You are on page 1of 14

Composite Structures 89 (2009) 575–588

Contents lists available at ScienceDirect

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

Failure behaviour of honeycomb sandwich corner joints and inserts


Sebastian Heimbs a,*, Marc Pein b
a
EADS Innovation Works, 81663 Munich, Germany
b
Hamburg University of Technology, Denickestraße 17, 21073 Hamburg, Germany

a r t i c l e i n f o a b s t r a c t

Article history: In nearly all sandwich constructions certain types of joints have to be used for assembly, but little is
Available online 11 December 2008 known about their failure behaviour. This paper deals with the investigation of the mechanical behaviour
of three different corner joints as a right-angled connection of two sandwich panels and of two different
Keywords: potted inserts as a localised load introduction in NomexÒ honeycomb sandwich structures with glass
Honeycomb sandwich fibre-reinforced composite skins. For this purpose, experimental test series were conducted including
Corner joint failure behaviour shear tests and bending tests of the corner joints and pull-out as well as shear-out tests of the threaded
Insert failure behaviour
inserts. The failure mechanisms and sequences are described for each load case and the influence of the
Finite element modelling
different designs and of the loading rate is discussed. Based on these characteristics, finite element sim-
ulation models were developed in LS-DYNA, which are able to represent the respective failure behaviours.
Ó 2008 Elsevier Ltd. All rights reserved.

1. Introduction very seldom documented and can be found for insert type a
according to Fig. 1 in [1] and for type f in [2–4]. Raghu, Battley
Sandwich structures with composite skins and a honeycomb and Southward [5] investigated the influence of potting diameter
core are widely used, especially in the aerospace industry, due to on the pull-out failure behaviour of insert type f. A recent study
their superior weight-specific bending stiffness and strength prop- [6] investigates the influence of honeycomb core height, density
erties. The failure behaviour of such sandwich panels is rather and skin thickness on the failure behaviour of insert type f under
complex and has been investigated in numerous research studies pull-out and shear-out loading. The fatigue behaviour of insert
in the past. However, in virtually all technical sandwich construc- joints under pull-out load was treated in [7–9]. Further papers deal
tions these panels have to be connected to subcomponents or pan- with pull-out tests of inserts in foam core sandwich, like in [10]
els have to be joined, and these joints are potential locations of (types f, g and h), [11] (type f), [12] (type b) and [13] (partial metal
failure as well, which have not been adequately treated in the tech- inserts). Pull-out tests in balsa core sandwich structures are docu-
nical literature. mented in [14] (insert type f) and [15] (types b, f and h). The shear-
A number of different methods exist, how to introduce localised out failure behaviour of a metallic bolt in a foam core sandwich
loads into a sandwich structure, several of which are illustrated in was analysed by Mares et al. [16].
Fig. 1. Especially in aerospace design, threaded inserts, bonded into In addition to these few quasi-static experimental studies,
the cellular core, are classically used for this purpose. The major Thomsen [17,18] approached this topic analytically using a higher
task of such an insert is to adequately transfer the load into the order sandwich plate theory. As an example load case, he investi-
sandwich skins. In practice, tensile loads normal to the sandwich gated the normal loading of a potted insert in an aluminium hon-
surface and shear loads parallel to the surface are most relevant, eycomb sandwich structure. Numerical finite element analyses
since localised compression and bending loads are typically allow for the visualisation of stress distributions in the insert, core
avoided due to large mounting surfaces, and torsion only occurs and skins and have been performed for pull-out loads in honey-
during the assembly of the construction and not in service. Because comb core sandwich in [1,19–21] and for foam core sandwich in
of usually very thin sandwich skins, the aim is to transfer the load [22–26].
into a preferably large area and into both skins. In case of honey- Besides other general monographs on sandwich structures con-
comb sandwich structures this may be achieved by filling the cells taining information on insert design [27,28], the most comprehen-
with a potting compound in the insert installation area. Pull-out sive collection of failure mode descriptions, test recommendations
tests of inserts normal to the honeycomb sandwich structure are or design guidelines can be found in the Insert Design Handbook
[29] of the European Space Agency (ESA), which is also used as a
* Corresponding author. Tel.: +49 89 607 25884; fax: +49 89 607 23067.
reference in most of the other papers listed here. Besides numerous
E-mail address: sebastian.heimbs@eads.net (S. Heimbs). strength vs. core height diagrams for different load cases, insert

0263-8223/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compstruct.2008.11.013
576 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

Fig. 1. Overview of different methods for local load introductions into honeycomb sandwich structures.

