Load Transfer of Polycarbonate Blast Resistant Glazing Systems

Joshua Calnan, M.S., E.I.T., University of Kentucky

Braden Lusk, Ph.D., P.E., University of Kentucky
Kyle Perry, Ph.D., P.E., University of Kentucky

Abstract

This paper presents detailed information and data regarding the load transfer characteristics of Blast
Resistant Glazing Systems utilizing two different thicknesses of polycarbonate. A series of triaxial load
cells, along with surrogates of identical stiffness characteristics, were placed around the glazing
perimeter to directly measure the load transferred to the support system from explosions of varying
degrees. The blast loads were developed using an explosive driven shock tube administered by the
University of Kentucky Explosives Research Team (UKERT) in Georgetown, KY. Results presented
include distribution of loading, maximum load transferred to connections, blast pressure-time data,
glazing system deflection-time data, and final performance results of each system. The two glazing
systems respond differently to the applied blast loading. Differences and similarities in the load transfer
characteristics are discussed. The experimental results are then compared to a commonly used blast
resistant glazing system analysis software tool utilizing a Single Degree of Freedom (SDOF) method of
analysis for each system. Preliminary results show that currently used models do not correspond well
with experimental data.


Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 1 of 10
Introduction
Static design methods and computer modeling techniques tend to be overly conservative in predicting
the design strength of glazing systems. While this may be beneficial in ensuring that the glazing system
will withstand a given blast loading, it may present challenges in adequately determining the load
transferred to surrounding structural supports. Current structural design methodology uses a load path
approach, with the required strength of a given element being calculated based on the loads transmitted
to it by connecting elements. If the transmitted loads are underestimated, it may be possible to overload
elements further down the load path. Such is the case with blast resistant glazing systems. If the glazing
material does not yield as predicted, greater loading than anticipated will be transmitted to the framing
material and connecting structural elements. If supporting members are not of adequate strength, severe
structural damage could occur.
The purpose of this research is to investigate the reaction forces transmitted by a blast resistant glazing
system (BRGS) to the surrounding support members when subjected to blast loading. Two phases of
testing were conducted on two different polycarbonate samples with thicknesses of one quarter (0.25)
inch (6.35 mm) and one half (0.50) inch (12.70 mm).
The first phase of testing characterized the reaction forces around the perimeter of the BRGS. Triaxial
load cells were placed at the corners and midpoints of all four edges of a rectangular window to
determine the magnitude and distribution of the reaction forces. The second phase of testing took the
samples to failure. This phase measured the peak reaction forces transmitted to the support members
prior to failure of the BRGS.
Direct measurements of load, pressure, and deflection allows for a direct comparison to commonly used
software programs which estimate the loads and deflections from given pressure versus time waveforms.
Instrumentation and Equipment Setup
To gain an accurate and complete understanding of the events occurring during testing of the two
glazing systems, a full suite of instrumentation was employed, including pressure sensors, deflection
gauges, triaxial load cells, and high speed video.
The pressure time history of each blast event was characterized using piezoelectric dynamic pressure
sensors. Two sensors were placed at the midpoint of the vertical window span, one on either side of the
window. Window deflection measurements were obtained through the use of a laser distance gauge.
This gauge has the ability to provide non-contact measurements of the window surface.
The major focus of this research was to characterize the reaction forces around the perimeter of
polycarbonate blast resistant glazing systems during blast loading. W.C. Wedding (Wedding, 2010)
developed the methodology and much of the equipment required to instrument the experiments. Triaxial
load cells with a Z-axis capable of measuring loads of 10,000 lbf (44.48 kN) at a sampling rate of 10
kHz. were selected for this test series.
To provide adequate support for the glazing system, attachment points were positioned on eight inch
centers, a commonly used commercial standard, around the perimeter of the window. Since it is not
economically feasible to populate all 26 attachment points with triaxial load cells, proxy load sensor
surrogates were used to occupy attachment points where there was not a load cell present. These proxy
sensors were designed and tested by W.C. Wedding (Wedding, 2010) to provide the same stiffness
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 2 of 10
characteristics as the load cells which they were intended to imitate, minimizing the effect the proxy
sensors would have on the overall readings.
Sensor brackets were bolted to the shock tube endplate with wood shims used to adjust the opening to a
final size of 47 in. (1.21 m) wide x 66 in. (1.68 m) tall. The endplate used for this test series is shown
in Figure 1, fully prepped for testing.
Data acquisition was accomplished with the use of a pair of Datatrap II data acquisition devices in a
master and slave configuration, allowing the 12 load cell channels, 2 pressure channels, and 1 deflection
channels to remain on the same time scale. The pressure sensor signals were routed to the master and
served as the trigger for the system.

