MONOTONIC AND CYCLIC LOAD TESTING OF PARTIALLY AND

FULLY ANCHORED WOOD-FRAME SHEAR WALLS

Peter Seaders
President
MSS, Inc
Corvallis, OR 97333

Rakesh Gupta{
Professor
Department of Wood Science and Engineering
Oregon State University
Corvallis, OR 97331

Thomas H. Miller*
Associate Professor
School of Civil and Construction Engineering
Oregon State University
Corvallis, OR 97331
(Received December 2008)

Abstract. The objectives of this study were to evaluate the performance of wood-frame shear walls
under monotonic and cyclic loads by: 1) determining variability of shear wall performance; 2) comparing
performance of walls under each loading protocol; 3) evaluating effects of anchorage on wall perfor-
mance; and 4) evaluating performance of walls qualitatively and quantitatively with respect to code-
defined performance measures. Sets of tests consisting of eight partially and two fully anchored walls
were conducted using both the ASTM E564 monotonic protocol and CUREE cyclic-test protocol for
ordinary ground motions for a total of 20 walls. Statistical comparisons of parameter variance and mean
values were made between partially anchored walls tested under different protocols and performance
comparisons were made between partially and fully anchored walls. Cyclic tests on partially anchored
walls generally exhibited a coefficient of variation that was lower than for monotonic tests. Failure mode
of fully anchored walls was different than that for partially anchored walls because hold-downs changed
the load path. Comparison of test results with ASCE 41 m-factors suggests that ductility of partially
anchored walls is below the acceptance criteria for shear walls with structural panel sheathing.
Keywords: Seismic performance, wood-frame, shear wall, cyclic testing, CUREE protocol, monotonic
protocol, prescriptive design, code performance.

INTRODUCTION very significant. In 1998, there were 14 tropical

storms and 10 hurricanes in the US that caused
Historically, light-frame residential and com-
$3.6 billion in damage and 32 deaths. Damage
mercial wood structures have performed quite
to wood-frame construction after the Northridge
well during earthquakes. However, assessments
earthquake dominated in all three basic cate-
performed after several natural disasters dem-
gories of earthquake loss: 1) casualties—24 of
onstrated that damage to residential wood
the 25 fatalities in the Northridge earthquake
structures, including residences designed in ac-
were caused by building damage that occurred
cordance with today’s building codes, can be
in wood-frame structures; 2) property loss—
one-half or more of the $40 billion in property
* Corresponding author: [email protected] damage was associated with wood-frame con-
{ SWST member struction; and 3) functionality—48,000 housing

Wood and Fiber Science, 41(2), 2009, pp. 145–156

# 2009 by the Society of Wood Science and Technology
146 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

units, almost all of them in wood-frame build- Many times, the same walls designed under the
ings, were rendered uninhabitable by the earth- IBC would require hold-downs to be installed to
quake (Seible et al 1999). resist tension forces in the shear-wall chords.
Some of the losses may be from gaps in know- Most of the literature (Pardoen et al 2000; Gatto
ledge that require testing to be more representa- and Uang 2002) describes tests of walls with hold-
tive of conditions in actual construction under downs (fully anchored) to resist tension forces in
more realistic loadings (Zacher 1999). Dolan the wall chords, thus simulating an engineered
(2000) presented a history of the design values design. This project focuses on partially anchored
for shear walls. All shear-wall design values in (no hold-downs), prescriptive shear walls that are
the United States are based on monotonic tests more typical in residential construction.
(ASTM 2000) of 2.4
2.4 m walls. In these
standard tests, not only are the loads not repre- The literature also does not fully address how
sentative of the short duration, random, and re- shear-wall performance is affected by material
versing loads experienced in an earthquake and/or construction variability. The test standards
or hurricane, but the shear walls in the tests for monotonic testing, ASTM E564-00 (ASTM
are not comparable to those used in residential 2000), and cyclic testing, ASTM E2126-02a
and commercial construction. Because design (ASTM 2003), used to establish acceptable design
values are based on static monotonic tests, data values require only two tests unless the peak load
from these tests do not represent a dynamic values are more than 15% different for monotonic
event. The overall goal of this project is to ad- tests or 10% for cyclic tests, in which case a third
dress this problem by testing shear walls under test is required. This standard has been followed
actual earthquake records and compare the in most studies. Most researchers have used only
results with walls tested under standard proto- two or three walls for any given wall treatment.
cols. This article presents the results of the Pardoen et al (2000) tested three walls per config-
monotonic and cyclic testing program. The full uration under the city of Los Angeles–University
results of the dynamic testing conducted can be of California Irvine shear-wall test program. The
found in Seaders (2004) and White (2005). CUREE-Caltech wood-frame project also tested
two shear walls for each configuration (Gatto and
There are currently two available design Uang 2002). Although this may be acceptable in
approaches for a wood-frame residential struc- many instances, it leaves open the question of
ture to resist wind and seismic lateral loads. The how much variability should be expected based
most common is a prescriptive approach gov- on materials and construction. Folz and Filiatrault
erned by the International Residential Code (2001) emphasized the need to quantify the varia-
(IRC) (ICC 2006b). Buildings with structural bility in the response of shear walls under load.
elements that do not meet the prescriptive guide-
lines of the IRC must instead be analyzed under Several researchers have compared the perfor-
the governing engineered code requirements. In mance of wood shear walls under various load-
the Pacific Northwest, this is currently the Inter- ing protocols. Dinehart and Shenton (1998)
national Building Code (IBC) (ICC 2006a). compared the sequential phased displacement
(SPD) protocol with monotonic tests and found
A primary difference between these two the SPD gave a 12% lower ultimate load and
approaches is the IRC assumes shear walls resist much lower (42%) displacement at ultimate
overturning moments based only on the overturn- load. They also found the SPD caused more
ing resistance resulting from dead load and adja- nails to fracture or pull out compared with
cent perpendicular walls. As a result, prescriptive monotonic tests in which nails tended to pull
shear-wall elements (brace panels) are not re- away from the framing with the sheathing, caus-
quired to have hold-downs installed unless ing more framing damage. As a result of a 30%
they are between 810 and 1220 mm wide. reduction in load between the first and fourth
 Seaders et al—MONOTONIC AND CYCLIC TESTING OF WOOD-FRAME SHEAR WALLS 147

cycles of repeated cycles with equal peak dis- This article presents the first part of a two-phase
placement, they recommended that a 25% re- research project to investigate the performance
duction in allowable unit shear based on of walls under monotonic, cyclic, and various
monotonic tests be adopted. He et al (1998) earthquake loading protocols. The overall
tested five walls using three different cyclic objectives for the project are as follows:
loading protocols (FCC, CEN–long, and CEN–
short). They found the CEN–long protocol gave 1. To examine the behavior of shear walls
results most similar to those under simulated under standard static test (ASTM E564)
earthquake loading but that improvements were and cyclic test (CUREE) protocols for later
still required to properly model response of a comparison with the behavior of shear
wall under actual earthquake loading. Karaca- walls under various actual dynamic loading
beyli and Ceccotti (1998) compared five dif- records; and
ferent cyclic protocols (SPD, CEN–long, CEN– 2. To understand the behavior (load-deflection
short, FCC, and ISO) and compared them response, strength, failure mode, ductility,
with monotonic and pseudodynamic tests. They energy dissipation characteristics, and so on)
found that different protocols produced different of shear walls under various actual dynamic
failure modes. Specifically, the SPD and FCC loading records: a) subduction zone, long-
protocols produced nail-fatigue failure from duration earthquakes for Oregon/Washing-
the higher energy demands. When comparing ton; and b) earthquakes, including sequences,
the cyclic protocols with a pseudodynamic test, from specific sites in California.
they concluded the allowable unit shear could
This article presents the results of the monotonic
conservatively be based on monotonic curve
and cyclic testing conducted in Phase I of the
results, contrary to Dinehart and Shenton
project. Specific objectives for this article are to:
(1998). Most recently, Cobeen et al (2004) con-
cluded there was no evidence to support a re-
1. Estimate the variability of shear-wall perfor-
duction in allowable unit shear values based on
mance under monotonic and cyclic tests;
the results of the CUREE Caltech Wood Frame
2. Evaluate the effects of anchorage on wall
Project (Seible et al 1999).
performance;
3. Compare the performance of walls under
The limitation of these studies (Dinehart and
monotonic and cyclic loading protocols; and
Shenton 1998; He et al 1998; Karacabeyli and
4. Evaluate the performance of the walls quali-
Ceccotti 1998) for this project is that they did
tatively and quantitatively with respect to
not incorporate the CUREE cyclic protocol
code-defined performance measures.
(Krawinkler et al 2001), a widely used standard
for cyclic testing. Also, larger sample sizes Results of preliminary earthquake testing in
(greater than 2 – 3 walls per treatment) are Phase I are given in Seaders (2004), and the
needed to establish a difference in average remainder of the earthquake testing and results
values for ultimate load or displacement at ulti- are given in White (2005).
mate load with a meaningful level of con-
fidence. Thus, the question of whether the
MATERIALS AND METHODS
monotonic curve can be used to conservatively
establish allowable unit shear values or whether Load Frame and Test Equipment
the cyclic backbone curve provides a better ap-
All tests were conducted at Oregon State Uni-
proximation of shear-wall earthquake response
versity. The loading frame used for monotonic
remains unanswered. The project investigators
and cyclic testing is shown in Fig 1.
hope to contribute to this question by comparing
earthquake shake-table testing with monotonic Specimens were bolted to a fabricated steel
and cyclic testing conducted in this project. beam solidly attached to the strong floor to
148 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

