BOARD OF EDITORS
Jacky MAZARS, France
Sidney MINDESS, Canada
Hans-Wolf REINHARDT, Germany
Folker H. WITTMANN, Switzerland
SCIENTIFIC COMMITTEE – BOARD OF ADVISERS
Paul ACKER, France
Pierre-Claude AÏTCIN, Canada
Mario COLLEPARDI, Italy
Ario CECCOTTI, Italy
Félix DARVE, France
François de LARRARD, France
Francis DUTRUEL, France
Hermann W. FRITZ, Switzerland
Yeb M. de HAAN, The Netherlands
Pierre HABIB, France
Hubert K. HILSDORF, Germany
Marcel MARIOTTI, France
SCIENTIFIC COMMITTEE – BOARD OF THE CHAIRMEN
OF RILEM TECHNICAL COMMITTEES
Mark ALEXANDER, South Africa
Carmen ANDRADE, Spain
Ginette ARLIGUIE, France
Peter J. M. BARTOS, UK
Zdeněk P. BAŽANT, USA
Harald BEITZEL, Germany
Arnon BENTUR, Israel
Luigia BINDA, Italy
Ignacio CAROL, Spain
Lennart ELFGREN, Sweden
Louis FRANCKEN, Belgium
Caspar de GROOT, The Netherlands
Ahmed HAKIMI, Morocco
Charles F. HENDRIKS, The Netherlands
Jörg KROPP, Germany
Michael LACASSE, Canada
Pritpal S. MANGAT, UK
Harald S. MULLER, Germany
Dan J. NAUS, USA
Philip J. NIXON, UK
Kai ÖDEEN, Sweden
1359-5997/99 © RILEM
Federico MAZZOLANI, Italy
Hirozo MIHASHI, Japan
Antoine E. NAAMAN, USA
Adam M. NEVILLE, UK
Duncan POMEROY, UK
Pierre ROSSI, France
Vijai M. SHARMA, India
Angelo DI TOMMASO, Italy
Dionys VAN GEMERT, Belgium
Michel VALLES, France
Jan G. M. VAN MIER, The Netherlands
Yoshihiko OHAMA, Japan
Gyan PANDE, UK
Moshe PUTERMAN, Israel
Alpo RANTA-MAUNUS, Finland
Hans-Wolf REINHARDT, Germany
Ian G. RICHARDSON, UK
Raymundo RIVERA-VILLARREAL, Mexico
Walter J. ROSSITER, USA
Asko SARJA, Finland
Ulrich SCHNEIDER, Austria
Max J. SETZER, Germany
Surendra P. SHAH, USA
Jan P. SKALNY, USA
Åke SKARENDAHL, Sweden
Rupert SPRINGENSCHMID, Germany
Ferenc D. TAMÁS, Hungary
Wayne TRUSTY, Canada
Luca UZIELLI, Italy
Lucie VANDEWALLE, Belgium
Andreas WOLF, Belgium
II
PUBLISHED BY :
RILEM Publications
Sarl au capital de 50 000 F
RCS : CRETEIL B 410 772 834
N° SIRET : 410 772 834 00017
CODE APE : 221E
EDITORIAL OFFICE :
Ms Pascale Callec
RILEM Publications
F-94235 Cachan Cedex, France
Tel. + 33 1 47 40 28 80
Fax. + 33 1 47 40 01 13
E-mail :
ms@rilem.ens-cachan.fr
RILEM WEB SERVER :
http://www.rilem.ens-cachan.fr
• Internet access to this journal :
http://www.rilem.ens-
cachan.fr/matstr.html
• List of papers accepted for
publication :
http://www.rilem.ens-
cachan.fr/manus.html
TYPESET BY :
APE
F-24750 Champcevinel
PRINTED BY :
IMPRIMERIE MONTLIGEON
F-61400 La Chapelle Montligeon
COPYRIGHT © RILEM 1999
COMMISSION PARITAIRE :
1202 T 52678
DÉPOT LÉGAL : Mars 1999
DIRECTEUR DE PUBLICATION :
Michel BRUSIN
SUBSCRIPTION RATES
for Volume 32 (1999) - 10 issues a
year (800 pages per volume)
For the printed version of the
journal only: 2134.00 FRF
For both the printed version
of the journal and the on-line
version: 2534.00 FRF
Mailing costs: 250.00 FRF
Single issues of the journal
(available to subscribers
only): 250.00 FRF
Subscriptions are on an annual basis
only. Claims for issues not received
must be made within three months of
publication.
ISSN: 1359-5997
 MATERIALS AND STRUCTURES
MATÉRIAUX ET CONSTRUCTIONS
MARCH MARS
VOL. 32 – No. 216

C O N T E N T S S O M M A I R E
EDITORIAL ..................................................................................................................................................... 82
SCIENTIFIC REPORTS RAPPORTS SCIENTIFIQUES
The effects of age up to 18 years under various exposure conditions on the tensile properties
of a polypropylene fibre reinforced cement composite
by D. J. Hannant .................................................................................................................................. 83
Bond characteristics of corroding reinforcement in concrete beams
by M. S. Elgarf and P. S. Mangat ........................................................................................................................... 89
Development of a multiple linear regression model for predicting the 28-day compressive strength
of Portland pozzolan cement
by G. CH. Kostogloudis, J. Anagnostou, CH. Ftikos and J. Marinos ............................................................................. 98
Lifetime of concrete dam models under constant loads
by F. Barpi, G. Ferrara, L. Imperato and S.Valente ................................................................................................... 103
Shear stress contributions and failure mechanisms of high strength reinforced concrete beams
by S. Sarkar, O. Adwan and B. Bose ...................................................................................................................... 112
Evaluation expérimentale des caractéristiques modales et ajustement du modèle dynamique
d'un bâtiment de portiques
(Experimental evaluation of the modal characteristics of a framed building and adjustment
of the mathematical model)
par C. Genatios, G. Cascante, M. Lafuente, O. López, P. Bellan et M. Lorrain ........................................................... 117
A model for cracking in fibre-reinforced concrete structures
by P. Rossi ......................................................................................................................................................... 125
Rheological properties of SEBS, EVA and EBA polymer modified bitumens
by X. Lu, U. Isacsson and J. Ekblad ....................................................................................................................... 131
TECHNICAL REPORTS RAPPORTS TECHNIQUES
Assessment of glass fibre reinforced cement by in-situ SEM bending test
by P. Trtik and P. J. M. Bartos .............................................................................................................................. 140
Fracture assessment of a high strength structural steel on the basis of Eurocode 3
by A. Valiente, M. Elices, J. Ruiz, J. A. Alvarez, J. Gutiérrez-Solana, J. González, F. J. Belzunce and C. Rodríguez ....... 144
LOCAL MATERIALS MATÉRIAUX LOCAUX

Caractérisation des bétons de sable routiers compactés : Application au cas du Sénégal

(Characterisation of compacted road sandcretes: Study applied to Senegal)
par I. K. Cisse, M. Laquerbe, A. Gaye et M. Diene ................................................................................................. 151
RILEM BULLETIN BULLETIN RILEM

Materials and technologies for sustainable construction

reported by C. Sjöström ......................................................................................................................................... 158

