ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 102-S59

Axial Capacity Model for Shear-Damaged Columns

by Kenneth J. Elwood and Jack P. Moehle

This paper introduces a model to estimate the axial capacity of a knowledge, it is frequently not economically feasible to
column that has previously experienced shear failure. The model protect all columns in an existing reinforced concrete
is applicable to existing reinforced concrete building columns building from shear failure during strong ground motion.
vulnerable to shear failure during earthquakes. The axial load on Given the lack of understanding of how the axial loads will
a shear-damaged column is assumed to be supported by a combi-
nation of compression of the longitudinal reinforcement and force be supported after shear failure, engineers have resorted to
transfer through shear-friction on an idealized shear-failure plane. installing secondary gravity load support systems to ensure
The effective coefficient of friction from the classical shear-friction that shear failure of individual columns does not lead to
equation is related to the drift ratio at axial failure using the collapse of the building.8 Hence, a better understanding of
results from 12 full-scale pseudostatic column tests. The model, column axial load capacity after shear failure may lead to a
which represents the general observation from experimental tests reduction in seismic retrofit costs.
that the drift ratio at axial failure of a shear-damaged column is
inversely proportional to the magnitude of the axial load and
directly proportional to the amount of transverse reinforcement, EXPERIMENTAL EVIDENCE
provides relations among axial load, transverse reinforcement, and To address the need for data after the loss of lateral load
the interstory drift at axial load collapse. capacity for columns typical of those constructed prior to the
mid 1970s in the western United States, pseudostatic planar
Keywords: axial load; column; seismic; shear. tests on 12 full-scale shear-critical reinforced concrete
columns were conducted by Lynn9 and Sezen10 up to the
INTRODUCTION point of axial failure. Figure 1 illustrates a typical test
Most tests of reinforced concrete columns under seismic column configuration. Specific column characteristics,
load conditions have been terminated shortly after loss of material properties, and measured responses are summarized
lateral load capacity. The resulting data are useful for in Table 1. With the exception of two specimens, the loading
columns considered as part of the lateral-force-resisting routine subjected each column to nominally constant axial
system. Considering traditional notions of safety (that is, once compression and maintained nominally zero rotation
shear failure begins, axial load collapse cannot be far behind), between column ends while the column was subjected to a
the data also probably define a practical upper-bound series of lateral displacements at increasing amplitude, with
displacement capacity even for columns not considered part three cycles at each amplitude. The two exceptions were
of the lateral-force-resisting system in new building designs.
For existing buildings, whether being evaluated for seismic
resistance or for seismic retrofit, a less conservative
approach may be required by economic and functionality
considerations. If a column can reliably carry gravity load
after its lateral strength degradation begins, it may be
possible to achieve considerable savings by considering the
column as a secondary component. The model developed in
this paper provides practicing engineers with a means of
estimating the drift ratio at which a shear-damaged column
can be expected to lose the ability to support axial loads.
Several pseudostatic tests have been performed in Japan to
investigate the axial capacity of shear-damaged columns.1-6
Several important observations should be noted from these
tests, namely, sliding along the diagonal shear cracks was
often observed prior to axial failure, axial failure occurred
when the shear capacity was reduced to approximately zero,
and the drift at axial failure decreased with increasing axial
stress. Kato and Ohnishi2 calibrated a plastic drift capacity
model, based on the results from 32 column specimens, to Fig. 1—Typical column test specimen.9,10 All dimensions
estimate the drift at axial failure; however, the model does in mm.
not compare well with the measured drifts from the column
database used in this study.7
ACI Structural Journal, V. 102, No. 4, July-August 2005.
MS No. 04-086 received March 11, 2004, and reviewed under Institute publication
RESEARCH SIGNIFICANCE policies. Copyright © 2005, American Concrete Institute. All rights reserved, including the
Engineers involved in the seismic retrofit of buildings in making of copies unless permission is obtained from the copyright proprietors. Pertinent
discussion including author’s closure, if any, will be published in the May-June 2006
California have found that, given the current state of ACI Structural Journal if the discussion is received by January 1, 2006.

578 ACI Structural Journal/July-August 2005


ACI member Kenneth J. Elwood is an assistant professor at the University of British
Columbia, Vancouver, British Columbia, Canada. He received his PhD from the
University of California, Berkeley, Berkeley, Calif. He is a member of ACI Committees
369, Seismic Repair and Rehabilitation, and 374, Performance-Based Seismic Design
of Concrete Buildings, and Joint ACI-ASCE Committee 441, Reinforced Concrete
Columns. His research interests include the behavior and performance-based design
of reinforced concrete structures under seismic loading.
ACI member Jack P. Moehle is Professor of Structural Engineering and Director of
the Pacific Earthquake Engineering Research Center at the University of California,
Berkeley. He is member of ACI Committees 318, Structural Concrete Building Code,
and 369, Seismic Repair and Rehabilitation, and Joint ACI-ASCE Committee 352,
Joints and Connections in Monolithic Concrete Structures. His research interests
include earthquake resistance of reinforced concrete buildings and bridges.
Column 2CVD12, which had variable axial load ranging from
250 kN tension to 2670 kN compression (with an axial load
of 1473 kN just prior to axial failure), and Column 2CLD12M,
which, after cycles below the yield displacement, was
subjected to monotonic lateral loading until axial failure.
Because the test setup did not allow for any redistribution of
the applied axial load, once axial failure was initiated, the
tests were terminated.
Figure 2 plots drift ratios corresponding to significant
events for the 12 columns reported by Lynn9 and Sezen.10
For columns having lower axial loads, the tendency is for
axial load failure to occur at relatively large drifts, regardless
of whether shear failure had just occurred or whether shear
failure had occurred at much smaller drift ratios. For
columns with larger axial loads, axial load failure tended to
occur at smaller drift ratios, and might occur almost
immediately after the loss of lateral load capacity. Note also
that the drift ratios at axial load failure tend to be lower for
columns with larger spacing of the transverse reinforcement
(dashed lines). The next section presents a shear-friction
model that can be used to represent the general observations
from Fig. 2, namely that the drift at axial load failure is
inversely related to the magnitude of axial load and directly
related to the amount of the transverse reinforcement.