brace-panel construction. Specimens were

2440
2440 mm as shown in Fig 2. Walls were
constructed using 38
89 mm #2 & Better
kiln-dry Douglas-fir framing. Studs spaced at
610 mm were nailed to a top plate and sill plate
with 2 – 16d (3.3
83 mm) nails per stud, and a
second top plate was nailed to the first using
1 – 16d nail at 610 mm on center. Two 32/16
APA rated 1220
2440
11 mm structural
OSB panels were installed vertically with nail-
ing as shown in Fig 2. All nails were full round-
head, smooth-shank, strip-cartridge SENCO
nails driven using a SENCO SN 65 pneumati-
cally driven framing nailer. For a test specimen
with the most realistic shear-wall performance,
12 mm regular gypsum wallboard (GWB) was
installed opposite the structural panel sheathing.
Figure 1. Monotonic and cyclic testing frame. International Residential Code (ICC 2006b)
brace panel construction using structural panel
simulate a fixed foundation. Specimens were sheathing (Method 3) requires 12 mm anchor
loaded using a 490 kN servo-controlled hydrau- bolts at least every 1800 mm, but does not re-
lic actuator with a 250 mm total stroke. The quire hold-downs. Thus, for the basic test speci-
hydraulic actuator was attached to the strong men, 12 mm A307 anchor bolts were installed
wall and supported by a 100 mm hydraulic at 300 mm from each end of the wall. Walls
cylinder. The cylinder was charged with an oil- having only anchor bolts installed without
over-air accumulator with a pressure of approxi- hold-downs are referred to as partially an-
mately 690 kPa. This allowed the actuator to chored. A modified wall design was included in
raise and lower freely during the test without the test matrix to account for the effects of full
creating additional vertical loading on the wall. anchorage and for comparison with strength
A steel C-channel was attached to the load cell values published in the literature. The modified
and actuator and was laterally braced to the design was identical to the basic design except
strong wall. The C-channel was connected to that SIMPSON Strong-Tie PHD-2A hold-
the top plate of the wall using four evenly spaced downs were installed at the ends of the wall.
12 mm A307 bolts installed through both top
plate members; 13 mm holes were drilled in the
top plates after the walls were positioned, ensur-
ing a tight, nonslip bolted connection. The cyclic
driving function was generated by an Analogic
2020 Polynomial Waveform Synthesizer. Data
were collected using a personal computer with
an AMD 550 MHz processor running National
Instruments LabVIEW 6.1.

Wall Specimens
Wall specimens were designed as Internatio-
nal Residential Code (ICC 2006b) prescribed Figure 2. Test specimen schematic.
 Seaders et al—MONOTONIC AND CYCLIC TESTING OF WOOD-FRAME SHEAR WALLS 149

Hold-downs also necessitated installing an addi- tests was discarded as a result of operator error
tional stud at each end of the wall. Walls with that caused additional vertical loading on the
both anchor bolts and hold-downs installed are wall. Two walls were tested for each of the fully
referred to as fully anchored. anchored monotonic and cyclic tests. Two par-
tially anchored walls were also tested monotoni-
cally with dead load applied. Table 1 shows a
Loading Protocols
complete listing of the testing conducted and the
Monotonic tests were based on the ASTM E564- labeling used for each wall treatment.
00 (ASTM 2000) test protocol, which requires
that ultimate load be reached in less than 5 min.
Partially anchored walls were tested at a loading Data Analysis
rate of 0.5 mm/s and fully anchored walls were
Backbone analysis. An analysis of the back-
tested at 0.76 mm/s. This corresponded to a time
bone curve (envelope curve) provides a useful
to failure of approximately 8 min for partially and
tool for comparing results from monotonic and
12 min for fully anchored walls.
cyclic tests. The backbone curve for cyclic tests
Cyclic tests were conducted according to the is derived from the hysteresis curves by drawing
CUREE protocol for ordinary ground motions a line between the consecutive points of peak
developed as part of the CUREE-Caltech Wood load of each primary cycle. Figure 3 shows an
frame project (Krawinkler et al 2001). The pro- example of points of interest used to derive
tocol consists of primary cycles at increasing backbone parameters. The equivalent energy-
displacement amplitudes followed by 2 – 5 trail- elastic-plastic (EEEP) curve is an elastic-
ing cycles at 75% of the primary cycle ampli- perfectly-plastic curve that is defined by the
tude. Tests were conducted continuously up to initial stiffness (Ge), area under the curve equal
1.5 times the reference displacement, Dref.
Subsequent groups of cycles were separated by Table 1. Test matrix and labeling.
a short 15-s pause to allow the test to be stopped Protocol
in a controlled manner in the event of wall col- Treatment Monotonic (ASTM E564) Cyclic (CUREE)
lapse. All tests were conducted at 0.1 Hz. Partially anchored PA-MT (N = 7) PA-CT (N = 8)
Partially anchored walls were tested to a maxi- Fully anchored FA-MT (N = 2) FA-CT (N = 2)
Partially anchored
mum displacement of 3.0 Dref using 49 fully with dead load DL-MT (N = 2) —
reversed cycles. The reference displacement N = number of tests.
was 19 mm. Reference displacement for fully
anchored walls was 60 mm, and tests were lim-
ited to 2.0 Dref by the stroke length of the hy-
draulic actuator (130 mm). Fully anchored
walls were tested to a maximum displacement
of 2.0 Dref using 43 fully reversed cycles.

