Composite Structures 203 (2018) 101–109

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Numerical modelling of partially potted inserts in honeycomb sandwich T

panels under pull-out loading
⁎
Ralf Seemanna, , Dieter Krauseb
a
Diehl Aviation Hamburg GmbH, Hein-Saß-Weg 41, 21129 Hamburg, Germany
b
Institute of Product Development and Mechanical Engineering Design, Hamburg University of Technology, Denickestraße 17, 21073 Hamburg, Germany

A R T I C LE I N FO A B S T R A C T

Keywords: Sandwich panel inserts are prone to pull-out loading, while predicting the pull-out strength is challenging due to
Nomex honeycomb the multitude of damage mechanisms involved. The present paper describes the implementation of two state-of-
Partially potted insert the art non-linear finite element models for predicting the pull-out strength. The two models include a 3D-
Explicit FEA continuum model where the core was modelled using 8-node brick elements with a homogenized material model
Virtual testing
and a detailed meso scale model with accurate honeycomb cell representation using 4-node shell elements. The
simulation results are benchmarked against experimental data of the reference configuration. Both models en-
able to predict the strength accurately, while the 3D-continuum model can be considered favourable for most
applications due to the reduced computational effort.

1. Introduction
Potted inserts are standard fastening elements in honeycomb sand-
wich construction [1,2]. According to the Insert Design Handbook [3]
there are three basic types of potted inserts – partially potted, fully
potted and the through the thickness. The three types are illustrated in
Fig. 1 a). In addition, the primary loading conditions for sandwich in-
serts are depicted in Fig. 1 b). The different types have their specific
field of application. Partially potted inserts add little mass to a sand-
wich panel regardless the panel thickness. However, they also generally
have the lowest load bearing capability. Fully potted inserts enable
higher load introduction due to the increased interface area with the
core and the direct bond to the opposite face sheet. Through-the-
thickness (TTT) inserts, can be applied for bushings and they enable
symmetric and thus higher in-plane (shear) load introduction. How-
ever, in general sandwich inserts are more prone to out-of-plane loading
such as pull-out. In addition, these loads typically also lead to a more
complicated failure behavior. Therefore, pull-out loading generally re-
ceives more attention when dimensioning sandwich panel inserts and it
is also subject of the present study.
Due to their importance in sandwich construction, inserts have been
investigated extensively in experimental studies in the past. Kim and
Lee [4] investigated the load transfer of partially potted inserts in
composite sandwich panels for different insert geometries. Song et al.
[5] studied different configurations of potted inserts in Nomex honey-
comb sandwich panels with carbon epoxy face sheets. They reported a
strong relationship between insert pull-out strength and core height,
core density as well as face sheet thickness, while the shear-out strength
was dominated by the face thickness. Demelio et al. [6] tested different
combinations of honeycomb sandwich panel fasteners under shear and
pull-out static and fatigue loading. They concluded, that the face sheet
strength and core height dominate the fatigue strength.
In addition to the previously introduced purely experimental stu-
dies, there are multiple publications on sandwich panel inserts, which
implement simulation models for a better understanding of the failure
mechanisms or for the prediction of failure. The available models can
be distinguished in analytical and finite element models. The analytical
models are based on the higher order sandwich theory (HOST).
Prominent examples for such models include the works of Thomsen and
Rits [7,8], Frostig et al. [9] and Bozhevolnaya et al. [10]. These models
address different insert types and make different assumptions regarding
boundary conditions and higher order sandwich theory. However, they
have in common that they assume isotropic material behaviour and
their application is limited to early design estimations and for deriving
design guidelines. Additional examples for the application of HOST
models include Bull and Thomsen [11], who developed a design tool for
initial dimensioning of inserts in sandwich panels and compared the
model performance with experimental data as well as finite element
analysis (FEA) predictions. A more recent study is the work of Smith
and Banerjee [12], who applied the Thomsen model to perform relia-
bility analyses on the strength of sandwich panel inserts comparing
different reliability analysis methods.

