You are on page 1of 8

New Test Method for Obtaining Softening Response of

Unnotched Concrete Specimen Under Uniaxial Tension

by Z. Li, S.M. Kulkarni and S.P. Shah

ABSTRACT--A new method was developed to obtain the men. This is mainly due to control difficulties in direct
complete stress-deformation response of an unnotched uniaxial tension test.
concrete specimen. This method employs a digitally con- To obtain a stable response in the post-peak region for a
trolled closed-loop testing machine and five control chan- uniaxial tension test, one can use an elastic load-sharing
nels: stroke LVDT and four LVDTs mounted on the
specimen. The test portion of the specimen was fully system parallel to the specimen (in the loading direction).
spanned by these four LVDTs. The outputs of these control In this way, unstable failure near the peak load (resulting
channels and the load cell were monitored during the test from sudden release of stored elastic energy) is avoided.
by a computer equipped with LabTech's Notebook soft- Such a test was conducted by Evans and Marathe, l who
ware. A 'C' language program was written to enable quick obtained the post-peak response for plain concrete by
switching of the mode of control from one LVDT to another. pulling the specimen together with a steel frame. Peters-
The problem of uncertainty in the location of the major crack 2 . . . . . . .
son carned out a slmdar experiment using surf aluminum
was tackled by the said LVDT arrangement and the com- columns. In his test, however, the load was indirectly
puter program. It was demonstrated that it is always possi- applied to the concrete specimen by heating the aluminum
ble to obtain stable post-peak response provided one
columns. Schorn and Berger-Bocker3 followed Evans and
ensures that at any time during the test, the feedback used
is the LVDT that exhibits, at that time, the largest slope of Marathe's idea and conducted a uniaxial tensile test with
the response-time curve. An acoustic emission (AE) meas- the help of steel bars parallel to the specimen to get the post
urement system of six channels was also used in the peak response. A major drawback of the method of using
experiments. Monitoring of signals from the AE transducers a parallel load-sharing system is that accuracy of the results
provided valuable information which helped in making the obtained is limited. Load shared by concrete is obtained by
decision to switch control. subtracting the contribution of the load-sharing system
from the total load, implying subtraction of two relatively
Introduction large numbers to obtain a small number. It may be noted
here that significant difference exists in the test results
Compared with its compressive behavior, concrete's reported by Evans and Marathe, and Petersson.
tensile behavior has received little attention in the past; Noting the disadvantages of the parallel load-sharing
partly because it is common practice to ignore tensile method, attempts were made to perform tensile tests using
resistance of concrete in reinforced concrete design. Inter- closed-loop control. Tensile failure of concrete is the result
est in tensile properties has grown substantially in recent of opening of a single 'major' crack. If the opening of the
years. Some of the important reasons for this are: introduc- crack is controlled in a closed-loop manner, gradual failure
tion of fracture mechanics in the field of concrete struc- can be obtained and instability near the peak load can be
tures, advent of powerful computers, and increasing avoided. However, since the location of the major crack
availability of useful numerical techniques. However, cannot be known a priori, choice of an appropriate control
available information on the complete load-deformation channel is difficult if an unnotched specimen is used. A
response of plain concrete and fiber-reinforced concrete pair of displacement sensors (one on each side of the
under tension is scarce, especially for an unnotched speci- specimen) having a large gage length (perhaps covering
the entire specimen) can be used and a control signal can
be obtained by averaging the signals of the two sensors.
Z. Li and S.M. Kulkarni are Research Assistants, and S.P~ Shah (SEM
Member) is Walter P. Murphy Professor of Civil Engineering and Direc- However, such a signal may lead to unstable failure near
tor of NSF Science and Technology Center for Advanced Cement-Based the peak load as described later in this paper. Essentially,
Materials, Northwestern University, Evanston, IL 60208. this signal consists of two parts: (1) a monotonically in-
Original manuscript submitted: June 29, 1992. Final manuscript re- creasing part corresponding to the opening of the major
ceived: February 25, 1993. crack and (2) another part which increases in the prepeak

