Engineering Structures 33 (2011) 3290–3298

Contents lists available at SciVerse ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Response of pierced fixed corrugated steel roofing systems subjected to

wind loads
David J. Henderson ∗ , John D. Ginger
Cyclone Testing Station, James Cook University, Townsville, QLD, Australia

article info abstract

Article history: The low-cycle fatigue response of corrugated metal roof cladding to fluctuating wind loads was studied by
Received 2 November 2010 subjecting cladding specimens to a series of static, cyclic and simulated ‘‘real’’ cyclonic wind loads using
Received in revised form a Pressure Loading Actuator (PLA), and measuring fastener response using a x–y–z load cell. The overall
31 July 2011
performance of cladding including crack initiation, propagation and patterns, and cycles to failure was
Accepted 2 August 2011
Available online 29 September 2011
found to be similar to previous tests that used line-loads to simulate wind pressure. The reaction at a
fastener to spatially varying pressures was assessed by analysing the influence coefficients, to show that
Keywords:
it is predominantly influenced by local loads acting along the screwed crest. In addition, the response of
Wind roofing specimens subjected to fluctuating cyclonic wind pressures replicated failures observed in the
Load field. The fastener response varied with the load level and the response spectrum followed the wind
Roof load spectrum up to 5 Hz even with deformation and cracking of the cladding showing that these higher
Cladding frequency wind ‘‘load cycles’’ were transferred into the supporting structure via the fastener.
Cyclone © 2011 Elsevier Ltd. All rights reserved.
Fatigue
Test
Damage
Housing

1. Introduction at a level near the ultimate static capacity. The number of cycles
to failure is also dependent on the load ratio defined as R =
Roofing of low-rise buildings (e.g., houses, warehouses, indus- Smin /Smax , where the load varies between a maximum, Smax and a
trial sheds) are subjected to large temporally and spatially vary- minimum Smin . Low-cycle fatigue is commonly defined as failure
ing wind pressures during wind storms such as a tropical cyclone. within 10,000 load cycles [3] with the magnitude of loading cycles
These fluctuating pressures are generated by turbulence in the ap- exceeding the yield stress.
proaching wind flow and flow around the building. The pressures The fatigue performance of roof cladding has been studied by
on parts of the roof vary with the changes in wind speeds and di- Beck and Stevens [3], Mahendran [4], and Xu [5], by subjecting
rections as a cyclone tracks past the building. Roofs clad with light specimens to cyclic loads. The data was analysed in terms of the
gauge profiled metal sheeting are susceptible to low-cycle fatigue, peak reaction at the fastener (Smax ) vs number of cycles to failure
as was seen in Darwin, Australia after Cyclone Tracy in 1974, where (N) or S–N curves. They found that the fatigue behaviour of the
more than 90% of houses and 70% of other structures suffered sig- thin (0.42 mm BMT) but high yield strength G550 grade cladding
nificant loss of roof cladding [1]. Following Cyclone Larry which is dependent on the load causing local plastic deformation (LPD),
impacted on Innisfail, Australia in 2006, some cases of fatigue fail- seen as dimpling under the screws around the fastener holes,
ure of metal cladding were also observed [2]. shown in Fig. 1. The fatigue life of corrugated cladding increases
A component fails in fatigue following the application of markedly if the load per fastener is kept below this local plastic
fluctuating loads of lower magnitude than the component’s static deformation load [3–5].
load capacity (strength). Metal roof cladding is susceptible to low- The interaction between the cladding and fastener is a crucial
cycle fatigue cracking at the fastener with failure by the cladding part of the cladding’s fatigue response to the applied load [6,
disengaging over the head of the screw. Fatigue failure may occur 7] and is affected by the thinness of the material relative to
from a large number of low level load cycles or from a few cycles the large region of plastic deformation near the screw, the load
history, and the range of stresses applied spanning the non-linear
S–N curve. Hence, typical predictive or analytical tools such as
∗ Corresponding author. Miner’s Rule, linear elastic fracture mechanics (LEFM) or crack
E-mail addresses: [email protected] (D.J. Henderson), tip opening methods (CTOD/CTOA) have limited capability for
[email protected] (J.D. Ginger). estimating fatigue performance [8–10]. Physical testing of cladding
0141-0296/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.engstruct.2011.08.020
 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298 3291

The typical corrugated cladding is rolled from a 0.42 mm base

metal thickness (BMT) G550 coil, which refers to the specified
characteristic yield strength of 550 MPa. The longitudinal tensile
strengths of the coils used in tests by Mahendran [20] and Xu [14]
were 690 MPa and 720 MPa, respectively. The transverse strengths
are larger than the longitudinal due to the initial rolling to reduce
billet thickness and the roll forming to shape the cladding profile.
Tests in this study were conducted using cladding with a mean
longitudinal tensile strength of 780 MPa.

