Abstract

Composite materials have been used for many years in industries such as the wind power, automotive and aeronautics to design lightweight structures with high mechanical performances, i.e. stiffness and strength. However, the microstructural complexity of the composite materials challenges the estimation of the load carrying capabilities of composite structures. Failure in laminated composites is usually caused by inter-laminar fractures, such as delamination or adhesive joint debonding, promoted by or coexisting with intra-laminar damage mechanisms, like matrix cracking and fibre failure. This paper presents a method for modelling fatigue-driven delamination using finite elements and a cohesive zone model approach. The fatigue damage rate is linked to a Paris? law-like expression for the crack growth rate evaluated using the energy release rate and constant load ratio. A J-integral formulation that defines an integration path across the length of the cohesive zone allows accurate computation of the energy release rate. Integration paths are tracked following the growth driving direction at each integration point, which is computed as the gradient of the ratio of the total performed work to the interfacial fracture toughness. Only the maximum loads of the fatigue cycles are simulated and a cycle jump algorithm is applied to avoid prohibitive CPU time together the adaptive cycle integration algorithm in order to account for static damage during the cycle jump. The jump criterion is based on a target crack growth increment chosen by the user. This method was implemented in the solver Simcenter Samcef. To validate the implementation, a batch of two double cantilever beam (DCB) specimens made of a non-crimp fabric laminate were tested applying a constant mode I bending moment on each arm. The specimens included a partial reinforcement at the mid-width to promote a curved delamination front during propagation. The experimental crack front shape evolution was well reproduced by the simulation method. The crack front propagation was slightly underestimated compared to the experimental results. However, the difference between experimental and simulation results at the end of the analysis is comparable to the variation in crack propagation among the different demonstrator specimens experimentally tested.