Fatigue Strength of Concrete

COE Report No.
24
FATIGUE OF CONCRETE SUBJECTED TO BIAXIAL
LOADING IN THE COMPRESSION REGION
By
Technical Report of Research
Supported by the
Under Cooperative Agreement
DOT 05-C-AT-UIUC
FEDERAL AVIATION ADMINISTRATION
DEPARTMENT OF CIVIL AND ENVIRONMENTAL ENGINEERING
UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
ADVANCED TRANSPORTATION RESEARCH AND
ENGINEERING LABORATORY (ATREL)
January 2004
S. P. SHAH
Bin MU
Final report (April 1, 02 March 31, 03) 2

ACKNOWLEDGEMENT/DISCLAIMER

The final report is prepared from a study conducted in the Center of Excellence for
Airport Pavement Research. Funding for the Center of Excellence is provided by the
Federal Aviation Administration under Research Grant Number 03-128/DOT-95-C-
001/A18. The Center of Excellence is maintained at the University of Illinois at Urbana-
Champaign and is in partnership with Northwestern University and the Federal Aviation
Administration. Dr. Patricia Watts is the FAA-COE Program Director and Dr. Satish
Agrawal is Manager of the FAA Airport Technology R&D Branch. The authors also
acknowledge the support from the NSF Center for ACBM, Northwestern University,
during the course of this investigation.

The contents of this report reflect the views of the authors who are responsible for the
facts and accuracy of the data presented within. The contents do not necessarily reflect
the official views and policies of the Federal Aviation Administration. This report does
not constitute a standard, specification or regulation.


Table of Contents:

1. Project Summary.......4
2. Research Background......6
3. Experimental Details for Concrete Cylinders under
Static and Fatigue Compression........11
4. Failure Mechanisms of Concrete Cylinders under
Static and Fatigue Compression17
5. Instrumentation and Test Control for Concrete Hollow Cylinders under
Combined Compression and Torsion.........20
6. Experimental Results for Concrete Hollow Cylinders under
Combined Compression and Torsion.........27
7. Model for Fatigue Failure of Concrete in c-C-T Region.....36
8. Conclusions....50

References
Appendix I EXPERIMENTAL DATA Compressive test
Appendix II EXPERIMENTAL DATA Combined compressive & torsional test
Appendix III Publications from This Project


1. PROJECT SUMMARY

This is the final report for the project: Fatigue of Concrete Subjected to Biaxial Loading
in the Compression Region, sponsored by the FAA Center of Excellence for Airport
Technology (FAA-COE). The major objective of this study is the material property
characterization in the biaxial c-C-T region, where the principal compressive stress is
larger in magnitude than the principal tensile stress. The aim of this project is to
understand the response of concrete subjected to high amplitude, low cycle biaxial
fatigue loading in the c-C-T stress space and verify/extend the previously proposed
analytical model in/to the whole Compression-Tension (C-T) space. In the project, a great
deal of attention has been placed on the interaction between characterization of the
material response under static loading and establishment of the failure mechanism under
fatigue loading. This research is important in view of the fact that 3D tensile-compressive
stress states occur in airport rigid pavements under traffic loads. The work of this project
can be summarized as below:
Static and fatigue tests were conducted on concrete cylinders and hollow cylinders
subjected to pure compression, and torsion superimposed on compression,
respectively. Closed-loop testing methods to avoid the test instability immediately
after the peak load were systematically studied. Static test results were compared with
fatigue test results with respect to compliance, crack length, or inelastic displacement,
as applicable.
The failure mechanism for pure compressive tests was explained by the Band
Damage Zone model (Jansen and Shah 1997). In the combined compressive and
torsional tests, the failure mechanism was modeled using fracture mechanics.
The previously proposed methods, theories and models in the t-C-T region were
verified with the new experimental results in the c-C-T region. The previous research
conclusions in the t-C-T region were found to be applicable to the whole biaxial C-T
region.

The details will be illustrated in the following chapters. Chapter 2 gives a review of
the research background. In Chapter 3, the detailed experimental procedures for the static
and fatigue compressive tests are introduced. The failure mechanism of the specimens
under the static and fatigue compressive loading is discussed in Chapter 4. Experimental
instrumentation and the controlling method of concrete hollow cylinders subjected to
static combined compression and torsion are presented in Chapter 5. Chapter 6 presents
detailed experimental results for both static and fatigue compressive and torsional tests.
Chapter 7 discusses the two typical failure modes in the c-C-T region. The previously
proposed models in the t-C-T region are also verified with the new results in this chapter.
Chapter 8 is the final chapter containing a summary and conclusions.


2. RESEARCH BACKGROUND

Fatigue behavior of concrete recently has received considerable attention from
investigators. The present state of the art in designing structures against fatigue distress is
largely empirical, gained by many years of experience. A more rational design procedure
requires a clear understanding of the behavior of material subjected to fatigue thus
permitting design for newer applications.

FF
TT
Pavemen
Subgrade
Figure 2.1 A concrete pavement subjected to triaxial and a biaxial
stress state at the top and bottom surfaces

Airport pavements are subjected to repeated high stress amplitude loads due to
passing aircraft. In addition, introduction of newer heavier aircraft subjects these
pavements to increased magnitude of fatigue stresses. Using Westergaard theory a
pavement structure subjected to point load shows the state of stress to be triaxial at the
top and bottom surfaces (Fig. 2.1). When resolved in terms of principal stresses, the
stresses are predominantly in the compression-tension (C-T) region of the triaxial stress
space, where the principal stresses have opposite signs. It is vital to understand behavior
of concrete under multi-axial stresses, such as those present in airport pavements, to be
able to assess reduction in stiffness due to the repeated loading of the structure. A
thorough understanding of the fatigue response of concrete subjected to such loading
would also enable determination of the remaining service life of the pavement.
Characterization of the material response to such loading will enable determination of the
response of the pavement to increasing loading in the future. The information could be

used in a mechanistic design of future airport pavements and for assessing the need for
rehabilitation and/or replacement of old airport pavements. Considering the complexity
of the 3-D fatigue loading case, the investigation began with the 2-D fatigue loading case.
One of the tensile stresses was eliminated for simplicity (Fig. 2.1). Then the concrete
pavement is subjected to biaxial fatigue loading in the C-T region.
Considerable work has been done to understand the response of concrete
subjected to uniaxial stress (monotonically increasing and fatigue loads). Tensile,
compressive and tensile-compressive loading schemes considering constant amplitude,
variable amplitude cyclic loading and random loading have been studied. Fatigue curves,
which predict the number of cycles to failure for a given loading scheme, have been
formulated. The failure is understood to be caused by the propagation of cracks in
tension. Cracks are formed from pre-existing flaws and these cracks propagate due to
cyclic or continued monotonic loading, resulting in increased deformation and
redistribution of stresses. After the crack is initiated, the propagation is studied using
fracture mechanics. Fracture mechanics has been applied to gain insight into the failure
mechanism of concrete under tensile loading and successfully applied to predict the
response of structures subjected to monotonic and fatigue loading. Relatively few studies
have been conducted to obtain the response of concrete subjected to multiaxial loading.
The difficult and tedious experiments required for fatigue research under multiaxial
loading are the reason for the scarcity of experimental results under such loading. Most of
the experimental results in the literature pertain to biaxial and triaxial compressive
loading (both monotonic and fatigue). The first systematic investigation into the
performance of concrete subjected to biaxial stresses was conducted by Kupfer et al.
(1969, 1973). In their investigation concrete plates were subjected to in-plane loading for
different biaxial stress combinations spanning the entire biaxial stress space. The results
from this investigation continue to be used today for developing biaxial constitutive
relationships of concrete. The mode of failure as seen in the crack patterns is different in
the three regions of biaxial stress space (C-C, C-T, and T-T). In the C-C region,
microcracking parallel to the free surface is the main damage mechanism. The behavior
of concrete in the C-T region is similar to that in the C-C region as long as the tensile
stress is less than 1/15
th
of the compressive stress. For larger tensile stresses concrete fails

by a single crack forming perpendicular to the maximum principal tensile strain. Similar
behavior is also observed for specimens in the T-T region. Several continuum mechanics
based theories have been used to model the material response under such loading
conditions (Kupfer and Gerstle 1973; Gerstle 1981). These constitutive formulations are
valid for monotonic loading up to ultimate stress (peak stress). No data are available in
the literature for the behavior of concrete subjected to biaxial loading after the peak
stress. Controlled experiments are needed to obtain the post-peak part of the response in
the C-T region. Complete understanding of fatigue response of concrete in the C-T region
of the biaxial stress space is still lacking. A thorough understanding of the material
response in the C-T space under a monotonically increasing load is essential to establish
the mechanism of failure of concrete. The evolution of damage under fatigue can then be
studied and a fatigue damage law can be established.

| | | |
2 1
>
0 0
1 2
> =
| | | |
2 1
=
0 0
2 1
< =
| | | |
2 1
<
FF
TT
v
iv
F
ii
TT
iii
F
i
FF
TT
1

t-C-T c-C-T

Figure 2.2 Biaxial C-T region (Subramaniam 1999)