SHEAR-FRICTION MODEL
Equilibrium equations
The column shown in Fig. 3 was damaged during the 1999
Kocaeli Earthquake. Any axial load supported by the
damaged column must be transferred across the obvious

Fig. 2—Column drift ratios as function of axial load for

columns tested by Lynn9 and Sezen.10 (Dashed and solid
lines indicate columns with transverse reinforcement
spacing of 457 and 305 mm, respectively.) Fig. 3—Damaged column from 1999 Kocaeli Earthquake.

Table 1—Database of columns with axial failure data

Specimen b, mm d, mm dc , mm db , mm nbars ρlong Ast, mm2 s, mm f c′ , mm fyl , MPa fyt , MPa P/Ag f c′ ∆y /L, % ∆s/L, % ∆a/L, %
9
Lynn
3CLH18 457 393 371 32.3 8 0.03 142 457 26 331 400 0.09 0.65 1.03 2.07
2CLH18 457 397 371 25.4 8 0.02 142 457 33 331 400 0.07 0.51 2.59 3.10
3SLH18 457 393 371 32.3 8 0.03 142 457 26 331 400 0.09 0.53 1.03 3.10
2SLH18 457 397 371 25.4 8 0.02 142 457 33 331 400 0.07 0.44 2.07 3.62
2CMH18 457 397 371 25.4 8 0.02 142 457 26 331 400 0.28 0.56 1.03 1.03
3CMH18 457 393 371 32.3 8 0.03 142 457 28 331 400 0.26 0.77 1.03 2.07
3CMD12 457 393 371 32.3 8 0.03 245 305 28 331 400 0.26 0.66 1.55 2.07
3SMD12 457 393 371 32.3 8 0.03 245 305 26 331 400 0.28 0.77 1.55 2.07
Sezen10
2CLD12 457 392 365 28.7 8 0.025 245 305 21 441 469 0.15 0.89 2.59 5.00
2CHD12 457 392 365 28.7 8 0.025 245 305 21 441 469 0.61 0.68 0.86 1.90
2CVD12 457 392 365 28.7 8 0.025 245 305 21 441 469 0.34 0.71 1.90 2.93
2CLD12M 457 392 365 28.7 8 0.025 245 305 22 441 469 0.15 0.91 2.84 5.09
Note: b = square column width; d = depth to centerline of tension reinforcement; dc = depth of column core (centerline to centerline of transverse reinforcement); db = diameter of
longitudinal reinforcement; nbars = number of longitudinal bars; ρlong = longitudinal reinforcement ratio; Ast = area of tie steel; s = tie spacing; f c′ = concrete cylinder strength; fyl =
longitudinal steel yield strength; fyt = transverse steel yield strength; P/Ag f′c = axial load ratio; ∆y/L = drift ratio at yielding of longitudinal reinforcement (based on strain gauge
data); ∆s/L = drift ratio at shear failure (at 20% loss in peak shear); ∆a/L = drift ratio at axial failure; L = clear height of column (L = 2946 mm for all columns in table).