3. Tributary area for a cladding screw

The uplift load at a fastener FZ (t ) at a time, t, is given by Eq. (1):

where p(x, y, t ) is the uplift pressure acting on the cladding at a
Fig. 1. Local plastic deformation of corrugated cladding under the head of the distance x and y from the screw, I (x, y) is the influence coefficient
fastener (hex-head screw). for the load on the fastener for the pressure at (x, y), and dA is the
area over which p acts.
systems is therefore an important means for both research and ∫
product evaluation. FZ (t ) = p(x, y, t ) · I (x, y)dA. (1)
Mahendran [11] and Xu [8] developed test programs which
used a then state of the art test rig to simulate pressures acting The influence coefficient is the value of the uplift load on the
on cladding specimens. Their systems used a double span set-up screw for a spatially varying unit uplift load applied to the cladding
that simulated the pressure acting on the cladding by applying specimen at (x, y). Influence coefficients for fastener loads were
a line-load across the profile in the middle of each span. Wind determined from a series of point load tests on cladding specimens.
loads were simulated using a sinusoidal load pattern on test The influence coefficient is defined as the ratio of fastener load
specimens that did not incorporate cladding laps. Engineering FZ to the load that is applied at various locations on the cladding
mechanics theory was used to replicate the line-loaded double specimen, at a location (x, y) from the fastener.
span with that of a uniformly loaded double span, by equating the The influence coefficients for FZ were obtained by applying an
central support reactions and moments for the two loading cases. upwards point load via a S-type load cell to a triple span cladding
Thus, they approximated a uniformly loaded 900 mm double span system as described in Henderson [19], and reproduced in Table 1.
cladding system to their line loaded 650 mm double span set-up. The loads on the fastener were measured using an S-type load
However, the response of a cladding specimen is constrained by cell in a similar manner to that described in Section 4. These
limitations of the load application which can either restrain the coefficients reveal that the majority of load was transferred along
cladding specimen or not conform to the profiled cladding shape. the screw fixed crests, with about 5% of load ‘‘felt’’ by the adjacent
This can result in non-representative local failure modes and screw fixed crests. Therefore, most of load applied to the cladding is
capacities [12–14]. These comprehensive line-load test programs transferred via the stiff cross sectional profile (i.e., along the crests)
underpin building product design standards and current low-cycle with a small percentage of the load carried to adjacent screws from
fatigue test criteria, such as AS4040.3 [15] and the Building Code of the cladding acting as a plate.
Australia [16]. Henderson [19] used wind tunnel data from a low pitch, low-
rise building, with a very high density of pressure taps, to explore
With the recent advent of Pressure Loading Actuators (PLA)
the effects of area averaging pressures for a range of wind approach
[17,18] capable of generating realistic wind pressures, the fatigue
angles, on an elongated rectangle representative of the tributary
response of corrugated metal roof cladding to fluctuating wind
area supported by a fastener. The pressures acting on the crest
loads were studied by subjecting cladding specimens to series of
near the fastener mostly influences the reaction (as shown by
static, cyclic and ‘‘real’’ cyclonic wind loads by Henderson [19]. The
influence coefficients) and hence the pressure acting away from
responses of the cladding and fasteners are reported in this paper
that fastener’s crest has limited affect on the fastener in question.
by measuring fastener reactions using a 3 axis (x–y–z) load cell, and
Henderson [19] showed that the spatially and temporally varying
are compared to previous line-load studies.
loads on the cladding system can be satisfactorily represented by
This paper analyses the response of a fastener to spatially
applying spatially-uniform fluctuating loads on a sheet of cladding
varying wind loads by measuring the force (or reaction) on
in an air chamber. However, this may not be the case for standing-
the fastener. These results are used for identifying the effective
seam roof systems with wider fastener spacings, as shown by
cladding tributary area for a typical fastener and its extent of
Surry [21]. The standing-seam fastener’s typical tributary area was
influence on adjacent fasteners. The adequacy of applying a 0.92 m2 , that is a 610 mm spacing between seams (ribs) and
uniform load across the test specimen is discussed with respect 1520 mm between rows of clips (e.g. purlins). The typical tributary
to the spatial and temporal variations occurring on the roof of a area for the corrugated pierced fixed roofing is a much smaller
building. elongated area of 0.14 m2 (152 mm × 900 mm).