Figure 1: Buck fully prepared for testing

Experimental Methodology
Load data was collected at the corners and midspan of each side, resulting in data being collected at 14
points. Based on the results found by Wedding (2010), it was shown that instrumenting these locations
instead of the entire perimeter would still allow for accurate load distribution measurements while
limiting the number of tests per BRGS. Limiting the number of tests conducted on a single
polycarbonate sample is important as polycarbonate tends to build up residual stresses, which can have a
negative effect on test results.
The BRGS samples tested were each 66 in. (1.68 m) tall by 44.75 in. (1.21 m) wide and of identical
construction other than polycarbonate thickness. The first sample tested had a nominal polycarbonate
thickness of 0.25 in. (6.35 mm). The second sample had a nominal thickness of 0.50 in. (12.70 mm).
Both were bonded to an extruded aluminum frame using structural glazing tape.
The first phase of testing populated 14 of the 26 attachment points to characterize the load distribution at
the corners and midpoints of each BRGS. In the final phase of testing, the charge size was increased
until the BRGS failed. This was done to determine the peak loading exerted by the system as it failed.
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 3 of 10
The attachment points were assigned alphabetical labels starting in the lower left corner and proceeding
around the perimeter in a clockwise fashion, lettered A through Z. Figure 2 shows the location of these
points as seen from outside the shock tube. For perimeter testing, four different setups were utilized to
characterize the glazing system. The test configurations are referred to in the following manner.
Setup 1 positions A, D, E, and H populated
Setup 2 positions H, I, K, and M populated
Setup 3 positions N, Q, R, and U populated
Setup 4 positions U, V, X and Z populated


Figure 2: Attachment point designations (Wedding, 2010)
Perimeter Testing Results and Anal ysis
Peak pressure results from the two pressure sensors were averaged together and remained very
consistent throughout the test series for both the one quarter inch sample and one half inch sample. The
average pressure and impulse for the one quarter inch test series was 4.657 psi (32.11 kPa) and 20.336
psi-ms (140.21 kPa-ms), respectively. The one half inch test series provided similar results with an
average pressure of 4.814 psi (33.19 kPa) and impulse of 20.655 psi-ms (142.41 kPa-ms).

Deflection results in the one quarter inch test series were consistent with minimal variation. Deflections
away from the explosion location are recorded as negative since the gauge is placed outside the shock
tube with the glazing moving toward the deflection gauge. Deflections ranged from -4.510 in. (-114.6
mm) to -3.944 in. (-100.2 mm) with a peak average of -4.113 in. (-104.5 mm). The time of peak
deflection varied from 9.8 ms to 10.5 ms.
Deflection results for the one half inch test series were again consistent with an average maximum of -
3.002 in. (-76.25 mm). Deflections varied from -2.846 in. (-72.29 mm) to -3.085 in. (78.36 mm) with
peak deflection times varying from 11.9 ms to 12.7 ms. Following the perimeter tests, each sample was
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 4 of 10
measured to look for residual deformation of the polycarbonate. A residual deformation of 0.75 in. (19
mm) was noted at the end of the test series for the one quarter inch polycarbonate, however, the one half
inch polycarbonate showed no signs of permanent deformation.

Reaction force data collection posed somewhat of a challenge. The results by the reaction measurements
are less consistent than that of pressure and deflection. Although the results are not as clear, some
general patterns can still be observed. A major challenge presented during the Z-axis analysis was the
result of faulty instrumentation. One of the four triaxial load cells produced erroneous results on its Z-
axis. These errors consisted of many large voltage spikes that did not correspond with the blast event.
This error went unnoticed until after the test series was completed and data analysis began. Z-axis data
from this load cell could not be salvaged. Therefore, Z-axis data was not available for positions A, M,
and U.

