ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 104-S52

Test of High-Rise Core Wall: Effective Stiffness

for Seismic Analysis
by Perry Adebar, Ahmed M. M. Ibrahim, and Michael Bryson

A test was conducted on a large-scale model of a concrete variation of cracking with increasing lateral load before
shearwall from the core of a high-rise building. The wall was reinforcement yielding.
slender (height-to-length ratio of 7.2), had a flanged cross section, FEMA 3561 recommends a factor of 80% for previously
a low percentage of vertical reinforcement (0.45%), and was uncracked concrete walls and a factor of 50% for previ-
subjected to a constant axial compression of 0.1fc′ Ag to simulate the ously cracked walls. The commentary to the 1995 New
effect of gravity loads. Approximately 70% of the flexural resistance Zealand concrete code2 recommends a factor of 25% for a
of the wall was due to the applied axial compression. The main concrete wall with no axial force and a factor of 35% for a
emphasis of the test was to investigate the influence of cracking on
effective stiffness for seismic analysis. Of particular interest was
concrete wall with an axial compression force equal to 10%
recovery of uncracked-section stiffness due to axial compression of fc′ Ag. Paulay3 recommends an effective stiffness equal to
from gravity loads closing flexural cracks in walls with a low the nominal strength divided by the yield displacement.
percentage of vertical reinforcement. Extensive measurements Before the current study, it was not known which of these
were made of concrete strains over the cracked region of the wall, recommendations were most appropriate for high-rise core-
and these were used to calibrate a nonlinear flexural stiffness model wall buildings.
for high-rise concrete shearwalls. The final failure mode was The current study began by developing a simple
buckling of an unsupported vertical reinforcing bar leading to nonlinear flexural model that could be used to determine
concrete spalling and bar fracture after a few post-buckling cycles. the effective flexural stiffness of high-rise concrete shear-
The maximum global drift of the wall was 2.4%. walls. The model accounts for recovery of uncracked-
section stiffness due to axial compression from gravity
Keywords: cracking; effective stiffness; flexure; seismic analysis; shear- loads closing flexural cracks. Test results were needed to
wall; test. calibrate the model.
Table 1 summarizes examples of previous tests4-13 on
INTRODUCTION concrete shearwalls with height-to-length ratios (hw /lw)
Concrete shearwalls in high-rise buildings are often greater than 2.0. Numerous tests not included in the table
located around the perimeter of elevator and stairway shafts, have been done on squat walls with hw /lw less than or equal
and are connected together to form a rectangular tube called to 2.0. The most slender wall in Table 1 has a hw /lw of 3.4.
a core that extends over the height of the building. When a None of the previously tested walls in Table 1 has all the
core is subjected to concentric lateral loading in a principal characteristics of a high-rise core wall including large hw /lw ,
direction, wall segments parallel to the lateral load primarily significant axial compression, flanged cross section, and low
resist shear, while wall segments perpendicular to the lateral percentage of vertical reinforcement. To calibrate the nonlinear
load act as flanges and primarily resist over-turning bending flexural model, detailed strain readings were needed over the
moments. As cores are usually located near the building cracked region of wall, which is difficult to obtain from
center, and gravity load columns are usually located around previous tests reported in the literature. Thus, a test was
the perimeter, core walls usually support significant gravity conducted on a large-scale extensively instrumented
loads. Due to the combination of axial compression and large concrete wall with the characteristics shown in the last row
flanges increasing over-turning resistance, and higher modes of Table 1.
reducing over-turning demand, high-rise core walls often
require a relatively low percentage of vertical reinforcement. RESEARCH SIGNIFICANCE
An important part of seismic design of high-rise buildings Results are presented from a large-scale test of a very
is estimating building displacements due to design earth- slender wall with a flanged cross section, low percentage of
quake forces. The design displacement is needed to confirm vertical reinforcement, and subjected to significant axial
building drifts are within acceptable limits, to determine compression. While many previous tests have been done on
whether confinement reinforcement is required in plastic concrete shearwalls, none have had this combination of
hinge regions of walls, and to ensure punching shear failures characteristics that are particularly important for high-rise
will not occur in slab-column connections of gravity load core walls. The test results are used to calibrate a nonlinear
frames connected to walls. The design displacement of flexural model14 suitable for push-over analysis, and to calibrate
concrete shearwalls must be determined using a structural recommendations for equivalent stiffness15 suitable for
model that accounts for the effect of concrete cracking. The
effective stiffness used in linear seismic analysis of a ACI Structural Journal, V. 104, No. 5, September-October 2007.
concrete shearwall is normally determined from the MS No. S-2006-106 received March 10, 2006, and reviewed under Institute publication
policies. Copyright © 2007, American Concrete Institute. All rights reserved, including
uncracked section stiffness multiplied by a single factor that the making of copies unless permission is obtained from the copyright proprietors. Pertinent
discussion including author’s closure, if any, will be published in the July-August
accounts for variation of cracking over the height, and 2008 ACI Structural Journal if the discussion is received by March 1, 2008.

ACI Structural Journal/September-October 2007 549

 providing additional vertical wall reinforcement below the
Perry Adebar, FACI, is a Professor of civil engineering at the University of British
Columbia, Vancouver, British Columbia, Canada. He is a member of ACI Committees 341, construction joint at this location (refer to Fig. 1).
Earthquake Resistant Concrete Bridges; 374, Performance-Based Seismic Design of Due to limited height of the laboratory, the wall was
Concrete Buildings; and E803, Faculty Network Coordinating Committee; and Joint
ACI-ASCE Committees 441, Reinforced Concrete Columns, and 445, Shear and Torsion.
constructed and tested in a horizontal position. The direction
His research interests include the seismic design of high-rise shearwall buildings.

ACI member Ahmed M. M. Ibrahim is a Senior Structural Engineer at the California

Department of Transportation (CALTRANS). He received his BSc from Cairo University,
Cairo, Egypt; his MASc from Concordia University, Montreal, Quebec, Canada; and
his PhD from the University of British Columbia. He is a member of ACI Committee 341,
Earthquake Resistant Concrete Bridges. His research interests include seismic design
of concrete structures, thermal analysis of bridges, and joint shear in integral bent cap.

Michael Bryson is a Project Engineer with Nishkian Chamberlain, Los Angeles,

Calif. His research interests include the analysis and design of concrete buildings.

linear dynamic analysis. The test results will also be useful

for other researchers to calibrate their models for high-rise
concrete shearwalls. While emphasis is on effective stiffness,
data is also presented about drift capacity.

EXPERIMENTAL PROGRAM
Description of test specimen
Figure 1 presents a summary of the test specimen, which
was 12.2 m (40 ft) high from the base and 1625 mm (5 ft, 4 in.)
long with a flanged cross section. The web of the wall was
1219 mm (4 ft) long and 127 mm (5 in.) thick, while the
flanges were 203 mm (8 in.) long and 380 mm (15 in.) thick.
Vertical reinforcement in each of the two flanges consisted of
5M to 10M reinforcing bars enclosed by No. 3 hoops spaced
at 64 mm (2.5 in.) in the lower 3 m (10 ft) of wall and spaced
at 152 mm (6 in.) over the remaining height of wall. Clear
cover to the No. 3 hoops was 6 mm (0.25 in.). The web was
reinforced with 10M bars spaced at 305 mm (12 in.) vertically
and horizontally (ρ = 0.27%). The horizontal bars were
hooked around a vertical reinforcing bar in the flanges.
In most high-rise buildings there are large perimeter walls
below grade that are connected to tower walls by diaphragm
action of concrete slabs. As a result, the location of the
maximum bending moment in tower walls is normally at
grade level, and there is no pullout of vertical reinforcement
from the foundation as occurs when the plastic hinge is
located immediately above the foundation. To simulate this
somewhat in the test, the critical section for bending moment Fig. 1—Details of test wall specimen: (a) elevation; and
was located 426 mm (17 in.) up from the wall base by (b) cross section.

Table 1—Examples of previous tests on concrete shearwalls

Reference no. and date Shape* Height, m Aspect ratio hw /lw Reinforcement,† % Axial compression, % fc′ Ag Lateral load‡ No. of specimens
4 (1973) R 3.7 to 6.4 1.9 to 3.4 0.27 to 3.0 5.5 to 7.2 M 6
5 (1975) R, F, B 4.57 2.4 0.50 to 2.27 0.3 to 0.4 M and RC 8
6 (1979) F, B 4.57 2.4 1.27 to 2.27 7.6 to 14.1 RC 6
7 (1981) R 5.49 2.9 1.00 1.0 RC 1
8 (1984) F 4.57 2.4 0.83 to 0.96 0 to 8.9 RC 5
9 (1986) B 4.50 3.2 0.68 4.9 RC 1
10 (1988) R 1.30 2.0 2.62 0 to 20.0 M 6
11 (1990) B 3.56 2.9 1.07 8.0 RC 4§
12 (1995) R, T 3.66 3.0 1.10 to 1.20 7.0 to 10.0 RC 4
13 (1999) R 4.56 2.3 0.67 to 1.54 5.0 to 12.0 RC 6
Current study F 12.0 7.4 0.45 10.0 RC 1
*R = rectangular; F = flanged; B = barbell; C = C-shape; and T = T-shape.
†Total vertical reinforcement as ratio of gross area of wall cross section.
‡
M = monotonic, and RC = reverse cyclic.
§
One solid, three with openings.
Note: 25.4 mm = 1 in.

550 ACI Structural Journal/September-October 2007

that would be vertical if the wall had been constructed in the total axial compression applied to the wall, as well as to
normal position is referred to as vertical throughout this confirm that the compression was uniform.
paper. The flange of the wall was supported on two sliding The top wall displacement was measured by linear variable
bearings located 2.97 m (9 ft, 9 in.) from the wall base and differential transformer (LVDT) Displacement Transducer
two bearings at 9.83 m (31 ft, 3 in.) from the base (refer to DT1 within the hydraulic actuator, and was verified by DT2
Fig. 1). Before casting the wall base, a test was conducted to and DT3. The lateral displacement of the wall at 5.08 m
determine the lateral force required to overcome friction in (200 in.) from the base was measured by DT5. The axial
the Teflon bearings due to the wall dead weight. The lateral displacement of the wall was measured by DT4 at 5.08 m
load required to maintain movement of the wall was 3.65 kN (200 in.) up from the wall base. Displacement Transducers
(0.83 kip), which is 2% of the maximum load applied in the DT6, DT7, and DT8 were used to monitor movement of the
test. The wall base was cast around shear pins inserted into wall base.
the strong-floor and was post-tensioned to the floor to To measure axial strains of the wall at numerous locations
prevent movement; however, as discussed in the following, over the height, metal targets were glued to the wall and a
some movement of the base was measured during the test. digital caliper was used to measure the relative displacement
of the targets. On each side of the wall, 11 targets were
Material properties spaced at approximately 500 mm (20 in.) over the cracked
The wall test was conducted 140 days after the concrete in region of the wall. Details of the target locations near the
the wall above the construction joint was cast from a single construction joint are shown in Detail A of Fig. 2.
batch of ready mixed concrete. At that time, the average Four strain gauges (SG1, SG2, SG3, and SG4) were
cylinder compressive strength of the concrete was 49 MPa attached to 10M vertical reinforcing bars at the four outside
(7.1 ksi). The measured stress-strain responses from three corners of the wall. The strain gauges were located exactly at
cylinders indicate a secant modulus of elasticity (at 40% fc′ ) the construction joint to measure the maximum reinforcing
equal to 30,000 MPa (4350 ksi), while the initial tangent bar strains.
modulus was 30% larger. The strain at peak compression
stress was 0.0022.
All 10M reinforcing bars used to construct the wall came
from a single heat of reinforcing steel. Randomly selected
samples of bars were tested. Based on a nominal cross-
sectional area of 100 mm2 (0.16 in.2), the average yield
strength was determined to be 455 MPa (66.0 ksi), while the
average ultimate strength was found to be 650 MPa (94.3 ksi).
The bare bar strain at onset of strain hardening was 0.009,
while the tangent modulus at that point was 6000 MPa
(870 ksi). The bare bar stain at ultimate strength was 0.08.

Testing procedure
A constant axial load of 10% fc′Ag = 1500 kN (342 kip)
was applied to the wall throughout the test. Two hydraulic
actuators located below the wall base were used to pull on
four Dywidag bars connected to the top of the wall (refer to
Fig. 2). The hydraulic actuators were subjected to a constant
oil pressure during the test (load control), and had sufficient
stroke to allow the required movement of the Dywidag bars
Fig. 2—Location of instrumentation (25.4 mm = 1 in.).
during cycling of the lateral load.
The lateral load was applied to the top of the wall using an
MTS hydraulic actuator with a maximum stroke of ±305 mm
(12 in.). The actuator was operated at all times at a constant
rate of total displacement of 1 mm/second (0.04 in./second).
The lateral load was applied 11.76 m (38 ft, 7 in.) from the
wall base, and 11.33 m (37 ft, 2 in.) from the construction
joint (refer to Fig. 1). Four complete displacement cycles
(from zero to maximum positive to maximum negative to
zero) were performed at each of 13 different displacement
levels (Table 2). The total displacement at the 13 levels
varied from 15 to 300 mm (1/2 to 12 in.), and are discussed
in detail in the Experimental Results section.

Instrumentation
Figure 2 summarizes the instrumentation used to measure
lateral and axial loads, displacements, and strains. Load
Cell LC1 was used to measure the lateral force applied by
the hydraulic actuator. Four load cells (LC2, LC3, LC4, and
LC5) were installed on the Dywidag bars used to apply axial
compression to the wall. These were used to determine the Fig. 3—Photograph of wall during test.

ACI Structural Journal/September-October 2007 551

 All load cells, LVDT displacement transducers, and strain additional level at ±48 mm (1.9 in.) total displacement with
gauges were recorded five times per second throughout the four complete cycles was added.
test. During the first and last cycle of each displacement level, The lateral force applied to the wall was calculated as the
the displacement was held constant at the maximum value difference between the actuator force and the friction force in
while crack patterns and crack widths were documented and the Teflon bearings that supported the wall dead weight. A
relative target displacements were measured (refer to Fig. 3). constant friction force of 3.7 kN (0.83 kip) was assumed to act
opposite the direction of wall movement, that is, the friction
EXPERIMENTAL RESULTS force was assumed to reverse at peak deflections. This was
Corrections confirmed by the relationship between uncorrected lateral
force and deflection, which indicated a reduction of 2 × 3.7 kN
Displacement Transducers DT6, DT7, and DT8 did
(0.83 kip) was required to reverse the wall movement.
measure some rotation of the base during the test. The
measured displacements were used to determine the rigid
body motion of the wall, including the lateral displacement Observations
The relationship between lateral force applied to the wall
component at the top of the wall. Throughout this paper,
and wall displacement (both corrected values) is shown in
measured displacements including rigid body motion of the
Fig. 5. To improve clarity, three separate plots are provided.
wall are referred to as total displacements, while corrected
There was typically a small difference between the first and
displacements with the rigid body rotation eliminated are
second cycles to the same displacement level. There was
referred to as wall displacements. Figure 4 compares these
typically no visible difference between the second, third, and
two displacements.
fourth cycles to the same displacement level. The only
Table 2 summarizes the 13 displacement levels in terms of exception was when the wall was displaced to 2.4% drift when
total displacement and wall displacement. Due to a problem significant degradation occurred during all cycles (Fig. 5(c)).
with the data acquisition system, the data was lost for 30 and There was no concrete cracking observed during the first
45 mm (1.2 and 1.8 in.) total displacement levels. Thus, an three displacement levels. The third level was at a total
displacement of 24 mm (1 in.) and a wall displacement of
Table 2—Test protocol displacement levels 18 mm (0.7 in.). The maximum lateral force applied during
Displacement Total Wall
level* displacement,† mm displacement, mm Wall drift,‡ %
1 ±15 12 –11 0.10 –0.09
2 ±18 14 –12 0.12 –0.10
3 ±24 18 –17 0.15 –0.14
4§ ±30 21 –21 0.18 –0.18
§ ±45 33 –34 0.28 –0.29
5
6 ±48 34 –38 0.29 –0.32
7 ±60 46 –49 0.39 –0.42
8 ±75 60 –63 0.51 –0.54
9 ±90 74 –77 0.63 –0.65
10 ±120 105 –104 0.89 –0.88
11 ±150 132 –138 1.12 –1.17
12 ±200 182 –187 1.55 –1.59
13 ±300 281 –241 2.39 –2.05
*
Four cycles to each displacement level.
†
Total displacement includes rigid body movement of wall and base.
‡
Global drift based on wall height of 11.76 m (12.86 yd).
§Data lost for these cycles
Note: 25.4 mm = 1 in.