⁎
Corresponding author.
E-mail address: ralf.seemann@diehl.com (R. Seemann).

https://doi.org/10.1016/j.compstruct.2018.07.028
Received 7 April 2018; Received in revised form 3 June 2018; Accepted 4 July 2018
Available online 05 July 2018
0263-8223/ © 2018 Elsevier Ltd. All rights reserved.
R. Seemann, D. Krause
a) Partially potted
Fully potted Through-the-thickness
Fig. 1. a) Typical potted insert configurations (based on [3]) and b) primary loading conditions for inserts.
However, the majority of available simulation models are based on
FEA with greatly varying level of detail. Early FE-models were often
based on axisymmetry, disregarding the irregular potting to honeycomb
interface similar to the HOST models [11,13,14]. Later, with increasing
computational capabilities more detailed 3D FE-models emerged. Bu-
nyawanichakul et al. [15,16] tested countersunk titanium fasteners
bonded in honeycomb sandwich panels on a test rig, which allows to
apply pre-stress to the fastener. In addition, they developed a FE-model
supported by constituent tests on the potting material and the honey-
comb core, while the honeycomb core was modelled using 3D-con-
tinuum elements. Nguyen et al. [17] studied various foam sandwich
panel joints in experiments and developed a FE model, while comparing
different failure modeling methods. Heimbs and Pein [18] tested dif-
ferent configurations of inserts in Nomex honeycomb sandwich panels.
They derived 3D-continuum models based on spotweld elements for an
implementation in a global non-linear model of aircraft interior com-
ponents. They also implemented a detailed FE-model of a honeycomb
sandwich insert, where the hexagon core geometry is modeled accu-
rately. Such detailed meso-scale models of sandwich panels with
structured cores have seen increasing applications, in particular for
impact and crushing analyses of sandwich panels [19–25]. However,
there are additional applications for sandwich panel inserts evident.
Bianchi et al. [26], investigated hot and cold bonded procedures of
honeycomb sandwich insert manufacturing. They implemented a linear
detailed FE-model and concluded that the stiffness of the potting ma-
terial has a significant impact on the insert joint strength. Roy et al.
[27] conducted experimental studies on honeycomb sandwich panel
inserts and derived the orthotropic material properties of the Nomex
cell wall material using a detailed meso-model of the joint. Silmane
et al. [28] investigated the sensitivity of the actual potting shape of
aluminum honeycomb inserts depending on the position of the insert in
the hexagon grid. In a previous study of Seemann und Krause [29],
detailed meso-scale models were developed for fully-potted insert
configurations under pull-out loading. Further examples of detailed
honeycomb joint models include the analysis of the thermal coupling of
sandwich inserts used in satellites [30] as well as an investigation of the
potting shrinkage during insert manufacturing [31].
Fig. 2 illustrates the four prevailing types of simulation models for
ŶĂůLJƟĐĂůŵŽĚĞůƐ Quasi 3D-continuum
Ⱥ
f1 (axisymmetric)
T c
M
f2
Q r
h
FEͲDŽĚĞůƐ
Fig. 2. Types of simulation models for sandwich panel inserts.
102
Composite Structures 203 (2018) 101–109
b)
Pull-out
(out-of-plane) Shear (in-plane)
sandwich panel inserts based on the reviewed literature. The available
computational models have in common that they were either applied
for preliminary design studies or for understanding the mechanical ef-
fects that lead to failure. However, the potential of the available
modelling approaches for an accurate strength prediction, including not
only the start of non-linearity but also the ultimate strength of sandwich
panel inserts has not been demonstrated yet. The present work was
intended to explore the capabilities of state of the art non-linear FE-
models for predicting the damage progression of sandwich panel inserts
under pull-out. Furthermore, the present study adds the example of a
partially potted insert to the existing literature on experimental and
computational analysis of sandwich panel inserts, which predominately
covers fully potted or through-the-thickness configurations.
2. Materials
The investigated materials represent a typical partially potted insert
configuration in aircraft interior applications. The face sheets com-
prised a single layer of glass fiber fabric reinforced phenolic resin
prepreg. The used prepreg corresponds to the Airbus standard
ABS5047-07 [32]. The core was a Nomex honeycomb core with 3.2 mm
cell width and a density of 48 kg/m3. The base panel had a nominal
thickness of 26 mm and was manufactured industrially using co-curing
flat press moulding. A light weight insert of the type ABS1005 was cold
bonded in the panel using a two-component epoxy adhesive. This insert
is made of a plastic body with a bonded steel thread. The average
potting depth was about 11 mm. The investigated material configura-
tion along with key geometric dimensions are illustrated in Fig. 3.
3. Experimental study
In the experimental study a total of 12 specimens were tested using
a typical insert pull-out fixture as suggested by the Insert Design
Handbook [3]. This setup is illustrated in Fig. 4. The circular cut-out of
the fixture had a diameter of D = 80 mm, while the specimens had a
cross section of 100 mm × 100 mm. The tests were conducted on a
Galdabini Quasar 100 universal testing machine, which was equipped
with a HBM S9M-10 kN load cell. The displacement was recorded
3D-continuum Detailed meso-scale
R. Seemann, D. Krause Composite Structures 203 (2018) 101–109

