Resumen de Towards a high-performance computing finite element simulation framework for virtual testing of composite structures
In many industrial applications, and especially in the automotive and aeronautical sectors, Virtual Testing (VT) of composite structures is foreseen as one of the strategies aimed at reducing the number of tests required to certificate new components. Conceptually, VT consists of the simulation of experimental tests through reliable numerical methods. By this strategy, it is intended to minimise the monetary and temporal costs associated with the certification and design processes and, at the same time, acquire a deeper level of understanding about the mechanical behaviour of composite structures. Thus, the VT ranges simulation from coupon level to subcomponent and component levels.
One of the most used numerical methods to build frameworks for VT of composite structures is the Finite Element Method (FEM). In part, this is because this method provides a versatile framework with reliable predictions in comparison with alternative modelling methods. Although these qualities, the VT of composite materials is a complex nonlinear numerical problem. The different damage mechanisms arising during the failure process increase the complexity of the models involved during the simulation. Moreover, the length scale concerning the damage phenomena imposes the use of thin meshes and refined time discretisation. These facts increase the computational costs of the simulations. Therefore, the VT of composite structures has usually an enormous computational cost, which, if not appropriately managed, can even preclude the simulation.
This doctoral thesis aims at developing and implementing a computational framework for Virtual Testing of Composite Structures in a High-Performance Computing (HPC) environment. By this objective is pretended to overcome the limitation regarding the management of the computational costs. In this sense, this thesis presents different constitutive models, which are based on the continuum damage mechanics theory, and their implementation in the HPC-based base FE simulation code called Alya. The verification and validation of the models are performed comparing the numerical predictions with analytical and experimental data. These comparisons demonstrate not only the reliability of models but also the potential of the usage of an HPC-based FE simulation code for Virtual Testing of COmposite STructures. Thus, the outcome of this thesis is both the formulation of novel fracture models and the numerical framework named Alya-VITECOST.