Fig. 5—Measured lateral load-wall displacement relationships:

Fig. 4—Relationship between total displacement and wall (a) up to 0.4% drift before yielding; (b) up to 1.1% drift;
displacement (25.4 mm = 1 in.). and (c) all cycles (1 kN = 0.228 kip; 25.4 mm = 1 in.).

552 ACI Structural Journal/September-October 2007

this level was 90 kN (20.5 kip). First cracking occurred when
the lateral force was approximately 100 kN (22.8 kip), and
the wall displacement was approximately 21 mm (0.86 in.).
The data for both the 30 and 45 mm (1.2 and 1.8 in.) total
displacement levels were lost. The dashed line in Fig. 5(a)
shows the approximate envelope of the missing curves.
Significant yielding of the vertical reinforcement occurred
at the construction joint during the first cycle of loading
to 60 mm (2.4 in.) total displacement, which corresponds to
+46 mm (+1.8 in.) wall displacement and 0.4% global drift.
There was typically very little residual displacement when
the lateral force was reduced to zero. Even after the wall was
displaced to over 180 mm (7 in.) top wall displacement
(Fig. 5(c)), which is equal to 1.5% global drift and approxi-
mately four times the yield displacement of the wall, the
residual displacement under zero lateral force was less than
10 mm (0.4 in.). The combination of low percentage of
vertical reinforcement and large axial compression force due
to gravity loads caused the flexural cracks to close when the
lateral load is removed.
Figure 6(a) presents a drawing of the final crack pattern
observed in the wall. The drawing was traced from a
photograph of the cracks. East is the positive displacement
direction, and west is the negative displacement direction.
The portion of wall shown is from the base to 3.6 m (11.8 ft)
up from the base. The first crack to occur was at the construction
joint and is labeled C in Fig. 6(a). Cracks A, B, and D formed
next in that order. A, B, and C are horizontal cracks, while
the cracks labeled D are somewhat inclined. In general, the
crack pattern is not perfectly symmetrical, in part because at
each displacement level, the wall was always displaced first
Fig. 6—Appearance of wall after test: (a) crack pattern; and
to the east and then to the west.
(b) photograph of failure zone near construction joint showing
The next cracks to form were the inclined cracks labeled I buckled vertical bar and spalled concrete.
and J on the west side, and the horizontal cracks labeled F
and G on the east side. Table 3 summarizes the width of all
these cracks, which are the only ones that existed when the Table 3—Measured widths (mm) of
wall was displaced to +46 and –49 mm (1.8 and –1.9 in.). cracks labeled in Fig. 6
Additional cracks occurred as the wall displacement was Crack Wall displacement, east (mm)
increased. The last cracks to form were the vertical cracks in (west face) +46 +74 +105 +132 +182
the lower portion of the web, which did not form until the J 0.3 0.3 0.3 0.3 0.3
wall displacement was approximately ±200 mm (7.9 in.). I 0.15 0.2 0.4 0.4 1
The largest crack throughout the test was located at the H — — 0.25 0.3 0.4
construction joint (refer to Table 3). It is interesting to note G — — 0.25 0.5 1.25
how at a wall displacement of +132 mm (5.2 in.), the width
F — 0.35 0.5 1.25 1.5
of the crack along the west end of the construction joint
E — 0.5 1.0 2.0 3
reduced to only 0.3 mm (0.012 in.) because of two cracks
immediately above the construction joint (both labeled D), D 0.5 0.8 1.25 1.5 3
which were each 1.5 mm (0.059 in.) wide. The crack along D — — 0.5 1.5 1.5
the construction joint was always wider at the east end than C 0.8 1.25 1.5 0.3 4
at the west end, and final failure of the wall was concentrated B 0.1 0.2 0.2 0.3 0.3
at the east end of the construction joint. Due to the additional A 0.15 0.2 0.25 0.3 0.3
vertical reinforcement provided below the construction joint Wall displacement, west (mm)
Crack
(Fig. 1), Cracks A and B never became wider than 0.3 mm (east face) –49 –77 –104 –138 –187
(0.012 in.). I — — — 0.6 0.3
A total of 15 primary flexural cracks occurred in the wall H — 0.1 0.25 0.4 0.4
flanges over a 3.6 m (11.8 ft) height, resulting in an average
G –0.3 0.5 0.4 0.6 0.8
crack spacing of 240 mm (9.5 in.). Near the middle of the
F 0.15 0.4 0.4 1.0 1.2
web, the vertical spacing of cracks was wider due to the
lower percentage of vertical reinforcement (0.65% in E — 0.6 0.5 1.0 1.5
flanges, 0.25% in web) providing less crack control. D 0.25 0.6 0.8 1.25 3
A significant amount of flexure-shear diagonal cracking C 0.8 1.5 3 3.5 7
was observed considering that the maximum shear stress V/blw B 0.1 0.15 0.1 0.15 0.25
was only 0.78 MPa (113 psi). This corresponds to 0.11√fc′ in A 0.15 0.2 0.25 0.3 0.3
MPa units, and 1.3√fc′ in psi units, which is much less than Note: 25.4 mm = 1 in.

ACI Structural Journal/September-October 2007 553

the concrete contribution Vc for a member subjected to an Strains
axial compression of 0.10fc′Ag. The wall was very slender Figure 7(a) presents the relationship between the strain of
and shear deformations due to diagonal cracks were small; vertical reinforcement and the wall displacement measured
however, diagonal cracks did appear to have an influence on by Strain Gauge SG-1 on the east side (compression side for
the strain of the vertical reinforcement. positive lateral force). The cycles before reinforcement
Some concrete cover on the compression face fell off at yielding clearly show two different slopes—when reinforcement
approximately 180 mm (7.1 in.) wall displacement (1.5% and surrounding concrete is in compression and when
global drift), and a much larger amount of concrete cover fell reinforcement is in tension. The curves are offset from the
off during the first cycle to 280 mm (11.0 in.) wall displacement origin due the applied axial compression. During the first
(2.4% global drift). It was observed that the vertical reinforcing cycle to 60 mm (2.4 in.) total displacement (46 mm [1.8 in.]
bar in the middle of the flange had buckled out (refer to wall displacement), the reinforcement yielded and the strain
Fig. 6(b)), and that the bar had pushed off the concrete cover. increased suddenly. This point is referred to as first yielding
The hoops in the flanges were spaced at approximately 6 times in Fig. 5(b). During the next three cycles to 60 mm (2.4 in.)
the diameter of the vertical reinforcing bars, which is sufficient total displacement, the vertical reinforcement strain remained
to prevent significant bar buckling; but the rectangular hoops greater than the yield strain throughout the entire cycle. As
were not effective in preventing buckling of the unsupported the wall was subjected to further cycles and the compression
bar midway between the corners of the rectangular hoops. strain of concrete increased, the tension strain of the vertical
The buckled vertical reinforcing bar fractured while subjected reinforcement reduced. For clarity, data is not shown beyond
to tension during the third cycle to 280 mm (11.0 in.) wall the 90 mm (3.5 in.) total displacement level; however, the
displacement (2.4% global drift). strains measured by all four strain gauges at the peak
Due to loss of concrete and fracturing of a vertical reinforcing displacement levels are summarized in Table 4.
bar, the wall stiffness reduced significantly and the wall strength The relative displacements of targets glued to the flexural
reduced by 8% in just one cycle. The test was stopped after compression face and flexural tension face of the wall were
the fourth cycle to 280 mm (11.0 in.) wall displacement used to examine the average axial strains over the cracked
because the displacement capacity of the hydraulic jack had region of the wall. The difference in relative target displacement
been reached; however, the wall displacement capacity had on the two wall faces divided by the wall length between the
also been reached. faces equals the rotation over the wall height between the
targets (typically 500 mm [20 in.] as shown in Fig. 2). The
average curvature over each segment is equal to the rotation
divided by the vertical distance between targets. Figure 8
summarizes these average curvatures over the lower 5 m (16 ft)
of wall. The first curvature value near the wall base is for the
segment from Target 1 to Target 3 (refer to Detail A in Fig. 2).
This segment was 665 mm (26 in.) long and included the

Table 4—Axial strains at base of wall during first

cycle to specified total displacement
Total displacement, mm
Gauge* 15 18 24 48 60 75 90 120 150 200
Strain: positive lateral load—tension face
W1-W3 0.38 0.49 0.72 1.80 2.74 4.10 4.48 5.97 5.93 9.77
W3-W4 –0.02 –0.02 0.04 0.90 1.48 2.19 3.86 6.48 9.36 14.72
SG 3 0.00 0.25 0.95 2.12 >9.0 >9.0 >9.0 >9.0 >9.0 >9.0
SG 4 –0.18 –0.01 0.65 1.94 8.22 6.40 4.00 3.95 4.01 >9.0
Strain: positive lateral load—compression face
E1-E3 –0.32 –1.04 –0.76 –0.80 –0.87 –1.10 –1.29 –1.80 –1.84 –2.14
E3-E4 –0.61 –0.65 –0.58 –0.90 –1.11 –1.19 –1.25 –1.42 –1.52 –2.03
SG 1 –0.49 –0.55 –0.64 –0.71 –0.87 2.32 0.99 0.50 4.19 4.63
SG 2 –0.37 –0.39 –0.46 –0.59 –0.72 0.30 –0.55 –1.23 –0.50 >9.0
Strain: negative lateral load—tension face
E1-E3 –0.11 0.06 0.00 1.84 2.58 4.21 5.52 7.96 11.29 15.84
E3-E4 –0.29 –0.33 0.09 0.17 0.56 1.84 3.13 4.74 7.63 11.45
SG 1 –0.02 0.02 0.48 2.93 >9.0 >9.0 >9.0 >9.0 >9.0 >9.0
SG 2 0.00 0.05 0.56 2.73 5.94 7.81 2.10 3.07 >9.0 >9.0
Strain: negative lateral load—compression face
W1-W3 –0.11 –0.34 –0.19 –0.59 –0.57 –3.72 –2.21 –1.49 –1.91 –2.4
W3-W4 –0.07 –0.33 –0.40 –0.65 –0.77 –0.40 –0.84 –1.02 –1.25 –1.52
Fig. 7—Measured axial deformations: (a) local strain of SG 3 –0.53 –0.56 –0.66 –0.89 2.77 0.77 –0.72 >9.0 2.25 >9.0
vertical reinforcement at construction joint; and (b) axial
SG 4 –0.58 –0.63 –0.74 –0.99 1.25 –1.52 –0.72 –1.31 –1.46 –0.88
displacement over lower 5.0 m (16 ft) of wall during first
*
cycle to each total displacement level (1 kN = 0.228 kip; Refer to Fig. 3 for locations of concrete surface displacement targets W1 to W4, E1
to E4, and strain gauges on reinforcement SG1 to SG4.
25.4 mm = 1 in.). Note: 25.4 mm = 1 in.

554 ACI Structural Journal/September-October 2007

construction joint. While the bending moment varied compression strain capacity is 0.003, the curvature capacity
linearly over the height, the curvatures varied approximately of the wall is predicted to be 0.003/135 mm = 22 rad/km
bilinearly because of the larger curvatures in the cracked (566 μrad/in.). The largest average curvature measured during
portion of the wall. the test was approximately 12 rad/km (309 μrad/in.) (refer to
The average relative target displacement on the two wall
faces divided by the 500 mm (20 in.) initial target spacing is
the average axial strain. Figure 9 summarizes these over the
lower 5 m (16 ft) of wall. At low lateral displacement levels,
the axial strains were compressive due to the large compression
force applied to the wall. As the peak lateral displacement
levels were increased, the tension strains on one side of the
wall became significantly larger than the compression
strains on the other side of the wall.
Figure 7(b) presents the total axial displacement measured
over the lower 5.08 m (200 in.) height of wall. For clarity,
only the first cycle is shown for each peak lateral displacement
level. The axial displacement is the integral of axial strains
shown in Fig. 9. Figure 7(b) shows the variation of axial
displacement (average strain) as the lateral displacement is
increased, while Fig. 9 shows the variation of axial strain
over the height at one particular lateral displacement level.
Initially, the axial displacement was a constant compressive
value due to the applied axial compression. As the lateral
load was increased, significant axial expansion of the wall
occurred. At 200 mm (7.9 in.) total lateral displacement,
which corresponds to 185 mm (7.3 in.) lateral wall displacement
and 1.6% global drift, the axial deformation was approximately
10 mm (0.4 in.), which corresponds to an average axial strain
of approximately 0.002. Figure 9 indicates that at this lateral
displacement level, the axial strain varied from a maximum
of approximately 0.007 near the base to –0.0005 further up
the wall.
Table 4 summarizes axial strains measured at the wall Fig. 8—Average curvatures over lower 5.0 m (16 ft) of wall
base from two sources—average concrete strains between (25.4 mm = 1 in.).
Targets W1 to W4 and E1 to E4, and reinforcement strains
from Gauges SG1 to SG4 (refer to Fig. 2 for locations of all
targets and strain gauges). The average concrete strains
included concrete tension stiffening effect, while strain
gauges measured maximum reinforcement strain at the crack
(construction joint). The strain gauges had an upper-range of
9.0 × 10–3. Strain measurements between Targets W1 and
W3 on the west side and E1 and E3 on the east side include
displacements at the construction joint.
The average concrete strains in Table 4 are plotted versus
wall displacement in Fig. 10 to extrapolate to the average
strains at maximum wall displacement of 280 mm (11 in).
Figure 10(a) indicates the maximum compression strains were
approximately 3.0 × 10–3, which is the traditionally assumed
failure strain of unconfined concrete. Figure 10(b) indicates
the maximum tension strains were between 15 × 10–3 and
approximately 25 × 10–3, which is considerably less than the
50 × 10–3 (5%) average strain normally expected at fracture
of reinforcement. A bare reinforcing bar will fracture at
approximately 10% strain. Tension stiffening of concrete
reduces average strain of a bar cast within concrete to
approximately 5%. Fracture of reinforcement occurred at an
average strain of only 2.5% due to the additional local strain
caused by bar buckling.