ABS1005
ABS 5047-07
Epoxy adhesive
Nomex honeycomb
3.2mm – 48kg/m³

11 mm
26 mm ABS1005
Ø14 mm
Plastic body
8 mm Steel thread

Fig. 3. Materials of investigated insert configuration.

ØD

Fig. 6. Identified damage mechanisms.

Fig. 4. Sketch of experimental setup according to Insert Design Handbook [3].

after the peak force. The governing damage mechanisms were in-
vestigated based on external inspection during and after the tests as
directly from the machine crosshead.
well as based on cut section of the specimens. The four identified da-
The experimental results for all specimens are given as force-dis-
mage mechanisms are depicted in Fig. 6. Shear buckling of the cell
placement curves in Fig. 5. The pull-out strength in terms of peak force
walls at the interface between potting and core is well known as first
ranged between 1600 N and 1300 N, which corresponds to a scatter of
damage mechanism for typical honeycomb sandwich insert configura-
about 20%. It is assumed that this considerable scatter largely origi-
tions [5,27,29,33] and this was also confirmed in the present study. The
nates from the varying effective potting radius in the specimens, which
rapid stiffness degradation and thus catastrophic failure is initiated by
has significant impact on the pull-out strength of sandwich inserts [3].
tensile failure of the core in out-of-plane direction beneath the potting.
The varying potting radius is a result of the random number of hon-
Face sheet failure and debonding of core and face only occurred well
eycomb cells being opened by the borehole during insert manu-
beyond the peak force. Unlike the previously studied fully potted and
facturing. This effect was studied in detail by Silmane et al. [28] and it
through-the -thickness inserts [5,27,29,33], the partially potted insert
was not subject of the present work. Despite the scatter, the curve
of the present study did not indicate significant further load increase
progression was consistent for all specimens. It was characterized by
after cell wall shear buckling. It is assumed that this is due to two ef-
three stages, linear elastic deformation followed by quadratic curve
fects. Firstly, the tensile strength of the core is lower than the adhesive
flattening from about 1000 N and rapid stiffness degradation shortly
bond of the potting to the lower face in the fully potted and through-the
-thickness configurations. Secondly, the potting-core interface area is
reduced in case of partially potted inserts due to the limited depth of the
potting. This leads to reduced shear load transmission, which in turn
results in increased normal stress beneath the potting. Both effects lead
to premature failure in the load path below the potting, when com-
paring fully and partially potted inserts.

4. Numerical models

The investigated insert configuration consists of four constituents.

In the following, the modelling approach for each constituent is de-
scribed before the overall model is introduced. All simulations were
performed using ABAQUS/Explicit 6.14. Despite the quasi-static nature
of the problem, the explicit solver was eventually selected, since im-
plicit time integration using ABAQUS/Standard did not converge due to
the multitude of damage mechanisms.

Fig. 5. Experimental results as force-displacement relationship.