ACI Structural Journal/July-August 2005 579

shear failure plane. Such transfer of load can be modeled by
a mechanism known as shear friction. Shear-friction models
evaluate the shear stress that can be transferred across a
crack as a function of the normal stress on the crack surface.
The normal stress results from the elongation of reinforcement
crossing the crack and/or applied forces normal to the crack
surface. For the column shown in Fig. 3, the transverse
reinforcement crossing the shear failure plane and the axial
load carried by the column combine to provide a normal
force and, hence, a shear transfer across the shear failure plane.
Figure 4 shows the free body diagram for the upper portion
of the column from Fig. 3. The inclined free surface at the
bottom of the free-body diagram is assumed to follow a
critical inclined crack associated with shear damage. In this
presentation, the critical crack is one that, according to the
idealized model, results in axial load failure as the shear-
friction demand exceeds the shear-friction resistance along
the crack. Dowel forces from the transverse reinforcement
crossing the inclined crack are not shown; instead, the dowel
forces are assumed to be included implicitly in the shear-
friction force Vsf along the inclined plane. Equilibrium of the
forces shown in the free body diagram results in the
following equations
Ast fyt d c
ΣF x → N sin θ + V = Vsf cos θ + -----------------
- tan θ + n bars V d (1)
s Eq. (2) and (3) requires models for the critical crack angle θ,
ΣF y → P = N cos θ + Vsf sin θ + n bars Ps (2)
where nbars is the number of longitudinal bars crossing the
shear-failure plane; dc is the depth of the column core from
centerline to centerline of the ties; s is the spacing between
the transverse reinforcement; Ast and fyt are the area and
yield strength of the transverse reinforcement; and the forces
P, V, N, Vsf, Ps, and Vd are shown in Fig. 4. The number of
hoops crossing the inclined crack is (dc/s)tanθ.
The shear resistance due to dowel action of the longitudinal
bars, Vd, is dependent on the spacing of the transverse rein-
forcement. As shown in Fig. 5, the upper concrete block will
bear against the longitudinal bar on one side of the crack and
the transverse steel will restrain the bar on the other side. As
Fig. 4—Free body diagram of column after shear failure.
580
the distance between these forces increases, the effectiveness
of the dowel action will diminish. Due to the large spacing of
transverse reinforcement in many shear-critical columns of
interest in this study, the distance between the forces will
most likely be too large to develop any significant dowel
action. Note that the dowel action may be more effective for
longitudinal reinforcement along the side face of the column
(that is, parallel to the direction of applied shear) since these
bars will be restrained by concrete above and below the
failure plane. Any limited resistance to sliding from the
dowel action, however, can be considered as incorporated in
the shear due to shear-friction Vsf acting on the shear failure
plane. Hence, the forces due to dowel action will be ignored
in the derivation of the axial capacity model. Furthermore,
based on experimental evidence suggesting that the shear
capacity drops to zero when axial load failure occurs,1-6,9,10
when considering the stage of axial load failure the external
shear force V can be set equal to zero.
In light of the previous discussion, Eq. (1) can be rewritten
as follows
Ast fyt d c
N sin θ = V sf cos θ + -----------------
- tan θ (3)
s
Further development of an axial capacity model using
the axial capacity of the longitudinal reinforcement Ps, and
the relationship between N and Vsf. Each of these models
shall be discussed in turn in the following sections.
Critical crack angle
Few reliable models exist for estimating the inclination θ of
the shear failure plane. A basic principles approach is to define
θ as the angle of the nominal principal tension stress at the
instant when it reaches the tensile capacity of concrete under
combined shear and axial load, using a Mohr’s circle represen-
tation of the state of stress. This approach, however, invariably
results in an angle steeper than those observed in tests.
A model proposed by Kim and Mander11 estimates the
crack angle based on minimizing the external work due to a
unit shear force. For the columns tested by Lynn9 and
Sezen,10 the critical crack angle estimated by the model
ranges from 65 to 71 degrees, with an average of 68 degrees.
Figure 6 plots the observed average angle of critical shear
cracks from the tests. (The angles were subjectively estimated
from photographs.) The angle could be approximated as
65 degrees relative to horizontal (the dashed line in the
Fig. 5—Dowel action in longitudinal bars.
ACI Structural Journal/July-August 2005
figure), or could have the linear variation suggested by the the bar ends, is greater than Asl fyl, suggesting that the lightly
solid line in the figure, that is loaded columns in the test series will not experience buckling
of the longitudinal reinforcement upon axial failure. Based
θ = 55 + 35P/Po (4) on these observations, the longitudinal reinforcement axial
capacity was evaluated as follows:
• For columns where Ps < Asl fyl , Ps is based on the plas-
Po is the axial capacity of the undamaged column given by
tic axial-load strength of the longitudinal reinforcement
Po = 0.85 fc′ (Ag – Asl) + fyl Asl, where fc′ is the concrete
in the deformed configuration; and
compressive strength; Ag is the gross concrete area; Asl is the
area of longitudinal steel; and fyl is the yield strength of the • For columns where Ps ≥ Asl fyl , Ps is based on the plas-
longitudinal reinforcement. (The outlying datum in Fig. 6 at tic strength in the deformed configuration, but limited
P/Po ≈ 0.21 was for Column 3CMH18. That column had a by the plastic buckling capacity.
critical crack that was somewhat less steep over most of its The following paragraphs discuss each of these cases in turn.
length, with a vertical segment near column mid-depth, The plastic strength of the longitudinal reinforcement in
resulting in the relatively large reported critical crack angle.) the deformed configuration is illustrated in Fig. 8. Assuming
Considering the difficulties of accurately determining the there is no dowel force, the plastic moment capacity of the
critical crack angle given the state of many of the columns at reinforcing bar and the axial capacity are related by
the end of the tests, and the lack of improvement observed in
the prediction of the drift at axial failure when Eq. (4) is used Ps ∆ = 2Mp (5)
in place of a constant crack angle of 65 degrees, only the
constant crack angle model will be used in the development Using the decomposition of the stresses in the fully plastic
of the axial failure model presented herein. section of a reinforcing bar shown in Fig. 8, the plastic
All of the columns tested by Lynn9 and Sezen10 had a moment can be determined as follows
height-to-width ratios greater than 6.0. For columns with low
height-to-width ratio, it is expected that the maximum crack M p = 2A tens fyl z tens (6)
angle will be limited by the aspect ratio of the column (that
is, θmax = tan–1(height/width)). This may be considerably less
than 65 degrees. where Atens is the area of the reinforcing bar in tension, and
ztens is the distance from the centroid of Atens to the centroid
Longitudinal reinforcement axial capacity
Based on observations of the final state the column longi-
tudinal reinforcement from the static tests by Lynn9 and
Sezen,10 it is assumed that the longitudinal reinforcement
will support a portion of the axial load nbars Ps up to a
maximum load defined by either the buckling or the plastic
capacity of the reinforcing bars. Columns with an axial load
greater than the pure axial plastic capacity of the longitudinal
reinforcement (Asl fyl) experienced a deformed shape of the
longitudinal reinforcement after axial failure indicative of a
buckling failure (for example, refer to Fig. 7(a)). In contrast,
most of the columns with an axial load less than Asl fyl expe-
rienced a deformed shape of the longitudinal reinforcement
after axial failure that did not suggest a buckling failure of
the longitudinal reinforcement (for example, refer to Fig. 7(b)).
Note that the elastic buckling capacity of the longitudinal
reinforcement, calculated using a buckling length equal to
the spacing of the ties and assuming full rotational fixity at
Fig. 7—Deformed shape for longitudinal bars at loss of
axial load capacity for column with: (a) high axial load;
and (b) low axial load.10

Fig. 6—Relation between observed angles of critical cracks Fig. 8—Plastic strength of longitudinal reinforcement in
and axial load. deformed configuration.

