You are on page 1of 17

Experimental determination of the stress-crack opening curve for concrete in tension

RILEM TC 187-SOC: Final report – May 2007

CHAPTER 3. INDIRECT TESTS FOR STRESS-CRACK OPENING


CURVE
Prepared by Jaime Planas (1), Gustavo V. Guinea (1), Jaime C. Gálvez (2), Beatriz Sanz (1)
and Adel M. Fathy (1)

(1) Departamento de Ciencia de Materiales, Universidad Politécnica de Madrid, Spain


(2) Departamento de Ing. Civil Construcción, Universidad Politécnica de Madrid, Spain

1. INTRODUCTION
Indirect tests use geometries and loadings that lead to complex states of stress, strain, and
crack opening. Their advantage is that they are easier to carry out than the uniaxial tension
test; their disadvantage is that they require relatively complex data reduction or inverse
analysis procedures.
In principle, any test with clearly defined boundary conditions (which must, moreover, be
easy to implement in the laboratory), in which a single crack grows in a stable, controlled
manner, can be used as an inverse test provided enough information is drawn during the test.
However, inverse analysis is known to be ill-posed in many circumstances and the
uniqueness of the solution may be an issue to consider. In fact the feasibility of the inverse
analysis is contingent not only on the specimen geometry and boundary conditions, but also
on the material properties, most particularly on its brittleness length, to be defined next.
It is well known, since Petersson’s thesis [1], that the stress crack opening curve for
concrete is highly nonlinear and basically smooth, as roughly depicted in Figure 1. The
softening curve has two basic parameters, the tensile strength ft at which the crack starts to
open, and the fracture energy GF, which equals the area between the curve and the axes.
All fracture processes, including fracture tests, exhibit a strong size effect, which is
controlled by the ratio of the specimen size D (beam depth, cylinder diameter, e.g.) to some
material length that can be built with the elastic and fracture parameters [2]. Hillerborg’s
characteristic length lch is defined as [3]
EGF
lch = (1)
ft 2
where E is the elastic modulus. This is a length that stems from the comparison of the elastic
stored energy and the total energy that the crack can absorb, and characterizes the brittleness
for complete fracture. However, in many circumstances the structure fails (i.e., reaches the
peak load) well before the crack is fully developed. In such a case, the response is as if the
stress vs. crack opening curve were given by its initial linear approximation as shown by the
dashed line in Figure 1, and a more appropriate material length, called the brittleness length l1,
may be defined as in [4-6] :

13
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

Ew1
l1 = (2)
2 ft

Figure 1. Stress versus crack opening curve an initial linear approximation.

As a rule of thumb, lch ≈ 2l1 and l1 = 100-150 mm for concrete; so we can notice that
these material lengths are of the order of the size of the specimens we test in the laboratory.
This is important because the ability to get complete information about the softening curve
depends on the ratio D/l1, which is larger for more brittle behavior. This implies that a
particular combination of experiment and data reduction method may be valid only over a
particular range of the ratio D/l1.
As limiting examples of this very general property, we have the tests on precracked and
smooth specimens of very brittle, non-plastic materials (e.g., fine ceramics) in which, for the
specimen size that may be tested in the laboratory —several millimeters, say—, we have
D/l1 >> 1. For precracked specimens of such a material, it is well known that LEFM applies
and, if measurements are only carried out on the surface of the specimen (load, displacement,
CMOD, ...) the only parameter that can be extracted from those test is the fracture energy GF
(or the fracture toughness KIC); nothing can be said about the tensile strength or the shape of
the softening curve. Conversely, if the specimen has no macro-crack, it will fail (usually
catastrophically) as soon as a crack is initiated at the worst location in the specimen, and thus,
the only information that can be obtained is ft ; nothing can be said about the fracture energy
or the shape of the softening curve.
Therefore, indirect tests and the corresponding data reduction techniques must be defined
and validated over specific brittleness ranges for which the method has been theoretically
proved to be viable.
In the following, two indirect tests are analyzed for which the theoretical validity range has
been assessed by thorough computational means. In Section 2, the test based on diametric
splitting of cylinders is discussed, for which very specific recommendations can be set. Then,
in Section 3, the test based on stable three-point-bending of notched specimens is presented
from the pure experimental point of view, with emphasis on recent improvements of the
experimental techniques and on the issues that are still open to discussion. The present
contribution is closed, in Section 4, with an outline of the explicit data reduction technique
that allows determination of a bilinear approximation of the stress-crack opening curve from

14
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

results of diametric splitting tests and stable three-point-bending tests of notched beams. A
similar procedure could be devised for other fracture specimens, such as the wedge-splitting
specimen, but the computational effort required to support such a method has not been carried
out yet.
2. DIAMETRIC SPLITTING TESTS
Diametric splitting tests of concrete cylinders and prisms were extensively studied by
Rocco et al. in the light of the cohesive crack model [7-9]. It was shown that, despite all a
priori objections that can be raised against the splitting strength (especially because of the
mixed tension-compression biaxial state along the potential crack line), it could be accurately
described by the cohesive crack model, as shown in Fig. 2.
This figure further shows that the splitting strength is strongly influenced by the size of the
specimen and by the width of the load bearing strips. The larger the specimen and narrower
the loading strip, the closer is the splitting strength to the true tensile strength of the cohesive
model. Moreover, for a given specimen dimension D, cylinders perform better than square
prisms in obtaining a splitting strength closer to the true tensile strength.
The work of Rocco showed moreover that the results depend on the stress rate during the
test, with best results for small stress rates; a limit of 1.0 MPa/min was proposed as a
reasonable compromise between accuracy and duration of test. From this, a compact set of
recommendations can be obtained as follows:
• Use cylindrical specimens manufactured according to the applicable national or
international standard. Curing the specimens submerged in lime saturated water at
23±2°C is the preferred mode of conservation.
• Test the cylinders for splitting strength according to the applicable national or
international standard with the following specific modifications:
o The width of the load bearing strip shall be between 4 and 8% of the specimen
diameter; 10 mm for 150 mm diameter and 5 mm for 75 mm diameter
cylinders may be a reasonable choice.
o Keep the specimen surface wet throughout the test.
1.4 1.4
granite microconcrete
1.3 0.16 1.3 0.16
b
b
1.2 1.2 D
theory D 0.08 theory
t
t

f /f
f /f

ts
ts

1.1 0.08 1.1


0.04
0.04
1.0 1.0
experiment, b/D = experiment, b/D =
0.16 0.08 0.04 0.16 0.08 0.04
0.9 0.9
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
D/l D/l
1 1
Figure 2. Comparison of the theoretical and experimental results of the splitting tensile
strength, fts, versus the specimen size, according to the results of Rocco et al.[8,9].

15
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

o Keep the stress rate between 0.5 and 1.0 MPa/min all the way up to the peak.
A record of load versus time should be taken to verify compliance with this
specification.
• Use the splitting test in combination with fracture tests on notched specimens and
check that the size of the specimen is large enough for the splitting strength fts ≅ ft . A
reasonable condition, according with the curves in Figure 2 for cylindrical specimens
is D ≥ 0.8l1 , which guarantees a theoretical error less than 2%.

Note: Rocco’s conclusions were drawn from tests on concrete, mortar and granite for which
the compressive strength is roughly ten times larger than the tensile strength, thus ensuring
that the crushing zone below the loading strips is kept small. This test may not be adequate
for materials with enhanced tensile strength, in particular for fiber reinforced concrete.

