Issue 51

F. Fabbrocino et alii, Frattura ed Integrità Strutturale, 51 (2020) 410-422; DOI: 10.3221/IGF-ESIS.51.30 417 Figure 6 : Crack tip speed vs crack tip position, comparisons with numerical ([26]and experimental [25]data: (a) L1 loaded configuration; (b) L2 loaded configuration. Figure 7 : (a) Crack tip speed vs crack tip position: comparisons with numerical [11,26] and experimental [25] data; (b) Variation of dynamic stress intensity factor vs time: comparisons with experimental data [25]. In Figs. 6a-b and 7a, results in terms of crack velocity normalized on the Rayleigh wave speed r c a function of the crack tip position normalized on the specimen length ( L ) obtained by the proposed model are compared with existing experimental [25] and numerical [11,26] data. The curves show high values of the crack tip speed especially during the initiation phase. Once the crack tip moves, an oscillatory behavior is observed until the crack arrest phenomenon is achieved. In Fig. 7b, the dynamic SIF time histories, computed by means of the proposed model and for the three loading configurations, are compared with experimental data [25]. The numerical results are in good agreement with the data arising from the experimental [25] and numerical [11, 26] data taken from the literature. All the computations are performed on a Xeon processor running on a Windows 10 system. The governing equations are solved by using an implicit time integration scheme based on a variable step-size backward differentiation formula (BDF). The total computational time used by the proposed numerical scheme is about 700 s, while it is about 360000 s for the re- meshing technique developed by [26]. Thus, this formulation allows to save much of the total computational time. In order to verify the computational efficiency of the proposed model, the influence of the mesh discretization as a function of both the mesh dependency and computational costs are discussed. To this end the following mesh discretization are considered: