Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/287208426

Experimental and numerical evaluation of mechanical behaviour of composite


structural insulated wall panels submitted to edgewise compression

Conference Paper · September 2015


DOI: 10.1201/b20057-111

CITATIONS READS

2 230

2 authors:

Łukasz Smakosz Ireneusz Kreja


Gdansk University of Technology Gdansk University of Technology
17 PUBLICATIONS 105 CITATIONS 103 PUBLICATIONS 753 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Łukasz Smakosz on 11 January 2018.

The user has requested enhancement of the downloaded file.


Experimental and numerical evaluation of mechanical behaviour of composite
structural insulated wall panels under edgewise compression

Ł. Smakosz & I. Kreja


Faculty of Civil and Environmental Engineering, Gdansk University of Technology, Gdańsk, Poland

ABSTRACT: A composite structural insulated sandwich panel (CSIP) is a quite novel approach to the idea of sandwich
structures. A series of natural-scale experimental test is required each time a change in panel’s geometry is planned and
a reliable computational tool is required to precede actual laboratory testing with virtual simulations. An attempt of cre-
ating such a tool has been made with use of a commercial FEM code ABAQUS, in order to predict behaviour of a spe-
cific kind of CSIPs with magnesium-oxide board facings and expanded polystyrene core. Results obtained from simula-
tions taking into account geometrical nonlinearity as well as material nonlinearity of both core and facing materials are
presented and compared with experimental data from natural-scale edgewise compression tests of wall CSIPs.

1 INTRODUCTION

Composite structural insulated sandwich panel is an ad-


vanced approach to the idea of sandwich structures. It is
based on the main principle of combining two materials
with diametrically different properties – a light-weight,
thick core sandwiched between two high-strength, dura- Figure 1. Analysed CSIP’s cross-section: (a) view and (b) scheme.
ble and thin facings – joined together by an adhesive lay-
er. CSIP’s facings are made of composite materials, mak- The paper focuses on predicting a behaviour of the se-
ing a light and easy to handle prefabricated element lected type of CSIPs under an edgewise compression
considerably stronger, immune to biological corrosion with the use of mentioned FEM model.
and more durable to weather conditions. Such improve-
ments make CSIPs very attractive alternative to classical
SIPs with OSB facings and broaden their field of applica- 2 MATERIAL PROPERTIES
tion (FAS 2009, Lstiburek 2008).
High facing-core elastic moduli ratio causes CSIPs to The analysed panel’s cross-section is shown in Figure 1.
be challenging in design. Due to a large number of possi- It consists of two 11 mm thick MgO boards, reinforced
ble failure modes, natural-scale experimental tests are re- with a fiberglass mesh on top and bottom surfaces, a 152
quired with each change in panel’s cross-section. A relia- mm thick EPS core layer with volumetric density of 19.5
ble computational tool is required to precede laboratory kg/m3 and a thin layer of polyurethane adhesive joining
tests with numerical simulations. Abaqus (2010) software the layers together, adding up to a total thickness of 174
was used to create a suitable computational model for the mm.
CSIPs with magnesium-oxide board (MgO board) facings Material properties of EPS are well recognized and
and expanded polystyrene (EPS) core. To observe behav- strongly dependent on its density, while, the available
iour and failure modes of this kind of CSIPs a series of characteristics of the MgO board are very limited. To de-
natural- and small-scale tests was performed on samples termine the mechanical properties of both materials the
with a specific cross-section (Fig. 1). A comparable FEA Authors performed a series of experimental tests on
was performed and a comparison of computational results small-scale samples described in Smakosz & Tejchman
with experiments was presented in (Smakosz & Kreja (2014).
2014, Smakosz & Tejchman 2014). To improve the process of parameter identification,
small-scale bending and compression tests of CSIP sam-
ples composed of core and both facings were performed.
Based on results presented in Figure 6 and in Smakosz &
Tejchman (2014), a set of values for the application in
FEM analysis was established.

3 EDGEWISE COMPRESSION TEST


3.1 Experimental set-up
Wall panel edgewise compression tests under an eccen-
tric load were carried out on two test samples in natural
scale. The dimensions of the tested panels were 2750 mm
× 1000 mm × 174 mm and two values of the compressive
force eccentricity were taken as e=d/6 and e=d/3, where
d=174 mm is the panel’s total thickness. Due to technical
limitations, the tests were performed in a horizontal posi-
tion (Fig. 2a) with both ends supported on hinges (Fig.
2b, c). In order to obtain relatively uniform distribution
of stresses at the edges, the panels were inserted into steel Figure 2. Experimental set-up for wall panel compression tests: (a)
profiles and sealed with concrete in first case and polyu- general view, (b) loading point and left hinge, (c) right hinge.
rethane foam in the other. The steel profile's height was
165 mm and the total distance between the support points
was 3080 mm. The complete experimental set-up is
shown in Figure 2.

