A systematic experimental research was carried out to investigate the conventional materials and mixing, placing and cun
the mechanical properties of highperformance concrete (HPC). A practices." Because HPC is much superior to ordin~
total of 9] plate specimens were tested in four different loading concrete in strength, toughness, workability, and durabili
conditions including uniaxial tension, uniaxial compression, which makes HPC particularly suitable for construction
biaxial compression, and tension compression with the help of a
tall buildings, longspan bridges, and other special structu
closedloop testing machine. Uniaxial and biaxial complete stress
such as nuclear reactor containment, thus considerab
strain curves were obtained under a strain control loading scheme.
Based on the stochastic process theory, the average curves and attention has been attracted to relevant researches in the p
standard deviation curves were obtained for uniaxial loading two decades (Rougeron and Ai:tcin 1994; Gutierrez a
conditions. The rehardening of concrete after having entered the Canovas 1996; Persson 1996; !ravani 1996; Neville
softening stage was observed in this experiment. The ultimate Aikin 1998; Candappa et al. 1999; Pfeifer 2000; Chu
strength envelopes in both stress and strain space were developed 2002; Aikin 2003). On the other hand, most conct
through parameter identification of the complete stressstrain (including HPC) materials used in modem structures, such
curves. This research laid a foundation forfurther research on the those mentioned previously, generally suffer multiaxial str
multiaxial constitutive law of concrete. and damages. So it is reasonable and necessary to empl
nonlinear material models verified by multiaxial experime
in the refined analysis of concrete structures (Kupfer et
1969; Buyukozturk et al. 1971; Nelissen 1972; Tasuji et'
INTRODUCTION 1978; van Mier 1986; Yin et al. 1989; Li and Guo 199
Although concrete is currently the most widely used Traina and Mansour 1991; Candappa et al. 1999; Niels
construction material in the world, its mechanical properties 1998; Hussein and Marzouk 2000; Lee et al. 2003; Seow a
when suffering damages have not been fully understood, Swaddiwudhipong 2005).
which may be the reason that, up until now, no generally In the case of biaxial loading experiments on concrel
accepted nonlinear theory had been available for the analysis although started as early as in the 1930s, the results we
and design of concrete structures. In fact, the complexity of found valid only after the influence of the friction betw
concrete material originated from two essential characteristics, loading platens and concrete specimens on testing resul
that is, nonlinearity and randomness (Li 2004). As the had been correctly considered, which was first done
nonlinear continuum mechanics and finite element theory Kupfer et al. (1969), who adopted a brush platen to minimi
have achieved great progress in recent years, the accuracy of the friction as much as possible. The essential considerati
the nonlinear analysis of concrete structures will greatly is still the paradigm until recently. For example, Li and G
depend on the development of constitutive models of (1991) reported a series of biaxial experiments on C20 ordin
concrete. Always a fundamental problem in concrete concrete with the 28day compressive strength of 20
research, the constitutive relationship has drawn intensified (2.9 ksi). Sixty 100 mm (3.94 in.) cubic specimens we
study and many models have been proposed, among which loaded under compressioncompression, compressio
the fracture mechanicsbased models (Hillerborg et al. 1976; tension, and tensiontension and a piecewise continuo
Bazant and Oh 1983; Bazant 1984; Bazant and Ozbolt 1990; function of the biaxial envelope was obtained by data regressio
Carol et al. 2001), and the continuum damage mechanics It should also be noted that three layers of butterlubricat
based models (Ortiz 1985; Mazars 1986; Mazars and aluminum foil were employed in the tests as the antifrictio
PijaudierCabot 1989; Iu 1989, Lee and Fevens 1998; Wu et Lee et al. (2003) experimentally investigated the biaxi'
al. 2006) are the hotspots at present. On the other hand, the behavior of concrete used in nuclear containment buildin
stochastic theory has been successfully introduced into the Specimens 200 x 200 x 60 mm (7.87 x 7.87 x 2.36 in.) we
structural analysis of reinforced concrete structures and loaded in the stress ratios of 0, 0.2, 0.5, and 1.0 for compressio
several stochastic constitutive models have already been compression; 0.05, 0.1, 0.2, and 0.3 for compressio
developed (Krajcinovic and Silva 1982; Carmeliet and Hens tension; and 00, 5, 2, and 1for tensiontension. In addition, 0.1
1994; Kandarpa and Kirkner 1996; Li and Zhang 2001). A (0.00394 in.) tetrafluorethylene pad was used to erase t
lack of systematic and accurate experimental data for verification, friction at the interface of the concrete and loading platens.
however, restricts the acceptance of these new models.