Upon first examination of the Z-axis data gathered during perimeter testing of the one quarter inch
polycarbonate, the data appears to be rather inconsistent. However, after closer examination, some
patterns begin to appear. A summary of the maximum positive Z-axis loading is presented in Table 1.
The long edges (sides) are shown in white and the short edges (top and bottom) are highlighted in blue.
Looking at each position individually, the first pattern realized is the distribution of loads along the long
edges, in positions A-H and positions N-U. The loading appears to be greatest at the top in positions H
and N, and decrease as it continues towards the midpoint. The peak loadings are 1173.69 lbf (5.2208
kN) at position H and 1138.77 lbf (5.0655 kN) at position N. The long side cannot be fully
characterized as data is not available for positions A and U, but it is assumed that these positions would
exhibit loadings similar to that of the top, approximately 1100 lbf (4.89 kN). The next pattern becomes
distinguishable after the loads for each edge are averaged. Of particular interest is the fact that the
average loading of the left and right sides differ by only 4.21 lbf (0.0187 kN), or less than 1%. A greater
difference is noticed between the top and bottom edges, with the top edge average force equaling 416.05
lbf (1.8507 kN) and the bottom edge equaling 260.31 lbf (1.1579 kN) for a difference of 62%.
Table 1: Summary of one quarter inch polycarbonate peak positive Z-axis loading

(lbf) (kN) (lbf) (kN) (lbf) (kN) (lbf) (KN) (lbf) (kN)
A 1 - - 2 - - 3 - - - -
D 1 297.26 1.3223 2 327.63 1.4574 3 318.22 1.4155 314.37 1.3984
E 1 406.13 1.8066 2 490.85 2.1834 3 420.18 1.8690 439.05 1.9530
H 1 1206.95 5.3688 2 1199.00 5.3334 3 1164.64 5.1806
H 4 1104.97 4.9152 5 1200.67 5.3408 6 1165.90 5.1862
I 4 569.18 2.5318 5 617.71 2.7477 6 617.71 2.7477 601.53 2.6758
K 4 219.95 0.9784 5 226.69 1.0084 6 245.08 1.0902 230.57 1.0256
M 4 - - 5 - - 6 - - - -
N 7 1110.49 4.9397 8 1134.90 5.0483 9 1170.93 5.2085 1138.77 5.0655
Q 7 336.74 1.4979 8 420.18 1.8690 9 447.85 1.9921 401.59 1.7864
R 7 375.96 1.6724 8 361.42 1.6077 9 384.94 1.7123 374.11 1.6641
U 7 - - 8 - - 9 - - - -
U 10 - - 11 - - 12 - - - -
V 10 260.48 1.1587 11 491.02 2.1842 12 363.99 1.6191 371.83 1.6540
X 10 234.99 1.0453 11 216.69 0.9639 12 247.77 1.1021 233.15 1.0371
Z 10 171.76 0.7640 11 181.82 0.8088 12 174.28 0.7752 175.95 0.7827
2.8387
1.1579
5.2208
Force Force Avg Force
1/4 Inch Polycarbonate Perimeter Testing
Maximum Positive Z- Axis Loading
Avg Force per
Edge Test # Position Test # Test #
Force
1173.69
642.37
416.05
638.16
260.31
2.8574
1.8507
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 5 of 10
Examination of the one half inch polycarbonate yielded similar results. For an unknown reason, the
peak loading experience by position N on Test 9 was dramatically lower than the previous two tests at
that position. Loads experience by positions Q and R were higher than previously recorded. For
analysis, the values recorded on Test 9 are not used. A summary of values recorded for one half inch
polycarbonate perimeter testing are shown in Table 2 with long edge values in white and short edge
values in blue.
Table 2: Summary of one half inch polycarbonate peak positive Z-axis loading


