86
5 Mechanical characterization on structural element level
Yet here, the simulation results collectively underachieve the experimental results. However in case of B19-L101, the experiments suggest inconsistent failure modes leading to
high scatter. The two simulation models accurately match the single test specimen,
which failed due to local indentation. As for B19-W101, the detailed model suggests
about 15% lower bending strength if compared to the experiments. Lastly there are the
B19 configurations with three prepreg layers. In case of B19-L003, the strength of both
simulation models is considerably lower than in the test, while the failure mode of local
indentation is predicted correctly. This is possibly due to the thick phenolic resin fillet
layer of this configuration (see Figure 57 on p. 75), which is not considered in the simulations models. It is assumed, that this fillet layer postpones local core indentation in the
actual specimens. Regarding B19-W003, the 3D-continuum model appears to correctly
predict the initiation of core shear failure. However, the homogenized hexahedron elements are not capable to represent the distributed large-scale core shearing due to cell
wall buckling. Instead, the 3D-continuum elements shear locally underneath the load
cylinders, leading to premature failure. This is illustrated in Appendix B2. In contrast, the
detailed model represents the actual failure pattern well (see Figure 64 b) and thus also
reproduce the experimental force-displacement curve progression. However, the simulation model exceeds the force plateau of the experiments by about 15%. This can be
explained by the fact that the calibrated detailed honeycomb core model, which is described in section 4.1.3, also exceeds the experimental strength in case of WT-shear on
constituent level (see Figure 44). This is because the derived detailed constituent core
model represents a trade-off between multiple loading conditions.
Summary
Generally the detailed modelling approach enables a good overall match of the experimental results in terms of both, visual damage patterns (Figure 64) and force-displacement curve progression (Figure 62 and Figure 63). Therefore, the previously determined
material models and properties could be validated. Discrepancies between simulation
and test are due to remaining limitations in the model detail and general material and
test uncertainties. In contrast, the 3D-continuum modelling approach reveals several
short comings in the performed bending study. The derived homogenized core model
tends to result in premature local core indentation. However, this is not evident in all
configurations. Furthermore, this modelling approach does not appear to be suitable to
model the complicated stress state in the vicinity of the loading cylinders once shear
failure is initiated. However, based on the performed study the 3D-continuum approach
seems well suited for preliminary conservative predictions or optimization studies, especially considering that the computational time is up to 10 times lower if compared to
the detailed models.
5 Mechanical characterization on structural element level
Yet here, the simulation results collectively underachieve the experimental results. However in case of B19-L101, the experiments suggest inconsistent failure modes leading to
high scatter. The two simulation models accurately match the single test specimen,
which failed due to local indentation. As for B19-W101, the detailed model suggests
about 15% lower bending strength if compared to the experiments. Lastly there are the
B19 configurations with three prepreg layers. In case of B19-L003, the strength of both
simulation models is considerably lower than in the test, while the failure mode of local
indentation is predicted correctly. This is possibly due to the thick phenolic resin fillet
layer of this configuration (see Figure 57 on p. 75), which is not considered in the simulations models. It is assumed, that this fillet layer postpones local core indentation in the
actual specimens. Regarding B19-W003, the 3D-continuum model appears to correctly
predict the initiation of core shear failure. However, the homogenized hexahedron elements are not capable to represent the distributed large-scale core shearing due to cell
wall buckling. Instead, the 3D-continuum elements shear locally underneath the load
cylinders, leading to premature failure. This is illustrated in Appendix B2. In contrast, the
detailed model represents the actual failure pattern well (see Figure 64 b) and thus also
reproduce the experimental force-displacement curve progression. However, the simulation model exceeds the force plateau of the experiments by about 15%. This can be
explained by the fact that the calibrated detailed honeycomb core model, which is described in section 4.1.3, also exceeds the experimental strength in case of WT-shear on
constituent level (see Figure 44). This is because the derived detailed constituent core
model represents a trade-off between multiple loading conditions.
Summary
Generally the detailed modelling approach enables a good overall match of the experimental results in terms of both, visual damage patterns (Figure 64) and force-displacement curve progression (Figure 62 and Figure 63). Therefore, the previously determined
material models and properties could be validated. Discrepancies between simulation
and test are due to remaining limitations in the model detail and general material and
test uncertainties. In contrast, the 3D-continuum modelling approach reveals several
short comings in the performed bending study. The derived homogenized core model
tends to result in premature local core indentation. However, this is not evident in all
configurations. Furthermore, this modelling approach does not appear to be suitable to
model the complicated stress state in the vicinity of the loading cylinders once shear
failure is initiated. However, based on the performed study the 3D-continuum approach
seems well suited for preliminary conservative predictions or optimization studies, especially considering that the computational time is up to 10 times lower if compared to
the detailed models.