ABSTRACT
The existing models of concrete degradation are being challenged by the sustainability-driven emergence of new chemical compositions and exposure conditions, such as from high fractions of cement replacements in new concretes or from service life extension of existing structures. This article identifies three challenges to address for concrete degradation models to become relevant also for unconventional compositions and exposure conditions: (i) being able to draw physical predictions of macroscale behaviours starting from chemo-mechanical processes at the microscale; (ii) conceiving experiments that are model-oriented, away from the current paradigm where full degradation experiments are run first, and then models are calibrated and validated on them without feeding back to the experiments; (iii) distilling the complexity of the microscale simulations into simpler, engineering-oriented tools describing the constitutive behaviour of the concrete. The first challenge, on modelling micro-chemo-mechanical processes, is discussed in detail, presenting the structure and capabilities of a recently developed simulator, MASKE, based on off-lattice, coarse-grained, kinetic Monte Carlo. Sample results are presented, from simulations of stress-induced dissolution, dissolution-induced strain-rate dependence of stress, crystallisation pressure, and cement paste carbonation. Potential solutions to the other two challenges are then discussed, envisaging a possible interplay between reverse-engineered experiments and short-term model predictions, and the use of machine learning to surrogate the results of the microscale simulations.