The term "highperformance concrete" (HPC) was coined
in 1980, although sporadic researches started much earlier ACI Materials Journal, V. 105, No.6, NovemberDecember 2008.
MS No. M2007320.R I received May 12, 2008, and reviewed under [nsti
(Aikin 1998). In 1993, ACI cleared the definition ofHPC as publication policies. Copyright © 2008, American Concrete Institute. All rights res<:
"concrete which meets special performance and uniformity including the making of copies unless permission is obtained from the copyright propri
Peninent discussion including authors' closure, if any, will be published in the Septem
requirements that cannot always be achieved by using only October 2009 ACI Materials Journal if the discussion is received by June 1,2009.
Xiaodan ReD is a PhD Candidate in the School of Civil Engineering at Tongji University,
Shanghai, China. His research interests include theoretical and experimental research
of the constitutive law of concrete and nonlinear analysis of structures.
Constituent Weight, kglm3 (lb/ft3)
Portland cement 204 (12.75)
WeizhODg Yang is an Associate Professor in the School of Civil Engineering, Zhengzhou
Fly ash 204 (12.75)
University, Zhengzhou, China. His research interests include the constitutive law of
concrete, analysis of concrete structures, and design a/masonry structures. Groundgranulated blastfurnace slag 102 (6.38)
Water 175 (10.94)
Yong Zhou is a Lecturer in the School of Civil Engineering at Tongji University. He
received his PhD in 2005 from Northwestern University, Evanston, IL His research Highrange waterreducing admixture 15.3 (0.956)
interests include the constitutive law and durability of concrete. Natural siliceous sand 640 (40)
Jie Li is the Distinguished Professor and Director of the Institute for Building Structures Crushed limestone 1100 (68.75)
in the School of Civil Engineering at Tongji University. He received his PhD in 1988
from Tongji University. His research interests include structural and seismic engineering,
essential mechanical properties of concrete, and the theory of stochastic structure
analysis and modeling.
Table 2Strength comparison of cut and uncut
specimens
Although the biaxial behavior of concrete has been
Dimension, Peak load, Peak stress, Mean, Remar
experimentally investigated comprehensively, most of the mm(in.) kN (kips) MPa (ksi) MPa (ksi) k
experiments only involved ordinary concrete. Except some 178 x 178 x 57.4 511 49.7
remarkable early works (Scavuzzo et al. 1983; van Mier (6.20) Cut
(7.00 x 7.00 x 2.26) (114)

1986; Vonk 1992), almost all of the previous biaxial 180 x 180 x 59.3 534 50.0
(120) Cut
experiments only concern the strength envelopes, that is, (7.09 x 7.09 x 2.33) (7.25) 51.3

curves of the peak stresses, instead of complete stressstrain 180 x 180 x 59.6 520 48.6 (7.43)
Cut
curves, which contain more abundant information about the (7.09 x 7.09 x 2.35) (117) (7.05)

multiaxial behavior of concrete. 179 x 179 x 56.5 576 56.8
Cut
(7.05 x 7.05 x 2.22) (129) (8.24)
The objective of the current research was to experimentally
200 x 200 x 57.5 548 47.8
investigate the mechanical properties of HPC materials (7.87 x 7.87 x 2.26) (123) (6.93) Uncut
under uniaxial and biaxial loading conditions. A total of 
200 x 200 x 58 559 48.3 48.6
91 squareshaped plates made of HPC were tested in four Uncut
(7.87 x 7.87 x 2.28) (126) (7.00) (7.05)

different loading conditions, including uniaxial tension, 200 x 200 x 58 576 49.8
Uncut
uniaxial compression, biaxial compression, and tension (7.87 x 7.87 x 2.28) (129) (7.22)
compression. The full uniaxial and biaxial stressstrain
curves and the strength envelope were obtained after
analyzing the experimental data. Although the load carrying waterreducing admixture was used to improve the workability
capacity and the strength still play important roles in the ofHPC. Table 1 shows the detailed mixture proportions ofHPC.