diameters and honeycomb core types, also analytical equations This paper intends to gain insight into the failure behaviour of
for the estimation of pull-out and shear-out strength values are inserts and corner joints of sandwich structures and to cover as-
given. pects that have not been treated previously. The material used in
Compared to these few studies and information on inserts in this study is the most relevant sandwich structure used in the air-
honeycomb sandwich structures, almost no literature is published craft industry with NomexÒ aramid paper honeycomb cores and fi-
on the failure behaviour of sandwich corner joints, also called L- bre-reinforced composite skins. Two different insert types and
joints. Some industrial relevant design options of corner joints, three different corner joints are tested under various loading con-
which are illustrated in Fig. 2, are shown in [27,28,30,31]. In the ditions including pull-out, shear and bending, and their failure
textbook of Noakes [32] a comprehensive overview on the manu- behaviour is characterised in detail. In addition to static testing,
facturing of folded corner joints is given, which is a state-of-the- also the influence of the loading rate is addressed. Furthermore,
art technique in modern sandwich design, often found in aircraft analytical calculations of the failure loads are presented and com-
interior components. Such folded corner joints of type f according pared to the experimental results. Subsequently, within a numeri-
to Fig. 2 in aluminium honeycomb sandwich specimens were cal analysis with the explicit finite element code LS-DYNA,
tested by Joulia and Grove [33], providing the only information simulation models are generated and modelling methods to cover
on failure loads. However, the state of stress in their combined the failure behaviour are derived. In this context, both detailed
bending-shear test is not clear and no comparison to other joint meso-scale models and rather simple macro-models are covered.
designs is given. Furthermore, no detailed description of the failure
behaviour is provided in this study. Carruthers [34] investigated
the crash performance of foam core sandwich crash boxes with 2. Failure behaviour of inserts
two different corner joints (type b and e) experimentally without
characterising the single corner joints separately. The only finite 2.1. Materials and specimen manufacturing
element analysis of three different L-joint designs of type b in alu-
minium honeycomb sandwich is documented in [35]. However, The following study is focused on honeycomb sandwich panels
not the failure characteristics but only the dynamic behaviour with two different standard metallic, threaded inserts used in the
was investigated in a modal analysis. aerospace industry. The 15 mm thick sandwich structure consists

Fig. 2. Overview of different methods of corner joints in honeycomb sandwich structures.


S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 577

of a NomexÒ honeycomb core with a density of 48 kg/m3 and a cell drilled in the middle of the potted area, the insert was positioned
size of 3.2 mm (type Schütz Cormaster C1-3.2-48). The skins are inside this hole, and an epoxy resin (type Huntsman AralditeÒ
made of one single prepreg ply of woven (satin weave 1/7) E-glass 2011) was injected through an opening for the fixation of the insert
fibre-reinforced phenolic resin (type Stesalit PHG 600-68-50) with (Fig. 3). The final test specimens were cut to a size of
an average cured ply thickness of 0.24 mm and a fibre volume frac- 127 mm  127 mm, each having one insert in the middle.
tion of 48%. The material properties are shown in Table 1. Before
curing of the sandwich specimens, the honeycomb cells within a 2.2. Experimental procedures
circle of approx. 38 mm diameter were filled with an epoxy-based
potting material (type Cytec BRÒ 632 P4) for the later positioning The pull-out and shear-out behaviour of these sandwich and in-
of the inserts. In the co-curing process no additional adhesive film sert configurations was to be investigated. However, no standard-
but only the resin of the skin prepregs was used for the skin-core ised test methods for this purpose exist. Some recommendations
bonding. The specimen plates were cured in an autoclave using a can be found in the ESA Insert Design Handbook [29] and some air-
flat mold and a vacuum bag (2 bar, 125 °C, 90 min curing). Two dif- craft manufacturers have developed their own test methods [38].
ferent self-locking steel inserts, specified by the US National Aero- Based on these data, two test rigs were designed and fabricated
space Standards NAS 1833-C3-370 [36] (diameter 14 mm, height within this study, shown in Fig. 4. The pull-out test rig features a
9.4 mm) and NAS 1835-C3-430 [37] (diameter 17.4 mm, height circular hole of 100 mm diameter to reduce edge effects, under
11 mm), were used. The latter one is a so-called floating insert with which the specimen is placed. The insert is pulled out vertically.
a moveable nut inside the insert housing for compensating assem- For the fixation of the specimens in the shear-out test rig, the hon-
bly inaccuracies. After curing of the sandwich panels a hole was eycomb core at the specimen sides had to be replaced by a solid
material with clearance holes. Once the specimen is bolted to the
test rig, the insert is loaded in the sandwich plane by pulling a
shear plate. In this case, the weft direction of the skin’s woven fab-
Table 1
ric was oriented parallel to the loading direction. For both experi-
Mechanical properties of sandwich specimens in this study.
ments, bolts of the type 0.190-32 UNJF-3A were used, a new one
Skin properties Glass–fabric reinforced phenolic for each test. The test rigs were mounted on an Instron 5566
Eskin 20 GPa mskin 0.059
10 kN universal testing machine. To investigate the influence of
Gskin 1.8 GPa rskin 150 MPa
tskin 0.24 mm the loading rate on the failure behaviour, the two cross-head
speeds of 1 mm/min (static) and 500 mm/min (dynamic) were
Core properties NomexÒ honeycomb
EL 0.58 MPa GLT 41.9 MPa tested. This led to failure after 200 s and 0.4 s, respectively.
EW 0.33 MPa GWT 25.5 MPa
ET 84.8 MPa GLW 0.31 MPa 2.3. Experimental results and failure process
sWT 0.9 MPa sLT 1.21 MPa
hcore (insert test) 14.6 mm hcore (corner test) 9.5 mm
The pull-out test results of six specimens for each insert type
Potting properties Epoxy and load case showed a high level of reproducibility, representative
Epotting 1.05 MPa mpotting 0.3
curves are shown in Fig. 5. The curves of both insert types are very
rpotting 68.7 MPa rpotting 19 mm
similar. In general, two distinct points can be identified, character-

