82
5 Mechanical characterization on structural element level
modelled as tied contact. In case of the meso-scale model, the top cell wall edges coincide with the face sheet. It is therefore a node to surface contact, while the 3D-continuum model is implemented with a surface to surface contact. Applying a tied contact
appears reasonable, since there was no core-face debonding evident in the experiments.
Lastly, the load cylinders are modelled as rigid bodies with an element size of 1.0 mm,
thus enabling a good geometry approximation. The implemented virtual testing framework is illustrated in Figure 61 a) in case of the meso-scale core and four-point bending
setup. The 3D-continuum model is essentially the same, just with 8-node hexahedron
elements instead of the detailed core. In all simulations, mass scaling and increased loading rates are implemented based on sensitivity studies in order to reduce the computational effort.
Results
The simulation results in terms of force-displacement curves are given in Figure 62 and
Figure 63 in comparison to the experimental results for each of the investigated bending
configurations. Figure 64 illustrates the failure modes of the detailed simulation approach in comparison to the experiments. A comparative study on the visual damage
mechanisms of the two implemented modelling approaches is given in Appendix B2.
Analogous to the previously described simulations on constituent level, all simulations
were carried out on a six core Intel Xeon X5680 workstation and computational effort
varies greatly between the different models and modelling approaches. In case of the
3D-continuum approach the computation time ranges between 20 minutes in case of
smaller models with little deformation until failure (e.g. B19-L101) and 2 h in case of
larger models with large deformations (e.g B26-W010). In contrast the computation time
of the detailed models varies between 2 h and 14 h.
Discussion
The simulated bending stiffness is consistent for both models in all configurations and
matches the experimental results well. The only exceptions are the two B07 configurations, where both simulation models lead to a 5% reduced bending stiffness if compared
to test results. However, regarding the failure mode and the bending strength, both
models partly differ considerably depending on the configuration. Generally, the detailed model tends to result in equal or higher strengths if compared to the 3D-continuum model. In case of the two W020 configurations, which were tested using the 4Pbending setup including loading plates, both models predict the correct failure mode
with comparable strength matching the experimental results well (deviation below
10 %). This is due to the fact that these configurations are dominated by face sheet failure, which could be calibrated by setting the compressive face sheet strength of the material model accordingly. The same applies for the thin B07 configurations.
5 Mechanical characterization on structural element level
modelled as tied contact. In case of the meso-scale model, the top cell wall edges coincide with the face sheet. It is therefore a node to surface contact, while the 3D-continuum model is implemented with a surface to surface contact. Applying a tied contact
appears reasonable, since there was no core-face debonding evident in the experiments.
Lastly, the load cylinders are modelled as rigid bodies with an element size of 1.0 mm,
thus enabling a good geometry approximation. The implemented virtual testing framework is illustrated in Figure 61 a) in case of the meso-scale core and four-point bending
setup. The 3D-continuum model is essentially the same, just with 8-node hexahedron
elements instead of the detailed core. In all simulations, mass scaling and increased loading rates are implemented based on sensitivity studies in order to reduce the computational effort.
Results
The simulation results in terms of force-displacement curves are given in Figure 62 and
Figure 63 in comparison to the experimental results for each of the investigated bending
configurations. Figure 64 illustrates the failure modes of the detailed simulation approach in comparison to the experiments. A comparative study on the visual damage
mechanisms of the two implemented modelling approaches is given in Appendix B2.
Analogous to the previously described simulations on constituent level, all simulations
were carried out on a six core Intel Xeon X5680 workstation and computational effort
varies greatly between the different models and modelling approaches. In case of the
3D-continuum approach the computation time ranges between 20 minutes in case of
smaller models with little deformation until failure (e.g. B19-L101) and 2 h in case of
larger models with large deformations (e.g B26-W010). In contrast the computation time
of the detailed models varies between 2 h and 14 h.
Discussion
The simulated bending stiffness is consistent for both models in all configurations and
matches the experimental results well. The only exceptions are the two B07 configurations, where both simulation models lead to a 5% reduced bending stiffness if compared
to test results. However, regarding the failure mode and the bending strength, both
models partly differ considerably depending on the configuration. Generally, the detailed model tends to result in equal or higher strengths if compared to the 3D-continuum model. In case of the two W020 configurations, which were tested using the 4Pbending setup including loading plates, both models predict the correct failure mode
with comparable strength matching the experimental results well (deviation below
10 %). This is due to the fact that these configurations are dominated by face sheet failure, which could be calibrated by setting the compressive face sheet strength of the material model accordingly. The same applies for the thin B07 configurations.