design and analysis of concrete structures, this study is primarily A slump of 200 to 220 mm (7.87 to 8.66 in.) and a slump
concerned with more elaborate behaviors ofHPC during biaxial flow of 450 to 500 mm (17.71 to 19.68 in.) were obtained for
loading process through experimental approaches. HPC made according to the aforementioned mixture
proportions. And the 28day compressive strengths
RESEARCH SIGNIFICANCE measured by using the cubic (150 x 150 x 150 mm [5.90 x
As HPC has currently been extensively used in many parts 5.90 x 5.90 in.]) and prismatic (100 x 100 x 300 mm [3.94 x
of the world and is generally in a multi axial stress state 3.94 x 11.8 in.]) specimens were 71.3 MPa (10.34 ksi) and
during its service life, there is an urgent need for a comprehensive 63.3 MPa (9.18 ksi), respectively.
and accurate multi axial constitutive law for the precise and
reliable evaluation of the nonlinear behavior of modem Specimens
concrete structures. This experimental research aims to Specimens used in this experiment were prepared in two
provide necessary data for the validation and improvement steps. First, large plate specimens with dimensions of 500 x
of the existing multiaxial constitutive laws for HPC as well 500 x 50 mm (19.68 x 19.68 x 1.97 in.) were cast in wooden
as ordinary concrete. Besides the routine strength envelope molds and carefully compacted with a shake table. The
for biaxial experiments, the complete stressstrain curves specimens were demolded after 48 hours and then moved to
were particularly obtained from this research with biaxial a curing room where the temperature was 20°C (68 OF)and
proportional strain controlled loading scheme, which had not the relative humidity exceeded 95%. Second, after 28 days
been previously reported. of curing, the specimens were cut into small pieces of 150 x
150 x 50 mm (5.90 x 5.90 x 1.97 in.) with the help of a high
precision cutter. The cutting surfaces were smooth and neat
Materials so that no further preparation of the loading surfaces was
In this experiment, three cementitious materials were necessary. Some preliminary tests were conducted to determine
used. The first was ASTM Type I cement with a 28day whether the cutting process would introduce initial damages
compressive strength of 42.5 MPa (6.16 ksi). The second into the HPC specimens (Table 2). It can be found that small
was the first grade fly ash, with specific surface being specimens directly cast using wooden molds of 150 x 150 x
6000 cm2jg (4.21 x 105 in.2/lb). The third was ground 50 mm (5.90 x 5.90 x 1.97 in.) have the similar compressive
granulated blastfurnace slag (GGBFS). strength as specimens cut from large specimens of 500 x 500
The fine aggregate was natural siliceous sand. The coarse x 50 mm (19.68 x 19.68 x 1.97 in.). It can therefore be
aggregate was crushed limestone with a maximum size of concluded that the influence of the cutting process on
15 mm (0.59 in.), which is less than onethird of the shortest mechanical properties of HPC specimens is negligible in the
dimension of specimens. A commercially available highrange following tests and analyses.
40
(5.80)
10
(1.45)
~mm(l.97in.)
Interlayer
Concrete near the loading surfaces was confined by fri
between the concrete specimen and the loading platens u
compression. As well known, Kupfer et al. (1969) reae
Strain au e the conclusion that such confining effect would result in
I 100x5mm apparent increase of the compressive strength of cone
I
I (3.94 x 0.197in.) To avoid the restraint effect under compressive loading,
Specimen : slices of tetrafluorethylene pads were inserted between con
~" specimens and steel platens in this experiment. The thic
" of each tetrafluorethylene pad was 0.1 mm (0.00394 in.).