ACI Structural Journal/July-August 2005 581

of the bar section. Given an axial load in the reinforcing bar,
Atens can be determined as follows
P plastic capacity of the longitudinal reinforcement without
A tens = ---  A bar – -----s
1
(7)
2  fyl 
where Abar is the cross-sectional area of one longitudinal
reinforcing bar. Equations (5) through (7) can be used to
determine a theoretical relation between the axial load Ps and
the lateral displacement at which the plastic capacity of a
reinforcing bar is fully developed. The results for the three
bars used as longitudinal reinforcement in the tests by Lynn9
and Sezen10 are shown in Fig. 9.
The curves shown in Fig. 9 must be determined by iteration
due to the nonlinear moment-axial load interaction diagram
that results from solving Eq. (6) and (7). If the linear,
conservative, approximation to the interaction diagram shown
in Fig. 10 is used with Eq. (5), the axial plastic capacity of
the longitudinal reinforcement can be related directly to the
story drift, without iteration, as follows
Ps db ⁄ L
--------------- = ----------------------
- (8)
A bar fyl 3 ∆ db
--- π --- + -----
4 L L
Fig. 9—Longitudinal reinforcing bar axial plastic capacity
using Eq. (5) through (7).
Fig. 10—Axial load-moment interaction diagram for
reinforcing bar (L/db = 100).
582
where L is the clear height of the column, and db is the diameter
of the longitudinal reinforcing bars. Equation (8), shown in
Fig. 11, provides a conservative approximation of the axial
iteration and will be used in the further development of the
axial capacity model.
For heavily loaded columns (P ≥ Asl fyl), the axial capacity
of the longitudinal reinforcement given by Eq. (8) will be
limited by the plastic buckling capacity. Evaluation of the
plastic buckling capacity requires estimation of the tangent
modulus of the reinforcement and the effective buckling
length. Based on tensile coupon tests of typical reinforcing
bars, the tangent modulus is estimated as 0.07Es, where Es is
the elastic modulus of the reinforcement. (Alternatively, an
equivalent modulus model12 could be used to estimate the
material stiffness; however, such a model predicts that the
plastic buckling load does not control the axial capacity of
the longitudinal reinforcement for any of the columns in the
database. Considering the observed buckled deformed shape
of longitudinal reinforcement for two of the tests [for example,
refer to Fig. 7(a)], the equivalent modulus model was not
adopted for this study.) Based on the observed deformed
shape of the longitudinal reinforcement shown in Fig. 7(a),
the effective buckling length should be 1.0s > Leff > 0.5s (that
is, shorter than a pinned-pinned condition at the ties and
longer than a fixed-fixed condition at the ties). An effective
buckling length of 0.8s is selected for this investigation.
Based on these assumptions, the axial capacity of the
longitudinal reinforcement is given by
Ps db ⁄ L (9)
--------------- = ----------------------
- if P < A sl fyl
A bar fyl 3 ∆ db
--- π --- + -----
4 L L
2
db ⁄ L π E s I bar n bars
- < 0.1 -------------------
= ---------------------- - ------------ if P ≥ A sl fyl
3--- π ∆
--- + -----
d b s
2 A sl fyl
4 L L
Table 2 gives the axial capacity of the longitudinal
reinforcement at axial failure of the column, estimated based
on Eq. (9), for the columns tested by Lynn9 and Sezen.10 The
values in Table 2 were calculated based on the measured
Fig. 11—Longitudinal reinforcing bar axial plastic capacity.
Comparison of iterative method and approximation of
Eq. (8) (L /db = 100).
ACI Structural Journal/July-August 2005
column drift ratio at axial failure. Note that only 2CMH18 is Vsf = Nµ (10)
controlled by the plastic buckling load. For most of the
columns, the above formulation results in approximately where N is the compression force acting normal to the crack,
25% of the axial load being carried by the longitudinal as shown in Fig. 4. Because the shear transfer mechanism
reinforcement at the point of axial failure. For two of the includes aggregate interlock and dowel action in addition to
columns with a low axial load (P = 0.09Ag f c′ ) and a relatively pure friction, values for µ must be higher than that for pure
high longitudinal reinforcement ratio of 3% (3CLH18 and friction across a concrete interface in order to match Eq. (10)
3SLH18), Eq. (9) estimates that the longitudinal bars are with test data.
supporting over 50% of the axial load (that is, 76 and 54%,
respectively). Due to the relatively low axial stiffness of the Substitution of Eq. (10) into Eq. (2) and (3), and eliminating
longitudinal bars, this seems unreasonably high and suggests the case where µ = tanθ, gives the following expression for
that a limit on the fraction of axial load supported by the the axial capacity of the column illustrated in Fig. 4