Fig. 3. Cross-section of insert types NAS 1833 (a) and NAS 1835 (b).
578 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

Fig. 4. Test rigs for pull-out (a) and shear-out (b) testing.

Fig. 5. Force–displacement results of pull-out tests and post-failure cross-sections.

ising the failure process under vertical loads. In the beginning, the 1835. This may be ascribed to the lower surface of the insert being
load increases due to elastic deformations. The first damage at bonded to the potting material, which is not the case for insert type
point A, leading to a slight drop of the load level, is attributed to NAS 1833 (see Fig. 3), therefore, failure occurs earlier. After point B
a transverse shear failure of the honeycomb core adjacent to the the post-damage behaviour is mainly characterised by a peeling/
potting mass. This can be confirmed by a view on the lower spec- debonding of the upper skin and friction effects while pulling out
imen surface at that time. The whole potted area exhibits a vertical the insert. In case of the dynamic loading, the curves show the
displacement with respect to the rest of the panel, which is only same characteristics as in the static tests, and therefore the same
possible under a large shear deformation of the core. Then the load failure process, but a higher load level. Also the shear failure occurs
increases again up to point B, where it drops because of a tensile at a higher load. This may result from the rate dependency of the
rupture of the potted honeycomb cells combined with a shear fail- NomexÒ honeycomb core structure, which was described by
ure at the cell wall interfaces, which can be seen in Fig. 5. The aver- Feichtinger [39] and Heimbs et al. [40,41]. Due to micro-inertial ef-
age static peak loads of both insert types at that point show almost fects, the shear strength under transient loading is significantly
equal values with 2220 N (standard deviation: 106.1 N) and higher than in the static load case.
2194 N (standard deviation: 105.7 N). However, the drop of the The results of the shear-out tests are shown in Fig. 6. As in the
load level occurs at a larger displacement for insert type NAS pull-out tests, both insert types show similar load curves. The aver-
S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 579

Fig. 6. Force–displacement results of shear-out tests and post-failure images.

age maximum force in the static tests is higher for type NAS 1835 2pr potting hcore sWT
F pull-out ¼ ¼ 1630 N ð1Þ
with 4360 N (standard deviation: 553.2 N), compared to 3921 N of CK max
type NAS 1833 (standard deviation: 149.5 N). This can once more b hcore
be explained by the connection of the lower insert surface to the with c ¼ ; b¼ ð2Þ
bþ1 tskin
potting material. Again, two characteristic points in the load–dis-  rffiffiffiffiffiffiffiffiffiffiffiffiffi 
rpotting r s max vðrpotting rs max Þ
placement diagrams are noticeable. After an elastic deformation, K max ¼ 1 e ð3Þ
r s max r potting
a tensile cohesive failure of the epoxy bond occurs at point A fol-
rpotting
lowed by a first drop of the load level. The final drop to almost zero rs max ¼ n ð4Þ
at point B happens because of a bearing failure with the upper ekðvrpotting Þ
1sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
composite skin and the potting compound below being sheared 1 Gskin   b 2
off (Fig. 6). The whole block of potting material in front of the insert v¼ 12 1  mskin 2
þ1þ ð5Þ
t skin Eskin 2 3b
in load direction is separated from the surrounding material and
the lower skin, which remains intact. One important aspect is the
k ¼ 0:931714
dependency of the failure process on the number of filled honey-
comb cells in front of the insert in load direction, or in other words
n ¼ 0:262866
the insert position, which explains the relatively high values of the The calculated pull-out force at shear failure is 1630 N, which is 23%
maximum forces’ standard deviations. In case of only a small quan- lower than the experimental results and therefore a rather conser-
tity of potting material, the bearing failure occurs earlier – some- vative result.
times even earlier than the cohesive failure – and therefore at a The calculation of the maximum shear-out force according to
lower load level. A number of different configurations with more the Insert Design Handbook [29], based on the radius of potting
or less potting material were investigated in this framework. For mass rpotting, the skin’s thickness tskin and compressive strength
insert type NAS 1835 the failure loads varied from 3649 N (just rskin as well as the core’s shear strength sWT, leads to 3967 N.
one filled cell ahead of insert) to 4920 N (five filled cells ahead of
insert). The comparison of the static and dynamic tests again F shear-out ¼ 8r 2potting sWT þ 2t skin r potting rskin ¼ 3967 N ð6Þ
shows a rate dependency with higher load levels for the transient The correlation to the experimental values of 3921 N and 4360 N for
loading. both insert types is much better than in case of the pull-out
strength with a maximum deviation of 10%.
2.4. Analytical investigation The analytical calculation of the shear-out force can therefore
be used for an estimation of the insert strength, while the calcula-
An analytical calculation of the insert pull-out force was con- tion of the pull-out force is very conservative. But possible varia-
ducted with the equations provided in the Insert Design Handbook tions and uncertainties of the parameters used in these equations
[29]. This calculation assumes a core shear failure as the limiting have to be kept in mind.
load and is based on the radius rpotting of the potting mass, the
thicknesses tskin and hcore of skins and core as well as their elastic 2.5. Numerical simulation
moduli and the core shear strength sTW. Confirming the experi-
mental findings, the insert type, diameter or height have no influ- Meso-scale models of the insert and honeycomb sandwich
ence on the pull-out strength, which is only determined by the structure require a large amount of modelling work and long calcu-
sandwich configuration and potting size. For the following calcula- lation times and are not suitable for an implementation into a lar-
tion the values given in Table 1 were used. ger structure’s model. However, they may be used to analyse stress
580 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

