Determination of Double-K Criterion For Crack Propagation in Quasi-Brittle Fracture, Part I: Experimental Investigation of Crack Propagation

International Journal of Fracture 98: 111–149, 1999.
© 1999 Kluwer Academic Publishers. Printed in the Netherlands.
Determination of double-K criterion for crack propagation

in quasi-brittle fracture, Part I: Experimental
investigation of crack propagation
SHILANG XU and HANS W. REINHARDT

Institute of Construction Materials, University of Stuttgart, 70550 Stuttgart, Germany
e-mail: [email protected]
Received 10 July 1997; accepted in revised form 17 August 1998
Abstract. The results of experimental investigations using laser speckle interferometry on small size three-point
bending notched beams and using photoelastic coating and the strain gauges on very large size compact tension
specimens of concrete are presented in detail. The investigations showed that there exists a stage of stable crack
propagation before unstable fracture occurs. The results are in agreement with other researchers’ investigations us-
ing moire interferometry, holographic interferometry, dye-impregnation method and microscope. Further detailed
study shows that the three different states, i.e., crack initiation, stable crack propagation and unstable fracture can
be distinguished in the fracture process in concrete structures. In order to predict the crack propagation during
the fracture process in quasi-brittle materials a double-K criterion is proposed. The double-K criterion consists of
two size-independent parameters. Both of them are expressed in terms of the stress intensity factors. One of them
reflects the initial cracking toughness, denoted with Kcini , which can be directly evaluated by the initial cracking
load, Pini , and the precast crack length, a0 , using a formula of LEFM. The other one refers to the unstable fracture
toughness, denoted with Kcun , which can be obtained inserting the maximum load, Pmax , and the effective crack
length, ac , into the same formula of LEFM. The values of the two parameters, KIc ini and K un , obtained from
Ic
the small size three-point bending notched beams and the large size compact tension specimens show that KIc ini
un
and KIc are size-independent. Evaluating with the KR -resistance curves obtained from the same test data, it is
found that the proposed double-K criterion is equivalent to it in basic principle, but, the double-K criterion can be
applied more easily than the KR -resistance curve. Finally, as a practical example, the application of the double-K
criterion to the prediction of the crack propagation in a concrete dam is discussed.
Key words: Double-K criterion, crack propagation, quasi-brittle fracture, experimental investigation, laser speckle
interferometry, photoelastic coating, strain gauges measurement, three-point bending, compact tension, large size,
concrete, KR -curve, double-K fracture parameters KIc ini and K un .
Ic
1. Introduction
In recent decades, many have attempted to study the fracture process zone in concrete
materials and its softening properties so that their influence on fracture characteristics of
concrete structures can be modelled. The experimental investigations showed that the fracture
process in concrete structures includes three different stages: crack initiation, stable crack
propagation and unstable fracture (or failure). Usually, the whole region of crack initiation
and stable crack propagation is called the fracture process zone. In order to predict the crack
propagation and to reflect the influence of the fracture process zone on fracture characterist-
ics of concrete materials, several fracture models, like the fictitious crack model (FCM) by
Hillerborg et al. (1976), the crack band model (CBM) by Bazant and Oh (1983), the two-
parameter fracture model (TPFM) by Jenq and Shah (1985), the effective crack model (ECM)
112 S. Xu and H.W. Reinhardt
by Karihaloo and Nallathambi (1990) and Swartz and Refai (1987) and the size effect model
(SEM) by Bazant, Kim and Pfeiffer (1986) have been proposed. In the models, different
material parameters are introduced to describe the cracking properties of concrete materi-
als. Correspondingly, the test methods to determine the corresponding fracture parameters
which are GF defined in FCM, KIcs and CTODc in TPFM and Gf and cf in SEM have been
recommended by RILEM (1985, 1990a, 1990b).
In engineering practice, one may expect that a fracture model is accurate enough for pre-
dicting the behaviour of cracked structures and is simple for evaluating the corresponding
fracture parameters introduced in the model. The fracture parameters should only depend
on the properties of materials and not depend on the structural geometry and size. It is also
expected that the fracture parameters can be easily determined and evaluated using a direct
measuring method. For a normal structure, one may only need to predict its failure or unstable
fracture under given loading or displacement conditions accurately. But for some special and
important structures, for example, for a concrete pressure vessel or a high concrete dam,
accurate prediction of both failure and crack initiation are required. In some cases, accurate
prediction of the crack initiation is more important.
The fictitious crack model and the crack band model are nonlinear fracture models that use
the softening traction-separation law to model the fracture behaviour of the fracture process
zone. The two models can be utilized to predict crack initiation, crack propagation and failure
of a concrete structure using a finite element code. The softening traction-separation law
is completely determined by three material parameters which are the fracture energy GF ,
the tensile strength ft , and the crack opening at zero stress wc . According to the method
recommended by RILEM (1985), the fracture energy GF , can be experimentally determined
by three-point bending notched beams. However, it was found by many researchers that the
obtained values of the fracture energy GF are size-dependent (Hillerborg, 1985; Wittmann et
al., 1990; Xu and Zhao, 1989, Zhao et al., 1991). Moreover, if the softening traction-separation
law (σ -w curve) is approximated by different types (linear, bilinear and exponential types
etc.), the determined values of wc are different even though the values of GF and ft are
constant.
As the fracture parameters in the two-parameter fracture model (TPFM), the critical stress
intensity factor KIcs , and the critical crack tip opening displacement CTODc , are introduced
(Jenq and Shah, 1985). For determining them, an unloading and reloading procedure in a test
was performed so that an unloading compliance cu can be used to evaluate the effective crack
length ac . The nonlinear behaviour of concrete fracture is reflected by the evaluation of the
effective crack length ac . Then, the measured value of the peak load Pmax , and the evaluated
value of the effective crack length ac , are inserted into a formula of LEFM to determine the
critical stress intensity factor KIcs . This approach was recommended by RILEM (RILEM,
1990a). It can underestimate the effective crack length ac , because the inelastic part of CMOD
is not taken into account. It was found that the inelastic part, or the permanent deformation part
of CMOD has an important influence on crack propagation (Bazant, 1996) and the nonlinear
fracture characteristics of concrete is mainly associated with the fracture process zone or the
stable crack propagation. In addition, the stable unloading procedure in tests requires a closed-
loop testing system. The advantage of this method is to use only a single size of three-point
bend beams. All of KIcs , CTODc , ac and CMODc in TPFM model can be directly measured. So,
it is possible that the properties of size-independence of KIcs and CTODc claimed by Jenq and
Shah (1985) which are evaluated by the method described in (Jenq and Shah, 1985; RILEM,
1990a) could be further justified by the results that are directly measured.
Determination of double-K criterion for crack propagation in quasi-brittle fracture, I 113
Recently, Tang, Ouyang and Shah have pointed out that there is ‘somehow restricted
application of TPFM’ (Tang et al., 1995). Instead the method of determining fracture para-
meters KIcs and CTODc described in (Jenq and Shah, 1985) and recommended by RILEM
(1990a), they proposed a peak load method (Tang et al., 1995). But, by the recently proposed
peak load method, several three-point bend beams or several splitting tension cylinders are
needed to be tested in the experiments.
The effective crack model by Karihaloo and Nallathambi (1990) and Swartz and Refai
(1987) attempts to determine the critical stress intensity factor KIc , by inserting the measured
peak loads Pmax , and the evaluated effective crack length ac , into the formula of LEFM using
a test on a single size three-point bend beam. The effective crack length ac , is determined by
the measured peak load Pmax , and the corresponding deflection δp .
Differing from the material parameters introduced in the two-parameter fracture model and
the effective crack model, the critical energy release rate Gf and the critical effective crack
extension cf , for infinite specimen are introduced as the two material parameters into the size
effect model by Bazant, Kim and Pfeiffer (1986). A method to determine the values of Gf
and cf by testing several three-point bending notched beams of at least three different sizes
and a similar geometry has been recommended by RILEM (1990b).
All the above-mentioned fracture models that attempted to adapt LEFM to concrete struc-
tures can only be used to predict the unstable fracture of concrete structures. But, the fracture
models cannot be used to predict the crack initiation which has been observed by many
researchers with different investigating methods. Therefore, for some special cases the ap-
plications of the fracture models are somehow restricted.
The phenomena manifested in many experimental investigations show that the fracture
process in concrete structures includes crack initiation, stable crack propagation and unstable
fracture. Therefore, a new analytical fracture model is required for practical purposes. The
results of numerical calculations using the fictitious crack model of Hillerborg et al. (1976)
and the crack band model of Bazant and Oh (1983) also show that crack initiation, crack
propagation and failure such three different states during a fracture process until the maximum
load is reached. It could be imagined that as a new fracture model which can be analytically
used to predict the whole fracture process may be connected closely with the analytical solu-
tion of the fictitious crack although there is no valid analytical solutions of the fictitious crack
for concrete at present yet. So, recently an attempt is set up for a new fracture model from the
test data collected in the experimental tests to propose an easy testing method for determining
the fracture parameters introduced in the new fracture model by three-point bending notched
beams, compact tension specimens and wedge splitting specimens. To solve the analytical
solution of the fictitious crack for the quasi-brittle materials is one of the research purposes
too.
In order to respond to the different states in concrete fracture, a double-K is proposed. In
the double-K criterion, the two fracture parameters, (KIcini and KIcun ) are introduced. Both of
them are given in terms of stress intensity factor. The double-K fracture parameters can be
measured directly in experiments by three-point bending tests and compact tension tests. KIcini
is called initiation toughness and its value is determined by inserting the initial cracking load
Pini , and the initial crack length a0 , into a formula of LEFM. KIcun is called unstable fracture
toughness or the critical stress intensity factor and its value is determined by inserting the
measured maximum load Pmax , and the measured critical effective crack length ac , into the
same formula of LEFM. It is found that KIcini and KIcun are size-independent for the tested
specimens. Furthermore, the conventional KR -resistance curves for the tested specimens are
P
D
a0
S B
Figure 1. Three-point bending beam investigated by laser speckle interferometry.
obtained. Comparison with the double-K criterion shows that the proposed double-K criterion
can be regarded as a simplification of the conventional KR -resistance criterion. In the sequels,
the other parts will be presented.
2. Crack propagation in the small size three-point bend beams investigated by laser
speckle interferometry
In order to determine the starting point of stable crack propagation and the critical unstable
point of crack propagation, to investigate the shape of the fracture process zone and to measure
its length for different size specimens of concrete, four investigating techniques had been
studied at Dalian University of Technology. The four investigating techniques are laser speckle
interferometry, acoustic emission, photoelastic coating and strain gauges. The range of depth
of the tested specimens is from 0.1 meters to 3.6 meters. The details of these investigations
are described in (Xu, 1988; Xu and Zhao, 1989a; 1989b; 1991a; 1991b; 1991c; 1991d). In
this section, some results pertaining to the crack propagation in the small size three-point
bending beams are given. In the next section, the results of the crack propagation in large size
CT-specimens observed by photoelastic coating will be presented in detail.
2.1. S PECIMENS AND TEST PROCEDURE
Two groups of three-point bending notched beams with different sizes are used in the ex-
perimental tests to investigate crack initiation and stable propagation before unstable fracture
occurs, by laser speckle interferometry. The specimen dimensions are 700 × 150 × 100 mm3
and 515 × 100 × 100 mm3 . The specimen configuration is shown in Figure 1.
The initial crack/depth ratios a0 /d, are 0.4 and the span/depth ratios are 4. All concrete
specimens were cast with the mix proportion of 1:1.87:3.36:0.57 (cement: sand: aggregate:
water by wt.) with maximum aggregate size of 20 mm. These specimens were tested between
28-32 days after they were made. The cube compressive strength of the concrete material fcu ,
is 35.2 MPa, the splitting tensile strength ft , is 2.45 MPa and the modulus of elasticity Ec , is
31000 MPa.
For increasing the reflectivity some aluminum powder was applied and a net with 10 ×
10 mm2 meshes was drawn on the laser exposed side surface of each specimen. The loading
point, the right point and the left point of the notch tip were marked so that the loading point
displacements δ, and the crack tip opening displacements CTOD, can be specially gained. In
order to justify the investigated results of the crack propagation by the speckle interferometry
used in this study, twenty strain gauges were attached in front of the precast notch in the other
side of one specimen numbered with S-1, of which, the details can be seen in (Xu, 1988).
In the fracture test on each small size three-point bending beams, the speckle interfero-
metry is used to measure the in-plane displacement on a specimen surface. It requires double
Laser
Recorded plate with double

