Vous êtes sur la page 1sur 10

Materials and Structures/Matériaux et Constructions, Vol.

33, July 2000, pp 381-390

Round-Robin analysis of over-reinforced concrete beams –

SCIENTIFIC REPORTS
Comparison of results
J. G. M. van Mier1 and J. P. Ulfkjær2
(1) Delft University of Technology, Faculty of Civil Engineering and Geo-Sciences, Microlab, P.O. Box 5048, 2600 GA Delft, The
Netherlands
(2) Joint Research Centre, ISIS-SMU, I-21020 Ispra (VA), Italy

Paper received: December 16, 1999; Paper accepted: February 4, 2000

A B S T R A C T R É S U M É

Following the Round-Robin test on compressive À la suite des essais croisés inter-laboratoires sur le
softening, RILEM TC 148-SSC (“Strain Softening of ramollissement en compression, le TC 148-SSC de la
Concrete”, see [1]) has proposed a second Round- RILEM a proposé une deuxième série d’essais croisés
Robin on the application of strain-softening data from concernant l’analyse des poutres en béton armé, basée sur
simple uniaxial compression tests for the analysis of over- les résultats de ramollissement issus d’expériences faites en
reinforced beams. The beams were tested at Aalborg compression simple. Ces expériences ont été effectuées en
University in Denmark in 1996. The Round-Robin is 1996 à l’Université d’Aalborg, au Danemark. Les essais
carried out in collaboration with ACI-ASCE committee croisés inter-laboratoires s’opèrent en collaboration avec le
447 (“Finite Element Analysis of Reinforced Concrete comité ACI-ASCE 447 (Analyse des structures en béton
Structures”). The results of the beam tests have been armé par éléments finis). Tandis que les résultats expéri-
kept confidential, and researchers were invited to analyze mentaux sont restés confidentiels, les chercheurs ont été
the beams given the compressive and tensile properties invités à analyser les poutres en tenant compte des proprié-
of the concrete and the reinforcement. In total six con- tés en tension et en compression du béton et de l’armature.
tributions were submitted, which were compared to the Au total, six articles ont été soumis et leurs conclusions ont
exper imental results in a workshop at the 3rd été comparées aux résultats expérimentaux. Cette compa-
International Conference on Fracture Mechanics of raison a été présentée lors de la Troisième Conférence sur la
Concrete and Concrete Structures, in Gifu, Japan in « Mécanique de Rupture du Béton et Constructions en
October 1998. In this report the outcome of the differ- béton » (FraMCoS), qui a eu lieu à Gifu (au Japon) en
ent analyses are summarized and the results are com- 1998. Dans ce rapport, les résultats des analyses sont
pared to the experimentally obtained peak loads, ductil- résumés et comparés aux résultats expérimentaux, tels que
ity and size effect. le chargement maximal, la fragilité et l’effet d’échelle.

1 INTRODUCTION as well has become an important issue since the introduc-


tion of numerical methods in engineering. The use of
In model codes formulas and recommendations are finite element methods allows for a description of the
included for the stress-strain behaviour of concrete sub- complete fracture behaviour of concrete structures. The
jected to uniaxial compression. Sometimes a bi-linear basic assumption is that the structure can be broken
diagram is proposed, but also more accurate formulations down in a number of elementary building blocks. These
appear based on the measured shape of the stress-strain building blocks include the constitutive laws for the con-
curve. The question is whether the information included crete, the steel and the bond zone between steel and con-
in the model codes is sufficiently accurate to warrant a crete. Because concrete is a quasi-brittle fracturing mater-
reliable analysis of reinforced concrete structures. The ial, the constitutive laws should include a reliable fracture
demand for accurate constitutive models, not only under criterion. The development over the years is quite well
uniaxial compression but under generalized triaxial stress known. Starting from classical strength of materials, via

Editorial Note
Prof. Jan G. M. van Mier is a RILEM Senior Member and a Member of the Management Advisory Committee, as well as RILEM's National
Delegate for the Netherlands. He also participates as a Corresponding Member in RILEM TC EAS 'Early age shrinkage induced stresses and cracking in
cementitious systems'.

