Proceedings of the International Association for Shell and Spatial Structures (IASS)
Symposium 2015, Amsterdam
17 - 20 August 2015, Amsterdam, The Netherlands
Structural assessment of a composite glass panel sandwich
based on stress analysis simulation and structural testing
Niloufar EMAMI *, Harry GILES a
* University of Michigan
2000 Bonisteel Blvd. Ann Arbor MI 48109
*, a University of Michigan
“Envelope” is a generic term that can be used to describe the total enclosure that separates the interior
from the exterior of a building. The envelope defines the architectural appearance of the building,
provides views to the inside and outside, resists wind loads, bears its self-weight, modulates heat
transfer between internal and external temperatures and transmits light to the interiors. In addressing
these functional parameters, various facade system solutions have drastically altered the appearance of
the building envelope.
This study is focused on an innovative passive integrated shading panel system for building
envelopes, where the internal shading is formed out of core elements that are structurally layered
between the two glass panes. The composite panel consists of two outer skins and an inner core,
which are assembled by bonding the two skins to the intermediate core. The skins are relatively
strong, stiff materials while the core is a less stiff material. This integrated composite panel system
modulates daylighting through the shading panels in addition to providing increased structural
efficiency, compared to a simple double glazed system.
This research studies the behavior of samples of composite glass beams, which are considered
representative of an entire composite glass panel. Two glass types, annealed and tempered, form the
outer layers while the intermediate core materials are assembled out of thin acrylic sheets. The three
layers are chemically bonded and cured prior to testing. The structural performance of the composite
glass beams were evaluated through structural analysis simulation in the Ansys, FEM program and
also load tested in flexure under static loading. The aim of the study was to both predict the structural
behavior of the composite glass beams as well as validate the analysis results through physical testing,
till structural failure occurs. The paper describes the evaluation process and discusses the results of
this study, with conclusions that inform the evaluation of the structural behavior of glass composite
panels applied to large scale building envelopes.
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Keywords: building envelope, glass composite panel, structural performance, structural simulation,
When it comes to all-glass building envelopes, it might be assumed that a building will receive the
maximum amount of daylighting; however, this comes with a penalty since it will also lose the
maximum amount of energy through the glass, with the potential for high glare problems. This has led
to the need to design various daylighting control systems for transparent skins, including blind
systems, light shelves, etc. Most of these systems are not integrated with the envelope and operate by
adding an additional external layer to the facade. This paper presents the results of a study that
investigated the structural behavior of a composite panel system which integrates shading devices
within the cavity of a double-layered glass system. (Giles and Kim ) This fixed inner structural
core, which connects to the inner and outer transparent window material, can meet multiple
specifications including enhanced thermal performance of the facade as well as playing a structural
role for the whole system. (Figure 1)
Figure 1: The integrated system with horizontal and vertical shading devices (Giles and Kim )
Composite panels consist of two outer skins and an inner core, which are assembled by bonding the
two skins to the intermediate core. The skin needs to resist weather conditions, while the inner core is
preserved from direct contact of those effects. Composite materials are widely used in engineering
fields such as aerospace and automobile, as well as building industries such as Structural Insulated
Panels (SIP’s), because of their relative structural efficiency and potential for thermal insulation.
Past studies explored a split in the façade into an upper daylight window and a lower vision window,
thus providing comfort and higher lighting energy savings. (Reinhart and Selkowitz ) On a separate
note, commercial outdoor shading enclosures are mostly driven by uniformity (“one design fits all”)
and often fail to cast shadow over the desired target-shade area. (Adriaenssens et al. ) It is essential
that customized shading devices can be designed and manufactured to provide the best answer for
every specific building.
The shading panel system under discussion in this study, provides the opportunity to not only design
an integrated building component for greater performance efficiency, but also as a flexible and
adaptable building product which addresses different building programs and design variations. This
can be attained by varying the placement and density of the grid towards designing a functionally
graded system. (Figure 2) The geometric grading responds to and reflects the internal layout of the
building on a macro scale. (Emami and Giles )
This study looks at the structural performance of a sample composite glass beam (considered as one
strip of the glass panel), through structural simulation and experimental study. It studies how different
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
glass types, namely annealed and tempered glass, fail under loading. In addition, it correlated the
results of simulation analysis and experimental studies. More importantly, the study provides the
researcher with some evidence of the composite panel’s advanced structural behavior under
demanding load conditions and verifies the actual strength of the glass beam in bending, towards the
potential for its application in large scale building applications.
