Multi-scale modelling strategy for textile composites based on stochastic reinforcement geometry

Abstract

The quality of high-performance composite structures is difficult to predict. Variability in the macroscopic performance is dominated by the spatial randomness in the geometrical characteristics of the reinforcement, especially for textile composites. This work provides a roadmap for generating realistic virtual textile specimens spanning multiple unit cells, which are required to perform high-fidelity simulations. First, the geometrical variability in the reinforcement structure is experimentally quantified on the meso- and macro-scale in terms of average trends, standard deviations and correlation lengths. Next, each reinforcement parameter is modelled by the sum of its average trend and its zero-mean deviations, which are both determined by analysing experiments. Virtual specimens are then created using advanced simulation techniques that match the experimental statistics. Depending on the nature of measured correlations, the simulation technique is either a Monte Carlo Markov Chain method, a cross-correlated Karhunen–Loève Series Expansion technique or a Fourier Transform method used in combination with a Markov Chain algorithm. In a last step, a virtual representation of the textile geometry is constructed in geometrical modelling software, such as the commercially available WiseTex software.The multi-scale framework is validated using data for a carbon–epoxy 2/2 twill woven composite produced by resin transfer moulding: the simulated tow deviations trends replicate the target statistics.