1359-5997/00 © RILEM 381


Materials and Structures/Matériaux et Constructions, Vol. 33, July 2000

plasticity theories, we have now arrived in the field of that this process also takes place in a beam loaded in
fracture mechanics. For tensile fracture, Hillerborg and bending. The energy dissipation in the compression
co-workers [2] proposed a variant to the well-known zone must be similar to what happens in a concrete
Dugdale-Barenblatt plastic crack tip model [3, 4]. Instead prism loaded in uniaxial compression in a rigid testing
of using a constant yield stress over a frontal process-zone machine. The main difference is however that the
of the crack tip, it was proposed that the softening behav- boundary conditions are relatively quite well known in
iour measured in a uniaxial tension test should be used as terms of boundary friction and rotational freedom of the
a closing pressure on the crack tip. However, the question loading platens in the ‘simple’ uniaxial compression test.
remains if the model is physically sound. Close scrutiny In the reinforced beam this is not so much the case, and
of uniaxial tensile tests has shown that the fracture some assumptions must be made about these effects.
process is dependent on the boundary conditions adopted Therefore making a proposal for a test method for mea-
in the test [5]. A main characteristic of the tensile soften- suring strain softening under uniaxial compression
ing behaviour is that all tensile deformations tend to (including all the problems caused by localization in the
localize in a narrow zone (ultimately a crack) after the post-peak regime), must be accompanied by verification
maximum stress has been exceeded. In 1984 Van Mier [6] analyses whether a new method would work or not.
showed that a similar localization of deformations occurs Thus, the main question to be posed is: “How should a
under uniaxial compression. This means that fracture compressive stress-strain diagram be measured to obtain the most
mechanics principles also would apply in compression, optimal result in structural engineering?” This is the main
but it is hard to imagine a shear band propagating and a reason for initiating the Round-Robin test [1] and the
frontal ‘process zone’ with a closure pressure similar to current Round-Robin analysis on compressive softening
the compressive softening diagram. In fact, such an anal- by the RILEM committee 148 SSC. From the results of
ogy with the tensile fracture model of Hillerborg, and the the Round-Robin test [1], it was confirmed that local-
plastic crack tip model of Dugdale/Barenblatt, would be ization of deformations occurs in the descending branch
the logical consequence of the above line of reasoning to under uniaxial compression. Moreover, it was confirmed
arrive at an extension for the Fictitious Crack Model for that the use of friction reducing measures, such as com-
compression. Based on the observations by Van Mier [6] binations of Teflon and grease layers between the speci-
a compressive model with localization was proposed by men and the steel loading platens of the compressive test
Hillerborg [7] for computing the rotational capacity of machine lead to a constant compressive strength, irre-
reinforced concrete beams failing in compression. The spective of the specimen slenderness. However, when
model was extended by Markeset [8], and more models friction-reducing measures are used, the post-peak
try to incorporate the localization phenomenon in com- behaviour becomes more brittle, and the softening curve
pression. becomes steeper. What is the effect of this on the out-
To the authors’ opinion, fracturing of concrete is a come of structural analyses? Tests with 0.25 ≤ h/d ≤ 2.0
process that may take different forms in structures of dif- performed by Van Vliet & Van Mier [14] showed that
ferent size and shape. The simple compression test the scatter in results decreased when longer specimens
between rigid loading platens must also be regarded as a were selected. The scatter band of the stress-strain curves
structural test. Kotsovos [9] clearly showed that the type was more narrow for 200 mm high specimens than for
of loading platens used in the compression test affects the 50 mm and 25 mm high specimens. The reason can only
peak load and the descending portion of the stress-strain be guessed. It should be noted however that a careful
curve. As mentioned, Van Mier [6] showed localization specimen manufacturing method is essential to obtain
of deformations through the dependency of softening on low scatter results in the first place.
specimen slenderness in uniaxial compression. The post- The above overview indicates the reason for the
peak stress-deformation diagram is independent of speci- Round-Robin analysis on over-reinforced beams. It was
men length, as was confirmed by many others, e.g. [1, 10, attempted to design a beam that would fail in compres-
11]. Vonk [10] showed that the post-peak behaviour also sion. In this way an excellent bench mark would become
depends on the rotational freedom of the loading platens, available to check whether the compressive strain-soft-
quite similar to the effect observed in tension [12]. Jansen ening curve from either low friction compression tests or
& Shah [11] extended the range of slenderness to h/d = 6 from high friction compression tests (between steel
and demonstrated that with a specially designed control platens) would be the better choice for a new standard
system stable post-peak behaviour - irrespective of the test for measuring strain-softening.
large energy release leading to global snap-back behaviour In this paper a short summary of the Round-Robin
- could be obtained. The independence of post-peak problem is given. The beam size and geometry, rein-
stress-deformation diagrams on slenderness was con- forcement layout, as well as the compressive strain-soft-
firmed even for these very slender specimens. ening properties and steel properties are described. After
Among the first to establish a relation between crack- that a brief overview of the main results from the beam
ing and the shape of the stress-strain curve in compres- tests is included. Next, a summary is given of the com-
sion were Hsu et al. [13]. A continuously increasing putational models used in the six submitted analyses.
crack density was measured in specimens loaded up to a Thereafter the main results (i.e. peak loads, mid-span
certain compressive level, and around peak-stress the def lections at peak load, size/scale effect on peak load,
cracks started to form continuous patterns. It is argued and post-peak ductility of the beam) are compared to the

382
Van Mier, Ulfkjær

results from the experiments.


Finally, the consequences of
this Round-Robin on the
new test standard for com-
pressive strain-softening will
be debated.