3. STABLE BENDING TESTS ON NOTCHED BEAMS

3.1. Specimen geometry


The stable three-point-bending test on notched beams of rectangular cross section is one of
the most popular methods to study fracture behavior of concrete (as well as other materials).
The pioneering work of RILEM TC 50 « Fracture Mechanics of Concrete » was based on this
geometry [10], with, however, a very large span-to-depth ratio S/D = 8.0, which was
recommended to facilitate obtaining stable tests under stroke control.
Nowadays, most testing machines come with closed-loop control, and shorter specimens
can be used by testing under CMOD control. A span-to-depth ratio of 4.0 became very
popular in the nineties, since this was the ratio in use in testing fracture toughness of metals
and the formulae for the stress intensity factor and CMOD as functions of crack depth were
available for this particular ratio [11].
In 1998 formulae were devised for stress intensity factor, load-point displacement, and
CMOD, that extend the range to any span-to-depth ratio larger than 2.5 [12]. Therefore, in
accordance with the consensus reached in the discussions during the NSF Workshop on
‘Workshop to Select a Fracture Toughness Test Standard for Concrete’ held in Captiva
Island, Florida, November 18-20, 2001, we propose to use beams with a span-to-depth ratio
of 3.0, which is customary in bending tests of concrete. This reduces specimen weight and
facilitates the reuse of molds and testing devices.
The notch depth adopted by RILEM TC 50 was one half of the beam depth, which leads to
more stable tests in stroke control. For the reasons discussed before, a shorter notch can be
used in CMOD control, with the advantage that the ligament, and thus the directly sampled
material, is larger. A reasonable compromise seems to be a notch of 1/3 of the beam depth.
The notch must be cut centered with its front line perpendicular to the planes that were
horizontal during casting.
The notch must be narrow enough to act like a crack in the specimen, and thus its width
must be much less than (1) the specimen depth, and (2) the material brittleness length l1 ; there
is no need to make the notch width small relative to the maximum aggregate size. To the
authors’ knowledge, a systematic analysis of the influence of the notch width on the test
results does not exist (especially in relation to the brittleness length, which is crucial).

16
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

Without further information, and taking into account that the brittleness length is not known
beforehand for a particular material, some a priori compromise must be adopted. Since, as
pointed out in the Introduction, the brittleness length for concrete and mortar is usually of the
same order of the beam depth (for mortars, which are more brittle, one tends to use smaller
specimens), and based on the availability of diamond disks of the appropriate thickness,
notches with a nominal width of 2% of the beam depth (±0.25 mm) are recommended.
The specimen size must be in accordance with the maximum aggregate size da. It is
advisable that the beam depth be as large as possible relative to da, but it is also advisable that
the specimen be as small as possible to limit its weight because heavier specimens are more
difficult to handle and more error prone. Therefore a minimum beam depth of five times the
maximum aggregate size is suggested as an absolute limit, which makes a ligament size of
3.33 times the maximum aggregate size, with a preferred limit of 6 times the aggregate size,
which makes a ligament of 4.0 times the aggregate size.

3.2. Testing rig


The test must be carried out in a closed loop machine, with control in CMOD. The bending
rig must provide torsion- and friction-free supports. The most stable device used by the
authors incorporates torsion-free hinges at the central loading support and at one of the lower
supports (Figure 3). Previous devices with anti-torsion hinges at both lower supports were
much less stable.
Frictionless supports are of paramount importance in this test. The simplest and, possibly,
best way to achieve low friction is to use hardened steel rollers on hardened steel plates as
sketched in Figure 3, a design borrowed from fracture toughness testing of metals (e.g., [11]).
While the rollers can be clamped during the test setup, it is essential to unclamp them before
starting the test. It is also essential to keep the rolling surfaces clean of any debris that can be
produced during handling; see Figure 4 for an illustration.

Figure 3. Sketch of the testing rig.

17
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

Figure 4. Picture of the authors’ implementation of the proposed device. Note that the rollers
are clamped with rubber bands and that the left support is soiled; rubber bands must be
removed and supports must be cleaned before initiating the test.

3.3. Weight compensation


The test procedure described in the following is devised to obtain the maximum possible of
information, including the far tail of the load-displacement curve, which requires self-weight
compensation.
The most stable and compact way to compensate the self weight of the specimen, in the
authors’ experience, is by clamping two prismatic steel counterweights of about the same
cross-section as the concrete specimen at both ends of the specimen, as sketched in Figure 5
and shown in the picture of Figure 4. Other systems used by the authors include prestressed
springs and hanging dead weights (see, e.g., [2], Sec.7.3.3), but these are more difficult to use
without disturbing the test, especially for heavy specimens.
Note : Assuming a relative density of 2.4 for concrete and 7.85 for steel, the minimum length
of each prismatic counterweight is about 0.54D for beams with a total length L = 4D, and
about 0.19D for beams with a total length L = 5D. The dimensions in the photograph in
Figure 4 correspond to the latter case, and the sketch in Figure 5 to the former one.