Similar to the one quarter inch data, the first pattern noticed when looking at the one half inch data is the
distribution of forces along the long edges. The peak loadings were recorded at positions H and N, with
values of 965.09 lbf (4.2929 kN) and 893.80 lbf (3.9758 kN), respectively. One outlier in the
distribution of loading on the long edges is observed at positions D and E, with position D exhibiting a
higher loading than E. However, loadings on the opposite side, positions N-R show greater loadings at
the top and decreasing loads towards the lower midpoint position. Again, the long edges cannot be fully
characterized due to the lack of data in positions A and U. A very noticeable similarity occurs when the
loadings along each edge are averaged. The left and right sides have nearly identical loadings with
values of 537.53 lbf (2.3911 kN) and 537.80 lbf (2.3923 kN), respectively. The top and bottom edges
differ by 153.84 lbf (0.6843 kN) or 58%, similar to the 62% noticed on the one quarter inch sample.

As mentioned previously, data was not recorded at the lower most positions on the long edges.
Therefore, a simple supplemental test series involving five tests was conducted to try to fill in these
holes. Load cells were placed in the lower and upper most positions on both long edges. Although the
results were not definitive, a better idea of what is occurring can be observed from these results. A
summary of the supplemental test series is shown in Table 3. To prevent skewing of data, the results
from the supplemental testing are not combined with the original perimeter testing. Some assumptions
based on the data can be made, but stating the results are definitive would not be justified at this time. It
is noted that the values recorded for right side positions N and U are within 10% of each other.
(lbf) (kN) (lbf) (kN) (lbf) kN) (lbf) (kN) (lbf) (kN)
A 1 - - 2 - - 3 - - - -
D 1 361.42 1.6077 2 357.57 1.5906 3 361.42 1.6077 360.14 1.6020
E 1 249.04 1.1078 2 289.48 1.2877 3 323.54 1.4392 287.35 1.2782
H 1 847.93 3.7718 2 976.54 4.3439 3 958.53 4.2637
H 4 1055.30 4.6942 5 989.53 4.4016 6 962.71 4.2824
I 4 429.54 1.9107 5 466.16 2.0736 6 488.29 2.1720 461.33 2.0521
K 4 264.76 1.1777 5 272.46 1.2119 6 277.59 1.2348 271.60 1.2081
M 4 - - 5 - - 6 - - - -
N 7 993.30 4.4184 8 794.30 3.5332 9 345.20 1.5355 893.80 3.9758
Q 7 469.99 2.0906 8 446.57 1.9865 9 522.35 2.3235 458.28 2.0385
R 7 264.76 1.1777 8 257.91 1.1473 9 382.38 1.7009 261.33 1.1625
U 7 - - 8 - - 9 - - - -
U 10 - - 11 - - 12 - - - -
V 10 293.75 1.3067 11 226.20 1.0062 12 275.37 1.2249 265.11 1.1793
X 10 256.78 1.1422 11 239.32 1.0646 12 233.12 1.0370 243.07 1.0812
Z 10 142.15 0.6323 11 129.06 0.5741 12 117.92 0.5246 129.71 0.5770
1.6301
2.3923
0.9458
4.2929
1/2 Inch Polycarbonate Perimeter Testing
Maximum Positive Z- Axis Loading
Avg Force
Avg Force per
Edge Position Test #
2.3910
Force
Test #
Force
Test #
Force
965.09
537.53
366.47
537.80
212.63
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 6 of 10
However, the values recorded on the left side in positions H and A differ by approximately 60%. Based
on this information, the only conclusion that the authors feel safe stating is that the values observed at
the top and bottom of the long edges are likely higher than those observed at the mid span.

Table 3: Summary of one half inch poly Z-axis loading during supplemental perimeter testing

When comparing the reaction forces of the polycarbonate panels side by side, interesting and possibly
very important patterns emerges. The most notable of these differences is that the magnitude of Z-axis
loading is less for the one half inch polycarbonate when compared to the one quarter inch polycarbonate.
Next, the average reaction force for each of the edges was compared to the average reaction force of the
left edge and expressed as a percentage. Figures 4 and 5 are graphical representations of the load
distributions for the one quarter inch and one half inch samples. From this comparison, it can be seen
that on a percentage basis, the load distributions for each of the sides is very similar between the
polycarbonate samples.