Drift capacity
A plane sections analysis of the test wall subjected to an
axial compression of 1500 kN (337 kip) gives a compression
stress block depth a = 115 mm (4.5 in.), and a corresponding Fig. 9—Average axial strain over lower 5.0 m (16 ft) of wall
neutral axis depth c = 135 mm (5.3 in.). Assuming the concrete (25.4 mm = 1 in.).

ACI Structural Journal/September-October 2007 555

Fig. 8); but this was at a global drift of only 1.6%. Using the be uniform over half the height of the linearly varying curvature,
extrapolated strains in Fig. 10 to determine the maximum and two such inelastic curvature distributions result in
curvature at the maximum drift gives (0.025 + 0.003)/ approximately the same inelastic wall displacement. At 200 mm
1625 mm = 17 rad/km (438 μrad/in.). As these strains are (8 in.) total displacement (1.6% drift), the inelastic curvature
average strains measured over a certain height, 17 rad/km varied linearly from maximum at 426 mm (1.4 ft) from the
(438 μrad/in.) is an average curvature over a 500 mm (19.7 in.) base to zero at approximately 2 m (6.6 ft) from the base,
height of the wall. The maximum curvatures within the 500 mm which is over a distance equal to approximately the wall
(19.7 in.) will be significantly higher, and thus it can be length lw = 1625 mm (5.3 ft). Thus, the equivalent uniform
concluded that when the wall reached the drift capacity of inelastic curvature would be over approximately 50% lw.
2.4%, the maximum curvature was approximately equal to Figure 8 indicates the height of linearly varying inelastic
the predicted curvature capacity of 22 rad/km (566 μrad/in.). curvatures increase as the curvatures increase. At the
Inelastic drift capacity of a flexural wall is normally maximum wall drift, the curvature increased approximately
predicted from a simple model where inelastic curvature is 40% from the curvature at 1.6% drift. Thus, the height of
assumed to be uniform over a certain height called the plastic linearly varying inelastic curvature and the height of equivalent
hinge length. It is interesting to determine what plastic hinge uniform inelastic curvature would correspondingly increase
length is indicated by the test results. The maximum wall at the maximum wall drift. In summary, both methods of
displacement was 281 mm (11.1 in.), and the elastic portion estimating the equivalent length of uniform inelastic
was 46 mm (1.8 in.). Thus, the inelastic wall displacement curvatures, that is, the plastic hinge length, suggest lp ~ 2/3lw.
was Δi = 235 mm (9.3 in.). The curvature capacity of the wall
is 22 rad/km (566 μrad/in.), and Fig. 11(a) indicates the EFFECTIVE STIFFNESS
elastic portion is 2 rad/km (51 μrad/in.). Thus the inelastic Simple nonlinear flexural model
curvature capacity is φi = 20 rad/km (509 μrad/in.). The height The most general model for calculating nonlinear flexural
from base of plastic hinge to level of wall displacement h = response of any reinforced concrete section is a fiber (layer)
11.33 m (37.2 ft). Substituting these values into the expression model, in which stress-strain relationships for concrete and
Δi = φilp(h – lp/2) and solving for the plastic hinge length
gives lp = 1.09 m (3.6 ft), which is 67% of wall length lw.
Figure 8 presents total curvatures measured at displacement
levels less than the displacement capacity of the wall. As
the elastic portion of the curvature is a constant 2 rad/km
(51 μrad/in.), the distribution of inelastic curvatures can be
visualized by shifting the vertical axis in Fig. 8 by the same
constant. The measured curvature distributions indicate the
inelastic curvatures were not uniform over a plastic hinge
length, but varied linearly. A linearly varying inelastic
curvature distribution gives the same inelastic rotation as
calculated from the maximum inelastic curvature assumed to

Fig. 11—Comparison of experimental results with nonlinear

model predictions: (a) bending moment-curvature data; (b)
Fig. 10—Measured average strains at base of wall: (a) peaks of load-displacement cycles; and (c) loading curves
compression strains; and (b) tension strains (25.4 mm = 1 in.). to various drift levels (1 kN = 0.228 kip; 25.4 mm = 1 in.).

556 ACI Structural Journal/September-October 2007

reinforcement are used to determine the stress in fibers stress in the vertical reinforcement equal to the yield stress,
(layers) of the cross section at a given strain level. The that is, strain-hardening is ignored. The nominal flexural
results from a fiber model16 were used to develop a simple capacity of the wall specimen based on the measured yield
nonlinear flexural model for typical high-rise concrete shear- strength of the vertical reinforcement is Mn = 1594 kNm
walls where the nonlinear compression strains of concrete (1175 k-ft). If the axial compression P applied to the wall is
are relatively small.14 The simple model is well suited for assumed to be zero, Mn reduces to approximately 30% of this
nonlinear static (pushover) analysis of concrete walls, and value, thus approximately 70% of the wall overturning
leads directly to the development of a linear (effective stiffness) capacity Mn is due to P.
model for high-rise concrete walls. The nonlinear model is Figure 11(a) compares predicted and measured bending
described and compared with the measured wall response moment-curvature relationships. As the test wall was
first, and discussion of the linear model follows. subjected to uniform axial compression and had uniform
The stress-strain response of concrete and post-yielding vertical reinforcement over the height, all curvatures shown
stress-strain response of reinforcement depend on strain in Fig. 8 can be plotted on one diagram by determining the
history; thus, the cyclic flexural response of reinforced associated bending moment at the height where the curvature
concrete depends on the maximum strains reached in was measured. As the curvatures were measured at peaks of
previous load cycles. Rather than deal with many different load cycles, predicted upper-bound curves are shown. The
relationships depending on load history, the concept of difference between the trilinear curve and the nonlinear
upper-bound and lower-bound bending moment-curvature curve, which was determined using a fiber model, is small
response can be used. The upper-bound response corresponds to compared with the scatter in experimental data.
a previously uncracked wall loaded monotonically to failure, Figures 11(b) and 11(c) compare predicted lateral force-
while the lower-bound response corresponds to a wall that is displacement (pushover) predictions with the measured
severely cracked during previous load cycles. response. While the prediction is based on a trilinear bending
The upper-bound and lower-bound bending moment- moment-curvature relationship, the resulting load-deflection
curvature relationships of a typical concrete wall can be relationship is nonlinear. Using an 11.76 m (38 ft, 7 in.) high
approximated by trilinear curves14 defined by four parameters: cantilever model of the wall resulted in a higher uncracked
1) slope of first straight line segment assumed to be EcIg; stiffness of the wall than measured due to deformations of
2) slope of second straight line segment assumed to be equal the wall base. Rather than using a more complex model to
to the cracked section flexural rigidity Ec Icr ; 3) bending moment account for deformations of the base, a fixed-ended canti-
Ml defining intersection of first and second linear segments lever with an increased effective height of 12.08 m (39 ft, 7.6 in.)
(value of Ml distinguishes between upper-bound and lower- was used to obtain the correct uncracked stiffness of the wall,
bound cases); and 4) flexural capacity of section Mn. and this model was used to do all push-over predictions.
For the test specimen Ec = 30,000 MPa (4350 ksi), Ig = Figure 11(b) compares the predicted upper-bound
98.0 × 109 mm4 (235 × 103 in.4) and the gross section flexural response with the measured envelope of the load-deflection
rigidity EcIg = 2.94 × 109 kNm2 (1.84 × 109 k-ft2). The cracked curves defined by the peak of the first cycle of loading to
section flexural rigidity EcIcr of the wall can be determined each displacement level in both the positive and negative
from first principles (initial slope of bending moment-curvature loading directions. There is generally very good agreement
relationship for zero axial load and no concrete tension stresses), except at large displacements where there is a strength
or can be estimated from a chart.14 For the test wall: EcIcr = increase due to strain hardening of reinforcement that is not
0.294 × 106 kNm2 (0.217 × 106 k-ft2), Icr = 9.7 × 109 mm4 accounted for in the prediction. The maximum bending
(23 × 103 in.4), and Icr/Ig = 0.10. moment applied at the construction joint was 18% larger
The following expression14 can be used to estimate the than the predicted nominal flexural capacity of the wall. This
bending moment defining the intersection of first and second increase is relatively small considering that the maximum
linear segments strength of the reinforcement is 40% greater than the yield
strength used to predict the strength. The reason the flexural
capacity of the wall increased by only 18% is because most
M l = ⎛ 1.5βf r + -----⎞ S g + 0.08Pl w
P
(1) of the flexural capacity of the wall is due to the axial
⎝ A g⎠
compression, not the reinforcement.
The predicted lower-bound response is compared with
where β = 1 for upper-bound and 0 for lower-bound, fr = results from the second cycle of loading to larger drift levels
0.3√fc′ = cracking strength of concrete subjected to uniform in Fig. 11(c). The indicated drift is the top wall displacement
stress in MPa units (fr = 4√fc′ in psi units), Sg is the gross divided by 11.76 m (38 ft, 7 in.) height of wall. The proposed
section modulus (Ig/yt, where yt refers to tension side of trilinear bending moment-curvature model assumes the
wall), P is the applied axial compression, Ag is the gross initial portion of all loading cycles will follow the original
cross-sectional area of wall, and lw is the wall length. Note uncracked wall response until the applied bending moment
that if 1.5β is taken as 1, and the term 0.08Plw is removed, exceeds the lower-bound Ml value of 776 kNm (572 k-ft),
Eq. (1) gives the bending moment at first cracking Mcr. For which corresponds to a lateral force of 66 kN (15 kip). The
the test wall: fr = 2.3 MPa (0.33 ksi), Sg = 120.1 × 106 mm3 experimental results do indicate an initial linear response up
(7.3 × 103 in.3), P = 1500 kN (337 kip), Ag = 310 × 103 mm2 to approximately that lateral load level; but the slope of the
(481 in.2), lw = 1626 mm (64 in.); resulting in Mcr = 857 kNm initial linear portion reduces as the wall is pushed to higher
(632 k-ft), upper-bound Ml = 1190 kNm (877 k-ft), and drift (damage) levels. The reduction in initial stiffness is
lower-bound Ml = 776 kNm (572 k-ft). accompanied by a corresponding increase in residual
The final parameter required to define the trilinear bending displacement so that at a lateral load of approximately 50 kN
moment-curvature relationship of a wall is the nominal flexural (11 kip), the predicted and observed displacements are equal
capacity Mn, which is normally calculated using a maximum except at 1.55% drift. The residual displacement and

ACI Structural Journal/September-October 2007 557

corresponding reduced initial stiffness is due to damage at at the wall base.15 Both methods account for the initial
the cracks preventing cracks from closing perfectly. A uncracked-section rigidity EcIg before cracking and the retention
refined model requires crack closing springs; however, of an initial stiffness approximately equal to the uncracked-
considering the good agreement between the simple trilinear section stiffness after cracking due to axial compression
model prediction and the measured response, the additional from gravity loads closing previously developed cracks.
complexity is not warranted. The large reduction in initial To develop a general method for effective stiffness of
stiffness to 1.55% drift is due to concrete spalling in the cantilever concrete walls, a number of simplifying assumptions
plastic hinge region of the wall. were used. The uncracked-section (initial) rigidity EcIg and
cracked-section rigidity EcIcr of the wall were assumed to be
Linear model: effective flexural stiffness uniform over the wall height. Many high-rise walls have
Linear dynamic analysis can provide a good estimate of relatively uniform geometry over the height, but even when
displacement demand of high-rise concrete walls provided this is not the case, the error will be small if the flexural rigidities
that an appropriate effective stiffness is used to account for near the wall base are used because wall deflections are
the effect of cracking. The trilinear bending moment-curvature primarily due to curvatures near the base. The value of Ml
model described previously was used to develop a general was assumed to vary linearly from the value at the base to
method for estimating the upper-bound and lower-bound zero at the top. This is reasonable because axial compression
effective flexural rigidity of concrete walls accounting for in a wall will vary approximately linearly from maximum at
the magnitude of axial compression, amount, and distribution the base to zero at the top. The effective flexural stiffness
of vertical reinforcement, concrete geometry, and all parameters was determined from the slope of the elastic portion of an
that affect the flexural capacity of the wall.15 In addition, equivalent elastic-plastic load-displacement curve that has
simple expressions were developed to estimate the upper- the same area under the curve as the actual nonlinear load-
bound and lower-bound effective flexural stiffness of displacement relationship. The lateral force was assumed to
concrete walls based only on the level of axial compression vary linearly over the height of the wall, and the reference
lateral displacement was taken at the top of the wall. The
resulting expression for effective moment of inertia is

M a M 1.6a
I e = I cr + 3 ⎛ -------l ⎞ – 2 ⎛ -------l ⎞ ( I g – I cr ) ≤ I g (2)
⎝ M n⎠ ⎝ M n⎠

where a = 1.1(Icr /Ig)–0.4. For the test wall: Icr/Ig = 0.10,

upper-bound Ml/Mn = 0.75, and lower-bound Ml/Mn = 0.49,
thus from Eq. (2), upper-bound Ie/Ig = 0.79 and lower-bound
Ie/Ig = 0.39.
The effective moment of inertia given by Eq. (2) for a wide
range of different walls were examined,16 and it was
observed that the most important parameter is the axial
compression applied at the wall base. The following simplified
expressions give reasonable estimates of upper-bound and
lower-bound effective moment of inertia of concrete walls

I e = ⎛ 0.6 + -----------⎞ I g ≤ I g
P
(3)
⎝ f c ′A g⎠

I e = ⎛ 0.2 + 2.5 -----------⎞ I g ≤ 0.7I g

P
(4)
⎝ f c ′A g⎠

For the test wall, which was subjected to an axial compression

of P/fc′Ag = 0.10, Eq. (3) gives an upper-bound Ie = 0.70Ig,
while Eq. (4) gives a lower-bound Ie = 0.45Ig. Equation (4)
generally gives a lower result than the lower-bound obtained
from Eq. (2); however, in this particular case, it gives a
slightly higher value.
Figure 12(a) presents the envelopes of the load-displace-
ment relationship for the two directions of loading along with
two equivalent bilinear load-displacement relationships with
equivalent area under the curve up to the point that the
Fig. 12—Comparison of experimental results with linear nonlinear curves cross the nominal capacity. The results indi-
model predictions: (a) peaks of load-displacement cycles; cate an effective elastic stiffness ke = 0.79kg when the wall was
(b) loading curve to 1.12% drift; and (c) effective stiffness pushed east and ke = 0.74kg when the wall was pushed west,
during loading to different drift levels (1 kN = 0.228 kip; which compares well with the upper-bound predictions of Ie =
25.4 mm = 1 in.). 0.79Ig from Eq. (2) and Ie = 0.70Ig from Eq. (3).