The biaxial C-T region is bounded by uniaxial compression at one end and by
uniaxial tension at the other end (Fig. 2.2). The C-T region in the biaxial stress space
represents a transition from a cracking type failure to a crushing type failure as the ratio
of the tensile principal stress to the compressive principal stress is varied from infinite to
zero. There is also a significant change in the strength of the material associated with a
change in the mode of failure in this region; the strength of concrete subjected to uniaxial

tensile stresses is approximately a tenth of the uniaxial compressive strength. Hence it is
of interest to investigate the behavior of concrete in the C-T region and identify the
failure mechanisms at different stress ratios.
In a previous FAA-COE sponsored project (Fatigue of Concrete Subjected to
Biaxial Loading in the Tension Region, 1999) an investigation to characterize the quasi-
static and low-cycle fatigue response of concrete subjected to biaxial stresses in the
tensile-compression-tension (t-C-T) region, where the principal tensile stress is larger in
magnitude than the principal compressive stress, was performed at the Center for
Advanced Cement Based Materials (ACBM) of the Northwestern University. The
damage imparted to the material was examined using mechanical measurements and an
independent nondestructive evaluation (NDE) technique based on vibration
measurements. The failure of concrete in the t-C-T region was shown to be a local
phenomenon under quasi-static and fatigue loading, wherein the specimen fails owing to
a single crack. The crack propagation was studied using the principles of fracture
mechanics. It was shown that the crack propagation resulting from the t-C-T loading
could be predicted using mode I fracture parameters. It was observed that crack growth in
constant amplitude fatigue loading is a two-phase process: a deceleration phase followed
by an acceleration stage. The quasi-static load envelope was shown to predict the crack
length at fatigue failure. A fracture-based fatigue failure criterion was proposed, wherein
the fatigue failure can be predicted using the critical mode I stress intensity factor. A
material model for the damage evolution during fatigue loading of concrete in terms of
crack propagation was proposed. The crack growth acceleration stage was shown to
follow the Paris law. The model parameters obtained from uniaxial fatigue tests are
shown to be sufficient for predicting the considered biaxial fatigue response. Details of
the experimental work and the analytical model for the material response are available in
the references (Subramaniam et al. 1998; Subramaniam et al. 1999, Subramaniam et al.
2000, and Subramaniam et al. 2002).
The proposed analytical model can be implemented in mechanistic or numerical
pavement response prediction models. However, for proper implementation of the
numerical procedures, a material model that predicts the material response under any
biaxial stress state in the Compression-Tension (C-T) region is required. Hence, it is of

interest to extend the model to the compression-Tension-Compression (c-T-C) stress state
(where the magnitude of the principal compressive stress is larger than that of the
principal tensile stress) within the Compression-Tension (C-T) stress space. In this
research project, the existing test setup was utilized to generate biaxial stresses in the c-
C-T region, by applying torsion with superposed axial compression to hollow concrete
cylinders. By varying the magnitudes of the torsional and axial forces the entire tension
compression stress state can be obtained. The crack formation and propagation under
controlled conditions for both static and fatigue loading were studied. The research
started with the uniaxial compressive test (point - iv in Fig. 2.2). Then the behavior of
concrete subjected to combined compressive and torsional test (point-v) in the c-C-T
region was studied.


3. EXPERIMENTAL DETAILS FOR CONCRETE CYLINDERS
UNDER STATIC AND FATIGUE COMPRESSION

Concrete cylinders (4-in diameter and 8-in length) were used in the static and fatigue
compressive test. The mixture proportions by weight of the constituents were, cement:
water: coarse aggregate: fine aggregate = 1.0:0.5:2.0:2.0.
In the static test, the samples were unloaded and reloaded in the post-peak period
at 90%, 80% and 70% of the peak strength, respectively to obtain the compliance of the
specimen in the post peak period. A closed-loop testing procedure was used with a
circumferential extensometer mounted around the cylinders until the first unloading
started at 90% of the peak strength in the post-peak period. The circumferential
displacement was used as the feedback signal in the MTS. At the first unloading process
at 90% of the peak strength, the control mode was shifted to displacement for the rest of
the testing region.
In the fatigue test, three different load ranges (90% - 5%, 80% - 5% and 76% -
5%) were employed in a sinusoidal waveform at a frequency of 2 Hz. The test was in
load control. To prevent the data file from becoming too large, a continuous data
acquisition process was designed such that data were acquired every seven load cycles, or
every time the stiffness of the specimen changed by a certain threshold value. Within
each load cycle data were acquired in a time increments of 0.05 s.
For each loading case at least three samples were loaded in a Material Testing
System (MTS). The typical load-displacement relationships for both the static and fatigue
tests are shown in Fig. 3.1. From Fig. 3.1(a), there is a progressive decrease in the
structural stiffness due to crack initiation and propagation as evidenced by the decrease in
the slope of the loading and unloading curves in the post peak. In the fatigue result (Fig.
3.1(b)), it can also be observed that there is a steady decrease in the axial stiffness of the
specimen with repeated loading as seen by the decreasing slope of the load-unloading
curves.


0
20
40
60
80
100
0.00 0.02 0.04 0.06 0.08 0.10
Axial Displacement (in)
L
o
a
d

(
k
i
p
)
C
i
C
r
0
20
40
60
80
100
0 0.02 0.04 0.06 0.08 0.1
Axial Displacement (in)
L
o
a
d

(
k
i
p
)
(a) (b)
Figure 3.1 Load-displacement response of a cylinder in compressive test
(a) Static test (b) Fatigue test

The axial normalized stiffness/compliance represents the secant
stiffness/compliance calculated between the minimum and maximum load levels divided
by the initial stiffness/compliance (K
r
/K
i
or C
r
/C
i
). The damage evolution in concrete
during compressive fatigue, in term of the measured stiffness/compliance, follows a
three-stage trend (Fig. 3.2). The normalized cycles in Fig. 3.2 are obtained from the
current cycle divided by the total number of fatigue cycles (N/N
f
). The curve is S-shaped.
A remarkable reduction in the stiffness during the first few cycles (region I) is followed
by a region of gradual, almost linear change (region II). This is followed by a sharp
decrease in stiffness prior to failure (region III). Fig. 3.3 shows a plot of the slope of the
decrease in compressive stiffness, dK/dN, in stage II of fatigue response versus the
fatigue life, N
f
, for all the specimens (where K, N and N
f
are stiffness, cycle and fatigue
life, respectively). The response of all the fatigue specimens tested at three different
loading range follows a linear trend. The best-fit relationship between the slope of stage
II and the fatigue life is found to be:

log(N
f
)=-0.9444*log(dK/dN)+2.5134 (3.1)

0.6
0.8
1
1.2
0 0.3 0.6 0.9 1.2
Normalized cycle
N
o
r
m
a
l
i
z
e
d

s
t
i
f
f
n
e
s
s
sample-1
sample-2
sample-3
sample-4
I
II
III

Figure 3.2 Normalized stiffness vs. normalized cycles (76%-5%)

0
1
2
3
4
5
-2 -1 0 1
Log(dK/dN)
L
o
g
(
N
f
)
90%-5% load
80%-5% load
76%-5% load
Log(N
f
)=-0.9444Log(dK/dN)+2.5134

Figure 3.3 Relationship between the decrease rate of stiffness
in stage II of the fatigue response and fatigue life

It is interesting to observe that the relationship between the slope of stage II and
fatigue life is independent of the load range (Fig. 3.3). This suggests that the fatigue life
can be determined or predicted from Eq. (3.1) for any load range if the slope in stage II of
the fatigue response is known. In practice, stage I may represent the first few months of
use and stage II may represent a long period of time before the airport pavement needs to
be repaired. Since stage I is quite short, and the slope of stage II shows a nearly linear
trend, engineers are able to predict the fatigue life of the airport pavement by conducting
the fatigue test over just a few cycles. Whenever stage II and the stiffness change rate in
stage II are determined, Eq. (3.1) can be employed.

60
70
80
90
100
1 1.5 2 2
Normalized compliance
%

p
o
s
t
-
p
e
a
k

l
o
a
d
static
fatigue: 90%-5%
fatigue: 80%-5%
fatigue: 76%-5%
0
20
40
60
80
100
0 0.02 0.04 0.06 0.08 0.1
Displacement (in)
L
o
a
d

(
k
i
p
)
static
fatigue: 90%-5%
fatigue: 80%-5%
fatigue: 76%-5%
.5
(a) (b)
Figure 3.4 A comparison of static and fatigue response
(a) Load displacement (b) Load - compliance
A comparison of the static response and fatigue response is shown in Fig. 3.4,
both in terms of axial displacement (Fig. 3.4(a)) and axial compliance (Fig. 3.4(b)). In
Fig. 3.4(a) the curves represent the static load-displacement and each data point
represents a different specimen tested in fatigue. It can be seen that the axial
displacement at fatigue failure, which is defined as the maximum displacement at the

higher fatigue loading level, is not comparable to post-peak displacement obtained from
the static response; the axial displacement at fatigue failure are considerably smaller than
the axial displacement at the corresponding load in the static post-peak. Therefore, a
failure criterion based on static deflections may not be suitable for plain concrete
subjected to uniaxial compressive fatigue. Similar observations have been reported for
the concrete airport pavement in the biaxial t-C-T region. However, from Fig. 3.4(b) it
can be seen that the compliance at the fatigue failure compares favorably with the
compliance at the corresponding load in the static post peak period response. This
suggests that the compliance at fatigue failure can be obtained from the static response.
Hence the static response acts like an envelope to the fatigue response when framed in
terms of compliance.

