Obtención de la forma estable de crecimiento de grieta mediante modelos numéricos e inteligencia artificial

It is commonly accepted that the shape of the crack front during induced fatigue crack growth is characterized by three phases: an initial transitional phase (short crack), a stable phase (most of the growth stage) and a final phase (associated to the remanent ligament). In the literature, there are different approaches related to the geometrical shape of crack front that must be considered for predicting fatigue crack growth. Said approaches, vary from the consideration of straight crack fronts advancing with an unaltered shape (i.e., most analytic or 2D numeric models) to 3D model calculations which intrinsically include crack front shape evolution. However, these models do not achieve a comparison across the crack front, between the critical parameter associated to the propagation (Δ?, Δ????…) and the values leading to a stable crack propagation. In this paper, a model based on neural networks and finite element calculations is proposed to produce metamodels allowing for the prediction of crack front geometries under stable fatigue growth conditions. The model proposed can predict the stabilization of the crack front shape during fatigue, for different fatigue crack growth parameters, also helping to evaluate the parameter adequacy, through comparisons between stable crack shapes and experimental results.