Abstract: The complex behavior of dissipative solids remains a challenging subject in computational mechanics. Particularly intriguing are microstructural effects leading to highly non-linear behavior, irreversible processes, and material failure. Examples include shear bands in granular media driven by complex grain kinematics, damage and fracture in quasi-brittle materials governed by microcracks, and ductile failure in metals emerging from the interplay between dislocations and microvoids. In this context, the field of continuum solid mechanics has seen significant progress in the past half-century, with numerous constitutive models developed to address these problems. Nevertheless, most frameworks are either too simplified to accurately predict the intricacies of dissipative materials or too heuristic to be framed within a rigorous mathematical framework. Hence, this talk will explore energetic/variational formulations for coupled models that describe various complex failure processes while remaining mathematically sound and physically interpretable. Then, despite their scalability, it is argued that most (if not all) constitutive modeling approaches inevitably introduce uncertainty through phenomenological assumptions and tuning parameters. Hence, we pivot towards micromechanical platforms, such as molecular dynamics or DEM, to model granular media relying solely on physics-based grain-scale interactions. In particular, the use of discrete element methods with shape capabilities (LS-DEM) has made it possible to describe granular materials with unprecedented fidelity and to seamlessly capture complex emergent behaviors such as breakage mechanics, cyclic loading effects, and multiphysics processes. Finally, we argue that micromechanical models are predictive but not computationally scalable. Therefore, we embrace the recently developed concept of data-driven computing to exploit the capabilities of micro- and continuum mechanics in a unified computational platform. Scalability is ensured by remaining at the continuum scale and enforcing the fundamental laws of continuum mechanics. Predictiveness is ensured by relying on empirical data, obtained from experiments or high-fidelity simulations, to capture the generally unknown constitutive behavior. We will explore recent developments in data-driven mechanics that allow tackling challenging problems such as granular breakage or strain localization in the absence of phenomenological models. This framework further suggests a paradigm shift in computational mechanics for various complex problems with unique interdisciplinary synergies.
Bio: I am currently a postdoctoral scholar at Caltech, supervised by José E. Andrade. Previously, I obtained my Ph.D. in 2022 from KU Leuven, Belgium, under the supervision of Stijn François and Geert Degrande. My research work has spanned various areas of computational mechanics, mainly involving the modeling of dissipative mechanisms in solid materials including plasticity, damage, fracture, fatigue, and multi-physics phenomena. My current work seeks to integrate the micromechanics of complex systems with continuum scale simulations, with emphasis on data-driven computing as a multi-scale computational paradigm.