distributions and understand cell wall deformations and failure model with an edge length of 50 mm had a total of 62000 ele-
mechanisms of the insert under load, since this insight is not pos- ments. The nodal boundary conditions with a circular support were
sible during the experiment. Such a meso-scale model with quar- applied corresponding to the test rig. The bolt was pulled using a
ter-symmetry was developed for dynamic pull-out simulations time-dependent linear displacement function. This model could
using the explicit finite element code LS-DYNA (Fig. 7). Both the be used to prove the theory that a core shear failure under trans-
skins and the honeycomb cell walls were modelled with 4-node verse pull-out loads occurs as the first failure mode, see Fig. 7.
shell elements and the orthotropic composite material model In addition to this meso-scale model, an alternative modelling
MAT54. For the rest of the model, i.e. the potting mass, epoxy resin, approach on a macro-scale was developed, which is much more
steel insert and steel bolt, 6-node wedge or 8-node brick elements feasible to be implemented into a larger sandwich structure model
with the isotropic material model MAT24 were used. This quarter with a higher number of inserts. It should be as simple as possible,

Fig. 7. Models of insert pull-out in honeycomb sandwich: meso-scale (a) and macro-scale (b).
S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 581

Fig. 8. Comparison of experiment and simulation for pull-out and shear-out loading.

while still being able to represent the failure behaviour correctly. The simulation results of the simplified macro-model are shown
For this model, the honeycomb core was homogenised with 8-node in Fig. 8. For the pull-out loading, the curves match with a good
continuum elements, and the orthotropic honeycomb material correlation. Especially the core shear failure prior to total rupture
model MAT126 in LS-DYNA was used. The library of LS-DYNA of- is covered correctly thanks to the honeycomb material model.
fers different options for joint modelling. In this case, the spotweld Also in case of the shear-out loading the elastic behaviour and
option was used for the insert modelling. Such spotweld elements failure in the simulation correlate to the experiment. However,
are solid elements, which connect two surfaces and can be posi- the residual strength after first damage, which was observed in
tioned mesh-independently using a spotweld-contact formulation. the test results, can not be covered by the numerical model. The
Different failure criteria can be adopted for these spotweld ele- reason for this is the spotweld model being no physical represen-
ments, making a pull-out or shear-out failure modelling possible. tation of the real failure process with a bonding failure. It simply
In this study, a quadratic failure criterion with respect to the nor- uses the defined maximum shear force and deletes the element
mal force FN and shear force FS in the element was chosen. The when reaching it. No post-damage behaviour can be covered.
maximum values Fpull-out and Fshear-out could directly be taken from In both curves, the drop of the load level after reaching the de-
the experimental results, since only one spotweld element was fined maximum forces happens to some extent earlier than in the
used. experiment, which corresponds to a slightly lower energy absorp-
 2  2 tion at insert failure. To change this, either the maximum force lev-
FN Fs els need to be increased or the post-damage shear behaviour of the
þ 1 ð7Þ
F pull-out F shear-out honeycomb material model needs to be adjusted, since these two
factors influence the load curve.
The final model as a quarter section is shown in Fig. 7b. With this Nevertheless, this simple insert modelling approach with spot-
modelling approach, only 100 elements are necessary for the same weld elements is feasible of representing the normal and shear
specimen size as for the meso-scale model before. Besides pull-out failure load levels sufficiently. Even strain rate effects can be
simulations, the same model was also used for shear-out simula- covered, corresponding to the experimental findings. In this case,
tions by simply changing the boundary conditions. a different failure criterion has to be chosen in the LS-DYNA

Fig. 9. Overview of manufacturing methods of the three different corner joints investigated in this study.
582 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

Fig. 10. Test methods for bending (a) and shear testing (b) of corner joints.