2. Pierced fixed metal roofing 4. Air-chamber set-up and testing

The metal roof cladding is fixed to battens typically spaced Tests were conducted on corrugated cladding specimens
about 900 mm apart. In cyclonic regions, the cladding is attached to installed in a 2000 mm long × 890 mm wide × 300 mm deep air-
the batten or purlin at intervals of 150–200 mm, using either self- chamber, shown in Fig. 2, as detailed by Henderson [19]. A range
drilling timber or metal screws depending on the type of support. of pressures; ramp, sinusoidal and simulated wind were generated
The cladding screws have a washer, which acts as a weather seal, in the air chamber and applied to the cladding specimen to study
of approximately 12 mm diameter between the screw’s 16 mm the reaction at fasteners and the condition of the cladding. The
diameter flange and the cladding, as shown in Fig. 1. Pressure Loading Actuator (PLA) and computer control systems
3292 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298

Table 1
Influence coefficients for fastener load FZ at (x = 152 mm, y = 900 mm).
Fastener spacing, x (mm) Influence coefficients
Distance from first purlin, y (mm)
0 225 450 675 900 1125 1350 1575 1800
a a
0 0 0.02 0.04 0.04 0 (S1) 0.04 0.02 0.02 0a
152 0a 0.27 0.55 0.76 1a (S2) 0.73 0.47 0.21 0a
304 0a 0.02 0.05 0.05 0a (S3) 0.07 0.04 0.04 0a
a
Inferred values tabled since point loads were not able to be applied at screw locations.

Fig. 3. Air-chamber in Configuration B prior to installation of cladding (i.e. the

exterior face of the cladding to be facing inwards with a negative pressure applied
in the air-chamber). Note: Screw stubs mounted into threaded brackets.

Fig. 2. Air-chamber in Configuration A prior to installation of cladding to be screw

fixed to Z -purlin (i.e. the exterior face of the cladding to be facing upwards with a
positive pressure applied to the air-chamber).

supplied and controlled the air pressure in the chamber. The

cladding forms the ‘‘lid’’ of the air-chamber. Therefore the PLA
supplied pressure acts on the inside surface, that is, the cladding
surface facing the air-chamber.
The ability of the PLA, to generate negative and positive pres-
sures allowed two test configurations to be designed. Configura-
tion A used the exterior surface of the cladding installed facing Fig. 4. X , Y and Z axis JR3 load cell and adjacent ‘‘S-type’’ load cell with screw stubs
upwards, that is a normal installation for a roof with screws driven inserted into screw holes from removed cladding screws.
through the cladding into battens or purlins. Configuration A, with-
out the cladding installed is shown in Fig. 2. Conversely, Config- beam S-type load cell. Screw stubs were connected to each
uration B had the exterior surface of the cladding facing into the load cell as detailed in Henderson [19]. The JR3 load cell is a
air-chamber, thereby allowing a negative pressure (i.e. suction) to proprietary doughnut shaped, instrumented aluminium block that
act on the cladding, allowing the ‘underside’ of the cladding (now can measure loads in the X , Y and Z directions, as shown in
facing up) to be observed. This gave an unobstructed view of the Fig. 4. The adjacent S-type load cell is used to measure reaction
development of the cladding deformation and cracking adjacent to in the Z direction. The screw stubs, including the compressible
the screw shank. Fig. 3 shows Configuration B prior to the instal- sealing washer, were standard cladding fasteners with the length
lation of the cladding specimen into the air-chamber. Before in- shortened by cutting off the thread and leaving the shank such that
stallation, the cladding specimen is placed into a template of the the stubs were not long enough to touch the purlin. The heads
air-chamber so that the fasteners are installed then removed giv- of the screw stubs were braised to threaded rods connected to
ing a known location for screw holes such that the specimen was the load cells allowing height adjustment to ensure correct screw
then placed over the screw stubs shown in Fig. 3. tightness on the compressible seal.
A three axis load cell was used to analyse the relationship In addition to the full sheet width of 860 mm, two edge strips
between the loads at a fastener and the changing damage state of were used to simulate the laps on either side of the sheet as shown
the cladding. Pressure transducers, displacement transducers and in Fig. 5. Both strips were approximately 100 mm wide and were
strain gauges were used in the tests and a digital camera mounted cut from the corresponding edges of another sheet. The cladding
above a test specimen linked to the data acquisition system was was fixed to the purlins at alternate crests and the fasteners on the
used to measure lateral movement of the screw and measure the central purlin identified as S1 · · · S6.
cladding crack lengths.
The reaction at the cladding fastener is a critical parameter 4.1. Static uniform pressure loading
in assessing the performance of the cladding system under
wind loading. The reactions at selected cladding fasteners were Under a steadily increasing, spatially-uniform uplift pressure,
measured by a three axis JR3 load cell and a standard shear the cladding undergoes a transformation resulting in changes to
 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298 3293

the z-load at the screws for the applied pressure is proportionally

less than prior to the cladding deforming. This is similar to the
results from line-load testing by Xu [14], and Xu and Reardon [22],
where the reaction coefficient drops to that of a hinged support
but then increases through membrane action of the cladding with
increasing load. Xu [8] noted that the response, of the cladding to
static loads provides an insight to the cladding system’s fatigue
Fig. 5. Cross section of cladding showing edge strips and fasteners on central purlin.
performance.
From the static load trials, the formation of the creasing
(buckling) of the cladding adjacent to the screws typically occurred
at loads between 600 and 650 N.