1359-5997/99 © RILEM 81
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, p 82
I E
n reviewing the contents of this issue of “Materials
and Structures”, I was reminded of a recent discus-
sion that I had with some of my students. They
had argued that it was difficult to attract students to the
study of building materials because there was nothing
much new in current research on the materials of
construction; compared to other fields of engineering,
most notably those related to computer science, our
research efforts appeared to have stagnated. Indeed, if
one looks only at the titles of the current papers in this
(and in most other journals devoted to civil engineering
materials) they are much the same as those that appeared
in these journals 25 years ago. Nonetheless, as I haste-
ned to point out to my students,
things really have changed over
the past quarter century. While
the titles may be similar, the EDITORIAL
authors have changed; they lar-
gely represent a new generation of researchers, who take
a much more sophisticated approach to research, in
terms of both experimental techniques and analysis. We
are much closer to understanding the underlying pheno-
mena which govern the behaviour of modern construc-
tion materials, and we can now largely design special
materials for very specific applications.
However, I was left with the feeling that there was at
least a kernel of truth in what my students had said.
Though “Materials and Structures” continues to maintain
its very high standards in the papers that it publishes, this
is not universally true of other publications. There are
too many journals, and at least some of the material that
is published unfortunately represents “the transference
of bones from one cemetery to another”. There is
“research” that is reported that merely confirms what we
have long known, or which is so specific as to the mate-
rials and methods used that the results cannot be genera-
lized in a meaningful way. This is made worse by the
current reward systems in most universities, which tends
to favour quantity over quality.
Despite these problems, civil engineering materials
research remains alive and well, as a reading of the
insightful papers in this issue will show. We continue to
solve the problems that arise from new materials deve-
lopment, and from modern construction techniques.
What is urgently needed now is a better way of transmit-
ting what we have learned, not only to practitioners in
the field, but to our own students as well, so that they
can observe for themselves what a vital and fascinating
career awaits them.
Sidney MINDESS
1359-5997/99 © RILEM
n parcourant le sommaire de ce numéro de
Matériaux et Constructions, je me suis sou-
venu d’une conversation récente avec quelques-uns
de mes élèves. Ils faisaient valoir l’argument selon lequel il
était difficile d’attirer les jeunes vers l’étude des matériaux
de construction, en raison de l’absence de réelle nouveauté
dans la recherche. Comparés aux autres domaines de l’ingé-
nierie, et plus particulièrement ceux liés à l’informatique ou
aux méthodes de calcul, nos efforts de recherche leur semble
avoir stagné. En effet, si l’on regarde uniquement les titres
des articles de ce numéro (ainsi que dans la plupart des
autres journaux spécialisés dans les matériaux de génie
civil), ils sont quasiment identiques à ceux publiés dans ces
journaux 25 ans plus tôt. Mais j’ai
tenu à préciser à mes élèves que les
choses ont évolué réellement au cours du
dernier quart de siècle. Même quand les
titres se ressemblent, les auteurs ont
changé : ils représentent une nouvelle génération de cher-
cheurs, qui ont une approche beaucoup plus élaborée tant en
termes de techniques expérimentales que d’analyses. Nous
nous sommes rapprochés de la compréhension des phéno-
mènes qui gouvernent les matériaux de construction
modernes, et nous pouvons désormais concevoir des maté-
riaux pointus répondant à des applications spécifiques.
Cependant, je suis resté avec le sentiment qu’il y avait un
soupçon de vérité dans les propos de mes élèves. Si
Matériaux et Constructions continue à maintenir ses cri-
tères de qualité pour ses articles, il n’en va pas toujours de
même dans la presse scientifique. Trop de journaux prolifè-
rent, et une partie de la matière publiée ne représente pas
d’apport nouveau, se contentant d’exhumer d’anciennes
connaissances. Il existe une « recherche » dont le compte-
rendu ne fait que confirmer ce que nous connaissons de longue
date, ou qui est si spécifique aux matériaux et aux méthodes
utilisées qu’elle interdit toute généralisation des résultats obte-
nus. Ceci est aggravé par le système de reconnaissance actuel-
lement en vigueur dans la plupart des universités, système
qui tend à privilégier la quantité à la qualité.
En dépit de ces problèmes, la recherche sur les maté-
riaux utilisés en génie civil reste vivante et bien vivante,
comme le démontrent les articles de ce numéro. Nous conti-
nuons à résoudre les problèmes liés au développement de
nouveaux matériaux, ainsi qu’aux nouvelles techniques de
construction. Il est essentiel désormais de trouver une
meilleure façon de transmettre ce que nous avons appris,
non seulement aux praticiens dans leur domaine, mais
aussi à nos propres étudiants, pour qu’ils puissent observer
eux-mêmes quelle carrière passionnante s’offre à eux.
Sidney MINDESS
82
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 83-88
The effects of age up to 18 years under various exposure
conditions on the tensile properties of a polypropylene
fibre reinforced cement composites
D. J. Hannant
Professor, Department of Civil Engineering, University of Surrey, Guildford GU2 5XH, UK.
Paper received: April 28, 1998; Paper accepted: September 16, 1998
A B S T R A C T
An 18 year test programme has been completed on
the tensile stress-strain performance of cement compos-
ites reinforced with two types of polypropylene fibril-
lated networks. Three storage conditions have been used
- natural weathering, storage inside in laboratory air, and
under water storage. Some composites have been pre-
cracked before exposure and others left uncracked before
testing. The parameters measured included the elastic
modulus and cracking stress of the matrix, complete ten-
sile stress-strain curves and crack distribution in the
composite, and the effects of the different weathering
conditions on the tensile strength, strain to failure and
bond strength of the polypropylene reinforcement.
In general terms, it was shown that where the fibre
volume remained above the critical fibre volume, the
strength, ductility and toughness of the composite was
maintained regardless of exposure conditions for very
long periods of time. In particular, the material has
remained ductile, with a failure strain in excess of 5%
after 18 years under water which is unusual for fibre
reinforced cement composites.
1. INTRODUCTION
It is essential for all new cement based construction
materials, particularly those containing f ibres and
intended for roofing or cladding applications, that real-
time weathering data is obtained for as long a period as
possible. This is because the validity of accelerated test
procedures to predict fibre-cement interactions is always
open to question. Accelerated tests are often used to pre-
dict long term fibre degradation but the changes which
occur in the cement matrix and at the fibre-matrix inter-
face are much more difficult to accelerate because they are
very slow in natural weathering and depend on the simul-
taneous interactions of carbonation, continued hydration,
1359-5997/99 © RILEM
SCIENTIFIC REPORTS
R É S U M É
Un programme d’étude d’une durée de 18 ans vient de
s’achever sur le fonctionnement en tension des composites
ciment renforcés avec deux types de réseaux de polypropylène
fibrillé. Trois conditions de conservation ont été utilisées –
vieillissement naturel en plein air, conservation ambiante
dans le laboratoire, et conservation sous l’eau. Quelques-uns
des composites ont été pré-fissurés avant d’être exposés et
d’autres sont restés non fissurés jusqu’à l’essai.
Les paramètres mesurés ont compris le module élastique
et la résistance en traction de la matrice, des courbes de
contrainte-déformation et la distribution des fissures dans le
composite, et les effets de diverses conditions de vieillissement
sur la résistance en traction, la déformation à la rupture et la
force d’adhérence de l’armature en polypropylène. Il a pu être
démontré que lorsque le volume de fibres dépasse le volume
critique, la résistance, la ductilité et la dureté du composite
ont été maintenues quelles que soient les conditions de
l’exposition pendant de très longues périodes. En particulier,
le matériau est resté ductile, avec une déformation à la rup-
ture augmentée de 5% après 18 ans sous l’eau, ce qui est
rare pour des composites ciment renforcés de fibres.
freezing and thawing and autogenous healing which are
very difficult to accelerate accurately.
Also, with the increasing use in concrete of a variety
of types of polypropylene fibres, it is important to know
how the tensile strength and bond characteristics of the
polypropylene are likely to change with time and storage
conditions.
A long term programme of tests was therefore initi-
ated in 1979/80 to examine the tensile characteristics of
cement based sheets containing two-dimensional net-
works of polypropylene film [1]. These tests have now
been completed after 18 years and the results give useful
guidelines on the durability and stability of cement based
polypropylene fibre composites.
83
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
2. TEST PROGRAMME
2.1 Variables
Fibre cement composites often operate in the cracked
state and therefore half of the specimens were precracked
before exposure to natural weathering at the Building
Research Establishment site at Garston, Hertfordshire or
in laboratory air at the University of Surrey. The crack-
ing procedure involved the application of a tensile strain
of 5000 microstrain which resulted in 22 cracks in the
100 mm strain measuring length with an average crack
width of 7 microns after unloading to a residual strain of
about 1600 microstrain. A number of uncracked sam-
ples were stored under water at 20°C.
The two types of continuous networks of opened
polypropylene fibrillated film were used supplied by
Bridon Fibres and Plastics Ltd. Type A (code named
1952/R1/7) containing no additives other than process-
ing additives; and Type B (code named 1952/R1/9)
containing an ultraviolet stabiliser.
Test ages were 1, 7 and 13 months, 2, 3, 5, 9 or 10
years and 17 or 18 years depending on the exposure con-
dition. Eight specimens were tested for each fibre type
Fig. 1 – Averaged tensile stress-strain curves for up to 18 years of natural weather-
ing compared with 1 month water stored samples. Film type A. (a) no pre-crack-
ing; (b) pre-cracked an reloaded.
84
and exposure condition at each age, half of the speci-
mens being precracked except for the under water stor-
age condition.
2.2 Manufacture and testing of specimens
The mortar matrix consisted of ordinary Portland
cement, pulverised fuel ash and sand in the proportions
1.0 : 0.25 : 0.19 respectively with a total water/cement
ratio of 0.34 and 2.4% by weight of cement, of Melment
L10, a sulphonated melamine superplasticizer.
About 350 specimens, 300 mm long by 25 mm wide
and nominally 6 mm thick were cut from 30 separate
sheets. The sheets, which were laid up by hand, con-
tained 51 layers of film networks, 34 layers being in the
test direction and 17 layers at right-angles to this, giving
a film volume of about 6% in the test direction and 3%
in the orthogonal direction. The actual film volume (Vf)
was determined for each specimen by acid dissolution in
diluted hydrochloric acid after testing to failure.
All tests were carried out in uniaxial tension using an
Instron 1122 testing machine, the strains being measured
by a clip-on strain measuring device outputing to X-Y-Y
recorders [1]. The rate of loading was con-
trolled by the strain rate, which was about
4% per minute. All specimens which
reached 5% strain were unloaded to check
whether strain recovery varied with time
and exposure conditions. The crack spac-
ing was determined after failure, using a
microscope, and the values for crack spac-
ing were used to calculate the fibre-matrix
bond strength.
3. RESULTS AND DISCUSSION
3.1 Tensile stress-strain curves
3.1.1 Natural weathering at the Building
Research Establishment
Figure 1 shows the average stress-strain
curves for film type A at 1 month, 10 years
and 18 years of natural weathering. The
results for the uncracked material are shown
in Fig. 1(a) and for the pre-cracked compos-
ite in Fig. 1(b). There is little difference
between the curves at 10 years and 18 years
indicating that neither the matrix nor the
fibre had changed properties during the last
8 years of natural weathering including the
effects of freezing and thawing.
Failure strains of more than 5% were
maintained because the f ibre volume
remained above the critical fibre volume for
tensile strengthening even though the matrix
strength had considerably increased from 1
month. One specimen out of the eight
tested at 18 years failed at less than 5% strain
Hannant

because the fibre volume was only 5.4% in

the stress direction rather than an average of
6.4% for the other samples. The critical
fibre volume at 18 years was more than 5.6%
and hence there was insufficient fibre to sup-
port the mean matrix cracking stress of more
than 14 MPa in this sample. A significant
feature of the 10 and 18 year curves is that
the cracking of the matrix continued until
the maximum stress was reached. According
to the theory of Aveston et al. [2, 3], if the
matrix has a constant cracking stress and
there is only frictional bond with the fibres,
the multiple cracking region will be horizon-
tal, which is a reasonable approximation to
the curve after one month curing in water.
Therefore, either natural weathering has
resulted in a more variable matrix strength or
more complex bonding mechanisms exist as
proposed by Ohno [4]. The relatively steep
initial part of the curve for the pre-cracked
samples in Fig. 1(b) shows that some autoge-
nous healing had taken place allowing tensile
stresses up to 8 MPa to be sustained before
re-cracking. This was less than the initial
cracking stress of about 12.8 MPa for the
uncracked composite in Fig. 1(a). However,
after about 1% strain, the two types of curve,
whether uncracked or pre-cracked, are very
similar, showing that the films had not signif-
icantly deteriorated across cracks after 18
years of natural weathering.
For film type B, the average failure strain Fig. 2 – Averaged tensile stress-strain curves for up to 17 years of inside storage
of the composite was 3.3% but this strain compared with 1 month water stored samples. Film type A. (a) no pre-cracking;
was greatly affected by 4 out of the 8 samples (b) pre-cracked an reloaded
having film volumes of 5.2% or less, which
was less than the critical fibre volume of more than 5.7% 3.1.3 Storage under water
at 3% strain. Samples with film volumes greater than 6% Samples were only tested in the uncracked state, the
generally showed similar performance to type A samples. average curve for 5 samples containing film type A being
Thus the importance of designing the composite to con- shown in Fig. 3. Surprisingly, the shapes of the curves
tain more than the critical fibre volume at all ages in the were remarkably similar to the air stored samples in Fig.
life of the product was clearly demonstrated. 2(a) with similar cracking stresses at given strains. Thus,
the continued availability of water throughout the 18 year
3.1.2 Storage in laboratory air period had not greatly improved the tensile properties of
Figure 2 shows tensile stress-strain curves for the the matrix compared with air storage where carbonation
same composite as for the naturally weathered material was possible. An important feature of Fig. 3 is that com-
in Fig. 1. There are significant differences in shape posite failure strains remained at about 6% giving excellent
between the two figures. For instance, the composite retention of toughness and no signs of embrittlement as
cracking stress in laboratory air at 0.5% strain was 9.5 sometimes occur with glass fibre or cellulose fibre rein-
MPa (Fig. 2(a)) compared with 12.5 MPa for natural forced cements. The implication is that both bond
weathering in Fig. 1(a). Also, the less jagged nature of strength and fibre strength are relatively unaffected by very
the curves for air curing, the lack of autogenous healing long periods under water.
in Fig. 2(b), and the earlier completion of cracking in air
storage all indicate that insufficient moisture has been
present for continuing hydration of the matrix even 3.2 Elastic modulus of the composite
though there is a small increase in cracking stress com-
pared with 1 month water curing. There has been virtu- The elastic modulus of the composite in tension was
ally no change in composite properties beween 9 years calculated from an amplified initial portion of the stress-
and 17 years in laboratory air storage. Composites made strain curve averaged from strain measurements on both
with both types of film showed a similar performance. sides of the uncracked samples. The water stored and nat-

85
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

moduli could be determined at later ages.