558 ACI Structural Journal/September-October 2007

 Figure 12(b) presents the load-displacement relationship fracture of the bar after a few post-buckling cycles. The
of the first cycle of loading to a 1.12% drift, and an equivalent maximum global drift of the wall was 2.4%.
bilinear relationship based on an equal area under the elastic
portion of the curve. For this drift level, the effective elastic REFERENCES
stiffness is 39% of the uncracked stiffness, which is exactly 1. American Society for Civil Engineers, “FEMA 356 Prestandard and
the lower-bound prediction from Eq. (2). This same procedure Commentary for the Seismic Rehabilitation of Buildings,” Reston, Va.,
Nov. 2000.
was done at all drift levels and the results are summarized in 2. NZS 3101:Part 2, “Concrete Structures Standard Part 2—Commentary on
Fig. 12(c). The reduction in effective elastic stiffness up to the Design of Concrete Structures,” Standards New Zealand, 1995, 264 pp.
the drift levels shown is due to cracking, and the lower- 3. Paulay, T., “Seismic Response of Structural Walls: Recent Developments,”
bound of these results agree well with the predicted lower- Canadian Journal of Civil Engineering, V. 28, No. 6, Dec. 2001, pp. 922-937.
4. Cardenas, A., and Magura, D., “Strength of High-Rise Shear Walls—
bound. At approximately 1.5% drift, concrete cover spalled Rectangular Cross Sections,” Response of Multistory Concrete Structures
from the wall specimen causing the effective stiffness of the to Lateral Forces, SP-36, American Concrete Institute, Farmington Hills,
wall to reduce to approximately 30% of the uncracked stiffness. Mich., 1973, pp. 119-150.
Additional concrete fell off the wall at a drift of approximately 5. Oesterle, R. G.; Fiorato, A. E.; Johal, L. S.; Carpenter, J. E.; Russell,
2.4% and the effective stiffness reduced further. Obviously, H. G.; and Corley, W. G., “Earthquake Resistant Structural Walls—Tests of
Isolated Walls,” Report to National Science Foundation, Construction
the effective stiffness of a wall will reduce if concrete is Technology Laboratories, Portland Cement Association, Skokie, Ill., Nov.
removed due to spalling, but up to approximately 1.5% drift, 1976, 315 pp.
the reduction in stiffness was primarily due to concrete 6. Oesterle, R.; Aristizasbal-Ochoa, J. D.; Fiorato, A. E.; Russell, H. G.;
cracking and the predicted lower-bound is very reasonable. and Corley, W. G., “Earthquake Resistant Structural Walls—Tests of
Isolated Walls—Phase II,” Report to National Science Foundation,
Construction Technology Laboratories, Portland Cement Association,
CONCLUSIONS Skokie, Ill., Oct. 1979, 325 pp.
A test was conducted on a large-scale model of a concrete 7. Shiu, K. N.; Daniel, J. I.; Aristizasbal-Ochoa, J. D.; Fiorato, A. E.; and
shearwall from the core of a high-rise building. Due to the Corley, W. G., “Earthquake Resistant Structural Walls—Tests of Walls
With and Without Openings,” Report to National Science Foundation,
combination of significant axial compression (simulating Construction Technology Laboratories, Portland Cement Association,
gravity loads) and a low percentage of vertical reinforcement, Skokie, Ill., July 1981, 120 pp.
horizontal cracks closed whenever the lateral force was 8. Oesterle, R. G.; Aristizabal-Ochoa, J. D.; Shiu, K. N.; and Corley, W. G.,
removed, and there was generally very little residual lateral “Web Crushing of Reinforced Concrete Structural Walls,” ACI JOURNAL,
Proceedings V. 81, No. 3, May-June 1984, pp. 231-241.
displacement. Even after the wall was displaced to over 9. Morgan, B. J.; Hiraishi, H.; and Corley, W. G., “U.S.-Japan Quasi-
1.5% global drift—approximately four times the drift at Static Test of Isolated Wall Planar Reinforced Concrete Structure,” Report
yield of the vertical reinforcement—the residual drift under to National Science Foundation, Construction Technology Laboratories,
zero lateral force was less than 0.1%. When the wall was Portland Cement Association, Skokie, Ill., Aug. 1986, 111 pp.
reloaded after being severely cracked, the initial stiffness of 10. Lefas, I., “Behaviour of Reinforced Concrete Walls and its Implication
for Ultimate Limit State Design,” PhD thesis, University of London, UK, 1988.
the wall with closed cracks was similar to the stiffness of the 11. Aejaz, A., and Wight, J. K., “R/C Structural Walls with Staggered
uncracked wall. Door Openings,” Journal of Structural Engineering, ASCE, V. 117, No. 5,
The test results were used to calibrate a simple nonlinear 1991, pp. 1514-1531.
flexural model14 for high-rise concrete walls. The model 12. Thomsen, J. H., and Wallace, J. W., “Displacement-Based Design of
RC Structural Walls: An Experimental Investigation of Walls with Rectangular
assumes that due to axial compression closing cracks, the and T-Shaped Cross-Sections,” Report to National Science Foundation,
initial stiffness of a previously cracked wall is equal to the Clarkson University, Potsdam, N.Y., June 1995, 353 pp.
stiffness of an uncracked wall. The model could be refined by 13. Dazio, A.; Wenk, T.; and Bachmann, H., “The Testing of Reinforced
accounting for residual crack opening, but the added complexity Concrete Walls under Static-Cyclic Actions,” IBK Report No. 239, Institute
of Structural Engineering, The Swiss Federal Institute of Technology, Zurich,
does not appear to be warranted. As the rational model has Switzerland, 1999. (in German)
been calibrated with this large-scale test, it can be used to 14. Adebar, P., and Ibrahim, A. M. M., “Simple Nonlinear Flexural Stiffness
predict the response of a variety of high-rise concrete walls. Model for Concrete Shear Walls,” Earthquake Spectra, EERI, V. 18, No. 3,
There was a surprising amount of flexure-shear diagonal Aug. 2002, pp. 407-426.
15. Ibrahim, A. M. M., and Adebar, P., “Effective Flexural Stiffness for
cracking in the wall considering the low level of shear stress. Linear Seismic Analysis of Concrete Walls,” Canadian Journal of Civil
Diagonal cracking did not appear to influence the overall Engineering, V. 31, No. 4, Aug. 2004, pp. 597-607.
response of the wall other than increasing the spread of 16. Ibrahim, A. M. M., “Linear and Nonlinear Flexural Stiffness Models
inelastic curvatures. A discussion on how to account for for Concrete Walls in High-Rise Buildings, PhD thesis, University of British
shear strains in a concrete wall can be found elsewhere.17 Columbia, Vancouver, British Columbia, Canada, 2000, 143 pp.
17. Gerin, M., and Adebar, P., “Accounting for Shear in Seismic Analysis of
The flexural failure mode of the wall was buckling of a Concrete Structures,” 13th World Conference on Earthquake Engineering,
vertical reinforcing bar causing spalling of concrete and then Vancouver, British Columbia, Canada, Aug. 2004, 13 pp.

ACI Structural Journal/September-October 2007 559

 DISCUSSION
Disc. 104-S51/From the September-October 2007 ACI Structural Journal, p. 542

Thermal Movements in Parking Structures. Paper by Mohammad Iqbal

Discussion by Gary Klein and Richard Lindenberg

ACI member, Wiss, Janney, Elstner Associates, Inc., Northbrook, IL; Davis Bowan & Friedel, Salisbury, MD

This paper discusses a very important but frequently parking structures. Based on the data, the author recommends a
misunderstood aspect of parking structure design. The paper movement factor of 0.6 for precast parking structures
provides very interesting data on actual movements of post- (compared with 0.8 for post-tensioned parking structures);
tensioned and precast parking structures. The discussers are that is, the author recommends calculating thermal shortening as
the principal investigators for an ongoing PCI-funded study 60% of the theoretical temperature strain times the length
on volume change effects in precast buildings, which is contributing to movement. Given the extreme scatter of the
referred to by the author. data, higher or lower movement factors are equally justified.
Under the discussion of thermal coefficient of expansion, The author further recommends a coefficient of thermal
the author contends that PCI reduces the temperature strains expansion of 7.5 microstrains per degree Fahrenheit, which
by a factor of 1.5 when computing thermal shortening. is somewhat higher than the value of 6 microstrains per
Although the PCI Handbook5 is not as clear as it perhaps degree Fahrenheit used in the PCI Handbook5 and other
should be, it does not recommend using the 1.5 factor for references. Interestingly, taken together, the author’s
reducing thermal shortening; rather, the factor is used to recommendations result in a calculated thermal shortening
estimate equivalent volume change strains for purposes of for precast concrete structures that is identical to that
calculating volume change forces. As indicated in Section computed using the current handbook, which includes PCI’s
3.4.3.1 of the handbook, “…since the shortening takes place 25% reduction for thermal lag. The author’s approach is an
gradually over a period of time, the effect of the shortening improvement in the sense that it acknowledges the effect of
on the shears and moment of the support is lessened because incremental movements at precast connections in reducing
of creep and micro cracking of the member and its support.” the movement demand at expansion joints. Both the PCI
In other words, the 1.5 factor is solely for the purpose of Handbook5 and author’s method provide a reasonable estimate
calculating member forces based on equivalent shortening. of thermal movement for expansion joint design. As
Appropriately, the design examples in the handbook do not confirmed by the author’s observations, actual movement
use the 1.5 factor for calculating temperature shortening. will vary greatly. The paper’s documentation of this variability
The author correctly points out that the PCI Handbook5 is a significant contribution toward a better understanding of
recommends a 25% reduction in computed thermal strains the nature of volume change movement. Managing volume
due to the thermal lag effect. The author argues that thermal change forces, however, is a problem of even greater
lag effects are very small and should be conservatively complexity and variability. Designers must accommodate
ignored when the computing volume change movement. both movements and forces without damaging the structure.
Referring to the PCI study that the discussers are now The report on the PCI research, due summer of 2008, will
completing, the author notes that summer temperatures highlight the importance of flexible connections and ductile
outside parking structures are on the order of 10 °F (5.6 °C) members in managing volume change forces and precluding
hotter than ambient temperatures inside the parking structure. damage due to volume change movement.
Because thermal shortening is even more critical than
thermal expansion, it should have also been noted that AUTHOR’S CLOSURE
ambient temperatures inside parking structures are some- The author appreciates the discussers’ interest in the paper.
what warmer than the outdoor temperatures in winter The paper provides an empirical underpinning for the design
months, especially in the lower levels. As a result of these of expansion joints and proposes an equation (Eq. (1) of the
temperature differences, the thermal response of parking paper) to estimate expansion joint thermal movements. The
structures is less extreme than would be predicted based on discussion focuses on two design factors in Eq. (1), namely,
ambient temperatures. In other words, the thermal lag effect design temperature and the movement factor (or M-factor).
is primarily due to the less extreme microclimate inside the Since the publication of this paper, the author has carried out
structure. The discussers agree with the author that the lag further investigation on expansion joint design. Both factors
associated with the time it takes for a structural element to will become clearer after the author’s subsequent paper on
reach ambient temperature is indeed not very significant. expansion joints is published. Regarding the value of design
Nonetheless, some reduction in thermal strains calculated temperature, the author recommends using the temperature
based on extreme outdoor temperatures is appropriate. A values recommended by the Federal Construction Commission
reduction of 25% approximates the combined microclimate and reproduced by ACI 224.3R-95.12 Any further reduction
and thermal lag effects at the lowest level of parking structures, in design temperature values would result in an overstress of
where volume change forces are most significant. The expansion joint assemblies, as shown in Fig. 3.
reduction at the upper levels is much less and, as suggested The discussers point to the obvious scatter in the test data
by the author, may be conservatively neglected. to compute the M-factors. As stated in the paper, the data
Referring to Fig. 9 of the paper, it is interesting to note the includes all readings recorded using constructed parking
extreme scatter of observed thermal movements in precast facilities in service, and not on laboratory samples tested in

ACI Structural Journal/July-August 2008 507

a control environment. The lateral-force-resisting systems and predicting volume change cracking in a structure. An
(LFRS) in the facilities had beam-column frames of various M-factor of 0.8 means that a structure moves 80% of the total
story heights, member sizes, and concrete strengths. These unrestrained movement with the remaining 20% movement
properties affected the LFRS stiffnesses and the restraint to consumed in the structural restraint. The 80% movement
thermal movement to a varying degree. As defined in the level also indicates the degree of restraint post-tensioned
paper, the movement factor M is the ratio of actual movement structures may tolerate while performing reasonably well.
and the calculated unrestrained movement under volume Therefore, the M-factor helps predict whether a facility
changes. The M-factor indicates a structure’s intrinsic capability would perform well under volume change effects. If a
to move under volume changes, and as such it serves as an structure’s LFRS is designed with an M-factor of 0.8 and
index. The M-factor values range from 0 and 1. Under an higher, it would exhibit minimal volume change cracks. If the
ideal condition in which a diaphragm is free to move under LFRS is stiffer and has a lower M-factor, however—say 0.7—
volume changes, its M-factor will be unity. On the other the facility is likely to exhibit more volume change cracking. An
hand, a diaphragm that is fully restrained by a rigid LFRS extremely stiff structure having an M-factor of zero or near-zero
extending the full length of the diaphragm has an M-factor would be the most prone to volume change cracking.
of zero. The pretopped precast structures data (Fig. 9) A correction to the paper is also cited by the author—on
exhibited more scatter than post-tensioned structures’ data p. 546, third line from the bottom of the right-hand column,
(Fig. 8) due to the presence of precast joints in the precast the number in the parenthesis is inadvertently noted as 4.8 m.
construction. The M-factors were computed to represent the It should be corrected to read 2.4 m.
upper bound of the movements.
REFERENCES
For post-tensioned and field-topped precast parking 12. ACI Committee 224, “Joints in Concrete Construction (ACI 224.3R-95),”
structures, the M-factor serves a useful index in quantifying American Concrete Institute, Farmington Hills, MI, 1995, 42 pp.