exposed speckle pattern Diffraction halo
Figure 2. The illustration of the point-by-point filtering set-up.
exposures of a photographic plate to the speckle intensities corresponding to the two states
of deformation of the specimen surface. The small size three-point bending beams were
loaded by a screwed loading frame on a vibration isolated table inside a darkroom. Each
specimen was loaded step-by-step increasingly until fracture. The specimen surface painted
by aluminum powder is illuminated with an expanded beam of laser light and is photographed,
of which the detailed arrangement was shown in (Xu, 1988), so that an appearance of speckles
of the specimen surface can be recorded on a photographic image. Before and after a load step
a photographic plate was exposed twice. The corresponding increase of deformation caused
by the increase of load will be recorded on the resulting negative because each pair of speckles
from the first and the second exposures on the resulting photographic plate encodes and
memories the deformation information of the corresponding point at the illuminated region on
the specimen surface before and after loading. Each time interval between two loading steps
was about twenty minutes so that the last photographed plate was developed and examined
for Young’s fringe formation. The amount of the next increase of load was adjusted according
to the number of the Young’s fringes formed in the last photographic plate. By the way, for
each specimen a number of photographic plates were obtained corresponding to the number
of loading intervals.
2.2. C RITICAL CRACK TIP OPENING DISPLACEMENT, CTODc
The point-by-point filtering set-up (shown in Figure 2) was used to visualize the speckle
displacement field which was frozen in the photographic emulsion.
The diffraction halo was formed in such an optical system of point-by-point filtration
through a recorded plate that is of double exposed speckle pattern. In the halo parallel equidistant
Young’s fringes were visualized due to interferometry of a pair of speckles from the first
and the second exposures. The direction of Young’s fringes is perpendicular to the speckle
displacement vector in the illuminated region. And the number of the Young’s fringes is
proportional to the in-plane displacement of the point. By the way, the measured load versus
initial crack tip opening displacement P-CTOD curves of all notched beams are drawn in
Figure 3(a). And the measured load versus loading point displacement (i.e., deflection) P -δ
curves are given in Figure 3(b). Corresponding to the maximum loads, the values of the critical
crack tip opening displacement CTODc are presented in Table 1.
It can be seen that the values of CTODc are size-independent which are directly measured
by laser speckle interferometry in the tests on the small size three-point bend beams. The
size-independence of the values of CTODc measured in our tests is in agreement with those
evaluated by the two-parameter model (KIcs and CTODc ) proposed by Jenq and Shah (1985).
The two-parameter model was based on the results of tests on small size three-point bending
beams. The range of specimen sizes used in our tests is similar to those used in (Jenq and
Shah, 1985). At least, in the small size range of the three-point bending beams, the size-
independence of CTODc evaluated by the two-parameter model that was claimed in (Jenq and
Shah, 1985) can be observed in our direct measurement.
Table 1. The results measured from the small size beams by laser speckle interferometry
Nos. of S×D×B a0 Pini Pmax CTODc δP 1ac

specimen (mm) (mm) (N) (N) (mm) (mm) (mm)
L-1 600 × 150 × 100 60 2402 4000 0.0207 0.169 53.8

L-2 600 × 150 × 100 60 2444 4158 0.0181 0.178 48.0
L-3 600 × 150 × 100 60 2500 4413 0.0177 0.162 33.0
Average 2449 4190 0.0188 0.170 45.0
S-1 400 × 100 × 100 40 1520 2435 0.0191 0.090 26.0

S-2 400 × 100 × 100 40 2285 3302 0.0176 0.109 29.4
Average 1903 2869 0.0184 0.100 27.7
Measured P-CTOD curves Measured P-δ curves

5000
5000
4000
4000
Load (N)
Load (N)
3000 3000
S-1 specimen
2000 S-1 specimen 2000
S-2 specimen S-2 specimen
L-1 specimen L-1 specimen
1000 1000
L-2 specimen L-2 specimen
L-3 specimen
L-3 specimen
0 0
0 5 10 15 20 25 0 50 100 150 200
CTOD (mm) Loading-point displacement, δ (mm)
(a) (b)
Figure 3. (a) P -CTOD curves and (b) P -δ curves measured from the three-point bend beams.
2.3. I NVESTIGATION OF THE CRACK PROPAGATION OR THE FRACTURE PROCESS ZONE
In our tests the main attention was focused on the investigation of the process of crack
propagation. In order to achieve this aim the technique of point-by-point filtering was used
to visualize the process of crack propagation and the shapes of fracture process zones in the
tested specimens.
It is well known that the displacement field of a body under action of an external load will
be continuous when no crack propagates in the body. Once a crack starts to propagate, the
displacement field in front of the propagating crack tip will be distorted. This characteristic
will be reflected on the speckle displacement field when the point-by-point analysis is used.
Due to the influence of cracking, the orientation and the period of the Young’s fringes of
diffraction halos in the illuminated regions at both sides of a new cracking zone will have
a sudden jump; the Young’s fringes of the diffraction halo at the new cracking zone will
be smeared. In our tests on the three-point bending notched beams, typical three types of
photographs of the Young’s fringes of the diffraction holes that were separately taken at the
Figure 4. The illustration of visualization of appeared crack by changed orientation and period of the Young’s
fringes of the diffraction halos: (a) the photographed diffraction halo pattern at the left side of the visualized
crack; (b) the halo pattern just at the trajectory of the visualized crack; (c) the halo pattern at the right side of the
visualized crack.
left side, the right side of a crack and just at the crack are presented in Figure 4. They are useful
for visualizing the path of crack propagation. It was found that there are two phenomena of
crack propagation. When the applied load is small, a single crack may appear. With increase
of the applied load, a cracking zone that is usually called a fracture process zone may start
to form. Therefore, the propagating path of a singe crack can be detected and positioned by a
jump in the orientation and the period of Young’s fringes. If a fracture process zone appears,
the boundary of its region can be detected and positioned by the smeared halo.
By the way above, the developing paths of the micro-cracks and the shapes of the fracture
process zones on the specimen surfaces of the tested beams were measured in details by the
point-by-point analysis and are shown in Figure 5 to Figure 9 for each specimen separately. In
order to save space of pictures, the whole section of a beam was not drawn in the Figures 5 to
9. The measured lengths of the fracture process zones corresponding to the recorded last step
loads are presented in Table 1 too. By the same method of the point-by-point analysis, Pflug
observed successfully the crack development of an unnotched concrete beam (Pflug, 1979)
and found that these observations by the speckle photography are in agreement with those
observed by the direct microscopic investigation.
From Figure 5. to Figure 9 it can be seen that there is an evident crack initiation and
stable propagation before unstable fracture occurs. The shapes of the fracture process zones
are irregular, long and narrow bands. And the average maximum width of the fracture process
zones measured in these specimens is 9.6 mm which is equal to about half of the maximum
aggregate particle size. The observed shapes and widths of the fracture process zones in our
investigations by the laser speckle interferometry are in conformity with those observed using
the moire interferometry by others (Du et al., 1989).
However, due to the step by step loading procedure, it could be expected that some of the
recorded last step load should be smaller than the real maximum load. It could lead to a larger
dispersion of the testing data like the maximum loads and the lengths of the fracture process
zones in the tests.
2.4. D ETERMINATION OF CRACK INITIATION POINT, Pini
From the results measured by strain gauges on the other side of the specimen S-1 (Xu, 1988), it
can be observed that when the applied load was small, the neutral axis of the net cross section
of the beam coincided with its horizontal central axis. And with increase of the applied load
Figure 5. The process of crack propagation and the development of fracture process zone in a surface of specimen
numbered with L-1 (length unit: mm)
-10 0 10 -10 0 10 -10 0 10 -10 0 10

40 40
30 30
20 20
10 10
0 0
-10 -10
(a) P=1373 N (b) P=1677 N (c) P=2226 N (d) P=3100 N

-10 0 10 -10 0 10 -10 0 10 -10 0 10
50 50
40 40
30 30
20 20
10 10
0 0
-10 -10
(e) P=3432 N (f) P=3677 N (g) P=3913 N (h) P=4158 N
-10 0 10 -10 0 10 -10 0 10 -10 0 10
40 40
30 30
20 20
10 10
0 0
-10 -10
(a) P=1834 N (b) P=2098 N (c) P=2500 N (d) P=2942 N

-10 0 10 -10 0 10 -10 0 10 -10 0 10
40 40
30 30
20 20
10 10
0 0
-10 -10
(e) P=3334 N (f) P=3726 N (g) P=4020 N (h) P=4413 N

-10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10

40 40
30 30
20 20
10 10
0 0
-10 -10
(a) P=637 (b) P=1050 (c) P=1520 (d) P=1990 (e) P=2435
numbered with S-1 (length unit: mm)
the neutral axis of the net cross section of the beam began to move upward and away from its
horizontal central axis. It was examined that the loading value at which the neutral axis of the
net cross section of the beam starts to move away the horizontal central axis of the net cross
section of the beam is the load value at the end point of the linear segment on the P -CTOD
curve of specimen S-1.
Although a very thin crack had appeared on the surface of the beam before the load was
achieved the end point of the linear segment of P -CTOD curve, the appearance of one thin
surface crack cannot be considered as that a substantial crack initiation has been occurred in
the beam because the surface of a concrete beam is easier to crack than the internal material of
the beam and such a fact that the length of the subcritical crack is different along the width of
the specimen had been observed using dye-penetrant technique (Bascoul et al., 1989). It was
-10 0 10 -10 0 10 -10 0 10 -10 0 10
40 40
30 30
20 20
10 10
0 0
-10 -10
(a) P=1137 N (b) P=1500 N (c) P=1853 N (d) P=2285 N

-10 0 10 -10 0 10 -10 0 10 -10 0 10
40 40
30 30
20 20
10 10
0 0
-10 -10
(e) P=2608 N (f) P=2814 N (g) P=3069 N (h) P=3305 N
numbered with S-2 (length unit: mm)
P max = 3.95 kN
A 3
4.0
4
3.5
2 1--- Load versus crack mouth opening

3.0 displacement curve (P-CMOD curve);
P (kN) 2--- Load versus loading point displacement
curve (P-d curve);
2.5 3—Load versus acoustic emission pulse rate
curve (P-dN/dt curve);
B 4—Load versus acoustic emission pile-up
2.0
pulse curve (P-N curve)
1
1. B x D x L = 15 x 15 75 cm
a = 6.75 cm
1.0
0.5
0.0 2 0.04 0.0 6 0.08 0.10 0.12 CMOD (mm)