Figure 2: System wide graded geometry concept applied to functional facades (Emami and Giles)
2. Glass composite beam
Historically the use of glass evolved from purely decorative applications to architectural windows that
provide light and transparency to the outdoors. (Emami ) Today, architects are pushing the
envelope of glass skins, towards more demanding applications, where the glass is used structurally in
beams and columns for walls, ceilings and roofs. The new application demands have challenged
engineers to integrate glass in load-bearing elements such as beams, columns and roof structures. The
concept of a glass beam was “in the air” in the 1980 states Nijsse. (Nijsse ) Wurm states that
“depending on the number and arrangements of supports, a glass beam can act as simply supported
span, a continuous beam or a cantilever”. Also because of the high slenderness of glass beam cross
sections, buckling is more likely to occur compared to other types of materials cross sections such as
profiled steel beams. Stronger and stiffer interlayer materials could greatly improve buckling behavior
of glass beams. (Wurm)
Glass is a brittle material which is strong in compression but its failure occurs at relatively lower
tensile stress. Although the theoretical strength of glass is high, its practical strength is much lower.
The reason for this is that glass is a brittle material and more susceptible to surface defects as tensile
stresses build up under load. The tensile stresses cause existing microscopic cracks to open, which
easily extend until they reach a free edge to cause material failure at lower than theoretical strength
values. (Wurm) A qualitative comparison of “stress/strain” graphs for three materials (glass, steel
and wood) are presented in Figure 3 that demonstrates the brittle nature of glass, as well as other
parameters that also affect the tensile strength of the glass.
Proceedings of the International Association for Shell and Spatial Structures (IASS) Symposium 2015, Amsterdam
Figure 3: Materials’ stress/strain graphs (left) Factors influencing the tensile strength (right) (Wurm)
2.2 Structural Evaluation Process
A simulation analysis was carried out for the glass beam in the ANSYS FEM program for static
structural analysis. This simulation provides a theoretical estimate of the behavior of the glass beam
under static loading, assuming linear elastic material properties. A total number of six samples were
physically load tested using three point loading and support. Three of these samples comprised
annealed glass and the three were tempered glass. Conducting the experiment on three identical
samples eliminated random errors and provided a consistent basis on which to compare the respective
results between each of the test samples. Simulation and experimental results were compared to
validate the theoretical predicted behavior of the samples.
The sizes used for the sample compositely fabricated glass beams were 610 mm long x 76 mm wide
(24” by 3”) and consisted of two 3.175 mm (1/8”) glass skins adhered to two 3.175 mm (1/8”)
vertical acrylic webs. The acrylic webs were placed in a double arrangement to avoid any instability
of the beam during testing. The plan and section view of the glass beam is illustrated in figure 4.
Figure 4: The plan and section view of the glass beam
Size of pane
Age of pane
(air moisture content)
3. Numerical study of glass beam
3.1. Boundary conditions and material properties
A solid CAD was imported into Ansys for analysis. The web cross-section areas were selected as a
“z” axis vertical support, and their movement was free in “x” and “y” plane. A vertical load was
applied to the top of the web, at the intersection between the webs and glass surfaces. (Figure 5)
Annealed glass material was assigned to the skin and Acrylic material to the core. The material
properties including Young’s Modules, Poisson ratio, density, and ultimate compressive strength were
extracted from a comprehensive data base on material properties, CES Edu Pack, developed by Granta
Design. The allowable surface and edge stress levels were extracted from ASTM E1300.The material
properties that were used are shown in Figure 6.
Figure 5: Boundary and loading conditions in Ansys
Figure 6: Material Properties used in Ansys analysis
Density (kg/ m^3)
Young’s Modulus (MPa)
Tensile yield strength (MPa)
Allowable surface stress for 3
seconds load (MPa)
Allowable edge stress for seamed
Compressive ultimate strength (MPa)
Load on webs
Web top surface
3.2. Initial simulation results
The glass was simulated under different loading levels, and the deflection (figure 7) and safety factor
were recorded until the safety factor reaches a number below 1. Knowing that the analysis is linear
elastic, the load-deflection relationship came out perfectly linear. However, since Ansys determines
safety factor based on the location of stresses, very high local stresses may not be representative of
what happens in reality, and some local stress may be dependent on how the structure is modeled in
Ansys. Therefore, this method was only a quick technique to guide through the number of simulation
iterations rather than providing the exact safety factor. After the safety factor fell under 1, loading the
beam was no longer simulated and the last deflection under that load was recorded. Based on this, the
tempered glass failed under 350 lbf ≈ 1556 Newton, compared to the annealed glass (220 lbf ≈ 978
Figure 7: Deflection plot of the beam under point load
The crucial point, however, is to discuss the average high stress levels, which are identified on the
stress plot. These values are later compared with experimental results, to see whether Ansys results
show failure stresses with associated load values that compare well with the tested results. This is
studied in the forthcoming sections.
