| Abstract |
Granular materials are ubiquitous in nature and indispensable in various industrial applications. However, their complex multi-scale behaviour remains poorly understood. The macroscopic mechanical response of granular media arises from intricate microscale particle interactions, which become even more complex when considering thermal behaviour. In the scenario of thermo-mechanics analysis, thermal cyclic loading (TCL) introduces additional challenges, requiring the integration of transient, small-scale effects with long-duration phenomena, often spanning weeks or months.
Predicting the thermo-mechanical behaviour of granular media under TCL is essential for applications such as optimising thermal energy storage systems and geostructures, ensuring the stability of soils and grain-filled silos, and tailoring the properties of granular media for specific industrial uses.
Current experimental and numerical approaches face significant limitations in addressing the long-term and multi-scale effects of TCL. Purely experimental methods struggle to extrapolate small-scale results to large-scale systems and are hindered by time constraints that prevent extensive TCL analyses. Moreover, repeated experiments are required for varying initial conditions of granular media, sometimes producing conflicting data.
Computational methods also fail to comprehensively represent the thermo-mechanical response of granular media under TCL. Specifically, continuous numerical methods cannot capture the particulate nature of granular media, relying heavily on phenomenological models. Conversely, discrete methods accurately model particle interactions but are computationally expensive. Hybrid continuum-discrete methods aim to combine the strengths of both approaches but face challenges such as complex calibrations and prohibitive computational costs for real-world applications
.
GRACY seeks to overcome the challenges of TCL analysis by developing an innovative, multi-scale, data-driven computational paradigm that bridges the gap between microscale particle interactions and large-scale, long-term analyses. This paradigm will integrate cutting-edge numerical methodologies to identify the most relevant material parameters and physical phenomena governing the thermo-mechanical response of granular media subjected toTCL. An experimental campaign will also be conducted to validate the numerical results using controlled data.
GRACYs methodology circumvents computational cost limitations without sacrificing accuracy through a clever subdivision of offline and online stages. In the offline phase, multiple microscale analyses will be performed using the Discrete Element Method on representative volume elements with different initial conditions and thermal expansion processes. A machine-learning tool will then use this database to create a surrogate model for the thermomechanical behaviour of the granular media. This surrogate model will be integrated into a continuum online solver and used to update relevant historical
variables of the granular body. The accuracy of the multi-scale, data-driven methodology will be evaluated through a targeted experimental campaign involving different granular assemblies subjected to TCL scenarios.
The resulting computational framework will advance the understanding of this intricate coupled problem, which is of significant interest to the scientific community and has growing industrial applications. |