0
0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 δ (mm)
Figure 10. The illustration of onset of the stable crack propagation and the critical unstable fracture in concrete
beam detected by an acoustic emission method: the first peak value of the pulse rate, point B, demonstrates the
onset of stable crack propagation and the maximum peak value of the pulse rate, point A, represents the onset of
critcal unstable fracture in concrete materials.
found that the appearance of a surface crack was earlier than an inside crack. And the length
of the surface crack was longer than of the inside crack. This means that one should measure
the lengths at a lot of measured points along the width of specimen, and then get an average
value of them. However, in normal practice it is not so easy.
According to our knowledge, most experimental investigations by many researchers mainly
concerned the measurement of the crack propagation on the specimen surface. And only a
few investigations dealt with inside crack propagation. After further comparing Figure 3 with
Figure 5 to Figure 9 it can be found that when only a single thin micro-crack appeared on the
surface of beam, the corresponding loading regions on the load-loading point displacement
curves (P -δ curves), or on the load-crack tip opening displacement curves (P -CTOD curves),
were within their linear segment regions. When the applied loads were beyond the end points
of the linear segments of either P -δ curves or P -CTOD curves, the fracture process zones
with a band shape began to appear. Combining the characteristics of either P -δ curves or P -
CTOD curves with the development plotting of the fracture process zones, the starting point
of nonlinear segment on a measured P -δ curve or P -CTOD curve can be taken as the initial
point of stable crack propagation aross the whole cross section of a tested specimen. This
point corresponds to the starting point of an observable fracture process zone with a band
shape. It is in agreement with the investigated results in the tests by John and Shah using a
crack gauge technique (John and Shah, 1986) and by Xu using an acoustic emission method
(Xu, 1982). In the investigation by John and Shah it was found that the starting point of the
nonlinear segment of the load-time curve corresponds to such a point of strain-time curve at
which the strain begins to rise rapidly. The same initial cracking phenomenon can be observed
in the tests by Xu using the acoustic emission method. The typical measured results by the
acoustic emission method are shown in Figure 10. It can be seen that there are two significant
peak values on the measured load versus acoustic emission pulse rate curve.
The smaller peak value (point B) that occurs at the starting point of the nonlinear segment
of P -CMOD curve means onset of the stable crack propagation. And the largest peak value
(point A) that occurs corresponding to the maximum loading point at P -CMOD curve means
onset of unstable fracture. From Figure 10 it can be seen that the P -CMOD curve is more
sensitive to the stable crack propagation than the P -δ curve. The reasons are that both the
stable crack propagation and the unstable fracture belong to a localized behaviour of strain
concentration in a structure and the crack mouth opening displacement is a local variable
which is mainly effected by the crack opening. Therefore, to describe the crack propagation
one may prefer to use P -CMOD curve (or P -CTOD curve). So, the starting point of the
nonlinear segment on the measured P -CMOD (or P -CTOD) curve can be taken as the onset
point of stable crack propagation. By this way, the load values corresponding to the crack
initiation point that are denoted with Pini are determined and presented in Table 1 for the
tested specimens.
3. Crack propagation in very large size CT-specimens observed by photoelastic coating
Since Kaplan investigated the stable crack propagation before unstable fracture by dye-pene-
trant method in his famous concrete fracture experiment in 1961 (Kaplan, 1961), more and
more researchers have studied the fracture process zone by various methods. It has been con-
firmed by enough experimental evidence that due to the heterogeneity inherent in cementitious
composite materials, there is a stable crack propagation before unstable fracture. Now, it has
been widely accepted that the linear elastic fracture mechanics can be adapted to concrete
structures when the influence of stable crack propagation is taken into account. Some fracture
parameters that are based on the stress intensity factor are determined by taking the length
of stable crack propagation into account (Jenq and Shah, 1985; Karihaloo and Nallathambi,
1990; Refai and Swartz, 1987). The fracture parameters are termed ‘critical stress intensity
factor, KIcs ’ in the two-parameter model (Jenq and Shah, 1985) and ‘fracture toughness, KIce ’
in the effective crack model (Karihaloo and Nallathambi, 1990; Refai and Swartz, 1987) re-
spectively. It is shown that the obtained values of KIcs and KIce are size-independent in a region
of small specimen size used in normal laboratories (Jenq and Shah, 1985; Karihaloo and
Nallathambi, 1990; Refai and Swartz, 1987; Karihaloo and Nallathambi, 1991). Thereupon,
RILEM has proposed a recommendation of the determination of concrete fracture parameters
of KIcs and CTODc (RILEM, 1990a).
In practice, the sizes of concrete structures are far larger than that used in tests in normal
laboratories. Specially, the sizes of some important concrete structures, like concrete dams
and prestressed reinforced concrete pressure vessels, are very large. Practical engineers may
take a great interest in understanding the behaviour of the crack propagation in large size
concrete specimens and may hope to know whether the characteristics of the fracture para-
meters obtained by small size concrete specimens are the same as those measured from enough
large size specimens. So, it is necessary to investigate the crack propagation in enough large
size concrete specimens. As we know, the published literature contains only few results of
observations dealing with the crack propagation in large size concrete specimens. In 1987,
the detailed investigations of the crack propagation in very large size CT-specimens, of which
the maximum plane dimensions are 3.6 × 3.0 m2 had been carried out in Dalian University of
Technology, but are not very well known because they were published in Chinese (Xu, 1988).
Later, experiments on wedge-splitting specimens (WS-specimens), of which the maximum
plane dimensions are 1.52 × 1.52 m2 were carried out to measure the fracture parameters of
dam concrete, without detailed investigation into crack propagation (Saouma et al., 1991). In
order to understand the crack propagation in very large size concrete specimens, this section
will present the main investigated results of crack propagation observed in the very large size
CT-specimens of concrete. The detailed descriptions are given in (Xu, 1988).
3.1. S PECIMEN PREPARATION
In order to observe the crack propagation in very large size specimen, the geometry of com-
pact tension specimen (CT-specimen) was utilized in the tests. The dimensions of the CT-
specimens, S × D × B (length × height × thickness) are 3000 × 3000 × 200 mm3 , 2500 ×
2500×200 mm3 , 2000×2000×200 m3 , 1500×1500×200 mm3 , 1000×1000×200 mm3 and
500 × 500 × 200 mm3 separately. The specimens were made with the same mix proportions
of 1:1.87:3.36:0.57 (cement: sand: coarse aggregate: water by wt). The maximum size of the
coarse aggregate which is crushed limestone is 20 mm and that of the river sand is 5 mm. All
CT-specimens were cast vertically in the laboratory hall so that they can be hosted easily for
preparations of the tests on them. For keeping the specimens in a same curing condition, the
specimens were cured by sealing up them with plastic membranes. The notch was cast using
a greased steel plate with thickness of 2 mm.
In the tests, two approaches of investigation were used to measure the crack propagation
on two sides of each specimen respectively. One approach is photoelastic coating that was
employed to observe the crack propagation on the west side of each specimen, denoted as
B-side. The other approach used strain gauges at the east side of each specimen, denoted as
A-side, for the investigation of the crack propagation and strain fields. The maximum number
of the strain gauges on one specimen was 172.
Because the test preparation for such large size specimens and for so many investigating
contents is quite complicated and needs much time, it is impossible that all CT-specimens
were tested in exactly the same age. In order to enable to neglect the influence of concrete age
to the measured results, the experiments were carried out after 180 days since the specimens
were cast. The longest age for the specimens is 360 days. The measured cube compressive
strength, tensile strength, modulus of elastic and Poisson’s ratio corresponding to the concrete
age of 360 days are 38.1 MPa, 2.78 MPa, 32.4 GPa and 0.197, respectively.
3.2. P REPARATION AND APPLICATION OF COATINGS
The epoxy-based materials included dibutyl phthalate and diethylenetriamine with the given
mix proportions were chosen as the coating materials. The corresponding properties of the
coating materials are as follows: the modulus of elasticity Ec , is 4.54 GPa; the Poisson’s ratio
νc is 0.39; the optical strain coefficient k = λ/fε is equal to 0.104 and the material fringe value
in the terms of strain fε , is 700×10−5 mm/fringe. It can be seen that the physical properties of
the coating materials can basically meet the following requirements for this research purpose
(Zandman et al., 1977):
(1) a high optical sensitivity coefficient k to maximize the number of fringes per unit strain;
(2) a low modulus of elasticity Ec to minimize reinforcing effects;
(3) high creep resistance to maximize stability of the fringe indication with time;
(4) linear strain-optic response for ease of data-reduction;
(5) a good bondability to ensure perfect strain transmission at the interface between coating
and specimen;
(6) freedom from initial birefringence to avoid correction for this condition;
(7) good machinability for ease in fabrication complex configuration.
The sheets of coating were cast in a spacious template on which a square frame of 200 mm×
200 mm was rounded by trips of silicone rubber with the desired thickness of 2.7 mm. After
a sheet of coating was cast for 24 hours it was removed from the template and then it was
bonded on the cleaned and smooth surface of a CT-specimen.
For gaining a perfect bond between the coating and the specimen, the same materials used
in the coatings were chosen to be as the materials of adhesive so that both materials of the
coatings and the adhesive have a same strength. In order to increase the reflectivity of the
investigated surfaces of the specimens, the aluminium powder with 15 percent total amount
of the adhesive was added into the adhesive. Before binding, an area with 200 × 200 mm2 in
front of the notch tip in each a specimen was sanded to get a very good level-off. Then the area
to have been sanded was carefully cleaned with ethyl alcohol. Once the adhesive was spread
evenly over the surface of a specimen, the sheet of coating is applied. The procedure to apply
a coating to the surface of a specimen is that the coating was, firstly, positioned at one end
of the position and slowly rotated, then, pressed into position for working off the air bubbles
and heated by an infrared ray lamp for achieving a perfect bond between the coating and the
specimen. After 48 hours it was bonded, the experiment was carried out.
3.3. T EST PROCEDURE
The load was applied to the CT-specimens by using a self-designed loading system with
two symmetrical synchronous hydraulic jacks. The displacement at the loading point was
measured by a clip-gauge. The information of CMOD and the strain values measured from 30
measuring points of strain gauges as well as the corresponding loading values were collected
by a computer-controlled 32 channel data-acquisition system. The system was programmed
to deal with the data. The testing set-up and the data-acquisition procedure are illustrated in
Figure 11.
Hydraulic jack Applying load
Load
Load cell
Hydraulics recording
instrument
Precast notch
Coating sheet
with 200x200 mm² CT-specimen
Figure 11a. The illustration of the photoelastic coating to be attached to B-site of CT-specimen and the testing
set-up in the compact tension fracture tests.
Hydraulic jack
Clip-gage
Computer
controlled 32 channel
Load cell
Hydraulics data-acquisition
Strain gages
system
For CT-6-1specimen with
dimensions of 3.6x3.0x0.2 meters,
test date: Dec. 1,1987
Figure 11b. The illustration of strain gauges measure on A-site of CT-specimen, testing set-up and data-acquisition
system in the compact tension fracture tests.
The photograph of testing ground of a CT-specimen is shown in Figure 12. The two loading
holes were made by placing 2 mm thick steel tubes with 89 mm diameters and were reinforced
by steel bars around them in each specimen.
The self-weight of specimen was supported simply by two round steel bars under the
bottom of it. On A-side of each specimen, a amount of strain gauges that the maximum
number is 172 and the minimum number is 72 were attached to around the precast notch tip
to observe the crack propagation and to measure the strain field near the notch tip. The strain
gauges in front of the notch tip were mainly used for investigating to the crack propagation
during loading. For each specimen, the data measured from 30 strain gauges that were placed
at within a circle with a radius of 50 mm from the notch tip and at the extension line of
the axis along the notch were collected and put out by the computer-controlled 32 channel
data-acquisition systems. Other strain gauge readings were recorded manually by two strain
Figure 12. The photography of fracture test on large size CT-specimen (before test for CT-6-1 specimen).
indications with 80 measuring points. The loading was applied in increments and paused at
every load level for reading and recording manually the test data. On B-side of each specimen,
a photoelastic coating with the dimensions of 200 × 200 mm2 was attached to in front of the
notch tip in order to measure directly the distribution of strain field and the trajectory of stable
crack propagation before unstable fracture occurs. The position of coating is illustrated in
Figure 11(a) too. The fringe patterns of the photoelastic coatings were observed by a reflection
polariscope and taken by a portable videorecorder and a camera. The used optical system is
shown in Figure 13.
Because the loading system used in the tests was a symmetrical synchronous hydraulic
jack system, two same 100 kN cells that are suitable for the both loading cases of tension and
compression were filled into the loading system. One of the two cells was for the computer-
controlled 32 channel data-acquisition system to collect loading data. The another one is
recorded by an electronic load recording instrument to provide the loading readings for the
photoelastic coating investigation. At the same time of the loading pause to record the readings
of the strain gauges, the fringe-order and the fringe length appeared in the photoelastic coating
are observed by the reflection polariscope and recorded manually. And the fringe patterns
of the photoelastic coating were taken by the videorecorder and the camera so that detailed
analysis for those fringe patterns can be carried out after the experiment. According to the
purpose of our experiments and the configurations of the specimen and the loading, only
normal incidence investigation was arranged to gain the isochromatic-fringe pattern to reflect
the difference of principal strain, ε1 -ε2 . The oblique-incidence was not needed.
In our experiments the evident phenomena of the stable crack propagation before unstable
fracture were seen with the naked eye on A-sides of the specimens numbered with CT-3-1 and
CT-2-3 too. The maximum width of the crack propagation at the precast notch tips is 0.1 mm,
and the tips of the stable propagating cracks only was visible to the visualization. The lengths
of the stable propagating cracks corresponding to the different loading levels were recorded.
3.4. T HE INVESTIGATED RESULTS OF THE STABLE CRACK PROPAGATION IN THE LARGE