4. Experimental study of glass beam
In this section, the experimental process including structural testing in flexure and the outcomes are
described. The experimental study was intended to compare the results of the theoretical analysis with
the experimental results in order to calibrate the simulation input data for future analysis work.
4.1. Materials and geometry
Two glass types, annealed glass and tempered glass, were used to conduct structural tests. 6 pieces of
3.175 mm (1/8”) thick glass of each type were prepared with dimensions 610 x 76 mm (24”x 3”). The
goal was to make two different glass beam groups, with three of each type using annealed glass and
three of each type using tempered glass (a total of six glass beams). The core was made out of acrylic
3.175 mm (1/8”) thick, cut to a size of 610 x50.8 mm (24”x 2”) and the resulting total depth of the
beam once fully adhered together, was 57 mm (2 ¼”).
4.2. Structural testing
After manufacturing identical samples, they were 3 point load supported and loaded using a Baldwin
load testing machine as shown in Figure 8. The beam was set on wooden supports on each end. The
wooden supports had a circular top shape to avoid any premature crushing from high local bearing
stresses at the support locations, since the overall flexural behavior was of interest in this experiment.
A wooden block was similarly placed on top of the beam that spread the load from the load point to
the two webs on either side of the beam section, again to avoid any premature crushing from high
local bearing stressed at the load location. A dial gauge that measured deflections to an accuracy of
0.0127mm (1/2000 inch) was installed at the bottom of the beam (Figure 8). After loading to failure,
the total deformation of the beams was visually noticeable. It noteworthy that although the beam
assembly was made from brittle materials, the adhesive was sufficiently plastic to allow the beam to
deflect significantly and still hold itself together, despite brittle fractures in both the glass and the
acrylic. This implies a good safety feature of this type of assembly, similar to that observed in the
behavior of laminated glass sheets.
Figure 8 : Load testing arrangement a) overal beam placement b) beam edge at the support c)
completed test showing final failed deflected beam
Beam number 1, 2 and 3, the annealed glass beams, were loaded first. They failed under 710 N, 710 N
and 533 N (160, 160 and 120 pounds) load respectively. In all of the beams, the top glass skin cracked
and failed. Two of the beams cracked in the middle while one (beam 1) cracked at the support. The
associated photos are shown in Figure 9.
Beam number 4, 5 and 6, the tempered glass beams, were loaded next. They failed under 1023 N, 845
N, 934 N (230, 190 and 210 pounds of load) respectively. In two beams, the adhesive failed while the
glass skin was still carrying load. In one of the beams (beam 5), the acrylic web cracked while the
glass skin was still carrying the load. The associated photos are presented in figure 10. It is worth
noting that after loading beam 4 and when its adhesive failed, the beam was reloaded to see if the
glass or web may fail. In the reloading test, the acrylic web broke under 400 N (90 lbs) in the middle.
This experiment has been referred as “4a” in the table and graph. (Figure 10)
Figure 9: Annealed glass composite beam experiment (beam 1 to 3 from left to right)
Figure 10: Tempered glass composite beam experiment (beam 4 to 6 from left to right)
4.2. Structural results
In this section, deflection versus load is been plotted for all six beams. (Figure 11) There is a
consistent trend in the behavior of the glass beams. Beam 4 to 6, the tempered glass beams, failed in
the load range 850 – 1000N, while the annealed glass beams failed under lower loads in the range 500
-750N. Beam 4a is also shown as a special case for reloading as previously described.