103
R. Seemann, D. Krause
4.1. Insert
The applied insert is made from a plastic body, which is reinforced
with a metal threaded bushing. The exact material of the insert is un-
known. Since the experimental study did not indicate any damage of
the insert, a simple isotropic elastic material model with an estimated
Young’s modulus of E = 5000 MPa was implemented. The threaded
bushing was modelled as rigid element, which also served as load in-
troduction point.
4.2. Potting/adhesive
In sandwich panel insert manufacturing, the adhesive, which is used
to bond the insert in the panel, fills a core cavity and thus forms a solid
component after curing. The material parameters of cured adhesives are
generally not given by manufacturers. However, Bianchi et al. [26]
showed that the mechanical properties of the potting have a noticeable
effect on the insert pull-out behavior. Therefore, the applied epoxy
adhesive was characterized in tension and compression tests based on
ASTM D638 [34] and ASTM D695 [35] respectively. A suitable material
model was derived and calibrated, using the experimentally determined
stress-strain behavior. The resulting isotropic elasto-plastic material
model has a different plastic stress-strain progression in compression
and tension. Fig. 7 illustrates the implemented material model graphi-
cally in terms of stress-strain relationship.
4.3. Core
Based on the experimental study the core can be considered as most
crucial constituent for the overall failure behavior, since it not only
governs the first damage mechanism of cell wall shear buckling but also
catastrophic failure due to tensile core failure beneath the potting. With
the objective of accurately predicting the pull-out strength of sandwich
inserts, advanced modelling approaches are appropriate. With regards
to Fig. 2, this corresponds to 3D-continuum and detailed meso-scale
models. These two approaches mainly differ in the implemented core
model. 3D-continuum models require considerably less modelling and
computational effort, while the capabilities to represent core damage
are limited by its homogenized nature. Detailed meso-scale models, on
the other hand, enable to represent accurate cell wall deformations and
damage mechanisms at the cost of greatly increased effort. Therefore,
the questions arises whether 3D-continuum models are sufficient or
detailed models are required in order to predict the pull-out strength of
honeycomb sandwich inserts. This was addressed in the present study
by comparing the two approaches.
Both approaches are well described in the literature. In case of
40
20
Tension
Stress [MPa]
0
-20
-40
Compression
-60
-80
-0.08 -0.06 -0.04 -0.02 0 0.02 0.04
Strain [-]
Fig. 7. Stress-strain relationship of implemented material model for the ad-
hesive/potting.
104
Composite Structures 203 (2018) 101–109
Meso scale 3D-continuum
Fig. 8. Core modelling approaches in case of compression.
detailed models, there is a comprehensive study on numerical model-
ling of the same Nomex honeycomb such as investigated in the present
work. In this previous study of Seemann and Krause [23], different
meso-scale modelling approaches were benchmarked based on com-
pression, tension, and transverse shear tests. They recommended a
single layer orthotropic plastic material model for the cell wall mate-
rial. In the present work, this material model with the same material
properties was implemented for the detailed core. In addition, a 3D-
continuum material model was derived and calibrated with the same
experimental results. A homogenized orthotropic plastic material model
was implemented, while the calibration was done using a single 8-node
element. Fig. 8 illustrates both modelling approaches in case of com-
pression after damage occurred. It can be seen that the detailed meso
scale model is capable to accurately represent the buckling of the
honeycomb cell walls, while the 3D-continuum model simply re-
presents plastic deformation as defined in the material model.
Fig. 9 depicts the stress-strain curves of both implemented core
models in comparison to the experimental results, which were used for
model calibration. Since the transverse shear properties of honeycombs
depend on the orientation, two shear plots are given, while standard
honeycomb nomenclature is used to refer to the load direction [2]. It
becomes apparent that the meso scale model reflects the actual curve
progression from the experiments well. Differences remain regarding
the plateau strength in compression and the shear strengths, which are
slightly overestimated in case WT-shear and underestimated for LT-
shear.
The 3D-continuum model allows to calibrate the strength in each
loading direction using independent variables. Therefore, the strengths
are matched accurately. The following curve progression was simply
approximated using few additional stress–strain data points. Regarding
the remaining in-plane core properties as well as Poisson’s ratios no
experimental reference data was available. However, it is common
practice to make assumptions regarding these properties. Bitzer [2]
suggests to use 1% of the known out-of-plane properties for the re-
spective in-plane properties, while he suggests a Poisson’s ratio of 0.1 in
all directions. This approach was applied in the present work.
In conclusion, it can be said that both implemented core models
represent the experimentally determined macroscopic core behavior
with regards to the decisive shear loading directions well.
4.4. Face
The glass fiber fabric reinforced phenolic resin faces were im-
plemented using an orthotropic material model based on the continuum
damage mechanics model of Johnson [36]. This material model is
available in ABAQUS/Explicit as pre-compiled user subroutine
(VUMAT). It enables progressive stiffness degradation due to fiber/
matrix cracking, while compressive and tensile strengths can be defined
independently in the two material directions. In addition, it is capable
to represent the shear plasticity of fabric reinforced composites. The
R. Seemann, D. Krause Composite Structures 203 (2018) 101–109

Fig. 9. Stress-strain relationship of both implemented core models in comparison to experimental results from a previous study [23].