Figure 3.5 Compliance change rate versus compliance
-8
-6
-4
-2
0
0.9 1 1.1 1.2 1.3 1.4 1.5
Normalized compliance, C
r
/C
i
L
o
g
(
C
/
N
)
sample-1 sample-2
sample-3 sample-4
sample-5 sample-6
sample-7 sample-8
sample-9 sample-10
deceleration acceleration
inflection point

The rate of compliance change ( N C / ) as a function of compliance for all of
the specimens is shown in Fig. 3.5. The rates of compliance follow a two-stage process; a
deceleration stage and an acceleration stage up to failure. There is a distinct inflection
point marking a critical compliance where the rate of compliance change crosses over

from deceleration to acceleration. This critical compliance in the fatigue test corresponds
to the compliance at the peak load in the static test, as can be seen from Fig. 3.6.

0.8
0.9
1
1.1
1.2
70 75 80 85 90 95 100
Max. load in fatigue as % of peak
N
o
r
m
a
l
i
z
e
d

c
o
m
p
l
i
a
n
c
e
at inflection in fatigue test
at the peak in static test

Figure 3.6 Comparison of the compliance at fatigue inflection
point and that at static peak load

All of the results at point-iv in the c-C-T region (Fig. 2.2) are similar to those in
the t-C-T region. This means that the theory, method and conclusions, except for the
terms related to the length of crack, obtained from the previous project can be easily
extended to the point-iv. As to the crack information, there is no simple way to
characterize such information in the compression test, because the cracks are not
distinctly localized as that in tension. However, using the concept of inelastic
displacement to replace the term of crack length, the previously proposed theoretical
model can still be applied to the compression test. The detailed modeling work will be
presented in Chapter 4 & 7.


4. FAILURE MECHANISMS OF CONCRETE CYLINDERS UNDER STATIC
AND FATIGUE COMPRESSION

In compression, the damage zone has a nonuniform distribution as shown in Fig. 4.1a.
From a previous study (Jansen and Shah 1997), this damage zone can be modeled as a
band (Fig. 4.2b). Using the band model, two distinct areas within the concrete sample can
be defined, bulk concrete and damage zone. Localization initiates at the peak stress
(Torrenti et. al. 1993) or just prior to the peak stress (Shah and Sankar 1987). In either
case, the shape of the stress-strain curve up to the peak can be considered approximately
the same in the bulk concrete and the eventual damage zone. During prepeak the same
amount of energy is dissipated in the bulk concrete and the damage zone due to
microcracking (Fig. 4.1c). During postpeak, the bulk concrete unloads. Additional energy
is dissipated in the damage zone. If one assumes that the total strain in the damage zone,
z
, is composed of two parts: one same as the bulk concrete,
b
and the additional
inelastic strain in the localized damage zone, then one can write
l
b z
/ + = (4.1)
where l is the length of the damege zone. The inelastic post-peak displacement is
obtained by substracting the prepeak displacement response of the concrete specimen
from the postpeak displacement. Displacement is an approximate measure of the
inelastic, localized deformations occuring during strain softening. It is assumed to be
independent of the length of the damage zone, as shown in Fig. 4.1c. The assumption that
is independent of l implies that the postpeak inelastic behavior of the concrete
specimen in compression is the same no matter the length of the specimen. This
assumption has been verified by the previous study (Jansen and Shah 1997) by pressing
concrete cylinders with different lengths.


l /
z
Postpeak
Energy
Prepeak
Energy
F
F
l
L
Band-damage zone
(a) (b) (c)
Figure 4.1 Band model for localization behavior in compression: (a) Distributed
damage; (b) Band-damage distribution; (c) Bulk concrete behavior and damage
zone behavior

This band damage zone model is employed to explore the fatigue failure
mechanism characteristic of concrete airport pavements. Fig. 4.2 gives the comparison of
these inelastic postpeak displacements at different loading levels from the static test and
the fatigue test. In a fatigue test, the inelastic post-peak displacement is calculated as the
difference between the failure displacement and the displacement at the compliance
inflection point. A good agreement is found between static and fatigue tests. This
suggests that the static postpeak displacements, which approximately describe the
localization displacements in static compression, govern the fatigue failure mechanism.
The fatigue failure under compression is also a localized phenomenon. The plot of the
load as a function of the inelastic postpeak displacement obtained from static tests acts as
the envelope curve for fatigue test.


0
20
40
60
80
100
0 0.01 0.02 0.03 0.04 0.05 0.06
Inelastic post-peak displacement, (in)
L
o
a
d

(
k
i
p
)
static
fatigue: 90%-5%
fatigue: 80%-5%
fatigue: 76%-5%

Figure 4.2 A comparison of static and fatigue response: Load post-peak
inelastic displacement in damage zone


5. INSTRUMENTATION AND TEST CONTROL FOR CONCRETE HOLLOW
CYLINDERS UNDER COMBINED COMPRESSION AND TORSION

5.1 Specimens, loading and instrumentation
Concrete hollow cylinders subjected to combined torsion and compression were used to
represent the loading condition of point-v (Fig. 2.2). Concrete hollow cylinders were 4-in
in outer diameter, 2.25-in in inner diameter, and 8-in in length. The mixture proportion is
the same as that used in the compressive test, i.e. cement: water: coarse aggregate: fine
aggregate = 1.0:0.5:2.0:2.0 (by weight). The ends of the hollow cylinders were reinforced
with 0.5 in x 0.5 in steel wire mesh cages that were embedded into the concrete to ensure
that failure happens in the gage section away from the ends. The mesh extends a length of
2.5 in from both ends and is continuous in the circumferential direction (Fig. 5.1a). The
mesh cage has a diameter of approximately 3.6 in. The average age of specimens used for
the test was over 6 months to ensure mature concrete.

T
P

(a) (b)
Figure 5.1 Hollow cylinder specimens (a) Geometry (b) Stress
state under compression and torsion

During the test, the ratio of compression to torsion was kept constant so as to keep
the ratio of the two principal stresses constant at 1.5 throughout the loading process (i.e.
proportional loading was used). For the hollow cylinder shown in Fig. 5.1b, the shear
stress at the mid-thickness of the cylinder due to applied torsion T, can be calculated as:
J Tr / = (5.1)
where J is the section modulus equal to and R 2 / ) (
4 4
0
i
R R
0
and R
i
are the inner and
outer radii, respectively, of the hollow cylinder. The radius at the mid-thickness, r, is
equal to (R
0
+R
i
)/2. The normal stress resulting from the applied axial force, P, can be
calculated as:

) (
2 2
0 i
R R
P
(5.2)
The ratio of shear to axial stress in a stress element at the mid-thickness of the cross-
section is given as:

) (
) (
2 2
0
0
i
i
R R P
R R T
+
+
=
(5.3)
The principal stresses can be calculated as:

2
2
2 , 1
2 2
= (5.4)
Hence, the ratio of the principal tensile and compressive stresses can be calculated as:
c =
+
+
=
) 1
2
1 (
1
2
1
2
2
2
1
(5.5)
Therefore,

) 1 (
c
c
(5.6)
where c . Substituting Eq. (5.6) into Eq. (5.3) yields: c =

) (
) (
) 1 (
0
2 2
0
i
i
R R
R R
c
c
P
T
+
+
= (5.7)

For the case of c equal to 0.667 (where the magnitude of principal compressive stress is
50% larger than that of the principal tensile stress), the ratio of axial force to torque is
2.88 ft
-1
.
Instrumentation was specially designed to measure rotational and axial
deformation of the gage section of the specimen. Two steel rings were mounted at the
ends of the 2.5 in long gage section. The relative movements of the two rings were
monitored using linear variable differential transducers (LVDTs) mounted on one of the
rings. Four equally spaced LVDTs along the circumference were used to measure the
relative axial displacement. The relative rotation of the two rings was measured by two
LVDTs mounted on one ring in the horizontal direction (tangential to the ring). The
measured change in length is converted to the change of angle between the two rings
(Fig. 5.2). The axial LVDTs have a range of 0.02 in and rotational LVDTs have a range
of 0.01 in.

(a) (b)
Figure 5.2 Experimental setup (a) LVDTs (b) Connection to MTS

The procedure to mount the LVDTs and specimens is described below.
Mount the two rings on the hollow cylinder. Ensure that the two rings are parallel
and level. Then mount the LVDT fixtures.

Cover the rings and fixtures with plastic sheet before applying epoxy, so the
epoxy will not flow on the rings and fixtures.
Attach the two steel caps to the MTS fixtures. The specimen will be connected
between these two steel caps by epoxy.
Apply epoxy cement to the specimen and the two steel caps. Slowly lower the
actuator of the MTS in load control and apply an initial compression of
approximately 0.2 kip. Shift to torque control and set the torque to zero. Leave it
in load and torque control for 24 hours.
After 24 hours, remove the plastic sheet, and reduce the applied axial load to zero.
Mount and zero the LVDTs.

5.2 Test control
The complete load-deformation response of unnotched concrete hollow cylinders
subjected to compression and torsion is difficult to obtain because of test instabilities.
The test instability is a result of snapback in the post-peak part of the torque-twist or
force-displacement response of these specimens. The three reasons for test instability are
summarized by Rokugo et al. (1986): a) the load is increased at a constant rate, or b) the
stiffness of the testing machine is lower than the specimen stiffness, even through the
displacement rate is kept constant, or c) the elastic energy in the specimen is larger than
the total fracture energy of the specimen. Reason a) is distinctly present in a test, while
reasons b) and c) can be either distinctly or both present in a test.
The sudden failure of specimen in cases a) and b) can be controlled by testing
concrete in a stiff testing machine at a constant displacement rate. In case c) the failure of
the specimen itself is unstable and the use of a stiff testing machine at a constant
displacement rate is not sufficient to obtain the complete load-displacement response.
The load-displacement response in case c) exhibits snapback in the descending branch of
the load-displacement wherein both force and displacement simultaneously decrease. The
snapback phenomenon is a result of localization of damage and simultaneous elastic
recovery of the unchanged portion of the specimen, and is dependent on the specimen
dimension. The total fracture energy required to break the specimen remains constant.
The elastic energy stored in the undamaged portion, however, increases with increasing

specimen length. Therefore the phenomenon of the simultaneous decrease of load and
displacement becomes more pronounced as the gage length is increased as shown in Fig.
5.3 (Jansen 1996).

