g., for reliability analyses, and enhance design efficiency to reduce material consumption, particularly in the context of concrete. By undertaking this multidimensional approach, our research aims not only to apply machine learning in geotechnical engineering but to fundamentally transform the field, ushering in a new era of efficiency, sustainability and resilience. Through collaboration and innovation, we aspire to make machine learning an integral and indispensable tool for addressing the complex challenges faced by geotechnical practitioners in the 21st century.

Granular materials are omnipresent in our daily life. The same granular material can behave like solid, and fluid, which poses formidable challenge to the constitutive models and numerical methods. Traditionally, constitutive models for the solid- and fluid-like behaviour have been developed for the respective flow regimes in different engineering/scientific disciplines with hardly any intersections. A single constitutive model capable of describing the transient behaviour during phase transitions in both solid-like and fluid-like regimes is a challenging task with enormous application potential. MOTRAN takes on this challenge with a simple yet genius ansatz by decomposing the stress rate into a frictional and collisional part. By adopting our hypoplastic model for the frictional part and non-Newtonian fluid for the collisional part, this ansatz gives rise to an unconventional constitutive model with the 2nd order strain rate similar to the acceleration of motion, which serves as an excellent classifier for steady and transient motions. This constitutive model is then augmented to include a length scale in micropolar continuum for multiscale analysis. Based on mixture theory, the field equations are established in rate form for the first time and discretised by multi-layer SPH model. For polydisperse granular flow with individual large particles, the SPH model is coupled with own developed Surface Mesh Represented DEM to simulate particles of arbitrary shapes. Advanced solution techniques are developed based on multi-GPU acceleration for high fidelity simulation of large-scale problems. The constitutive model is calibrated by laboratory experiments on natural granular materials and their transparent surrogate. The numerical model is validated by scaled model tests under elevated acceleration in centrifuge as well as real-world cases of our database. MOTRAN is an exciting endeavour with the potential to create a new paradigm that will revolutionise the way how transient granular flow is to be modelled.

LOC3G project seeks to advance the knowledge of multiscale and multiphysics localization phenomena in porous geological media, with the aim of creating new predictive models for geophysics, geohazards, and geoengineering. The consortium combines a diverse array of expertise, including geological surveys, constitutive modeling and numerical simulations, laboratory tests, and real-world applications such as CO2 storage and geo-resource/energy exploitation. The project will incorporate innovative research techniques and utilize advanced constitutive models and next-generation numerical approaches to investigate the localization of deformation in geological media. The ultimate goal is to provide cutting-edge knowledge and interdisciplinary training to improve the capacity for research and technology globally, and to provide practitioners with the tools they require to tackle relevant problems in their fields. Additionally, LOC3G is expected to have a significant impact on addressing EU energy crisis caused by geopolitical issues.