Modeling of delamination migration in DCB test on multidirectional composite laminates

The different failure mechanisms occurring in fibres reinforced polymer matrix composites usually interact with each other, making the overall failure of a composite structure a complex process, difficult to predict. The interaction between delamination and matrix cracks in laminated composites has been investigated here, using the case of a Double Cantilever Beam (DCB) test on multidirectional laminates. A Cohesive Zone based approach was employed in the finite element analyses [1] and validating experiments were carried out. Delamination and matrix cracks cause a redistribution of stresses inside the plies, eventually leading to fibres breakage in the primary load-carrying plies and, therefore, to the final failure of a structure. The interaction between delamination and matrix cracks causes delamination migration through the thickness of laminates. Delamination finds its way to the weakest interface in a laminate, through matrix cracks, finally causing the complete failure of a structure [2]. The process is well known in industrial practice as an important cause of failure in laminated composite structures. Nevertheless, it is usually neglected in finite element analyses, because of the difficulty in modelling the interaction between the different damage mechanisms when using the currently available finite element tools [3, 4]. In this work delamination migration through the thickness at ±θ interfaces was modelled using a full scale three-dimensional DCB test model on an aerospace grade carbon/epoxy composite. Hexahedral continuum elements were employed to represent a 40 ply laminate. Cohesive interface elements were employed to represent both matrix cracks and delamination in four groups of ±θ plies in the centre of the lay-up. Matrix cracks were modelled by bands of cohesive elements, parallel to the fibres orientation and equally spaced along the ply length. This allowed for the matrix cracks to initiate in any of the possible locations, according to the initiation criteria defined. Delamination was accounted for by 9 inter-ply layers of cohesive elements tied to the surrounding elements The models consisted of 540898 nodes, with 124668 cohesive elements. Numerical analyses were performed using the Abaqus/Explicit solver, requiring a solution time of 46 hours using 8 CPUs. The fracture surfaces obtained from experimental DCB tests showed the typical “crack jump” behaviour. The main features of the damage mechanisms observed were captured by the proposed model. Crack jumps initiated at the free edges of the sample and damage propagated to different interfaces through cracks within the plies. The model can predict onset and propagation of both delamination and matrix cracks. Above all, the interaction between intersecting cracks on different planes was correctly managed, also for a number of different crack initiation locations. Although a complete agreement with the experimental results has not yet been achieved, this approach represents a step forward in the prediction of delamination migration in full scale composite structures, since it is able to manage the simultaneous presence and interaction of multiple cracks and delaminations