Test Matrix
Two wall treatments (fully and partially an-
chored) were tested with two different loading
conditions (monotonic and cyclic). Eight mono-
tonic tests and eight cyclic tests were conducted
on partially anchored walls to provide large
enough sample sizes to estimate variability. How- Figure 3. Explanation of parameters from backbone
ever, one of the partially anchored monotonic analysis.
150 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

to the energy dissipated (E), and the calculated procedures for deformation-controlled (ie duc-
yield load (Pyield) as defined in ASTM E2126- tile) structures as:
02a (ASTM 2003) in which:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi mQCE  QUD ð2Þ

2E
Pyield ¼
failure  ð
failure Þ2   Ge ð1Þ where m is the component-demand modifier to
Ge account for expected ductility associated with
with Dfailure as the deflection at failure and the the selected structural performance level, QCE
initial stiffness (Ge) as the secant stiffness at is the expected strength of the component or
40% of peak load. element at the deformation level under consid-
eration, QUD is the total load resulting from
Qualitative analysis. Failures in the wall earthquake and gravity forces, and k is the
were primarily at connections. Thus, failure knowledge factor to account for uncertainty in
modes observed were classified as illustrated in strength of existing structures.
Fig 4: edge breakout, nail pull-through, nail
The m-factor for linear analysis procedures is
withdrawal, and sill-plate splitting.
determined with an idealized load-displacement
Gypsum wallboard screws also exhibited a brit- curve superposed on actual test data. The idea-
tle-fracture failure mode not illustrated in Fig 4. lized load-displacement curve is found by draw-
Three possible failure modes existed for the ing the elastic portion of the curve through the
GWB screws: edge breakout, localized gypsum point at 60% of peak load and then drawing the
crushing (similar to nail pull-through), and brit- remaining linear segments to minimize (visual
tle fracture. At the end of each test, overall approximation) the area between the idealized
condition of the test specimen was recorded curve and actual load-displacement data to cre-
along with the locations and types of connection ate equal areas under idealized and actual load-
failures and condition of sheathing and framing displacement curves.
members.
The acceptance criteria m-factor for a target
m-Factor analysis. ASCE 41 (ASCE 2007) performance level is calculated as 0.75 times
defines a ductility parameter for each type of the ratio of the drift for the desired performance
structural component called an m-factor. The level to the drift at the yield point on the idea-
m-factor is somewhat related to the R-factor lized curve. For example, the drift that defines
used in the IBC, except it applies to individual the m-factor for collapse prevention (CP) corre-
elements instead of an entire system, and it sponds to the failure point on the idealized
modifies the load-resistance balance by increas- curve. Thus, the m-factor would be calculated
ing the structural element strength instead as 0.75 times the ratio of the drift at the failure
of decreasing the applied loads. ASCE 41 def- point to the drift at yield. Life safety (LS) drift
ines the acceptance criteria for linear analysis is taken as 75% of the CP drift, and immediate

Figure 4. Failure modes observed.

 Seaders et al—MONOTONIC AND CYCLIC TESTING OF WOOD-FRAME SHEAR WALLS 151

occupancy drift is taken as 67% of the LS drift. be the result of the nature of the cyclic test pro-
Derivation of these parameters is illustrated in tocol with trailing cycles at 75% of the peak
Figs 2 – 3 of ASCE 41 (ASCE 2007). displacement of the preceding primary cycle.
The incremental loading and trailing cycles of
the CUREE protocol effectively permit the wall
RESULTS AND DISCUSSION
to “relax” after it has experienced damage. This
Variability Analysis and Protocol relaxation allows the localized internal stresses
Comparisons surrounding the nails and other connections to
be relieved before localized failure occurs.
Table 2 shows a comparison of the average
Therefore, in the subsequent loading cycle,
values for the monotonic and cyclic testing and
stresses can be redistributed more equally
p values for the variance tests (F-test) and mean
throughout the wall connections. The monoton-
tests (t-tests) performed on the data. The p value
ic loading protocol does not have this benefit
listed indicates the probability that the null
because the unidirectional gradually increasing
hypothesis, H0, should be accepted (H0 is the
loading causes localized internal stresses to
prediction that the variances or mean values are
build up until an individual element fails. When
equal). The p values were calculated for both the
a connection fails, stresses are immediately
mean test (t-test) assuming equal variance and
redistributed to the others. This sudden redistri-
the mean test (t-test) assuming unequal variance.
bution may have a chain reaction effect because
The p value from the appropriate mean test was
other connections already near failure are im-
used based on the results of the variance test (F-
mediately overstressed. As a result, a sudden
test) at a level of significance of 0.05.
decrease in load-carrying capacity is observed
As shown in Table 2, COV of each parameter, in the load-deflection curve (Fig 5) in monoton-
except Dpeak (displacement at peak load) and ic tests. It should also be noted that results may
upeak (uplift between foundation and stud at end or may not be the same for fully anchored walls
of wall at peak load), was lower for the cyclic because more connections are engaged during
tests than for monotonic tests. An F-test for loading. Behavior of partially anchored walls is
equal variance was performed on each backbone controlled by the sheathing-to-sill-plate nail
parameter. Variance for E, De, Dyield, and Ge connections. Because of the small number of
showed a difference at a level of significance of these connections, one failure causes a signifi-
0.05 (shown in bold). Although not statistically cant increase in load on the remaining fasteners.
significant, COV for Ppeak was lower for cyclic In contrast, the fully anchored wall engages
tests compared with monotonic tests. This may nearly all of the fasteners on the edges of the