Fig. 11. Force–displacement results of bending tests and post-failure images.


S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 583

spotweld definition, requiring the input of the maximum forces as layer (glass fibre cloth with Bakelite EPR L 43 epoxy resin) was
a function of the effective strain rate. laminated onto the surface to fixate the corner joint (Fig. 9c). The
advantage of this method is that the outer skin remains intact
3. Failure behaviour of corner joints and that even very complex constructions can be folded very effi-
ciently from one pre-cured sandwich plate.
3.1. Materials and specimen manufacturing All specimens had a side length of 140 mm, whereas at both
edges the honeycomb core was replaced in a length of 25 mm by
The experimental investigation of the corner joints was per- a solid block for load introduction purposes. The specimen width
formed on 10 mm thick sandwich specimens with the same No- was 250 mm. The ribbon direction of the core cells was oriented
mexÒ honeycomb core and the same GFRP skin material as in the perpendicular to the corner line.
insert study. A total of three different corner joints were analysed
and compared. 3.2. Experimental procedures
The first one is the simplest variant, a bonded butt joint accord-
ing to type a in Fig. 2. In one of the two pre-cured sandwich panels Standard test methods for the evaluation of the strength of
the upper skin and core were removed. The second panel was sandwich corner joints do not exist. Besides some investigations
bonded at this position with an epoxy-based 3M Scotch-Weld on the failure behaviour of sandwich T-joints [42–46], the study
9323 B/A adhesive resulting in two contact surfaces (Fig. 9a). in [33] seems to be the only reference, where such tests on sand-
The second variant is based on the ‘mortise and tenon’ method wich L-joints are documented. However, in those tests the corner
according to Fig. 2c. At the edges of both pre-cured panels 60 mm joints are exposed to both bending and shear loads, making a sys-
wide pockets were milled out while keeping the outer skin intact, tematic analysis of stress states and failure modes difficult. There-
resulting in tenons of 50 mm width (Fig. 9b). Both panels were fore, two test procedures were developed for separately
joined and the same adhesive as before was used for the bonding. investigating the specimens under bending and shear loads
The third corner joint features the ‘cut and fold’ technique as a (Fig. 10). In the bending test the sandwich edges were supported
very efficient and industrially relevant design. In contrast to the by linear bearings while the load was applied directly onto the cor-
other two methods, just one pre-cured sandwich plate is necessary. ner, bending the two sides apart. A silicone strap was used under
In the middle of this plate a 16 mm wide and 3.5 mm deep groove the loading plate to assure a load introduction over the whole spec-
was milled out from the upper surface. The open honeycomb cells imen length. In the shear test one specimen side was clamped onto
were filled with a Mankiewicz Alexit FST compound. Then the plate a rigid surface, also using a silicone strap, while the other one was
was folded about 90°. Afterwards, a 60 mm wide reinforcement pulled perpendicularly. This results in a shear loading of the corner

Fig. 12. Force–displacement results of shear tests and post-failure images.


584 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

joint. For each experiment at least six specimens of all three corner to the other corner joints. The damage tolerance of this folded cor-
joint types were tested. The loading rate in these tests was kept ner is superior, because as long as the delamination has not
constant at 50 mm/min. reached a critical extent, loads can still be transmitted through
the intact outer sandwich skin.
3.3. Experimental results and failure process The force–displacement curves of the corner joint shear tests
are almost congruent for the elastic regime until first failure occurs
Force–displacement curves of representative specimens in the (Fig. 12). The reason for this behaviour is a simple transverse shear
bending test are illustrated in Fig. 11, showing the pre- and post- loading of the honeycomb core in all three cases, followed by a core
damage behaviour. The butt joint corners are characterised by shear failure, independent of the corner joint type. The average
the lowest failure load. After a nearly linear elastic beginning, the maximum force Fmax for the butt joint specimens has a value of
bonding of the two panels fails abruptly on the inside of the joint. 2839 N (standard deviation: 90 N). With respect to the sheared
The post-failure residual strength is relatively low. The outer skin surface A, this results in a shear strength of 1.2 MPa:
is bent at a low load level under increasing displacements.
F max 2839 N
The ‘mortise and tenon’ specimens exhibit a similar bending sLT ¼ ¼ ¼ 1:2 MPa: ð8Þ
A 250 mm  9:5 mm
stiffness with a slightly higher failure load. At the peak load the
bonding of the tenons on the inside of the corner joint fails, while This value corresponds exactly to the data sheet value of the
the bonding on the outside remains intact. The residual strength is NomexÒ honeycomb core’s shear strength of 1.21 MPa. For the
higher than for the butt joint specimens and is characterised by the ‘cut and fold’ specimens the peak load is slightly higher with
bending of the outer sandwich skins and the pull-out of the tenons. 3310 N (standard deviation: 171 N). This effect may be attributed
The ‘cut and fold’ specimens show the highest failure loads and to the filling compound, which has also partly been filled into the
also the highest bending stiffness, which is the result of the solid shear-loaded honeycomb cells, leading to an increase of shear
filling compound in the corner. The failure mode is a rupture of strength. After this core shear failure the real loading and character-
the filled and neighbouring unfilled honeycomb cells on one side isation of the corner joint takes place. For the butt joint specimens a
of the corner joint, which cannot be avoided by the reinforcement debonding of the two sandwich panels occurs under a declining
layer. Afterwards, a bending of the outer skin occurs at the failed load curve. The mortised corner first shows a slight increase of
side, while on the opposite intact side the reinforcement layer is the load level due to the larger number of bonded surfaces, but once
continuously delaminated from the inner sandwich skin. These the debonding is initiated, the load level also declines. The folded
mechanisms lead to the highest post-failure load level compared corners show a different behaviour. Because of the reinforcement