4.2. Sinusoidal loading

Sinusoidal load tests, with pressure variations in form, range,

peak, and frequency, were carried out to study the response of
the cladding specimens, in terms of numbers of cycles to failure,
crack development and crack patterns, cladding deformation and
reaction at the fasteners.
Sinusoidal pressures were applied to cladding test specimens
until fatigue failure with the cladding pulling over at least one of
the screws. The applied pressures were typically at a frequency of
Fig. 6. Applied static step pressure (p), X , Y and Z load components at screw S3, 1.5 Hz. Fig. 7 shows the peak cyclic pressures (Pmax ) applied versus
and CR . the number of cycles to failure (Ni ) for a range of load ratios (R).
The uplift pressures are denoted negative to follow the convention
its response characteristics. Initially, the cladding deflection is used in wind load standards such as AS/NZS 1170.2 [23].
proportional to the applied pressure. As the pressure is increased, Sinusoidal load tests carried out with a similar peak load but
deformation (yielding) of the cladding under the screw head with a larger R withstood a larger number of load cycles than tests
results in an increase in rate of deflection with increase in pressure. with smaller R, as shown in Fig. 7. The increased life, that is, the
The z-load at the screw is given by Fz = CR × p × A: where larger number of cycles to failure, for trials with larger R is due to
CR is the reaction coefficient, p is the uniform pressure and A is the reduced flexure per cycle and the smaller change in stress state
the tributary area of the fastener (i.e., cladding span × fastener at the crack tip. The equivalent results for R = 0, double 650 mm
spacing). For a double span set-up (assuming a continuous beam span line-load tests obtained by Xu [8] are also presented in Fig. 7
spanning over an internal support in the middle of its length) CR and form a lower bound of the air-chamber R = 0 data.
equals 1.25. This is the approach used for converting the applied
line-load to an equivalent pressure by Mahendran [6] and Xu [14]. 4.3. Crack patterns
Fig. 6, presents the X , Y and Z loads, FX , FY and FZ respectively,
measured by the JR3, for a stepped static applied pressure, and non- The fatigue failures produced two typical types of crack
dimensional reaction coefficient (CR ). The CR is about 1.2 during patterns: crease and star type, as shown in Fig. 8(a) and (b),
the first three steps until a spike in FZ (at 48 s) coincided with the respectively. The crease and star type crack patterns are classified
start of the fourth pressure step, and a drop in CR to about 1.05, as: cracks initiating at crease points away from the screw hole and
when the cladding adjacent to the fastener creased and buckled. cracks initiating at screw hole. Similar observations were made by
The creasing of the screwed crests can also be detected as a spike Mahendran [4] and Xu [14] using line-load test loading systems.
in FX and to a lesser extent in FY . The lateral load (FX ) on the The crease cracks are sub-divided into two patterns ‘H’ and ‘T’, as
‘‘screw’’ is approximately 15% of FZ . During the fourth and fifth load described by Henderson [19].
steps, the cladding continued to deform with the unscrewed crests Fig. 9 shows the number of cycles at which a crack is initiated
deflecting more than the screwed crests and with a CR increase and the number of cycles at which failure occurred for Pmax . Fig. 9
for the final pressure step. The reduction in stiffness of the profile also shows crack types at failure for each test which can be related
across the central support reduces the reaction coefficient, that is, to the load ratio R and maximum load per cycle shown in Fig. 7.

Fig. 7. Peak pressure with numbers of cycles to failure with varying R values from the air-chamber trials and for Xu [8] line-load test data.
3294 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298

Fig. 8. Example of (a) ‘crease’ type and (b) ‘star’ type crack patterns.