Experimental Mechanics 9 181

region, but decreases near the peak load. Often, dominance chosen for displacement sensors (LVDTs) was such that
of the second part forces the signal to decrease during the the entire test-portion of the specimen was spanned by four
test, leading to unstable failure. This has been explained in LVDTs. (2) Readings of all four LVDTs were monitored
some more detail later in this paper. If the gage length is and a C language program, developed specially for the
reduced significantly, the probability of the major crack purpose of this test, was used to enable rapid switching of
forming within the span of the two sensors reduces. If the control among the control channels.
major crack forms outside the span of the sensors, the The experiment proceeded on an anticipatory basis.
specimen will fail in an uncontrolled manner near the peak From the monitored LVDT readings and acoustic event
load. Consequently, such an attempt has a discouragingly data (described later), a continuous judgement of location
small success rate. of the region of maximum damage (potentially the region
If the location of the major crack is predetermined by where the major crack may develop) was made, resulting
introducing a notch in the specimen, opening of the notch in the choice of a control channel. If the selected choice
can be used as a feedback signal. Gopalaratnam and Shah 4 was incorrect, that would be indicated by subsequent
conducted such a test by introducing a double edge-notch LVDT response as well as the acoustic-activity signals
in the test specimen. Their feedback signal was notch- acquired. In such a ease, another choice was made. It was
opening displacement averaged for two sensors on the two found that the time required for the judging and the switch-
sides of the specimen. Following Gopalaratnam and ing of the control mode was short enough (switching time
Shah's work, several researchers used notched specimens was about one second) to permit a stable post-peak re-
to obtain the comsPlete tensile response. For instance, Li, sponse for all the specimens tested.
Chan and Leaung conducted a test by using compact-ten- Available application programs such as TestWare SX
sion specimen to obtain tension-softening response of do not currently provide the facility of such quick and
mortar. Hordijk and Reinhardt 6 carried out a tension test arbitrary switching among control modes when a test is in
using a single-notched specimen. Cho, E1-Shakra and progress. Hence development of a new program was nec-
Gopalaratnam7 used the notched-specimen technique to essary. This program starts the test in stroke control by
study the toughening and failure mechanisms of fiber-re- applying a ramp function. At any point on this ramp,
inforced concrete. One of the problems associated with a depending on the user's judgement, it is possible to switch
notched specimen is that it forces the major crack to form control to any one of the four LVDTs by simply pressing
at a predetermined location. Also, the state of stress in the one of the 'Function keys' on the load unit control (LUC
specimen is not truly uniform. When a notched specimen Pod). After this, the user can continue the process of
is used, it is not possible to study accumulation of damage switching control among the four LVDTs. Each time
and strain localization--phenomena which are of great switching is desired, one needs to press a designated
interest in the failure process. function key on the LUC Pod.
In the work reported here the above problems were The program was complied by using Microsoft C 6.0 in
solved. A new test method to obtain the response of a protect mode. After compilation, the executable file was
concrete specimen under uniaxial tension (without either retrieved into TestStar through the System Administrator
a notch or a parallel load-sharing system) was developed. program.