Failure
zone
L L
1
L
f

+
L
1
-L
f
L-L
f
P
P
P,

L
f
L
1
L
P

e

=

f

P,
(a) (b) (c)
Figure 5.3 Test instability due to snapback in the post-peak part of force-
displacement response (a) Specimen with a failure zone of finite length, L
f
, (b)
Elastic response (P ~
e
) and response of failure zone (P ~
f
), (c) Combined load-
displacement response of specimen with different gage lengths, (
f e
+ = ).

In order to limit test instability and therefore have stable control in specimens that
exhibit snapback behavior, a measurable test parameter, which increases monotonically
throughout the destruction of the specimen is needed. In this way, using a failure
sensitive parameter as the feedback signal can eliminate the test instability. From Fig. 5.3
it can be seen that as the gage length decreases the elastic displacement, , reduces
proportionally without affecting the response of the failure zone. Therefore a test
parameter which monitors the response of a very small gage length, which includes the
failure zone would be an appropriate feedback signal to control snapback. Okubo and
Nishimatsu (1985) proposed a control method in which a linear combination of load and
e

displacement is used as the feedback signal in a closed loop servo controlled test
machine. This feedback signal was used to obtain the complete load-displacement
response of rock specimens that exhibited snapback when subjected to unaxial
compression. The feedback signal is defined as

0
K
F
Signal Feedback = (5.8)
where is axial displacement of the specimen, and K
0
is the initial tangent stiffness of
the entire specimen. F is axial force, and is a coefficient, which must be a positive
number less than 1.0. Such a signal has been also successfully put into use to obtain the
complete load-displacement response of high strength concrete specimens subjected to
compression and exhibiting snapback in the descending portion of the load-displacement
response (Jansen and Shah 1993). For a specimen of length L, where the damage
localizes in a region of finite length L ) 1 ( and the rest of the material L behaves
elastically, the calculated value of this feedback signal by Eq. (5.8) corresponds to the
displacement of the failure zone. The symbol physically represents the fraction of the
specimen that unloads elastically. Under tension, the failure in the concrete localizes
along a plane perpendicular to the applied force and the value of may be chosen close
to one. High strength concrete specimens subjected to uniaxial compression have a
failure zone with finite dimension and a value of equal to 0.6 was found to produce the
best results in 8-in high specimens (Jansen and Shah, 1993).
In this study of concrete hollow cylinders subjected to combined compression and
torsion, a feedback signal which is a linear combination of gage rotation and torque was
used to obtain the complete torque-gage rotation curves. It has the form

0
K
T
Signal Feedback = (5.9)
where K
0
is the initial tangent modulus of the torque-gage rotation curve and is
approximately 25,000 ft-lb/deg. The applied torque and the rotation of the gage section
are T and , respectively.
Theoretically, Eq. (5.9) can eliminate the elastic snapback at the peak point of a
torque-rotation curve as illustrated in Fig. 5.3. However, the selection of the parameter,
, is critical in the experiment. If is too small, elastic snapback may not be

completely eliminated (Fig. 5.4a) and if is too large, the feedback signal described in
Eq. (5.9) may not be monotonic (Fig. 5.4b). Since the initial tangent modulus, K
0
, can
only be estimated from the previous test results, it might also have influence on the value
of . In the current static test, a value of approximate 0.8 for is found to give a stable
control (Fig. 5.4c).

=
(1-)L
(1-)L
(1-)L
L
L
L
L
L
L
P
P

P
=
'
P
=
e
P
P
'
P

P
e

P

e

'

(a) (b) (c)
Figure 5.4 Test instability of force-displacement response
(a) too small , (b) too large , (c) suitable

In the test, two combined signals are defined as input signals (calculation signal
type) in the MTS configuration file. The first is given by Eq. (5.9) and a rate of 2.0x10
-5

deg/sec was used. The second was used to achieve a constant ratio of the principal

compressive stress to the principal tensile stress. The second input signal given by Eq.
(5.7) was defined on the axial control channel and was kept constant (close to zero)
throughout the loading process.


6. EXPERIMENTAL RESULTS FOR CONCRETE HOLLOW CYLINDERS
UNDER COMBINED COMPRESSION AND TORSION

6.1 Static combined compression and torsion
A typical torque vs. gage rotation curve for tests under combined compression and
torsion is shown in Fig. 6.1. The torque and gage rotation values have been normalized
with respect to peak torque and the gage rotation corresponding to the peak torque,
respectively. During the test, the samples were unloaded and reloaded in the post-peak
period at approximately 90%, 80% and 70% of the peak strength, respectively, to obtain
the compliance of the specimen in the post peak period. Closed-loop testing was used
with a combined signal given by Eq. (5.9) until the first unloading started at 90% of the
peak strength in the post-peak period. At the first unloading point at 90% of the peak
strength, the control mode was shifted to the rotation gage for the rest of the testing
process. The percentage increase in rotational compliance is calculated with respect to the
initial rotational compliance (the most linear part of the pre-peak torque-gage rotation
curve). The rotational compliance corresponding to each unloading-reloading loop
represents secant compliance calculated between the top and the bottom intersection
points of the unloading-reloading curves. The increase in rotational compliance in the
post-peak part can be attributed to the propagation of a crack in the gage region.
Fig. 6.2 shows the extension of the four axial LVDTs with respect to normalized
test time. The crack growth can be traced by the variation of the axial LVDTs. It can be
seen that LVDT 2 & 3 increased while LVDT 1 & 4 decreased after the peak load
(normalized time = 1.0). The difference of the extensions of the four axial LVDTs after
the peak load apparently indicated a misalignment, because no significant variation was
noticed before the peak load. This misalignment was produced by the initiation and
propagation of a single inclined crack as observed in the experiment. This crack formed
between LVDT 2 & 3. The crack was inclined at an angle of approximately 51
o
with
respect to the horizontal (Fig. 6.3).


0
0.2
0.4
0.6
0.8
1
1.2
0 1 2 3 4 5
Normalized gage rotation
N
o
r
m
a
l
i
z
e
d

t
o
r
q
u
e

Figure 6.1 Torque vs. gage rotation for tests in combined signal control

-0.004
-0.002
0
0.002
0.004
0.006
0.008
0 1 2 3 4 5 6 7 8
Normalized time
E
x
t
e
n
s
i
o
n

o
f

a
x
i
a
l

L
V
D
T
s

(
i
n
)
LVDT-1
LVDT-2
LVDT-4
LVDT-3
2
4 1
3

Figure 6.2 Extension of the four axial LVDTs


Figure 6.3 An inclined crack in the hollow cylinder specimen

0
0.0002
0.0004
0.0006
0.0008
1 1.5 2 2
Normalized rotational compliance
A
v
e
r
a
g
e

a
x
i
a
l

d
i
s
p
l
a
c
e
m
e
n
t

(
i
n
)
.5

Figure 6.4 Increase in rotational compliance versus the average axial deformation

The average axial displacement (from the four axial LVDTs) of the specimen
versus the increase in the rotational compliance is plotted in Fig. 6.4. The rotational

compliance is normalized to the initial rotational compliance. It can be seen that the
average axial deformation is monotonically related to the increase in the rotational
compliance of the specimen. This increase in the rotational compliance of the specimen
can be attributed to the propagation of a crack in the gage region.

6.2 Fatigue combined compression and torsion
Fatigue tests were conducted at three different torque ranges. Torque was applied
between two fixed torque levels in a sinusoidal waveform at a frequency of 2 Hz. The
tests were performed in a torque control. The lower limit of the torque was kept as 5% of
the average static peak torque value. Three different upper limits were used
corresponding to 90%, 85% and 80% of the average static peak torque.

0
200
400
600
800
0 0.01 0.02 0.03 0.04 0.05
Gage rotation (deg)
T
o
r
q
u
e

(
f
t
-
l
b
)
90%-5%

Figure 6.5 A typical fatigue response of a specimen

A typical fatigue response of a specimen is shown in Fig. 6.5. There is a steady
increase/decrease in the rotational compliance/stiffness as seen by the decrease slope of
the loading-unloading curves. The change in rotational compliance/stiffness during
fatigue loading for the same specimen is shown in Fig. 6.6. The number of cycles has
been normalized with respect to the number of cycles to failure (N
f
) for the specimen.

The rotational compliance/stiffness represents the secant rotational compliance/stiffness
of the specimen calculated between the minimum and maximum torque levels. The
damage evolution during the fatigue combined compressive and torsional loading, in
terms of the measured rotational stiffness, can be divided in to three stages. There is a
measurable drop in the stiffness in stage I, which lasts for a few cycles. Subsequently in
stage II, there is a gradual, almost linear change suggesting that the damage accrues at a
constant rate in this stage. Stage III is marked by a large and rapid reduction in stiffness
immediately followed by failure.