2. SUMMARY OF
ROUND-ROBIN
PROBLEM
The experiments for the
Round-Robin analysis on
over-reinforced concrete
beams were carried out at
Aalborg University in
Denmark, [15]. Next to the
beam tests, uniaxial compres-
sion tests, uniaxial tension
tests and three-point-bend-
ing tests were conducted to
characterize the three mate-
rials used in the experiments. Fig. 1 – Schematic drawing of the beams for the Round-Robin analysis. The large beam is twice the
size of the small beam. In the Figure the dimensions of the small beam are given by (S), those for the
The different concretes were large beam by (L). The cross-sections are not drawn to scale. Note that the upper reinforcement is
a Normal Strength Concrete interrupted in the mid-span region to enforce failure of the concrete compressive zone.
(NSC), a High Strength
Concrete (HSC) and a Fibre
Reinforced High Strength Concrete (FRHSC). Beams
Table 1 – Compressive strength for the three different
of two different sizes were tested, namely a small beam concretes tested
with a span of l = 3600 mm, and a large beam with span
of l = 7200 mm. The cross-section of the beams was rec- Type of Boundary condition fc [MPa] (st.dev. [MPa])
tangular. The size was 100 mm wide and 200 mm high concrete in test
for the small beam and double that size for the large NSC low friction 22.75 (0.42)
beams. The beams were loaded in four-point-bending, high friction 24.28 (0.22)
as shown in Fig. 1. HSC low friction 117.74 (10.91)
The distance between the two loads was a = 600 and high friction 162.69 (7.17)
1200 mm for the small and large beams respectively. The FRHSC low friction 114.41 (4.43)
main reinforcement consisted of three layers of stranded high friction 130.12 (2.30)
wires (small beams) or ribbed bars (in the large beams).
Along the top of the beams also two bars were placed,
but these were interrupted at the mid-span region to To illustrate the effect of boundary conditions on the
enforce compressive failure in the concrete compression compressive stress-strain curves two diagrams for high
zone. Moreover, modelling the beam would be simpler friction platens (steel platens) and low friction platens
as well. The parts of the beam between the support and (Tef lon and grease sandwich between steel loading
the load were reinforced with stirrups to avoid shear fail- platen and concrete specimen) are shown in Fig. 2.
ure of the beams. The results of Fig. 2 are representative for all con-
The invitation to the Round-Robin analysis, [16], cretes tested. A higher strength is measured under high
contained all the information to carry out an analysis. friction conditions, whereas a more brittle behaviour is
The concrete and steel properties were included, as well measured under low friction conditions. For high
as the beam geometries. As the tests were intended for a strength concrete it was not possible to measure a soften-
comparison between compressive softening diagrams ing branch under low friction conditions; the material
determined under low friction and high friction loading was very brittle indeed and the test-control system
platens, stress-strain diagrams according to these two should be modified in order to capture a full record of
experimental conditions were specified. It was left to the the post-peak softening behaviour [11].
analyst to make the choice for the stress-strain diagram The tensile strength and fracture energy were
that was expected to give the best result. Table 1 shows retrieved from the uniaxial tension tests and the three-
the peak compressive strengths from the low and high point-bend tests. In Table 2 the tensile strength and frac-
friction compression tests on cylinders with 200 mm ture energies are listed. The modulus of elasticity of the
length and a diameter of 100 mm. steel reinforcement was 2.22 105 MPa, the yield strength

383
Materials and Structures/Matériaux et Constructions, Vol. 33, July 2000

Table 2 – Tensile strength and fracture energy


Type of type of test*) ft [MPa] (st.dev.) Gf [N/m] (st.dev.)
concrete
NSC UT 1.89 (0.33) 97.1
TPB 3.18 (0.56)**) 85.0 (11.9)
HSC UT 4.83 (0.47) 112.0
TPB 8.76 (1.10)**) 125.0 (45.1)
FRHSC UT 7.14 (0.21) -
TPB 16.7 (0.22)**) 3780 (400)
*) UT = uniaxial tension test; TPB = three point bending test;
**) indicates bending tensile strength.

was 650 MPa. More detailed information about the


material properties can be found in [15] and [16].

3. OUTCOME OF THE BEAM TESTS


Fig. 2 – Compressive stress-strain diagrams for NSC for high
friction and low friction loading platens, after [16]. A full overview of the results of the beam tests is given
in [15]. The main results are shown in [17], and for rea-
sons of completeness, the main load-deflec-
tion diagrams are gathered in Fig. 3 for
NSC (a,b for small and large beams respec-
tively), for HSC (c) and for FRHSC (d).
The displacement is the beam def lection
measured in the middle of the beam, the
load is the total load on the beam. The
beam tests were repeated three-fold; the
three lines in each of the diagrams of Fig. 3
show the individual load-def lection dia-
grams. It was not attempted to translate the
results to average curves. The large and
small NSC beams showed a rather ductile
behaviour in comparison to the HSC small
beams. In the latter case, the load-deflection
diagrams were almost straight lines up till
peak, after which a severe snap-back behav-
iour was obser ved. Only part of the
descending branch was measured for one of
the HSC small beam tests, see Fig. 3c. The
fibres in the FRHSC made quite some dif-
ference, and the ductility was improved
considerably in comparison to the HSC
small beams. This can be seen from a com-
parison of Figs. 3c and 3d. The peak-loads,
peak def lections, bending moments and
rotations at peak are all gathered in Table 3.
The results of Table 3 form the main mater-
ial for comparison with the submitted
analyses.

4. SUBMITTED CONTRIBUTIONS
As mentioned, in total six contribu-
tions were received, namely from Bascoul
Fig. 3 – Load-deflection curves for the beam tests on normal strength concrete et al. [18], Kang et al. [19], König et al. [20],
small beams (a), and NSC large beams (b), HSC small beams (c) and FRHSC Kotsovos [21], Légeron et al. [22] and Ožbolt
small beams (d), after [17]. et al. [23]. Légeron et al. submitted their

384
Van Mier, Ulfkjær

Table 3 – Peak characteristics for the beam tests, after [17]