Details of Experiments Specimen Preparation

Three types of specimens--plain concrete, steel-fiber-
Closed-loop Digitally Controlled Testing reinforced concrete (SFRC) and polypropylene-fiber-rein-
Machine forced concrete (PFRC) were prepared. The plain concrete
specimens were prepared using a mix proportion of 1 : 0.65
Tests reported in this paper were conducted using a
: 2.6 : 2,6 (cement: water : sand : coarse aggregate) by
closed-loop digitally controlled testing system at North-
weight. The SFRC specimens were prepared using a pro-
western University. This system, marketed by MTS Sys-
portion of 1 : 0.6 : 2.45 : 2.45 (cement: water: sand : coarse
tems Corp., consists of a loading frame, a digital controller,
aggregate) by weight. 0.5 percent by volume of hooked
load unit control panel (LUC Pod), and a personal com-
steel fibers manufactured and provided by Bekaert Corp.
puter. Any test performed using the machine is controlled
(designation ZC 50/0.50) were added into the mixture. In
by the digital controller using the instructions of a com-
addition, superplasticizer (1.7 ml per pound of cement)
puter program (vendor-provided or user-written). MTS
was added to achieve reasonable workability. The PFRC
Corp. provides a basic control software called TestStar.
specimens were prepared with a mix proportion of 1 : 0.56
Some 'high-level' applications such as the TestWare SX
: 2.4:2.4 (cement: water: sand: coarse aggregate) by
program are also available which simplify the process of
weight. 0.5 percent by volume of fibrillated polypropylene
'programming' a test. A 790 C programming interface is
fibers (length = 50 mm) made by Fibermesh company were
also provided to allow users to write involved programs
added into the mixture. The mix proportions for fiber-re-
for unconventional tests.
inforced-concrete specimens were selected so as to match
a round-robin test program on fiber-reinforced concrete.8
C Language Program for Uniaxial Tension Test All the specimens were demolded 24 hours after casting
and left in a water bath for curing. The specimens were
In order to resolve the aforementioned difficulties in tested at the age of 28 days. Two days before testing, the
control, two major steps were taken: (1) The arrangement specimens were taken out of the bath and left to dry. Steel

182 9 September1993
330 turn

/ Spec~imen //~T -2'mm

Aligning pins I , , '!/'/~-~7 m~
\~,~ , AIi,qnfq,q ~ips e~ I"r .........
~~Loading end tabs~ . ~ _ _ _
/<--the bottom ~ "~-~6 5 mm

390 mm Base aluminunplate

Fig. lmProcedure for glueing steel end tabs to the

]~ ~ ~ SteelPlatefor-
I L~adTransfer
loading plates were then glued to the specimens. Before
glueing the loading plates, both the specimen surface and
the steel plates were polished with #400 sand paper and
cleaned with acetone. As shown in Fig. 1, to ensure good
alignment of the loading holes, two steel plates were first
fixed on the base aluminum plate through four aligning
pins. Well-mixed epoxy was then spread on the plates. The
specimen was placed on the plates and aligned carefully so
that the center line of the bottom plates and that of the
specimen coincided. Epoxy was spread on the exposed
Fig. 2--Test setup
(top) surface of the specimen, Next, the other two loading
plates were slid down on the top surface of the specimen
through the aligning pins. After squeezing out the air
bubbles, a 20-1b weight was placed on each of the top steel AE Data-acquisition System
plates to keep good contact during the epoxy hardening
process. The entire specimen size was 330 127 28 mm. Data-acquisition system shown in Fig. 3 was used to
After gluing the loading plates, the test portion left was 127 acquire AE data. The system consisted of transducers,
127 28 ram. preamplifiers and A-D modules. Six channels were em-
ployed. The transducers were purchased from Physical
Acoustics Corporation (model number micro 80). These
general-purpose transducers are made of piezoelectric
Test Setup
crystal and are hence called PZT transducers. They are
convenient to use because of their small size (9.5-ram
The setup for the test is shown in Fig. 2. Two identical diameter and I l-ram height), negligible weight (5 g) and
loading fixtures were used: one mounted on the machine's desirable sensitivity. These transducers have an essentially
actuator and the other connected to the load cell. In each flat amplitude response of approximately _+9 dB in the
fixture, a sliding steel block was provided which helped frequency range of 100 kHz to 1.2 MHz. The preamplifiers
center the specimen. This kind of setup provides boundary used were also purchased from Physical Acoustics Corpo-
conditions close to those provided by nonrotatable loading ration (model number 1220B). These preamplifiers have a
platens. 9 The thickness of the loading plates was gradually bandwidth of 20-1200 kHz and a gain of 40 or 60 dB
reduced (Fig. 1) from 9.5 mm to 1.6 mm to avoid a drastic (interchangeable). In the system, channel A was used as a
change in stiffness and consequent stress concentration. trigger channel and was connected to a 6103 amplifier
Four LVDTs with a working range of 0.635 mm and gage trigger module. The other five channels were used as
length of 69.85 mm were mounted on the specimen (Fig. working channels and were connected to five TR8837F
2). The signals from the LVDTs (including the stroke digitizer modules which have a digitizing rate of 16 MHz.
LVDT) and the load cell were acquired using LabTech A 6128 fan-out module was used to receive the signal from
Notebook software. The readings of all six sensors were the 6103 amplifier and to trigger all working channels
displayed in six individual windows (sensor reading as a simultaneously. A LeCroy 8901A interface board was
function of time) on the Notebook screen. These graphs used to communicate with an IBM-AT via a GPIB cable.
were useful in making decisions regarding switching of the The operation of the LeCroy system was controlled by
control mode. Six PZT transducers were also glued (Fig. Physical Acoustics's CATALYST program which also
2) on the surface of the specimen to monitor the acoustic stored the digitized data on the computer's hard disk. 3000
emission (AE) activity. data points were stored for each channel.