longitudinal bars may be appropriate.
As shown in Fig. 4, the axial load supported by shear- Ast fyt d c 1 + µ tan θ
- tan θ ------------------------ + n bars P s
P = ----------------- (11)
friction and the axial load supported by the longitudinal s tan θ – µ
reinforcement act in parallel. By using the ultimate axial
capacity of the longitudinal reinforcing bars in the equilibrium
The first term in Eq. (11) is the axial load carried through
equation (Eq. (1)) it is assumed that the ultimate shear-
shear friction, whereas second term is the axial load carried
friction capacity and the ultimate capacity of the longitudinal
by the longitudinal reinforcement (given by Eq. (9)). Note
reinforcement are reached at the same time. The ultimate
that values of µ greater than tanθ will result in a meaningless
shear friction capacity may be exceeded prior to full devel-
negative shear-friction capacity. For µ equal to zero, the
opment of the longitudinal reinforcement axial capacity,
shear friction term in Eq. (11) reduces to the same form as
thereby transferring axial load to the longitudinal reinforcement
the 45-degree truss model.
as sliding occurs on the shear failure plane. This transfer of
load to the longitudinal reinforcement may exceed the ultimate Recall that for the total capacity model, the shear friction
axial capacity of the reinforcement and subsequently lead to and longitudinal reinforcement terms are summed, as shown
axial failure of the column. In this case, axial failure should in Eq. (11), while for the maximum capacity model, only the
be defined by exceeding the shear friction capacity, and the maximum of the two terms is considered. Equation (11)
load carried by the longitudinal reinforcement should be can be rearranged to give the following expression for the
limited to some fraction of the total axial load. A limit of effective coefficient of friction for the total capacity model
Psnbars/P < 50% was selected because this improved the
correlation of the model with test data. Considering this limit, Ast fyt d c
the axial load supported by the longitudinal reinforcement P – n bars P s – ----------------- -
s
for Specimens 3CLH18 and 3SLH18 is reduced to 0.12Asl fyl. µ t = -------------------------------------------------------------------- (12)
( P – n bars P s ) Ast fyt d c
(These reduced values are shown in parentheses in Table 2.) - + ------------------ tan θ
-------------------------------
tan θ s
Maximum and total capacity models
Considering the expected transfer of axial load from the where the subscript t refers to the total capacity model. Using
shear failure plane to the longitudinal reinforcement after the a constant crack angle of 65 degrees and the longitudinal rein-
shear-friction capacity is exceeded, it may be appropriate to forcement axial capacity given in Table 2 (but limited to less
consider the axial load support from the longitudinal rein- than 50% of the axial load on the column, as discussed previ-
forcement independently of that for shear friction. In such a ously), the effective coefficient of friction for each of the test
model, the load carried by the longitudinal bars is removed columns can be calculated using Eq. (12). Figure 12(a) plots
from the equilibrium equation (Eq. (2)) and the capacity the calculated values for each column as a function of the
curves for the longitudinal reinforcement (Fig. 11) are lateral drift ratio at which the column could no longer
superimposed on the capacity curves from the shear-friction
model (to be developed in the next section). The axial Table 2—Calculated longitudinal reinforcement
capacity of the column is taken as the maximum of the axial capacity for columns in database
capacity from the longitudinal reinforcement and the
Specimen Ps/(Abar fyl) Psnbars/P
capacity from the shear-friction model. This model, referred
to as the maximum capacity model, will be developed further 3CLH18 0.18 (0.12) 0.76 (0.50)
in the following sections. The model based on summing the 2CLH18 0.11 0.29
ultimate capacity from the longitudinal reinforcement and 3SLH18 0.13 (0.12) 0.54 (0.50)
the ultimate capacity from shear-friction, in accordance with 2SLH18 0.09 0.25
the equilibrium equations, will be referred to as the total 2CMH18 0.11* 0.10
capacity model.
3CMH18 0.18 0.25
3CMD12 0.18 0.25
Application of classical shear-friction model
3SMD12 0.18 0.25
The literature documents several shear-friction models
2CLD12 0.08 0.27
that relate Vsf and N.13-15 The classical shear-friction model,
included in ACI 318 since 1977, idealizes the crack across 2CHD12 0.18 0.16
which shear must be transferred as a flat surface with an 2CVD12 0.12 0.19
effective coefficient of friction µ. The shear capacity is 2CLD12M 0.08 0.26
defined as *
Controlled by plastic buckling capacity.

