The state-of-the-art design approach of composite structures in Formula One (F1) rely on expensive full-scale crash tests and simple models that cannot describe the underlying physical phenomena during crushing. Although the models can describe the crush forces and resulting accelerations, their predictive capabilities are poor, and several iterations are usually required to arrive at a satisfactory design, causing the teams money and resources. Reductions in costs could be possible by reducing the number of full-scale experiments and increasing the amount of analysis. This thesis aims to evaluate the capabilities of a commercially available material model at two distinct modelling scales by conducting physical and virtual experiments on two different specimen geometries; tubular coupons and a new open-section channel specimen. Firstly, the new open-section specimen with a geometry representing the typical F1 crash structures was designed, manufactured and tested. The newly proposed specimen showed controlled failure modes, and displayed typical characteristics of composite structures under crushing loading. Secondly, the two specimen geometries were numerically analyzed on the macro-scale (laminate-level) and meso-scale (ply-level) using a commercially available material model implemented in the Abaqus finite element software. Mesh-dependency of the macro-scale approach was discussed and numerical aspects required to perform crush simulations on the meso-scale were identified and discussed. Lastly, calibration of the numerical models was performed to understand the effects of model inputs on the output responses, to evaluate the validity of the input parameters, and to assess two different calibration approaches; the first method is based on macro-scale modeling, while the second method is based on surrogate-based inverse parameter identification at the meso-scale. The numerical models could not reproduce the experimental crush responses of either geometry when using experimentally characterized material data. Following the second calibration approach, the qualitative and quantitative crush responses of the tubular specimen could be accurately described, but the channel specimen showed no improvements in their response. The macro-scale modelling approach with the current material model was judged to be impractical for large-scale analysis, and the meso-scale approach showed need for further developments.