Beam type Upeak [mm] Ppeak [kN] αpeak [mm-1] Mpeak [kNm]
NSC Small 21.55 16.92 1.69E-05 25.39
NSC Small 20.49 15.94 1.58E-05 23.90
NSC Small 20.96 16.37 1.65E-05 24.56
Mean 21.00 16.41 1.64E-05 24.61
FRHSCa 52.07 66.90 4.31E-05 100.3
FRHSCb 54.15 63.45 4.54E-05 95.17
FRHSCc 48.97 59.93 4.03E-05 89.90
Mean 51.73 63.42 4.29E-05 95.12
HSCa 57.25 54.49 4.39E-05 81.72
HSCb 52.88 55.48 4.29E-05 83.21
HSCc 57.04 59.18 4.64E-05 88.77
Mean 55.72 56.38 4.44E-05 84.57
NSC large 42.89 60.00 1.73E-05 90.01
NSC large 47.61 61.68 1.93E-05 92.50 Fig. 4 – Constitutive laws from the microplane model used by
NSC large 52.88 70.65 2.18E-05 106.0 Ožbolt et al. [23].
Mean 47.79 64.11 1.95E-05 96.17
the same as that in the surrounding concrete. In the analy-
sis successive equilibrium states are defined, increasing step
results at a very late stage when the experimental results by step the maximum concrete strain.
were released. Not everyone analyzed all the different Kang et al. [19] use a fibre beam element. Each beam
cases, but a clear view on several influence factors can be element is layered. The element is based on the classical
retrieved from this material. Moreover, König et al. sup- assumption that plane sections remain plane. Only half
plied some additional results after the workshop, which of the beam is modeled using four elements only. High
may be helpful in coming to a sound conclusion. First an accuracy and robustness is claimed for the approach that
overview is given of the different methods of analysis was used.
adopted by the different contributors, as well as the con- Ožbolt et al. [23] and Kotsovos [21] both use a com-
stitutive laws for compressive behaviour that were used. plete three-dimensional finite element code. Brick ele-
ments (eight-node solid elements) are used. Ožbolt uses
a code called MASA3, which uses the microplane model
4.1 Models used as constitutive law. Only one fourth of the beam is mod-
eled due to symmetry considerations. Steel bars are
All models used are based on some numerical modeled as truss elements. Kotsovos uses a model with
method. König et al. [20], used the so-called BDZ model 27-node brick elements. Steel is modeled by means of
(‘Biegedruckzonen modell’) in combination with the line elements without transversal stiffness.
Compression Damage Zone model of Markeset [8]. In
this (analytical) model localization in compression is
included by assuming different constitutive models over 4.2 Constitutive models for compression
certain parts of the compressed area. The damage zone
(unconfined), damage zone with activated transverse The data for the compressive stress-strain behaviour
reinforcement (or confined damage zone) and undis- were given in the form as shown in Fig. 2 for NSC. The
turbed zones (elastic) are distinguished. The model analysts were free to take any constitutive law that would
requires identification procedures for all the constitutive represent this material behaviour.
laws in these three regimes. Kang et al. [19] modeled the large NSC beam only.
Légeron et al. [22] have used two different approaches: They use the Kent and Park model [24] for the concrete
a curvature integration method (CIM) and a simplified compressive behaviour. A parabola from the origin to
finite element analysis (SFEM). In CIM, the concrete sec- the peak, followed by a linear softening branch until the
tion is divided into layers and steel is condensed in steel residual level of 0.2fc is reached characterizes the model.
layers. The moment curvature relation is determined by For larger strains the softening branch then follows a
calculating the depth of the neutral axis for a certain curva- constant stress. The compressive strength was taken
ture from equilibrium considerations. The SFEM is a from the low friction tests, namely fc = 22.75 MPa, at a
finite element program (EFICOS developed at LMT in strain at peak of 0.0014. The break point in the descend-
Cachan) that uses multilayer beam elements. Each layer is ing branch where the linear falling part joins the hori-
either a concrete layer or a homogenized steel-concrete zontal residual stress is characterized by the coordinate
layer. Plane sections are assumed to remain plane. (0.2 fc, εcu ). The value for εcu is varied in the analyses
Bascoul et al. [18] also use the finite element method from 0.00575, 0.008 to 0.10. A lower value implies a
with multilayered short elements. The moment-curvature more brittle behaviour.
relation is calculated under the assumption that plane sec- Ož bolt et al. [23] used the microplane model for
tions remain plane, and that strain in the reinforcement is concrete, [25]. The low friction results were used, and

