4.1 Sandwich core
47
nodes (‘node shaking’) [Hei08], assigning random material properties and/or wall thickness [Hei08], [Asp13], [Kar13], [Fis16], pre-buckled cell walls or cell wall waviness
[Hei08], [Roy13], [Fis16], angular deviation of the cells [Fis16] and actually determined
geometry imperfections using CT-scans [Hei08], [Fis09], [Fis16]. These approaches have
proven to effectively approximate local imperfections of actual honeycomb cells. A more
simplistic approach is applying cell geometry along with globally reduced material properties. Heimbs [Hei08] showed, that this approach enables a good correlation between
numerical and experimental results, while still reproducing realistic buckling patterns.
From the available literature, it can be concluded that to some extend model calibration
is generally required in order to match test results, independently of the level of model
detail or the implemented approach to modelling imperfections. Therefore, the latter
simplified approach is implemented in the present work, since it appears to be most efficient and thus well suited for engineering applications. As a result, all derived honeycomb material properties of the present thesis represent reduced properties including
local geometric and material imperfections. However, the imperfect global hexagon grid
of the investigated specimens is considered by the implemented models (a ≠ b and
ϕ ≠ 60° in Figure 30). This is required since this imperfection has a different effect on the
macroscopic mechanical behavior depending on the loading direction. For instance, in
case the hexagon grid is elongated in L-direction (ϕ < 60°), the superiority of the LT-shear
strength if compared to WT is enhanced. This cannot be captured by globally reduced
material properties.
Preliminary numerical studies
In case of the honeycomb core, additional preliminary studies were performed prior to
the material parameter calibration. These included the determination of a suitable
model scale as well as hexagon geometry. The performed preliminary studies are described in the following.
Model scale
A previous study from Wilbert et al. [Wil11] showed that despite the implementation of
an appropriate unit cell along with periodicity conditions prescribed on the free honeycomb edges, the macroscopic stress-strain progression obtained from simulation models
correlates with the number of included unit cells. They achieve convergence after a few
iterations. However, their results could not be transferred to the present study, since
different unit cells and boundary conditions were defined. In addition, the study of Wilbert et al. was restricted to out-of-plane compression. In the present work, an appropriate model scale was determined through convergence studies based on the previously
established boundary conditions for all three considered load cases. For a more detailed
description of the performed convergence studies it is referred to [See17]. The conclusively determined model scales for each load case are illustrated in Figure 37.
47
nodes (‘node shaking’) [Hei08], assigning random material properties and/or wall thickness [Hei08], [Asp13], [Kar13], [Fis16], pre-buckled cell walls or cell wall waviness
[Hei08], [Roy13], [Fis16], angular deviation of the cells [Fis16] and actually determined
geometry imperfections using CT-scans [Hei08], [Fis09], [Fis16]. These approaches have
proven to effectively approximate local imperfections of actual honeycomb cells. A more
simplistic approach is applying cell geometry along with globally reduced material properties. Heimbs [Hei08] showed, that this approach enables a good correlation between
numerical and experimental results, while still reproducing realistic buckling patterns.
From the available literature, it can be concluded that to some extend model calibration
is generally required in order to match test results, independently of the level of model
detail or the implemented approach to modelling imperfections. Therefore, the latter
simplified approach is implemented in the present work, since it appears to be most efficient and thus well suited for engineering applications. As a result, all derived honeycomb material properties of the present thesis represent reduced properties including
local geometric and material imperfections. However, the imperfect global hexagon grid
of the investigated specimens is considered by the implemented models (a ≠ b and
ϕ ≠ 60° in Figure 30). This is required since this imperfection has a different effect on the
macroscopic mechanical behavior depending on the loading direction. For instance, in
case the hexagon grid is elongated in L-direction (ϕ < 60°), the superiority of the LT-shear
strength if compared to WT is enhanced. This cannot be captured by globally reduced
material properties.
Preliminary numerical studies
In case of the honeycomb core, additional preliminary studies were performed prior to
the material parameter calibration. These included the determination of a suitable
model scale as well as hexagon geometry. The performed preliminary studies are described in the following.
Model scale
A previous study from Wilbert et al. [Wil11] showed that despite the implementation of
an appropriate unit cell along with periodicity conditions prescribed on the free honeycomb edges, the macroscopic stress-strain progression obtained from simulation models
correlates with the number of included unit cells. They achieve convergence after a few
iterations. However, their results could not be transferred to the present study, since
different unit cells and boundary conditions were defined. In addition, the study of Wilbert et al. was restricted to out-of-plane compression. In the present work, an appropriate model scale was determined through convergence studies based on the previously
established boundary conditions for all three considered load cases. For a more detailed
description of the performed convergence studies it is referred to [See17]. The conclusively determined model scales for each load case are illustrated in Figure 37.