The transportation industry and passenger cars in particular are strong emitters of gases that contribute to the climate crisis. For this reason, the automotive industry investigates opportunities to reduce emissions, such as reducing the weight of the car. Composite materials, due to their high strength, stiffness and energy absorption to weight ratio, are a suitable material choice to reduce weight. The challenge here is that composites do not satisfy the fast development times and low costs required by the car industry. An efficient design phase, using more simulation and less physical testing, allows for time and cost-savings. However, there is a lack of efficient computational models to help the design with composite materials, which is fundamental for a widespread usage of composites in the automotive industry. This thesis presents the development, improvement and validation of constitutive models for composites in crash, focusing on compressive damage modes, matrix compression and fibre compression. The material being modelled is a carbon fibre/epoxy uni-weave Non-Crimp Fabric (NCF) composite. The properties of the composite constituents are homogenized to the ply level for a more efficient modelling.
The matrix behaviour is modelled by combining damage and friction on the microcrack surfaces. The transverse mechanisms are modelled efficiently using a criterion for final failure, interaction of damage modes and a continuous response between compression and tension. The model is validated against 45- and 90-degree specimens. The fibre compression mode is fibre kinking growth, a very complex mechanism, responsible for high energy absorption. A homogenized 3D model based on Fibre Kinking Theory (FKT) is developed. It includes initial fibre misalignments and further rotations are governed by equilibrium with shear nonlinearity. The model is implemented in a commercial Finite Element (FE) software together with a mesh objective methodology. Furthermore, another formulation with similar physical principles but more suitable, efficient and robust for crash simulations is developed, implemented in an FE software and validated against experiments. The results show good qualitative and quantitative agreement. The proposed models allow for a reduction of physical testing required to develop crashworthy structures.