385
Materials and Structures/Matériaux et Constructions, Vol. 33, July 2000

by taking directly the stress-strain diagrams specified in


the invitation to the Round-Robin Analysis. Bascoul et
al. performed the analyses both for the high friction and
low friction compression results.
In the two models used by Légeron et al. [22], two
different constitutive models are used. In the CIM model
the stress-strain curve proposed by Cusson and Paultre
[27] is adopted. The stress in the ascending branch
depends on the strain following a power law. The
descending portion of the curve is modeled by means of a
formula of Fafitis and Shah. The parameters are directly
fitted to the given stress-strain data. In the SFEM analysis
a damage model proposed by Laborderie is used [28]. In
Fig. 5 – Constitutive laws for the CIM and SFEM models used by
this damage model six parameters must be identified. In
Légeron et al. [22]. Fig. 5 a comparison between the CIM and SFEM consti-
tutive laws and the high friction compression data is
the resulting stress-strain curves for the three concretes given. Curiously a 0.95 correction factor was applied (the
from the microplane model are shown in Fig. 4. A direct complete curve was scaled) to account for the small size
observation is that the models for HSC and FRHSC of the specimens tested. Note that the size of the com-
results are too ductile in comparison to the experimental pression zone in the beams should be of the same size of
stress-strain data for these materials. Note that the high the prisms for the uniaxial compression tests, in particular
strength stress-strain curves for high friction uniaxial for the small beams, but probably also for the larger
compression tests can be seen in Fig. 5. beams in an advanced stage of post-failure behaviour.
Kotsovos [21] is not very explicit about the constitu- König et al. [15] use in their model the Eurocode 2
tive equations used in his finite element code, and refers formulation for the ascending portion of the stress-strain
to [26] only. Interestingly, however, he also presented curve. The peak strength is the mean value between the
results from analyses according to the Eurocode. In these high friction and low friction test results. The strain at
latter analyses, he varied the ultimate strain of concrete peak is defined according a formula of Popovics from
in the compression zone, i.e. from 0.002 to 0.0035. 1973. For the post-peak behaviour the CDZ model of
Bascoul et al. [18] used, as mentioned, a multilayer Markeset is used. After the FraMCoS workshop, König
model. A uniaxial state of stress is assumed to be present et al. performed a number of additional analyses where
in each layer. The compressive behaviour is incorporated the low friction stress-strain diagrams and the high fric-
tion diagrams were used respectively. The results of these
analyses have been incorporated in this
overview as well.

5 COMPARISON WITH THE


EXPERIMENTAL RESULTS
The results of the different contribu-
tions are all given in the respective papers
[18-23], whereas the experimental results
are presented in [17]. For a first compari-
son a few data have to be considered only,
namely the peak load of the beams, the
deflection at mid-span at peak, the slope of
the softening curve of the load-def lection
diagram and the size effect of the large and
small beams of NSC. The additional results
of König et al. have been included for the
comparison of peak loads only.

5.1 Peak-loads and size effect


The peak loads could all be retrieved
from the papers submitted to the FraMCoS
Fig. 6 – Comparison of beam strength from the various analyses with the experi- workshop. In Fig. 6 the ratio of calculated
mentally determined peak loads. peak load (Pcalc ) over experimental peak load

386
Van Mier, Ulfkjær

Another general observation is that the


results for the beams made of the more
extreme materials HSC and FRHSC gave
the largest deviations from the experiments.
One of the reasons might be that the con-
stitutive laws in the models are based on
normal strength concrete data from past
experiments. The models might therefore

ˇ
be less suitable for the new high strength
materials, but this is just speculation.
From a simple analysis, i.e. by applying
simple elasticity or plasticity theories, it
can be deduced that the strength of the
large beams should be four times as large as
the load that can be carried by the small
beams. From the experiments on normal
strength concrete a ratio of 3.91 was mea-
sured. This reduction is what normally is
referred to as size/scale effect. In Fig. 7 the
Fig. 7 – Size effect from analyses and experiments.
deviations found in the different analyses
are shown. In general the deviations are
(Pexp) is shown for the beams made of three different con- within 10%, except for the SFEM analysis of Légeron et
cretes, and for two different sizes. In general the predicted al. [22]. For the conventional concrete it can be con-
maximum loads are within 15 % of the experimentally cluded that the results are quite consistent, and most
determined maximum load. Exceptions are the results from models are capable of representing the scale difference
the analyses by Kotsovos. The reason is not clear because no adequately. Of course it has to be seen how the scale
details were given of the constitutive model. Kotsovos var- variation works out for different materials, i.e., the HSC
ied the maximum concrete compressive strain in so-called and the FRHSC. Note that relative results are shown
‘code analyses’, and it was found that the beam strength only. Fig. 7 does not imply that absolute values of failure
would increase considerably when 0.35% was selected loads are correct as can be seen from Fig. 6.
instead of 0.2%. The code analyses were closer to the
experimental (i.e., for the code analyses Pcalc/Pexp = 0.97,
0.91, 1.06 and 0.95 for the NSC small beam, the NSC large 5.2 Beam deflection at peak load
beam and the HSC and FRHSC small beams respectively,
for 0.35% maximum compressive strain, see the encircled The deflection at mid-span of the beams at peak-load
data in Fig. 6). The results of Bascoul et al., König et al. and is generally more difficult to predict. All the non-linear-
Kang et al. are of interest because a variation of input para- ities from tensile cracking and crushing in the compres-
meters was studied. Bascoul et al. analyzed the beams both sion zone must be captured adequately. The variation in
for low friction and high friction concrete properties. It is results is found to be much larger than for the strength.
interesting to note that the low friction results always Fig. 8 shows the results from the different analyses. The
underestimate the experimental results of the beam tests deflection at peak from each analysis is plotted relative to
(up to 15% lower), whereas the high friction results are up the experimental outcome. Obviously the average
to 16% higher. This would suggest, without being to hastily deflection is taken from the experiments (Table 3), and
by to draw conclusions, that the truth lies somewhere in there is some scatter in those values as well. For the gen-
between of these two extremes. König et al. show the effect eral trend this is however not very important because the
of varying the compressive stress-strain diagram from low differences are larger in the analyses. When we first con-
friction and high friction tests as well. Their analyses show sider the conventional normal strength concrete, it can
the same tendency as was found by Bascoul et al. The beam be seen that most results deviate between 10% and 20%
strength calculated with the high friction stress-strain dia- from the experimental values.
grams is always higher than that found in the analyses with The high friction results of Bascoul et al. [18] are in
the low friction compressive stress-strain diagram. Where slightly better agreement with the experiments than the
the results of Bascoul et al. show a more or less similar low friction values. Also the results of Kang et al. indi-
increase of calculated beam strength for all different beam cate that the computational results improve when the
combinations, the relative increase in the analyses of König post-peak ductility in the compressive softening diagram
et al. is larger for HSC and FRHSC. Kang et al. varied the increases. In general it must be concluded, however, that
slope of the compressive softening diagram. They found the experimental mid-span def lection is not very well
that the peak load would be between 25% lower and 1.5% predicted by most models. Again, the largest deviations
higher than the experimental load on the large beams with are found for the new materials HSC and FRHSC,
increasing ductility of the compressive softening diagram. which suggests that more work is needed to improve the
The peak strength was kept constant in the analyses. constitutive models.