SIZE CT- SPECIMENS
Because the information about the stable crack propagation in very large size specimens
in published literature is scarce, the main attention in our experiments was focused on the
Specimen
Reflection Coating
surface sheet
QA _ Q P
A P
Camera Reflection
polariscope
Figure 13. An optical system with V -type used in the photoelastic coating investigation.
investigation into the fracture process that includes the crack initiation and the stable crack
propagation. Combining the results of investigations by means of the photoelastic coating
with those measured using the strain gauges and through the directly seeing with the naked
eye, detailed data of lengths of the stable crack propagation and the corresponding loads for
specimens numbered with CT-6-1, CT-6-2, CT-3-1 and CT-2-3 are obtained. As a result, the
crack initiation, the stable crack propagation and the critical crack fracture can be indicated.
For other specimens only the critical lengths of the crack propagation corresponding to the
maximum loads were measured.
In the investigations by the photoelastic coating the monochromatic fringe patterns for each
tested specimen at every loading level were taken. The non-uniform fringes (V -type) indicate
the presence of a crack (Zandman et al., 1977). The investigation reported in (Zandman et al.,
1977) showed that not only a crack in concrete can be detected by this method before it can
be observed with a microscope but also a crack which will appear at a point in a concrete
specimen can be predicted by focusing continuously on the point. The photoelastic patterns to
reveal the crack initiation, the stable crack propagation and the critical fracture corresponding
to the various loading levels were shown in Figure 14(a) to Figure 14(g), where, Figure 14(g)
indicates the trajectory of the crack propagation after unstable fracture occurred. They were
taken from the specimen numbered with CT-6-1. The dimensions (S × D × B) of it are 3000 ×
3000 × 200 mm3 . The coating had no precast notch. The upper edge of the coating was closely
positioned the precast notch tip of the specimen. It was found that the trajectory of practical
failure shown in Figure 14(g) was coincided with the trend of V -type fringes of the patterns
shown in Figure 14(a) to Figure 14(f) that revealed the crack initiation and the propagation
process. Therefore, it is confirmed that the photoelastic coatings can be used to detect and to
follow the crack propagation in concrete specimens.
According to the experimental results in (Xu, 1988), the value of 0.25% can be taken as the
limit tensile strain value for the concrete used in the experiments. Inserting the corresponding
parameters into the stress-optic law to consider the effects of coating thickness and Poisson’s
ratio mismatch, the fringe-order n of the coating corresponding to the crack initiation can be
determined as follows:
2(1 + νs )tc ε1s

n = h i
1 + ttcs EEcs (1+ν
(1+νs )
c)
fε
Figure 14. The photographs of photoelastic coating pattern taken in the test on CT-6-1 specimen (a) P = 0.515
Pmax ; (b) P = 0.615 Pmax ; (c) P = 0.704 Pmax ; (d) P = 0.778 Pmax ; (e) P = 0.859 Pmax ; (f) P = 0.926
Pmax ; (g) The crack trajectory after unstable fracture occurs.
Measured P-∆a curves

100
80
Load P (kN)
60
40
CT-2-3
20 CT-3-1
CT-6-1
CT-6-2
0
0 50 100 150 200 250
Crack extension ∆a (mm)

Figure 15. The relations of the load P versus crack extension 1a measured on large CT-specimens.
2 × (1 + 0.197) × 2.7 × 250 × 10−6

= h i = 0.23
2.7×4.54×103 (1+0.197)
1 + 200×32.4×10 3 (1+0.390) × 700 × 10−5
where, subscripts c and s refer to the coating and the specimen respectively, t denotes the
thickness of both the coating and the specimen and fε is the coating-material-fringe value in
the terms of strain.
According to the coloured fringe sequence (Zandman et al., 1977), when the coloured-
fringe order n is equal to 0.23, the corresponding colour is between black and grey and is very
close to grey (the fringe order for the black is 0 and the one for the grey is 0.28). Such, it can
be regarded as that once the colour on the coatings changed into the grey from the black, it
means the corresponding observed point in the specimen starts to crack. The position of the
tip of a grey fringe with V -type just is the one of the tip of cracking extension occurred in the
specimen. Therefore the lengths of crack propagation corresponding to the different loading
levels for several specimens can be measured in the observation by the photoelastic coatings
using the reflection polariscope directly.
The lengths of crack propagation on both sides of the specimens by two approaches of
the photoelastic coatings and the visualization are presented in Table 2. It can be seen that
Table 2. The lengths of the crack propagation on both sides of CT-3-1 and CT-2-3 specimens
measured by the photoelastic coating and the visualization separately
Nos. of loading level The measured lengths of the stable crack propagation
specimens
P P /Pmax by photoelastic coating by visualization
(kN) (mm) (mm)
CT-3-1 51.97 0.866 65 65

54.91 0.915 85 100
57.95 0.966 110
58.05 0.967 127 130
58.64 0.977 155 164
60.01 1.000 207
CT-2-3 23.42 0.752 50 65

27.04 0.868 110 110
30.61 0.983 193 160
30.90 0.992 195 180
31.08 0.998 200 200
31.14 1.000 210 210
Table 3. The results of Pini , Pmax , 1ac measured from the four large size CT-specimens.
Nos. of specimens Dimensions of specimens a0 Pini Pmax 1ac

S × D × B (mm) (mm) (kN) (kN) (mm)
(1) (2) (3) (4) (5) (6)
CT-6-1 3000 × 3000 × 200 1200 20.0 84.68 210

CT-6-2 3000 × 3000 × 200 1200 20.0 93.21 200
CT-3-1 1500 × 1500 × 200 600 14.0 60.01 207
CT-2-3 1000 × 1000 × 200 400 11.5 32.29 210
the results of observation for the lengths of the crack propagation using the two observing
approaches are in approximate agreement.
Moreover, the crack propagating processes on A-sides of several specimens were measured
also by distinguishing the changing trends of strain fields and the load-strain curves measured
at the measured points. The detailed determination procedure of the crack propagation through
the measured strain fields and the load-strain curves were described in (Xu, 1988). Due to the
limited space in this paper it is impossible to present the detailed procedure.
For CT-specimens numbered with CT-6-1, CT-6-2, CT-3-1 and CT-2-3 the curves of the
load versus crack extension (P -1a curves) are given in Figure 15 which were measured by
above mentioned three approaches. The measured crack initiation points Pini , the maximum
loads Pmax , and the critical crack propagation lengths 1ac , for those four specimens are
presented in Table 3.
3.5. T HE STRAIN FIELD IN FRONT OF NOTCH FOR A LARGE SIZE CT- SPECIMEN
The strain fields on the two sides (A-side and B-side) of specimens are measured using the
photoelastic coatings and the strain gauges. Here, the strain fields measured from the two sides
of the specimen CT-6-1 is presented as the examples. First, the distribution of differences of
principal strains, ε1 -ε2 , for different loading levels measured on B-side of the large size speci-
men CT-6-1 by the photoelastic coating are presented. Then, as comparison, the distribution of
the principal strain field along the circle direction, ε0 , near by the precast notch tip measured
on the A-side of the same specimen by the strain gauges are shown too.
Since a coating can deform together with the surface of a specimen where the coating
is attached under the condition of perfect bond between the coating and the specimen, the
strain field in the surface of the specimen can be reflected on the coating. Considering the
situations in our tests that the plane dimensions of the CT-specimens are far larger than the
thickness of them, the stress state of the CT-specimens under loading is plane stress one. So it
is hardly to appear strain variation through the coating thickness (Zandman, 1977). According
to the method in (Zandman, 1977), when the reinforcing effect caused by the coating and the
influence of the Poisson’s ratio mismatch are taken into account, the relation to convert the
strains measured in the coated specimen (ε1c -ε2c ) to the strains ε1u -ε2u that would be developed
in the specimen without a coating can be expressed as the following form:
1
ε1u − ε2u = (ε1c − ε2c ),
c1
where
1 tc Ec (1 + νs )
=1+
c1 ts Es (1 + νc )
is a correction factor.
Inserting the thickness and the material parameters of the coatings and the specimens into
the expression of the correction factor 1/c1 , it can be got the value of 1/c1 as follows
1 2.7 × 4.54 × 103 (1 + 0.197)

=1+ = 1.0016.
c1 200 × 32.4 × 103 (1 + 0.390)
It can be seen that the influences of the coating reinforcement and the Poisson’s ratio
mismatch is less than 0.16 percent and can be neglected. Therefore, the strains (ε1c -ε2c ) to
be reflected on the coating are equal to the strains, ε1s -ε2s measured on the surface of the
specimen (Zandman, 1977).
nfε
ε1s − ε2s = ε1c − ε2c = .
2tc
Then, to insert tc = 2.7 mm and fε = 700 × 10−5 mm/fringe into the expression above, we
have
ε1s − ε2s = 700 × 10−5 n/(2 × 2.7) = 1297 × 10−6 n

Therefore one order of the fringe represents the values of strains, ε1s -ε2s , of approximate
1300 × 10−6 . For the specimen CT-6-1 the measured distribution of the principal strain ε1s -ε2s
corresponding to the different typical loading levels are presented in Figure 16. All of them
are of V -type patterns. They well represent the process of crack propagation and are basically
similar to those observed by the laser speckle interferometry for the small three-point bending
beams.
The information of the strain field measured directly using an amount of strain gauges on
very large size specimens will be beneficial for understanding the fracture properties, crack
propagation and fracture process zone in practical concrete structures. In the experiments re-
ported in (Xu, 1988) such information has been got. The measure approaches and the detailed
measured data of strain fields for the large size CT-specimens were shown in (Xu, 1988). As a
comparison with the investigated results obtained by the photoelastic coatings, the strain fields
measured by attaching an amount of strain gauges on A-side of CT-specimen numbered with
CT-6-1 at two typical loading levels were presented in Figure 17.
From Figure 17(a) it can be seen that when the applied load is not too large, the strain
concentration zone is positioned in a small region nearby the precast notch tip. The shape of
the strain concentration zone is an irregular, long and narrow band. The micro cracking zone
and the stable crack propagation will occur within the strain concentration zone. The pattern
of the high strain concentrating zone shown in Figure 17(a) measured in A-side of CT-6-1
specimen through using the strain gauges is similar to that measured on B-side of the same
specimen by the photoelastic coatings shown in Figure 16. And the shape of the high strain
concentrating zone in Figure 17(a) is also similar to the shapes of the fracture process zones
visualized by the laser speckle interferometry on the small size three-point bending beams
shown in Figure 5 to Figure 9.
In Figure 17(b) the corresponding load achieved at 0.95 Pmax before unstable fracture oc-
curs. Because the main crack propagates steadily, the strain values measured at the measuring
points which are along the extension line of the precast notch in the plotted region in Fig-
ure 17(b) have exceeded the limit values of the strain gauges attached. Those strain gauges are
made with glue (colloid)-matrix, their limit values can be over 10, 000 µ. Since new cracking
happens, the corresponding restrain was free. It engaged the stresses to relax on the both sides
of the new propagated crack to. The strain values on the regions in the both sides of the new
propagated crack will rapidly drop down. It leads that stresses in some regions are close to
zero.
4. Experimental evaluation of the initiation toughness KIcini , and the unstable fracture
toughness KIcun , for the crack propagation in concrete fracture
4.1. T HE EXPERIMENTAL EVALUATION OF THE DOUBLE -K FRACTURE PARAMETERS