3.2 Test results


In both cases the load eccentricity was positioned in the
lower half of the cross-section which resulted in a flexur-
al deformation with an arch tip directed upwards, oppo-
site to the eccentricity’s direction.
The samples failed by a facing cracking in the contact
zone between sample and steel profiles. In both cases, (a)
damage initiated as a horizontal crack appearing suddenly
Figure 3. Failure development: (a) bottom facing horizontal crack –
in the bottom facing (Fig. 3a), signalled with a barely au-
damage initiation, (b) top facing vertical crack – damage propagation,
dible snapping sound and a clearly visible drop of mid- (c) deterioration of damaged facing, (d) a gap between the steel pro-
span deflection. Further loading led to a gap forming be- file and the sample.
tween the damaged sample edge and the steel profile
(Fig. 3d) – this shift in sample’s position in relation to the
loading profile eventually caused a vertical crack in the 4 NUMERICAL SIMULATIONS
top facing (Fig. 3b). Slow, progressive deterioration of
bottom facing's material (Fig. 3c) and increase of the gap 4.1 Assembly’s description
size followed. No buckling behaviour – neither global A commercial FEM software – Abaqus (2010) – was
nor local – was observed throughout the test, and no used to build a numerical model of two tests described in
debonding occurred prior to the damage initiation. previous section. In order to capture the nonlinear re-
The applied load and the horizontal movement of the sponse of tested panels a fine mesh of solid elements in
piston were recorded and presented as force-displacement two-dimensional space was used to discretise full-length
relationships in Figure 7. samples.
Both samples were represented as plane stress sections
Table 1. Average properties obtained from small scale tests on specif- divided into partitions corresponding to core and facing
ic materials: E – elastic modulus, σy – yield stress.
________________________________________ as shown in Figures 4-5. A single FE mesh divided in
____ EPS MgO board
_________________________________ subsections with different material properties was em-
compr. tension compr. flexure ployed (Figs 5-6) because no pre-failure deboning was
MPa MPa MPa MPa
________________________________________ observed in the compression tests so no additional con-
E 6.1 10.4 1964 7043 tact interactions between the sample’s layers were need-
y 0.09 0.11 13.3 10.7
σ________________________________________ ed. Steel mounting profiles were substituted with a rigid
body planar wire sections. A gap of 1 mm was introduced
between the profiles and the sample (Fig. 5) to take into
account a small clearance present in the actual test setup.
These parts were assembled together and a penalty fric-
tion contact interaction with a coefficient of 0.05 was
10
used between the rigid supporting parts and the sample. F [kN]
Pinned boundary conditions were created at a single ref-
FEM,
erence point assigned to each rigid profile. One half of 8
E*1.5
the numerical assembly is presented in Figure 4 and a
magnified view of one of the sample's vertices is shown 6 test data
in Figure 5.
The assembly was loaded in two steps – first the dead 4
weight was taken into, next a horizontal displacement of
one of the supports was forced. Geometric nonlinearity
2
effects were taken into account. FEM,
E*1.0
A fine mesh of 4-node plane stress elements with re- u [mm]
0
duced integration and hourglass control was used. Global 0 0.1 0.2 0.3 0.4 0.5 0.6 (a)
mesh size of 5.5 mm resulted in 28 elements of 1:1 pro-
portions through the core thickness. In the facing area the 35 F [kN]
mesh was refined in one direction resulting in 4 elements
30
of 1:2 proportions through the facing thickness. An en- test data
larged view of a mesh section is presented in Figure 5. 25

20 FEM,
4.2 Material models E*1.5
15
Both, the MgO facings and the EPS core were de-
scribed with an elastic-plastic Drucker-Prager material 10 FEM,
E*1.0
model. Since both layers show different behaviour in
5
compression and tension, two sets of properties were de-
u [mm]
fined. A user-defined field procedure was created to gen- 0
erate an additional field variable (FV) in each integration 0 0.5 1 1.5 2 2.5 (b)

point in such a way that it takes values from –1 (when


250 F [N] FEM,
both principal stresses are negative) to 1 (when both prin- E*1.5
test data
cipal stresses are positive). Elastic properties and a hard-
200
ening behaviour for compression are used for both mate- FEM,
rials when FV = –1, while the tensile properties are E*1.0

assigned in EPS, and flexural ones in MgO board for 150


FV = 1. For all the other values of FV the adequate mate-
rial properties are linearly interpolated between these two 100
data sets.
50

u [mm]
0
0 2 4 6 8 (c)

