COMPLAS 2023

Computationally Efficient Modelling of Short Fiber Composites Across Different Length Scales

  • Mirkhalaf, Mohsen (University of Gothenburg)
  • Castricum, Brian (Eindhoven University of Technology)
  • van Beurden, Tom (Eindhoven University of Technology)
  • Ekh, Magnus (Chalmers University of Technology)
  • Larsson, Fredrik (Chalmers University of Technology)
  • Fagerström, Martin (Chalmers University of Technology)

Please login to view abstract download link

Short Fibre Reinforced Composites (SFRCs) have high specific properties. Also, fabrication processes of these materials are efficient both time-wise and cost-wise. As a result, a trend of increasing usage of SFRCs is observed in different industries. Many microstructural properties, such as fibre volume fraction; fibre orientation distribution; fibre geometrical aspects etc., affect the macroscopic behaviour of these materials. Hence, to have an accurate structure-property relationship for these materials, it is crucial to take these microstructural properties into account in the modelling process. In this study, we have developed a two-step non-linear elasto-plastic micro-mechanical model for SFRCs. In the first step, a number of elasto-plastic Finite Element (FE) computations are conducted on a Unit Cell (UC) that is essentially a single fibre surrounded by matrix material. Based on these FE results, a surrogate constitutive model is calibrated. This calibrated surrogate model is then subsequently used in a second Orientation Averaging (OA) step, following Advani and Tucker [1]. Here, in addition to widely used interaction assumptions such as Voigt and Reuss that provide upper and lower bounds, a self-consistent interaction is also developed and evaluated. The developed model is not only capable of accurate predictions but also has a considerably better computational performance compared to computational homogenization approach. To facilitate practical and/or industrial use, a coupled multi-scale model is developed using FEM (at the macro-scale) and the micro-mechanical model (at the micro-scale). In order to reduce the computational cost of this model, a Bazant integration scheme [2] is adopted in the OA step. In addition, a more efficient integration approach for the calculation of the Eshelby tensor (for the self-consistent interaction) is employed. As a result, a remarkably improved computational enhancement is achieved in the coupled multi-scale model, and full-scale analysis of structures/specimens with a variety of micro-structural properties is enabled.