0.6
0.8
1
1.2
0 0.3 0.6 0.9 1.2
Normalized cycles, N/N
f
N
o
r
m
a
l
i
z
e
d

s
t
i
f
f
n
e
s
s
90%-5%
I II III

Figure 6.6 Decrease in rotational stiffness during the fatigue test

Fig. 6.7 shows a plot of the slope of the decrease in rotational stiffness, dK/dN, in
stage II of fatigue response versus the fatigue life, N
f
, for all the specimens (where K, N
and N
f
are stiffness, cycle and fatigue life, respectively). The response of all the fatigue
specimens tested at three different loading range follows a linear trend. The relationship
between the slope of stage II and the fatigue life was obtained by linear regression:
log(N
f
)=-0.8213*log(dK/dN)+2.7572 (6.1)
Similar to the previous research in the t-C-T region and compression tests of c-C-
T region, the relationship between the slope of stage II and fatigue life is independent of

the load range. This suggests that the fatigue life can be determined or predicted from Eq.
(6.1) for any torque-range if the slope in stage II of the fatigue response is known.

1
2
3
4
5
6
-4 -2 0 2
Log(dK/dN)
L
o
g
(
N
f
)
90%-5% load
85%-5% load
80%-5% load
Log(N
f
)=-0.8213Log(dK/dN)+2.7572

Figure 6.7 Correlation between the decrease rate in rotational stiffness
in stage II of the fatigue response and fatigue life

The axial displacements measured by the four axial LVDTs mounted on the same
specimen during the fatigue test are shown in Fig.6.8. The trends in axial displacement
correspond well with the observed change in rotation stiffness. The axial displacements
also show a three-stage change, I, II & III, similar to rotation stiffness. This suggests a
possible connection between the two phenomena. LVDT-3 shows an increase in axial
displacement while LVDT-1 shows a decrease in the measured axial displacement. In the
experiment, it was observed that an inclined crack initiated and propagated between
LVDT-2, 3 & 4 (Fig.6.8). From the static response, it has been established that the
different extensions of axial LVDTs are explained by crack propagation in the gage
section. Thus, formation and propagation of a crack in the gage section appears to be the
mechanism of failure for the fatigue loading.

The extensions of axial LVDTs suggest that the damage localized into a single
crack in the first few cycles of loading. This corresponds to a measurable reduction in
stiffness in the first few cycles (stage I). The crack propagates at a steady rate in the
subsequent cycles. This corresponds to gradual reduction in the rotational stiffness in
stage II. Failure occurs when the crack grows to a size that cannot safely support the
applied load.

-0.0001
0
0.0001
0.0002
0.0003
0.0004
0 0.3 0.6 0.9 1.2
Normalized cycles (N/N
f
)
E
x
t
e
n
s
i
o
n

o
f

a
x
i
a
l

L
V
D
T
s

(
i
n
)
LVDT-1
LVDT-2
LVDT-3
LVDT-4
2
4
3
I II III

Figure 6.8 Extensions of axial LVDTs during the fatigue test

6.3 Comparison between static and fatigue results
As known, for static test, the specimen can still sustain loading after the peak load. The
static failure displacement at a certain load level should be in the post-peak part of the
load displacement curve. A comparison of gage rotation at fatigue failure and that at the
corresponding load level in the post-peak part of the static response is shown in Fig. 6.9.
Each data point in the graph represents a different specimen loaded in fatigue. It can be
seen that the rotation at fatigue failure, which is defined as the maximum rotation at the
higher fatigue loading level, is not comparable to rotation obtained from the static post-
peak response; the rotation at fatigue failure are considerably smaller than the rotation at
the corresponding load in the static post-peak period. Therefore, a failure criterion based

on static rotation may not be suitable for plain concrete subjected to biaxial fatigue
loading in the c-C-T region. Similar observations have been reported for the concrete
airport pavement in the biaxial t-C-T region (Subramaniam 1999).

0
0.3
0.6
0.9
1.2
0 0.03 0.06 0.09 0.12
Gage rotation (deg)
N
o
r
m
a
l
i
z
e
d

t
o
r
q
u
e
Static
90%-5%
85%-5%
80%-5%

Figure 6.9 A comparison of static and fatigue response: Torque - rotation

1
1.2
1.4
1.6
75 80 85 90 95
% post-peak torque
N
o
r
m
a
l
i
z
e
d

r
o
t
a
t
i
o
n
a
l

c
o
m
p
l
i
a
n
c
e
Fatigue
Static

Figure 6.10 A comparison of static and fatigue response: Torque compliance

A comparison of rotational compliance at ultimate failure in fatigue and at
different points in the post peak part of the static response is shown in Fig. 6.10. The
normalized rotational compliance in fatigue was calculated with respect to the rotational
compliance in the first cycle. The figure indicates that for rotation compliance, the static
response is comparable to the fatigue response. This suggests that static response acts like
the failure envelope for fatigue behavior when change in compliance is considered. Since
change in compliance is associated with crack propagation, the fatigue response indicates
that the crack length at failure may be obtained from the static response.

-4
-3
-2
-1
0.8 1 1.2 1.4
Normalized rotational compliance
L
o
g
(
C
/
N
)
90%-5%
Inflection point

Figure 6.11 Rotational compliance change rate versus rotational compliance

The rates of compliance change as a function of compliance for a typical
specimen as shown in Fig. 6.11. The rates of compliance follow a two-stage process; a
deceleration stage followed by an acceleration stage up to the failure. There is a distinct
inflection point marking a critical compliance where the rate of compliance changes from
deceleration to acceleration. This critical compliance corresponds to the compliance at
the peak static loading.


7. MODEL FOR FATIGUE FAILURE OF CONCRETE IN c-C-T REGION

7.1 Combined compressive and torsional test
Experiments conducted on concrete under combined compression and torsion discussed
in Chapter 6 showed that for a point in the biaxial c-C-T region (point-v), where the
principal compressive stress is 50% larger than the principal tensile stress, both the static
and fatigue failures are due to initiation and propagation of a single inclined crack.
Experimental evidence suggests that the increase in rotational compliance during the
compressive and torsional test is a result of this crack growth. This failure is very similar
to those observed in the t-C-T region when loading is pure torsion. Hence, it is important
to obtain information about the crack growth due to combined compressive and torsional
loading. The problem can then be studied by fracture mechanics, and a fracture-based
crack growth criterion can be established for such loading. The previously proposed
models and methods in the t-C-T region can then be extended and verified.

i

o
R
i
=2.25 in
R
o
=4 in

(a) (b)
Figure 7.1 Stress state at the inner, middle and outer radius of the cylindrical wall

Direct measurements of crack length are impossible in the concrete hollow
cylinders. The relationship between the crack length and rotational compliance of the
specimen can only be established by the finite element method (FEM) using the
principles of linear elastic fracture mechanics (LEFM). Detailed FEM simulation for
concrete hollow cylinders under pure torsion was carried out by Subramaniam (1999). He
fully discussed the crack front profiles, stress intensity factors (K
I
, K
II
, K
III
), and crack

propagation criterion, and found that the crack growth in concrete hollow cylinders under
pure torsion was due to the principal tensile stress and was governed by the Mode I stress
intensity factor (SIF), K
IC
.
K
I
=K
IC
(7.1)
Using FEM, he presented the relationship between the Mode I SIF, K
I
, crack
length, a, and applied torque as:
K
I
= (2.6296x10
-6
(a) + 7.2064x10
-6
) * Torque (7.2)
where K
I
is in N/mm
3/2
, Torque is in N-mm, and a is in mm.

Torque

C
i
C
u
K
IC

Rotation
Figure 7.2 Failure analysis using Mode I SIF in the static
combined compressive & torsional test

Eq. (7.2) was obtained from the pure torsional case, i.e. (Fig.
5.1b). For this loading case, the crack surface was observed to be planar along the
thickness direction. However, under the combined compressive and torsional loading
case, the crack surface is no longer a plane as observed in the case of pure torsional
loading, because of the influence of the compressive stress (
o
45 and 0 = =
0 ). The crack surface
formed a helically curved surface in space. However, the angles of inclination of the
helical crack along the thickness direction ( in Fig. 7.1b) calculated from the elastic
theory were found to be quite close among each other: 50.76
o
: 52.91
o
: 49.53
o
(1: 1.04:
0.98) for the case of c=-2/3. Even for c=-1/4, the ratio of the calculated angles of
inclination of the crack is not very large: 63.43
o
: 68.08
o
: 60.17
o
(1: 1.07: 0.95). So it can

be assumed that the crack surface is approximately a plane in the thickness direction. The
inclined crack in the specimen was shown in Fig. 6.3 at an angle of approximately 51
o

with respect to the horizontal (for c=-2/3).
Since the failure of hollow specimens under the combined compression and
torsion is quite similar to those under the pure torsion (both due to the crack propagation
by the principal tensile stress), and the ratio of the inclined angles along the hollow
cylinder thickness direction is not large, it is reasonable to extend the relationship of Eq.
(7.2) to the combined compressive and torsional loading case. However, due to a change
of direction and magnitude of the principal tensile stress, the applied torque, which
generates the principal tensile stress, should be modified by multiplying a factor, c (Eq.
5.5 & 5.6). For example, at pure torsion,
1
=, and at combined compression & torsion, if
c=-2/3,
1
= c * = 0.817. Then Eq. (7.2) can be modified as
K
I
= (2.63x10
-6
(a) + 7.21x10
-6
) * c * Torque (7.2a)
where K
I
is in N/mm
3/2
, Torque is in N-mm, and a is in mm.