35 F [kN]

30
test data
Figure 4. Half of the FEM model assembly: boundary conditions and
section assignments. 25

20 FEM,
E*1.5
15

10 FEM,
E*1.0
5
u [mm]
0
0 0.5 1 1.5 2 2.5 (d)
Figure 5. FEM assembly vertex zoom – mesh size comparison, section
property assignment and contact zone. Figure 6. Comparison of FEA with small-scale tests used for parame-
ter identification: (a) MgO board uniaxial compression, (b) CSIP
sample edgewise compression, (c) MgO board bending, (d) CSIP
beam three point bending.
200 F [kN] el’s bearing capacity is lost after damage initiation, the
test data
e=d/6 test data following progression process was not analysed.
FEM, e=d/6 e=d/3
160
E*1.5
FEM, e=d/6
5 CONCLUSIONS
120 FEM, e=d/3
E*1.5
FEM, e=d/3 The comparison illustrated in Figure 7 shows clearly that
80
the numerical model with MgO board's modulus of elas-
ticity in compression entered as a mean value of small-
scale uniaxial tests generally underestimates the perfor-
40 mance of an actual panel. By using strength properties es-
timated in an additional parametric study one can obtain
u [mm] results that are much closer to the experimental ones.
0
0 5 10 15 20 There are two possible explanations to this situation:
1) the performed experimental tests showed that MgO
Figure 7. Comparison of force-displacement curves from laboratory board's behaviour in compression can vary significantly
tests and numerical simulations. from batch to batch, but it is also possible that 2) material
properties obtained in the uniaxial edgewise compression
of small board samples might not correspond well to
conditions in the actual panel facing.
The lower strength obtained in both simulations can
be explained the same way. Since significant variation in
Figure 8. FEM simulation: deformed shape and initial yield zone sig- yield strength was observed during the small-scale labor-
nalling failure initiation. atory tests, an average value of compressive yield stress
and hardening curve were used which placed the numeri-
Elastic properties used to describe tensile and com- cal results on the safe side.
pressive properties of both materials were obtained in a The behaviour of simulated samples gives satisfactory
two-step parameter identification process. First mean results in terms of deformed shapes and failure initiation
values obtained from compression, tension and bending zone – they both match the ones observed in experiments.
tests (Tab. 1) were used directly and afterwards addition-
al simulations of analogical tests performed on small
composite CSIP samples (core and both facings) were REFERENCES
carried out to verify them (Fig. 6). These additional simu-
lations indicated that MgO board's averaged modulus of Abaqus 2010. Analysis User’s Manual (6.10). Dassault Systèmes.
elasticity obtained from uniaxial compression tests was Federation of American Scientists 2009. Report on Expanding the
Scope and Market of SIP Technologies: a history of SIPs and
considerably underestimated. A parametric study showed CSIP manufacturing, construction and market issues.
that by increasing this value by a multiplier equal to 1.5 Lstiburek J. 2008. Builder's Guide to Structural Insulated Panels
one obtains results being more accurate to the ones rec- (SIPs) for all Climates. Building Science Press.
orded in small-scale CSIP sample tests. A value obtained Smakosz Ł. & Kreja I. 2014. Experimental and numerical evaluation
this way is slightly greater than a maximum value ob- of mechanical behaviour of composite structural insulated panels.
In T. Łodygowski, J. Rakowski & P. Litewka (eds), Recent Ad-
tained in compression of separate MgO board samples
vances in Computational Mechanics: 269-276. London: CRC
(Fig. 6a). The following values of the Poisson’s ratios Press/Balkema.
were assumed: 0.11 for EPS and 0.18 for the MgO board. Smakosz Ł. & Tejchman J. 2014. Evaluation of strength, deformabil-
ity and failure mode of composite structural insulated panels. Ma-
terials & Design 54: 1068-1082.
4.3 FEA results
The comparison of numerical and experimental results
with natural-scale experimental data (force vs. horizontal
displacement curves) is presented in Figure 7. Two sets
of numerical results are presented: with MgO board’s
EMgO(FV = –1)=1964 MPa and EMgO(FV = –1)=3000 MPa. Re-
sults computed from FEM models with increased facing’s
Young’s modulus in compression are in better agreement
with experimental data. In both cases the simulations re-
flect panels’ elastic behaviour well but yield and collapse
appear sooner.
One half of the deformed sample is shown in Figure 8
with a yield zone signalising sample's failure. Since pan-

View publication stats

You might also like