Table 1
Implemented material properties of face sheet.
E1 [MPa] E2 [MPa] G12 [MPa] ν12 [–] σ1t [MPa] σ2t [MPa] σ1c [MPa] σ2t [MPa] Ply th. [mm]

23,200 21,500 4980 0.15 322 250 220 180 0.25

not given as single material parameter. Instead, the shear stress-strain

relationship of the implemented material model is illustrated in Fig. 10
for reference. The face sheets were modelled with a ply thickness of
0.25 mm.

4.5. Overall model

The previously described constituent models were assembled to two

Fig. 10. Shear stress-strain relationship of implemented face sheet material.
applied material properties are summarized in Table 1, where the 1-
material direction refers to the warp direction of the fabric. The uni-
directional material properties were determined in coupon tests of the
fabric as well as based on four-point bending tests of the base panel. The
shear properties were determined from picture frame shear tests of the
base panel. Due to the shear plasticity of the fabric, the shear strength is
overall models for insert pull-out. The two models only differ in the
core modelling approach, while both introduced core models, meso
scale and 3D-continuum, were implemented. The models feature an
irregular potting-to-core interface accounting for the hexagon grid
which leads to varying potting geometries. This interface was modelled
as tied contact, since no debonding was evident in the experimental
study. The same applied for the face-to-potting interface. In contrast,
there was debonding of the face sheet and core visible during the tests
and when inspecting the specimen after testing. However, this only
occurred well beyond the peak force in the post failure regime.
Therefore, this effect was neglected and the core-to-face interface was
modelled as tied contact as well. It is noteworthy, that the model can
represent mode I fracture beneath the core-to-face interface due to
tensile failure of the core. The fixture was modelled using 8-node
continuum elements and an elastic isotropic material model with a

105
R. Seemann, D. Krause Composite Structures 203 (2018) 101–109

Face Tx͕Ty͕Tz = 0 Face-Įxture interface 3D-conƟnuum core

S4R | Krth. fabric (VUMAT) Rx͕Ry͕Rz = 0 Penalty contact
0.4-2mm elemsize

Fixture Face-core bond

C3D8R | /sotr. elastic Tied contact
5mm elemsize

Tz = v
Tx͕Ty = 0 z
Meso scale core
y x

Honeycomb
S4R | Krth. plasƟc
0.4/1.5mm elemsize
C3D8R | Krth. plastic
2mm elem size
WoƫnŐ Insert WoƫnŐ-core bond
C3D10M | /sotr. bi-plastic C3D8Z ͮ ElasƟc Tied contact
0.8mm elemsize 0.5mm elemsize

Fig. 11. Implemented overall models with summary of implemented materials and model details.

Young’s modulus of 210,000 MPa. The interface between fixture and a) 2000
top face was modelled as penalty contact. This allowed the panel to Test scatter
separate from the fixture due to bending. Similar to previous publica- Test average
Meso scale
tions [16,18,27,29], the model was implemented as quarter model in 1500
3D-continuum
order to utilize the symmetry of the specimen and loading condition,
Force [N]

thus reducing the computational cost. The simulations were carried out
using mass scaling (initial time step t = 3.5e-6 s) and increased dis- 1000
placement rate (v = 10 mm/s). Both numerical parameters were set
based on convergence studies using the peak load as reference. The
same applied for the eventually implemented mesh sizes of the different 500
constituents. The overall model is illustrated in Fig. 11. The computa-
tional time for the 3D-continuum model was about 7 min, while the
detailed meso model required 87 min to run. The simulations were run 0
using four cores on a state of the art Intel® Core™ i7 workstation. 0 1 2 3 4
Displacement [mm]
5. Results and discussion
b) 2000
The simulation results of both models in comparison to the ex-
perimental results are given in terms of force-displacement relationship
in Fig. 12 a). The experimental results are plotted in the background, 1500
while also the scatter is indicated by light grey lines. For a clearer
Force [N]