ExperimentalMechanics~ 183
Test Results

Five plain concrete specimens, three PFRC specimens

-4 ~,m I
--I TrJc~t~er I
n| ,,,,i.
and three SFRC specimens were tested. Each test started
I E;128 I under stroke control with a ramp rate of 0.0156 mm of
I Fan-out I
_-r~. piston displacement per minute. At the beginning of the
"3 TR8837F |I loading [region A in Figs. 4(a) and 5(a)] the specimen
-J Di#i:'er

behaved elastically as expected and the outputs of the four
LVDTs increased linearly with time. Corresponding load-
deformation curves are shown in Figs. 4(b) and 5(b). The
readings of the LVDTs were almost identical in this region,
indicating uniformity of deformation throughout the speci-
men. As loading continued, (that is, at about 1/3 to 1/2 of
LeCroyData the peak load for plain concrete and 3/5 to 3/4 of the peak
I kquisition
System load for fiber-reinforced concrete) some AE events were
Data typically registered. Also, the readings of the LVDTs
started to exhibit some difference. The control mode was
then switched to the LVDT which exhibited the largest
Fig. 3--AE data-acquisition setup
time-rate of output. As the load increased further, more

S t e e l Fiber S p e c i m e n SFRC-F21 S t e e l f i b e r s p e c i m e n (SFRC-F21)

(0.5 p e r c e n t f i b e r by v o l u m e ) 20 ( 0 . 5 p e r c e n L by v o l u m e )


0.055 LVDT-3 LVDT-



I .~10

. A- r LVDT-4

A: Stroke Control
B: LYDT-2, 4 or 1 Control
C: LVDT- 1 Control
-0.025 0~ ~-~
2000 4000 6000 8000 -0.0064 0.0037 0.0137 0.0237
Time (second) Displacement (ram)

Fig. 4--(a) Deformation-time curves for SFRC specimen SFRC-F21 ; b) Load-deformation curves for
SFRC specimen SFRC-F21

Deformation-time curves for concrete Load d i s p l a c e m e n t c u r v e s

specimen C - M I 0 for s p e c i m e n C - M 1 0

B-C' r-4
~9= 0.029
.~ 1 0 ;
0 LVDT-1
LVDT-2 5;

A: Stroke control
B: LVDT 3 or 4 control
C: LVDT 4 control
0 2000 4000 6000 0000 -0.010 O.OOO 0.010 0.020 0,030
Time (second) Displacement (ram)

Fig. 5--(a) Deformation-time curves for concrete specimen C-MIO; (b) Load-deformation curves for con-
crete specimen C-MIO