Figure 3: Load distribution for one quarter inch
polycarbonate

Figure 4: Load distribution of one half inch
polycarbonate
Failure Testing Results and Anal ysis
Two tests were required to fail the one quarter inch polycarbonate system. Due to the extreme flexure of
the system, the screws holding the frame to the attachment points sheared, allowing the system to fall
back into the shock tube during the negative blast phase. The first test resulted in a maximum peak
pressure of 6.437 psi (44.38 kPa) and impulse of 47.809 psi-ms (329.63 kPa-ms). This caused screws in
(lbf) (kN) (lbf) (kN) (lbf) (kN) (lbf) (kN) (lbf) (kN) (lbf) (kN) (lbf) (kN)
A 1128.26 5.0187 1272.80 5.6617 1207.25 5.3701 1206.03 5.3647 1223.53 5.4426 1208.20 5.3743
H 707.61 3.1476 732.58 3.2587 722.51 3.2139 745.87 3.3178 745.87 3.3178 728.69 3.2414
N 1026.62 4.5667 1043.97 4.6438 1029.04 4.5774 1011.70 4.5003 990.72 4.4070 1020.44 4.5391
U 868.44 3.8630 963.83 4.2873 936.80 4.1671 949.13 4.2219 943.96 4.1989 925.41 4.1164
3.2414
4.3278
Avg Force
Avg Force per
Edge Position Test # Test # Test # Test # Test #
Force Force Force Force Force
728.69
972.92
1 2 3 4 5
1/2 Inch Polycarbonate Perimeter Testing - Supplemental
Maximum Positive Z- Axis Loading
642.37 lbf 638.16 lbf
2.8574 kN 2.8387 kN
100.00% 99.34%
64.77%
416.05 lbf / 1.8507 kN
40.52%
260.31 lbf / 1.1579 kN
537.53 lbf 537.8
2.3910 kN 2.3922 kN
100.00% 100.05%
366.47 lbf / 1.6301 kN
68.18%
39.56%
212.63 lbf / 0.9458 kN
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 7 of 10
nine of the 26 positions to shear, but the window remained in place. The second test, which resulted in
the failure of the glazing system, achieved a peak pressure of 8.287 psi (57.13 kPa) and impulse of
74.713 psi-ms (515.13 kPa-ms). It should be reiterated that glazing system itself did not fail, rather the
screws holding the frame to the attachment points failed, allowing the glazing system to fall into the
shock tube. The glazing system remains largely intact with some deformation of the aluminum framing.
A summary of the Z-axis reaction forces for the one quarter inch polycarbonate can be found in Table 4
along with the pressure and impulse associated with each of the two tests. The reaction forces increased
from Test 1 to 2 as anticipated with the exception of position D, in which case the peak loading
decreased from 1058.17 lbf (4.7070 kN) in Test 1 to 770.32 lbf (3.4265 kN) in Test 2. The reason for
this is unknown as the reaction force curve does not exhibit any abnormalities. It is possible that a
greater percentage of the loading was transferred to an adjacent attachment point but without
instrumentation this cannot be verified.
The one half inch polycarbonate system required three tests to cause failure. The mode of failure was
similar to that of the one quarter inch system. The flexure of the framing material caused a majority of
the screws to shear resulting in the system being sucked 14 ft. (4.3 m) back into the shock tube, flipping
in the process. Had the screws not failed, it is uncertain how much more loading the glazing system
may have been able to withstand. The first test resulted in a pressure of 6.799 psi (46.877 kPa) and
impulse of 49.591 psi-ms (341.918 kPa-ms). There was no noticeable damage to the glazing system
following this test. The next test resulted in a pressure of 8.355 psi (57.606 kPa) and impulse of 77.823
psi-ms (536.571 kPa-ms). Damage to the glazing system was limited to some slight bending of the
framing material concentrated near the attachment points. The final test, which achieved a peak
pressure of 10.454 psi (72.078 kPa) and impulse of 120.923 psi-ms (833.735 kPa-ms), resulted in the
glazing system failing and being sucked 14 ft. (4.3 m) back into the shock tube.
Table 4: Reaction force loading for one quarter Inch Polycarbonate Failure Testing

A summary of the Z-axis reaction forces for the one quarter inch polycarbonate can be found in Table 5
along with the pressure and impulse associated with each of the three tests. The reaction force results
are as expected with the loading increase for each test until the sample failed. There was a slight
deviation in the data for Test 2, position D, with the peak value decreasing slightly from Test 1 to Test 2.
However, the value did increase for Test 3. As with one quarter inch failure testing, the reason for this
decrease cannot be explained.