4.3.2. Star crack type

Star cracks are initiated in the cladding at the screw hole under
the compressible seal that fits between the underside of the screw
head and the cladding as a weather seal on the screw holes. Cracks
formed at lower pressures than the pressures measured for local
plastic deformation during the static loading tests.
The star crack develops with the local plastic deformation of the
cladding (flattening of crest) under the head of the screw, shown
in Fig. 1 through to the propagation of transverse and longitudinal
cracks (star crack pattern) as shown in Fig. 8(b).
As for the crease type cracks, the lateral loads of FX and FY
increase during the load cycling, albeit at a lower magnitude. These
reactions were dependent on the development of the cracks and
Fig. 9. Number of cycles to crack initiation and to failure for star and crease (‘H’ the location of the fastener in the hole. FX and FY drop from the
and ‘T’) type cracks. initial values of approximately 5% of Fz, with crack formation, but
increase with non-uniform crack growth applying an eccentric
Crack initiation is defined when a crack greater than about loading on the screw. At the completion of the test, rub marks were
1.0 mm was detected. The numbers of cycles for initiation and for typically observed on one quadrant of the shank with the opposite
failure may not be for the same screw in the specimen being tested. quadrant of the screw flange having pronounced rub marks. The
They represent the threshold numbers of cycles for initiation and disengagement of cladding over a fastener generates lateral loads
for failure of any one of the four ‘‘fully loaded’’ screws, shown in in the order of 25% of Fz in an adjacent fastener [19].
Fig. 5 as S2–S5, in each test specimen. The Reaction coefficient (CR ) for Fz obtained at the peak
pressures of each cycle reduced over the duration of the test.
A crack is initiated relatively earlier in the cycle counts for
The CR decreased from approximately 1.25 to 1.15, showing that
crease type cracks, with the cracks forming at the crease points
the cladding was losing stiffness with increasing region of plastic
away from the screw head. The load per cycle is greater for these
deformation and crack growth.
cracks than the star cracks that form under the screw head. In
addition, significantly longer cracks are required for failure to occur
from crease type cracks, with corresponding faster growth rates. 4.3.3. S–N relationship
Repeat sinusoidal tests were conducted with the same Pmax
and R, to compare the variations in numbers of cycles to failure.
4.3.1. Crease type crack group
These tests indicated that the star and crease crack patterns had
Crease points were formed near the edges of the deformed approximately 30%–40% variation in the number of cycles to failure
crests at loads greater than the local plastic deformation (LPD) with respect to the average number of cycles. The line-load tests by
strength of the corrugated cladding. Cracks were initiated at Mahendran [20] also show a large spread in the ‘‘number of cycles
these crease points and grow from these points forming an ‘X ’ to failure’’, in the order of 40%.
shape longitudinally and transversely. A typical crease type crack The peak load per fastener (Smax ) versus numbers of cycles to
formation and propagation recorded in test Configuration B is failure is shown in Fig. 12, for these PLA air-chamber tests and the
shown in Fig. 10(a)–(d): (a) formation of the crease points (at line-load tests by Mahendran [20]. The results obtained from these
33 cycles) followed by initial cracking at the crease points (at 50 two test methods for the range of R values is consistent within the
cycles), (b) crack growth in the shape of an ‘X ’ (at 140 cycles), variation of number of cycles to failure. The vertical bars in the
(c) crack growth with one side towards screw and the other air chamber PLA tests indicate the variation in fastener reaction
longitudinally (at 680 cycles), and (d) crack growth extending from resulting from the changes in the reaction coefficient CR .
screw hole prior to failure at 1040 cycles. Fig. 12 shows the lines of best fit in terms of Smax = C1 × ln(Ni )+
The corresponding loads on the fastener, FX , FY and FZ at C2 where C1 and C2 are constants, for R = 0. The change in slope
the start of the test are shown in Fig. 11. The spike in FX (at above and below the load of local plastic deformation highlights
approximately 18 s) was from the formation of the buckle and the differences in crack pattern and fatigue capacity. The rate of
crease followed by the reverse flexure as the profile tried to regain crack growth increases and conversely Ni decreases for load cycles
its original shape during the unloading phase of the cycle. The with Smax above the local plastic deformation load. Also, increasing
buckling of the crests applied a lateral load (FX ) to the screws, R for a given Smax results in an increase in Ni . For example, for load
typically in the order of 100 N. Cracks developed at the crease cycles where Smin is half Smax (i.e. R = 0.5), Ni is approximately an
points that formed by the buckling of the cladding crests. order of magnitude greater than for the same Smax at R = 0.
 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298 3295

Fig. 10. Development of ‘H’ type crease crack from sinusoidal loading; (a) at 50 cycles, (b) 140 cycles, (c) 680 cycles and (d) 1040 cycles.

and propagation. Miner’s rule, which is often used to predict

fatigue performance, was shown to be unsatisfactory for the
complexities of the cladding’s properties changing during the test
coupled with different load cycle magnitudes [8,24].
Using this S–N data, Henderson et al. [25] adopted a modified
version of Miner’s rule to evaluate the potential for cladding
to resist fatigue resulting from loads developed by a simulated
cyclone. Providing the loading history had peaks to induce crease
type cracks, the method was able to indicate potential for low-
cycle fatigue failure from a dynamic wind trace, but not able to
determine when failure would occur [19]. Holmes [26] used a
similar closed form solution to examine high cycle wind induced
fatigue of slender structures.