3.4. Measuring devices


During the test, load, deflection and CMOD have to be recorded. The load measurement shall
comply with the specifications of an appropriate national or international standard, and will
not be further discussed here.

18
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

STEEL
WEIGHTS

A B

Figure 5. Specimen self-weight compensation.

0.2D MAX.
h 0.05D
MAX.
0.25D 0.25D
MAX. MAX.

Figure 6. Detail of the attachment of the CMOD gage.

To measure the CMOD a clip-on gage or a knife-edge extensometer can be used. If the
CMOD record has to be used for quantitative data analysis, it would be desirable that the
knife-edges be located on the bottom specimen face. However, this is not practical, in general,
because it would require further machining of the notch mouth, and so knife-edges are located
at a distance h from the specimen face as depicted in Figure 6. This distance must be kept as
small as possible and be always less than 5% of the specimen depth.
Measuring the deflection is not trivial, because as discussed in [17] crushing and highly
localized deformations at the loading point and the supports should be excluded from the
measurements. The easiest way is to measure the relative vertical displacement between a
horizontal line defined by points on the top face of the specimen initially in the vertical of the
supports (points A and B in Figure 5), and a point on the bottom face of the specimen just in
the center (point C in Figure 5).

19
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

The reference line (or plane) can be implemented by means of a rigid frame with three
supporting conically tipped screws, two over one support and one centered on the other
support, as shown in Figure 4 and in the sketch in Figure 7a.
Unfortunately, the bottom center point coincides with the notch and it is difficult to attach
a reference there. To solve this problem, the following method has proven to be effective.
Two 6 mm thick plates were cut with the shape shown in Figure 7b. and fastened to one
side of the notch. The one on the right side of the notch has wings at both ends that protrude
out the specimen sides and through its center, thus providing an adequate surface to support
the tip of the displacement transducers attached to the rigid frame, fully symmetrical and
centered. In the authors’ implementation, the knife-edges for the CMOD gage are inserted in a
groove in the reference plate and in a smaller but similar plate on the other side of the notch.
In this way, only four screws are needed to hold the assembly, two on each side of the notch.
A bottom view of the specimen is shown in Figure 8 showing the plates and the CMOD gage.

(a)
(b)

STEEL
WEIGHTS

Figure 7. (a) Sketch of specimen with reference frame for displacements, LVDT, bottom
reference plate and CMOD gage. (b) Sketch of bottom view of specimen with measuring
plates.

20
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

Figure 8. View of the specimen from the bottom, with the CMOD gage in the center, and the
measuring plates, one longer and with wings to provide a support for the mobile tips of the
displacement transducers.

3.5. Procedure
The procedure to carry out the test can be summarized as follows :
− Take the specimen out of the bath, and keeping the surface wet all the time, mark with
pencil the central and loading cross-sections. Mark the position of the screws to fasten
the bottom plates, and drill the holes for the screws using water refrigeration.
− Cut the notch with a diamond saw and water refrigeration.
− Fasten the bottom plates (with the knife edges for the CMOD gage) to the specimen.
− Place the specimen on the testing rig, attach the counterweights, put the reference frame
for displacement measurement and the transducers in place and carefully center the
whole assembly (all devices must be centered within ±1 mm or 0.25% of span,
whichever is the less). At this stage the rollers may be clamped to easy the placement.
− Set all measuring channels to zero, and gently preload the specimen up to a load of 5 to
15% of the expected peak load. After the test, this pre-load must be checked not to have
exceeded 15% of the actual peak load. If this limit was exceeded, the test must be
considered invalid. During the pre-load, the hinges must be free (and even helped) to
accommodate any initial geometrical misfit.