picture on the damage initiation, Fig. 12 b) illustrates the results only

up until a displacement of 1.25 mm. Macroscopically, both simulation 1000
models yield similar results with regards to initial stiffness, peak force
and stiffness degradation. Both models match the initial stiffness of the
experiments well, while the simulated strength lies in the scatter of the Test scatter
500
Test average
experiments. The curve progression in the following stiffness degrada-
Meso scale
tion phase is not matched as accurately by the simulation models. The 3D-continuum
experiments were characterized by a continuous and moderate decrease 0
in stiffness, while the simulation models indicate a sudden load drop 0 0.5 1
shortly after the peak force. The carried load then stabilizes at a plateau Displacement [mm]
before entering a continuous stiffness degradation similar to the ex-
Fig. 12. Comparison of simulations and experiments in terms of force-dis-
periments. In general, load plateaus were also evident in the experi- placement relationship a) full displacement range and b) close up on damage
ments, however, they were not as distinct and at considerably lower initiation.
load levels. In total, the simulation models generally underestimate the
stiffness degradation after peak force.
slightly lower strength. In addition, it is characterized by a more brittle
When comparing the results of the two simulation models in more
failure. While the meso scale model clearly reproduces the
detail it becomes apparent that the 3D-continuum model indicates a

106
R. Seemann, D. Krause
Fig. 13. Force-time plot of both simulation models as reference for visual output in Fig. 14.
characteristic quadratic flattening resulting from core shear failure
before catastrophic failure, this effect is significantly less pronounced in
case of the 3D-continuum model. The following plateau is on the same
load level for both models. However, the continuous stiffness de-
gradation is initiated earlier in case of the meso scale model.
In order to retrace the damage mechanisms of the simulation models
and map them onto the curve progression, an additional force plot is
given in Fig. 13. However, here the force is plotted over the simulation
time. The given x-axis labels correspond to the five simulation stages
that are illustrated in Fig. 14. For each stage, a top view of the specimen
with an indication of face sheet damage (stiffness degradation in the
material directions 1 and 2) is given along with a cross sectional view,
which indicates the core damage evolution. In case of the 3D-con-
tinuum model, a shear stress contour is given in order to give better
indication the shear damage. In addition, several magnified views are
given for the detailed meso-scale model to illustrate local core damage.
In the following, the different stages are briefly analyzed considering
both, the force-time plot (Fig. 13) and the visual simulation results
(Fig. 14).
• 0.075 s Both models are shortly before catastrophic failure. No face
sheet or tensile core damage is evident. However, core shear damage
is visible in both models. In case of the meso scale model, the shear
damage is visible as buckling of the single cell walls adjacent to the
potting. Analogous, the 8-node elements connected to the potting
indicate shear distortion in case of the 3D-continuum model.
Furthermore, the single cell walls in the center beneath the potting
appear to have started buckling as well.
• 0.150 s Both models are past catastrophic failure in a plateau with
constant load. Tensile core failure beneath the potting is clearly
evident in both models, while the shear damage adjacent to the
potting progressed. In the meso scale model, the cell wall buckling
underneath the potting extended to the entire projected cross sec-
tion area of the potting. There is limited face sheet damage is in both
models, while no elements indicate fully degraded stiffness (100%
damage).
• 0.250 s The meso scale model is in the continuous stiffness de-
gradation phase, while the 3D-continuum model remains on the
plateau. This corresponds to the face sheet damage, which shows
many fully degraded elements in case of the meso scale model, while
the face sheets in the 3D-continuum model still indicate con-
siderably less damage. In both models core shear and tensile damage
further progressed and for the meso model tearing of the cell walls is
evident adjacent to the potting. The cell wall buckling below the
potting mostly disappeared.
• 0.350 s The 3D-continuum model also entered the final continuous
stiffness degradation phase and large scale face sheet damage is
evident for the first time. The damage in the meso model further
107
Composite Structures 203 (2018) 101–109
progressed and delamination of face and core becomes evident in
terms of tensile core failure surrounding the potting. The face sheet
damage in the 3D-continuum model appears to be more distributed
if compared to the meso model.
These observations support the general understanding of the da-
mage evolution in the experiments. The first damage mechanism is core
shear buckling of the honeycomb. No face sheet damage is evident
before catastrophic failure, which is initiated by tensile core failure.
Load drops in the post failure regime are due to suddenly progressing
cracks in the face sheet. A new effect, which has not been described
before is the buckling of single cell walls underneath the potting.
Despite the overall good representation of the damage progression of
the experiments, the simulation is not capable to reproduce the ex-
periments in terms of curve progression in the post failure regime. In
order to fully understand this, further investigations are required.
However, it is assumed that it is largely due the core, which dominates
the damage behavior of inserts under pull-out loading. At the same
time, Nomex honeycomb is a complex composite material, which is
challenging to represent in numerical models and the implemented
detailed meso scale model of the honeycomb is still characterized by
several simplifications [23].
6. Conclusion and outlook
In concluding remarks are summarized in the following.
• 3D-continuum and detailed meso scale models enable to represent
the decisive damage mechanisms, tensile and shear failure of the
core accurately and thus are able to predict the insert pull-out
strength based on material properties, which were calibrated using
standard sandwich core tests.
• The detailed meso scale model performed slightly better. This is
evident in the more accurate representation of the core shear da-
mage prior to catastrophic failure. The same applies for the stiffness
•
degradation following the peak force.
In the post failure regime, both models are not capable to capture
the continuous stiffness degradation as in the experiments
• Both models give similar results, while the 3D-continuum model
requires significantly reduced modelling and computational effort
(12 times less computing time). Therefore, the 3D-continuum model
is recommended for general engineering applications
Future investigations should focus on a more accurate representa-
tion of the core damage progression. In case of the 3D-continuum
model, a failure model could be implemented, allowing to delete ele-
ments at high strains and thus accounting for honeycomb tearing.
Similarly, for the detailed meso model there are more sophisticated
R. Seemann, D. Krause Composite Structures 203 (2018) 101–109