Disc. 104-S52/From the September-October 2007 ACI Structural Journal, p. 549

Test of High-Rise Core Wall: Effective Stiffness for Seismic Analysis. Paper by Perry Adebar, Ahmed M. M.
Ibrahim, and Michael Bryson

Discussion by Himat Solanki

Professional Engineer, Building Dept., Sarasota County Government, Sarasota, FL

The authors have presented an interesting concept on the Based on References 23 and 24, the authors’ Eq. (3) is again
effective stiffness of core/structural walls. The discusser inconsistent with the aforementioned values because, for
would like to offer the following comments: example, when P/fc′ Ag = 0.5, the authors’ Ie will be Ig instead of
1. The authors have proposed Eq. (3) based on a limited 0.80Ig or 0.90Ig. In addition, when the authors’ value of P/fc′Ag =
test series of walls having an hw /lw ratio equal to or greater –0.05, the authors’ Ie will be 0.55Ig instead of 0.40Ig or 0.50Ig.
than 3.0. Also, walls in the references studied by the authors 3. Based on the test series of Reference 20 (high axial
were rectangular, except for References 9 and 12, which load ratio), the Ie becomes equal to Ig, and for a wall tested
cover walls having end boundary elements and T-shaped in Reference 18, the Ie again becomes inconsistent with the
walls. It was very difficult to determine the effective stiffness test results.
of a wall with a T-shape due to the loading patterns. Also, 4. The discusser has studied walls having an hw /lw ratio equal
three walls in Reference 11 have a staggered opening, which to or greater than 3.0, as tested in References 1, 4, 7, 9, 12, and
further complicates the analysis. There are few references18-21 18-21, and the authors presented them in the paper. The
that address the wall hw/lw ratio of 3.0 or higher. following equation was derived by using Reference 22 concepts
2. In Reference 22, the following equation is suggested for
structural walls 24.5 Pu
I e = ---------- + ----------- I ≤ 0.8Ig (6)
fy f c′ A g g
14.5 Pu
I e = ---------- + ----------- I (5) Equation (6) considers not only the steel stress but also the
fy f c′ A g g
width of rectangular walls based on the recommended value
as specified in ACI 318-05.25
It can be seen that authors’ Eq. (3) is inconsistent with Eq. (5). 5. Based on the aforementioned studies, the discusser
3. If structural walls are considered as a wall-column believes that the authors should revisit Eq. (3) and (4) of their
concept,23 the Ie can be considered as outlined in Reference 24 paper so that they become consistent with other research and
and can be presented as follows other test data. The authors’ proposed Eq. (3) or (4) leads to
an overly conservative design for the practicing structural
engineer and will not be a cost-effective design, especially
Effective moment of inertia Ie for high-rise reinforced concrete buildings.
Range Recommended value
P/fc′ Ag ≥ 0.5 0.70Ig to 0.90Ig 0.80Ig REFERENCES
P/fc′ Ag = 0.2 0.50Ig to 0.70Ig 0.60Ig 18. Riva, P., et al., “Cyclic Behaviour of a Full-Scale RC Structural
Wall,” Technical Report N-18/2002, Dip. di Ingegneria Civile, Universitá
P/fc′ Ag = –0.05 0.30Ig to 0.50Ig 0.40Ig da Brescia, Via Branze, Brescia, Italy. (in Italian)

508 ACI Structural Journal/July-August 2008

 19. Lestuzzi, P., et al, “Dynamische Versuche Stahlbetontragwänden University of California, San Diego, CA, 1989.
auf dem ETH-Erdbebensimulator,” Bericht Nr. 240, Institut für Baustatik 23. Nakazawa, A., et al., “Experimental Study on Shear Behavior of R/C
und Kunstruktion (IBK), ETH, Zurich, Basel:Birkhäuser Verlag, 1999. Wall-Columns—Using High-Strength Reinforcement of 13,000 kg/cm2
(in German) Grade—Parts 1-4,” Summaries of Technical Papers, Annual Meeting of
20. Su, R. K. L., et al., “Seismic Behaviour of Slander Reinforced Concrete the Architectural Institute of Japan, Tokyo, 1995-1998, pp. 351-358. (in
Walls under High Axial Load Ratio,” Engineering Structures, V. 24, 2007, Japanese)
pp. 1957-1965. 24. Paulay, T., and Priestley, M. J. N., Seismic Design of Reinforced
21. Carvajal, O., et al., “Muros de Concreto Reforzades con Armadaro Concrete and Masonry Buildings, John Wiley & Sons, Inc., New York,
Minima,” Buletin Technico, Universidad Central de Venezuela, Facultad de 1991, 768 pp.
Ingenieria Ano 21, Enero-Diciembre, pp. 72-73. (in Spanish). 25. ACI Committee 318, “Building Code Requirements for Structural
22. Priestley, M. J. N., and Hart, G. C., “Design Recommendations for Concrete (ACI 318-05) and Commentary (318R-05),” American Concrete
Period of Vibration of Masonry Wall Building,” Research Report SSRP 89/05, Institute, Farmington Hills, MI, 2005, 430 pp.

Disc. 104-S52/From the September-October 2007 ACI Structural Journal, p. 549

Test of High-Rise Core Wall: Effective Stiffness for Seismic Analysis. Paper by Perry Adebar, Ahmed M. M.
Ibrahim, and Michael Bryson

Discussion by James M. LaFave and Dawn E. Lehman

ACI member, Associate Professor, Department of Civil and Environmental Engineering, University of Illinois, Urbana, IL; Assistant Professor, Department of Civil and
Environmental Engineering, University of Washington, Seattle, WA

Structural walls are a commonly-used seismic-force-resisting cracking would likely extend even higher in the prototype
system, and reliable models to simulate their response are structure’s wall than it did in the model test wall. Using the
certainly needed (especially for walls in high-rise structures). presented experimental data for lateral load at first visible
This paper provides important information related to this cracking, and also at the end of the test (when cracking
need, and the authors are to be commended for their valuable extended the equivalent of approximately four stories up the
contribution. Perhaps most important are the results of the model wall), the discussers were able to estimate the actual
scaled model test of a slender concrete structural wall that flexural cracking strength (that is, modulus of rupture) of the
represents part of the core of a high-rise building. The wall concrete used in the wall to be nearly 0.6√fc′ MPa (8√fc′ psi).
had a height-to-length (hw /lw) ratio of more than 7, contained Using that value, the flexural cracking would have most
a relatively modest amount of longitudinal reinforcement, likely extended upward by one or more additional stories if
and was subjected to a constant axial compression force the axial compression had actually decreased up the height
(from top to bottom) of approximately 10% of its gross axial of the wall. Consequently, the existing flexural cracks in the
strength. The lateral loading was applied using an actuator at approximate region of the second to fourth floors would have
the otherwise free top end of the nearly 12 m (39 ft) high probably been wider, the extent of reinforcement yielding
wall, which had a nominally fixed base. near the base of the wall might have been greater, and
Given the 1625 mm (64 in.) length of the test wall and its residual drifts would have been larger.
127 mm (5 in.) web thickness, one might consider this wall The aforementioned discussion can serve as a basis to
model to be at approximately 1/4 of full-scale. It could further explain, in part, why the overall experimental load-displacement
be assumed that the lateral load applied at the top of the wall behavior of the model wall appears to be so similar to that
might represent something akin to the resultant of a linearly normally associated with post-tensioned concrete walls,
varying (that is, inverted triangular) lateral-load distribution such as the hybrid walls described by Kurama26 and the
over the height of a real building (to achieve a reasonably partially prestressed wall tested by Holden et al.27 This has
correct proportion of moment and shear at the base of the perhaps been amplified by the method of application of the
wall). Under this assumption, the test wall model would axial compression load in the authors’ test (that is, via
represent an actual (full-scale) structural wall approaching external bars passing into and centered on the base of the test
18 m (59 ft) in height, which is something on the order of a specimen). This situation can benefit the structural behavior
21-story building. Could the authors briefly comment on in terms of both stiffness and strength along the entire height
whether this assessment is reasonable, and also as to whether of the wall (including minimizing crack formation and crack
an acting (gravity) axial load at the base of 0.1fc′ Ag is typical opening) without the tendency to contribute to any P-Δ
for such a wall or was simply selected for testing expedience? effects in the way that actual gravity loads would.
Assuming that the tested wall represents the bottom 14 At model wall test drifts of approximately 0.5% or more
stories (or so) of a 21-story building, then one can assess the (beyond the onset of flexural yielding), the discussers further
relative values of bending moment, shear, and axial force in wonder whether there could have been any additional contri-
the model compared with the theoretical prototype structure. bution toward system flexibility from shear distortions
Moving up the model wall from its base, the bending occurring near the base of the wall. It has previously been
moment drops linearly, whereas the rate of change in an noted by Kim et al.28 that, even in reinforced concrete structural
actual building could be less, depending on the building’s walls having relatively low shear stress demands and levels
response to any particular ground motion. Furthermore, of diagonal cracking, the shear deformations can contribute
whereas the axial compression in the prototype would be to the flexibility of the system. This deformation mode is
expected to drop almost linearly up the wall (due to the typically triggered after flexural yielding and would
decreasing number of stories above), the axial load in the physically involve slip along the wider flexure cracks,
model remains constant throughout, as does the shear force. resulting in behavior even more pinched than for just flexural
The result of these differences suggests that flexural wall yielding. From the figures in the paper, it does not appear that

ACI Structural Journal/July-August 2008 509

instrumentation to explicitly measure this phenomenon was Earthquake Engineering, Madrid, July 1992, pp. 3303-3308.
used in the test, therefore it is difficult to decipher any 36. Zhang, Y., and Wang, Z., “Seismic Behavior of Reinforced Concrete
Shear Walls Subjected to High Axial Loading,” ACI Structural Journal,
contribution of such shearing deformations in this test for V. 97, No. 5, Sept.-Oct. 2000, pp. 739-750.
comparison with other test data.
Finally, the discussers found some of the recommendations AUTHOR’S CLOSURE
for effective stiffness ratio (Ie/Ig) to deviate from their own The authors appreciate the interest in their paper, which
findings. The second discusser, along with a colleague, has had three main parts. First and foremost was the presentation
studied the effective stiffness as a function of drift from a of results from a test on a large-scale model of a concrete
variety of experimental structural concrete wall data found in shear wall from the core of a high-rise building. The
the literature.5-6,29-36 The considered tests comprise nearly thoughtful questions and comments by discussers LaFave
35 cases, divided roughly evenly between planar rectangular and Lehman regarding the test will be discussed first. The
and barbell wall shapes; nearly half of these tests are the second part of the paper compared the measured wall
same as some of those summarized in the authors’ Table 1. response with a nonlinear flexural stiffness model14 for
The tests examined by the discussers had wall aspect ratios concrete shear walls, and the final part was a summary of
(hw /lw) ranging from 1 to more than 6, with axial compression work that had previously been done15 on estimating effective
ratio values (P/fc′ Ag) between zero and approximately 0.35. flexural rigidity to account for cracking of concrete shear
For these cases, Ie/Ig was seldom more than 0.6 for drifts walls when using linear dynamic analysis. The comments on
beyond approximately 0.3%, nor more than approximately this topic will be briefly discussed at the end.
0.3 for drifts greater than 0.9%, with the exceptions being a The prototype wall was assumed to be 73.2 m (240 ft)
handful of cases where the axial compression ratio was high, and to have a height-to-length ratio of 11. The test wall
larger than 0.2fc′ Ag. These trends would tend to indicate was assumed to be approximately 1/4-scale, but rather than
somewhat lower effective stiffness values than those applying a varying lateral load over an 18.3 m (60 ft) high
presented by the authors in Fig. 12(c) of their paper. specimen, the resultant lateral point load was applied at
The discussers appreciate the great challenges and approximately 2/3 that height, and the height of the specimen
compromises that must be made in conducting large-scale was reduced accordingly. The level of axial compression,
laboratory tests and in then applying the results to practice. 10% fc′ Ag, was selected because it is a typical value of axial
Therefore, all of the aforementioned comments are made compression due to gravity loads in high-rise concrete walls.
primarily in an attempt to address a concern that structural It is true that the actual shear force and bending moment
designers might end up with a bit too optimistic of a view diagrams in the prototype wall could be different than what was
regarding the equivalent effective flexural stiffness of this applied in the test and depends on the building’s response to any
sort of a wall in practice, especially if these results are indeed particular ground motion. The bending moment variation
to be used across the full spectrum of stated design applications associated with a first mode distribution of lateral load is of
from push-over analysis to linear dynamic analysis. On the other particular interest as it causes maximum top wall displacements.
hand, actual walls could perhaps exhibit somewhat greater The bending moment variation from a point load applied at
energy dissipation than that seen in this test. approximately 3/4 of a cantilever wall height and a first mode
distribution of lateral load are almost identical from the base
REFERENCES of the wall to approximately midheight.
26. Kurama, Y. C., “Hybrid Post-Tensioned Precast Concrete Walls
for Use in Seismic Regions,” PCI Journal, V. 47, No. 5, Sept.-Oct.
In an actual high-rise wall, bending moments and shear
2002, pp. 36-59. forces vary in a complex way, both over the height of the wall
27. Holden, T.; Restrepo, J.; and Mander, J. B., “Seismic Performance of and from one instance to the next. Nonlinear analysis indicates
Precast Reinforced and Prestressed Concrete Walls,” Journal of Structural the shear force (bending moment gradient) will reverse
Engineering, ASCE, V. 129, No. 3, Mar. 2003, pp. 286-296. direction a number of times while the bending moment at the
28. Kim, T.-W.; Foutch, D. A.; and LaFave, J. M., “A Practical Model for
Seismic Analysis of Reinforced Concrete Shear Wall Buildings,” Journal
base of the wall is relatively constant. The objective of the
of Earthquake Engineering, V. 9, No. 3, May 2005, pp. 393-417. slow cyclic test on a large test specimen was not to try to
29. Lefas, I. D., and Kotsovos, M. D., “Strength and Deformation simulate this complex behavior, which can only be done on
Characteristics of Reinforced Concrete Walls under Load Reversals,” ACI a shake table, but to measure the fundamental flexural
Structural Journal, V. 87, No. 6, Nov.-Dec. 1990, pp. 716-726. behavior of a structural concrete wall with minimal contribution
30. Pilakoutas, K.; Elnashai, A. S.; and Ambraseys, N. N., “Earthquake
Resistant Design of Reinforced Concrete Structural Walls,” Report No.
from shear deformations. The bending moment-curvature
ESEE 4/91, Imperial College, London, UK, Apr. 1991. relationship that was measured during the test is that
31. Sittipunt, C.; Wood, S. L.; Lukkunaprasit, P.; and Pattararattanakul, fundamental property. The constant axial compression over
P., “Cyclic Behavior of Reinforced Concrete Structural Walls with Diagonal the height of the wall was not only easier to apply in the test,
Web Reinforcement,” ACI Structural Journal, V. 98, No. 4, July-Aug. but also meant that all bending moment and curvatures
2001, pp. 554-562.
32. Thomsen, J. H., and Wallace, J. W., “Displacement-Based Design of measured over the height of the wall were part of the same
Slender Reinforced Concrete Structural Walls—Experimental Verification,” bending moment-curvature relationship for the wall.
Journal of Structural Engineering, ASCE, V. 130, No. 4, Apr. 2004, It is certainly reasonable to question whether applying a
pp. 618-630. constant axial compression (rather than a linearly varying
33. Vallenas, J. M.; Bertero, V. V.; and Popov, E. P., “Hysteretic Behavior
of Reinforced Concrete Structural Walls,” Report No. UCB/EERC-79/20,
axial compression) had a significant influence on the load-
Earthquake Engineering Research Center, University of California, displacement behavior of the wall. The authors believe that it
Berkeley, CA, Aug. 1979, 266 pp. did not because the height of wall significantly contributing to
34. Wang, T. Y.; Bertero, V. V.; and Popov, E. P., “Hysteretic Behavior of wall drifts was small compared with the 18.3 m (60 ft) height over
Reinforced Concrete Framed Walls,” Report No. UCB/EERC-75-23, which the axial compression should have been reduced to zero.
Earthquake Engineering Research Center, University of California,
Berkeley, CA, Dec. 1975, 367 pp.
Due to the maximum lateral load at 2% drift, approximately
35. Yanez, F. V.; Park, R.; and Paulay, T., “Seismic Behaviour of Walls 15 flexural cracks occurred over a height of 3.6 m (11.8 ft)
with Irregular Openings,” Proceedings of the Tenth World Conference on above the base. Using this information, and assuming a