Figure 11: Load versus deflection for all six beams in experimental studies
It can be seen in the lower load range, up to 550N, the beams generally displayed linear behavior
except for beams 2 and 4a, which showed significant nonlinear behavior at an early stage of the
loading. Also beams 2 and 4a show a much lower stiffness due to a lower effective slope of the load
deflection plot. In the case of beam 2, no failure of the glass nor the acrylic was observed during these
early stages, therefore it can only be concluded that the adhesive joint interface had started to deform
or fail at an early stage, owing to discrepancies in the manufacturing process (which was in fact
observed during fabrication and isolated as a possible cause for different expected behavior), resulting
in less adhesive material and less adhesive bonding between the web and the flange of the beams
affected, causing the adhesive to delaminate at an early stage, resulting in the change in behavior
compared to the other beams. In the case of, beam 4a it was decided to reload beam 4, which had
previously failed as shown on the graph, to observe how a typical beam might behave after reloading
an already failed beam. It is worth noting that there still remained a large degree of stiffness and load
bearing capacity in the beam, approximately 40% of the original values, which suggests good post
failure safety as reference above in relation to laminated glass sheets. Other than beam 2 and case 4a,
there is a consistent trend among the glass beams’ behavior.
5. Structural verification among simulation and experiment
5.1. Beam stiffness
In the simulation, with every 44.48 Newton (10 lbs) of load, the beam deflected 0.03mm (0.0012”),
whereas in the test, with adding every 44.48 Newton (10 lbs) of load, the beam deflected around
0.15mm (0.006”). This has been demonstrated more clearly in figure 11. Looking at the graph at 500
Newton (110 lbs) load, Ansys results which are indicated with a dashed blue line, show 0.35mm
(0.013 inches) deflection; whereas the experiment results which are indicated with a dashed red line,
show 1.67 mm (0.066 inches) deflection (figure 11). In order to calculate beam stiffness (K), hand
calculations in addition to the simulation and experimental results were carried out. This was done by
using K=P/d, where “K” is the overall beam stiffness, “P” is the point load and “d” is the absolute
value for deflection. With this, beam stiffness was calculated and compared in the linear elastic region
of the test that is at approximately 500 N and below. For hand calculation, if we assume that the whole
beam is made of glass, deflection can be calculated by d= (P*L^3)/ (48 E*I) under a load of 500 N
(110 lbs), which is calculated to be around 0.09 mm (0.0035 inches). Then beam stiffness is calculated
(1). The same procedure can be done for an assumed all-acrylic beam (2). Since E of glass is 25 times
more than E of acrylic, it is expected that deflection of an all acrylic beam is 1/25 of the deflection of
an all-glass beam. Beam stiffness for the composite glass-acrylic beam is expected to be a value
between the two values, and closer to 5555 N/mm than 222, since the glass is in the flange and that
will contribute 80+% of moment of inertia for the entire section (3). Finally, K is calculated for Ansys
(4) and experimental (5) results:
Beam stiffness for an all-glass beam: K all-glass=P/d= 500/0.09 =5555 N/mm (1)
Beam stiffness for an all-acrylic beam: K all acrylic = 1/25 K all glass =5000/25 = 222 (2)
Beam stiffness for glass-acrylic beam: K glass-acrylic composite= 80% K all-glass = 5000*0.8 = 4444 (3)
For Ansys results: K Ansys= P/d = 500/0.35 = 1430 N/mm (4)
For experimental tests: K Experiments= P/d = 500/ 1.67 = 299 N/mm (5)
Looking at the results, there is a huge difference between hand calculations and the other results. This
difference may come from the different behavior as a result of different materials on such a short span
and the application of a point load. The “shear deformation” phenomena that occurs in short beams
with deep cross sections may be the reason for having higher deflections than predicted hand
calculations. The hand formula works for more slender beams with a span/depth ratio 0f 20-30,
whereas this beam has a span/depth ratio of 12. One conclusion from this is that for large scale
applications, the loading and the span/depth ratio relationships are very different and not as severe as
the results this test show. Therefore, more realistic comparisons between test and analysis data is
Comparing numerical and experimental studies results, the glass beams under experiment deflect 5
times more than the expected outcomes of the simulation in the linear elastic region for both cases
(they were more flexible). This can be explained by the fact that the adhesive between the web and the
skin, is a much more flexible material than glass and acrylic, which reduces the stiffness of the beam.
The simulation, did not attempt to model the effects of the lower adhesive stiffness, and given the test
results, this demonstrates that further study is needed to test different E (modulus of Elasticity) for the
adhesive, until a close result is achieved compared to the experimental results.
5.2. Failure loads
Looking at the experimental results, it was measured that the annealed and the tempered glass beams
failed in the region of 530- 710 N (120-160 lbs), and 850-1020 N (190-230 lbs) respectively.
Previously analyzed failure load in Ansys in section 3.2 which was based on safety factor, indicated
that the failure load is 978 N (220 lbf) and 1556 N (350 lbf) for annealed and tempered glass beams
respectively. The predicted load limits based on safety factor are nearly double the actual failure load.