the other hand, when subjected to tension, the concrete spec'
were glued directly to the steel platens using a structural adhes
with the tensile strength of 35 MPa (5.01 ksi). The adhesi
took less than 48 hours to reach the specified tensile stren
Experimental system Herein, both the tetrafluorethylene pads and the struct
According to the discussions of Krajcinovic (1996), to adhesive between the loading platen and concrete spec'
obtain the full stressstrain curves of concrete, especially in were referred to as interlayer. It is obvious that the interla
tension tests, a stiff loading system and a deformation would be deformed during the loading process. Theref
controlled loading scheme are the prerequisites. Thus, a stiff the output of extensometers actually consisted of deformati
framed servohydraulic closedloop testing machine was of the interlayer and the concrete specimens. In other wo
the original stressstrain curves collected by the data acquisili
adopted in the current study. The biaxial loading system
system were contaminated. To eliminate the contaminati
consisted of a vertical fourcolumn testing machine and a
to make the subsequent experimental data analysis accura
horizontal loading frame, both of which had the stiffness of
a simple uniaxial loading test with the concrete specim
more than 6 x 106 MPa (8.7 x 105 ksi), sufficient for
replaced by a steel plate of the same dimension was carri
preventing sudden rupture of specimens during the loading
out in advance. Then the stressdeformation relationship
process. In both the vertical and horizontal directions, two one interlayer was calculated from the following equation
extensometers were set up along the whole length of the
specimen to measure the deformations as well as global
strains (Fig. 1). To obtain the full stressstrain curves in each
loading direction, the outputs of extensometers in the same
direction were averaged and collected by the data acquisition where Li. was the deformation of the interlayer, EeCcr) was
system as the feedback signals. A microcomputer was global strain measured by extensometers, EgC cr)was the strain
employed for collecting the feedback signals and giving the steel plate measured by strain gauges, and L stood for t
command signals calculated in a real time way based on a length of the steel plate. Figures 3 and 4 show the stress
closedloop control algorithm. Besides the extensometers deformation curves and corresponding linear regressio
mentioned previously, strain gauges and a corresponding curves of the tetrafluorethylene pads and structural adhesive
data collection system (Fig. 2) were used to measure the respectively. It is obviously that a linear stressdeformatio
local strains of concrete specimens. model of the interlayer is accurate enough for furthe
The vertical direction was set as the primary loading direction data analysis.
because the transducer in that direction had a higher resolution.
Realtime communication between the vertical and horizontal Experiment design
loading systems was established to ensure the constant strain The behaviors of HPC specimens, with the dimension 0
ratio (the ratio of global strains in the two directions measured 150 x 150 x 50 mm (5.90 x 5.90 x 1.97 in.) subjected to
by extensometers) during biaxial loading experiments. uniaxial and biaxial loading were investigated in this research.
1.5
(0.212)
3.5 60
(0.507)
(8.70)
3.0
(0.435)
2.5
(0.362)
2.0
(0.290)
1.5
(0.212)
1.0
(0.145) J
0.5
(0.0725)
f
0.0
o 300 400 4000
& 10{i & lO{i
4.0
(0.580)
3.5
(0.507)
3.0
(0.435)
2.5
MPa
CF (ksi) Mean

Standard deviation
MeanStandard deviation
T Mean+Standard deviation
60
(8.70)
50
(7.25)
0'
MFa
(hI'
f~'
~
;;:;;..~ \
Mean
T
Standard deviation
MeanStandard deviatioo
Mean+Standard deviatiOi'
(0.362)
1/1 \\~\
2.0
(0.290)
1.5
30
(4.35) II ~I ~
(0.212)
1.0
(0.145) 4
20
(2.90) f .....:~~
0.5
(0.0725)
0.0
,
p' .
.
10
(145)
o
II...
__

..
o 300 o 3000
& 106 E lO{i
Fig. 7Statistical characteristics of complete stressstrain Fig. 9Statistical characteristics of complete stresssf
curve under uniaxial tension. curve under uniaxial compression.
where N is the number of curves, and the average stress is cr deviation, and coefficient of variation of the peak stress
a function of strain E. On the other hand, the standard deviation strain, of uniaxial tensioncompression experiments.