Time [s] 3D-continuum Meso scale

0.075

0.150

0.250

0.350

Fig. 14. Damage mechanisms in simulation at different stages.

108
R. Seemann, D. Krause Composite Structures 203 (2018) 101–109

modelling approaches available in the literature, which could be im- [17] Nguyen K-H, Park Y-B, Kweon J-H, Choi J-H. Failure behaviour of foam-based
plemented in the future. sandwich joints under pull-out testing. Compos Struct 2012;94(2):617–24.
[18] Heimbs S, Pein M. Failure behaviour of honeycomb sandwich corner joints and
inserts. Compos Struct 2009;89(4):575–88.
Acknowledgements [19] Castanié B, Aminanda Y, Barrau J-J, Thevenet P. Discrete modeling of the crushing
of nomex honeycomb core and application to impact and post-impact behavior of
sandwich structures. In: Abrate S, Castanié B, Rajapakse YDS, editors. Dynamic
The presented work was partially funded by the German Federal Failure of Composite and Sandwich Structures. Dordrecht: Springer, Netherlands;
Ministry for Economic Affairs and Energy, Germany, in the framework 2013. p. 427–89.
of LuFo V-2 as part of the research project ‘SYLVIA’. [20] Asprone D, Auricchio F, Menna C, Morganti S, Prota A, Reali A. Statistical finite
element analysis of the buckling behavior of honeycomb structures. Compos Struct
2013;105:240–55.
References [21] Fischer S, Drechsler K, Kilchert S, Johnson A. Mechanical tests for foldcore base
material properties. Compos A Appl Sci Manuf 2009;40(12):1941–52.
[22] Foo C, Chai G, Seah L. A model to predict low-velocity impact response and damage
[1] Zenkert D. The handbook of sandwich construction. Cradley Heath, West Midlands:
in sandwich composites. Compos Sci Technol 2008;68(6):1348–56.
Engineering Materials Advisory Services; 1997.
[23] Seemann R, Krause D. Numerical modelling of Nomex honeycomb sandwich cores
[2] Bitzer T. Honeycomb technology: materials, design, manufacturing, applications
at meso-scale level. Compos Struct 2017;159:702–18.
and testing. 1st ed. London, New York: Chapman & Hall; 1997.
[24] Giglio M, Manes A, Gilioli A. Investigations on sandwich core properties through an
[3] ESA. Space engineering Insert Design Handbook; 2011.
experimental–numerical approach. Compos B Eng 2012;43(2):361–74.
[4] Kim BJ, Lee DG. Characteristics of joining inserts for composite sandwich panels.
[25] Heimbs S. Virtual testing of sandwich core structures using dynamic finite element
Compos Struct 2008;86(1–3):55–60.
simulations. Comput Mater Sci 2008;45(2):205–16.
[5] Song K-I, Choi J-Y, Kweon J-H, Choi J-H, Kim K-S. An experimental study of the
[26] Bianchi G, Aglietti GS, Richardson G. Static performance of hot bonded and cold
insert joint strength of composite sandwich structures. Compos Struct
bonded inserts in honeycomb panels. J Sandwich Struct Mater 2011;13(1):59–82.
2008;86(1–3):107–13.
[27] Roy R, Nguyen KH, Park YB, Kweon JH, Choi JH. Testing and modeling of Nomex™
[6] Demelio G, Genovese K, Pappalettere C. An experimental investigation of static and
honeycomb sandwich Panels with bolt insert. Compos B Eng 2014;56:762–9.
fatigue behaviour of sandwich composite panels joined by fasteners. Compos B Eng
[28] Slimane S, Kebdani S, Boudjemai A, Slimane A. Effect of position of tension-loaded