387
Materials and Structures/Matériaux et Constructions, Vol. 33, July 2000

indicated with the symbol w, and the definition is


shown in the small diagram at the right of Fig. 9.
The smaller the difference between w at 75% and
50% of peak-load, the steeper the descending
branch. Negative values of w indicate snap-back
behaviour. A comparison with experiments was
made for the NSC analyses on small and large
beams only. The comparison was not made for
the FRHSC and the HSC because the compres-
sive softening diagrams for these materials were
not complete in the invitation to the Round-
Robin analysis [16]. As it turns out only the post-
peak deflection from the small beam analyses by
Bascoul et al. [18], Légeron et al. (SFEM) [22] and
Ožbolt et al. [23] could be used for comparison
with the experimental data. For the large beam
NSC analyses, also the results of Kang et al. [19]
could be compared to the experiments, but at
0.75 Pmax only. The small beam results are gath-
ered in Fig. 9, the large beam analyses in Fig. 10.
Let us first consider the small beams. On the
vertical axis the post-peak deflection w is plotted;
along the horizontal axis, the load levels at which
Fig. 8 – Comparison of deflection at peak load from analyses and experi- w is shown, i.e. 0.75 Pmax and 0.50 Pmax. A steep
ments. increase between 0.75 and 0.50 implies a rela-
tively large ductility. Obviously the experiments
[15] showed a larger ductility than was found in
any of the analyses. The SFEM results of Légeron
come closest to the experiments. Ož bolt and
Bascoul predict a too brittle behaviour. In the
analyses of Bascoul the concrete compressive
stress-strain diagram was varied, i.e. both the low
friction and high friction diagrams (Fig. 2) were
used, but this had a minor effect on the predicted
values only.
The situation becomes more severe for the
large NSC beams, as can be seen from Fig. 10.
Again, the SFEM analysis comes closest to the
experiments [15], but Ožbolt and Bascoul pre-
dict an even more brittle response than was
found in the small beam analyses. For both com-
pressive diagrams (LF and HF), Bascoul et al. [18]
predict snap-back behaviour, i.e. w < 0. From the
analyses by Kang et al. [19] only values at
0.75Pmax were available. Their analyses showed a
larger ductility with increasing maximum strain
Fig. 9 – Comparison of post-peak deflection w at mid-span from the analy- in the compressive softening diagram, which was
ses and the experiments on small beams of normal strength concrete. to be expected.

5.3 Ductility 6. DISCUSSION: EFFECT ON NEWLY PRO-


POSED TEST STANDARD
In the present beams the post-peak behaviour is com-
pletely governed by the behaviour of the concrete in the The results can be grouped in four categories: (1)
compression zone. The beams are highly over-reinforced, peak load, (2) deflection at peak load, (3) post-peak duc-
and the strains in the reinforcement do not exceed the tility, and (4) the shape of the load-deflection diagram. It
elastic limit. Because the results were submitted in widely is not possible to select the ‘best’ analysis. All results have
varying formats, it was decided to retrieve from the dif- some advantages and some drawbacks. Because the HSC
ferent contributions the post-peak deflection at 75% and and FRHSC analyses proved to give the largest devia-
50% of the maximum load. This post-peak deflection is tions from the experiments, the discussion will focus on