Fig. 13. Comparison of experiment and simulation for corner bending and shear test of butt joint specimens.
S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 585

layer being clamped in the specimen fixation and being loaded in ple as possible to be implemented into large-scale models of sand-
tension, no complete failure of the corner joint occurs. A crack be- wich constructions with a number of such joints. The same macro-
tween the upper connection of filled and unfilled honeycomb cells scale modelling approach with shell elements for the composite
as well as a delamination of the reinforcement layer develop. This skins (MAT54) and brick elements for the homogenised honey-
happens at a comparably high and increasing load level. After large comb core (MAT126) was adopted here. The material models and
deformations a crack occurs in the middle of the reinforcement mechanical properties are identical to the insert model.
layer, leading to a declining load level. The experiments have shown that the primary driver of the cor-
ner joint failure is debonding or the disconnection of bonded sur-
3.4. Numerical simulation faces. Therefore, the modelling of the corner joint failure is based
on contact formulations with a failure option. In LS-DYNA such
Just as for the inserts, a modelling method in LS-DYNA was contact definitions are called tiebreak contacts, since first they
developed, which on the one hand is able to represent the corner tie two surfaces together, and once the interface stresses meet a
joint’s failure behaviour correctly, and on the other hand is as sim- defined failure criterion, the contact breaks open and the surfaces

Fig. 14. Comparison of failure process in experiment and simulation for corner shear test of butt joint specimens.
586 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

are separated. In this case maximum normal stresses NFLS and In the model of the ‘cut and fold’ specimens an isotropic material
maximum shear stresses SFLS in the interface are combined in a model was used for the filled corner and the curvature of the edge
quadratic failure criterion was approximated in the mesh. Since in the experiment a discon-
nection of the filled and unfilled cells could be observed, tiebreak
 2  2
jrj jsj contact formulations between the isotropic material and the regular
þ  1: ð9Þ
NFLS SFLS honeycomb elements were generated in the finite element model.
Further contact definitions with different failure parameters were
Those maximum stress values for the bonded contact interfaces introduced between the sandwich skin and the reinforcement layer,
were unknown and had to be adjusted within parameter studies which was modelled as a separate layer of shell elements. Despite its
in correlation to the experimental results. simplicity, this simulation model allows for a good representation of
In case of the butt joint corner only two contact formulations the failure process. However, the difference in the load curves is
had to be defined in the model. The interface failure stresses could slightly higher (Fig. 15). This may be attributed to the same factors
be adjusted so that a satisfactory representation of both the failure as described before, i.e. possible imperfections and the neglect of the
modes and the load curves could be achieved (Fig. 13). However, an silicone strap. In addition, no exact material data were available for
exact compliance of the curves could not be achieved, which may the filling compound, which strongly influences the corner’s proper-
be ascribed to imperfections in the hand-made sandwich speci- ties. The failure process in the corner shear test simulation is shown
mens that are not fully covered by the rather ideal model. Further- in Fig. 16 and it correlates well with the experimental results. After
more, the silicone strap, used in both tests, was not included in the the core shear failure, a crack develops between the isotropic corner
simulation. One requirement for a good compliance is that the fail- elements and the honeycomb core, which is responsible for the first
ure process is reproduced correctly. This could be achieved with a load drop-off. Afterwards, the contact of the reinforcement layer
high degree of accuracy, which can be seen in the illustration of the fails successively as in the experiment. The simulation results and
corner shear test simulation in Fig. 14. At first, core shear failure the contact interface strength values are without doubt mesh size
occurs. Then the contact at the inner side of the joint opens, leading dependent. However, in this study only one mesh size was used.
to a crack. This crack grows until also the outer contact fails. Sim- Although the curves do not match exactly, this modelling approach
ilar results could be achieved for the ‘mortise and tenon’ speci- shows the potential to be a simple and efficient technique to imple-
mens. The difference here lies in the higher number of contact ment a corner joint failure possibility into a larger finite element
surfaces. model of a sandwich construction.

Fig. 15. Comparison of experiment and simulation for corner bending and shear test of ‘cut and fold’ specimens.
S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588 587

Fig. 16. Comparison of failure process in experiment and simulation for corner shear test of ‘cut and fold’ specimens.