S was adopted to represent the variability among experimental Strain localizationIt is well known that the nonlin
curves according to the statistics theory. behavior of concrete originates from the propagation
cracks inside the cement matrix and aggregates. After
peak in the descending branch, concrete specimens w
divided into several pyramids and wedgeshaped blocks
cleavage cracks. The global deformation was localized i
the opening and sliding deformation between both sides
A set of experimental curves under uniaxial tension are cracks. Although the global strain kept increasing due
plotted in Fig. 6. The corresponding average curve and standard external loading of the testing machine, the local str'
deviation curve are illustrated in Fig. 7. Figure 8 shows part recorded from separated blocks of specimen started
of the experimental curves under uniaxial compression. decrease. This phenomenon is known as strain localizati
Figure 9 displays the average curve and standard deviation which is discussed in Newman and Sigvaldason (1965), \
curve corresponding to Fig. 8. It is observed that the average Mier (1986b and 1991), and Torrenti et al. (1993).
curve is similar to the experimental curves but smoother than In Fig. 10, the axial stressstrain curves for uniaxial tens'
any individual curve. Both Fig. 7 and 9 present that the standard and compression are shown. It is observed that the I
deviation curve also has an ascending part and a descending strains recorded from the surface of specimen are smaller
part. Furthermore, the strain corresponding to the peak of the global strains recorded by extensometers between loa'
standard deviation curve approximates the strain corresponding platens. Therefore, in the descending branch of the stre
to the inflection point of the average curve. Tables 4 and 5 list strain curve, structural behavior becomes essential rather
the statistical characteristics, including the average, standard material behavior. The aforementioned results suggested
MPa
3.0 a
(0.435) (ksi)
2.5
(0.362)
2.0
(0.290)
1.0
(0.145)
0.5
(0.0725)
60
(8.70)
50
(7.25)
40
(5.80)
30
(4.35)
20
(2.90) (b) Failure schematic
10 Fig. iiSingle shear failure mode.
(1.45)
o
o 1000 2000 3000 4000 5000 6000 7000 Table 5Statistical characteristics of uniaxial
l: 10 6 compression experiments
(b) Uniaxial compression Uniaxial compression
Number Peak stress, MPa (ksi) Peak strain, ~e
2AC2 56.48 (8.19) 1906
2AC4 50.39 (7.31) 1778
2AC8 48.11 (6.98) 2395
Table 4Statistical characteristics of uniaxial
tension experiments 2AC9 45.45 (6.59) 2325
2AC1O 61.80 (8.96) 2185
Uniaxial tension
2ACl1 49.44 (7.17) 2171
Number Peak stress, MPa (ksi) Peak strain, ~e
2AC12 54.57 (7.91) 1820
2AT1 3.56 (0.516) 148.8
Average 52.32 (7.59) 2082
2AT2 3.27 (0.474) 127.6
Standard deviation 5.62 (0.815) 247.5
2AT3 3.43 (0.497) 119.2
Coefficient of variation, % 10.7 11.9
2AT5 2.85 (0.413) 106.6
Average 3.28 (0.476) 125.6
Standard deviation 0.309 (0.0448) 17.74 failure modes of concrete specimens and suggested four
Coefficient of variation, % 9.42 14.1 types of failure modes needed to be considered in the 3D
experiment and analysis of concrete material. In the
current study, all speci mens subjected to biaxial
the proper local averaging approaches should be carefully compression failed in single shear failure mode or
considered during the continuum modeling of concrete. multiple shear failure mode, alternatively.
In the case of single shear failure mode (Fig. 11), specimens
Biaxial tests
were divided into two triangular prisms by a unique diagonal
Failure modeAccording to previous discussions, even
crack that inclined at angles of 30 to 40 degrees to the load
though the specimen was under uniaxial or biaxial loading
conditions, the cracks propagation and the ultimate failure free surface of the square plate. Specimens generally failed
modes were actually three dimensional (3D), let alone the in this mode when deformations in the two loading directions
more intricate multiaxial compression. According to van differed from each other. For multiple shear failure modes
Mier (1986a), there were two primary failure modes under (Fig. 12), specimens were split into several triangular
multiaxial compression, that is, the planar failure and pyramids by several diagonal cracks that inclined at
columnar failure. Yang (1989) discussed the multiaxial angles of 30 to 40 degrees to the corresponding loading
According to foregoing discussion in this paper,
reasonable to define the stressdeformation relation
interlayer in a linear form
r ~
stage during uniaxial compression, the strains in the load .•• (ksi)
o
(
:~o8
loading directions, the shear dilation became more and more 0'., o 1000 2000 .J(XX) 4OX) 6000 eooo 7000 8000
conditions at certain strain ratios. Whether this will happen .•• (ksi) ..••• ~.r>.
in practical concrete structures requires further studies. (8.70 / \.