21
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

− Unclamp the rollers and run the test under CMOD control. The rate must be selected so
that the peak load is reached within 3 to 5 min. Keep the rate constant until the load past
the peak decreases to 33% of the recorded maximum. Then increase progressively the
rate until ten times the initial rate and keep it until the end of the test.
Note: If the rate is kept constant during all the test, the testing time may be very long (as
much as 3 to 5 hours). The solution given above is an expedient that can be used with
most testing machines. If the machine can be input any specified evolution of CMOD
vs. time, a good choice could be to use the piecewise function defined as CMOD = w0
t/t0 for t < t0, and CMOD = w0 exp[(t-t0)/t0], for t < t0 where t0 = 3-5 min is the
(nominal) time to peak and w0 the (nominal) CMOD at the peak. With this rate, the time
to test completion is around 5 times the time to peak (i.e. between 15 and 25 minutes,
which is reasonable). The value w0 has to be initially guessed and then corrected from
the results of a trial test, just as we do for the initial rate.
− The test can be stopped when the CMOD reading reaches 4D/300.

Note: The definition of the rate and the end of the test should depend on the fracture
properties themselves. In a sense, defining the initial rate in terms of time to reach the peak
does take into account the material properties. However, for the end of the test there is no
known way to make a choice in terms of fracture properties (for example in terms of GF/ft). In
trial tests simulations for 150 mm depth beams it has been found that an absolute value of
CMOD = 2 mm is a reasonable round number for the end of the test. This corresponds to 4/3
of a percent of the depth in terms of rigid body rotation of the two halves of the specimen,
which is the value adopted in the foregoing.

4. EXPLICIT INVERSE ANALYSIS FOR BILINEAR APPROXIMATION

4.1. General concepts


A bilinear approximation of the softening curve (Figure 9) has been extensively used to
approximate the stress-crack opening curve of concrete (see [2], Sec. 7.2.1, for a review of the
proposed bilinear curves). In 1994, a relatively simple method not requiring extensive
numerical computation was devised in [13], based on the results of diagonal splitting strength
and of stable bending tests on notched beams. The four parameters of the bilinear
approximation are derived by forcing the resulting approximate curve: (1) to predict (by
numerical computation) the same peak loads as those observed in both the splitting cylinders
and the notched beams, 2) to have same fracture energy as that measured on the bending tests,
and 3) to predict the same far-tail behavior as observed in the beam tests (this leads to the
determination of the position of the center of gravity of the area below the stress-crack
opening curve).
That initial work underwent several improvements [2,14-16], especially regarding the
development of closed-form, explicit formulas, and the extension to beams with a span-to-
depth ratio of 3.0 and to a wide range of notch-to-depth ratios. Here we briefly describe the
latest state of the procedure which relies on the explicit formula relating the brittleness length
l1 to the peak load in the bending tests [14]:

22
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

 c c  f Pu S
l1 = (1 − α 0m ) D  2 1 2 + 22  , with x = t and fp = (3)
 ( x − 1) 2 B ( D − a0 )
2
x  fp

where α0 = a0 /D is the notch-to-depth ratio, fp is the net plastic strength, defined in terms of
the peak load Pu and the geometry of the specimen, and m, c1 and c2 are parameters depending
on the span-to-depth ratio, which, for S/D = 3.0, take the values 1.7, 11.2 and 2.365,
respectively.
Note: in [14], the values of the constants were determined for a span-to-depth ratio of 4; the
present values were recomputed using the same procedure as in [14] for the span-to-depth
ratio of 3.0, which, as discussed in the previous Section, is standard in testing concrete prisms
in bending.
This formula is accurate within ±0.5% over the following ranges of depths of beams and
notches:

0.15 ≤ D/l1 ≤ 2.5 and 0.3 ≤ a0 /D ≤ 0.37 (3a)

Note : the formula can be used with reduced accuracy, within (-1% ,+0.5%), for 0.1 ≤ D/l1 ≤
3.0 and notch depths between 30 and 70% of the beam depth.

The foregoing equation assumes that when the peak load takes place, none of the points in
the cohesive zone have softened below the kink point on the softening curve. Therefore, such
a condition must be checked after the determination of the bilinear approximation. It has been
found through numerical simulations that the condition to be satisfied can be approximately
given by the simple formula
σk D
≤ 0.44 − 0.15ln (3b)
ft l1

Figure 9. Bilinear softening curve.