184 9 September1993
AE s i g n a l s oeeured at t= 2380 second
for the specimen shown in Fig. 5 mation was used to decide the mode of control in cases
4.0 where two or more choices existed on the basis of LVDT
readings. In such cases, information from the latest AE
2 Transducer E
80 ~a; events was used to reduce the number of choices to one.
In fact, AE was given somewhat higher importance in such
Transducer D decisions. Although this might appear highly subjective,
success rate of the experiments conducted (almost 100
Transducer C percent) indicates that such subjectivity is not a significant
i 1.0 problem.
Locations of AE events that were registered during the
[- /ran time indicated by region A of Fig. 5(a) are shown in Fig.
7(a). Region A corresponds to time t bounded as 0 < t <
1490 s. (Please note that this source-location analysis was
-1.0 -~ ............. i, ............. i done after the test and is exact within limitations of experi-
O.O0000 0.00006 0.00012
Time (second) mental measurements.) Roughly speaking, these locations
Fig. 6--AE signals for the specimen C-M10 at are closest to LVDT-3. At the border between regions A
t = 2380 se and B in Fig. 5(a), control mode was first switched to
LVDT-3 despite the fact that slopes of the curves of
LVDT-3 and LVDT-4 were almost identical, and that the
reading of LVDT-3 was actually smaller of the two.
Source locations at a later time in the region B of Fig.
AE events were recorded, As explained earlier, and as 5(a) are shown in Fig. 7(b). Region B corresponds to the
indicated in Fig. 4(a) and Fig. 5(a), the control mode had time interval 1490 s < t < 2500 s. During this time,
to be changed frequently at this stage in accordance with acoustic events continued to originate at locations near to
the current time-rates of outputs of the LVDTs. In many either LVDT-3 or LVDT-4, that is, roughly in the right half
cases, near the peak load, one of the LVDTs showed a of the specimen, but these locations were otherwise ran-
rather sudden jump in its reading. For example, in Fig. 4(a), dom. Owing to this randomness, the control mode had to
LVDT-1 exhibits such a sudden jump (in comparison with be changed frequently between LVDT-3 and LVDT-4. In
other LVDTs). At that time, control was quickly switched this context, it is instructive to examine readings of LVDT-
to LVDT-1. (A slight delay in this operation would have 3 and LVDT-4 in the region B. Notice that the slopes of
caused unstable failure. It is at this point that the ability of
quick switching of the control anode is critically impor-
tant). After this time the controlling LVDT continued to
show a linear increase in output. Responses of the other 120.00
LVDTs exhibited a significant change in slope (of the
response-time curve). In fact, the slopes changed from Cr ~z
positive to negative. It is important to observe that LVDT- v 80.00

3 which was located on the opposite side of the controlling

LVDT (that is, LVDT-1) showed the largest change in
slope. This implies that in the post-peak region, the speci- I 40.00
men tends to rotate in its plane as the major crack propa-
gates, mThe amount of rotation observed in a test obviously
depends on the boundary conditions provided by the load- r
0.00 . . . . . . . . , . . . . . . . . . . . . . .
ing fixtures used. 0.00 40.00 80.00 i 80.08
X-direction (mm)
AE e v e n t s d u r i n g stage A

D e f o r m a t i o a - t i m e curves for concrete

The Role of AE ~iraen C-MIO
0.089 y

Use was made of the AE signals to obtain qualitative

2 4X LVDT-4
information about cracking within the specimen. By in-
specting the signals on the CATALYST screen, it is pos-
sible to quickly deduce (albeit approximately) the location
of the AE source. An example is provided in Fig. 6. It is
clear from this figure that the source of this AE event is
closest to transducer B and farthest from transducer E. In 9e~
~ LVDT-2

other words, the source is closest to LVDT-1 and farthest

~:. Strolce c o n t r o l
from LVDT-4. Moreover, comparison of amplitudes of LVDT 3 o r 4 c o n t r o l
C: LVDT4 c o n t r o l
different signals obtained from two different events also --0.051 ~
, ......... ,. . . .
9 ....
Time (second)
gives the experimenter an idea of relative amounts of
energy release corresponding to the two events. Fig. 7(a)-- Locations of AE events reg-
Information from AE events was u'sed as a supplement istered during stage A for specimen
to the information from LVDT signals. Often, AE infor- C-M10