... I
r, .
Strength envelope in stress spaceNot only the mean (4.35 /
curve but also the mean curves plus/minus one time of the 15
(2.1')
standard deviation are shown in Fig. 16. Normalized mean
strength envelopes in the compressioncompression region o
o 1000 2000
measured in this experiment are shown in Fig. 17, which also 100
~ ."
(ksi)
1.• 0", ·1.
(0.261 '0", (ksi) (2.61) 0') (ksi)
.•.
.•........ .. ......... r.~'".•..•.
4 
1,S . cr 1"e:1 ·80 ' ~  _.
 I
(0.218)  (2.18) A_
(11.6)
1,2
6 0'2£2 .'
.•..•. .• ·12 :t
....•....•..•. "'.
..
HU74) (1.74)
./~
. •
•........•...
0_
(130)
({I. 130)
·60
', ..
06
(0.0870)
./
/
.•.......... (0.870)
(8,70)
Y
l·
03 ...•
(0.0.435) 1•• ""JJ..Io. (0.435)
.•.. e (10") ·40
•
\\
0.0 0
(580) Mean
0 30 60 90 120 150 160 _ Mean5tandard deviation
.•........
0
e, (10") A Mean+Standard deviation
....... ..
..........
(2,90)
I' MFa
0'2,(ksi)
=t ~ ..I 1 2
. .~
.•... .~.~. a_. 0
0 20
(2,90)
·40
(5,80)
.~
60
(8.70)
~
80
(11.6)
 e, (10")
(a) MFa
Fig. 16Biaxial strength envelope in stress space.
MFa
3.0
(0.435) 0", (ksi) 0", (ks") .,.
'.,.. (2.61)
2.S e(J1""E
·15
1,5
1
(0.]6)
A<J E ",/ (2.18) e
e
e__
.......
1 .•.• ···\ .•.
',0
(0.290)
2 2
,
...•.....•.•.,. ,./
·12
(1.14) ,/' ""
15
(0.217)
••... /*
,.J' ~ V
.•.....•....
.•
.,
(1.30)

'" /
U
1,2 .••........••• ...•
\d
10
(..().14S) , ~ •..•..•.•.•.••.••. <:.870)
O.S •...... b
0,9
(o'o72S)
0.0 l.~
0 50 100 200 250
e, (10") 0.435)
0
300
'50
e Present exp
O·r •.•.• " e
·200
...... e, (10") 0.6  x Guo ZH et.al.
e Kupfer et.al.
...•.•....  •• Tasuii et.al.
'. e
""" ..•....•' .. d
... .....
0,3
EOO
600
·1000

e, (10")
£1&2
I
......•.....•
.•.•.•.... 0,0
0.0 0.3 0.6 0.9
/ 1.2 1,5
b) (j1 I fc
MFa MPa
25
(oJ63) 0", (ksi) . CJ1"'e:1
(J
'~
I 10
••••• (I.4')
Fig. 17Normalized biaxial strength envelope in stress sp
.•.•..
/ I"...
2.0 ~cr2E2
(0.290) •••••• .,., (~.16)
·5000
.•
.•.. ~
I.S
(0.217)
1.0
f
'
/'
V
~I''•• ...,
.......•.• .....
(0.870)
..
(0.580) ~4000
~
/

I
(0.145)
V "i'0
os .••.... .•.~ ·2 ..
(o.072S)
00
1/ e, (10")
0
(0.290)
"''''.3000 ~ ~
1\\\
0 50 100 150 200 250 300 350 400 450
0
....• e (10")
..•..•. ....•. ·2000
..~..•.•10·•.•.•
·200
400
.•..... 1000
I~Mean
~
__ Mean  Standard deviation
Mean  Standard deviation ~\\
.....•.•..
600 ~&1 .•.E2 .•.• .•...•...• ).
0 i
600 ~10·LL ~ 0 1000 2000 3000 4000 5000
(C) E
1
(106)