388
Van Mier, Ulfkjær

come. As far as the post-peak deflection is con-


cerned, however, the computed values are much
lower than the experimental data, both for the
HF and LF analysis. Thus, the results are not
decisive if it comes to selecting a standard for a
compressive softening test.
The results of the compressive Round-Robin
test by RILEM TC 148-SSC [1] indicated that
low friction loading platens would yield results
with a smaller scatter. From the experimental
point of view this would be sufficient reason
already to decide to use the Teflon sandwiches as
interlayer between steel and concrete. The reason
is that more accurate data would be preferred
against slightly higher costs. The costs for making
the specimens is larger not only because the use
of Teflon, but also because the specimens must
be ground flat and plan-parallel to a high degree
of precision. Perhaps this is already the main rea-
son for obtaining low-scatter stress-strain curves.
The results from the experimental Round-
Robin as well as the findings from this Round-
Fig. 10 – Comparison of post-peak deflection w at mid-span from the Robin analysis have been incorporated in the
analyses and the experiments on large beams of normal strength concrete. new RILEM test recommendation for the mea-
surement of strain softening under uni-axial
the NSC small and large beams only. The most impor- compression [29].
tant conclusion which can be drawn from the HSC and From the analyses it can further be concluded that
FRHSC results is that still quite some work is needed to quite some work is needed to come to good predictions
come to constitutive models with improved predictive of structural behaviour, even for a ‘simple’ case as four-
qualities for these new high strength materials. point bending. Although large improvements were made
For NSC the analyses of Légeron et al., in particular over the past decade, there is still room for further gains,
the SFEM analysis seems to have merit. The peak loads in particular when it comes to the applications of new
and def lection at peak are rather well reproduced, materials like high strength concrete and fiber reinforced
although the latter exceeds the experimental value by materials. Moreover, there is room for improvements at
more than 15%. The prediction of the post-peak deflec- the computational front. Kang et al. investigated some of
tion is quite accurate for the SFEM model. It should be the computational issues involved with the beam prob-
mentioned, however, that a curious reduction of 0.95 was lem. Changing the number of integration points in the
made to the compressive stress-strain curve. The authors critical element had a large effect on the outcome
mention that this was done because of the small size of the already. At the same time, the procedures for measuring
prisms used for measuring the compressive softening dia- softening under compression for such materials must be
gram. It should be realized, however, that the compressive improved. Especially when snap-back behaviour is
zone, especially in the small beam, is not very much expected, as for many of the high strength materials, a
larger, and towards the end of the beam test even much critical re-evaluation of test procedures should be carried
smaller than the dimension of the prism used for deter- out. For example, do we still need large specimens when
mining the compressive stress-strain diagram. The size more homogeneous high strength materials are used, or
effect predicted by the SFEM method showed the largest can the size be reduced to allow for a more simple
deviation from the experiments (Fig. 7). method of measuring softening?
The results of Ožbolt et al. [23] appear to be most con- Blind Round-Robin analyses like the present beam
sistent. Their results always underestimate the experi- problem are to the authors’ opinion an excellent tool for
ments, but the same deviations are found for all materials improving the abilities of numerical tools and may indi-
and beam sizes. The largest deviations are found in the cate the best possible standard test. The present beam
deflection at peak load and the post-peak deflections, but problem is a small step in the right direction.
again the amount to which the experiments are underesti-
mated remains more or less the same.
Kang et al. [19], König et al. [20] and Bascoul et al. 7. CONCLUSIONS
[18] make an interesting variation of the effect of the
ductility in the compressive stress-strain diagram. A From the Round-Robin analysis not only the abili-
more ductile compressive stress-strain diagram leads to a ties of existing numerical programs to handle compres-
higher failure load. Also the def lection at peak load sive failure are elucidated, but the earlier obtained con-
increases, and comes closer to the experimental out- clusion that a modification of the existing standard test is

