Modelação de Alta-Fidelidade da Fratura Progressiva em Compósitos Avançados com Gradientes de Deformações Multi-Escala

The main objective of HOOPLA is to develop high-fidelity numerical tools that enable the prediction of the inelastic deformation and progressive failure of novel materials used in structural components of aircrafts, linking the micro-scale with the composite structural performance. These numerical models will allow the generation of reliable artificial databases to feed ML-models, paving the road to virtual design and certification in aeronautics based on digital twins. The numerical framework is focused on CFRTs, which make aircrafts more sustainable during their production (efficient manufacturing), operation (lightweight structures) and end-of-life (recyclability). The research plan addresses 3 main questions: Q1: What are the main microscopic mechanisms driving progressive failure of CFRTs and how to model them? Q2: How to capture their effect on the mechanical response at the meso-scale (ply level)? Q3: Can we generate artificial data allowing for Multi-Scale Digital Twins that consider the effect of micro-scale defects on the structural response? To address Q1, the main molecular mechanisms responsible for deformation and failure of polymers in regions undergoing large strain gradients will be investigated. For metals, the role of SSDs (statistically stored dislocations) and GNDs (geometrically necessary dislocations) on their plastic behaviour, dependent on strain-gradient, is now well-established [11]. However, this is still underexplored for polymer matrices, even though it is important to properly model the micro-scale behaviour of composites [8,9]. Constitutive models relating size effects under strain-gradients with the underlying molecular mechanisms occurring in the matrix, are needed in regions driving fracture initiation, and will be developed. These aspects are particularly critical for transverse failure and longitudinal compression causing fibre-kinking, where high-curvatures are observed on the fibres before failure. 3D FE models of the composite microstructure combined with these new constitutive models will provide descriptions of the microscopic mechanisms with unprecedented accuracy. The hypothesis associated with Q2 is that 2nd-order homogenisation is required since the microscopic damage mechanisms are intrinsically related to mesoscopic regions lying in fracture process zones, where meso-strain gradients become significant. Therefore, the 2nd-order formulation developed by the PI [1,12] must be further enriched to account for the transition of fracture across the scales, combining it with the work in [2]. This disruptive approach has not been addressed in the literature, hitherto. Dynamic effects must also be included in the multi-scale second-gradient continuum, to deal with possibly unstable fracture, which is another aspect lying beyond the state-of-the-art. The combination of multi-scale computational mechanics with ML and HPC allows answering Q3. The work in [1] enables considering the effect of micro-voids on the composite failure with strain-gradients for the first time. Even in standard homogenisation, studies on their effect are limited to the elastic properties [5]. The databases generated for ML-models predicting the strength of notched specimens have been generated with semi-analytical models [4]. Unprecedented high-fidelity models will be used to generate data in HPC frameworks, enabling multi-scale digital twins considering the effect of micro-voids on the coupons strength.