5. Cladding response to simulated cyclonic wind pressures

Fig. 11. Loads on fastener in X , Y and Z directions at start of test with change in FX
from formation of buckle and crease. Henderson [19] combined pressure coefficient time histories
obtained from the United States’ National Institute of Standards
The changing relationship of S–N is a function of the different and Technology (NIST) building database, described by Ho
fatigue response of the cladding due to its modes of crack initiation et al. [27] on various parts of roofs along with the wind speed and

Fig. 12. Maximum load on fastener vs number of cycles to failure for the previous line-load tests by Mahendran [20] and for PLA air-chamber tests.
3296 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298

cyclone: (a) start (0–45 min), (b) onset of deformation and

cracking, (90–135 min) and (c) at the end (225–270 min) with
cracks up to 20 mm in length. Fig. 14 shows the change in stiffness
and the response of the cladding system to the applied pressure
with the amount of permanent deformation and cracking accrued
by the cladding. The relationship of the reaction coefficient with
respect to pressure applied during each time period was very
similar to the cladding response during the sinusoidal loading
trials.
The spectra of the applied load (i.e. pressure × fastener
tributary area) and that of the fastener response for three time
spans (i.e. early, middle, and late during the ‘‘design’’ cyclone) are
shown in Fig. 15. These spectra indicate that the fastener response
follows the loading up to approximately 4 Hz at the start, 8 Hz at
Fig. 13. Net pressure variations on the windward roof corner (5° pitch, 5 m eaves the middle and 5 Hz at the end. As the mean pressure increased, the
height) with the passage of the ‘‘design’’ cyclone. fluctuating component of pressure also increased, as shown by the
increasing spectral density, along with an increase in the frequency
directions of a design cyclone described by Jancauskas et al. [28] to match between applied pressure and reaction. That is, the middle
produce a series of 5 h pressure traces. The PLA and air-chamber interval (90–135 min) had the larger mean and peak reactions
system was used to apply these dynamic pressure traces as a with the cladding subjected to membrane action increasing the
simulation of wind loads on cladding specimens. cladding system’s stiffness and keeping the cladding in contact
with the underside of the screw, longer than for the other two time
Fig. 13 shows such a simulated pressure trace that was applied
intervals. This pattern can also be observed by the changes in CR .
to a cladding specimen on the highly loaded corner region of a 5°
The relationship of FX , FY and FZ and the movement of the
roof slope, 5 m eaves height building with a dominant opening
screws were similar for those measured during the sinusoidal
occurring on the windward wall approximately 2.5 h into the
load trials. That is, the proportion of the loads in the lateral to
design cyclone, as described by Henderson [19]. The mean wind
vertical directions at the onset of buckling was similar to the crease
speed increases from 33 to 46 m/s and drops back to 33 m/s over
cracking during the cyclic load tests. Failures of the cladding at a
the 5 h duration.
fastener adjacent to the screw attached to the load cell also resulted
A dominant opening can significantly increase the internal in similar vertical and lateral reactions ratios to the sinusoidal load
pressure and also the net pressure for certain approach wind tests.
directions. For the building orientation considered, the net The response of the cladding system to a cyclone with a higher
pressures across the cladding are calculated for the scenario intensity (i.e. the maximum mean wind speed was increased from
that a dominant opening will be created on a windward wall, 46 to 50 m/s at 10 m height) resulted in the pressure trace shown in
near the peak of the cyclone. This is analogous to a windward Fig. 16. An increase in the mean wind speed increases the number
window breaking or a door blowing inwards at approximately the of cycles that will occur for the given time interval.
2.5 h mark. The internal pressures are taken to be equal to the Slight flattening of the crests under the fixings was observed
external pressure fluctuations at the opening. This is a reasonable at approximately 45 min into the trace. At approximately 75 min,
assumption based on the analysis by Ginger et al. [29]. there was a sudden severe distortion of the cladding with the
When subjected to these cyclonic wind loads, the cladding and fixed crests flattened as the valleys deflected up to the level of the
fastener responded in a manner similar to the sinusoidal loads. The fastener heads at a pressure of −5.8 kPa, in a similar manner to
response of the fastener to the applied pressure and increasing the static load tests. Buckling of cladding (creases) formed adjacent
damage of the cladding (crack initiation and propagation) were to the fastener heads and cracks were observed seven minutes
representative of the cladding being subjected to a range of later. The fatigued cladding failed by pulling over the screws at
sinusoidal loads. Fig. 14 shows the reaction coefficient (CR ) for approximately 123 min at a pressure of −6.6 kPa, shown in Fig. 16.
the applied pressures over three time periods during the ‘‘design’’ The crease type crack pattern shapes of ‘H’ and ‘T’ at failure shown