ACI Structural Journal/July-August 2005 583

sustain the applied axial load. The data apparently follow
a trend that can be approximated by
100 ∆
µ t = tan θ – ---------  --- ≥0 (13)
3  L axial
In selecting Eq. (13), the effective coefficient of friction
was set equal to tan(65 degrees) at zero drift to ensure the
shear-friction capacity remained positive for all valid drifts.
A plot similar to that shown in Fig. 12(a) can be developed
for the maximum capacity model by omitting the nbarsPs
term from Eq. (12) and recalculating the effective coefficient
of friction (refer to Fig. 12(b)). The data appear to have less
scatter when the capacity of the longitudinal reinforcement
is omitted. Based on Fig. 12(b)), the drift ratio at axial
failure appears to follow a straight-line trend that can be
approximated by
100 ∆
µ m = tan θ – ---------  --- ≥0 (14)
4  L axial
The data of Fig. 12 suggest that the effective shear-friction
coefficient is a function of the drift angle at axial failure.
This relation is plausible considering that increased defor-
mation (and increased sliding along the critical shear plane)
degrades the roughness of the shear plane and reduces the
effective friction. It is worth recalling that the increased
deformation capacities are associated with reduced axial loads
and increased amount of transverse reinforcement (Fig. 2).
Fig. 12—Relation between effective coefficient of friction
and drift ratio at axial failure for: (a) total capacity model;
and (b) maximum capacity model.
584
It is expected that the shear-friction coefficient will also be
inversely proportional to other parameters related to the
amount of sliding along the critical shear plane. Among
others, such parameters may include the displacement
ductility, the number of cycles past the yield displacement,
and a drift ratio based on the height of the damaged region of
the column. The interstory drift ratio (IDR) (based on the
clear height of the column) was selected for this investigation
to be consistent with research by other investigators into the
use of the maximum IDR as an appropriate engineering demand
parameter in a performance-based design methodology.16
Axial failure may be more closely related to a drift ratio
based on the height of the damaged region, or the concentrated
drift ratio (CDR), as defined in Fig. 13 (dc is the depth of the
column core from centerline to centerline of the ties). After
shear failure, most of the column deformations are concentrated
in the shear-damaged region of height h. As suggested by
Fig. 13, two columns of length LA and LB that experience the
same displacement will have different IDRs, but may have
the same CDR. For columns shorter than dctan65 degrees,
the height of the damaged region will be constrained by the
height of the column and the IDR will be equal to the CDR.
Because all of the columns tested by Lynn9 and Sezen10
have the same height-to-width ratio and a critical crack angle
of approximately 65 degrees is assumed for all specimens, the
CDRs for this database will be approximately equal to the
IDRs times a constant factor. For a more extensive data-
base, the CDR should be used to distinguish between
columns such as those illustrated in Fig. 13.
Drift capacities
The preceding sections presented the expressions that can
be used to establish relationships for the drift ratio at axial
failure in terms of the axial load, the transverse reinforcement,
and the longitudinal reinforcement.
For the total capacity model, Eq. (9), (11) (with θ = 65
degrees), and (13) are be combined to give the drift capacity
curves shown in Fig. 14(a). For high axial loads, the buckling
capacity of the longitudinal reinforcement is assumed to
govern according to Eq. (9), resulting in the sudden reduction
in drift capacity seen in Fig. 14(a). For low axial loads, the
drift capacity curves approach horizontal, suggesting that
there exists a lower bound axial load below which axial
failure is not expected to occur. Based on this model, the
lower bound axial load capacity is the sum of the axial load
supported by the 45-degree truss model and the longitudinal
bar capacity at large drifts. One must bear in mind, however,
Fig. 13—Concentrated versus interstory drift ratio.
ACI Structural Journal/July-August 2005
that this model is the result of simple theoretical concepts ation between the two models at low and high drifts is due to
applied to a small data set and hence may not capture all the changes in the longitudinal reinforcement capacity,
possible failure modes which may lead to a loss of axial load which influences only the total capacity model.
capacity for columns subjected to axial loads below the To convey a sense of the accuracy implicit in the relations
lower bound suggested in Fig. 14(a). of Fig. 14, those relations were used to estimate the drift
The maximum capacity model, shown in Fig. 14(b), takes capacity of the columns tested by Lynn9 and Sezen.10 The
the maximum axial load from either the longitudinal bar results are plotted in Fig. 16(a) for the total capacity model,
capacity (refer to Fig. 11) or the shear-friction capacity (that is, and Fig. 16(b) for the maximum capacity model. The mean
the first term of Eq. (11), with θ = 65 degrees and µ given by ratios of the measured to calculated drift at axial load failure
Eq. (14)). Note that longitudinal bar buckling does not based on the total and maximum capacity models are 1.02 and
influence this model because the buckling capacity of the 0.97, respectively; the coefficients of variation are 0.22 and
longitudinal reinforcement will always be less than the 0.26, respectively.
shear-friction capacity at low drifts. For the parameters
shown, the longitudinal bar capacity only governs for large
column drifts and low amounts of transverse reinforcement.
Given that the longitudinal bar capacity has such little effect
on the maximum capacity model, and that no data exist
beyond a drift ratio of 0.06 to support the claim that the
longitudinal bars will govern the capacity, the additional
complexity of including the longitudinal bar capacity may
not be warranted.
All of the plotted relations in Fig. 14 suggest the intuitive
result that drift capacity increases with increasing transverse
reinforcement and decreasing axial load. This is consistent
with the experimental observations discussed previously.
Figure 15 compares the drift capacity curves based on the
total and maximum capacity models. The very close agreement
between the two models is a result of selecting the relations
Fig. 15—Comparison of total and maximum capacity models.
between the effective coefficient of friction and the drift
ratio at axial failure based on the same data (Fig. 12). The vari-

Fig. 14—Drift capacity curves based on: (a) total capacity Fig. 16—Comparison of measured-to-calculated drift ratios
model; and (b) maximum capacity model. (Buckling capacity for tests by Lynn9 and Sezen10 based on: (a) total capacity
of longitudinal reinforcement assumed to be 0.1Asl fyl for model; and (b) maximum capacity model. (Dashed lined are
relations shown in (a)). plus or minus one standard deviation from mean.)