4. Conclusions tive connection surfaces. The ‘cut and fold’ corner showed the
highest failure loads and a superior post-damage behaviour.
The failure behaviour of different types of potted inserts and For both inserts and corner joints, simple and efficient model-
corner joints in NomexÒ honeycomb sandwich structures was ling methods in LS-DYNA based on spotweld elements or tiebreak
investigated experimentally. Insert pull-out tests showed a core contact formulations have been developed, which are able to cover
shear failure to occur first, before the potted cells fail under tensile the failure behaviour with an acceptable degree of accuracy and
rupture. Under shear-out loading, the potted cells together with can be implemented into large-scale models of sandwich
the upper skin fail in shear with the insert position within the pot- structures.
ted area having a significant influence on the results. However, the
insert type had no influence, only the potting diameter. The Acknowledgement
failure stresses in both experiments were affected by the loading
rate. This work was performed within the project INTECK, with par-
The failure behaviour of the corner joints under bending or tial funding by the Free and Hanseatic City of Hamburg, Germany.
shear loads was primarily driven by the debonding of the respec- Sincere thanks are given to COMTAS Composite, Hamburg for
588 S. Heimbs, M. Pein / Composite Structures 89 (2009) 575–588

manufacturing the specimens and Dipl.-Ing. Manfred Heimbs for [21] Bull PH, Thomsen OT. Development of a design tool for initial analysis of
inserts in sandwich structures. In: Eighth International Conference on
manufacturing the test rigs.
Sandwich Structures, Porto; 6–8 May 2008. p. 1036–47.
[22] Noirot F, Ferrero JF, Barrau JJ, Castanie B, Sudre M. Analyse d’inserts pour les
References structures sandwich composites. Mec Ind 2000;1(3):241–9.
[23] Roth MA. Strukturelles Nähen: Ein Verfahren zur Armierung von
[1] Bunyawanichakul P, Castanie B, Barrau JJ. Experimental and numerical analysis Krafteinleitungen für Sandwich-Strukturen aus Faser-Kunststoff-Verbund.
of inserts in sandwich structures. Appl Compos Mater 2005;12(3–4):177–91. PhD Thesis, Technische Universität Kaiserslautern, Germany; 2006.
[2] Bianchi G, Aglietti GS, Richardson G. Optimization of bolted joints connecting [24] Bozhevolnaya E, Lyckegaard A, Thomsen OT, Skvortsov V. Local effects in the
honeycomb panels. In: First CEAS European air and space conference, Berlin; vincinity of inserts in sandwich panels. Compos Part B 2004;35(6–8):619–27.
10–13 September 2007. p. 2179–86. [25] Tsouvalis NG, Kollarini MJ. Parametric study of stress concentrations caused by
[3] Cushman JP, Murphy RJ. Geometric considerations in the design of honeycomb inserts in sandwich panels. In: 12th European conference on composite
sandwich fasteners. J Spacecrafts Rockets 1966;3(11):1686–8. materials, Biarritz; 29 August – 1 September 2006.
[4] D’Amato E. Caratterizzazione Strutturale di Inserti per Pannelli in Composito [26] Burchardt C. Fatigue of sandwich structures with inserts. Compos Struct
Parte I: Dati Sperimentali. XXXV Convegno Nazionale AIAS, Ancona; 13–16 1998;40(3–4):201–11.
September, 2006. [27] Bitzer T. Honeycomb technology. London: Chapman and Hall; 1997.
[5] Raghu N, Battley M, Southward T. Strength variability of inserts in sandwich [28] Zenkert D. The handbook of sandwich construction. West Midlands: EMAS;
panels. In: Eighth international conference on sandwich structures, Porto; 6–8 1997.
May 2008. p. 558–69. [29] Insert Design Handbook, ESA PSS-03-1202, European Space Agency, Paris;
[6] Song KI, Choi JY, Kweon JH, Choi JH, Kim KS. An experimental study of the 1987.
insert joint strength of composite sandwich structures. Compos Struct [30] Funke H. Sandwich-Leichtbauteile in Wabenbauweise. Internet information
2008;86(1–3):107–13. brochure of R&G GmbH, Waldenbuch, Germany, http://download.r-g.de/
[7] Demelio G, Genovese K, Pappalettere C. An experimental investigation of static waben.pdf.
and fatigue behaviour of sandwich composite panels joined by fasteners. [31] Sandwich panel fabrication technology, Brochure LTU 018, Hexcel Composites,
Compos Part B 2001;32(4):299–308. Duxford; 1997.
[8] Kim BJ, Lee DG. Characteristics of joining inserts for composite sandwich [32] Noakes K. Successful composite techniques. London: Ospery Publishing; 1992.