23
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

Given the foregoing formula and conditions, the outline of the procedure, which solves for
material parameters one at a time, is as follows:
− Determine the approximate value of the tensile strength ft by approximating it as the
mean of the splitting tensile strength on a series of standard cylinders as described in
Section 1. (This assumes a priori that D/l1 ≥ 0.8, which must be verified a posteriori).
− Determine the brittleness length l1 from the peak load of each specimen by formula (3).
− Determine the elastic modulus E from the initial slope of the load-CMOD curve.
− Determine the fracture energy GF from the work of fracture in a classical way [10,17,2].
− Determine the center of gravity of the area below the softening curve (wG) by curve
fitting of the far end of the load vs. CMOD curve [14,2]. Indeed, the load in the far end
of the test can be expressed as a function of the deflection δ or the CMOD wM assuming
that the movement of the specimen can be approximately described as two rigid halves
linked by a hinge in the middle. Then δ = wM S /4D and P can be expressed as:
2
A  4D  A
P= =  2 (4)
δ 2
 S  wM
where A is a constant that was shown in [14] to be related to the fracture energy GF and
the centre of gravity of the area enclosed by the softening curve wG as
1
A= BSGF wG (5)
4
where B is the specimen thickness.

4.2 Details of the procedure


Based on the foregoing key ideas, the detailed calculation procedure is as follows (this
includes some ‘data cleaning’ due to self-weight compensation):
1. For a set of a minimum of three cylinders (six recommended) tested for indirect
splitting strength, determine the mean tensile strength for the concrete as follows:
a. Determine the splitting tensile strength for each specimen from the equation
fts = 2 Pu /(π DL) for the Brazilian test, where Pu, D and L are the ultimate load,
diameter, and length of the specimen, respectively.
b. Determine the (mean) tensile strength of the concrete, ft, as the mean of the
splitting tensile strength of all the cylinders.
2. For each specimen of a set of a minimum of three beams (six recommended) tested in
three-point-bending, carry out the following steps 3-7:
3. Determine the modulus of elasticity E of the specimen as follows
a. Select the test points with measured load P’ between 15 and 55% of the
maximum load on the raising branch of the P’-CMOD curve.
Note : the recorded load P’ does not include the combined effect of specimen
weight and steel counterweights ; this is accounted for in forthcoming steps.

24
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

b. Fit a straight line to the selected points by linear regression of CMOD vs. P’ to
calculate the initial compliance of the specimen Ci = ∆wm/∆P’.
c. Compute the elastic modulus E as
6Sa0 a +h
E= V (α '0 ) , with α '0 = 0
2 1
, (6)
Ci BD D+h
where h is the thickness of the steel knife used to clamp the CMOD gauge (Fig. 6) and
0.66
V1 (α ) = 0.8 − 1.7α + 2.4α 2 + +
(1 − α )
2

(7)
4D
+
S
( −0.04 − 0.58α + 1.47α 2 − 2.04α 3 )

4. Determine the constant A in Equations (4) and (5) as follows:


a. Exclude from the record the points with CMOD > 4D/300.
b. Write down the values of the CMOD and the measured load for the last point of
the record, say wMR and P’R.
c. Calculate a corrected load P1 as P1 = P’- P’R.
d. Draw the curve P1-CMOD and determine the CMOD value wMA for zero load
on the load-raising branch (Figure 10).
e. For all the points of the tail such that P1 < 0.05 P1u, where P1u is the maximum
corrected load, calculate the values X defined as

 4D 
2
 1 1 
X =   − 2 (8)
 ( wM − wMA ) ( wMR − wMA ) 
2
 S 

f. Plot P1 versus X and compute the best fit of the equation P1 = X(A+kX) to the
experimental points by least square fitting (most data-analysis packages will
carry out the required non linear regression computation; if this feature is not
available, carry out linear regression of Y = P1/X versus X).
g. The value of A so determined is recorded to 3 significant digits.

25
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

load P1, N

P1u

CMOD wM, mm
0
wMA wMR= 1.33D

Figure 10. Plot of corrected load P1 versus CMOD.