Table 2. Statistical hypothesis testing for partially anchored monotonic and cyclic tests.
p values
Monotonic (N = 7) Cyclic (N = 8) t-test mean test (H0: m1 = m2)
Ratio cyclic/monotonic
Parameter Units Mean m1 COV Mean m2 COV F-test variance test Equal variance Unequal variance (m2/m1)

Ppeak kN 9.65 14.9% 8.58 9.6% 0.17 0.048 0.058 0.89

Dpeak mm 23.4 15.6% 20.8 17.2% 0.951 0.099 0.099 0.89
upeak mm 16.9 18.5% 18.1 30.7% 0.184 0.315 0.309 1.07
E J 238 25.1% 183 13.8% 0.04 0.017 0.027 0.77
De mm 3.2 39.3% 2.9 17.0% 0.028 0.295 0.307 0.92
Pyield kN 8.27 15.6% 7.01 9.9% 0.127 0.016 0.023 0.85
Dyield mm 6.9 38.6% 6 16.9% 0.023 0.206 0.223 0.88
Dfailure mm 32 18.4% 29.1 10.4% 0.105 0.121 0.135 0.91
Ge kN/mm 1.33 32.7% 1.21 13.4% 0.02 0.246 0.261 0.91
Note: Performance parameters for monotonic tests were calculated directly from the load-deflection curves; parameters for cyclic tests from the backbone curve.
Bold values indicate significance at the 0.05 level.
152 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

(Dpeak) found in this investigation and 42% re-

duction found by Dinehart and Shenton (1998)
may be from differences between the CUREE
and SPD loading protocols. Langlois et al
(2004) concluded the performances in monoton-
ic and cyclic tests were equal; however, all of
their tests were on fully anchored walls. Com-
parison with dynamic tests must be made to
investigate how well these tests represent the
performance of walls under actual dynamic
loading. One possible reason the values of Ppeak,
Figure 5. Backbone curves of partially anchored mono- E, and Pyield may be lower under cyclic com-
tonic and cyclic tests. pared with monotonic tests is that the energy
demand of cyclic tests is higher. Discussion of
sheathing panels. Thus, because of the large Dinehart and Shenton (1998) by Karacabeyli
number of fasteners carrying load, failure of a et al (1999) suggested that this may be the rea-
single connection has a much smaller impact on son that the SPD protocol produced lower Ppeak
the overall performance of the wall. values. The same may be true in this case be-
cause the total energy dissipated (area enclosed
Because Dinehart and Shenton (1998) found by all hysteresis loops) in the cyclic tests to
peak load and displacement at peak load for Dfailure is approximately 1200 – 1300 J com-
cyclic tests to be equal or smaller than the pared with 238 J in the monotonic tests.
values obtained from monotonic tests, compar-
isons were made using a one-tailed t-test at a One observation from Fig 5 is that positive and
level of significance of 0.05. The t-test, assum- negative backbone curves for the cyclic tests are
ing equal variance, was used to calculate asymmetrical. The first explanation is when the
p values unless the F-test (for equal variance) wall experiences damage in one direction, its
indicated that the variance was different at a load-carrying capacity is reduced slightly as it is
level of significance of 0.05 (ie F-test p value racked in the opposite direction. The second
less than or equal to 0.05). In Table 2, p values possibility is that some additional vertical load-
for both tests are listed, but the bold values were ing was caused by the test equipment if the rate
used for comparisons. Average values of Ppeak, of uplift was faster than the hydraulic support
E, and Pyield were different at a level of signifi- cylinder (Fig 1) could compensate. The hydrau-
cance of 0.05. Average value of Ppeak from the lic cylinder that supported the hydraulic actuator
cyclic tests was 11% lower than from monoton- (Fig 1) allowed it to freely move vertically;
ic tests. Average value of E from the cyclic tests however, the rate at which it was able to move
was 23% lower than from monotonic tests, and was limited by the flow characteristics of
Pyield was 15% lower from the monotonic to the supply hose and accumulator. Thus, as
cyclic tests. These trends are also graphically long as the rate of uplift was low, there was
shown in the representative monotonic and no significant vertical loading on the wall.
cyclic curves in Fig 5. This is similar to what Figure 6 shows a comparison of the monotonic
Dinehart and Shenton (1998) found in compar- and cyclic curves for fully anchored walls.
ing the SPD cyclic loading protocol with static The hysteresis curve for a fully anchored wall
testing. They found a 12% reduction in ultimate also shows some asymmetry between the posi-
load-carrying capacity and a 42% reduction in tive and negative backbone curves. However,
displacement at peak load between monotonic it is unlikely that test equipment caused signifi-
and cyclic tests. The difference between the cant effects on the fully anchored tests because
11% reduction in displacement at peak load uplift was very low.
 Seaders et al—MONOTONIC AND CYCLIC TESTING OF WOOD-FRAME SHEAR WALLS 153

