Fibre-reinforced composite materials are increasingly being used in the engineering industry due to their good specific stiffness properties. However, their relatively brittle behaviour can cause interlaminar delamination to propagate in a rapid and unstable manner under impact loading conditions such as bird-strike. Several modelling studies at the coupon-scale exist in the literature however, very few address the challenges associated with delamination modelling at large-scale specimens. For instance, in such specimens damage propagation is influenced by fibre-bridging, blunting of crack-tips, microcracking in matrix and delamination migration and these effects lead to an increase in fracture toughness. Additionally, modelling at larger length-scales requires computationally efficient methods due to the substantial increase in the number of degrees of freedom when compared to the coupon-scale. To examine the large-scale delamination behaviour, high-velocity soft-body impact tests on tapered composite laminates were carried out. This test was modelled using Finite Element Analysis (FEA) with rotation-enabled continuum and cohesive elements that are computationally efficient when compared to linear elements. The continuum element can model bend-twist coupling within the element due to its C1 continuity and as a result fewer elements are required through the thickness. The vertex-rotations in the cohesive elements enable equivalent mid-side nodes and thus larger in-plane mesh sizes can be used. This modelling strategy is shown to be beneficial at the sublaminate-scale modelled. Gelatine characterisation and the verification of boundary conditions were performed by modelling impact tests on a steel specimen. Using the calculated impactor properties and the verified boundary conditions, impact tests on composite laminates were modelled at the three tested impact velocities: 145, 155 and 165 m/s. Comparisons between experiments and modelling are performed using displacement-time history and delamination area. The importance of modelling the R-curve behaviour in large-scale specimens is also examined.