5. Determine the brittleness length l1 and the horizontal intercept of the initial linear
approximation w1 as follows :
a. The effective maximum load Pu is calculated as
A
Pu = P1u + (9)
( wMR − wMA )
2

b. Compute the brittleness length of the specimen l1 from Equation (3); at this
point the mean tensile strength must be used for all the specimens (no
individual values of ft are available).
c. Compute w1 from Equation (2) as w1 = 2ft l1 /E..

6. The fracture energy GF is determined as follows:


a. The corrected load P1 is plotted against the deflection δ and the deflections δA
and δR (in the loading and unloading parts) corresponding to zero load are
determined (Figure 11).
b. For each specimen, the measured fracture work WFm is calculated as the area
enclosed by the curve P1 - δ and the δ axis.
c. Then, the total fracture work is calculated as
2A
WF = WFm + (10)
δR −δA
d. The fracture of energy GF is computed as
WF
GF = (11)
B( D − a0 )

26
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

load P1, N

P1u

WFm

load-point displacement δ, mm
0
δA δR

Figure 11. Plot of corrected load P1 versus load-point displacement δ.

7. The abscissa of the centre of gravity of the softening curve wG is determined by solving
for wG from Eq. (4), with the result wG = 4A/(BSGF).

8. For the whole set of beam specimens, determine the mean bilinear approximation of
the softening curve as follows:
a. Compute the mean values w1m , GFm , and wGm of, respectively, w1, GF, and wG
for the set of all bending specimens (minimum of three, six recommended)
b. Compute the characteristic crack opening wch as wch = GFm / ft.
c. Compute the critical crack opening of the bilinear approximation as:

3wGm − w1m  2 w ( 3wGm − 2 wch )( 2 wch − w1m ) 


wc = wch 1 + 1 − 1m  (12)
2 wch − w1m wch ( 3wGm − w1m )
2
 

d. The coordinates (wk, σk) of the angular point of the bilinear curve (as shown in
Figure 9) are then given by:
2wch − w1m wc − 2wch
σ k = ft and wk = w1m (13)
wc − w1m wc − w1m

9. Verify that the conditions of applicability of the data reduction techniques are fulfilled :
a. Check that, for the split cylinders, the diameter satisfies D ≥ 0.8 l1m, where l1m
is the mean value of the brittleness length determined in the beam tests.
b. Check that, for the beams, the beam and notch depths satisfy conditions (3a).
c. Check that the beam depth and the resulting softening curve satisfy condition
(3b).

27
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

4.3. Automatization and example


Although the foregoing procedure can be carried out with standard tools for data-analysis,
it is relatively easy to run into errors that may be difficult to trace since the intermediate
operations executed by the user may not be recorded.
One way to provide traceability to the operations is to drive the tools with a script
whenever possible. In this way the script itself can be exhaustively checked and the
operations repeated starting with the original experimental data record. However, not all the
utilities are scriptable, and, certainly, most of those that are scriptable, are not portable and
use proprietary systems.
Therefore, Fathy et al. [16] adopted the expedient of writing an object oriented program in
C++ that is fully portable. In its present version, the program reads from a text file the
geometrical data of the specimen, the tensile strength (previously determined) and the table of
experimental results. It then processes the data through steps 3-7 and writes the results to a
file. Once all individual bending tests have been processed, the program proceeds through
steps 8 and 9 and writes the final results to the file. The program is now running satisfactorily
on Mac OS X. Once it gets through the verification process on Windows and Linux it will be
made freely available through the net.
Figure 12 shows an example of the final comparison of the original mean experimental
load-CMOD curve and the curves computed using the foregoing procedure for two types of
concrete, one normal strength (40 MPa) and one high strength (85 MPa) [16]. For each
concrete 6 beam specimens of 100x100x500 mm were tested. Note that, as expected, the
prediction fits well the peak load, the area under the curve and the far tail. Taking into
account that a bilinear curve is a rough approximation of the actual, smooth softening, the
comparison can be considered excellent both for normal strength and for high strength
concretes.
5.0 10.0
normal strength concrete high strength concrete
4.0 8.0

3.0 6.0
P (kN)

P (kN)

Experiment, mean Experiment, mean


2.0 Numerical, bilinear 4.0 Numerical, bilinear

1.0 2.0

0.0 0.0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
CMOD (mm) CMOD (mm)

Figure 12. Comparison of the experimental load-CMOD curves and the curves computed
using the bilinear softening as determined by the procedure in Sec 4.2 (Adapted form [16]).