Table 1 shows that the elastic moduli of the
composites remained virtually the same for
both storage conditions which indicated
that 18 years of freezing and thawing under
UK winter conditions had not caused sig-
nif icant damage to the composite.
Although not shown in Table 1, the initial
elastic modulus of the naturally weathered
precracked samples had remained virtually
unchanged at about 30 GPa between 7
months and 18 years showing that the early
autogenous healing across the cracks was
also time stable.

3.3 Composite strength and

Fig. 3 – Averaged tensile stress-strain curves for up to 18 years of storage under
water of uncracked samples compared with 1 month water storage. Film type A. strength of the polypropylene film
For this type of composite, which con-
Table 1 – Effect of age and storage conditions on the tains more than the critical fibre volume, the composite
tensile elastic modulus of the uncracked composite (Ec) strength is controlled entirely by the film volume and
film strength because all the continuous fibres must frac-
Age Storage condition Number of Ec (GPa) S.D. (GPa)
after 1 month samples ture at the ultimate load. The film strength was there-
fore calculated by dividing the composite strength by the
1 month - 45 28.0 3.4 film volume.
5 years Natural weather 7 35.8 4.02
Under water 13 34.7 1.9 3.3.1 18 years of natural weathering
10 years Natural weather 8 31.7 1.02 Figure 4 shows that the strength of film A has decreased
Under water 10 29.3 3.6 by about 10% over 18 years of natural weathering and this
18 years Natural weather 7 32.6 1.9 difference is statistically significant at the 95% confidence
Under water 8 31.9 3.5 level. The strength reduction may have been influenced in
S.D.: denotes Standard Deviation. some samples by a local dissolving of the surface layer of
cement which could have exposed the two top film layers
to U.V. degradation resulting in polymer chain scission. In
urally weathered samples remained relatively straight the bulk of the specimen thickness, no such degradation
throughout the 18 years, but the specimens stored in labo- had occurred. The strength of film B is also significantly
ratory air became so warped that they had virtually different between 1 month and 18 years of natural weather-
reached the cracking strain on one face before loading, ing at the 95% confidence level, having decreased in
due to clamping in the Instron grips and hence no elastic strength from 279 MPa to 227 MPa. The reduction in
strength has mainly
occurred between
5 years and 10 years,
the 18 year strength
being identical to the
10 year strength.

3.3.2 17 years of stor-

age in laboratory air
Figure 5 shows that
there is about 15% dif-
ference in strength
between 1 month and
17 years for f ilm A.
Also, f ilm B has
reduced in strength
over the 17 year period
by about 12%. The
Fig. 4 – The effect of natural weather for up to 18 years on the strength of the polypropylene film. Error composite strengths are
bars denote 95% confidence limits. remarkably similar for

86
Hannant

material stored outside

or in laboratory air at
closely similar f ilm
volumes.

3.3.3 18 years of stor-

age under water
Figure 5 also con-
tains limited data on
underwater storage and
shows that there has
been little change in
the strength of film A
during 18 years under
water and a small Fig. 5 – The effect of inside storage for up to 17 years or under water storage up to 18 years on the
change, which is signif- strength of the polypropylene film. Error bars denote 95% confidence limits.
icant at the 95% confi-
dence level, for film B although there has been no signifi- 3.4 Crack spacing and bond strength
cant change for this film between 5 years and 18 years.
The ‘real’ or absolute value of bond strength, if such a
3.3.4 General comment value exists, is notoriously difficult to measure for fibre
No other data has been found in the literature for reinforced cements with circular fibres but is even more
measurements of the strength of polypropylene fibres in diff icult for split f ilms with hairy, irregular surface
a cement matrix over such long time scales. Therefore shapes. For polypropylene, there is thought to be no
this data should give some confidence regarding the chemical bonding at the fibre-cement interface and
long-term stability of polypropylene as a reinforcement therefore the bonding is mechanical in nature.
in highly alkaline conditions under a variety of weather- In order to measure the effects of time and weathering
ing conditions. The performance of film A without on the nominal bond strength (rather than the absolute
U.V. stabilisers has proved marginally better than film B bond strength) which is satisfactory for comparative pur-
with U.V. stabilisers so that the shielding effect of the poses, the average bond strength τ is calculated from the
cement matrix has been more than sufficient to enable final crack spacing C using equation (1). Assuming that
the cheaper film to be used. the bond is frictional in character
The effect of the increase in the matrix strength over
τ = 1.364 m ⋅ σmu ⋅ f
V A
long time scales on increasing the critical fibre volume for (1) Ref [2]
multiple matrix fracture, has been demonstrated and has V f C Pf
more general implications for a whole range of fibre rein- where Vm and Vf are the volumes of fibre and matrix
forced cements. These should be designed at the outset to respectively,
include sufficient fibre volume to ensure multiple matrix Af and Pf are the cross-sectional area and perimeter of
fracture throughout the lifetime of the product. Thus, the fibre respectively,
brittle fracture with a single crack can be avoided. σmu is the mean matrix cracking stress determined from
the complete stress-strain curve.
The quantity Pf /Af is effectively the specific surface
area of the film per unit volume of the
f ilm and this was determined by
Table 2 – Crack spacing and bond strength variations with time
Montedison spa using the absorption of
Condition Composite Average film σmu: MPa Average crack Average bond Krypton gas at liquid nitrogen tempera-
age volume: % spacing: mm strength MPa tures. The determination of σmu at later
Natural weathering 1 month 6.2 9.4 0.9 0.4 ages is itself a problem for the naturally
5 years 6.2 > 13.5 1.4 > 0.4 weathered material which continued to
10 years 6.2 > 14.5 1.6 > 0.4
18 years 6.4 > 14.9 1.7 > 0.3
crack at strains in excess of 5% and
hence only a lower limit to σmu could
Air storage 1 month 6.3 9.2 0.9 0.4 be estimated. With these reservations,
5 years 6.6 11.3 1.1 0.4
9 years 6.1 11.9 1.3 0.4 the average bond strengths for film A at
17 years 6.2 11.9 1.3 0.4 ages up to 18 years are shown in Table 2.
Underwater 1 month 6.0 9.4 0.9 0.5
It can be seen from Table 2 that in most
5 years 5.6 11.7 1.3 0.4 cases, the crack spacing and the mean
10 years 6.0 11.0 1.1 0.4 matrix cracking stress increased or
18 years 6.3 11.5 1.4 0.4 decreased in proportion to each other,
2 3
Surface area Pf /Af of film A is 482 mm /mm . and therefore the calculated bond
σmu is mean matrix cracking stress. strengths were effectively unchanged by