1
1.2
1.4
1.6
0 0.5
Crack length (in)
N
o
r
m
a
l
i
z
e
d

c
o
m
p
l
i
a
n
c
e
Experiment
Fitting curve
y=3.35x
2
-1.80x+1
1
0
0.2
0.4
0.6
0.8
1 1.2 1.4 1.6
Normalized compliance
C
r
a
c
k

l
e
n
g
t
h

(
i
n
)
Experiment
Fitting curve
y=sqrt(-4.02x
2
+11.32x-7.30)
(a) (b)
Figure 7.3 Normalized torsional compliance vs. crack length in static test


In the previous research on the t-C-T region, evidence was presented to show that
both static and fatigue failure are due to Mode I crack propagation and governed by the
Mode I SIF, K
IC
. K
IC
= 38 N/mm
3/2
. Under combined compression and torsion, the failure
observed was very similar to failures in the t-C-T region (due to the crack propagation).
Thus, it is assumed that under static combined compression and torsion the post-peak
torque vs. rotation curve was also governed by the Mode I SIF, K
IC
(Fig. 7.2). The crack
length at any point in post-peak part of the response can be calculated from the increase
in unloading/reloading compliance. From Fig. 6.1 and 7.2 and Eq. (7.2), the relationship
between crack length and compliance can be established (Fig. 7.3).
Torsional compliance = 3.35*(a)
2
- 1.80* (a) + 1 (7.3a)
a = SQRT(-4.02* (torsional compliance)
2
+ 11.32*(torsional compliance) - 7.30)
(7.3b)
where a is in inches.

(a) (b)
0
0.2
0.4
0.6
0.8
1
0 0.3 0.6 0.9 1.2
Normalized cycles, N/N
f
C
r
a
c
k

l
e
n
g
t
h

(
i
n
)
90%-5%
Inflection point
-4
-3
-2
-1
0.2 0.4 0.6 0.8
Crack length (in)
L
o
g
(
a
/
N
)
90%-5%
a
inflection
Figure 7.4 Crack growth during fatigue compressive and torsional test
(a) crack length vs. cycles (b) rate of crack growth

In Fig. 6.11, it was shown that under compression and torsion the compliances are
comparable between static and fatigue tests. Therefore, Eq. (7.3) and Eq. (7.2) are also
applicable for fatigue loading. This implies that the fatigue failure is also governed by the
Mode I SIF. Thus, fracture mechanics can be employed to analyze the failure mechanism
of concrete under fatigue compression and torsion.
The fatigue crack length vs. number of cycles is shown in Fig. 7.4a, and the rate
of fatigue crack growth (a/N) for the same specimen is shown in Fig. 7.4b. It can be
seen that similar to the change rate of torsional compliance, the crack growth rate also
follows a two-stage process: a deceleration stage followed by an acceleration stage up to
failure. There is an inflection point (minimum) corresponding to a critical crack length,
a
critical
, where the rate of crack growth changes from deceleration to acceleration. This
critical crack length is estimated to be 0.44 inch based on Fig. 7.4b. On the other hand,
from Eq. (7.2) it can be calculated that the crack length corresponding to the static peak
load, a
peak
, is approximately 0.48 inch. Then
peak critical
a a . Taking crack length as the
indicator, the inflection point in fatigue test corresponds to the peak point in the static
test. The deceleration stage typically accounts for the first 45% the fatigue life of the
specimen.

0
0.2
0.4
0.6
0.8
1
75 80 85 90 95
Load range (%)
C
r
a
c
k

l
e
n
g
t
h

(
i
n
)
Fatigue
Static

Figure 7.5 A comparison of static and fatigue response: Torque crack length

In Fig. 6.9, it was shown that a deflection based failure criterion is not suitable for
fatigue behavior of concrete under combined compression and torsion. The static failure
deformation (gage rotation) at the post-peak period is not comparable with that of the
corresponding fatigue test. However, as shown in Fig. 6.10, under the torsional
compliance, the static post-peak behavior is favorably comparable to the corresponding
fatigue behavior. This suggests the possibility of using crack/fracture based failure
criterion for predicting fatigue failure using the results from the static test. A comparison
of crack length at fatigue failure to those at different unloading/reloading points in the
static post-peak period is shown in Fig. 7.5. It can be seen that the crack lengths at fatigue
failure compare favorably with the crack lengths at the corresponding load in the static
post-peak response of the specimen. This suggests that the crack length at fatigue failure
can be obtained from the static response. Hence the static response acts as a failure-
envelope to the fatigue response if framed in terms of crack lengths. Further more, Eq.
(7.1) (K
I
=K
IC
) represents the failure criterion for fatigue loading.
From the previous research in t-C-T region, it is proposed that the crack growth
rate at the deceleration stage is governed by the increasing resistance (R curve, Eq.7.4a),
and it is governed by the Mode I SIF (Paris law, Eq.7.4b) in the acceleration stage. The
concept of the R-curve is defined by Shah et. al. (1995). In the deceleration stage, the
crack growth rate can be expressed as:

1
) (
0 1
n
a a C
N
a
=
(7.4a)
or ) ( ) ( ) (
0 1 1
a a Log n C Log
N
a
Log + =

In acceleration stage, the crack growth rate can be expressed as:

2
) (
2
n
I
K C
N
a
=
(7.4b)
or ) ( ) ( ) (
2 2 I
K Log n C Log
N
a
Log + =

Where a
0
is the initial crack length (approximately 2 mm). Log(C
1
), n
1
, Log(C
2
) and n
2

are constants. These parameters were calibrated from the previous study in the t-C-T
region (Table 7.1). The units of crack length, a, and SIF, K
I
, are mm and N/mm
3/2
,
respectively.

Table 7.1 Parameters in the models
Log(C
1
) n
1
Log(C
2
) n
2

t-C-T region -0.65~-2.05 -1.25~-1.7 -17.5~-42.5 9~24
c-C-T region (Tensile failure) 59.6 -68.8 -28.6 17.33
c-C-T region (Compressive failure) -10.25 -2.02 -6.46 21.33

A comparison of the experimental data from the combined compressive and
torsional tests and the model predictions from the deceleration and acceleration stages is
shown in Fig. 7.6. In the deceleration stage (Eq. 7.4a), the values of Log(C
1
) and n
1
were
taken to be 59.6 and 68.8, respectively. The Paris law constants (Eq. 7.4b), Log(C
2
) and
n
2
, were taken to be 28.6 and 17.33, respectively (the average value from the previous
research, Table 7.1). It can be seen that there is a reasonable match between the
experimental data and model predictions. So the previously proposed model can be
successfully extended to the c-C-T region. However, the parameters in the deceleration
stage are different from those calibrated from the flexural response of concrete. The
fatigue crack rate growth at the acceleration stage can be predicted using the uniaxial
material parameters.

(a) (b) (c)
Figure 7.6 A comparison between analytical prediction and experiment:
loading equal to (a) 90%-5% (b) 85%-5% and (c) 80%-5%

-6
-4
-2
0
2
5 10 15 20 25
Crack length (mm)
L
o
g
(
a
/
N
)
Deceleration
Accelaration
-6
-4
-2
0
2
5 10 15 20 25
Crack length (mm)
L
o
g
(
a
/
N
)
Deceleration
Aceleration
-6
-4
-2
0
2
5 10 15 20 25
Crack length (mm)
L
o
g
(
a
/
N
)
Deceleration
Acceleration
7.2 Pure compressive test
From the band-damage zone model (Fig. 4.1), it can be seen that the post-peak inelastic
displacements in the damage zone of the static test are comparable with those of the
fatigue test (Fig. 4.2). This suggests that the fatigue failure mechanism is the same as that
of the static failure mechanism. As mentioned by Shah and Sankar (1987), failure
localization initiates just prior to the peak stress. This phenomenon is clearly observed in
the experiment. For example, in the static compressive test, the pre-peak load-
displacement curve becomes nonlinear at approximately 70% of the peak load (Fig. 3.4a).
This means that the inelastic displacement of the specimen starts prior to the peak load.
However, since it is found that the total displacement for fatigue is not comparable with
the static test, it follows that the pre-peak inelastic displacements are also not comparable
between static and fatigue tests. This further implies that the deformation mechanisms are
different between the two tests in the pre-peak period. On the other hand, the pre-peak
compliances are comparable between the two tests. This provides a way to define the pre-
peak inelastic displacements for both static and fatigue test, based on compliance.

d
F
C
r
C
peak
C
i

Displacement
Compression

0

Figure 7.7 Definition of inelastic displacements
It is assumed this inelastic displacement starts at 70% of peak load, in both static
and fatigue tests. Prior to this load level, the specimen is elastic and the normalized
compliance (C
r
/C
i
) is unity. In the static test, the total inelastic displacement,
*
, consists
of two parts: pre-peak inelastic displacement,
0
, and post-peak inelastic displacement, :

(7.5) + =
0
where is distributed with in the band damage zone only, and
0
is distributed along the
whole length of the specimen including the band damage zone. The definition of the pre-
peak inelastic displacement,
0
, for the static test is illustrated in Fig. 7.7. It is equal to the
total displacement less the elastic displacement:

0
= d F*C
r
, (7.5a)
This pre-peak inelastic displacement,
0
, leads to the inelastic strain part in
b
(Eq. 4.1).
For the fatigue test, it is assumed that the inelastic displacement also consists two
parts: pre-peak inelastic displacement,
0
, and post-peak inelastic displacement, , (same
as that of static test).
(7.5b) + =
0
where
0
is a pre-peak inelastic displacement. The relationship between the static pre-
peak inelastic displacement,
0
, and fatigue pre-peak inelastic displacement
0
is
established through the comparable pre-peak compliances of the two tests.

y = 0.22x
3
- 0.92x
2
+ 1.26x - 0.57
0
0.004
0.008
0.012
0.016
0.02
0.024
1 1.2 1.4 1.6 1.8 2
Normalized compliance, C
r
/C
i
I
n
e
l
a
s
t
i
c

d
i
s
p
l
a
c
e
m
e
n
t
,

,

(
i
n
)
Experiment

Figure 7.8 Inelastic displacement vs. compliance in static compressive test
In pure compression, the cracks are distributed, unlike in the t-C-T region, where
individual cracks can be identified and measured, so it is not possible to verify the

previously proposed models in the t-C-T region, which are based on crack information.
To verify/extend the previous models for the pure compression point, this inelastic
displacement is used analogously to the crack length in the previous model.