Anchorage Effects energy dissipation parameter, which increased

by factors of 8.66 and 6.39 for monotonic and
The addition of hold-downs to the wall pro-
cyclic tests, respectively. The value of upeak was
duced a dramatic change in overall behavior
the only parameter to show a decrease between
and performance of the shear wall. Relevant
the partially and fully anchored walls, as
backbone parameters from the fully anchored
expected.
monotonic and cyclic tests are shown in Table 3.
Average value of uplift displacement at peak
By comparing Tables 2 and 3, it is apparent the load (upeak) decreased from 16.9 – 10.4 mm and
addition of hold-downs produced a large in- 18.1 – 6.6 mm for monotonic and cyclic tests,
crease in load-carrying capacity, deformation respectively. Thus, installation of hold-downs
capacity, and energy dissipation characteristics predictably reduces the uplift at peak load. For
of the shear-wall specimens. Average value both fully and partially anchored walls, uplift is
of Ppeak increased from 9.65 – 24.34 kN and proportional to the lateral displacement (Dpeak).
8.58 – 22.47 kN for monotonic and cyclic tests, Thus, if “uplift” is evaluated as a ratio of uplift
respectively. This represents an approximate to lateral displacement, this difference is even
2.5-factor increase between fully and partially more significant. Average uplift rate for partially
anchored walls. Similarly, values for Dpeak anchored walls was 0.72 mm/mm for vertical/
and Pyield increased by factors ranging from lateral displacement and 0.87 mm/mm for
2.1 – 2.8. The most dramatic increase between monotonic and cyclic tests, respectively. Aver-
partially and fully anchored walls was in the age uplift rate for fully anchored walls was 0.16
and 0.15 mm/mm for monotonic and cyclic tests,
respectively. This corresponds to a decrease in
uplift rate between the partially and fully an-
chored walls by approximately a factor of 5.
Although no statistical comparisons can be made
between tests because there are only two fully
anchored walls for each wall treatment, trends
similar to the partially anchored tests can be seen
in the data. Average value of Ppeak for the cyclic
tests was 7.7% lower than for the monotonic tests.
Surprisingly, the differences between many of the
other backbone parameters are much greater.
Average value for Dpeak is 33% lower for cyclic
Figure 6. Backbone curves of fully anchored monotonic than monotonic tests. This is much larger than
and cyclic tests. the 11% difference observed with the partially

Table 3. Fully anchored monotonic and cyclic backbone parameters.

Monotonic Cyclic
Test# FA-MT-1 FA-MT-2 Mean FA-CT-1 FA-CT-2 Mean Ratio cyclic/monotonic

Ppeak kN 22.75 25.92 24.34 22.81 22.13 22.47 0.923

Dpeak mm 58.5 74 66.3 47.9 41.1 44.5 0.671
upeak mm 8.4 12.4 10.4 7.1 6.2 6.6 0.635
E J 1792 2333 2062 1331 1010 1170 0.567
De mm 2.9 4 3.5 4 5.1 4.5 1.286
Pyield kN 20.39 22.98 21.68 19.58 19.01 19.29 0.890
Dyield mm 6.6 9 7.8 8.6 10.9 9.7 1.244
Dfailure mm 91.2 106 98.6 71.1 58.5 64.8 0.657
Ge kN/mm 3.09 2.56 2.83 2.28 1.75 2.02 0.714
154 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