87
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
age or type of storage. Since the matrix is very likely to
have densified around the film during 18 years of natural
weathering and underwater storage, the apparent lack of
change of bond strength indicates that fibre slip may be
mostly controlled by shearing within the film elements
themselves, which would be unaffected by changes in the
matrix. Photographic evidence of shearing within film
elements has been obtained by Ohno et al. [4] and Table 2
implies that sufficient mechanical bonding of the film
edges has been achieved by 1 month. The measured bond
strengths between 1 month and 18 years would then
depend on the intermolecular shear strengths of the
aligned polymer chains which are naturally time stable and
do not depend on the matrix microstructure. If this
hypothesis is correct, it is unlikely that the composite will
ever become brittle due to excessive bonding.
4. TOUGHNESS
It is often found that the toughness of fibre cements
reduces with time in natural weathering and with under-
water storage. For instance, early work on glass fibre rein-
forced cement showed reduced toughness after some years
[5] although this has been substantially remedied by atten-
tion to the matrix [6]. Cellulose fibre reinforced cements
become brittle due to petrification of the cellulose fibres
[7] and asbestos-cement also becomes brittle with age.
Toughness can be expressed as the energy absorbed to
failure by a unit volume of composite and so has units of
J/m3. The toughness values quoted in this paper were calcu-
lated from the areas under the appropriate tensile stress-
strain curves. For the composite made with film A after
1 month under water, the average energy absorbed up to
failure was 790 kJ/.m3 compared with about 770 kJ/m3 after
18 years natural weathering, 730 kJ/m3 after 17 years in lab-
oratory air and 750 kJ/m3 after 18 years under water. Thus,
provided that the fibre volume remains above the critical
fibre volume, embrittlement is unlikely to occur. These
values compare with about 5.5 kJ/m3 for asbestos-cement.
5. CONCLUSIONS
1. The uncracked tensile elastic moduli of the naturally
weathered and under water stored composites were similar
at 33 GPa and 31 GPa respectively at 18 years. This implied
that freeze-thaw damage had been minimal for the natu-
rally weathered samples. The initially pre-cracked samples
stored outside had healed across the cracks to give an elastic
modulus of about 31 GPa. No autogenous healing had
occurred for samples stored inside.
2. At 18 years, the stress at 0.5% strain for the uncracked
naturally weathered material averaged at 11.9 MPa com-
pared with about 9.8 MPa for dry air storage. Thus it
appeared that the effects of water availability combined with
carbonation had increased the tensile strength of the matrix
above other curing conditions.
3. The complete tensile stress-strain curves are signif-
icantly different at different exposure conditions.The
88
smoothest curves are for 18 years under water and the
most jagged are for natural weathering.
4. The strength of polypropylene film Type A which
had no additives apart from processing additives, had
decreased by about 10% over 18 years of natural weath-
ering. There was virtually no change in the strength of
this film after 18 years under water and about a 15%
reduction in strength after 17 years in dry air.
Film Type B , which contained a high U.V. stabiliser,
was generally slightly weaker than Type A with a strength
reduction in natural weathering of 19% between 1 month
and 10 years but no further reduction between 10 years
and 18 years. After 17 years in laboratory air, the strength
reduction was 12% whilst under water there was a 13%
reduction at 5 years and no further change between 5 years
and 18 years.
5. At 17 or 18 years, all composites with fibre vol-
umes greater than the critical fibre volume had strains to
failure in excess of 5%. A few samples with less than the
critical fibre volume failed at strains of about 3% show-
ing the importance of allowing, at the design stage,for
potential long term changes in the matrix and the fibre
in order that the critical fibre volume is always exceeded.
6. Bond strength between cement and polypropylene
as calculated from crack spacing was relatively unaffected
by exposure condition or time for periods up to 18 years.
There may have been a small decrease in bond strength
in under water storage but the experimental problems in
establishing the ‘true’ bond strength involved consider-
able errors and unknowns giving only limited confi-
dence in any apparent differences in performance over
such a long time period.
7. Overall, the composite has been shown to retain its
toughness and strength at a high level over many years,
giving increased confidence in the long term stability of
polypropylene in a cementitious environment.
REFERENCES
[1] Hannant, D. J., ‘Durability of cement sheets reinforced with
polypropylene networks’, Magazine of Concrete Research 35 (125)
(1983) 197-204.
[2] Aveston, J., Cooper, G. A. and Kelly, A., ‘Single and multiple
fracture’ in ‘The Properties of Fibre Composites’, Conference
Proceedings of the National Physical Laboratory (IPC Science &
Technology Press Ltd. Guildford, April 1971) 15-24.
[3] Aveston, J., Mercer, R. A. and Sillwood, J. M., ‘Fibre reinforced
cements - scientific foundations for specifications’, in
‘Composites, Standards, Testing and Design’, National Physical
Laboratory Conference Proceedings, (IPC Science &
Technology Press Ltd. Guildford, April 1974) 93-103.
[4] Ohno, S. and Hannant, D. J., ‘Modelling the stress-strain
response of continuous fibre reinforced cement composites’, ACI
Materials Journal 91 (3) (1994) 306-312.
[5] Majumdar, A. J. and Laws, V., ‘Glass fibre reinforced cement’,
BSP Professional Books, Oxford, UK.
[6] Marikunte, S., Aldea, C. and Shah, S. P., ‘Durability of glass fibre
reinforced cement composites’ Advanced Cement Based Materials
(5) (1997) 100-108.
[7] Bentur, A. and Akers, S.A.S., ‘The microstructure and ageing of
cellulose fibre reinforced cement composites cured in a normal
environment’, Int. J. of Cement Composites & Lightweight Concrete
11 (2) (1989) 99-109.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 89-97
Bond characteristics of corroding reinforcement in
concrete beams
P. S. Mangat, M. S. Elgarf
School of Construction, Sheffield Hallam University, Sheffield, S1 1WB, England
Paper received: February 2, 1998; Paper accepted: March 9, 1998
A B S T R A C T
The results of an experimental study of the bond char-
acteristics of reinforced concrete beams subjected to rein-
forcement corrosion are presented. Beam specimens rec-
ommended by the joint RILEM/CEB/FIP Committee
were used, which comprised two halves of a reinforced
concrete beam rotating about a hinge mechanism.
Corrosion was induced at different levels of rebar diameter
loss: 0, 0.3, 0.4, 0.5, 1, 2 and 5 percent by impressing
direct current of intensity 0.8 and 2.4 mA/cm2. The speci-
mens were tested under four point bending to induce
bond failure and load - free end slip curves were plotted.
The tests showed that at up to 0.4% degree of corro-
sion, no free-end slip occurred in the reinforcement bars
until complete breakdown of bond at failure. At higher
degrees of reinforcement corrosion, free-end slip com-
menced immediately upon application of load and
increased linearly with increasing load. The free-end
slip at maximum load was a function of the degree of
reinforcement corrosion. At small degrees of corrosion,
the bond strength increased with increasing degree of
corrosion, showing a maximum increase of over 25% at
0.4% corrosion. Higher degrees of corrosion led to a
sharp decrease in bond strength.
1. INTRODUCTION
Corrosion of reinforcing steel is the most important
single factor for reducing the bond strength in a reinforced
concrete member to such an extent that serviceability fail-
ure occurs [1-5]. The loss of bond strength is caused by
the opening of longitudinal cracks along the reinforce-
ment due to radial tensile stresses induced at the steel/con-
crete interface by the expansive products of corrosion.
Bond in conventional reinforced concrete is well under-
stood. Primary factors governing bond and its design pro-
cedures have been established [6-9]. Corrosion induced
bond deterioration and its effects on the strength and ser-
viceability of structural elements, however, is not well
understood [5, 10]; limited research has been reported
Editorial Note
Prof. P. Mangat is a Senior Member and the Chairman of RILEM TC LPC: Long term performance characteristics of fibre cement composites.
1359-5997/99 © RILEM
SCIENTIFIC REPORTS
R É S U M É
Cet article présente les résultats d’une étude expérimen-
tale sur les caractéristiques d’adhésion pour des poutres en
béton armé où l’armature est exposée à la corrosion. Des
échantillons de poutres recommandés par le comité joint
RILEM/CEB/FIP étaient employés sous forme de deux
moitiés d’une poutre en béton armé tournant autour d’un
pivot. La corrosion était induite à plusieurs niveaux de réduc-
tion de la section d’armature à savoir 0 ; 0,3 ; 0,4 ; 0,5 ;
1,2 et 5% par le biais d’un courant direct d’intensité 0,8 et
2,4 mA/cm2. Les échantillons ont été testés sous des condi-
tions de quatre points de flexion pour induire l’affaiblisse-
ment d’adhésion des courbes de glissement ont été tracées.
Les essais ont démontré que jusqu’à 0,4% de corro-
sion, il n’y avait pas de glissement par l’armature jusqu’à
rupture totale d’adhésion. Pour des niveaux de corrosion
élevés, le glissement a commencé dès l’application de la
charge augmentant linéairement avec l’augmentation de
celle-ci. Le glissement sous charge maximale est une fonc-
tion du degré des armatures. Pour un faible niveau de cor-
rosion, l’adhésion augmente avec le degré de corrosion, pour
un accroissement de l’adhésion de plus de 25% pour un
degré de corrosion de 0,4%. Des niveaux de corrosion éle-
vés induisent une réduction significative de l’adhésion.

which does not match the scale of the problem [11-13].
Some of the recent findings contradict assumptions made
in current practice - for example design specifications
often require steel reinforcement to remain free from cor-
rosion during service life of a concrete structure in order
to avoid loss of bond strength whereas, in tests, small
degrees of reinforcement corrosion have been shown to
develop higher bond strengths [11, 12]. Enhancement of
bond strength with initially rusted reinforcement has also
been reported [14] but the mechanism of bond improve-
ment in these situations is not the same as the case of rein-
forcement corroding inside concrete. The introduction
of ultimate strength design approach and the wide scale
use of high strength steel in modern reinforced concrete
construction compounds the seriousness of reinforcement
89
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

corrosion and the consequent reduction in

bond strength.
The limited amount of published data on
the effects of corrosion on bond strength are
conflicting [11-13] due to different bond test
techniques and different rates of corrosion
adopted in tests. Most research on bond has
been based on tests conducted on pull-out
specimens [15]. The pull-out test represents
bond behaviour in extreme cases of high
stress concentrations. There are two major
limitations of this test method. First, the dis-
tribution of stress on the rebar length is not Fig. 1 – Beam specimen for bond test.
uniform despite attempts to avoid the prob-
lem by using short embedment lengths. Second, it is the positioned to provide a cover of 25 mm to the centre of
steel bar which is stressed prior to the concrete, whereas, reinforcement. The reinforcement embedment length
in a reinforced concrete beam, it is the concrete which (Fig. 1) was the controlling experimental parameter to
transfers stresses to the steel through the interfacial ensure the precedence of bond failure in beams over flex-
bond. An alternative bond test conducted on beam ural and shear failure. The required embedment lengths
specimens is considered to be more representative for were obtained by using plastic sleeves to cover the
flexural members than the pull-out test [15-17]. In this unbonded lengths. Based on RILEM recommendations
type of test, the free-end slip of the steel bar relative to [18], the bonded (embedment) length of reinforcement in
the concrete is measured on beam specimens tested each of the two half-beam blocks was 10 φ, where φ is the
under flexure. The only uncertainty is the accuracy of diameter of reinforcement. Extended lengths of the plastic
locating the centre of rotation of the beam. The sleeves were provided at each rebar end to protect the
RILEM/CEB/FIP beam specimen for bond testing [18] externally protruding lengths of each rebar during the cor-
has been adopted, with some modif ications, for the rosion inducing phase which is detailed in section 2.4.
research reported in this paper. These plastic sleeves on the protruding length of reinforce-
ment were removed after the corrosion inducing process
was completed. As indicated in Fig. 1, the ends of each of
2. EXPERIMENTAL the internal plastic sleeve were sealed to the steel bar with
an elastic silicon sealant, carefully chosen so that it would
2.1 Beam specimens allow free movement of the rebar relative to the concrete.
Electrical connection was made with ultra-flex high-con-
Reinforced concrete beams of 910 mm length, and rec- ductive copper wires connected to one end of each rebar
tangular cross-section 150 mm deep and 100 mm wide by means of a self-amalgamating polyisobutylene tape and
were used (Fig. 1). Each specimen comprised of two half external collars, supplied by R S Component Ltd.
lengths of the beam, interconnected at the bottom by two
deformed steel bars, each of 10 mm diameter and
1100 mm length. A steel hinge was provided near the top 2.2 Materials
(Fig. 1) to allow for rotation of the concrete blocks,
according to the RILEM recommendations [18]. The Ordinar y Portland cement, supplied by Castle
steel hinge comprised of two high-toughness steel compo- Cement Limited, Stamford, Lincolnshire, f ine and
nents constituting together a pivot around which the two coarse aggregates, supplied by Tarmac Road Stone Ltd,
half-beam blocks rotated under loading. The gap
between the two half-beam blocks was maintained at
the casting stage by means of polystyrene packing
which was removed just prior to testing. The exact
location of each steel hinge in the mould was achieved
by bolting the two components of the hinge to a 3 mm
thick steel plate, which was fixed to the steel mould as
shown in Fig. 2a. The 3 mm thick steel plate was
bolted to the top flanges (top face) of the mould by
screws marked B while the hinges were located by
bolting them to the plate by screws marked A. The
reinforcement bars were located symmetrically at a
spacing of 50 mm centre to centre. Each of the rein-
forcement bars was threaded through two holes at each Fig. 2 – Positioning of hinge mechanism and reinforcement bars in the
end face of the steel mould and protruded 95 mm from mould: (a) Hinge mechanism attachment; (b) Location of reinforce-
each end face. As shown in Fig. 2b, the holes were ment bars.