KIcini AND KIcun
The observable evident obtained in the investigations of the crack propagation by the differ-
ent methods above showed that there are three different cases: crack initiation, stable crack
propagation and unstable fracture in quasi-brittle fracture. The stable crack propagation has
been widely noted and its influence on the fracture characteristics of materials has been recog-
nized well too. As the result, the critical length of the stable crack propagation corresponding
to the maximum load Pmax has been taken into account evaluation of the critical stress intensity
factors (Jenq and Shah, 1985; Karihaloo and Nallathambi, 1990; Refai and Swartz, 1987).
Figure 16. The distributions of principal strain difference, ε1s = ε2s , at varied loading levels measured from
CT-6-1 specimen by the photoelastic coating.
However, the crack initiation has not been considered in the evaluation of crack propagation in
concrete structures. For some concrete structures to protect the environment against pollution,
the exact evaluation of the crack initiation are more important. Even for some normal concrete
structures in some special cases, the evaluation of the crack initiation cannot be neglected
also. In order to evaluate the crack propagation in practical structures, it is necessary to
set up a suitable evaluation criterion and to measure the corresponding fracture parameters
defined in the criterion. Such a criterion is required to enable both to evaluate the safety of
practical structures with cracks adequately and to explain reasonably the phenomena observed
in experiments. The corresponding fracture parameters defined in the criterion are required to
60 40 20 0 10 mm Unit of length: mm
60 40 20 0 10 mm Unit of length: mm
40
40
Precast notch
0 20
62 20
20 2
5 20
51 10 35
85 132 105 15 12252 -1 8
98 61 0
0 15 5 441 12514 346 15 35 30 30 12 12252 12512 159 59 30 30
25
0
_ 29
64 12309 9
329 73 20 30
20 25 13 30
30 30 0
140 118 11020
25 264 20 20 30
20 30 25
20 -20 1
134 105 1512 15
25 40
40 134 35 -3
172 -30 16
98 11291 -12
167 59
271
60
60
-20
25 -65
10 28
0 11903
162 _ 80
80
-10
30
-50
20
100 13231
100
176 50
_ The trajectory of stable
crack propagation
120 Legends:
120 -- Position of strain gauge
Legends: 62 -- Strain value (µ )
-- Position of strain gauge
62 -- Strain valu (µ)
140
140 -10
20 -80
22 12910
31
160
160
(a) (b)
Figure 17. (a) A strain field at loading level of 0.344 Pmax , (b) strain distribution and trajectory of crack
propagation at loading level of 0.95 Pmax measured on the A-side of CT-6-1 specimen.
be useful in the engineering application and to be measurable in the experiments as well as to

be reasonable in the definitions. The characteristics of them should be only dependent on the
material properties.
In order to describe the three different situations during the crack propagation until the
critical fracture occurs, we attempt to define the initial situation and the critical situation of
the crack propagation using the terms of the stress intensity factor. At the initial situation of
the crack propagation, the crack does not start to propagate. The initial crack length is a0 .
Once the crack starts to propagate, the corresponding load is called initial cracking load Pini .
Accordingly, the initial cracking stress intensity factor created at the initial crack tip by the
initial cracking load Pini , is defined as the initiation toughness KIcini . The critical situation of
the crack propagation means that the crack propagates to its critical length ac , at which the
load arrives at its maximum value Pmax . Correspondingly, the critical stress intensity factor
created at the critical crack tip by the maximum load Pmax , is defined as the unstable fracture
toughness KIcun .
In fact, the initiation toughness KIcini is the inherent toughness of a material. It implies that
a crack does not propagate when the stress intensity factor at the initial crack tip is less than
the inherent toughness, i.e., initiation toughness KIcini . From Figure 3, it can be seen that the
initial cracking load Pini , is at the end of the linear segment of the measured P -CTOD or P -δ
curves. So, the initiation toughness KIcini , can be evaluated using a formulae developed from
linear elastic fracture mechanics.
For the three-point bending notched beams, the calculation of the initiation toughness KIcini ,
is performed using the following formulae (Tada et al., 1985)
PS √
K = 1.5 aF (α), (1)
BD 2
where P is the applied load; S is the beam length; B is the beam thickness; D is the beam
depth; a is the crack length and α = a/D.
1.99 − α(1 − α)(2.15 − 3.93α + 2.7α 2 )

F (α) = . (2)
(1 + 2α)(1 − α)3/2
For CT-specimens used in the experiments a formulae to calculate the stress intensity factor
is developed as follows:
P
K = 3.675 √ [1 − 0.12(α − 0.45)](1 − α)−3/2 (0.2 6 α 6 0.8). (3)
B D
The maximum errors between obtained by Equation (3) and the accurate results by boundary
collocations method in (Srawley and Gross, 1972) is less than 2.0 percent. It can meet the
accuracy required in practical engineering calculations.
Similarly, the unstable fracture toughness KIcun , can be evaluated by inserting the critical
crack length ac , and the maximum load Pmax , into the formulae above. This approach is the
same as that in the published literature, for example, in (Jenq and Shah, 1985; Karihaloo and
Nallathambi, 1990; Refai and Swartz, 1987) and in the recommended method of the determ-
ination of fracture parameters (KIcs and CTODc ) by RILEM (1990a). In the determination
method for the critical stress intensity factor of plain concrete using three-point bend tests
recommended by RILEM (1990a), the same expressions (1) and (2) for calculating the stress
intensity factor were proposed.
By inserting the measured Pini , a0 and Pmax , ac presented in Table 1 and Table 3 for the
small size three-point bending beams and the large size CT-specimens into the formulae
(1) and (3) respectively, the corresponding values of initiation toughness KIcini , and unstable
fracture toughness, KIcun , can be determined and are presented in Table 4. From Table 4 it
can be seen that the values of initiation toughness KIcini and unstable fracture toughness KIcun
are independent of the specimen sizes in the tested ranges. Since the physical properties,
like Ec , ft , fc , of the concrete materials used in the tests on the two geometry specimens are
very close, the values of KIcini and KIcun for the two geometry specimens are quite close too,
respectively. Due to the influence of the loading procedure in the laser speckle interferometry,
the testing data of the recorded maximum loads and the length of the fracture process zones
could have a larger dispersion. As the results, the evaluated values of the critical unstable
fracture toughness KIcun , for the tested three-point bending notched beams will have much
larger dispersion too.
In the experiments on the large size CT-specimens, the detailed data of the crack propaga-
tion were measured from four CT-specimens with different dimensions. They include the
initial cracking point, the several lengths of the stable crack propagation corresponding to
the several loading levels and the critical crack length at the maximum load. For other CT-
specimens, only the critical crack length ac , and the maximum load Pmax were measured.
The values of unstable fracture toughness KIcun , measured from all CT-specimens tested were
presented in (Xu, 1988).
To compare the values of KIcun presented in Table 4 measured from the four CT-specimens
with three different dimensions with those presented in (Xu, 1988) measured from 11 CT-
specimens with five different dimensions, the average value of KIcun in Table 4 collected
from the four CT-specimens that is 2.525 MPa m1/2 is very close to the average value of
2.556 MPa m1/2 collected from 11 CT-specimens shown in (Xu, 1988). It shows that the results
shown in Table 4 fall into the normal range of this type of concrete materials tested.
From Table 4, it can be found that KIcini and KIcun collected from the small size three-point
bending beams and the large size CT-specimens are only dependent on the properties of the
tested materials, not dependent on the specimen geometry and size. It manifests such a fact
that in the crack propagation process in concrete fracture there exists two size-independent
controlling fracture parameters KIcini and KIcun . One of them controls the crack initiation and
the other controls the unstable fracture. Because the critical crack lengths ac to be used to
calculate the value of KIcun are directly measured, the obtained values of KIcun can be regarded
as to be directly measured too.
Such information of KIcun measured directly from the small size three-point bending beams
and the very large size CT-specimens provides a justification for the determination of KIcs
recommended by RILEM (1990a). In the work of Jeng and Shah (1985) that was used as a
basis of the recommendation method for the two fracture parameters (KIcs and CTODc ) by
RILEM (1990a), the critical crack length ac , to be used to evaluate the critical fracture tough-
ness KIcs , is indirectly evaluated using the unloading compliance cu , not directly measured
in the experiments. So, the value of KIcs proposed by Jeng and Shah (1985) and the recom-
mendation of RILEM (1990a) is not obtained by a direct measurement. Because the definition
of KIcun proposed in this paper is the same as that of KIcs proposed by Jeng and Shah, the
directly measured results of KIcun presented in this paper provide a good confirmation for the
determination method of KIcs originally proposed by Jeng and Shah (1985) and recommended
by RILEM(1990a).
Of course, one may expect more tested data and the information of KIcini and KIcun measured
directly from the large size specimens. We can see that it is not easy to carry out so large
size CT-specimens and to measure directly the whole process of crack propagation. If inside
cracking in the testing specimens is considered, the obtained values of KIcini in Table 4 may be
unterestimated and those of KIcun may be overestimated.
4.2. A PRIMARY COMPARISON WITH THE ANALYTICAL SOLUTIONS OF THE FICTITIOUS

CRACK IN THE G RIFFITH CRACK AND IN THE SEMI - INFINITE CRACK
We have solved the analytical solutions of the fictitious crack at various loading stages for
both Griffith crack and semi-infinite crack problems. Due to the limit of space, all expressions
of them cannot be presented herein. However, a primary comparison between the results
measured and those predicted by the analytical solutions at the critical situation for both crack
problems will be made.
Until the critical situation of unstable fracture, the toughness of a loaded body increases
from the value of KIcini to the one of KIcun due to stable crack propagation. The toughness
increase during the stable crack propagation is only due to the cohesive force on the fictitious
crack. This contribution at the critical situation is called KIcc . This leads to
KIcini + KIcc = KIcun . (4)
Using the solutions for the two crack problems, the expressions of KIcc for them can be got.
ini and K un .
Table 4. The measured values of KIc Ic
Geometry of Nos. of Dimensions of Cube strength Tensile Modulus ini

KIc un
KIc
specimens specimens specimens of compression strength of elastic
S×D×B fcu ft Ec (MPa m1/2 ) (MPa m1/2 )
(mm) (MPa) (MPa) (GPa)
Three-point bend beams L-1 600 × 150 × 100 0.513 3.454

L-2 600 × 150 × 100 0.522 2.860
L-3 600 × 150 × 100 35.2 2.45 31.0 0.533 1.904
S-1 400 × 100 × 100 0.393 1.518
S-2 400 × 100 × 100 0.585 2.405
Average 0.509 2.428
Compact tension specimens CT-6-1 3000 × 3000 × 200 0.459 2.323

CT-6-2 3000 × 3000 × 200 0.459 2.534
39.14 2.79 32.4
CT-3-1 1500 × 1500 × 200 0.455 2.837
CT-2-3 1000 × 1000 × 200 0.457 2.404
Average 0.458 2.525