(lbf) (kN) (lbf) (kN)
D 1058.17 4.7070 770.32 3.4265
E - - - -
Q 646.23 2.8746 929.76 4.1358
R 440.30 1.9586 765.82 3.4065
1/4 Inch Polycarbonate Failure Testing
Maximum Positive Z- Axis Loading
Position
Test #1
Pressure/Impulse
Test #1 Force Test #2 Force Test #2
Pressure/Impulse
6.437 psi /47.696 psi-ms
44.38 kPa/329.63 kPa-ms
8.287 psi /74.713 psi-ms
57.13 kPa/515.13 kPa-ms
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 8 of 10
Table 5: Reaction force loading for one half Inch Polycarbonate Failure Testing


Comparison to HazL Modeling
Window Fragment Hazard Level Analysis, or HazL, is a software program developed by the United
States Army Corps of Engineers Protective Design Center. This program performs a single degree of
freedom (SDOF) analysis on a blast resistant glazing system to calculate the glazing response to blast
loading. Input parameters such as blast load, window geometry, and glazing type and thickness can be
entered. The program is then able to output the hazard level, glazing response, and reaction loads. This
information is commonly used to determine the loading on support members in structural design
calculations.
The first sample analyzed using the program was the one quarter inch polycarbonate blast resistant
glazing system. Using a representative blast waveform from the experimental test series as the input
blast loading yielded interesting results. With a peak pressure of 4.7 psi (32.4 kPa) and impulse of 21.0
psi-ms (145 kPa-ms), the model predicts failure of the sample with the glazing pulling out of the frame.
As was shown through multiple experimental tests, the one quarter inch sample survived this loading
with very minimal damage. In fact, a loading of 8.416 psi/74.319 psi-ms (58.03 kPa/512.4 kPa-ms) was
required to cause pullout of the screws holding the glazing system to the attachment points, in which
case the glazing still remained in the frame. Table 6 summarizes the model predicted reaction forces
versus the reaction forces determined experimentally. As shown in the table, the program greatly over-
predicted the reaction forces. Another interesting result presented by the program was peak deflection
of the glazing. Through experimental results it was shown that the average peak deflection was 4.113
in. (104.5 mm). The program predicted a peak deflection of 5.185 in. (131.7 mm) at this loading. A
number of different theoretical blast loading scenarios were run using the program until a loading was
found that did not cause failure. This loading had a pressure of 0.70 psi (4.8 kPa) and 5.25 psi-ms (36.2
kPa-ms) with a positive phase duration of 15 ms, in which case the model still predicted a peak
deflection of 3.58 in. (90.9 mm) at 22.17 ms. It is hypothesized that the program over-predicting the
deflection of the one quarter inch polycarbonate, leads it to predict premature failure of the glazing
system.
The one half inch polycarbonate glazing system was also analyzed using the program in the same
manner. Using the same typical blast waveform, the model predicted that the sample would survive;
however, the peak deflection was overstated at 3.939 in. (100.1 mm) compared to 3.002 in. (76.25 mm)
witnessed experimentally. A summary of the program predicted reaction forces versus experimental
reactions forces is shown in Table 7. In this case, the model predicted reaction forces roughly two times
higher than those predicted for the one quarter inch sample. However, experimental results showed that
the Z-axis reaction forces decreased from the one quarter inch sample to the one half inch sample.