90
Mangat, Elgarf
East Midlands, were used. The fine aggregate con-
formed to the medium zone of B.S.882: 1965. The
coarse aggregate was crushed grey granite of 10mm
maximum size (B.S. 882). 10 mm diameter deformed
high-yield steel bars of yield strength 520 N/mm2 and
modulus of elasticity 206 KN/mm2 were used
2.3 Mixing, casting and curing
A concrete mix of proportions 1: 2.24 : 3.22 (cement:
fine aggregate; coarse aggregate) with w/c = 0.5 was used.
The cement content was 425kg/m3. 1.0% (by weight of
cement), sodium chloride was added to the mix to
improve the concrete conductivity. The compressive
strength of the concrete after 28 days was 45 N/mm2, with
1 N/mm2 standard deviation. The workability of the con-
crete was high.
The dry constituents of concrete were mixed in a pan
mixer for 3 minutes, and a further 3 minutes after water
was added. The concrete was cast horizontally in steel
moulds with reinforcement and hinges positioned as
described in section 2.1. Compaction was carried out by
means of a poker vibrator. After placing and compacting
concrete, the specimen surface was smoothed with a steel
trowel. The moulds were then covered with polythene
sheets. Twenty four hours after casting, the steel plate that
connected the hinge to the mould was removed and the
specimen was demoulded. To avoid possible disturbance
of the steel hinge during handling of the beam specimen,
the polystyrene pieces that separated the two half-beam
blocks were left in place until the specimen was ready for
testing. In general, demoulding and transporting the spec-
imens were done with great care to avoid any relative dis-
placement of the two concrete blocks. Failure to do so
might have resulted in disrupting the adhesion of the bar to
the concrete. After demoulding, the specimens were
cured in a water spray chamber at 20°C and 90% relative
humidity for a period of 14 to 18 days. The specimens
were then transferred to the apparatus which was used to
induce corrosion in the reinforcement. The procedure for
inducing corrosion is given in section 2.4. The corrosion
Fig. 3 – Accelerated corrosion apparatus.
91
inducing period ranged between 1 and 7 days after which
the specimens were returned to the water spray chamber to
resume curing until the test age of 28 days.
2.4 Accelerated corrosion in reinforcement
After 15 days of casting (1 day in the mould and 14
days curing in the water spray chamber), each specimen
was subjected to accelerated galvanic corrosion by means
of two identical integrated systems shown in Fig. 3.
Each system incorporated a direct current power supply
with a built-in ammeter to monitor the cell current.
The corrosion process took place in a shallow steel tray
in which 3.5% NaC solution was used as the elec-
trolyte. The solution level in the tray was adjusted to
ensure submersion of the beam up to the top of rein-
forcement level. The current intensity was arranged so
that the reinforcing bars served as the anodes while the
steel tray was the cathode. For each beam the current
intensity and the corrosion period were adjusted to give
the desired degree of corrosion. Each selected degree of
corrosion provided a pre-defined percentage reduction
in the bar diameter within a short time scale. If R is the
corrosion in ‘mm/year’, T is the time in ‘years’ since the
initiation of corrosion and D is the diameter of rein-
forcement in ‘mm’, the degree of corrosion is defined as
2RT/D% (% reduction in rebar diameter). The rela-
tionship between corrosion current density and the
weight of metal lost due to corrosion was determined by
applying Faraday’s law [19] as follows:
∆ω = AIt (1)
ZF
where:
∆ω metal weight lost due to corrosion
A atomic weight of iron (56g)
I Corrosion current (amp)
t time elapsed (sec)
Z valency of the reacting electrode (iron) which is 2
F Faraday’s constant (96500 amp sec).
The metal weight loss due to corrosion can also be
expressed as:
∆ω = aδγ (2)
where
a rebar surface area before corrosion (cm2)
δ material loss (cm)
γ density of material (g/cm3)
The corrosion current can be expressed as:
I=ia (3)
where i = corrosion current density (amp/cm2)
substituting from equations (2) and (3) into (1) gives:
δ = Ait (4)
γZF
Considering steel reinforcement in concrete, the
density of iron γ = 7.86 g/cm3 and values for A, Z and F
from above can be substituted in equation (4).
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
Table 1 – Test programme
No of Degree of Corrosion
Reinforcement
specimens corrosion rate
2 0% - -
2 0.3% 0.8 mA/cm2
2 0.4% 0.8 mA/cm2
2 φ 10mm 2 0.5% 0.8 mA/cm2
2 1.0% 0.8 mA/cm2
2 2.0% 0.8 mA/cm2
2 5.0% 2.4 mA/cm2
If R is defined as the material loss per year (cm/year),
its value can be determined from equation (4) by substi-
tuting t = 1 year (in appropriate units) to give:
R = 1156 i (cm/year)
According to equation 5, the corrosion rate, R, that
is caused by a current density (i) of 1mA/cm2 works out
to be 11.56 mm/year. That means, a reinforcing bar of
10 mm diameter under impressed current density of
0.8 mA/cm2 will corrode to a degree of 0.5% reduction
in rebar diameter (2RT/D% = 0.5%) in a period of
24 hours (approximately). Six degrees of reinforcement
corrosion (2 RT/D%) were investigated; 0.3%, 0.4%,
0.5%, 1.0%, 2.0% and 5.0%. Each degree of corrosion
was represented by two replicate specimens as shown in
Table 1. Two beam specimens with no reinforcement
corrosion were employed as control beams, to which the
strengths of the other beams were related.
2.5 Flexural-bond test
At the age of 28 days after casting, the beam speci-
mens were tested under four-point bending to deter-
mine the load-slip curves. The arrangement of loading
and the monitoring device is shown in Fig. 4. Since no
shear reinforcement stirrups were provided inside the
specimens, shear failure was prevented by externally
reinforcing the shear zones by means of two external
steel collars as shown in Fig. 4. This ensured that the
Fig. 4 – Loading arrangement under four point bending.
mode of failure was through disruption
Pre-corrosion Corrosion
of interfacial bond between reinforce-
curing period duration (hrs) ment and concrete. Loading was
applied in increments of 5 KN; each
-
increment was achieved in half a
15 days 14.4 minute. After each increment, the load
15 days 19.2 was kept constant to stabilise the slip,
15 days 24.0
which normally occurred within 1
minute. The slip was measured at the
15 days 48.0 start and at the end of each incremental
15 days 96.0 stage of loading by means of four
15 days 80.0 Linear Var iable Displacement
Transducers (LVDT). Two transducers
(5)
Fig. 5 – Free end slip measuring device.
were connected to the right side of the beam (R1 and
R2) and two to the left side of the beam (L1 and L2) as
shown in Fig. 4. Each LVDT was secured to the free
end of one of the reinforcement bars by means of a steel
collar as shown in Fig. 5. The probes of the transducers
were in contact with steel discs which were firmly glued
to the end faces of the beam. The free slips at both ends
of each reinforcement bar, at every load increment, were
constantly logged into a six-channel data-logger and
transmitted to a 6-pen chart recorder. The set-up was
automatically controlled by a 386 s OPUS personal
computer, loaded with a computer soft-
ware which was developed by the authors.
Each test was continued until complete
bond failure of the bars occurred in either
half of the beam blocks.
2.6 Determination of the lever arm
Six beam specimens, in addition to
those listed in Table 1, were designed to
check the position of the centre of rotation
of the beam under four-point bending test.
Each reinforced concrete beam specimen
comprised of the two halves as shown in
Fig. 6. A steel hinge was provided near the
92
Mangat, Elgarf

sents the beam specimen (Fig. 1) under

four-point bending. The forces P/2 are the
externally applied forces, force F in the rein-
forcement is mobilised by the bond at the
concrete/reinforcement interface along the
embedment length (Fig. 1) and the hinge
mechanism provides equal and opposite
force F to maintain equilibrium.
Considering the equilibrium of forces about
the hinge, for each half of the mechanism,
gives the following expression:
Fig. 6 – Beam specimen for determining lever arm provided by the hinge
mechanism.
2 la
(
F = P a 1 − a2 ) (6)

top to allow rotation of the concrete blocks in exactly the where a is the lever arm, and a1 and a2 are the dimen-
same manner as the bond specimens in Fig. 1. The two sions shown on Fig. 7.
blocks were interconnected by continuously bonded rein- If fb and fs are the bond stress at the interface and the
forcement bars at the bottom with U-shaped anchorage tensile stress in the steel respectively at a total machine
hooks at the ends to prevent pull-out failure under four- applied load P, then:
point bending. This arrangement resulted in yield failure
of the reinforcement instead of pull-out failure. Three fb = F (7)
Ab
combinations of 6 mm diameter mild steel reinforcing bars
were employed - one bar, two bars and three bars. The and
bars were positioned symmetrically in the cross section at a
cover of 25 mm to the centre of reinforcement. Each of fs = F (8)
As
the three reinforcement combinations was represented by
two replicate beams. No corrosion was induced in the where Ab and As are the steel surface area and the steel
reinforcement and no plastic sleeves were located around cross-sectional area respectively. If φ is the rebar diame-
the reinforcement so that full bond was maintained with ter and n is the number of the reinforcing steel bars, Ab
the concrete. The specimens were tested under four-point and As can be expressed as follows:
bending, as described earlier in section 2.5 and shear failure Ab = nπ φ (10 φ) (9)
was prevented as previously (Fig. 4) by providing external
steel collars as shear reinforcement in the critical zone. In where (10 φ) is the rebar bonded length as shown in Fig. 1.
each test the load cell of the loading machine was con-
nected to a chart recorder to monitor the applied load, nπφ2 (10)
As =
which was increased at a constant rate of 5 KN/min until 4
the reinforcement fractured. Substituting for Ab from equation (9) and for F from
equation (6) into equation (7) gives:

3. RESULTS AND DISCUSSION fb =

(a1 − a2 ) ∗ P (11)
2 la 10 nπφ2
3.1 Analysis of the bond test mechanism Similarly, substituting for As from equation (10) and
for F from equation (6) into equation (8) gives:
Fig. 7 shows the forces acting on the Hinge-
Mechanism test for determining bond strength. It repre- fs =
( a 1 − a2 ) ∗ 4P
(12)
2 la nπφ2

3.2 Lever arm calculations

Fig. 8 shows the load-time cur ves
obtained for beam specimens tested under
four-point bending to determine the lever
arm as described in section 2.6. Each
curve represents the average of two beam
tests. The point at which the load-time
curve deviates sharply from linearity was
taken as the onset of yielding of reinforce-
ment. According to Fig. 7, the maximum
Fig. 7 – Forces of the hinge mechanism.