510 ACI Structural Journal/July-August 2008

concrete tensile strength in bending of 4.2 MPa (0.61 ksi), due to the P-Δ effect of gravity loads, and the load-deformation
additional flexural cracking would have occurred over a response of the test wall would have been very flat.
height of 1.0 m (3.4 ft) if the axial compression had reduced An important question is why the load-deformation
linearly from 10% fc′ Ag at the construction joint to zero at response of the current test wall was so different than has
18.3 m (60 ft) above the base. As the flexural cracks had an been observed with other reinforced concrete walls and so
average spacing of 240 mm (9.5 in.), four additional flexural similar to what has been observed with post-tensioned
cracks would have occurred in the wall. In judging whether these concrete walls. The explanation is not in the details of how
additional cracks would have significantly influenced the load- the test was conducted, but is the result of an important property
deformation response of the wall, it is important to note that of the wall—namely, the relative amount of flexural resistance
many of the upper flexural cracks in the wall were small under provided by axial compression versus bonded vertical
the maximum lateral load (refer to Table 3) and completely reinforcement. This explains the flat load-deformation
closed when the lateral load was reduced slightly. That is, these response discussed previously, and explains the pinching of
cracks were closed for most of the load-deformation response of hysteresis loops causing reduced energy dissipation. The
the wall. Most of the deformation of the wall was the result of latter behavior is related to the bending moment at which
approximately half the cracks over 2 m (6.6 ft) of the wall height flexural cracks near the base of the wall close due to the axial
(refer to Fig. 8 and 9). Over this height, the axial compression compression. An explanation follows.
should have been reduced by only 11%. If there had been no vertical reinforcement in the test wall,
LaFave and Lehman are correct that the method of preexisting flexural cracks in the wall would open and close
applying axial compression load in the test by external bars (the wall would rock) at an overturning moment equal to
attached to the top of the wall and passing through the base approximately (P/A)S = 857 kNm (632 kip-ft), which for the
of the wall did not model the P-Δ effect of gravity loads, and critical crack at the construction joint corresponds to a lateral
provided some artificial increase in stiffness and strength of load of 75.6 kN (17.2 kip). Vertical reinforcement in the wall
the wall. The question again is how significant was the error that has yielded in tension resists closing of the cracks and
introduced by the simplifications that were necessary to must yield in compression before the flexural cracks will
conduct the test in a laboratory. As drifts are small near the completely close. If the yield force of the vertical tension
base of cantilever walls where axial compression is large, the reinforcement is large in relation to the axial compression
P-Δ effect is generally much smaller in cantilever walls force, the flexural cracks will not close until the bending
compared with frames. The external bars, which passed moment reverses and flexural compression is applied to the
through large sleeves in the base of the wall, had an overall tension reinforcement. This behavior, which results in significant
length of 15 m (50 ft). Throughout most of the test, the bars residual displacements under zero lateral load, has been
seemed to pass freely through the base. That is, they did not observed in most previous tests of reinforced concrete
appear to be touching the sides of the sleeves. Using the structures. In the current test wall, the yield force of all
overall length of the external bars, the horizontal component vertical reinforcement in the tension flange and web is equal
to 410 kN (93.5 kip), which is only 27% of the applied axial
of the 1500 kN (342 kip) force in the external bars would be
compression force at the base of the wall. As the centroid of
4.6 kN (1.05 kip) at a top wall displacement of 46 mm
this vertical reinforcement is located at 0.4 m (1.3 ft) from
(1.8 in.) when significant yielding of the vertical reinforcement
the wall centerline, an axial compression force of 410 kN
occurred. This corresponds to 3.5% of the applied lateral
(93.5 kip) applied at the wall centerline and bending moment
load. Thus, this effect does not appear to have been significant in
of 164 kNm (123 kip-ft) is required to yield this reinforcement.
the elastic range of the wall. Thus, the overturning moment at which flexural cracks are
After yielding of the vertical reinforcement, the lateral predicted to close in the test wall is (1 – 0.27) × 857 – 164 =
displacement increased without a significant increase in 462 kNm (342 kip-ft). For the lowest flexural crack at the
lateral load. Thus, the percentage increase in lateral load due construction joint, this corresponds to a lateral load of 41 kN
to the horizontal component of the external bars used to (9.3 kip). As indicated by the change in slope of the
apply axial compression does increase. Also, it is possible unloading curves shown in Fig. 5, the flexural cracks were all
that the external bars did contact the sides of the sleeves at closed when the lateral load was less than approximately 40 kN
very large top wall displacements. At the maximum drift, the (9 kip). The residual displacements remaining after the
maximum lateral load applied to the wall was 18% larger flexural cracks closed were primarily due to misalignment of
than predicted based on the measured yield strength of the damaged crack surfaces.
vertical reinforcement; this increase, however, is partly due LaFave and Lehman question whether there could have
to strain hardening of vertical reinforcement. As the ultimate been any contribution toward system flexibility from shear
strength of the vertical reinforcement was 40% larger than deformations. The test was designed to measure flexural
the yield strength, and calculations indicate 30% of the over- response of a concrete shear wall with minimum contribution
turning moment was resisted by vertical reinforcement, the from shear deformations. As the maximum shear force that
maximum strength increase due to strain hardening was was expected to be applied in the test was considerably less
12%. Thus, it appears that the method of applying axial than the concrete contribution Vc for a member subjected to
compression in the test increased the wall strength by at least an axial compression of 10% fc′Ag, significant diagonal
6% and likely approximately 10% at maximum drift. This cracking was not expected in the test, and no instrumentation
error seems acceptable considering how difficult it would was installed to measure shear deformations. As reported in
have been to better simulate the influence of gravity loads. the paper, a surprising amount of flexure-shear diagonal
The most important conclusion from this discussion is that if cracking was observed in the web of the wall considering that
gravity loads had been exactly simulated, the small strength the maximum shear stress V/blw was only 0.78 MPa (113 psi)
increase due to strain hardening of vertical reinforcement and the axial compression stress P/A was 4.9 MPa (710 psi). The
would have been eliminated by the small strength reduction deformations at these cracks were carefully observed during

ACI Structural Journal/July-August 2008 511

the test (for example, measured crack widths are reported in curve, nonlinear dynamic analysis was used to determine the
Table 3). Very little shear displacement of the crack surfaces maximum displacement of a variety of concrete walls
was observed. Thus it was concluded that the diagonal cracking subjected to a variety of ground motions. The effective stiffness
did not influence the overall response of the wall other than is that which gives the correct estimate of average maximum
increasing the spread of inelastic curvatures (that is, increasing wall displacement using linear dynamic analysis. The nonlinear
plastic hinge length). load-deformation response of the walls was determined using
It is important to note that, when shear deformations the experimentally calibrated nonlinear flexural model.14
contribute to the flexibility of a wall, the flexural stiffness of The results of this recent work indicates that the effective
the wall is actually larger than what one would estimate by stiffness ratio (Ie/Ig) of a concrete shear wall is seldom less
ignoring this contribution from shear deformation. Most
than 0.5, and for tall walls with long initial fundamental
other wall tests that have been done to date have had significant
periods, is seldom less than 0.6. In addition to the characteristics
contributions from shear deformations. Because of the
of the ground motion, and initial period of the structure, the
strong influence of shear on the flexural compression stress
distribution and, hence, magnitude of flexural tension, it is ratio of maximum elastic force demand to strength of the
difficult to completely separate the influence of shear from wall (R factor) is an important parameter that influences
flexure. It is for this reason that tests on very slender walls, effective stiffness. The level of axial compression has much
such as the current test wall, should be used for estimating less influence than previously thought. Axial compression
the flexural stiffness of high-rise concrete shear walls. does delay the point that the loading curve becomes nonlinear;
but as discussed previously, if the flexural resistance of a wall is
Effective stiffness for linear dynamic analysis primarily provided by axial compression in concrete rather than
The final part of the paper was a summary of work that had tension in vertical reinforcement, much less energy is dissipated.
previously been done by the authors15 on what effective flexural This increases maximum wall displacements, which reduces
rigidity should be used in a linear dynamic analysis (such as effective stiffness. The stress at which the vertical reinforcement
response spectrum analysis) to account for cracked sections. yields is not an important parameter.
The approach taken in Reference 15 to arrive at the Discusser Solanki believes the authors should revisit Eq. (3)
recommendations summarized in the current paper was to and (4) because they are inconsistent with the recommendations
use the results from a fiber model to develop a simple for effective stiffness that he has summarized. First, the
nonlinear flexural model14 for typical high-rise concrete recommendations he has summarized are themselves
shear walls. This nonlinear model was verified by comparing inconsistent, as Eq. (5) gives a very different result from Eq. (6).
predictions with the results of the large-scale test presented Second, the lower-bound effective stiffness given by Eq. (4) is
in the current paper. The nonlinear flexural model was then consistent with Eq. (5) at low levels of axial compression, and is
used to predict the load-displacement response of many consistent with Eq. (6) at high levels of axial compression for a
different high-rise walls, and the results of these were used wall, which is no more than approximately 30% fc′ Ag. The
to develop the recommendations for effective stiffness. statement by discusser Solanki that Eq. (3) and (4) lead to
A fundamental assumption used in the previous work is that overconservative designs that are not cost-effective is
effective stiffness of a nonlinear member is the slope of the
nonsensical. The critical issue in the design of high-rise
elastic portion of an equivalent elastic-plastic load-displacement
concrete walls is maximum drift demands. Equations (3)
relationship that has the same area under the curve as the actual
and (4) give higher effective stiffnesses than the other
nonlinear load-displacement relationship. Other authors have
used other ways to determine effective stiffness such as recommendations for typical values of axial compression
assuming that the effective stiffness is equal to the secant in high-rise concrete walls, which results in lower
stiffness to some predefined point on the nonlinear load- maximum drift demands for a given structure or less
displacement relationship. Unfortunately, none of the discussers concrete structure for a given maximum drift.
have indicated how they determined effective stiffness. There were two typographical errors in the paper: 1) the
The first author and his recent students have continued the column in Table 1 entitled “Axial composition” should be
work on effective flexural stiffness of high-rise concrete “Axial compression”; and 2) the definition of the parameter
shear walls. Rather than use a simple definition of effective a in Eq. (2) given immediately below the equation should be
stiffness based on the shape of the nonlinear load-deformation a = 1.1(Icr /Ig)–0.4.

Disc. 104-S57/From the September-October 2007 ACI Structural Journal, p. 601

Justification of ACI 446 Proposal for Updating ACI Code Provisions for Shear Design of Reinforced Concrete
Beams. Paper by Zdenek P. Bažant, Qiang Yu, Walter Gerstle, James Hanson, and J. Woody Ju

Discussion by Evan C. Bentz

ACI member, Associate Professor, University of Toronto, Toronto, ON, Canada

The authors are to be congratulated for writing a compre- fully agree with the method being proposed, he welcomes
hensive paper that summarizes the opinions of ACI the increased attention to the issue that should result from
Committee 446 on fracture mechanics approaches to shear this paper. The discusser has the following questions and
design. First, it must be noted that the discusser agrees that comments on the paper:
the size effect should be included in the shear design 1. The authors state that it is inappropriate to trust empirical
methods of the ACI Code. Whereas the discusser does not relations for the effect of depth on shear strength. Despite this,

512 ACI Structural Journal/July-August 2008

they seem willing to accept empirical relations for the effect or REFERENCES
ρ, a/d, fc′ , and all other variables. Is this not inconsistent? 45. Bentz, E. C., and Buckley, S., “Repeating a Classic Set of Experiments
2. The results of the 1991 shear tests of Bažant and on Size Effect in Shear of Members without Stirrups,” ACI Structural
Journal, V. 102, No. 6, Nov.-Dec. 2005, pp. 832-838.
Kazemi4 are clearly important to the decision to apply a fracture 46. Bentz, E. C., “Empirical Modeling of Reinforced Concrete Shear
mechanics approach to shear. Is it appropriate, therefore, to Strength Shear Size Effect for Members Without Stirrups,” ACI Structural
neglect to mention the recent 2005 paper45 that presents the Journal, V. 102, No. 2, Mar.-Apr. 2005, pp. 232-241.
results of repeat tests that were inconsistent with the original 47. Collins, M. P.; Bentz, E. C.; and Sherwood, E. G., “Where is Shear
1991 tests? The authors refer to three of the discussions to Reinforcement Required? A Review of Research Results and Design
this 2005 paper21-23 but not the paper itself. It would seem Procedures,” ACI Structural Journal, V. 105, No. 5, Sept.-Oct. 2008.
unfair for the readers to be kept in the dark concerning questions
about the repeatability of these experimental results. AUTHORS’ CLOSURE
3. The paper includes an argument that the 1991 Bažant The authors deeply appreciate the thoughtful and stimulating
and Kazemi4 experiments should be included in the ACI 445-F questions of discusser Bentz, which provide an opportunity to
database used in this paper. For inclusion in this database,18 clarify points that were not specifically addressed in the
member width needed to be at least 2 in. (50 mm). The paper. His six points are answered as follows:
authors state that the 1991 beams missed this limit by a 1. The reason why empirical relations can be used for the
“mere 4%,” as they were 1.90 in. (48 mm) wide. Bažant and effects of steel ratio ρw, relative shear-span ratio a/d, and
Kazemi4 reported that their beams were 1.5 in. (38 mm) concrete compressive strength fc′ , but not for the effect of
wide, not 1.90 in. (48 mm) wide, as stated by the authors. size, is the difference between interpolation and extrapolation.
Thus, the member width limit for inclusion in the database Interpolation of experimental data can provide sufficient
was missed by over 30%, not 4%. It is disconcerting that a accuracy, but extrapolation cannot. Extrapolation can be
paper by Bažant and colleagues would accidentally base an trusted only if it is based on a good theory that must, of
argument on factual errors about their own experiments. course, be verified by experiments.
4. It is stated that there are only 11 size effect series of tests The experiments can be of different kinds, for example,
available for comparison with the fracture mechanics method. reduced scale tests with reduced aggregate size and tests of
A recent paper,46 however, includes 24 size effect series. specimens with different geometries. For ρw, a/d, fc′ , and da,
Recent work on a larger database47 has uncovered additional test beam shear tests can, and have been, conducted to cover the
series, giving a total of 38 published size effect series on shear entire practical ranges of these variables, sampling them with
strength. The authors may wish to expand their database and see an almost uniform statistical density of distribution. But for
if their conclusions apply to all published size effect tests. the effect of size, unfortunately, the existing database has an
5. The discusser is concerned with the argument being enormous statistical bias for small sizes. The practical range
made in relation to Fig. 5. Because the value of d0 is a statistical of interest extends from depth d ≈ 50 mm (19.7 in.) to at least
curve-fit parameter, the same individual set of experiments 15 m (or 50 ft). For example, the depth of the girder of the
could show up as fully consistent with the theory of fracture Koror-Babeldaob Bridge in Palau, which collapsed, was
mechanics whether they showed no size effect, a medium 14.2 m (or 46.5 ft). In the ACI 445 database,18 however, 86%
size effect, or a very strong size effect. Once natural of all the data pertain to d < 0.5 m (19.7 in.), 99% to d < 1.1 m
experimental scatter is included, it would appear that just (43.3 in.), and 100% for d < 2.0 m (79 in.).
about any set of experimental points measured with respect Therefore, the design for d > 0.5 m (19.7 in.) represents
to size can be declared fully consistent with the theory of predominantly, and for d > 1 m (39.4 in.) almost totally, an
fracture mechanics. This would mean that the theory is non- extrapolation. It is because of the cost of large-scale tests that
falsifiable in terms of its application to the shear strength of beams with d > 3 m (118.1 in.) have never been tested to
reinforced concrete beams. To demonstrate that the theory is failure, and generally for any size d > 1 m (39.4 in.), it is next
falsifiable and, therefore, a productive scientific theory, what to impossible to conduct failure tests to cover the entire range
predictions can the authors make based on the fracture of ρw, a/d, fc′ , and da. On the other hand, for small beams,
mechanics approach to shear that could be used to test the the ACI 445 database does cover the entire range of these
theory experimentally? influencing variables, and does so with reasonable uniformity.
6. Figure 8 shows the results of a fracture mechanics-based So, the semi-empirical relations for the effect of these four
finite element (FE) model predicting the shear strength of parameters, which are based mainly on the research of
members with stirrups based on a test from Toronto.11 The Shioya and Akiyama,7 Kani,8 Reineck et al.,18and Pauw,48
discusser is surprised at the predicted shear strength values are adequate and can be trusted. For the effect of size d,
shown in Fig. 8(b). Scaling off the figure, the theoretically empirical formulas could be trusted for d < 0.5 m (19.7 in.),
predicted shear stress at shear failure on the top part of the which is still interpolation, but not for extension to larger
figure is 189 psi (1.30 MPa). Is it a coincidence that the sizes d, which represent extrapolation, yet are of the main
ACI Code also predicts the same shear stress at failure of interest for practice. A well founded, experimentally verified
2√4870 + 50 = 189 psi (1.30 MPa)? It appears that the finite theory is the only one to trust.
element results have been curve-fit to match the ACI Code 2. There was no intention to “keep the readers in the dark”
for the particular details of the Toronto test. Should the reader about the Bentz’s 2005 paper.45 Although this paper
interpret this as a vote of validation for the ACI Code method presented very valuable new test results, it was not referenced
for members with stirrups? Is not the relatively complex because, if it were, some extraneous comments would have
statistical discussion on the page leading up to this figure been necessary. In particular, it would have been necessary
invalidated by having theoretical FE results in the paper to explain why the interpretation of the 1991 tests of Bažant
showing that the fracture mechanics approach gives the same and Kazemi4 was misleading, and the length limit of ACI
answers as the existing ACI Code for members of a large papers did not allow for adding such an explanation. Contrary to
depth with stirrups? what was claimed, the experiments that Bentz reported in that