0
0.004
0.008
0.012
0.016
0.02
0 0.3 0.6 0.9 1.2
Normalized cycles (N/N
f
)
I
n
e
l
a
s
t
i
c

d
i
s
p
l
a
c
e
m
e
n
t
,

,

(
i
n
)
76%-5%
Inflection point
-7
-6
-5
-4
0 0.005 0.01 0.015 0.02
Inelastic displacement,
, (in)
L
o
g
(
N
)
76%-5%
Inflection point
(a) (b)
Figure 7.9 Increment of inelastic displacement during fatigue compressive test
(a) inelastic displacement vs. cycles (b) rate of inelastic displacement increment

Fig. 7.8 shows the relationship between the compressive compliance and inelastic
displacement in the static compressive test. The fatigue inelastic displacement vs. number
of cycles is shown in Fig. 7.9a, and the rate of fatigue inelastic displacement increment
(
*
/N) for the same specimen is shown in Fig. 7.9b. (In Fig. 7.9, nominal pre-peak
inelastic displacement is used). Similar to the previous results of crack length vs. number
of cycles in t-C-T region and point-v in c-C-T region (Fig. 7.4a), a S-shape curve is found
in Fig. 7.9a. The inelastic displacement change rate also follows a two-stage process: a
deceleration stage followed by an acceleration stage up to failure. There is an inflection
point (minimum) corresponding to a critical nominated inelastic displacement, where the
rate of inelastic displacement increment changes from deceleration to acceleration. This
critical inelastic displacement corresponds to the critical compliance in Figs. 3.5 and 3.6.

This means that when analyzed in terms of inelastic displacement, the inflection point in
the fatigue test corresponds to the peak point in the static test. The deceleration stage
typically accounts for the first 40% of the fatigue life of the specimen for pure
compression loading case.
In Fig. 4.2, it was shown that in terms of the post-peak inelastic displacement, ,
the static post-peak behavior is similar to the corresponding fatigue behavior. This
suggests the possibility of using an inelastic displacement-based failure criterion for
predicting fatigue failure using the results from the static test. From the previous research
in t-C-T region and current research at loading point-v (combined compression and
torsion) in the c-C-T region, it has been seen that the crack growth rate in the deceleration
stage is governed by the increasing resistance (R curve), and it is governed by the Mode I
SIF (Paris law) in the acceleration stage (Eqs. 7.4a and 7.4b). Here, it is reasonable to
assume that for pure compressive loading the change rate of inelastic displacement in the
deceleration stage is governed by the increasing compressive resistance, which is a
function of the increment of the pre-peak inelastic displacement, and it is governed by the
compliance in the acceleration stage (post-peak inelastic displacement). Based on this
assumption the previously proposed model, Eqs. 7.4a and 7.4b, can be modified. In the
deceleration stage, the inelastic displacement change rate can be expressed as:

1
) (
1
n
C
N
(7.6a)
or ) ( ) ( ) (
1 1
+ =
Log n C Log
N
Log
In acceleration stage, the crack growth rate can be expressed as:

2
) (
2
n
i
r
C
C
C
N
=
(7.6b)
or ) ( ) ( ) (
2 2
i
r
C
C
Log n C Log
N
Log + =

Where * is the inelastic displacement in inches, and C
r
and C
i
are the reloading and
initial compliance, respectively.
A comparison of the experimental data from the pure compressive tests and the
model predictions from the deceleration and acceleration stages is shown in Fig. 7.10. In

the deceleration stage (Eq. 7.6a), the values of Log(C
1
) and n
1
were taken to be 10.25
and 2.02, respectively. In the acceleration stage (Eq. 7.6b), Log(C
2
) and n
2
, were taken
to be 6.46 and 21.33, respectively (Table 7.1). A reasonable match between the
experimental data and model predictions is found. So by substituting crack length with
the inelastic displacement the previously proposed model can be extended to the pure
compressive loading point in the c-C-T region.


-8
-6
-4
-2
0
0 0.005 0.01 0.015 0.02
, (in)
L
o
g
(
N
)
Deceleration
Acceleration

(a)
-8
-6
-4
-2
0
0 0.005 0.01 0.015 0.02
Inelastic displacement, *, (in)
L
o
g
(
*
/
N
)
Deceleration
Acceleration

(b)

-8
-6
-4
-2
0
0 0.005 0.01 0.015 0.02
Inelastic displacement, *, (in)
L
o
g
(
*
/
N
)
Deceleration
Acceleration

(c)
Figure 7.10 A comparison between analytical prediction and experiment in
compressive loading: (a) 90%-5% (b) 80%-5% and (c) 76%-5%

7.3 Transition point between tensile failure mode and compressive failure mode
As can be seen in the c-C-T region (Fig. 2.2), there are two typical failure modes caused
by two typical loading cases. One is the tensile failure mode similar to that observed in
the t-C-T region, which is dominated by crack initiation and propagation, as represented
by point-v (torsion & compression). The other is the compressive failure mode, which is
dominated by inelastic displacement, as represented by point-iv (pure compression). In
the case of the tensile failure mode, a crack based fracture mechanics criterion is
employed to explain the failure mechanism. In the case of compressive failure mode, an
inelastic displacement based criterion is employed.

0 0
2 1
< =
c-C-T region t-C-T region
2

| | | |
2 1
>
iii
| | | |
2 1
=
ii
iv
v
| | | |
2 1
<
Transition point
15
1
2
1
=
Inelastic displacement
dominated compressive
failure
Crack dominated tensile failure
i
0 0
1 2
> =

Figure 7.10 Transition point from tensile failure to compressive failure
Kupfer et. al. (1969), Kupfer and Gerstle (1973), and Nelissen (1972)
systematically investigated the performance of concrete subjected to biaxial stresses. In
their study concrete plates were subjected to in-plane loading for different biaxial stress
combinations spanning the entire biaxial stress space. The mode of failure was found to
be different in the biaxial stress space. It has been found that in the c-C-T region, the
failure is similar to that in compression as long as the principal tensile stress is less that
15
1
of the principal compressive stress. For larger principal tensile stress concrete fails by
a single crack forming perpendicular to the maximum principal tensile stress. Their
conclusions support the current research results. Hence it is reasonable to postulate that
the transition point between tensile failure mode and compressive failure mode is

determined by the ratio of the principal tensile to compressive stresses (Fig. 7.10), as
expressed as

15
1
=

stress e compressiv principal
stress tensile principal
(7.7)
So, a larger principal tensile to compressive stress ratio (>
15
1
) leads to the tensile
failure mode where failure is governed by the crack based fracture mechanics, while a
smaller principal tensile to compressive stress ratio (<
15
1
) leads to the compressive
failure mode where failure is governed by the inelastic displacement.

Load Load
A
Fatigue
Static envelope governed by K
I
=K
IC
0
/
0
'

70% peak load
critical

C
B
A
Fatigue inelastic displacement growth
*
Static envelope
Fatigue
a
critical

C
B
Fatigue crack growth
Crack length

No. of cycles
No. of cycles
(a) (b)
Figure 7.11 Schematic representation of (a) crack growth, and (b) inelastic
displacement growth in static and fatigue loading
In the tensile failure mode, the crack growth in the static and fatigue loading can
be schematically represented by Fig.11a, and in compressive failure mode, the inelastic
displacement growth in the static and fatigue loading can be schematically represented by
Fig.11b.