It should be emphasized that although the program may be overestimating the load transfer for a given
blast event, it is underestimating the overall survivability of the glazing system. As a result, this may
(lbf) (kN) (lbf) (kN) (lbf) (kN)
D 483.32 2.1499 474.34 2.1100 827.63 3.6815
E - - - - - -
Q 718.18 3.1946 742.87 3.3044 1186.01 5.2756
R 372.43 1.6567 430.25 1.9138 642.40 2.8575
Test #1 Force Test #2 Force
1/2 Inch Polycarbonate Failure Testing
Maximum Positive Z- Axis Loading
8.355 psi /77.823 psi-ms
57.61 kPa/536.57 kPa-ms
10.454 psi /120.923 psi-ms
72.078 kPa/833.735 kPa-ms
Position
Test #1
Pressure/Impulse
Test #2
Pressure/Impulse
Test #3
Pressure/Impulse
Test #3 Force
6.799 psi /49.591 psi-ms
46.88 kPa/341.918 kPa-ms
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 9 of 10
provide load transfer data that is in fact lower than what could be expected at the true failure point of the
glazing system.
Table 6: Comparison of model output to
experimental results
Table 7: Comparison of model output to
experimental results

Conclusions
Perimeter testing results have shown that different thicknesses of polycarbonate do exhibit the same load
distributions, albeit at different magnitudes. Another point presented was that testing of a single
polycarbonate sample should be limited as residual stresses are likely to accumulate, leading to
permanent deformation and possibly erroneous data. Finally, it was shown that it may be possible to
predict the load distribution around the perimeter of the glazing system by instrumenting only a limited
number of points. However, future testing of additional polycarbonate panels of differing thickness and
sizes would be required to validate this hypothesis.

A software tool was used to analyze each of the polycarbonate glazing systems and provided results that
did not correspond well with test data. In the case of the one quarter inch polycarbonate, the program
dramatically under-predicted the failure point, predicting that the sample would fail at blast loadings of
less than 1 psi, when in fact the sample survived 19 tests at pressures of 4.5 psi (31 kPa) or greater. The
program also over-predicted peak deflection and peak loading at failure. The analysis of one half inch
polycarbonate presented the same trends. If supporting members are designed using only the blast
loadings at which the program predicts polycarbonate glazing system failure, they may be dramatically
undersized. This could result in severe damage to the support members and transmit greater than
anticipated loads down the load path, resulting in progressive collapse of the structure.

Based on these conclusions, the only currently acceptable solution to accurately validate the design of
polycarbonate blast resistant glazing systems is to subject samples to blast loading to determine glazing
resistance. The tests should be conducted in manner consistent with real world applications, and the
samples should be tested to failure.

Bibliography
Calnan, J .T., 2013, Experimental Comparison Study of the Response of Polycarbonate and Laminated
Glass Blast Resistant Glazing Systems to Blast Loading,. Masters Thesis, College of
Engineering. University of Kentucky, Lexington, KY.
HazL, 1998, https://pdc.usace.army.mil/software/hazl/, Protective Design Center, Omaha District Corps
of Engineers, Omaha, NE.
Wedding, W. C.,2010, Experimental Study of Blast Resistant Glazing System Response to Explosive
Loading,. Masters Thesis, College of Engineering. University of Kentucky, Lexington, KY.
(l bf) (kN) (l bs/i n) kN/m (l bf) (kN) (l bs/i n) (kN/m)
Long Edge 959.64 4.2687 14.54 2.541 640.27 2.8481 9.70 1.695
Short Edge 627.44 2.7910 13.14 2.307 338.18 1.5043 7.08 1.243
Reacti on Force Reacti on Force Peak Loadi ng Peak Loadi ng
1/4 Inch Comparison
Experimental Results HazL Results
(l bf) (kN) (l bs/i n) (kN/m) (l bf) (kN) (l bs/i n) (kN/m)
Long Edge 2269.08 10.0934 34.38 6.008 537.66 2.3916 8.15 1.424
Short Edge 1298.32 5.7752 27.19 4.773 289.55 1.2880 6.07 1.064
Reacti on Force Reacti on Force Peak Loadi ng Peak Loadi ng
1/2 Inch Comparison
Experimental Results HazL Results
Copyright 2014 International Society of Explosives Engineers
2014G - Load Transfer of Polycarbonate Blast Resistant Glazing Systems 10 of 10