2001;32(4):299–308.
inserts on honeycomb panels used for space applications. Int J Interact Des Manuf
[7] Thomsen OT, Rits W. Analysis and design of sandwich plates with inserts—a high-
2017;16:131.
order sandwich plate theory approach. Compos B Eng 1998;29(6):795–807.
[29] Seemann R, Krause. D. Virtual testing of Nomex honeycomb sandwich panel inserts.
[8] Thomsen OT. Sandwich plates with ‘through-the-thickness’ and ‘fully potted’ in-
20th International Conference on Composite Materials 2015;19-24 July -
serts: Evaluation of differences in structural performance. Composite Structures
Copenhagen, Denmark.
1997;40(2):159–74.
[30] Boudjemai A, Mankour A, Salem H, Amri R, Hocine R, Chouchaoui B. Inserts
[9] Frostig Y, Baruch M, Vilnay O, Sheinman I. High-Order theory for sandwich-beam
thermal coupling analysis in hexagonal honeycomb plates used for satellite struc-
behavior with transversely flexible core. J Eng Mech 1992;118(5):1026–43.
tural design. Appl Therm Eng 2014;67(1–2):352–61.
[10] Bozhevolnaya E, Lyckegaard A, Thomsen OT, Skvortsov V. Local effects in the vi-
[31] Courteau-Godmaire H, Fotsing ER, Causse P, Lévesque M, Ruiz E. Modeling of Resin
cinity of inserts in sandwich panels. Compos B Eng 2004;35(6–8):619–27.
Shrinkage around Insert in Composite Sandwich Panels. 20th International
[11] Bull PH, Thomsen OT. Development of a Design Tool for Initial Analysis of Inserts in
Conference on Composite Materials; 2015. 19–24 July – Copenhagen, Denmark.
Sandwich Structures. 8th international conference on sandwich structures
[32] Airbus SAS. Engineering Directorate. ABS5047 – Aerospace series: Woven glass
2008(Porto).
fiber reinforced thermoset prepreg (e.g. phenolic) Flame retarded (FST). Issue 10.
[12] Smith B, Banerjee B. Reliability of inserts in sandwich composite panels. Compos
Blagnac, France; 2015.
Struct 2012;94(3):820–9.
[33] Park Y-B, Kweon J-H, Choi J-H. Failure characteristics of carbon/BMI-Nomex
[13] Tsouvalis NG, Kollarini MJ. Parametric study of stress concentrations caused by
sandwich joints in various hygrothermal conditions. Compos B Eng
inserts in sandwich panels. Proceedings to the 12th European Conference on
2014;60:213–21.
Composite Materials 2006;29 Aug – 1 Sep, Biarritz, France.
[34] ASTM International. D 638 – 02a Standard Test Method for Tensile Properties of
[14] Raghu N, Battley M, Southward T. Strength variability of inserts in sandwich panels.
Plastics (D 638 – 02a). West Conshohocken, PA, USA; 2002.
8th international conference on sandwich structures 2008;6–8 May, Porto.
[35] ASTM International. D 695 – 02a Standard Test Method for Compressive Properties
[15] Bunyawanichakul P, Castanie B, Barrau J-J. Experimental and Numerical Analysis
of Rigid Plastics(D 695 – 02a). West Conshohocken, PA, USA; 2002.
of Inserts in Sandwich Structures. Appl Compos Mater 2005;12(3–4):177–91.
[36] Johnson AF. Modelling fabric reinforced composites under impact loads. Compos A
[16] Bunyawanichakul P, Castanié B, Barrau J-J. Non-linear finite element analysis of
Appl Sci Manuf 2001;32(9):1197–206.
inserts in composite sandwich structures. Compos B Eng 2008;39(7–8):1077–92.

109