ACI Structural Journal/July-August 2008 513

paper were not a realistic reproduction of the 1991 experiments, theoretically unjustifiable, thermodynamically impossible,
as explained in the cited discussion of his paper. and has no experimental support.
3. As Bentz pointed out, the width of the 1991 test specimen 6. The discusser is not right in saying that “the finite
was indeed 38 mm (1.5 in.),4 not 48 mm (1.90 in.), and we element results have been curve fit to match the ACI Code,”
regret the error. However, Bentz’s argument for excluding nor is he right in regarding the shear strength values in Fig. 8(b),
these tests from the database is nonetheless invalid. It has computed for beams with stirrups by a fracture mechanics-
been well documented that the beam width increase has a based finite element code, as a “vote of validation of ACI
negligible effect on the shear strength if it exceeds approximately Code method for beams with stirrups.” The problem is that
four aggregate sizes, and the width of the 1991 test specimens he, like most researchers in this field, overlooked the covert
was eight aggregate sizes. Therefore, the beams 38 mm safety margin that is hidden in the ACI Code provisions (as
(1.5 in.) wide must have given the same results for shear discussed in detail in Reference 41), and is not uncovered unless
strength vc as beams 254 mm (10 in.) wide, and thus there one carefully examines the database used in setting up the code
was no reason to exclude them from the ACI database. This equation. The safety margins separating the ACI Code equation
exclusion has been especially unfortunate for the advocates from the experimental database must be the same as those
of fracture mechanics-based theory of size effect on beam separating the Toronto experiments from those equations (or
shear strength because these reduced-scale tests provide the else that covert safety margin would be unnecessary and could
best experimental support for the applicability of this theory. be removed from the ACI Code or, in other words, the factor of
2 in the formula vc = 2 f cr′ could be increased).
4. Bentz claims that 24, rather than 11, size effect test
series are available for comparison to the fracture The finite element calculations were based on the mean
mechanics-based theory.46 From careful examination, however, compression strength of concrete (which was reported as fc =
it transpires that his expansion of the size-effect database is 33.6 MPa [4870 psi]). The discusser does not take into
mainly due to questionable loosening of the requirement for account that the ACI Code equation vc = 2 f cr′ was not set
geometric similarity and questionable narrowing of the to match the mean of the database points (as seen in Fig. 1 of
requirement for breadth of the size range. The required range is the paper). Rather, it was set to lie at the lower margin of this
database, representing a 5% probability cutoff (as calculated
proportional to the scatter band width (in the log-log plot); it
from a Gaussian distribution of vc fitted to the database). This
must exceed it by a factor of at least 6 (this follows from the
cutoff equals 65% of the mean vc (as marked in Fig. 1 of the
requirement that ωy|x/ωx < 0.1549). Otherwise, the regression
paper; for more detail, refer to Fig. 1(a) of Reference 41).
line has a great error.49,50 In the size effect tests used in the
paper, most of the test series have a size range greater than Consequently, the value of vc = 0.96 MPa (140 psi), which
1:8, which is barely sufficient to obtain unambiguous results. Bentz calculated from fcr′ according to the ACI Code equation,
In the size effect tests collected by Bentz, over half of the test is not the expected experimental value of vc (in the sense of the
expectation E in statistics). Rather, the expected experimental
series have a size range close to 1:4, which is statistically
value is vc = 0.96/0.65 = 1.48 MPa (215 psi), and this value is
insufficient for determining the size effect. The acceptance
1.54 (= 1/0.65) times higher than the experimental result in the
of data with insufficient size range is what led Bentz to
Toronto tests. Thus it must be concluded that the ACI Code
questionable conclusions about the size effect.46
equation, when applied to the beams of the size tested in the
5. Regarding Fig. 5, it must be emphasized that d0 is not a Toronto tests, overpredicts vc by 54%.
curve-fitting parameter. Based on fracture mechanics, d0 can The value of vc is, of course, random. Were it possible to
be calculated from the dimensionless energy release function test not one but hundreds of beams of the same large depth,
g(α) (obtainable from handbooks) and the material characteristic with many different concretes, different shear spans, and
length identified from tests of fracture specimens of different different longitudinal and stirrup steel ratios, covering the
sizes. To calculate function g(α) for shear of a reinforced entire ranges of these variables, surely a statistical distribution
concrete beam, a complete analysis of propagation of diagonal, of vc with a large standard deviation would be found. To
flexural, and compressive shear cracks would be needed. suppose that the Toronto result lies below the 5% cutoff of
This is a difficult problem, not yet solved. It is for this reason this undocumented distribution, however, would be dangerous,
that the size effect formulation proposed in the paper could wishful thinking.
not be based on a theory in its entirety. To get a clue where the vc value from the Toronto large
Bentz is right that data exhibiting size effects of different beam test is positioned on this distribution, one would need
slopes in the log-log plot could be matched to the proposed to carry out, with the same concrete, geometrically similar,
size effect equation just by adjusting d0, so that the data reduced-scale tests for depth d between 0.152 and 0.254 m
would get positioned on the portion of the size effect curve (6 and 10 in.). Within that range, many test results are available
that has the appropriate slope. Such ambiguous, or falsifiable, to determine the statistical distribution of vc (which may be
matching, however, is possible only for test data of limited size assumed to be Gaussian), and many more can be generated at
range (<1:4). The match becomes unique for data of a broad low cost. It may now be checked to which cutoff probability the
enough size range for which the transition from a near-zero slope reduced-scale test corresponds on this distribution (marked
to the slope of –1/2 in a log-log size effect plot is clearly visible. in Fig. 4 of Reference 51 by distance a from the mean). It is
Data whose range is broad enough to show the transition from then logical to assume that the Toronto large beam test would
zero slope to slope –1/2 cannot be fitted by other proposed size correspond to the same probability cutoff (distance a from
effect equations, including the power law of exponent –1/3 (as the mean) on the distribution of vc if it could be tested for that
proposed in ACI 445) or –1/4 (according to the JSCE equation). size. In this way, the error of the ACI Code equation could be
Also, they certainly cannot be fitted by the modified assessed. For beams without stirrups, this was demonstrated
compression field theory, which gives a size effect curve in Reference 51. It must be concluded that the ACI Code
terminating, in the log-log plot, with the slope of –1 (that is, a equation overpredicts the shear stress vc in the large beam
power law of exponent –1). Such an excessive slope is by almost 54%. This is serious indeed. So the calculation

514 ACI Structural Journal/July-August 2008

in the paper is not a “vote of validation for the ACI Code 49. Bažant, Z. P., and Pfeiffer, P. A., “Determination of Fracture Energy
method.” Rather, it raises a serious question about the factor from Size Effect and Brittleness Number,” ACI Materials Journal, V. 84,
No. 6, Nov.-Dec. 1987, pp. 463-480.
of safety of the ACI Code equation when applied to large 50. Bažant, Z. P., and Planas, J., Fracture and Size Effect in Concrete and
beams with stirrups. If the size effect is ignored, the safety Other Quasibrittle Materials, CRC Press, Boca Raton, FL, (textbook and
margins are reduced. reference volume, 616 + xxii pp.).
51. Bažant, Z. P., and Yu, Q., “Consequences of Ignoring or Misjudging
the Size Effect in Concrete Design Codes and Practice,” Concrete
REFERENCES Technology (Taiwan), V. 1, No.1, 2007, pp. 29-55 (authorized republication,
48. Pauw, A., “Static Modulus of Elasticity of Concrete as Affected by with updates, from Proceedings of the 3rd Structural Engineers World
Density,” ACI JOURNAL, Proceedings V. 57, No. 12, Dec. 1960, pp. 679-688. Congress, Bangalore, 2007).

Disc. 104-S58/From the September-October 2007 ACI Structural Journal, p. 611

Investigation of Deep Beams with Various Load Configurations. Paper by Michael D. Brown and
Oguzhan Bayrak

Discussion by Dipak Kumar Sahoo, Bhupinder Singh, and Pradeep Bhargava

Research Scholar, Indian Institute of Technology, Roorkee, India; Assistant Professor, Indian Institute of Technology; and Professor, Indian Institute of Technology

The authors are to be complimented for their comprehensive The order, as seen in Table 1 and Fig.7, is
and elaborate investigation of the behavior of deep beams
under various load configurations. In the context of the UL-0-8.5 < UL-0-0 < UL-17-17 < UL-8.5-0b < UL-17-0 < UL-8.5-0a
investigation, the authors would like to respond to the The order, as per the normalized ultimate loads presented
following issues: in Table 3, is
1. At the outset, attention is drawn to what seem to be
printing errors in the 7th column of Table 1. The south UL-0-0 < UL-0-8.5 < UL-17-17 ≈ UL-8.5-0b ≈ UL-17.-0 < UL-8.5-0a
reaction for Specimen UL-0-8.5 should be 204 kN (45.9 kip)
instead of 2040 kN (458.6 kip) and that for Specimen CL-8.5-0 The authors may like to address this inconsistency in
should be 124 kN (27.9 kip) instead of 1240 kN (278.8 kip). specimen behavior.
b. A comparison between Specimens CL-0-0 and CL-8.5-0
2. For a more realistic assessment and comparison of the
shows that normalized ultimate load and normalized cracking
test results, normalized ultimate load (ultimate load per unit
load of the former having no web reinforcement are
concrete strength), normalized cracking load (cracking load
significantly higher than those of the latter satisfying web
per unit concrete strength), and effective transverse reinforcement
reinforcement requirements of Eq. (A-4), in ACI 318-05,
ratio calculated on the basis of the corrected Eq. (A-4) of Appendix A.1 The authors have attributed this aberration to
ACI 318-05, Appendix A, have been tabulated in Table A. the large amount of scatter usually associated with results of
Keeping in mind that the strut efficiency is not expected to shear behavior of concrete beams. The ratios of cracking
significantly vary with concrete strengths in the range of 16.3 load to ultimate load for Specimens CL-0-0 and CL-8.5-0,
to 22.3 MPa (2364.1 to 3234.3 psi) (Table 1), the following however, are 0.89 and 0.51, respectively (Fig. 7). The relatively
observations are made with respect to the normalized loads of lower ratio of 0.51 in the case of Specimen CL-8.5-0 indicates
Table 3: that web reinforcement has significantly improved the
a. Among the specimens with distributed loading, the post-cracking load-carrying capacity of this specimen.
ascending order based on ultimate strengths with reference to c. On the basis of the observed shear behavior of
ACI-STM provisions for strut efficiency factors ought to be Specimens UL-0-0 and UL-0-8.5, the authors state that the
horizontal shear reinforcement did not positively affect the
UL-0-0 < UL-17-0 < UL-17-17 = UL-0-8.5 = UL-8.5-0a = UL-8.5-0b shear strength of the specimens, and the large variation in

Table A—Normalized ultimate and cracking loads

Effective transverse reinforcement Effective transverse reinforcement
ratio, as per ACI 318-05, Eq. (A-4) ratio (corrected)
A A 2

Specimen ID
Concrete strength,
MPa (psi)
= ∑ --------
bs si
si
sin α i = ∑ --------
bs si
si
sin α i Normalized ultimate Normalized cracking
load, kN/MPa (lb/psi) load, kN/MPa (lb/psi)
UL-8.5-0a 16.8 (2440) 0.0031 0.0022 56.9 (88.3) 17 (26.4)
UL-8.5-0b 18.2 (2640) 0.0031 0.0022 46.4 (72) 17.1 (26.5)
UL-0-0 22.3 (3230) 0 0 36.6 (56.8) 16.7 (25.9)
UL-0-8.5 18.2 (2640) 0.0031 0.0022 41.5 (64.4) 25.3 (39.3)
UL-17-17 18.3 (2660) 0.0031 0.0022 46.1 (71.5) 20.4 (31.7)
UL-17-0 18.3 (2660) 0.0016 0.0011 46.5 (72.2) 20.1 (31.2)
2C-8.5-0 22.1 (3210) 0.0031 0.0022 33.3 (51.7) 8.4 (13)
2C-0.0 22.1 (3210) 0 0 24.1 (37.4) 13.8 (21.4)
CL-8.5-0 17.8 (2580) 0.0034 0.0024 26.7 (41.4) 13.6 (21.1)
CL-0-0 16.3 (2370) 0 0 34.5 (53.5) 22.6 (35.1)