ExperimentalMechanics 9 185


BO.O0 r
c) I 40.00

I 40.00

4w~ ~,r 0.00

0.O0 40,00 80,00 120.00
X-direction (ram)
AE events during stage C
0.00 40.00 ao.oo 1~o.oo Deformation-time curves for concrete
X-direction mm) ~eoimen C-MIO

AE events during stage B 0.969

I . ' 3
D e f o r m a t i o n - t i m e curves for concrete LVDT 4
specimen C - k i 1 0
0.069 -~ 8.029

~ L V D T - l
9~ 0.029 LVDT-2

DT-3 ~: Stroke control

LVDT3 or 4 control
CI LVDT 4 control
~ ~ B IC "~_ _LVDT-t 2000 4000 6000 8000
Time (second)
c~ LVDT-2
7(c)-- Locations of AE events regis-
tered during stage C for specimen C-
A: Stroke control
B: LVDT 3 or 4 control M10.
C: LVDT 4 control
-0.051 ................... , ......... , . . . . . . . . .

0 2000 4000 6000 8000

Time (second)
Fig. 7(b)-- Locations of AE events regis- Y
tered during stage B for specimen C-MIO

the curves corresponding to LVDT-3 and LVDT-4 are

almost identical and that the reading of LVDT-3 is smaller
in magnitude compared to that of LVDT-4. ,,~
Near the end of region B, LVDT-4 reading was substan-
tially higher than all other LVDT readings. This observa-
tion, coupled with the then most recent AE event locations
(obtained approximately and in real time) suggested that Major crackposition
LVDT-4 be retained as the controlling LVDT at the end of for concretespecimen
region B. C-M10
Figure 7(c) shows source locations of AE events registered 7(d)-- Major crack position for
during the time t > 2500 s. Tendency of the AE sources to concrete specimen C-M10
concentrate in a rather narrow zone in the bottom right comer
is easily seen from this figure. This indicates formation of a
major crack. Trace of the major crack visually observed on
the surface of the specimen at the end of the test shown in Fig.
7(d) supports this conclusion.
when a given control mode is used. Results obtained from
this work can be used to examine stability of some control
Some Other Control Methods and Stability modes that may be considered intuitive choices in a tension
of Tests test. We consider two such choices: (1) the signal obtained
by averaging the signals of the 'controlling' LVDT and the
Obtaining stable failure in a tes[ depends on the feed- LVDT situated on the opposite side (for example, LVDT-2
back signal used. The phrase 'stability of control mode' and LVDT-4). (2) Signal obtained by averaging the signals
will be used in the following discussion to describe of all four LVDTs. (This is the same as total extension of
whether a specimen undergoes gradual (stable) faiIure the specimen.)

186 9 September1993
Deformation-time curves for concrete S n a p b a c k c a u s e d by a v e r a g e o u t p u t
specimen C-)/iI 0 of f o u r LVDTs for s p e c i m e n C-MIO
i J



o ,, . . . . . , , i , , . . . . . . . i . . . . . . .
ooo 4000 0ooo ~176 0.0o0 0,005 0,010 o.015 0.020
Time ( s e c o n d ) Average o u t p u t of f o u r LVDTs ( m m )

Fig. 8~Three possible feed-back sig- Fig. 1 O---Snapback caused by average

nals for concrete specimen C-MIO of signals from all four LVDTs for speci-
men C-MIO