From another point of view and by looking at the equivalent stress levels, these were then compared
with material properties retrieved from ASTM E1300. For example beam 4 presented in figure 12
broke at 1000N (230lbs) and the comparable stress from Ansys under 978 N (220 lbs) as shown in
figure 12 is about 35MPa at the top of the beam and 17MPa at the bottom. The experiment did show
the glass cracking at the top (where the local stress in compression was shown to be the greatest) and
did not fail at the bottom, where the predicted stresses are less (in tension). It is surprising that the
glass failed at a stress level that is lower than the rated yield strength for tempered glass ( 73MPa edge
strength value) compared to annealed glass (16.6MPa edge strength value). Local stress effects like
this can be unpredictable with a material like glass which is brittle since edge defects could play a
significant role in strength reduction and in such a small sample such as this test, could prove to be
more significant that in a larger structure where local effects are less and defects are more distributed.
According to hand calculation, the moment (6) and average bending stress (7) at this load level is:
M = (P*L)/4 = (1000N* 609mm)/4 = 152250 N.mm (6)
Bending stress= (M*c)/I = ((152250 N.mm)*(28.5 mm))/491611 mm^4 = 8.82 MPa (7)
Looking at the Ansys plot in figure 13, the green color indicates average normal stress levels and not
account for high local stresses that occur under point loads. Therefore, the normal stresses in the range
of 5-11 MPa are in line with the hand calculation results.
Figure 12: Ansys equivalent stress plot
Figure 13: Normal stress plot in x-direction in Ansys
The aim of the study was to both predict the structural behavior of the composite glass beams as well
as to validate the analysis results through physical testing, till structural failure occurs. The composite
glass beam was loaded in flexure to observe its bending behavior. In the case of annealed glass beams
with acrylic core, they failed in the middle at maximum moment or flexural stress. The brittle nature
of the glass resulted in cracking of the top glass pane as well as the acrylic web. In the case of the
tempered glass beam, which also carried a higher load before failure, the adhesive failed first which
caused the web and the surface to come apart and then the acrylic web cracked in some cases. This
study demonstrates that adhesive connections in the glass structural elements is a viable structural
solution for composite glass-acrylic panels, Further studies are needed to advance the application of
structural adhesives and suitable modeling assumptions, for large structural applications. The key
conclusions / observation here:
1. One has to use a generous factor of safety in the design of the glass components.
2. The acrylic provided significant stiffness to the structure (although less that an all glass beam).
3. The adhesive provided a measure of safety that kept the beam together even after measured failure
and also showed how the beam could be reloaded and carry reload without breaking catastrophically.
4. The combination of glass / acrylic / plastic adhesive is a good combination for future composite
structural panels at a large scale.
5. The adhesive appears to have a significant effect on reducing the overall beam stiffness and this
will need to be taken account of in future simulation studies.
6. Due to the nonlinear behavior noted in the test results, the materials elastic properties would need to
also be modeled with non-linear elastic properties. This will be the subject for future research in this
The work carried out in the paper was facilitated through the funding from pioneering research carried
out by Professor Harry Giles in the fabrication of composite glass wall systems, previously sponsored
by the Environmental Protection Agency (EPA).
 Adriaenssens, Sigrid et al. “Dialectic Form Finding of Passive and Adaptive Shading Enclosures.” Energies
7.8 (2014): 5201–5220. Web. 29 Aug. 2014.
 Emami, N. “Glass Structures , from Theory to Practice.” ICSA. N.p., 2013. Print.
 Emami, Niloufar, and Harry Giles. “Engineering without the Engine: An Integrated Panelized Passive
Shading System for Transparent Façades.” ARCC (The Architectural Research Centers Consortium).
Honululu, Hawaii: N.p., 2014. Web.
 Giles, Harry, and Kyoung-hee Kim. “Transparent Façade Panel Typologies Based on Hybrid Bio-Composite
and Recyclable Polymer Materials.” ARCC (The Architectural Research Centers Consortium). N.p., 2008.
 Nijsse, Rob. Glass in Structures : Elements, Concepts, Designs /. Basel: Birkhauser, 2003. Web.
 Reinhart, Christoph, and Stephen Selkowitz. “Daylighting—Light, Form, and People.” Energy and Buildings
38.7 (2006): 715–717. Web. 20 Mar. 2014.
 Wurm, Jan. Glass Structures : Design and Construction of Self-Supporting Skins. Basel : Boston : Birkhauser,