y y
ft ft ft
σs(CTODc) σs(CTODc)
σs(CTODc)
CTOD c
CTODc
0 x 0 x
2a0
∆ac
2ac
(a) (b)
Figure 18. The cohesive force distributions on the fictitious crack zones at the critical situation: (a) Griffith crack
problem; and (b) Semi-infinite crack problem.
Figure 19. The comparison between the results of formulae (5) and (6): (a) for beam specimen; and (b) for
CT-specimen.
The cohesive force distribution on the fictitious crack zones in the Griffith crack is illus-
trated in Figure 18(a) and the one in the semi-infinite crack is in Figure 18(b).
Therefore, for the case of Griffith crack the expression of KIcc is solved as follows:
2ft p p
KIcc = ac /π{(β − a0 /ac ) arccos(a0 /ac ) + (1 − β) 1 − (a0 /ac )2 } (5)
1 − a0 /ac
and for the case of semi-infinite crack we have:
2p p
KIcc = 2/π (2 + β)ft 1ac , (6)
3
where ft is the tensile strength of the tested material a0 is the initial crack length ac is the
critical crack length and 1ac is the critical crack extension length, β is the ratio of the
value of cohesive force at the initial crack tip σs (CTODc ), to the tensile strength ft , i.e.,
β = σs (CTODc )/ft .
If the values of KIcc expressed by formulae (5) and (6) are independent of the geometry and
sizes of specimens, it is confirmed that the double-K fracture parameters KIcini and KIcun are
independent of the geometry and sizes too. It implies two meanings. There is no significant
difference between values of KIcc calculated from the formulae (5) and (6) and they can apply
to other geometries when their application conditions can be met.
Now, a comparison is made to investigate whether there is a significant difference between
values of KIcc calculated from the formula (5) and (6) when they apply to both three-point
bending beam geometry and CT-specimen geometry. The two examples in the comparison
are given in Figure 19. One is a three-point bending beam with parameters of D = 400 mm,
a0 = 120 mm, 1ac = 40 mm, ft = 2.58 MPa and β = 0.163. Another one is a larger size
CT-specimen. The corresponding parameters used in the comparison for formulae (5) and (6)
are: D = 3000 mm, a0 = 1200 mm, 1ac = 200 mm, ft = 2.79 MPa and β = 0.183. The
curves calculated by formulae (5) and (6) are presented in Figure 19(a) for the beam specimen
and in Figure 19(b) for the CT-specimen where the crack extension amount 1a increases from
zero until its critical value 1ac . More examples in the comparison are presented in Table 5.
The corresponding relative errors between the values of KIcc evaluated by formulae (5) and
(6) have been checked. The maximum relative errors between values of KIcc evaluated are
less than 2.0 percent. Considering the aim of practical application, the relative errors between
the values of KIcc could be neglected for the considered examples. So, the formula (6) can
be taken as a very good approximation and an alternative of the formula (5) due to the very
simple expression of the formula (6).
Large size CT-specimens and their long notches could allow to apply the formula (6). As
the examples, the values of KIcc evaluated by formulae (6) are presented in Table 5. To make a
comparison between the results calculated using the two formulae for the practical specimens,
the corresponding results obtained by the formula (5) is shown in the same table too.
Both Figure 19 and Table 5 show that there is no significant difference between the cohe-
sion toughness KIcc due to the same cohesive force distribution on the same length fictitious
cracks evaluated by the two formulae (5) and (6) for the two different crack problems re-
spectively. So, for a same concrete its critical length of effective crack should be the same.
In such a case, it has been proven that the cohesion toughness KIcc is not dependent on the
specimen geometry. In addition, from Table 5 it can be observed that the values of KIcc are
size-independent.
However, it can be found that the values of KIcc evaluated by both formulae (5) and (6) are
less than the differences between KIcun and KIcini presented in Table 4 that are directly measured
by the used approaches. Theoretically speaking, the values of KIcc predicted by the formula (5)
or (6) should be equal to those measured directly by the approaches. There are two causes of
the deviation of the results predicted by the analytical solutions from those measured directly
by the used approaches. The first cause is that configurations of both the Griffith crack and
the semi-infinite crack problems studied herein are not the same as those of the specimens
tested. The second one could be explained by the fact that the used approaches only can
measure the crack propagation on the specimen surface. Since the surface crack of a specimen
develops earlier than the inside crack of the specimen, it leads to an underestimate for KIcini
and an overestimate for KIcun . As the results, it could lead to an overestimate for the difference
between KIcun and KIcini .
5. KR -resistance curve for the crack propagation in concrete fracture
In order to describe the slow crack growth in the middle-lower strength metal materials, the
current concept of the so-called crack growth resistance curve (R-curve) uses a unique rela-
tionship between the amount a crack growth and the applied stress intensity factor in which the
crack growth resistance R increases with crack length. This concept was proposed by Irwin in
1960 (ASTM, 1960), instead of the original concept in 1954 by Irwin and Kies, in which the
crack growth resistance R remains constant or decreases with crack length (Irwin and Kies,
1954). When the crack growth resistance R is termed the crack driving force G, it is called
GR -curve with unit of N/m. And when it is termed the applied stress intensity factor K, with
c by equations (5) and (6).
Table 5. The comparison between values of KIc
Nos. of Length Depth Thickness Initial Crack β c

KIc c
KIc relative
specs. S D B crack extension (Mpa m1/2 ) (Mpa m1/2 ) errors
(mm) (mm) (mm) length, a0 1ac by equation (5) by equation (6) (9)/(8)
(mm) (mm)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
CT-6-1 3000 3000 200 1200 200 0.191 1.463 1.454 0.994
CT-6-2 3000 3000 200 1200 200 0.189 1.465 1.451 0.990
CT-5-2 2500 2500 200 1000 210 0.212 1.520 1.504 0.990
CT-4-3 2000 2000 200 800 200 0.168 1.456 1.439 0.988
CT-3-1 1500 1500 200 600 210 0.164 1.495 1.472 0.985
CT-2-3 1000 1000 200 400 210 0.168 1.501 1.469 0.980
Crack driving force curves
Stress intensity factor, K

K c t , onset of unstable
Crack extension
crack extension resistance curve
K(P max , a c )
K c ini , initiation of
crack extension
K(P ini , a 0 )
a0 ac
Crack length, a
Figure 20. The illustration of KR -curve.
unit of MPa m1/2 , it is called KR -curve. The later, KR -curve was often used. In studies for the
stable crack propagation in concrete materials and fibre reinforced cementitious composite,
R-curve have been employed too (Karihaloo, 1989; Mai, 1984).
According to the definition of the KR -curve, it can be seen that the KR -curve is the tra-
jectory of the stress intensity factor created by the extra load at the tip of a growing crack. It
is usually regarded as a function of the crack extension 1a or a function of the instantaneous
crack length a = a0 + 1a, i.e.,
KR = KR (1a) or KR = KR (a).
If a relationship between the applied extra load and the amount a crack growth is measured
in a test, the KR -curve is only dependent on the properties of the tested materials, not depend-
ent on the specimen geometry and size, KR -curve can be employed to represent the properties
of materials.
For distinguishing the onset of unstable crack extension, one needs to combine the crack
growth resistance curve (KR -curve) which represents the materials properties (Ripling and
Falkenstein, 1973) with the crack driving force curves which represent the characteristics of a
structure with a crack. In terms of Ki , the crack driving force curves (Ki -curves) are regarded
as the function of the crack length a, and the applied extra load Pi , i.e., Ki = Ki (Pi , a). The
procedure for distinguishing the onset of unstable crack extension by combining the KR -curve
and Ki -curves are illustrated in Figure 20.
From Figure 20 the onset of unstable crack extension can be described as follows:
(1) When the partial derivative of Ki with respect to a, ∂Ki /∂a, is less than the partial deriv-
ative of KR with respect to a, ∂KR /∂a. In other words, when the following inequality is
met:
∂Ki ∂KR
< (7)
∂a ∂a
the crack extension will be stable;
(2) When ∂Ki /∂a is equal to or larger than ∂KR /∂a, i.e.,
∂Ki ∂KR
> (8)
∂a ∂a
the crack will propagate unsteadily.
Measured KR-∆ a curves
3
Crack growth resistance K

(MPam )
2
CT-2-3
1 CT-3-1
CT-6-1
CT-6-2
0
0 50 100 150 200 250
Crack extension ∆ a (mm)

Figure 21. KR curves measured from large CT-specimens with different sizes.
The equilibrium condition to be involved in equation (8), i.e.,
∂Ki ∂KR
=
∂a ∂a
represents the onset of unstable crack propagation. This point is the point of tangency Kct , of
KR -curve to Ki -curve in Figure 20.
Now we can compare the measured KIcini and KIcun with the measured KR -curves to invest-
igate whether there is a determined corresponding relationship between the measured KIcini and
KIcun and the measured KR -curves. It will be helpful to understand the theoretical background
of the measured KIcini and KIcun from the experiments. As examples, the investigated results
from CT-specimens of CT-6-1, CT-6-2, CT-3-1 and CT-2-3 are utilized in the comparison.
The detailed results of the applied load Pi and the corresponding instantaneous propagating
crack length ai , that are measured from the four CT-specimens are presented in Table 6. By
inserting the test data of Pi and ai into formula (3), the data of the crack growth resistance KR ,
are obtained and listed in Table 6 too. Then KR -curves measured from the four CT-specimens
with three different sizes are plotted in Figure 21. Accepting usual scatter of material and test-
ing, Figure 21 shows that the measured KR -curves are size independent. This is in agreement
with the investigated results of others (Ripling and Falkenstein, 1973).
Furthermore, using the conditions of equations (7) and (8), Ki (Pi , a)-curves calculated
by formula (3) and KR -curve for every CT-specimen are drawn in Figure 22 to Figure 25
separately, where Ki (Pi , a) is a function of the propagating crack length a = a0 + 1a, under
conditions of the given constant load parameters Pi , with rising load increment.
To employ the condition of equilibrium given in equation (8), the corresponding critical
point, i.e., the tangency Kct , of KR -curves to Ki -curves in Figure 22 to Figure 25 can be
determined. It is found that the values of Kct determined by KR -curve approach are equal to
the values of KIcun measured in the experiments. And the values of the initial cracking points
on KR -curves shown in Figure 22 to Figure 25 are equal to that of KIcini . Because the points
of the initial cracking and the critical crack propagation on KR -curve are the most important
two points that represent the fracture properties of materials, it may be said that there is an
equivalent corresponding relationship between KR -curves and the double-K parameters KIcini
Table 6. The measured results of stable crack propagation, 1a and KR .
Nos. of Dimensions of loading levels Lengths of stable KRi

specimens specimens Pi (N) crack propagation (MPa m1/2 )
S × D × B (mm) 1ai (mm)
CT-6-1 3000 × 3000 × 200 20.0 0 0.459

43.594 50 1.042
52.863 75 1.290
59.59 105 1.491
65.862 125 1.676
68.999 130 1.764
72.762 140 1.876
78.487 170 2.077
81.543 190 2.197
84.680 200 2.323
CT-6-2 3000 × 3000 × 200 20.0 0 0.459

27.832 10 0.644
45.986 30 1.082
55.928 40 1.326
65.25 45 1.554
68.357 50 1.634
71.464 60 1.723
77.678 80 1.904
82.083 127 2.093
83.898 140 2.163
87.000 150 2.263
90.107 170 2.385
93.214 200 2.534
CT-2-3 specimen
Crack growth resistance KR,
3
Stress intensity factor KI,
K3(P3 =32.49 kN)

(MPam )
1/2
2
KR
K2(P2 =22.5 kN)

1
K1(P1 =11.5 kN)
0
0 50 100 150 200 250

Figure 22. The stability analysis of crack propagation for CT-2-3 specimen by KR -curve method.
Table 6 (contd.) . The measured results of stable crack propagation, 1a and KR .
Nos. of Dimensions of loading levels Lengths of stable KRi

specimens specimens Pi (N) crack propagation (MPa m1/2 )
S × D × B (mm) 1ai (mm)
CT-3-1 1500 × 1500 × 200 14.0 0 0.455

39.51 30 1.347
41.59 35 1.429
43.57 45 1.522
51.97 65 1.879
54.87 75 2.018
57.95 110 2.268
58.05 130 2.358
58.64 164 2.542
60.01 207 2.837
CT-2-3 1000 × 1000 × 200 11.5 0 0.457

18.29 30 0.783
21.32 40 0.936
24.43 50 1.101
26.23 80 1.281
26.6 90 1.336
27.84 95 1.418
27.83 100 1.438
28.21 110 1.500
28.71 120 1.573
29.45 130 1.663
30.07 140 1.752
30.57 160 1.899
31.18 180 2.072
31.75 190 2.185
32.24 195 2.258
32.42 200 2.312
32.49 210 2.404
and KIcun in basic principle. But, for the purpose of engineering application, the double-K
parameters KIcini and KIcun are more practical than KR -curve. If the conventional KR -curve
approach is used, one has to measure a KR -curve and to obtain the partial derivative of Ki
with respect to crack length a for practical structures. It is not convenient for the practical use
of it.
6. Double-K fracture criterion for crack propagation and its primary application to a
concrete dam
It is clear that there exist three different stages which include crack initiation, stable crack
CT-3-1 specimen
3

K3(P3=60.01 kN)
KR
(MPam )
2
1/2
K2(P2=40 kN)
K1(P1=14 kN)
0
0 50 100 150 200 250