ACI Structural Journal/July-August 2008 515

ultimate strength between these two specimens is ascribed to mentioned by the discussers differ only in the order of the
the differences in concrete strengths. The ultimate loads two weakest specimens. The authors are not aware of any
reported for Specimens UL-0-0 and UL-0-8.5 in Table 1 inconsistency among the specimens. As indicated in the
are 817 and 755 kN (184 and 170 kip), respectively, but paper, the authors suggest that only small amounts of transverse
when these loads are interpreted in terms of normalized ultimate reinforcement are necessary to enhance the failure loads of
loads, the corresponding values for Specimens UL-0-0 and the specimens. Hence, specimens without transverse
UL-0-8.5, as reported in Table A, are 36.6 and 41.5 kN/MPa reinforcement failed at loads less than those with transverse
(56.8 and 64.4 lb/psi), respectively, which clearly reflect the reinforcement. That point is evident in either of the two
influence of horizontal reinforcement on the shear strength of lists of specimens presented by the discussers.
the specimens. Kong et al.11 have reported that, depending on b. Specimen CL-8.5-0 failed at an ultimate load less than
span-depth ratio L/D and shear span-depth ratio a/D, both Specimen CL-0-0. The authors are unaware of the basis for
vertical as well as horizontal web reinforcement influence the discussers’ claim that Specimen CL-8.5-0 had a “signif-
the load-carrying capacity of deep beams. icantly improved” post-cracking strength.
3. It is appreciated that reinforcement will not have a
c. The greatest shear span-depth ratio of the specimens
significant influence on diagonal cracking load. Because,
tested by Kong et al.11 was 0.7. This value is much smaller
prior to cracking, shear is resisted by concrete alone, the
diagonal cracking load should be dependent on the crushing than any of the specimens tested by the authors. Due to
strength of concrete. With reference to the data in Fig. 8, dissimilarity in the specimens, it is unclear how to compare
although the concrete crushing strength of Specimen CL-0-0 the differing conclusions regarding the effectiveness of
is 26% less than that of Specimen 2C-0-0, the former has horizontal reinforcement. It should be noted, however, that
21% higher diagonal cracking load compared with the latter. the uniformly loaded specimen with only horizontal web
4. It would be instructive if the authors could clearly reinforcement (UL-0-8.5) ranked near the weakest specimen
identify which of the following failure modes mentioned in in either of the discussers’ lists in the previous comment.
the paper correspond to which of the beams used in their These results indicate that horizontal web reinforcement
investigation: a) diagonal tension; b) crushing of concrete; appears to be less beneficial than vertical web reinforcement.
and c) splitting failure of the strut. 3. The authors are unclear as to the point the discussers
5. Calculation of effective transverse reinforcement have raised. The cracking of concrete is related to its tensile
requirement and the minimum strut width for a diagonal strut strength rather than its compressive strength. The relationship
depends on the inclination of the diagonal strut that, in turn, between those two types of strength is approximate at best.
depends on the size of the bearing plates. Therefore, the The authors did not measure the tensile strength of the
authors may like to indicate the size of the bearing plates concrete as part of this study; thus, no analysis of the relationship
used at the loading points. between cracking load and tensile strength can be made.
4. It should be noted that the authors consider diagonal
REFERENCES tension and splitting failure of the strut to be a single failure
11. Kong, F.-K.; Robins, P. J.; and Cole, D. F., “Web Reinforcement mechanism. Based on the photos presented in the paper,
Effects on Deep Beams,” ACI JOURNAL, Proceedings V. 67, No. 12, Dec.
1970, pp. 1010-1017.
crushing of concrete is clearly visible in Specimens UL-17-0,
UL-805-0a, and UL-80.5-0b. All other specimens failed
AUTHORS’ CLOSURE through splitting of the strut.
The authors thank the discussers for their interest in the 5. As described in the original paper, the bearing plates
paper and critical evaluation of the work presented therein. were 6 x 6 in. (152 x152 mm) at the north reaction and 6 x 8 in.
The authors would like to provide answers for or comment (152 x 203 mm) at the south reaction. For specimens
on each one of the points raised by the discussers using the subjected to uniform loading, the loading plates were 3.5 in.
same reference/numbering scheme: (89 mm) in length. There was a 0.5 in. (13 mm) gap between
1. The discussers are correct. The errors they have indicated adjacent plates. For specimens subjected to a single
are typographical. concentrated load, the bearing plate at the loading point was
2. The author’s are unaware of any lack of realism in the 8 in. (203 mm) in length. For specimens subjected to a pair
analysis of the data as presented in the paper. of concentrated loads, the bearing plates were 6 in. (152 mm)
a. It should be noted that the lists of specimens based on in length. All bearing plates covered the full width of the
the actual failure loads and the normalized failure loads beam specimens.

Disc. SP-246–10/From ACI Special Publication 246, Structural Implications of Shrinkage and Creep, p. 167

Effect of Shrinkage on Short-Term Deflections of Reinforced Concrete Beams and Slabs. Paper by Peter H.
Bischoff and Ryan D. Johnson

Discussion by Gintaris Kaklauskas, Viktor Gribniak, and Darius Bacinskas

Professor, Head of Department, Vilnius Gediminas Technical University (VGTU), Vilnius, Lithuania; Researcher, VGTU; and Associate Professor, VGTU

The authors are to be congratulated for raising an important, magnitude that well exceeds the cracking strain. Due to
but unjustly neglected, issue. In general practice, effects of restraining action of reinforcement, shrinkage-induced
shrinkage and creep are taken into account in prestress loss tension stresses in concrete may significantly reduce crack
and/or long-term deformation analysis. Even at first loading, resistance and increase deformations of reinforced concrete
however, free shrinkage strain of concrete may be of such (RC) members subjected to short-term loading. Most of the

516 ACI Structural Journal/July-August 2008

 Fig. A—Shrinkage effect on deformation behavior of reinforced concrete beams.

techniques, however, do not take into account these effects in stiffening relationships from test data of flexural RC
the short-term analysis. members. This method is based on plane section hypothesis
In a series of publications, the first author and his associates and employs moment-curvature or moment-average strain
have shown influence of shrinkage effect on tension stiffening (at any layer) diagrams. Using the equilibrium equations, the
and deformations of RC members subjected to short-term stress-strain (tension stiffening) relationships are computed
axial tension. Bischoff has proposed three analytical incrementally for the extreme fiber of tensile concrete. In this
techniques based on load sharing and tension stiffening strain inverse approach, the previously computed portions of the
concepts. Using test data of shrunk RC tension members, stress-strain relationship at each load increment are used to
Fields and Bischoff4 have derived a tension-stiffening stress- compute the current increments of the stress-strain relationship.
stress relationship free of shrinkage effects. Therefore, the In present analysis, moment-curvature diagrams were
paper under discussion dealing with bending members was a obtained from the test moment-deflection relationships
logical continuation of the research. The proposed analytical shown in Fig. A(a). The conversion error due to the disregard of
technique as well as new experimental data is a significant shear effects was insignificant, as the beams were tested
contribution to the state of the art of the issue. under a four-point bending scheme. Tension stiffening
The authors have reported results on tests of four RC relationships derived by the aforementioned method are
beams with identical nominal geometrical parameters and shown in Fig. A(b) with the normalized stresses where the
cast from the same batch of concrete. One couple of the tension strength is according to Eurocode 2.13 Though a
beams was protected from shrinking and tested at 14 days, good match between the curves of the twin specimens was
whereas the second couple after 14 days of wet curing was obtained, disagreement for the shrunk and nonshrunk beams
exposed to drying condition and tested at 62 days. The test was significant. The shrunk beams have shown much less
moment-deflection diagrams are shown in Fig. A(a) along tension stiffening effect than the nonshrunk beams. The
with the analysis results performed by the ACI 318-051 and maximal stresses for the shrunk beams were well below the
Eurocode 213 techniques. The analysis was based on the tension strength. The cracking resistance was reduced by the
reported cylinder strength of concrete. As shown in Fig. A(a), shrinkage-induced tension stresses in concrete.
the code techniques accurately predicted the behavior of The aforementioned differences were due to the coupling
shrunk beams, whereas deflections of nonshrunk beams of tension stiffening with shrinkage in the shrunk beams. In
were overestimated. This suggests that the codes indirectly this study, shrinkage effect was eliminated from the
assess the effect of shrinkage. tension stiffening relationships. The analysis was based on
Applicability of code methods is limited to simple cases of the layer approach that combined the direct15 and the
loading and structural shapes. A simple approach, extensively inverse14 techniques, assuming reverse (expansion)
used in numerical modeling, is based on a smeared crack shrinkage strain. A free shrinkage strain was calculated by
model and use of a stress-strain relationship of cracked the Eurocode 213 technique assuming normal indoor
tensile concrete. Stress in the concrete is taken as the conditions.16 The computed tension stiffening relationships
combined stress due to tension stiffening and tension softening, with eliminated shrinkage effects are shown in Fig. A(c).
collectively called the tension stiffening. Most the tension The relationships obtained from the shrunk and nonshrunk
stiffening relationships were derived in a straightforward members have approached each other and most parts have
manner from tensile tests of RC members.4 Kaklauskas and practically coincided. Only small portions with maximal
Ghaboussi14 have proposed a method for deriving tension stresses (corresponding to the initiation of cracking) differed,

ACI Structural Journal/July-August 2008 517

but they practically had little effect on overall load- response prediction for the members that were allowed to
deflection behavior of the beams. It should be noted that shrink. Similar reasoning explains why the deflection
these curves were in good agreement with the tension response of the beams without any shrinkage restraint would
stiffening relationship proposed by Fields and Bischoff4 be overestimated.
(refer to Fig. A(c)). As noted, the latter was derived from Code values for the rupture modulus represent a lower
tension RC members by eliminating shrinkage effect. It bound on tensile strength, and fr can vary anywhere from
should be remembered that the shape of tension stiffening 0.62 f c′ to 1.0 f c′ MPa (7.5 f c′ to 12.0 f c′ psi).8 Hence,
relationships may depend on a number of parameters such the fact that the observed cracking moment in the preshrunk
as reinforcement bar diameter and bond characteristics, beams corresponded to the ACI computed value is fortuitous.
reinforcement ratio and bar distribution, section height, The first author of the paper has often carried out tests with
and cover. concrete having a measured rupture modulus closer to
A load-deflection diagram calculated by the layered 0.62 f c′ (7.5 f c′ ) and, in this instance, the code approach
model15 using the relationship proposed by Fields and would have underestimated member deflection of beams
Bischoff4 along with the originally reported experimental data allowed to shrink before loading because the experimental
is shown in Fig. A(d). The calculated moment-deflection (restrained) cracking moment would then be less than the
diagram well predicted deflections of the nonshrunk beams. ACI computed value.6
In deriving most tension-stiffening relationships, the Despite the apparently good fit using either the ACI or
shrinkage effect was neglected. Shrinkage may significantly Eurocode 2 approach for computing the response of the two
change the shape of the tension-stiffening relationship. beams that were allowed to shrink prior to loading (Fig. A(a)),
Therefore, future tests should either eliminate shrinkage or the discussion fails to reconcile the fact that the measured
carry out shrinkage and associated creep recordings for response of these beams actually crosses over the Icr
subsequent numerical elimination of this effect. response because of shrinkage (refer to Fig. 5 of the paper).
Both the ACI and Eurocode 2 response gradually approach
REFERENCES (but never cross) the Icr response as tension stiffening
13. Comité Européen de Normalisation, “Eurocode 2: Design of Concrete decreases under increasing load. The effect of having the
Structures—Part 1: General Rules and Rules for Buildings,” CEN, Brussels,
2001, 230 pp.
member response cross over the bare bar response is much
14. Kaklauskas, G., and Ghaboussi, J., “Stress-Strain Relations for easier to observe in axial tension members (Fig. 2 of the
Cracked Tensile Concrete from RC Beam Tests,” Journal of Structural paper) and an explanation of this phenomenon constitutes
Engineering, V. 127, No. 1, Jan. 2001, pp. 64-73. one of the main messages in the paper. While the lower
15. Kaklauskas, G., “Flexural Layered Deformational Model of Reinforced cracking moment that results from restraint to shrinkage is
Concrete Members,” Magazine of Concrete Research, V. 56, No. 10, Dec.
2004, pp. 575-584.
easily taken into account when computing deflection,6 the
16. Gribniak, V.; Kaklauskas, G.; and Bacinskas, D., “Shrinkage in corresponding shift in the Icr response is less well understood
Reinforced Concrete Structures: A Computational Aspect,” Journal of and most often ignored for this reason.
Civil Engineering and Management, V. 14, No. 1, Mar 2008, pp. 49-60. Tension stiffening represents tension carried by the
concrete between cracks and is characterized by the difference
AUTHORS’ CLOSURE between the measured member response and Icr response, as
Interest expressed in the authors’ paper is appreciated and shown in Fig. 5 for the four beams tested. Based on this
the need to consider shrinkage and creep effects is reiterated observation, the beams that are allowed to shrink will appear
in the discussion. The discussers go on to compare the to exhibit less tension stiffening unless the Icr response is
authors’ test results with design approaches used by ACI shifted over to the right (away from the member response).
318-05 and Eurocode 2, followed by a discussion of numerical In fact, tension stiffening becomes negative once the member
modeling and development of a stress-strain relationship for response crosses over the Icr response and indicates that the
cracked concrete that was originally introduced by concrete in the tensile zone is in compression, which of
Scanlon17,18 to account for tension stiffening of concrete. The course is ridiculous. Figure A(c) from the discussion
observed correlation between tension stiffening results from confirms there is little difference in tension stiffening
axial tension members and flexural members is encouraging. between beams with or without shrinkage as long as the analysis
Tension stiffening in beams is controlled by the cracking takes shrinkage into account. Hence, any discrepancies or
moment, and member stiffness is reduced when a lower conflicting results that are sometimes observed between
cracking moment is used in either the ACI 318-05 or different research programs could well arise from failure to
Eurocode 2 approaches for computing deflection. Hence, consider shrinkage effects; further work is needed in this
accurate prediction of deflection is dependent on having the area as pointed out in the discussion.
correct cracking moment.6 The cracking moment depends on
the rupture modulus fr of the concrete, and the authors’ REFERENCES
control tests gave a measured value of fr equal to 5.2 MPa 6. Scanlon, A., and Bischoff, P., “Shrinkage Restraint and Loading
(750 psi) (corresponding to 0.78 f c′ MPa [113.1 f c′ psi) History Effects on Deflection of Flexural Members,” ACI Structural
that was 25% greater than the ACI computed value of Journal, V. 105, No. 4, July-Aug. 2008. (Publication information updated
0.62 45 = 4.2 MPa (7.5 6500 = 600 psi). Development of since publication of SP-246 and is presented herein for reference).
tensile stresses from restraint to shrinkage probably reduced 17. Scanlon, A., “Time-Dependent Deflections of Reinforced Concrete
Slabs,” PhD thesis, University of Alberta, Edmonton, AB, Canada, 1971.
the apparent cracking stress to a value close to the ACI 18. Scanlon, A., and Murray, D. W., “Time-Dependent Reinforced Concrete
computed value, which is the reason why the discussers Slab Deflections,” Journal of the Structural Division, ASCE, V. 100,
found that the code techniques gave a fairly reasonable No. ST9, Sept. 1974, pp. 1911-1924.

518 ACI Structural Journal/July-August 2008