28
Experimental determination of the stress-crack opening curve for concrete in tension
RILEM TC 187-SOC: Final report – May 2007

REFERENCES
[1] Petersson, P.E. “Crack Growth and Development of Fracture Zone in Plain Concrete and Similar
Materials”. Report No. TVBM-1006, Division of Building Materials, Lund Institute of
Technology, Lund, Sweden (1981).
[2] Bažant, Z.P. and Planas, J, Fracture and Size Effect in Concrete and Other Quasibrittle Materials,
CRC Press, Boca Raton, FL (1998).
[3] Hillerborg, A., Modéer, M. and Petersson, P.E. “Analysis of crack formation and crack growth in
concrete by means of fracture mechanics and fracture elements”, Cement and Concrete Research,
6 (1976) 773-782.
[4] Planas, J. and Elices, M. ‘Shrinkage eigenstresses and structural size-effect’, Fracture Mechanics
of Concrete Structures, Z. P. Bazant, ed., (Elsevier Applied Science 1992) 939-950.
[5] Elices, M., Guinea, G.V., Gómez, J. and Planas, J. ‘The cohesive zone model: advantages,
limitations and challenges’, Engineering Fracture Mechanics, 69 (2002) 137-163.
[6] Planas, J., Elices, M., Guinea, G.V., Gómez, F. J., Cendón, D.A. and Arbilla, I. ‘Generalizations
and specializations of cohesive crack models’, Engineering Fracture Mechanics, 70(14) (2003)
1759–1776.
[7] Rocco, C., Guinea, G. V., Planas, J. and Elices, M. ‘Mechanisms of rupture in the splitting test’,
ACI Materials Journal, 96(1), 52-60 (1999).
[8] Rocco, C., Guinea, G.V., Planas, J. and Elices, M. ‘Size effect and boundary conditions in the
Brazilian test: experimental verification’, Materials and Structures, 32 (1999) 210-217.
[9] Rocco, C., Guinea, G.V., Planas, J. and Elices, M. ‘Size effect and boundary conditions in the
Brazilian test: theoretical analysis’, Materials and Structures, 32 (1999) 437-444.
[10] RILEM (1985) ‘Determination of the fracture energy of mortar and concrete by means of three-
point bend tests on notched beams’, Materials and Structures, 18, 285-290 (1985) (RILEM Draft
Recommendation, TC 50-FMC Fracture Mechanics of Concrete.)
[11] ASTM E 399 Standard Test Method for Linear-Elastic Plane-Strain Fracture Toughness KIc of
Metallic Materials, American Society For Testing And Materials, Philadelphia, PA.
[12] Guinea, G. V., Pastor, J. Y., Planas, J. and Elices, M. ‘Stress intensity factor, compliance and
CMOD for a general three-point-bend beam’, Int. J. Fracture, 89 (1998) 103-116.
[13] Guinea, G.V., Planas, J. and Elices, M., ‘A general bilinear fitting for the softening curve of
concrete’, Materials and Structures, 27 (1994) 99-105.
[14] Planas, J. , Guinea, G.V. and Elices, M. ‘Size Effect and Inverse Analysis in Concrete Fracture’,
International Journal of Fracture, 95 (1999) 367-378.
[15] Planas, J., Guinea, G.V. and Elices, M. “Standard Test Method for Bilinear Stress-Crack Opening
Curve of Concrete,” Proposal submitted to ACI Committee 446, (2002), revised (2005).
[16] Fathy, A.M., Sanz, B., Sancho, J.M. and Planas, J. ‘Modelización de hormigones de alta
resistencia utilizando elementos finitos con fisura cohesiva embebida’, Anales de Mecánica de la
Fractura, 24 (2007) 289-294. (In Spanish.)
[17] Elices, M., Guinea, G. V. and Planas, J. ‘On the measurement of concrete fracture energy using
three point bend tests’, Materials and Structures, 30 (1997) 375-376.

29

You might also like