anchored walls and is very similar to results Fully anchored walls exhibited a much different
found by Dinehart and Shenton (1998). Similarly, behavior. Here, uplift forces are transferred
there was a large decrease in E of 43% between from the sheathing into the studs and then
the monotonic and cyclic tests. directly into the foundation through the hold-
down and anchor bolt. Thus, the sheathing-to-
sill-plate connections carry an equal portion of
Effect of Dead Load Application the shear loading as other sheathing to stud con-
nections in the wall and are not required to carry
Table 4 shows a comparison of the partially
uplift forces. This allows the sheathing connec-
anchored walls tested monotonically with dead
tions to resist shear forces as intended and the
load (PDL) applied to the average values
GWB connections to act as a group. GWB con-
obtained from the partially and fully anchored
nections, individually very weak, can have
monotonic tests. As expected, the partially an-
appreciable strength when combined.
chored walls with dead load had higher Ppeak
(41 – 88%) and Dpeak (22 – 74%) values than Thus, there are noticeably different failure
similar walls without dead load applied. Greater modes for fully anchored walls subjected to
increases resulted from larger applied dead each test protocol. The monotonic test protocol
loads with the fully anchored configuration primarily exhibited nail pull-through and some
serving as an upper bound for these increases. edge breakout in the nailed sheathing connec-
tions. GWB connections exhibited crushing
around the screws with little damage to the
Failure Mode Comparisons screws themselves. Failure modes observed in
cyclic testing exhibited more nail withdrawal
Failure modes of partially anchored walls were
than monotonic tests and occasional fractures
almost identical for both test protocols. In every
in GWB screws.
case, failures were confined to the connections
between the sill plate and bottom edge of OSB Before each test, careful inspection of each wall
and GWB panels. In several cases, sill-plate was made and each fastener that was overdriven
splitting occurred; however, in most cases, nail or too close to the panel edge was noted in the
and screw connections failed first. The failure pretest report. Comparing posttest damage with
mode was so consistent for partially anchored pretest inspections, it was apparent that fastener
walls with different test protocols because edge distance controlled whether a connection
the connections along the bottom of the wall would fail by edge breakout or a combination of
had a much higher demand placed on them withdrawal and pull-through. Because of limit-
than any of the other connections. The higher ed space requirements at adjoining panel edges
demand is because all uplift forces must be and difficulty in placing nails accurately with a
transferred through these connections into the pneumatic nailer, it was not uncommon to have
sill plate and then through the anchor bolt several fasteners with less than the required
into the foundation. 9.5 mm edge distance. Sheathing connections
typically failed by edge breakout. Thus, special
Table 4. Comparison of monotonic tests with dead load attention should be given to shear walls when
applied. adjoining sheathing panel edges are joined on
Anchorage PA DL DL FA 38 mm thick framing members.
No. of tests Units 7 1 1 2

PDL kN — 10.7 17.8 —

Ppeak kN 9.65 13.62 18.11 24.34 Code Comparisons
Dpeak mm 23.4 28.5 40.7 66.3
E J 238 476 778 2062 An m-factor analysis was performed for each
Ge kN/mm 1.33 1.11 1.10 2.83 monotonic and cyclic test. Average values of
Note: PDL is the total applied dead load the m-factor from testing of partially and fully
 Seaders et al—MONOTONIC AND CYCLIC TESTING OF WOOD-FRAME SHEAR WALLS 155

Table 5. m-Factors for monotonic and cyclic tests. at 40% of peak load (De), yield displace-
Acceptance criteria
(ASCE 41 Table 8-3)
N=7
PA-MT
N=8
PA-CT
N=2
FA-MT
N=2
FA-CT
ment from EEEP curve (Dyield), and initial
stiffness (Ge) were significantly different.
IO 1.7 1.21 1.19 2.28 1.58
LS 3.8 1.80 1.78 3.41 2.36
The lower variances may be the result of
CP 4.5 2.40 2.37 4.54 3.15 a redistribution of loads during the cyclic
IO, immediate occupancy; LS, life safety; CP, collapse prevention. protocol that results in more consistent
properties related to failure such as peak
anchored walls are shown in Table 5. Fully an- load and energy dissipation than for mono-
chored monotonic tests had an average m-factor tonic tests.
(ductility) greater than that provided for wood 2. Comparisons of average values between
shear walls with wood structural panel sheath- monotonic and cyclic tests of partially an-
ing in ASCE 41 (ASCE 2007) for linear proce- chored walls using a one-tailed t-test show
dures at the immediate occupancy and collapse that backbone parameters for peak load,
prevention performance levels. Partially an- yield load, and energy dissipation are signifi-
chored walls clearly did not meet the accep- cantly different at an alpha level of 0.05.
tance criteria. 3. Performance parameters for fully anchored
walls exhibited increases over partially an-
From Table 5, it appears that there is almost no chored walls by a factor of about 2.5 for peak
difference between the ductility of partially an- load and displacement at peak load and a
chored walls tested using the monotonic and factor of almost 9 for energy dissipation.
cyclic protocols. The monotonic tests of fully Other backbone parameters also exhibited
anchored walls, however, appear to have a no- increases.
ticeably higher ductility than cyclic tests. This 4. Failure mode of fully anchored walls was
suggests the acceptance criteria should be re- different than for partially anchored walls
vised to reflect differences in ductility between because hold-downs changed the load
fully and partially anchored walls and that the path. Partially anchored walls failed only in
acceptance criteria for fully anchored walls the sheathing-to-sill-plate nail connections
may be revised to better reflect shear-wall and in the sill plate itself, irrespective of
performance based on cyclic testing. These con- loading protocol. No other fasteners in
clusions, however, are only based on the prelim- the wall experienced any visible damage or
inary testing in this project and more testing is displacement.
needed to make a final recommendation for 5. Sheathing and gypsum wallboard fasteners
design. Moreover, the following aspects deserve of fully anchored walls experienced different
further study: 1) effects of different cyclic test failure characteristics when subjected to
protocols on these observations/conclusions; 2) various loading protocols. Monotonic tests
interactions between wall aspect ratio and the caused primarily nail pull-through-type fail-
degree of anchorage; and 3) effect of end walls ures in the sheathing connections and crush-
to partially offset the lack of hold-downs in ing of the gypsum in the screwed GWB
partially anchored systems. connections. Fully reversed cycling of
CUREE tests caused some nails to withdraw
and GWB screws to fracture.
CONCLUSIONS
6. Partially anchored walls with dead load ap-
1. Cyclic tests on partially anchored walls plied experienced increases in load-carrying
generally exhibited a coefficient of varia- capacity that were approximately proportion-
tion lower than monotonic tests. Compari- al to the magnitude of the dead load resisting
sons of variance using an F-test at an moment applied. Fully anchored walls repre-
alpha level of 0.05 also indicated that var- sent an upper bound for the performance of
iances of energy dissipation, displacement walls with dead load.
156 WOOD AND FIBER SCIENCE, APRIL 2009, V. 41(2)