Fig. 14. Reaction coefficient (CR ) over three time spans of a design cyclone.
 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298 3297

Fig. 15. Spectral density of equivalent applied load (pressure × tributary area) and load response at fastener during early, middle and late time spans of the ‘‘design’’ cyclone.

steadily increasing ‘‘static’’ step pressures, cyclic pressures and

fluctuating ‘‘real’’ wind pressures, using a PLA.
This study has demonstrated through the application of cyclic
pressures from the PLA that the corrugated cladding response
from low-cycle fatigue is within the variability of previous line-
load tests. This validates the basis for the current cladding test
standards which were derived using line-load tests. The initiation
and propagation of cracks, the type of cracks and the number of
cycles to failure, were similar for a given fluctuating load in each
test method (i.e. the line-load and the PLA).
The tests using the PLA were also able to establish that the peak
load during a cycle and its amplitude mainly govern crack initiation
and growth, more than the cycling rate and form as indicated by
similar fatigue cracking modes during sinusoidal and realistic wind
loads. Loads measured at a cladding fastener using a three axis load
cell showed that the lateral loads could be a significant proportion
of the uplift load, during various scenarios, such as: the initial
Fig. 16. Applied pressure trace showing failure at −6.6 kPa at 123 min.
buckling of the cladding crest at the screw, unequal crack growth
and with failure of an adjacent fastener.
in Fig. 17 (a) and (b), were similar to failures of cladding during The number of cycles to failure depends on the load on the
cyclone Tracy, documented by Beck [30]. The cladding had resisted fasteners and its range, with the numbers of cycles to failure
a pressure of −8.3 kPa nine minutes prior to the failure at −6.6 kPa significantly reducing for cycle loads exceeding the local plastic
highlighting the nature of fatigue damage accumulation (crack deformation load of the cladding. This characteristic is also seen
growth) in that failure during wind loading is not necessarily in the cladding response to the applied cyclonic wind loads.
associated with the largest peak but is also a function of load This study has shown that the spatial variation of wind loads
history as was the case in sinusoidal load tests. has very little effect on the reactions on a fastener for corrugated
cladding, with most of the reaction being generated by the load
acting on the elongated tributary area given by the batten spacing
6. Conclusions × width of the corrugation between fasteners. Notwithstanding
the temporally and spatially varying wind pressure across the
Typical corrugated pierced fixed roof cladding specimens were building envelope, typical product testing can be carried out
subjected to free air pressure load tests. The loading included satisfactorily by applying uniform loads on the cladding specimens.

Fig. 17. Test roofing specimen ‘‘failed’’ by cladding cracking and pulling over head of screw with (a) ‘H’ crack pattern and (b) ‘T’ crack pattern.
3298 D.J. Henderson, J.D. Ginger / Engineering Structures 33 (2011) 3290–3298