CT-6-1 specimen
3
K4(P4=84.68 kN)
2
(MPam )
1/2
K3(P3=64 kN)
KR
1 K2(P2=42 kN)
K1(P1=20 kN)
0
0 50 100 150 200 250

propagation and critical unstable fracture during a complete fracture process in the concrete
fracture experiments on the small size three-point bending beams and on the large size CT-
specimens. By summing up the test data of the crack propagation collected from the small size
three-point bending beams and the large size CT-specimens, the two size-independent fracture
parameters KIcini and KIcun were found. KIcini can describe the crack initiation and KIcun can do
the critical crack propagation. The two fracture parameters directly measured from the ex-
periments are equivalent to the two controlling points on KR -curves measured from the same
experiments. Therefore, if the two fracture parameters KIcini and KIcun are combined, we can
reduce them to a double-K fracture criterion for crack propagation in quasi-brittle materials.
Since both the parameters, KIcini and KIcun , are in terms of K, it is called the double-K fracture
criterion. The double-K fracture criterion can be described as follows: (1) a preformed crack
begins to start cracking initiation when the applied stress intensity factor K at the preformed
crack tip arrives at the value of initial cracking toughness KIcini of the material; (2) the propagat-
ing crack extends steadily when the applied stress intensity factor K is less than the value of
unstable fracture toughness KIcun after the stress intensity factor K at the propagating crack tip
exceeded KIcini of the material; (3) when the stress intensity factor K at the propagating crack
tip arrives at the value of unstable fracture toughness KIcun , the propagating crack begins to
entry an unstable propagating situation.
They can be expressed in terms of mathematics as follows:
(1) when K = KIcini , the preformed crack begins to crack initially;
(2) when KIcini < K < KIcun , the propagating crack develops steadily;
(3) when K > KIcun , the crack propagates unsteadily.
For a structure containing a crack with an arbitrary geometrical configuration under ar-
bitrary loading condition, the stress intensity factor K at the crack tip can be evaluated by
an analytical solution, or a numerical approach using a finite element code. Then, no matter
what case applies, for example, whether the crack will begin to crack initially, or the crack
propagates steadily, or the crack will arrive at the onset of its unstable propagation, it can be
adjusted by using the double-K fracture criterion. For the same problem, if one uses a finite
element code combining the fictitious crack model or the crack band model, one needs more
CPU time. If one uses the two parameter model or the effective crack model, only the critical
situation can be described.
In practical structures the type of crack propagation also includes a mode II, a mode III
and a mixed mode besides the mode I above, specially, the mixed mode crack is more often
met. In principle, the double-K fracture criterion is not only suitable for the mode I crack
propagation, but also could be suitable for the mode II, the mode III and the mixed mode
crack propagation.
The suitability of the double-K fracture criterion to a mixed mode where mode II is pre-
dominant can be confirmed by the experimental evidence provided by a plaster model test
for simulating the fracture mechanism of a gravity dam in (Zhao, 1982). In the plaster model
test a preformed short horizontal crack at the heel of the simulated gravity dam under action
of water pressure on the upstream face of the gravity dam was studied to simulate the pure
mode II fracture which occurred on the short horizontal crack. In the test, the horizontal load
Q, that is simulated as the action of water pressure on the upstream face of the gravity dam
and the horizontal shear displacement u, were measured. It was found that before 0.477Pmax ,
Q-u curve is linear. At 0.477Pmax , there was a sudden extension of the horizontal shear
displacement u, (a small terrace) on Q-u curve. After 0.477Pmax , the measured Q-u curve
was nonlinear until up the maximum load Qmax . In (Zhao, 1982) the author reported that
from the beginning of loading until the final failure the crack propagation can be divided
into such three different stages: crack initiation, stable crack propagation and unstable failure.
It was found that the initial cracking occurred when the load Q was arrived at 0.477Pmax .
The corresponding stress intensity factors at the crack tip under the load of 0.477Pmax were
KI = 0.155 N/mm3/2, KII = 1.85 N/mm3/2. It was observable that KII was predominant
when crack initiation occurred. After initial cracking, the stable crack propagation was in a
pattern of the mixed mode I and mode II. The final failure was due to the common action of
the horizontal shear force Q and the self-weight of the gravity dam. It can be seen that the
observed three different stages in the model test of the gravity dam are in agreement with
those described by the double-K fracture criterion.
Recently, quite enough experimental evidence to prove the double-K fracture criterion
had been provided in the mode II fracture tests on mortar punch-through shear specimens by
Davies (1995). For the specimen studied in (Davies, 1995) with H = 30 mm and a/D =
0.4, KII /KI ratio was found to be in the order of 30. Such a result should be regarded as
predominance of mode II (shear) type of failure. The two sets of photographs, Figure 5 and
Figure 6 shown in the paper and an example of typical load-defection curves obtained from
the experiments indicated that there exist three different crack propagation stages of the initial
crack evolution, stable cracking process and unstable fracture. As the author reported, it also
can be observed in Figure 10 of the paper that the initial crack evolution begins at about 40
percent of the maximum load. When the crack begins to propagate initially, the corresponding
load was called the first crack load by the author in (Davies, 1995). She found that in both
cases studied the progressive cracking together with crushing of the crack face bridges took
place between the first crack load and the maximum load and the cracking process was fairly
stable. After the maximum load was reached the fracture process became unstable and very
fast. Such crack propagation phenomena are in agreement with those predicted by the double-
K fracture criterion. Similar phenomena have also been observed in the wood mode II fracture
tests (Xu et al., 1996).
Therefore, it could be expected that the double-K fracture criterion may be extended to
describe the mode II and the mixed mode I and mode II crack propagation. Regarding the
suitability of the double-K fracture criterion to the mode III crack propagation, it needs to
examine in further studies. Of course, in the further studies, it is also required to solve how to
ini un
determine the corresponding double-K fracture parameters, like KIIc and KIIc in the mode II
ini un
and KIIIc and KIIIc in the mode III as well as the detailed expression of the double-K fracture
criterion in terms of mixed crack mode.
Now the double-K fracture criterion is applied to a concrete gravity dam using the nu-
merical results of the stress intensity factors at the tip of horizontal crack on the bottom
position of the upstream of the gravity dam calculated by a finite element code in (Zhao,
1982). Under the conditions of different water levers H on the upstream of the dam, the
corresponding stress intensity factors at the horizontal crack tip, of which the length is equal
to 1 m, were given in Figure 8 in (Zhao, 1982). In order to illustrate the basic procedure, only
KI is considered. It is assumed that the concrete materials used in the bottom position of the
gravity dam is the same as that used in the tested CT-specimens so, KIcini = 0.46 MPa m1/2 and
KIcun = 2.55 MPa m1/2 are regarded as the corresponding fracture parameters of the concrete
material, where other factors are neglected. It can be known from (Zhao, 1982) that when
the height of water level H is 87 m, the mode I stress intensity factor KI is 0.46 MPa m1/2,
and when H = 95 m, KI = 2.55 MPa m1/2 . Therefore it could be predicted that under the
condition that the length of a horizontal crack is 1 m, if the height of water level H is less
87 m, the crack does not propagate further; if the height of water level is in such a region of
87 m 6 H 6 95 m firstly, the crack propagates initially, then, the crack propagates steadily,
finally, critical fracture happens; if H > 95 m, the crack will unsteadily develop. The critical
fracture of the crack is the most dangerous for the safety of the dam. If the mode II stress
intensity factor at the crack tip were taken into account in the instance, the corresponding
predicted values of H may be different from those mentioned above.
On the other hand, the double-K fracture criterion could be utilized to determine the al-
lowed crack lengths under conditions of different water levels for several controlling sections
of the dam. In the instance above, when the height of a water level H is 87 m, a = 1 m just is its
allowed crack length on the bottom position of the dam. So, the fracture parameter KIcini can be
regarded as a material parameter to evaluate the allowed crack length. And KIcini 6 K < KIcun
can be regarded as a warning region to protect the dam against unstable fracture; and KIcun
can be as a material parameter to control whether unstable fracture occurs in the dam. Such
a warning region may be more useful for judging the safety of some special structures like
prestressed pressure vessels of a nuclear reactor.
7. Conclusion
The large evidence of experimental investigations on the large size CT-specimens by the
photoelastic coatings and on the small size three-point bending beams by the laser speckle
interferometry showed that there exist three stages of crack propagation: crack initiation,
stable crack propagation and unstable fracture in quasi-brittle materials like concrete. The
initial cracking point and the critical fracture point can be distinguished from a complete
process of crack propagation. Because the initial cracking point represents the onset of stable
crack propagation and the critical fracture point represents the onset of unstable fracture,
they are of different characteristics. The complete process of crack propagation with different
characters in the quasi-brittle materials cannot be described by a single material parameter
like KIcs , CTODc and KIce etc. Summing up the test data collected from the experiments on
the small size three-point bending beams and the large size CT-specimens, it was found that
there exist two material parameters. Both of them are in terms of the stress intensity factor,
corresponding to the two distinguishably characteristic points observed in the investigations
on crack propagation in the experiments. The initial cracking stress intensity factor created at
the preformed crack tip by the initial cracking load Pini, is defined as the initiation toughness
KIcini . The critical stress intensity factor created at the critical crack tip by the maximum load
Pmax , is defined as the unstable fracture toughness KIcun . It was found that KIcini and KIcun are only
dependent on the material properties, not dependent on the tested specimen geometry and size.
By combining the two fracture parameters KIcini and KIcun , a double-K fracture criterion might
be manifested. The double-K fracture criterion could be employed to describe such a crack
propagation with two distinguishable characteristic points observed in the quasibrittle material
fracture tests.
As known, the role of stable crack propagation in concrete fracture has been recognized
well and has been taken into account as the critical stress intensity factor, like KIcs in (Jenq
and Shah, 1985) and KIce in (Karihaloo and Nallathambi, 1990). However, the crack initiation
cannot be described by both parameters of KIcs (Jenq and Shah, 1985) and KIce (Karihaloo
and Nallathambi, 1990). In addition the effective crack length ac used in the evaluation for
both KIcs and KIce was not measured directly in the experiments. Therefore, the double-K
fracture criterion that is summed up from the experiments can be regarded as a complement
for the fracture models in the published literature. Furthermore, the size-independence of KIcun
that is found in the directly measured test data can be used to confirm some aspects of KIcs
(Jenq and Shah, 1985) and KIce (Karihaloo and Nallathambi, 1990; Refai and Swartz, 1987)
that are determined by the indirectly measured methods. KR -curves had been employed by
many researchers to study the properties of the crack propagation in concrete fracture tests.
The examination for KR -curve by the collected test data from CT-specimens illustrated that
there exists a good corresponding relationship between the measured values of the unstable
fracture toughness KIcun in the double-K fracture criterion and the tangency Kct of K-curves
to KR -curves, and the one between the measured values of the initiation toughness KIcini and
the onset of stable crack propagation on KR -curves. So, in principle, the double-K fracture
criterion can be taken as a simplification of KR -curves because at the two distinguishably
characteristic points the double-K fracture criterion is equivalent to KR -curves.
The given instance of primary application of the double-K fracture criterion to a concrete
gravity dam showed that the double-K fracture criterion could be employed as the controlling
parameters to describe initial cracking, stable crack propagation and unstable fracture in con-
crete structures. Specially the double-K fracture criterion can provide a warning for some
important concrete structures, like the concrete protection shell of a nuclear reactor.
For understanding the essential properties of the crack propagation in concrete materials
and setting up a suitable fracture criterion, the collection of test data of KIcini and KIcun that
are measured by a direct way is necessary, but, the determination methods of KIcini and KIcun
in such a direct measure way are not convenient as an ordinary testing method. Especially
the determination of the initial cracking point and the critical crack propagation length ac
is not easy. Furthermore, the measure approaches used in this paper only can observe the
crack propagation on the specimen surface. As the result, it leads to an underestimate for KIcini
and an overestimate for KIcun after such a fact that the surface crack appears earlier than the
inside crack is considered. It is also in agreement with the primary prediction of analytical
solutions of the fictitious crack in the Griffith crack problem and the semi-infinite crack prob-
lem. Thereupon, a task to develop a simple, convenient and indirectly measured method for
determination the double-K fracture parameters KIcini and KIcun is presented. The analytically
evaluated methods for three-point bending notched beams, CT-specimens and wedge splitting
specimens will be presented in the subsequent series of papers.
Acknowledgements
The experiments reported in this paper were performed at Dalian University of Technology.
The first author (SX) is grateful for the continuous support from Dalian University of Tech-
nology and his former colleagues there, as well as National Natural Science Foundation of
China.
References
ASTM Special Committee on Fracture Testing of High-Strength Materials, ASTM Bulletin (1960), 29–40.
Bascoul, A., Kharchi, F. and Maso, J.C. (1989). Concerning the measurement of the fracture energy of a micro-
concrete according to the crack growth in a three points bending test on notched beams. Fracture of Concrete
and Rock (Edited by S.P. Shah and S.E. Swartz), Springer-Verlag, New York, 396–408.
Bazant, Z.P. (1996). Analysis of work-of-fracture method for measuring fracture energy of concrete. Journal of
Engineering Mechanics ASCE 122, 138–144.
Bazant, Z.P. and Oh, B.H. (1983). Crack band theory for fracture of concrete. RILEM, Materials and Structures
16(93), 155–177.
Bazant, Z.P., Kim, J.K. and Pfeiffer, P.A. (1986). Determination of fracture properties from size effect tests.
Journal of Structural Engineering. ASCE, 112 (2), 289–307.
Davies, J. (1995). Study of shear fracture in mortar specimens. Cement and Concrete Research 25(5), 1031–1042.
Du, J.J., Kobayashi, A.S. and Hawkins, N.M. (1989). An experimental-numerical analysis of fracture process zone
in concrete fracture specimens. Engineering Fracture Mechanics 35(1/2/3), 15–27.
Hillerborg, A. (1985). Results of three comparative test series for determining the fracture energy GF of Concrete.
Ibid. 18(107), 407–413.
Hillerborg, A., Modeer, M. and Petersson, P.E. (1976). Analysis of crack formation and crack growth in concrete
by means of fracture mechanics and finite elements. Cement and Concrete Research 6, 773–782.
Irwin, G.R. and Kies, J.A. (1954). Welding Research Supplement 19, 193–198.
Jenq, Y.S. and Shah S.P. (1985). Two parameter fracture model for concrete. Journal of Engineering Mechanics,
ASCE, 111(10), 1227–1241.
John, R. and Shah, S.P. (1986). Fracture of concrete subjected to impact loading. Journal of Cement and Concrete
Aggregation 8(1), 24–32.
Karihaloo, B.L. and Nallathambi, P. (1990). Effective crack model for the determination of fracture toughness
(KIcs ) of concrete. Engineering Fracture Mechanics 35(4/5), 637–645.
Karihaloo, B.L. and Nallathambi, P. (1991). Notched beam test: Mode I fracture toughness. Fracture Mechanics
Test Methods for Concrete, Report of RILEM Technical Committee 89-FMT (Edited by S.P. Shah and A.
Carpinteri), Chapman & Hall, London, 1–86.
Kaplan, M.F. (1961). Crack propagation and the fracture of concrete. Journal of ACI 58(5), 591–610.
Karihaloo, B.L. (1989). Do plain and fiber-reinforced concretes have an R-curve behavior? Fracture of Concrete
and Rock (Edited by S.P. Shah and S.E. Swartz), Springer-Verlag, New York, 96–105.
Mai, Y.W. (1984). Fracture measurements of cementitious composites. Application of Fracture Mechanics to
Cementitious Composites (Edited by S.P. Shah), NATOARW, 399–429.
Pflug, L. (1979). Proceedings of 7th National Congress of the Italian Society for stress Analysis, Supplement,
Cagliari, A.I.A.S., 5–26.
Refai, T.M.E. and Swartz, S.E. (1987). Fracture Behavior of Concrete Beams in Three-Point Bending Considering
the Influence of Size Effects. Report No. 190, Engineering Experiments Station, Kansas State University.
RILEM Technical Committee 50-FMC (1985). Determination of the fracture energy of mortar and concrete
by means of three-point bend tests of notched beams, proposed RILEM draft recommendations. RILEM,
Materials and Structures 18(106), 285–296.
RILEM Technical Committee 89-FMT (1990a). Determination of fracture parameters (KIc s and CTOD ) of plain
c
concrete using three-point bend tests, proposed RILEM draft recommendations. Ibid. 23(138), 457–460.
RILEM Technical Committee 89-FMT (1990b). Size-effect method for determining fracture energy and process
zone size of concrete, proposed RILEM draft recommendations. Ibid. 23(138), 461–465.
Ripling, E.J. and Falkenstein, E. (1973). Measuring KR -curves for thin sheets. Fracture Toughness Evaluation by
R-Curve Methods ASTM STP 527, American Society for Testing and Materials, 36–47.
Saouma, V.E., Broz, J.J., Brühwiler, E. and Boggs, H.L. (1991). Effect of aggregate and specimen size of fracture
properties of dam concrete. ASCE, Journal of Materials in Civil Engineering 3(3), 204–218.
Srawley, J.E. and Gross, B. (1972). Stress intensity factors for bend and compact specimens. Engineering Fracture
Mechanics 4, 587.
Tada, H., Paris, P.C. and Irwin, G.R. (1985). The Stress Analysis of Cracks Handbook. Paris Productions
Incorporated, St. Louis, Missouri, USA.
Tang, T.X., Ouyang, C.S. and Shah, S.P. (1995). A Simple Method for Determining Material Fracture Para-
meters from Peak Loads. NSF Center for Science and Technology of Advanced Cement-Based Materials,
Northwestern University, Evanston, Illinois, USA.
Wittmann, F.H. and Mihashi, H. and Nomura. Size effect on fracture energy of concrete. Engineering Fracture
Mechanics 35(1/2/3), 107–115.
Xu, Shilang. (1988). The Fracture Mechanism of Concrete. Ph.D. Thesis (in Chinese), Dalian University of
Technology, Dalian, China.
Xu, Shilang and Zhao, Guofan (1989a). A study on fracture process zones in concrete by means of laser speckle
photography. Brittle Matrix Composites 2 (Edited by A.M. Brandt and I.H. Marshall), Elsevier Applied
Science, London, 333–341.
Xu, Shilang and Zhao, Guofan (1989b). The stable propagation of crack in concrete and the determination of
critical crack tip opening displacement. Journal of Hydraulic Engineering (in Chinese), Beijing (4), 33–44.
Xu, Shilang and Zhao, Guofan (1989c). Determination of fracture toughness and the fracture energy of concrete.
Fracture Toughness and Fracture Energy (Edited by H., Mihashi et al.), Balkema, Rotterdam, 197–203.
Xu, Shilang and Zhao, Guofan (1991a). Researches on the Fracture Mechanics of Concrete (in Chinese). The
Press of Dalian University of Technology, Dalian, China.
Xu, Shilang and Zhao, Guofan (1991b). Concrete fracture toughness of huge specimens and criterion of fracture
toughness for judging cracks in high concrete dam. China Civil Engineering Journal (Quarterly, in Chinese),
Beijing 24(2), 1–9.
Xu, Shilang and Zhao, Guofan (1991c). The investigation on the propagation process of a crack in the concrete by
means of photoelastic coatings. Journal of Hydraulic Engineering (Quarterly, in Chinese), Beijing (3), 8–18.
Xu, Shilang et al. (1991d). Fracture energy and strain field near the tip of a notch in huge concrete specimen under
compact tension. Ibid. (11), 17–25
Xu, Shilang, Reinhardt, H.W. and Gappoev, M. (1996). Mode II fracture method for highly orthotropic materials
like wood. International Journal of Fracture 75, 185–214.
Xu, Shilang. (1982). Analysis and test on the fracture toughness of concrete. Journal of Hydraulic Engineering
(in Chinese), Beijing (6), 51–56.
Zandman, F., Redner, S. and Dally, J.W. (1977). Photoelastic coatings. Society for Experimental Stress Analysis
Monograph (3), SESA, The Iowa State University Press, Ames, Iowa, USA.
Zhao, Zhongji (1982). Fracture mechanical analysis of horizontal crack on the upstream face and heel of gravity
dams. Journal of Hydraulic Engineering (in Chinese), Beijing (6), 51–56.
Zhao, Guofan, Hui, Jiao and Xu, Shilang (1991). Study on fracture behavior with wedge splitting test method.
Fracture Process in Concrete, Rock and Ceramics (Edited by J.G.M. van Mier et al.), RILEM Proceedings 13,
E & FN Spon, London, 789–799.