389
Materials and Structures/Matériaux et Constructions, Vol. 33, July 2000

needed to measure the softening diagram of concrete capacity and rotational capacity of reinforced concrete beams’,
under uni-axial compression [1], seems more securely Eng. Fract. Mech. 35 (1990) 233-240.
[8] Markeset, G., ‘Failure of concrete under compressive strain gradi-
founded. ents’, PhD thesis, Norges Tekniske Ho/ yskole, Trondheim,
The following conclusions can be drawn from the Norway (1993).
results presented in this paper: [9] Kotsovos, M. D., ‘Effect of testing techniques on the post-ulti-
(1) Using conventional models for the uni-axial compres- mate behaviour of concrete in compression’, Mater. Struct. 16
sive behaviour of NSC concrete, the peak load of over- (1983) 3-12.
[10] Vonk, R. A., ‘Softening of concrete loaded in compression’,
reinforced concrete beams loaded in four-point bending PhD thesis, Eindhoven University of Technology (1992).
can be ‘predicted’ with an accuracy of 10 to 15%. [11] Jansen, D. C. and Shah, S. P., ‘Effect of length on compressive
(2) Predicting the beam deflection at peak-load as well as strain softening of concrete’, J. Eng. Mech. (ASCE) 1 (1998) 25-35.
the ductility proved to be more difficult and generally [12] Van Mier, J. G. M., Schlangen, E. and Vervuurt, A., ‘Tensile
larger deviation from the experiments is observed. cracking in concrete and sandstone: Part 2 - Effect of boundary
rotations’, Mater. Struct. 29 (1996) 87-96.
(3) In particular the predictions for HSC and FRHSC [13] Hsu, T. T. C., Slate, F. O., Sturman, G. M. and Winter, G.,
beams seem hampered by insufficient knowledge of the ‘Micro-cracking of plain concrete and the shape of the stress-
mechanical behaviour of these new materials. Most pro- strain curve’, J. Amer. Conc. Inst. 60 (2) (1963) 209-224.
nounced seems the lack in knowledge of post-peak [14] Van Vliet, M. R. A. and Van Mier, J. G. M., ‘Experimental
behaviour. investigation of concrete fracture under uniaxial compression’,
Mech. Coh. Frict. Mat. 1 (1996) 115-127.
(4) Overestimating the ductility of the compressive [15] Pedersen, L. H., Bundgaard, K. and Morch, T., ‘Numerical and
strain-softening diagram leads to an increase of the pre- experimental investigation of compression failure of concrete
dicted failure load and the deflection at peak load. beams’, Graduation thesis, June 1996, 162 pp.
(5) In general the most consistent results are obtained [16] Ulfkjær, J. P., Van Mier, J. G. M. and Stang, H., ‘Invitation to a
when stress-strain curves from ‘low-friction’ tests are competition on modelling of over-reinforced concrete beams’,
RILEM Technical Committee 148-SSC “Strain Softening of
used. A lower bound for the beam strength and deflec- Concrete” and ACI/ASCE Committee 447 on “Finite Element
tion at peak-load is found. Moreover, the ‘trend’ from Analysis of Reinforced Concrete Structures” (1997).
analyses with the same model of beams made of different [17] Ulfkjær, J. P., ‘Experimental investigation of over-reinforced con-
materials seems to be more consistent, although the dif- crete beams of three different types of concrete and at two different
ference is hard to express in solid numbers. size scales’, in Proc. FraMCoS-3, ed. H. Mihashi and K. Rokugo,
AEDIFICATIO Publishers, Freiburg, (1998) 1253-1260.
[18] Bascoul, A., Duprat, M. and Pinglot, M., ‘Load deflection dia-
gram of over-reinforced concrete beams’, in Proc. FraMCoS-3,
ACKNOWLEDGEMENTS ed. H. Mihashi and K. Rokugo, AEDIFICATIO Publishers,
Freiburg, (1998) 1211-1222.
The authors would like to thank all the contributors [19] Kang, H. D., Spacone, E. and Willam, K. J., ‘A study of com-
pressive failure in over-reinforced concrete beams’, in Proc.
who took the trouble of analyzing the beam problem. FraMCoS-3, ed. H. Mihashi and K. Rokugo, AEDIFICATIO
Without their effort the improved insight in the use of Publishers, Freiburg, (1998) 1195-1210.
compressive softening diagrams in structural analysis [20] König, G., Meyer, J. and Sint, A., ‘Round Robin analysis on
would not have been possible. modelling of over-reinforced concrete beams - Calculation of the
load deformation behaviour of concrete beams with the BDZ
model’, in Proc. FraMCoS-3, ed. H. Mihashi and K. Rokugo,
AEDIFICATIO Publishers, Freiburg, (1998) 1241-1251.
REFERENCES [21] Kotsovos, M. D., ‘Round Robin finite element analysis of over-
reinforced concrete beams, Report, Athens University, (1997) 9 pp.
[1] Van Mier, J. G. M., Shah, S. P., Arnaud, M., Balayssac, J. P., [22] Légeron, F., Mazars, J. and Paultre, P., ‘Prediction of the behavior
Bascoul, A., Choi, S., Dasenbrock, D., Ferrara, G., French, C., of over-reinforced concrete beams with two levels of simplified
Gobbi, M. E., Karihaloo, B. L., König, G., Kotsovos, M. D., approach’, in Proc. FraMCoS-3, ed. H. Mihashi and K. Rokugo,
Labuz, J., Lange-Kornbak, D., Markeset, G., Pavlovic, M. N., AEDIFICATIO Publishers, Freiburg, (1998) 1223-1232.
Simsch, G., Thienel, K-C., Turatsinze, A., Ulmer, M., Van [23] Ožbolt, J., Li, Y.-J. and Eligehausen, R., ‘3D finite element
Geel, H. J. G. M., Van Vliet, M. R. A., Zissopoulos, D., ‘Strain- analysis of over-reinforced beams’, in Proc. FraMCoS-3, ed. H.
softening of concrete in uniaxial compression - Report of the Mihashi and K. Rokugo, AEDIFICATIO Publishers, Freiburg,
Round-Robin Test carried out by RILEM TC 148-SSC’, Mater. (1998) 1233-1240.
Struct. 30 (198) (1997) 195-209. [24] Kent, D. C. and Park, R., ‘Flexural members with confined
[2] Hillerborg, A., Petersson, P. E. and Modéer, M., ‘Analysis of concrete’, ASCE J. Struct. Div. 97 (1971) 1964-1990.
crack formation and crack growth in concrete by means of frac- [25] Ožbolt, J., Li, Y. and Kozar, I., Microplane model for concrete -
ture mechanics and finite elements’, Cem. Conc. Res. 6 (6) (1976) mixed approach’, (1997) submitted to Int. J. Solids Struct.
773-782. [26] Kotsovos, M. D. and Pavlovic, M. N., Structural Concrete: Finite
[3] Dugdale, D. S., ‘Yielding of sheets containing slits’, J. Mech. Phys. Element Analysis for Limit-State Design, Thomas Telford (1995).
Sol. 8 (1960) 100-108. [27] Cusson, D. and Paultre, P., ‘Stress-strain model for confined high
[4] Barenblatt,G. I., ‘The mathematical theory of equilibrium cracks strength concrete’, ASCE J. Struct. Div. 121 (1995) 468-477.
in brittle fracture’, Adv. Appl. Mech. 7 (1962) 55-129. [28] Laborderie, C., Phénomènes unilatéraux dans un matériau
[5] Van Mier, J. G. M., ‘Fracture Processes of Concrete - Assessment endommageable : modélisation et application à l’analyse de struc-
of Material Parameters for Fracture Models’, CRC Press, Boca tures en béton’, PhD thesis, LMT Cachan (1991).
Raton (FL), USA, 448 pp. [29] Van Mier, J. G. M. and Shah, S. P., ‘Test method for measure-
[6] Van Mier, J. G. M., ‘Strain-softening of concrete under multiaxial ment of the strain-softening behaviour of concrete under uniaxial
loading conditions’, PhD thesis, Eindhoven University of compression, RILEM Draft Recommendation, Mater. Struct. 33
Technology (1984). (230) (2000) 347-351.
[7] Hillerborg, A., ‘Fracture mechanics concepts applied to moment

390

Vous aimerez peut-être aussi