The application of fluctuating pressures representing realistic [11] Mahendran M. Simulation of cyclonic wind forces on roof claddings by random
wind loads using the PLA enabled the assessment of cladding block load testing. Cyclone testing station technical report 38. Townsville:
James Cook University, JCU; 1993.
response to these dynamic pressures and a comparison to [12] Henderson D, Reardon G, Ginger J. Performance of light gauge metal
sinusoidal load tests. Cladding crack patterns generated with the roof cladding subjected to cyclonic wind loading. In: 9th Australian wind
simulated wind trace from the PLA were similar in shape and engineering workshop. 2001.
[13] Seccombe C, Carne S, Martin B. Improvements in system evaluation under
length to crack patterns reported in damage surveys. As well, static and cyclic wind loads. In: 1996 National engineering conference.
the loads at which the claddings deformed and buckled were Darwin: I.E. Aust; 1996.
of similar magnitude to those in the cyclic load tests. From [14] Xu Y. Behaviour of different profiled roofing sheets subject to wind uplift.
Cyclone testing station technical report 37. Townsville: James Cook University,
monitoring the reaction at a screw, the screw was receiving the JCU; 1992.
pressure fluctuations up to a frequency of approximately 5 Hz, [15] Standards Australia. AS4040.3 Methods of testing sheet roof and wall cladding
demonstrating that the cladding transfers the wind fluctuations to – Resistance to wind pressures for cyclone regions. Sydney, (NSW): Standards
Australia; 1992.
the screws and, therefore, justifying the application of cyclic load
[16] Australian Building Codes Board. Building Code of Australia. Canberra,
tests for cladding, fixings and supports (e.g. roof battens). Australia; 2006.
[17] Kopp GA, Morrison MJ, Gavanski E, Henderson D, Hong HP. Three Little Pigs
project: hurricane risk mitigation by integrated wind tunnel and full-scale
Acknowledgments laboratory tests. ASCE Nat Hazards Rev 2010;11:11.
[18] Morrison MJ, Kopp GA. Performance of toe-nail connections under realistic
This research was conducted at James Cook University (JCU). wind loading. Eng Struct 2011;33:69–76.
[19] Henderson DJ. Response of pierced fixed metal roof cladding to fluctuating
D.J. Henderson and J.D. Ginger gratefully acknowledge the support wind loads. Ph.D. thesis. Townsville (Australia) James Cook University;
of an Australian Research Council Linkage Grant and Stramit 2010.
Building Products as the industry partner. The PLA was provided [20] Mahendran M. Fatigue behaviour of corrugated roofing under cyclic wind
loading. Cyclone testing station technical report 35. Townsville: James Cook
on loan to JCU by the University of Western Ontario. University, JCU; 1989.
[21] Surry D, Sinno RR, Nail B, Ho TCE, Farquhar S, Kopp GA. Structurally effective
static wind loads for roof panels. J Struct Eng ASCE 2007;133:871–85.
References
[22] Xu YL, Reardon GF. Test of screw fastened profiled roofing sheets subject to
simulated wind uplift. Eng Struct 1993;15:423–30.
[1] Walker G. Report on cyclone Tracy – effect on buildings – Dec 1974. Australian [23] Standards Australia. AS/NZS1170.2:2002 Structural design actions part
Dept of Housing and Construction; 1975. 2: wind actions. AS/NZS11702:2002. Sydney (NSW, Australia): Standards
[2] Ginger JD, Henderson DJ, Leitch CJ, Boughton GN. Tropical cyclone Larry: Australia; 2002.
estimation of wind field and assessment of building damage. Aust J Struct Eng [24] Mahendran M. Towards an appropriate fatigue loading sequence for roof
2007;7:209–24. claddings in cyclone prone areas. Eng Struct 1995;17:476–84.
[3] Beck V, Stevens L. Wind loading failures of corrugated roof cladding. Civil [25] Henderson DJ, Ginger JD, Morrison MJ, Kopp GA. Simulated tropical cyclonic
Engineering Transactions IE Aust; 1979. winds for low cycle fatigue loading of steel roofing. Wind Struct 2009;12:
[4] Mahendran M. Fatigue behaviour of corrugated roofing under cyclic wind 381–98.
loading. Civil Engineering Transactions IE Aust; 1990. [26] Holmes JD. Fatigue life under along-wind loading – closed-form solutions. Eng
[5] Xu Y. Wind induced fatigue loading on roof cladding of low-rise buildings. Struct 2002;24:109–14.
Cyclone testing station technical report 41. Townsville: James Cook University; [27] Ho TCE, Surry D, Morish D, Kopp GA. The UWO contribution to the NIST
1993. aerodynamic database for wind loads on low buildings: part 1. Archiving
[6] Mahendran M. Behaviour and design of crest-fixed profiled steel roof format and basic aerodynamic data. J Wind Eng Ind Aerodyn 2005;93:30.
claddings under wind uplift. Eng Struct 1994;16:368–76. [28] Jancauskas ED, Mahendran M, Walker GR. Computer-simulation of the fatigue
[7] Xu YL. Fatigue damage estimation of metal roof cladding subject to wind behaviour of roof cladding during the passage of a tropical cyclone. J Wind Eng
loading. J Wind Eng Ind Aerodyn 1997;72:379–88. Ind Aerodyn 1994;51:215–27.
[8] Xu YL. Fatigue performance of screw-fastened light-gauge-steel roofing [29] Ginger JD, Holmes JD, Kopp GA. Effect of building volume and opening size on
sheets. J Struct Eng ASCE 1995;121:389–98. fluctuating internal pressures. Wind Struct 2008;11:361–76.
[9] Hussain K. Short fatigue crack behaviour and analytical models: a review. Eng [30] Beck V. Appraisal of roof cladding under dynamic wind loading – cyclone Tracy
Fract Mech 1997;58:327–54. Darwin 1974: Appendix 4. In: Walker G, editor. Report on cyclone Tracy –
[10] Li WZ, Siegmund T. An analysis of crack growth in thin-sheet metal via a effect on buildings – Dec 1974. Australian Dept of Housing and Construction;
cohesive zone model. Eng Fract Mech 2002;69:2073–93. 1975.