Snapback caused by average output Specimen C-M10

of LVDT 2 and 4 for specimen C-MIO

i 1
,.~ 4.00 t 1~~

~3 ~3.00



0 r ........ , ......... ~......... ,,,, ...... , ......... 0.00 ~ ....... i .... , .... ~ ~ T
0,000 0.005 0.010 0.015 0,020 0.025 0.000 0.020 0.040 0.060
A v e r a g e o u t p u t o f LVDT 2 a n d (mm) O u t p u t of LVDT-4 ( r a m )

Fig. 9--Snapback caused by average of Fig. 11--Stress-deformation curve for

signals from LVDT-2 and LVDT-4 for LVDT-4 of specimen C-MIO
specimen C-MIO

From the point of view of closed-loop control theory, Signal from LVDT-4 does not lead to snapback and is
feed-back signals for which mechanics of the specimen- hence suitable for control.
machine system dictates reversal of sign of the slope of
response-time curve will lead to unstable failure at the
point where the sign reversal happens. If the sign of the Stress-deformation Curves
slope remains unchanged, then a well-tuned closed-loop
The stress-deformation curves obtained are shown in
control mode is stable. Fig. 12 for three plain concrete specimens. The fact that
For the point of view of structural mechanics, any these curves are virtually coincident implies that the results
system for which the load-displacement curve exhibits of this test method are quite reproducible. 'Deformation'
snapback is unstable under either load or displacement refers to displacement measured by the LVDT which was
control. in control in the post-peak region. The curves can be
Stability of the two control modes mentioned above can divided into three parts: linear-elastic part, nonlinear pre-
be discussed with reference to Figs. 8-11. Data from plain peak part and post-peak part. In the first part, concrete
concrete specimen C-M10 is used in these diagrams. In behaves elastically. Linear-elastic part is characterized by
Fig. 8, the averaged signals exhibit reversal of the sign of uniform deformation and 'global' behavior of the material.
slope, whereas the signal from LVDT-4 increases mono- In the second part, due to the damage (indicated by occur-
tonically as the test proceeds. If the averaged signals are rence of AE events) in the specimen, the modulus of the
used, they will lead to unstable failure. Figures 9 and 10 material starts to reduce and thus nonlinearity appears in
show stress-displacement diagrams corresponding to the the stress-deformation curves. Since damage does not hap-
averaged signals. Figure 11 shows the stress-displacement pen uniformly in the specimen, behavior of the material
diagram corresponding to the signal of LVDT-4 (the con- ceases to be 'global.'
trolling LVDT). The fact that the averaged signals lead to After the peak-load, a major crack develops in the speci-
snapback implies that they are not suitable for control. men. Behavior of the specimen in this region can be explained

Experimental Mechanics ~ 187

Uniaxial Teasiola Test Results: U n i a x i a t T e n s i o n Test. R e s u l t s :
Plain Concrete Specimens Plain and Fiber Reinforced Concrete
4.00 I - - S p e c i m e n SFRC-F21
5.00 -- S p e c i m e n PFRC-FI8
Specimen C-FI4

S.O0,l ~ ~3.00
2.00 er

-- S p e c i m e n C-F14
U)2,001.I 00 ~.o~Pl
S p e c i m e n C-MI3
S p e c i m e n C-k126
0,00 ~ ~ " 0.00
0.00 0.02 0.04 0.06 0.08 0.10 0,000 0.020 0.040 0.060 0.080 0.100
Displacement (turn) Displacement (rnm)
Fig. 12--Stress-deformation curves for Fig. 13--Comparison of typical stress-defor-
three of the concrete specimens tested mation curves for the three materials tested