93
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

measured to the nearest 0.05 mm by means

of a crack-measuring microscope.
Table 2 shows that the maximum width
of corrosion-induced cracking is dependent
on the degree of reinforcement corrosion.
The maximum crack width did not exceed

94
Mangat, Elgarf
Fig. 11 – Load - slip relationships of the flexural bond tests (degree of corrosion:
0, 0.3, 0.4, 0.5, 1, 2, 5%).
Fig. 12 – Relationship between maximum load at bond failure and degree of cor-
rosion (2 RT/D%).
resents specimens which underwent 0.4% reinforcement
corrosion. It shows that no free-end slip occurred up to
the point of complete breakdown of bond. At higher lev-
els of corrosion (e.g. 2% in Fig. 10), however, an approxi-
mately linear increase in free-end slip is observed up to the
maximum load. Figs. 9 and 10 also show that complete
breakdown of bond occurs in one half of the beam which
leads to ultimate failure.
The load versus free-end slip curves of all beam speci-
mens at different degrees of reinforcement corrosion (0-
5%) are plotted in Fig. 11. Each plot is an average of two
test specimens. No free-end slip until complete break-
down of bond is recorded at corrosion degrees 0% to
0.4%. At higher degrees of corrosion, free-end slip com-
mences at the onset of loading and the rate of increase with
load is greater for higher degrees of corrosion. Beyond the
maximum load (representing break down of bond) and up
to ultimate failure large increases in free-end
slip, occur. The process of bond breakdown
in corroding beams (degree of corrosion >
0.4%) comprises three sub-process. The first
sub-process is dominated by the loss of
mechanical interlock due to corrosion of the
rebar ribs. This results in free-end slip com-
mencing at very low loads as indicated in Fig.
11. The second sub-process is the loss of
frictional resistance due to the flaky corro-
sion products at the concrete/steel interface.
This results in the free-end slip increasing
almost linearly with an increase in the load.
The third sub-process causes a sharp devia-
tion of the load-slip curve at maximum load
resulting in rapid increase in free-end slip
with the load remaining fairly constant (or
dropping at high degrees of corrosion). This
is caused by a decrease in rebar confinement
due to the opening of longitudinal cracks
along the reinforcement.
3.4.2. Bond strength
Fig. 12 illustrates the effect of the degree
of reinforcement corrosion on the maximum
f lexural load at bond failure of the beams.
The maximum load at bond failure corre-
sponds to the point in Fig. 11 at which the
free-end slip increases sharply without signifi-
cant increase in the corresponding load.
Assuming the bond stress to be uniform over
the bond length, the bond strength, fb is cal-
culated by substituting the maximum load in
equation (11). The resulting bond strength is
plotted against degree of rebar corrosion in
Fig. 13. Each point in Fig. 13 represents the
average of two beam tests. Figs. 12 and 13
show that bond strength increases at small
degrees of corrosion (< 0.4%). This is due to
the increased radial pressure, caused by the
expansive corrosion products at the interface,
which enhance the holding capacity of the
bar. The formation of small cracks along the
reinforcement (width < 0.05 mm, Table 2) does not pre-
vent such increase in bond strength.
A 25% increase in bond strength is evident at 0.4%
corrosion which was induced at a rate of 0.8 mA/cm2
(Table 1). Beyond 0.4% degree of corrosion, Fig. 13
shows decreasing bond strength with higher degrees of
corrosion. Other research based on direct pull-out bond
tests [12] has reported similar observations of increasing
bond strength at low levels of corrosion. Bond strength
was found to increase by about 50% at a corrosion degree
of 1% [12]. It has also been shown that the increase in
bond strength at low levels of corrosion is dependent on
the corrosion rate [14]; no increase in bond strength being
evident at corrosion rates greater than about 0.25 mA/cm2.
This contradicts the results in Fig. 13 which show over
25% increase caused by 0.4% degree of corrosion induced
at a rate of 0.8 mA/cm2 (Table 1).
95
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

The tensile stress in the reinforcement

bars at maximum load from the bond tests,
was calculated from equation (12). The
stress is plotted against degree of corrosion
in Fig. 14 which shows that the maximum
stress is less than the yield stress (520
N/mm2). This enabled the reinforcement
to mobilise full bond strength before beam
failure.

3.4.3. Critical free-end slip

Fig. 15 shows the relationship between
the degree of corrosion (2 RT/D%) and
the free-end slip of reinforcement at bond
failure (maximum load from load - slip dia-
gram). Each point on Fig. 15 represents the
average of two beam tests. For the beams
with 0.3% and 0.4% reinforcement corro-
Fig. 13 – Relationship between bond strength and degree of corrosion (2 sion no rebar free-end slip was recorded up
RT/D%). to the point of complete bond breakdown
(Fig. 1). For the beams with reinforcement
corrosion of 0.5%, an average free-end slip
of 0.12 mm was recorded at maximum load. Beams
which had corrosion levels higher than 0.5% showed an
increase in rebar slip with increasing degree of corrosion
(Fig. 15). A linear regression analysis yielded the follow-
ing relationship between degree of corrosion 2 RT/D%
and free-end slip, xf, at bond failure (maximum load):
 
x f = 0.12 − 0.056 0.5 − 2RT % (15)
 D 
Equation (15) shows that the critical slip (rebar free-
end slip corresponding to the maximum bond stress) is
corrosion dependent. Therefore, a universal value of criti-
cal slip reported in literature, and used in design [12, 20],
needs to be reviewed.

Fig. 14 – Maximum tensile stress in reinforcement during bond

tests at different degrees of corrosion. CONCLUSIONS
The investigation presented in this
paper leads to the following conclusions:-
1. Bond strength of corroding reinforce-
ment increases with increasing degree of cor-
rosion (2 RT/D%) of up to 0.4%. A maxi-
mum increase of 25% occurs at 0.4%
corrosion. Longitudinal cracks induced by
corrosion along the reinforcement do not
exceed 0.05 mm at degrees of corrosion up
to 0.4%.
2. Beyond 0.4% degree of corrosion,
bond strength decreases sharply. This is
caused by longitudinal cracks of maximum
width > 0.2 mm, which result in a loss of
rebar confinement by the surrounding con-
crete.
3. At degrees of corrosion (2 RT/D%)
greater than 0.5%, the free end slip of the
reinforcement commences upon load appli-
Fig. 15 – Relationship between degree of corrosion (2 RT/D%) and free end slip. cation and increases steadily with increasing

96
Mangat, Elgarf
load. At lower degrees of cor rosion
(< 0.5%), negligible free-end slip occurs until a break-
down of bond takes place at maximum load.
4. The free end slip at maximum load remains con-
stant (0.12 mm) at low degrees of corrosion (up to 0.4%)
and then increases linearly with increasing degree of cor-
rosion (2 RT/D%).
ACKNOWLEDGEMENTS
The paper presents some of the results of an EC
funded Brite/Euram project BREU P3 - 91 “Assessment
of Performance and Optimal Strategies for Inspection
and Maintenance of Concrete Str uctures using
Reliability based Expert Systems”. The partners in the
project were:
Sheffield Hallam University/Aberdeen University, UK
Labein, Bilbao, Spain
CSR, Aalborg, Denmark
Instituto Superior Technico, Lisboa, Portugal
Jahn Ingenieurs bureau, Brielle, Netherlands
REFERENCES
[1] Takewaka, K and Matsumoto, S., ‘Behaviour of reinforced con-
crete members deteriorated by corrosion of reinforcement’, Proc.
of JCI 6 (1987) 177-180.
[2] Everett, L. H. and Treadaway, K. W., ‘Deterioration Due to
Corrosion in Reinforced Concrete’, Building Research
Establishment Information Paper 1P 12/80.
[3] Lin, C. Y., ‘Bond Deterioration Due to Corrosion of Reinforcing
Steel’, Performance of Concrete in Marine Environment, SP-65,
American Concrete Institute, Detroit (1980) 255-269.
[4] Brezny, F. S., Kemp, E. L. and Wilhelm, W. J., ‘The Influence of
Loose Rust and Mill Scale on the Bond Characteristics of
Deformed Reinforcing Bars’, Civil Engineering Studies, Report
No. 2004, West Virginia University, Morgantown (1969) 129.
[5] Mangat, P. S. and Elgarf, M. S., ‘Strength and serviceability of
97
repaired reinforced concrete beams undergoing reinforcement
corrosion’, Magazine of Concrete Research (In Press).
[6] Furguson, P. M., ‘Bond stress: The state-of-the-art”, Report by
ACI Committee 408, ACI Journal Proceeding 63 (11) (1966)
408.1-408.22.
[7] ACI Committee 408, ‘State-of-the-art-bond under cyclic load-
ing’, ACI Materials Journal 88 (6) (Nov-Dec 1991) 669-673.
[8] CEB Task Group VI, ‘Bond Action Behaviour of
Reinforcement-State-of-the-Art Report’ (Dec. 1981) 153.
[9] Tepfers, R., ‘Cracking of concrete cover along anchored deformed
reinforcing bars’, Mag. Concr. Res. 21 (106) (1979) 3-12.
[10] Mangat, P. S. and Elgarf, M. S., ‘Flexural strength of concrete
beams with corroding reinforcement’, ACI Structural Journal (In
Press) (Jan.-Feb. 1999).
[11] Maslehuddin, M., Allam, I., Al-Sulaimani, G., Al-Mana, A. and
Abdeljauwad, S. ‘Effect of rusting of reinforcing steel on its
mechanical properties and bond with concrete’, ACI Materials
Journal 87 (5) (Sep.-Oct. 1990) 496-502.
[12] Al-Sulimani, G. J., Kaleemullah, M., Basunbul, I. A. and
Rasheeduzzafar, ‘Influence of corrosion and cracking on bond
behaviour and strength of reinforced concrete members’, ACI
Structural Journal 87 (March-April 1990) 220-231.
[13] Murphy, F. G., ‘The Effect of initial rusting on the Bond
Performance of Reinforcement’, CIRIA Report 71, pp 36,
(London 1977).
[14] Saifullah, M. and Clark, L. A., ‘Effect of Corrosion Rate on the
Bond Strength of Corroded Reinforcement’, Proceedings of
International Conference on Corrosion and Corrosion
Protection of Steel in Concrete, Vol. 1, (1994), Editor R N
Swamy, pp 591-602.
[15] ASTM C 234-86, ‘Standard Test Method for Comparing
Concrete on the Basis of the Bond Developed with Reinforced
Steel’, ASTM Annual Book of Standards, Section 4,
Construction, 1988.
[16] Abrishami, H. H. and Mitchell, D., ‘Simulation of uniform
bond stress’, ACI materials Journal 89 (2) (March-April 1992)
161-168.
[17] Bažant, Z. P. and Sener, S., ‘Size effect in pull-out tests’, Ibid. 85
(5) (Sept-Oct 1988) 347-351.
[18] RILEM/CEB/FIP Recommendation, ‘Bond test for reinforcing
steel’, Mater. Struct. 6 (32) (1973).
[19] Fontana, M. B. and Green, N. D., ‘Corrosion Engineering’,
McGraw Hill, New York, (1967).
[20] Mathey, G. R. and Watstein, D., ‘Investigation of bond in beam
and pull-out specimens with high-yield-strength deformed bars’,
ACI Journal, Proceedings 57 (9) (March 1961) 1071-1090.
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 98-102
Development of a multiple linear regression model for
SCIENTIFIC REPORTS
predicting the 28-day compressive strength of Portland
pozzolan cement
G. CH. Kostogloudis1, J. Anagnostou1, CH. Ftikos1, J. Marinos2
(1) Lab. of Inorganic Materials Technology, Department of Chemical Engineering, National Technical University of Athens, 9 Heroon
Polytechniou Str., Zografou Campus, GR-157 80 Athens, Greece
(2) HERACLES General Cement Company, P. O. B. 3500, GR-102 20 Athens, Greece
Paper received: December 24, 1997; Paper accepted: April 21, 1998
A B S T R A C T
A multiple linear regression model was developed for
the prediction of the 28-day compressive strength of
Portland Pozzolan Cement (PPC) containing Santorin
Earth as an admixture. The independent variables of the
model were: (1) the compressive strength at the age of
2 days, (2) the compressive strength after autoclave
hydrothermal treatment at 2.07 N/mm2 (300 psi) and
214°C for 3 h, (3) the C3S/C2S ratio of the clinker, (4) the
insoluble residue of the cement, (5) the pozzolanic activity
factor and (6) the C3A content of the clinker. The evalua-
tion of the proposed model was performed by various sta-
tistical tests, all of which were successful. These tests
included: multiple correlation, test of the significance of
coefficients (t-test), estimation of confidence intervals for
coefficients, test for outliers and unusual residuals, test for
influential points, conditional sums of squares, R-squared
and analysis of variance. There was very good agreement
between the strength predicted by the multiple regression
model and experimental results.
1. INTRODUCTION
Many efforts have been made to identify the relation-
ship between various physicochemical and mineralogical
parameters of cement clinker and the strength of cement
[1]. Although such relationships have not been fully jus-
tified [2], many models have been proposed for the pre-
diction of the compressive strength of Portland cement
[1-11]. A carefully designed model can serve as a valu-
able tool during the production of cement and the con-
trol of its quality. However, the use of such models is
restricted to the specific applications for which they are
designed. In the case of blended cements, which have
drawn much attention over the past two decades mainly
due to energy saving reasons, little information is avail-
able concerning strength prediction [12]. In the present
work, a multiple linear regression model was developed
for the prediction of the 28-day compressive strength of
Portland Pozzolan Cement (PPC) containing Santorin
1359-5997/99 © RILEM
R É S U M É
Un modèle de régression multiple linéaire a été developpé
pour la prédiction de la résistance en compression du ciment
pozzolanique Portland, contenant de la Terre de Santorin
comme mélange. Les variables indépendantes du modèle
étaient : (1) la résistance en compression pour l’âge de
2 jours, (2) la résistance en compression après un traitement
de 3 heures dans un autoclave à 2,07 N/mm2 (300 psi) et
214°C, (3) le rapport C3S/C2S du clinker, (4) la quantité
d’insolubles du ciment, (5) le facteur d’ activité pozzola-
nique et (6) la quantité de C3A du clinker. L’évaluation du
modèle proposé à été effectuée par une série d’essais statis-
tiques : une corrélation multiple, un essai de signification des
coefficients (t-test), estimation des intervalles de confiance
pour les coefficients, essai pour les substances étrangères et les
résidus inhabituels, essai pour les points d’ influence, somme
des moindres carrés conditionnels, carré et une analyse de
variance. Un très bon accord entre la résistance prévue par le
modèle de régression et les résultats expérimentaux, a été
constaté.
Earth as an admixture. The evaluation of the proposed
model was performed by various statistical tests.
2. EXPERIMENTAL
Seventeen industrial PPC samples were used for the
development of the model. The natural pozzolan material
used as an admixture for the production of PPC was
Santorin Earth. The sampling was random, and it was per-
formed during a period of three months at the place of pro-
duction. The chemical analysis, potential mineral composi-
tion and ratios of the Portland clinker used for the
production of PPC, are shown in Table 1. The fineness
(Blaine), R90µm, R32µm, SO3 content, insoluble residue
(IR) and loss on ignition (LOI) of all PPC samples are also
shown in Table 1. It should be noted that the IR of the PPC
cement refers to the pozzolan content of the cement, and
should not exceed 20%, according to Greek specifications.
98
Kostogloudis, Anagnostou, Ftikos, Marinos