Determination of Double-K Criterion For Crack Propagation in Quasi-Brittle Fracture, Part I: Experimental Investigation of Crack Propagation

Uploaded by

Copyright:

Available Formats

Determination of Double-K Criterion For Crack Propagation in Quasi-Brittle Fracture, Part I: Experimental Investigation of Crack Propagation

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Determination of Double-K Criterion For Crack Propagation in Quasi-Brittle Fracture, Part I: Experimental Investigation of Crack Propagation

Uploaded by

Copyright:

Available Formats

International Journal of Fracture 98: 111–149, 1999.

© 1999 Kluwer Academic Publishers. Printed in the Netherlands.

Determination of double-K criterion for crack propagation

SHILANG XU and HANS W. REINHARDT

Received 10 July 1997; accepted in revised form 17 August 1998

Figure 1. Three-point bending beam investigated by laser speckle interferometry.

2.1. S PECIMENS AND TEST PROCEDURE

Recorded plate with double

Figure 2. The illustration of the point-by-point filtering set-up.

2.2. C RITICAL CRACK TIP OPENING DISPLACEMENT, CTODc

Nos. of S×D×B a0 Pini Pmax CTODc δP 1ac

L-1 600 × 150 × 100 60 2402 4000 0.0207 0.169 53.8

Average 2449 4190 0.0188 0.170 45.0

S-1 400 × 100 × 100 40 1520 2435 0.0191 0.090 26.0

Average 1903 2869 0.0184 0.100 27.7

Measured P-CTOD curves Measured P-δ curves

2.3. I NVESTIGATION OF THE CRACK PROPAGATION OR THE FRACTURE PROCESS ZONE

2.4. D ETERMINATION OF CRACK INITIATION POINT, Pini

-10 0 10 -10 0 10 -10 0 10 -10 0 10

(a) P=1373 N (b) P=1677 N (c) P=2226 N (d) P=3100 N

(e) P=3432 N (f) P=3677 N (g) P=3913 N (h) P=4158 N

(a) P=1834 N (b) P=2098 N (c) P=2500 N (d) P=2942 N

(e) P=3334 N (f) P=3726 N (g) P=4020 N (h) P=4413 N

-10 0 10 -10 0 10 -10 0 10 -10 0 10 -10 0 10

(a) P=1137 N (b) P=1500 N (c) P=1853 N (d) P=2285 N

(e) P=2608 N (f) P=2814 N (g) P=3069 N (h) P=3305 N

2 1--- Load versus crack mouth opening

0.0 2 0.04 0.0 6 0.08 0.10 0.12 CMOD (mm)

3. Crack propagation in very large size CT-specimens observed by photoelastic coating

3.1. S PECIMEN PREPARATION

3.2. P REPARATION AND APPLICATION OF COATINGS

3.3. T EST PROCEDURE

Hydraulic jack Applying load

For CT-6-1specimen with

dimensions of 3.6x3.0x0.2 meters,

test date: Dec. 1,1987

3.4. T HE INVESTIGATED RESULTS OF THE STABLE CRACK PROPAGATION IN THE LARGE

2(1 + νs )tc ε1s

Measured P-∆a curves

Crack extension ∆a (mm)

2 × (1 + 0.197) × 2.7 × 250 × 10−6

CT-3-1 51.97 0.866 65 65

CT-2-3 23.42 0.752 50 65

Nos. of specimens Dimensions of specimens a0 Pini Pmax 1ac

CT-6-1 3000 × 3000 × 200 1200 20.0 84.68 210

1 2.7 × 4.54 × 103 (1 + 0.197)

ε1s − ε2s = 700 × 10−5 n/(2 × 2.7) = 1297 × 10−6 n

4.1. T HE EXPERIMENTAL EVALUATION OF THE DOUBLE -K FRACTURE PARAMETERS

be useful in the engineering application and to be measurable in the experiments as well as to

1.99 − α(1 − α)(2.15 − 3.93α + 2.7α 2 )

4.2. A PRIMARY COMPARISON WITH THE ANALYTICAL SOLUTIONS OF THE FICTITIOUS

KIcini + KIcc = KIcun . (4)

Geometry of Nos. of Dimensions of Cube strength Tensile Modulus ini

Three-point bend beams L-1 600 × 150 × 100 0.513 3.454

Average 0.509 2.428

Compact tension specimens CT-6-1 3000 × 3000 × 200 0.459 2.323

Average 0.458 2.525

5. KR -resistance curve for the crack propagation in concrete fracture

Nos. of Length Depth Thickness Initial Crack β c