8. CONCLUSIONS

The objective of this work is to characterize the static and fatigue response of concrete
subjected to biaxial stresses in the c-C-T region, where the principal compressive stress is
larger in magnitude than the principal tensile stress. By extending the previous methods
and results to the biaxial compression region, a full description of fatigue in the C-T
region was obtained.
Previous research on high amplitude fatigue response of concrete subjected to
biaxial stresses in t-C-T region, where the principal tensile stress is larger in magnitude
than the principal compressive stress, suggested the following: (a) the change rate of
structural compliance and crack length in constant amplitude fatigue loading is a two-
phase process: a deceleration phase followed by an acceleration stage; (b) a static load
envelope was shown to predict the crack length or structural stiffness/compliance at
fatigue failure. The primary mode of failure in this biaxial stress region was shown to be
crack propagation. The fatigue crack growth models were developed using fracture-based
parameters.
An experimental investigation of material behavior in the biaxial c-C-T region
was conducted. Two typical loading cases were selected to represent the biaxial c-C-T
loading cases: compression and combined compression and torsion. The experimental
setup consisted of two test configurations: concrete cylinders subjected to pure
compression, and hollow concrete cylinders subjected to torsion with a superimposed
axial compressive force. The damage imparted to the material was measured by
mechanical means.
In the investigation, the static load response can be visualized as a failure
envelope curve, where each point in the post-peak region is an equilibrium point
representing the maximum load that can be supported for a given level of damage.
Therefore every point on the post-peak load envelope can be characterized by a given
damage level. Further, it can be implicitly assumed that the change in compliance of a
specimen is due to accruing damage in the specimen and the increase in compliance is

indicative of the increase in level of damage. In both tests, the favorable comparison
between the percentage increase/decrease in compliance, or crack length (if applicable),
or inelastic displacement (if applicable) in fatigue and static post-peak load indicates that
the damage level is comparable for the two loadings. Some specific conclusions of this
project are listed below:
The previously proposed methods, theories and models in the t-C-T region are
applicable in the whole C-T region if a suitable length/displacement is selected to
describe the failure mode. For tensile failure mode, the crack length is used, and for
compressive failure mode, the inelastic displacement is used.
A deflection based fatigue failure criterion is not suitable for concrete subjected to
biaxial C-T loading because the displacement at fatigue failure is not comparable to
the static post-peak displacement at the corresponding load.
In the biaxial C-T region, the static response acts like an envelope to constant-
amplitude, low-cycle fatigue response when framed in terms of compliance, or
crack length, or inelastic displacement.
The change rate of compliance, or crack length, or inelastic displacement is a two-
phase process in the C-T region: a deceleration stage followed by an acceleration
stage up to failure.
The damage evolution in C-T region, in term of the measured stiffness, follows a
three-stage trend. Fatigue life can be predicted by the slope of the stiffness change
rate in stage II, and the relationship between the slope of stage II and fatigue life is
independent of the fatigue load range.


REFERENCES

Gerstle, K. H., (1981). Simple Formulation of Biaxial Concrete Behavior, ACI
Materials Journal, 78, 62-68.
Jansen, D. C. and Shah, S. P., (1997). Effect of Length on Compressive Strain Softening
of Concrete, Journal of Engineering Mechanics, ASCE, 123(1), 25-35.
Jansen, D. C., (1996). Postpeak Properties of High Strength Concrete Cylinders in
Compression and Reinforced Beams in Shear, Ph.D thesis, Northwestern
University.
Jansen, D. C. and Shah, S. P., (1993). Stable Feedback Signals for Obtaining Full Stress
Strain Curves of High Strength Concrete, Proceedings Utilization of High
Strength Concrete, V.2, 20-24, Lillehammer, Norway.
Okubo, S. and Nishimatsu, Y. (1985). Uniaxial Compression Testing Using a Linear
Combination of Stress and Strain as the Control Variable, International Journal
of Rock Mech. Min. Sci. & Geomech. Abstr. 22(5), 323-330.
Rokugo, K., Ohno, S., and Koyanagi, W., (1986). Automatic Measuring System of
Load-Displacement Curves Including Post Failure Region of Concrete
Specimens, In Fracture Toughness and Fracture Energy of Concrete, Edited by
F. H. Wittman, Elsevier Applied Science.
Kupfer, H. B., Hilsdorf, H., and Rusch, (1969). Behavior of Concrete Under Biaxial
Stresses, ACI Materials Journal, 66(8), 656-666.
Kupfer, H. B., and Gerstle, K. H., (1973). Behavior of Concrete under Biaxial Stresses,
Journal of Engineering Mechanics, ASCE, 99(4), 853-866.
Nelissen, L. M. J., (1972). Biaxial testing of normal concrete, Heron (Delft), 18(1).
Shah, S. P., Swartz, S. E., and Ouyang, C., (1995). Fracture Mechanics of Concrete:
Applications of Fracture Mechanics to Concrete, Rock and Other Quasi-Brittle
Materials, Wiley, New York.
Shah, S. P. and Sankar, R., (1987). Interanl cracking and strain-softening response of
concrete under uniaxial compression, ACI Materials Journal, 84(3), 200-212.

Subramaniam, K. V., Popovics, J.S., and Shah, S.P., (2002). Fatigue Fracture of
Concrete Subjected to Biaxial Stresses in the Tensile CT Region, Journal of
Engineering Mechanics, ASCE, 128(6), 668-676.
Subramaniam, K. V., ONeil, E., Popovics, J.S., and Shah, S.P., (2000). Flexural Fatigue
of Concrete: Experiments and Theoretical Model, Journal of Engineering
Mechanics, ASCE, 126(9), 891-898.
Subramaniam, K. V., (1999). Fatigue of Concrete Subjected to Biaxial Loading in the
Tension Region, PhD dissertation, Northwestern University, Evanston, IL.
Subramaniam, K. V., Popovics, J.S., and Shah, S.P., (1999). Fatigue Behavior of
Concrete subjected to Biaxial Stresses in the C-T Region, ACI Materials Journal,
96(6), 663-669.
Subramaniam, K. V., Popovics, J.S., and Shah, S.P., (1998). Testing Concrete in
Torsion: Instability Analysis and Experiments, Journal of Engineering
Mechanics, ASCE, 124(11), 1258-1268.
Torrenti, J. M., Benaija, E. H. and Boulay, C., (1993). Influence of boundary conditions
on strain softening in concrete compression test, Journal of Engineering
Mechanics, ASCE, 119(12), 2369-2384.


APPENDIX I EXPERIMENTAL DATA Compressive test

Table A-1 Test data at the peak loads in static compressive test
Specimen
Peak load (kip) Axial displacement (in) Normalized compliance (C
r
/C
i
)
1 76.06 0.035 1.13
2 75.29 0.033 1.04
3 75.83 0.034 1.09
Average 75.73 0.034 1.05

Table A-2 Post-peak normalized compliance (C
r
/C
i
) and axial displacement in static
compressive test

Specimen
Pre-peak
C
i
/C
i
(Disp,
in)
90% post-peak
load
C
r
/C
i
(Disp, in)
80% post-peak
load
C
r
/C
i
(Disp, in)
70% post-peak
load
C
r
/C
i
(Disp, in)
1 1.0 (0.026) 1.21 (0.038) 1.43 (0.040) 1.56 (0.041)
2 1.0 (0.026) 1.12 (0.038) 1.39 (0.040) 1.59 (0.043)
3 1.0 (0.027) 1.18 (0.040) 1.43 (0.043) 1.56 (0.045)
Average 1.0 (0.026) 1.16 (0.038) 1.41 (0.041) 1.56 (0.043)
Note: i) The axial displacement at the initial compliance stage was measured at the
starting point of the nonlinear load-displacement curve in pre-peak load
period.
ii) The axial displacement at x% post-peak load was measured at the starting
unloading point in the load-displacement curve.


Table A-3 Results from fatigue tests for the load cycle of 90% ~ 5% of the average static
compressive strength
At fatigue failure At inflection
Specimen

Cycles to
failure
Axial disp.
(in)
Normalized
compliance
Normalized
compliance
1 3621 0.038 1.27 1.03
2 338 0.034 1.28 1.03
3 66 0.031 1.21 1.02
Average 1342 0.034 1.25 1.03

Table A-4 Results from fatigue tests for the load cycle of 80% ~ 5% of the average static
Specimen

Cycles to
failure
Axial disp.
(in)
Normalized
compliance
Normalized
compliance
1 1966 0.035 1.32 1.05
2 1390 0.032 1.24 1.06
3 3984 0.038 1.35 1.10
Average 2447 0.035 1.30 1.07

Table A-5 Results from fatigue tests for the load cycle of 76~ 5% of the average static
Specimen

Cycles to
failure
Axial disp.
(in)
Normalized
compliance
Normalized
compliance
1 15256 0.035 1.39 1.04
2 15906 0.035 1.72 1.04
3 18656 0.037 1.54 1.03
4 11174 0.033 1.43 1.04
Average 15248 0.035 1.52 1.04


APPENDIX II EXPERIMENTAL DATA Combined compressive & torsional test

Table AA-1 Test data at the peak torque static combined compressive & torsional test
Specimen
Peak torque (ft-lb) Gage rotation (deg)
1 838 0.0371
2 829 0.0411
3 835 0.0408
Average 834 0.0397

Table AA-2 Cycles for fatigue combined compressive & torsional test

Cycles to failure

Cycles to failure

Specimen
(90%-5%) (85%-5%)

Cycles to failure
(80%-5%)
1 1458 14162 90642
2 128 46074 -
3 274 3370 -
4 2134 - -
Average 999 21202 90642


APPENDIX III Publications from This Project

1. Bin Mu, Kolluru V. Subramaniam and S. P. Shah, (2003). Failure Mechanism of
Concrete under Fatigue Compressive Load, Journal of Materials in Civil
Engineering, submitted (MT/2003/022651).

2. Bin Mu and S. P. Shah, (2003). Failure Mechanism of Concrete Under Biaxial
Fatigue Load, Fifth International Conference on Fracture Mechanics of Concrete
and Concrete Structures, April 12 16, 2004, Vail Colorado, accepted.

3. Bin Mu and S. P. Shah, (2003). Fatigue Behavior of Concrete Subjected to
Biaxial Loading in the Compression Region, Materials and Structures,
submitted.

Fatigue Strength of Concrete

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Fatigue Strength of Concrete

Diunggah oleh

Hak Cipta:

Format Tersedia

COE Report No.

Anda mungkin juga menyukai