7. Comparison of test results with ASCE 41 Folz B, Filiatrault A (2001) Cyclic analysis of wood shear
(ASCE 2007) m-factors shows the ductility walls. J Struct Eng 127(4):433 – 441.
He M, Lam F, Prion GL (1998) Influence of cyclic test
of partially anchored walls is below accep- protocols on performance of wood-based shear walls.
tance criteria for shear walls with structural Can J Civil Eng 25(3):539 – 550.
panel sheathing. Gatto K, Uang CM (2002) Cyclic response of woodframe
shearwalls: Loading protocol and rate of loading rate
effects. CUREE Publication No. 13. Richmond, CA.
ACKNOWLEDGMENTS ICC (2006a) International building code. International
Code Council, Whittier, CA.
This project was supported by the National Res- ——— (2006b) International residential code. International
earch Initiative of the USDA Cooperative State Code Council, Whittier, CA.
Research, Education and Extension Service, Karacabeyli E, Ceccotti A (1998) Nailed wood-frame shear
walls for seismic loads: Test results and design consid-
grant number 2003-35103-12918. We thank erations. Structural Engineering World Wide 1998,
Milo Clauson of Oregon State University for as- ISBN: 0-08-042845-2. Paper reference: T207-6.
sistance in conducting the shear-wall tests and ———, Dolan JD, Ceccotti A, Ni C (1999) Comparison
Simpson Strong-tie for donating hold-downs. of static and dynamic response of timber shear walls.
Discussion. J Struct Eng-ASCE 125(7):796 – 797.
Krawinkler H, Parisi F, Ibarra L, Ayoub A, Medina R (2001)
Development of a testing protocol for woodframe struc-
REFERENCES tures. CUREE Publication No. W-02. Richmond, CA.
Langlois JD, Gupta R, Miller T (2004) Effects of reference
ASCE (2007) Seismic rehabilitation of existing buildings. displacement and damage accumulation in wood shear
ASCE/SEI 41-06. American Society of Civil Engineers, walls. J Struct Eng 130(3):470 – 479.
Reston, VA. Pardoen GC, Kazanjy RP, Freund E, Hamilton CH, Larsen D,
ASTM (2000) Standard method of static load test for shear Shah N, Smith A (2000) Results from the City of Los
resistance of framed walls for buildings. ASTM E Angeles-UC Irvine shear wall test program in Proc World
564-00. American Society of Testing and Materials, Conference on Timber Engineering. Paper 1.1.1 on CD.
West Conshohocken, PA. Seaders PJ (2004) Performance of partially and fully an-
——— (2003) Standard test methods for cyclic (reversed) chored wood frame shear walls under monotonic, cyclic
load test for shear resistance of framed walls for build- & earthquake loads. MS Thesis, Oregon State University,
ings. ASTM E 2126-02a. American Society of Testing Corvallis, OR.
and Materials, West Conshohocken, PA. Seible F, Filiatrault A, Uang C-M (eds.) (1999) Proc Invi-
Cobeen K, Russell J, Dolan DJ (2004) Recommendations tational Workshop on Seismic Testing, Analysis and
for earthquake resistance in the design and construction Design of Woodframe Testing. CUREE Publication
of woodframe buildings. CUREE Publication No. W-30b. No. W-01. Richmond, CA.
Stanford University, Stanford, CA. White KBD (2005) Performance of wood frame shear
Dinehart DW, Shenton HW III (1998) Comparison of static walls under earthquake loads. MS Thesis, Oregon State
and dynamic response of timber shear walls. J Struct Eng University, Corvallis, OR.
124(6):686 – 695. Zacher EG (1999) Gaps in information for determination of
Dolan JD (2000). Code development of seismic design of performance capabilities of light woodframe construc-
woodframe structures: Testing needs. Pages 9 – 14 in tion. Pages 1 – 2 in F Seible, A Filiatrault, and C-M
Proc Invitational Workshop on Seismic Testing, Analysis Uang, eds. Proc Invitational Workshop on Seismic Test-
and Design of Woodframe Testing. CUREE Publication ing, Analysis and Design of Woodframe Construction.
No. W-01. Richmond, CA. CUREE, Richmond, CA.