ACI Structural Journal/July-August 2005 585


Fig. 17—Axial capacity model (Eq. (15)) plotted with test
data from Lynn9 and Sezen.10
Fig. 18—Shear failure modes with bearing support for
axial loads.
Given the close agreement between the models (Fig. 15),
the lack of influence from the longitudinal reinforcement on
the maximum capacity model (Fig. 14(b)), and the reasonable
accuracy of the two models (Fig. 16), it is recommended that
the maximum capacity model based on the shear-friction
capacity alone should be used to assess the drift ratio at
which axial failure is expected to occur. Based on the test
data, the accuracy of such a model is equivalent to that
shown in Fig. 16(b) for the maximum capacity model since
the longitudinal reinforcement capacity did not control at the
drift ratios recorded in the tests. Such a model only requires
information on the transverse reinforcement and the axial
load, and can be expressed as follows
∆---
4 1 + ( tan θ )
= --------- ---------------------------------------------------------
 L axial
tan θ + P  -----------------------------
100
 Ast fyt d c tan θ
where θ was assumed to be 65 degrees during the derivation of
the model. The axial capacity model can be plotted as a single
curve and compared with the test data, as shown in Fig. 17.
Note that the uppermost data points (2CMH18 and
3CMH18) in Fig. 17 differ only by the amount of longitudinal
reinforcement. Based on the longitudinal reinforcement axial
capacity model, the column with the lower drift ratio at axial
failure and lower longitudinal reinforcement ratio (2CMH18)
is expected to experience buckling of the longitudinal
reinforcement. Therefore, the difference in measured drifts
at axial failure for the two specimens may be explained by
the reduction in drift capacity for columns susceptible to
586
longitudinal bar buckling according to the total capacity
model (refer to Fig. 14).
SHORTCOMINGS OF PROPOSED MODEL
The shear-friction model described above significantly
simplifies a very complex problem, hence, several deficiencies
in the model can be expected. Some of the deficiencies
include: the reliance on full anchorage of the transverse
reinforcement, not accounting for direct bearing of concrete
components, the dependence on a distinct shear failure plane,
and the limited data set upon which the model is based.
The shear-friction model assumes that the full yield
capacity of the transverse reinforcement can be achieved and
maintained after shear failure of the column. This assumption
is only valid if the transverse reinforcement has sufficient
anchorage. Because 90-degree hooks are common for the
ties of older reinforced concrete columns, such anchorage
cannot always be relied on. It is recommended that future
modifications to the model include a coefficient that reduces
the contribution of the transverse reinforcement. Such a
reduction factor has been proposed for the calculation of the
shear capacity of older reinforced concrete columns.17
Some shear failure modes, illustrated in Fig. 18, result in
axial support provided by the bearing of concrete against
concrete across a shear failure plane. This mechanism of
axial load support is not considered in the shear-friction model.
Further study is required to develop methods by which the
formation of such failure modes can be reliably predicted.
The shear-friction model assumes that the shear failure
plane is continuous and distinct. However, the complex
behavior of a column during shear failure can result in a
disjointed failure plane where the principle sliding
surface is intercepted by multiple cracks at various
angles. Due to damage to the column core, the failure
surface may, in fact, consist of several blocks of concrete
bearing against one another.
It must be recognized that the axial failure model derived
above is based on data from only 12 columns. All of the
columns were constructed of normal strength concrete, had
the same height-to-width ratio, and were designed to yield
the longitudinal reinforcement prior to shear failure. Only
limited variation in the spacing and type of transverse rein-
forcement was possible. The axial failure model presented
herein may not be appropriate for columns for whom the test
specimens are not representative.
Furthermore, all of the columns in the database were tested
2
(15) under unidirectional lateral load parallel to the one face of
the column. With the exception of two tests, the loading
s
routine was standardized, with each column subjected to
nominally constant axial compression and a series of lateral
displacements at increasing amplitude (three cycles at each
amplitude). During earthquake excitation, columns can
experience bidirectional loading and a wide variety of
loading histories, which may consist of a single, large pulse
or many smaller cycles prior to shear and axial load failure.
It has been demonstrated that an increase in the number of
cycles past the yield displacement can result in a decrease in
the drift capacity at shear failure.18 While it is anticipated
that an increasing number of cycles has a similar impact on
the drift capacity at axial failure, not enough test data are
available to support or refute this hypothesis. Further
testing of reinforced concrete columns to the stage of axial
failure is needed to supplement the current database.
ACI Structural Journal/July-August 2005
 CONCLUSIONS
Based on shear-friction concepts and the results from
12 columns tested to axial failure, a model (Eq. (15) and
Fig. 17) has been developed to estimate the drift at axial
failure for a shear-damaged column. The mean ratio of the
measured to calculated drift at axial load failure based on
Eq. (15) is 0.97; the coefficient of variation is 0.26. The
relatively large scatter may be a product of inherent
randomness associated with the complicated failure mech-
anism. The results from the model, however, are consistent
with the general observation from experimental tests that the
drift at axial failure is directly proportional to the amount of
transverse reinforcement and inversely proportional to the
magnitude of the axial load. The model suggests that for low
axial loads, or high amounts of transverse reinforcement, a
shear-damaged column can maintain its axial loads well
beyond a drift of 2%.
The proposed model provides a good estimate of the mean
drift at axial failure for the 12 columns included in the data-
base. As such, approximately half of the columns from the
database used to develop the model experienced axial failure
at drifts less than those predicted by the model. Furthermore,
actual columns have configurations and loadings that
differ from those used in the tests reported herein, so that
some additional scatter in results beyond that demonstrated