with the help of fracture-mechanics theory. One of the which arbitrary and quick switching among multiple con-
results from this theory is that stiffness of a specimen trol channels is needed for stable response.
(defined as the ratio of load to load-point displacement)
decreases as a result of growth of a crack. In the tests
described here, the test machine was issued a command to
apply load in such a way that the response of the control- This research is sponsored by a grant from U.S. Air Force
ling LVDT increased linearly with time, To do this in the Office of Scientific Research managed by Dr. Spencer T. Wu.
post-peak region, the machine must decrease the load on Support from the NSF Center for Science and Technology of
the specimen in order to maintain equilibrium. Such reduc- Advanced Cement-Based Materials is also appreciated.
tion in load was explained in earlier works by hypothesiz-
ing that the material becomes 'soft' in the post-peak
region. Today, it is understood that 'softening' response in References
tension is actually a manifestation of growth of a single 1. Evans, R.H. and Marathe, M.S., "Microcracking and Stress-strain
crack. Curves for Concrete in Tension," Mat. and Struct, Research and Testing,
Figure 13 shows stress-deformation curves for the three RILEM, Paris, 1 (1), 61-64 (Jan./Feb. 1968).
materials tested: plain concrete, PFRC and SFRC. General 2. Petersson, P.-E., "Crack Growth and Development of Fracture
characteristics of the curves are similar for these three Zones in Plain Concrete and Similar Materials," Rep. No. TVBM-
1006, Lund lnst. of Tech., 174 (Dec. 1981).
materials. However, fiber-reinforced concrete shows a 3. Schorn, H. and Berger-Bocker, T., "Test Method for Determining
more ductile behavior (greater area under stress-deforma- Process Zone Position and Fracture Energy of Concrete," EXPERIMENTAL
tion curve) as expected. TECHNIQUES,29-33 (June 1989).
4. Gopalaratnam, V.S. and Shah, S.P., "Softening Response of Plain
Concrete in Direct Tension," ACI J., 310-323 (May~June 1985).
Conclusions 5. Li, V.C., Chan, C.M. and Leaung, C., "Experimental Determination
of the Tension Softening Relations for Cementitious Composites," Ce-
A method to obtain the complete tensile response of an ment and Concrete Res., 441-452 (1993).
unnotched concrete specimen has been developed. Results 6. Hordijk, DA. and Reinhardt, H.W., "Fracture of Concrete in
of the tests reported in this paper are believed to provide Uniaxial Tension Experiments as Influenced by Curing Conditions,"
Eng. Fract. Mech., 35 (4/5), 819-826.
important information about deformation history of four 7. Cho, B.S., EI-Shakra, Z.M. and Gopalaratnam, V.S., "Failure of
different portions of the specimen, The test method re- FRC in Direct and Indirect Tensile Test Configurations," Proc. Int.
quires a digitally controlled testing machine and five con- Symp. on Fatigue and Fracture in Steel and Concrete Structures, Oxford
trol channels. The work reported here also demonstrates & 1BH Publishing Co. PVT., Ltd, New Delhi, India, ed. A.G. Madhava
Rao and T.V.S.R. Appa Rao, 587-601 (Dec. 1991).
usefulness of information from acoustic emission events
8. Gopalaratnam, V.S., Shah, S.P., Batson, G.B., Criswell, M.E.,
as a valuable guidance in controlling a tensile test al- Ramakrishnan, V. and Wecharatana, M, "Fracture Toughness of Fiber
though, on a real-time basis, such information is of a Reinforced Concrete," ACI Mat. J., 88 (4) 339-353 (July/Aug., 1991).
qualitative nature. A limitation of the experimental proce- 9. Hordijk, D.A., "Local Approach to Fatigue of Concrete," PhD
dure suggested here is that owing to the absence of fast and Thesis, Univ. of Delft (1991).
reliable analysis procedure and computing power, it is not 10. Bazant, Z.P., Tabbara, M.R. and Kazemi, M.T., "Stable Path
of Interacting Crack Systems and Micromechanics of Damage," 3,
possible to obtain exact source locations of AE events in Advances in Fracture Research, Proc. 7th Int. Conf. on Fract., ed.
real-time. The basic control philosophy of this test method Salama, Ravi-Chandar, Taplin and Rao, Pergamon Press, Oxford, 2141-
has the potential to be used in a variety of other tests in 2150 (1989).

188 9 September1993