panels. Compos Struct 2008;86(1–3):55–60. [33] Joulia N, Grove SM. Strength of sandwich panel ‘cut and fold’ joints. Adv
[9] Klein B, Gänsicke T, Schönfelder T. Lebensdauerbestimmung von Inserts in Compos Lett 2003;12(6):243–6.
Aluminium–Sandwichplatten. Automobiltechnische Zeitschrift ATZ [34] Carruthers JJ. Some aspects of the energy absorption of composite materials.
1995;97(2):102–7. PhD Thesis, The University of Sheffield, UK; 1997.
[10] Roth MA, Himmel N, Johannes M, Cuervo Pinera V. Entwicklung von in [35] Petrone F, Garesci F, Lacagnina M, Sinatra R. Dynamical joints influence of
Dickenrichtung armierten Krafteinleitungen (IDAK) für Sandwich-Strukturen sandwich structures. In: Third European nonlinear oscillations conference,
aus Faser-Kunststoff-Verbund. Konstruktion 2005;57(5):81–5. Copenhagen; 8–12 August 1999.
[11] Skelton TJ, Adams DO. Localized stitching to increase insert pull-out load in [36] NAS 1833: Insert, molded in, threaded, self-locking, nonself-locking,
sandwich composites. In: 49th SAMPE international symposium, Long Beach, sandwich panel. National Aerospace Standard, Aerospace Industries
CA; 16–20 May 2004. Association; 1999.
[12] McGarva LD, Aström BT. Insert integration in thermoplastic-based foam core [37] NAS 1835: Insert, molded in, blind threaded, self-locking, floating, sandwich
sandwich components. J Sandwich Struct Mater 1999;1(3):214–29. panel. National Aerospace Standard, Aerospace Industries Association; 1999.
[13] Kepler J. Partial insert in sandwich panels with thin face-sheets. In: [38] AITM 1-0056, Airbus test method, vertical insert shear, Airbus S.A.S., France;
Composites and Sandwich Structures: Nordic Consortium for Composites 2004.
and Sandwich Construction, Stockholm; 22–23 October 1997. p. 205–16. [39] Feichtinger KA. Test methods and performance of structural core materials –
[14] Youngquist WG, Kuenzi EW. Stresses induced in a sandwich panel by load IIa. strain rate rependence of shear properties. In: Fifth autumn INERN
applied at an insert. Report No. 1845, Forest Products Laboratory, United States conference – composite materials with sandwich structure, Lorient; 16–17
Department of Agriculture, Wisconsin, USA; 1955. October 1990.
[15] Roth A, Himmel N. Konstruktion, Auslegung und Prüfung von Krafteinleitungen [40] Heimbs S, Schmeer S, Middendorf P, Maier M. Strain rate effects in phenolic
in Faser-Kunststoff-Verbund-Sandwich-Strukturen. In: Fifth internationale composites and phenolic-impregnated honeycomb structures. Compos Sci
AVK–TV Tagung, Baden-Baden; 17–18 September 2002. p. C2.1–C2.11. Technol 2007;67(13):2827–37.
[16] Mares J, Wald F, Sokol Z. Modelling of joints of sandwich panels. In: [41] Heimbs S. Sandwichstrukturen mit Wabenkern: Experimentelle und
Banitopoulos CC, Wald F, editors. The paramount role of joints into the numerische Analyse des Schädigungsverhaltens unter statischer und
reliable response of structures. NATO Science Series. Series II, vol. kurzzeitdynamischer Belastung. PhD Thesis, Technische Universität
4. Dortrecht: Kluwer Academic Publishers; 2000. p. 387–95. Kaiserslautern, Germany; 2008.
[17] Thomsen OT. Sandwich plates with ‘through-the-thickness’ and ‘fully potted’ [42] Dodkins AR, Shenoi RA, Hawkins GL. Design of joints and attachments in FRP
inserts: evaluation of differences in structural performance. Compos Struct ships’ structures. Mar Struct 1994;7(2–5):365–98.
1998;40(2):159–74. [43] Rosander M. T-joints in sandwich structures. Licentiate thesis, Royal Institute
[18] Thomsen OT, Rits W. Analysis and design of sandwich plates with inserts – a of Technology, Stockholm, Sweden; 1997.
higher-order sandwich plate theory approach. Compos Part B [44] Toftegaard H, Lystrup A. Design and test of lightweight sandwich T-joint for
1998;29(6):795–807. naval ships. Compos Part A 2005;36(8):1055–65.
[19] D’Amato E. Caratterizzazione Strutturale di Inserti per Pannelli in Composito. [45] Theotokoglou EE, Moan T. Experimental and numerical study of composite T-
Parte II: Modelli Numerici. XXXV Convegno Nazionale AIAS, Ancona; 13–16 joints. J Compos Mater 1996;30(2):190–209.
September 2006. [46] Yeh HY, Nguyen SV, Yeh HL. Failure analyses of polymer matrix composite
[20] Thompson RW, Matthews FL, O’Rourke BP. Load attachment for honeycomb (PMC) honeycomb sandwich joint panels. J Reinf Plast Comp
panels in racing cars. Mater Design 1995;16(3):131–50. 2004;23(9):923–39.

You might also like