herein seems likely. A model used for design or assessment
might be adjusted accordingly to provide additional
conservatism as may be appropriate. Future testing of
reinforced concrete columns to the stage of axial failure is
needed to supplement the limited data set used to develop the
proposed model and provide verification of the assumptions
used to develop the model.
ACKNOWLEDGMENTS
This work was supported in part by the Pacific Earthquake Engineering
Research Center through the Earthquake Engineering Research Centers
Program of the National Science Foundation under Award No. EEC-
9701568. This support is gratefully acknowledged. The experimental studies
were conducted in the research laboratories of PEER at the University of
California, Berkeley, Berkeley, Calif.
NOTATION
Abar = area of one longitudinal reinforcing bar
Ag = gross concrete area of column
Asl = area of longitudinal reinforcement
Ast = area of transverse reinforcement
Atens = portion of reinforcing bar in tension due to plastic
deformations
db = diameter of longitudinal reinforcing bar
dc = depth of core (centerline to centerline of ties)
f ′c = concrete compressive strength
fyl = yield strength of longitudinal reinforcement
fyt = yield strength of transverse reinforcement
L = clear height of column
Mp = plastic moment capacity of a single longitudinal reinforcing
bar
N = normal force on inclined shear-failure plane
nbars = number of longitudinal reinforcing bars
P = axial load
Po = axial capacity of the undamaged column, 0.85f c′ (Ag –
Asl) + fylAsl
Ps = axial load supported by longitudinal reinforcement
s = spacing of transverse reinforcement
V = shear
Vd = shear resistance due to dowel action of longitudinal steel
Vsf = shear-friction force along inclined shear-failure plane
∆/L, (∆/L)axial = drift ratio of column, drift ratio at axial failure
θ = angle from horizontal of critical shear-failure plane
µ, µt , µm = effective coefficient of friction, coefficient of friction for
total capacity model, coefficient of friction for maximum
capacity model
ACI Structural Journal/July-August 2005
REFERENCES
1. Kabeyasawa, T.; Tasai, A.; and Igarashi, S., “An Economical and
Efficient Method of Strengthening Reinforced Concrete Columns Against
Axial Load Collapse during Major Earthquake,” PEER Report 2002/02,
Third US-Japan Workshop on Performance-Based Earthquake Engineering
Methodology for Reinforced Concrete Building Structures, Pacific
Earthquake Engineering Research Center, University of California, Berkeley,
Calif., 2002, pp. 399-412.
2. Kato, D., and Ohnishi, K., “Axial Load Carrying Capacity of R/C
Columns under Lateral Load Reversals,” PEER Report 2002/02, Third US-
Japan Workshop on Performance-Based Earthquake Engineering
Methodology for Reinforced Concrete Building Structures, Pacific
Earthquake Engineering Research Center, University of California, Berkeley,
Calif., 2002, pp. 247-255.
3. Nakamura, T., and Yoshimura, M., “Gravity Load Collapse of Reinforced
Concrete Columns with Brittle Failure Modes,” Journal of Asian Architecture
and Building Engineering, V. 1, No. 1, 2002, pp. 21-27.
4. Tasai, A., “Residual Axial Capacity and Restorability of Reinforced
Concrete Columns Damaged due to Earthquake,” PEER Report 1999/10,
US-Japan Workshop on Performance-Based Earthquake Engineering
Methodology for Reinforced Concrete Building Structures, Pacific
Earthquake Engineering Research Center, University of California, Berkeley,
Calif., 1999, pp. 191-202.
5. Tasai, A., “Residual Axial Capacity of Reinforced Concrete Columns
during Shear Deterioration,” PEER Report 2000/10, Second US-Japan
Workshop on Performance-Based Earthquake Engineering Methodology
for Reinforced Concrete Building Structures, Pacific Earthquake Engineering
Research Center, University of California, Berkeley, Calif., 2000, pp. 257-267.
6. Yoshimura, M., and Yamanaka, N.,“Ultimate Limit State of RC Columns,”
PEER Report 2000/10, Second US-Japan Workshop on Performance-
Based Earthquake Engineering Methodology for Reinforced Concrete
Building Structures, Pacific Earthquake Engineering Research Center,
University of California, Berkeley, Calif., 2000, pp. 313-326.
7. Elwood, K. J., and Moehle, J. P., “Shaketable Tests and Analytical
Studies on the Gravity Load Collapse of Reinforced Concrete Frames,
PEER Report 2003/01, Pacific Earthquake Engineering Research Center,
University of California, Berkeley, Calif., 2003.
8. Holmes, W. T., “Risk Assessment and Retrofit of Existing Buildings,”
Paper No. 2826, Proceedings of the 12th World Conference on Earthquake
Engineering, New Zealand Society for Earthquake Engineering, Upper
Hutt, New Zealand, 2000.
9. Lynn, A. C., “Seismic Evaluation of Existing Reinforced Concrete
Building Columns,” PhD dissertation, Department of Civil and Environmental
Engineering, University of California, Berkeley, Berkeley, Calif. 2001.
10. Sezen, H., “Seismic Response and Modeling of Reinforced Concrete
Building Columns,” PhD dissertation, Department of Civil and Environ-
mental Engineering, University of California, Berkeley, Berkeley, Calif.,
2002.
11. Kim, J. H., and Mander, J. B., “Truss Modeling of Reinforced Concrete
Shear-Flexure Behavior,” MCEER-99-0005, Multidisciplinary Center for
Earthquake Engineering Research, Buffalo, N.Y., 1999.
12. Pantazopoulou, S. J., “Detailing for Reinforcement Stability in RC
Members,” Journal of Structural Engineering, V. 124, No. 6, 1998, pp. 623-632.
13. Mattock, A. H., and Hawkins, N., “Shear Transfer in Reinforced
Concrete—Recent Research,” Journal of Prestressed Concrete Institute, V. 17,
No. 2, 1972, pp. 55-75.
14. Mattock, A. H., reader comments of “Influence of Concrete Strength
and Load History on the Shear Friction Capacity of Concrete Members” by
J. Walraven, J. Frenay, and A. Pruijssers, PCI Journal, V. 33, No. 1, 1988,
pp. 165-166.
15. Mau, S. T., and Hsu, T. T. C., reader comments of “Influence of Con-
crete Strength and Load History on the Shear Friction Capacity of Con-
crete Members” by J. Walraven, J. Frenay, and A. Pruijssers, PCI Journal, V.
33, No. 1, 1988, pp. 166-168.
16. Krawinkler, H.; Medina, R.; and Alavi, B., “Seismic Drift and
Ductiltity Demands and Their Dependence on Ground Motions,” Engineering
Structures, V. 25, No. 5, 2003, pp. 637-653.
17. Moehle, J. P.; Elwood, K. J.; and Sezen, H., “Gravity Load Collapse
of Building Frames during Earthquakes,” S. M. Uzumeri Symposium: Behavior
and Design of Concrete Structures for Seismic Performance, SP-197, S. A.
Sheikh and O. Bayrak, eds., American Concrete Institute, Farmington
Hills, Mich., 2002, pp. 215-238.
18. Pujol, S., “Drift Capacity of Reinforced Concrete Columns Subjected
to Displacement Reversals,” PhD thesis, School of Civil Engineering,
Purdue University, West Lafayette, Ind., 2002.
587