Table 1 – Analysis of Portland clinker and Portland pozzolan cement (PPC)

Sample number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Portland clinker chemical analysis (%)

SiO2 21.21 21.80 21.27 21.27 21.47 21.40 22.38 21.47 21.36 21.40 21.19 21.28 21.10 21.38 21.28 21.49 21.38
Al2O3 5.94 5.88 5.75 5.68 5.62 6.07 5.05 5.49 5.49 5.68 5.85 5.49 5.94 6.07 6.32 6.42 5.97
Fe2O3 3.10 3.00 3.05 2.80 3.10 3.20 3.15 3.05 2.95 3.20 2.85 3.05 2.90 3.05 3.10 3.10 3.15
CaO 62.86 62.30 63.63 63.00 62.72 63.28 63.42 63.63 63.56 63.42 63.91 63.14 63.28 63.42 61.95 61.53 63.63
MgO 3.38 3.43 3.08 3.58 3.43 3.18 2.97 3.18 3.43 3.03 3.33 3.13 3.18 3.08 2.88 3.23 3.48
K2 O 0.97 0.88 0.93 0.92 1.04 0.88 0.78 0.83 0.97 0.77 0.97 0.91 0.98 0.92 1.00 0.92 0.95
Na2O 0.83 0.74 0.73 0.76 0.92 0.75 0.60 0.62 0.88 0.64 0.81 0.83 0.80 0.80 0.76 0.74 0.74
SO3 0.62 0.64 0.64 0.35 0.58 0.72 0.38 0.64 0.41 0.78 0.43 0.34 0.54 0.69 0.56 0.60 0.34
CaOf 1.35 1.12 1.80 1.71 1.71 2.25 1.44 1.26 1.12 1.44 1.26 1.67 1.57 1.44 1.35 1.22 1.08
Portland clinker potential mineral composition (%)
C3S* 43.03 37.69 45.17 44.55 41.29 38.33 42.65 47.59 49.51 44.65 49.33 46.23 43.77 42.65 36.41 32.85 46.50
C2S 28.43 34.15 26.98 27.48 30.48 32.51 32.08 25.73 23.97 27.75 23.62 26.22 27.55 29.20 33.62 36.91 26.30
C3A 10.50 10.51 10.08 10.32 9.65 10.68 8.06 9.39 9.56 9.64 10.68 9.39 9.15 10.93 11.51 11.77 10.50
C4AF 9.42 9.12 9.27 8.51 9.43 9.73 9.58 9.27 8.97 9.73 8.67 9.28 8.82 9.28 9.43 9.43 9.58
Portland clinker ratios
LSF* 89.12 86.70 89.68 89.48 87.99 87.36 87.20 90.15 90.88 89.28 91.41 89.84 89.22 88.96 87.03 85.69 90.21
SR 2.35 2.45 2.42 2.51 2.46 2.31 2.73 2.51 2.53 2.41 2.43 2.49 2.14 2.34 2.26 2.26 2.34
AR 1.92 1.96 1.89 2.03 1.81 1.90 1.60 1.80 1.86 1.78 2.05 1.80 2.05 1.99 2.04 2.07 1.90
C3S/C2S 1.51 1.10 1.67 1.65 1.36 1.18 1.33 1.85 2.07 1.61 2.09 1.76 1.59 1.46 1.08 0.89 1.77
PPC analysis
Blaine
4360 4500 4540 4780 4400 4580 4460 4520 4500 4420 4530 4440 4440 4420 4680 4570 4420
(cm2/g)
R90µ m (%) 2.0 2.4 2.4 2.0 2.0 2.8 2.4 2.8 1.6 1.6 1.6 2.8 2.8 3.2 2.0 3.2 3.2
R32µ m (%) 28.8 28.4 32.0 32.6 28.6 37.0 32.0 32.6 30.0 28.6 30.6 34.6 37.4 33.8 32.2 32.2 35.7
SO3 2.98 2.87 2.70 2.80 3.20 2.80 2.80 2.80 2.80 2.80 2.80 3.00 2.90 2.70 2.90 2.90 2.98
IR 15.94 14.65 15.80 15.90 17.00 14.10 14.20 16.00 14.90 15.10 14.10 15.20 13.70 13.90 13.90 14.08 15.10
LOI 3.70 3.50 3.50 3.40 3.30 4.00 3.70 3.30 2.90 2.90 3.40 3.40 3.50 4.10 4.40 3.70 3.70
* The CaO-CaOf content was used for the calculation.

Table 2 – Compressive strength development, strength after autoclave hydrothermal treatment,

and pozzolanic activity factor (PAF) of PPC
Sample number
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Compressive strength of PPC (2, 7 and 28 days)
S2 (N/mm2) 13.8 12.6 12.2 13.8 11.6 12.2 11.6 10.7 11.9 11.3 13.5 11.9 12.2 11.6 11.9 10.1 10.1
S7 (N/mm2) 23.9 23.0 22.0 24.5 23.9 23.0 22.0 21.1 23.9 22.4 25.4 21.7 23.6 22.4 22.4 20.2 19.9
S28 (N/mm2) 36.1 35.5 34.6 37.4 36.1 34.9 35.8 34.3 37.4 35.8 38.9 34.9 36.4 34.6 34.6 31.5 32.5
Compressive strength of PPC after autoclave hydrothermal treatment
Sa (N/mm2) 34.20 37.30 35.10 35.60 41.80 33.70 35.10 37.50 39.70 39.90 38.20 35.90 37.50 35.40 35.00 35.30 35.60
28-day compressive strength of standard Portland cement/pozzolan mixture (S28sp), and PAF of PPC
S28sp
41.0 40.7 40.7 41.0 40.7 42.0 44.1 42.3 40.4 41.0 41.7 41.7 39.5 41.0 42.6 42.6 42.9
(N/mm2)
PAF 0.50 0.47 0.47 0.50 0.47 0.60 0.82 0.63 0.44 0.50 0.57 0.57 0.35 0.50 0.66 0.66 0.69

The compressive strength of the PPC samples was
measured at the ages of 2, 7 and 28 days, according to
EN specifications, and the results are shown in Table 2.
The compressive strength of PPC mortar specimens
after hydrothermal treatment at 2.07 N/mm2 (300 psi)
and 214°C for 3h was measured. The results of this test
are also illustrated in Table 2.
The activity of the pozzolan used for the production
of each cement was tested. The test involved the mea-
surement of the 28-day compressive strength of a mix-
ture of a standard high strength Portland cement of
known strength (80%), and the material of which the
pozzolanic activity was to be determined (20%). The
pozzolanic activity factor (PAF) was then calculated by:

99
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Table 3 – Correlation coefficients. Values in parentheses

S28sp − S28ss (1)
PAF = are P values for the hypothesis of zero correlation
S28s − S28ss
S2 Sa C3S/C2S IR PAF C3A
where S28sp is the 28-day strength of the standard
S2 -0.1070 0.1953 0.0980 -0.4239 0.0094
Portland cement (80%)/pozzolan (20%) mixture, (0.6826) (0.4525) (0.7083) (0.0900) (0.9714)
S28ss is the 28-day strength of the standard Portland
Sa -0.1070 0.3518 0.3531 -0.4268 -0.3355
cement (80%)/AFNOR sand (20%) mixture (Blaine: (0.6826) (0.1661) (0.1644) (0.0875) (0.1880)
3810 cm2/g, R90µm = 6.8%, S28s s= 36.1 N/mm2),
and S28s in the 28-day strength of the standard C3S/C2S 0.1953 0.3518 0.2699 -0.2520 -0.3995
(0.4525) (0.1661) (0.2947) (0.3292) (0.1122)
Portland cement (Blaine: 3820 cm2/g, R90µm=2%,
S28s = 45.9 N/mm2). The preparation of the speci- IR 0.0980 0.3531 0.2699 -0.2101 -0.2437
(0.7083) (0.1644) (0.2947) (0.4183) (0.3460)
mens was performed according to EN specifications.
The results on the measured S28sp and calculated PAF -0.4239 -0.4268 -0.2520 -0.2101 0.0099
PAF values are presented in Table 2. (0.0900) (0.0875) (0.3292) (0.4183) (0.9700)
C3A 0.0094 -0.3355 -0.3995 -0.2437 0.0099
(0.9714) (0.1880) (0.1122) (0.3460) (0.9700)
3. DEVELOPMENT OF THE MULTIPLE
REGRESSION MODEL Table 4 – Estimated coefficients for the multiple regression model
and their confidence intervals
3.1 Formulation of the model Standard 95% confidence limits
Estimate error t value P value Lower Upper
The multiple linear regression model for Constant 9.0302 3.3930 2.66 0.0238 1.4702 16.690
the prediction of the 28-day compressive S2 1.4526 0.0930 15.62 0.0000 1.2455 1.6598
strength (S28) of PPC had 6 independent vari- Sa 0.43196 0.0505 8.55 0.0000 0.31938 0.54455
ables, and it is described by the following equa- C3S/C2S 0.98120 0.2873 3.42 0.0066 0.34110 1.6213
tion: IR -0.44481 0.0957 -4.65 0.0009 -0.65796 -0.23165
S28 = b0 + b1⋅S2 + b2⋅Sa + b3⋅C3S/C2S PAF 2.7257 0.9711 2.81 0.0186 0.56202 4.8893
(2)
+ b4⋅IR + b5⋅PAF + b6⋅C3A C3A -0.31589 0.1050 -3.01 0.0131 -0.54980 -0.08197
R-squared = 97.76%
where bi are the coefficients to be determined. Standard error of estimation = 0.33366
The independent variables of the model were: Mean absolute error = 0.21191
(1) the compressive strength at the age of 2 days
(S2), (2) the compressive strength after autoclave 3.3 Estimation of coefficients, test of signifi-
hydrothermal treatment at 2.07 N/mm2 (300 psi) and cance and confidence intervals
214°C for 3 h (Sa), (3) the C3S/C2S ratio of the clinker,
(4) the insoluble residue of the cement (IR), (5) the poz- The estimated coefficients for the multiple regression
zolanic activity factor (PAF), and (6) the C3A content of model are shown in Table 4. The P values correspond to
the clinker. tests of the hypotheses that the coefficients are equal to
The determination of the coefficients of the model zero. Values of P less than 0.05 indicate statistically sig-
and all statistical tests were performed using the EXE- nificant nonzero coefficients at a 95% confidence level.
CUSTAT statistical computer package [13]. In Table 4, all coefficients have P < 0.05. The standard
error of estimation provides an estimate of the standard
deviation of the residuals around the regression model,
3.2 Multiple correlation equal to 0.33366.
Table 4 also shows the 95% confidence intervals for
Before the determination of the coefficients of the each of the coefficients in the fitted model. These inter-
model it is necessary to ensure that the independent vals quantify the sampling variability in the estimation of
variables do not correlate with each other. In practice the the coefficients and are useful for determining how well
question is whether there is a significant correlation each coefficient has been determined. Since none of the
among the independent variables. A table was con- intervals includes 0, there is an actual linear relationship
structed (Table 3) with the correlation coeff icients between the dependent variable and each of the inde-
between pairs of variables. The correlation coefficient pendent variables, and so, all variables should be
measures the strength of the linear relationship between included in the model.
two variables on a scale of -1 to +1. The P values, which
are included in Table 3, are used to test whether the
coefficients are significantly different from zero. If the P 3.4 Test for outliers and unusual residuals
value is less than 0.05, there is a significant correlation
between the pair of variables with 95% confidence. As Fig. 1 shows the Studentized residuals as a function
can be seen, P > 0.05 for all pairs, indicating that all vari- of the predicted S28 values. The residuals show the dif-
ables are independent from one another. ference between the actual values and the predictions,

100
Kostogloudis, Anagnostou, Ftikos, Marinos

the residuals, shown in Fig. 2, can be used to judge

whether the residuals could reasonably be considered to
follow a normal distribution, and may also be helpful in
detecting outliers. The residuals fall fairly well along a
straight line, while no outliers can be observed.
In Fig. 1, it can be seen that none of the residuals
have Studentized values greater than 2 or less than -2.
Consequently, no unusual residuals appear for this
model.

3.5 Test for influential points

Fig. 1 – Plot of Studentized residuals versus predicted S28 values.
A point may be considered as influential if it has an
unusually large impact on the fitted model. This is mea-
sured through a statistic called “leverage”, which mea-
sures each point’s inf luence on the coefficients of the
fitted equation. The average leverage was calculated to
be 0.4118. As none of the points has a leverage more
than 2.5 times the average leverage, there are no influen-
tial points.

3.6 Conditional sums of squares

Fig. 2 – Normal probability plot of residuals.
The conditional sums of squares, shown in Table 5,
test the significance of each variable when added sequen-
tially to the fit. These sums of squares are useful for
quantifying the additional contribution of each variable
after accounting for the contribution of those entered
earlier. Since all variables have small P values (less than
0.05), they are all added significantly to the fit when
they enter the model. S2 is the most significant variable,
while the significance of Sa is also high. The contribu-
tion of the other four variables is far lower than that of
the previous two.

and the Studentized residuals, express each deviation in 3.7 Analysis of variance
terms of how many deviations it is away from the fitted
line. The Studentized residuals that were calculated, are The analysis of variance (ANOVA) table (Table 6)
based on the estimated residual standard deviation if the decomposes the total variability in S28 into two compo-
fitting were performed without that data value. This nents: one due to the regression, and the second due to
kind of residuals is particularly useful for detecting out- deviations around the fitted model. The R-squared sta-
liers (i.e. points that do not follow the same pattern as tistic, based upon the ratio of the model sum of squares
the others). As can be seen in Fig. 1, the plot appears rea- divided by the total (corrected) sum of squares, indicates
sonably random, and none of the residuals is noticeably that the model accounts for 97.76% of the variation of
distinct from the others. The normal probability plot of S28. The mean squared error estimates the variance of
the deviations around the model to be equal to 0.1113.
Since the P value corresponding to the F-ratio is less
Table 5 – Conditional sums of squares than 0.05, the model as a whole is statistically significant.
Source Sum of squares D.F. Mean square F-ratio P value
S2 30.1193 1 30.1193 270.55 0.0000
Table 6 – Analysis of variance
Sa 12.2284 1 12.2284 109.84 0.0000
C3S/C2S 1.8539 1 1.8539 16.65 0.0022 Source Sum of squares D.F. Mean square F-ratio P value
IR 2.0998 1 2.0998 18.86 0.0015 Model 48.682 6 8.1137 72.88 0.0000
PAF 1.3727 1 1.3727 12.33 0.0056 Error 1.1133 10 0.1113
C3A 1.0079 1 1.0079 9.05 0.0131 Total
49.7953 16
Model 48.682 6 (corr.)

101
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
4. DISCUSSION
The multiple regression model that was developed in
the present work, was shown to satisfy all the statistical
tests used for its evaluation. Fig. 3 shows a plot of the
observed (experimental) values of S28 versus the values
predicted from the fitted model. A line with slope equal
to 1 is also included. All points of the plot fell very close
to this line, while they show approximately the same
scatter around the line everywhere. Therefore, the pro-
posed model has a very good predicting ability.
The choice of the independent variables of the model
was based on statistical as well as physical considerations. A
large number of variables with physical importance was
considered, and those satisfying certain statistical criteria
were actually selected. The variable S2 is the most signifi-
cant among those used, and its presence in the model is
essential. The lack of a satisfactory mechanism for the
strength development of blended cements [14], does not
yet permit the substitution of S2 with other physicochemi-
cal parameters. The variable Sa is also of high significance.
This was expected, since the strength development under
conditions of accelerated hardening provides a good mea-
sure of the actual hydraulic behavior of the cement. The
C3S/C2S ratio is an important variable, which has also been
used in other works [3]. It incorporates trends arising from
the variation of other important parameters that affect the
C3S and C2S contents. These parameters include the CaOf
content, the alkalis occurring in solid solution with the
above calcium silicate phases, and the degree of burning of
the clinker. The IR is also an important variable, since it is
directly related to the concentration of pozzolan in the
blended cement. The negative dependence of S28 on IR
can be explained by the fact that an increase of the poz-
zolan content is known to reduce the compressive strength
of blended cements. The PAF is a measure of the contri-
bution of the pozzolan in the strength development of the
cement, and it is also included as a variable in the model.
However, it should be pointed out that the need of know-
ing this coefficient may induce some delay, as compared to
the rest of the input parameters. The C3A content is the
less significant variable among those used. It reflects the
composition of the liquid phase of the clinker, and incor-
porates factors related to the burning process and the cool-
ing rate.
5. CONCLUSION
The developed multiple regression model was shown
to be able to predict with adequate accuracy the 28-day
compressive strength of Portland pozzolan cement
(PPC) containing Santorin Earth as an admixture. The
model was evaluated by various statistical tests, which
were all successful. The usefulness of the proposed
model to the cement industry, is based on its ability to
make predictions on the 28-day strength of PPC, only 2
days after its production. This gives the opportunity to
cement engineers to make the necessary adjustments to
improve the quality of the product in due time.
102
Fig. 3 – Plot of observed versus predicted values of S28.
ACKNOWLEDGMENTS
The authors wish to thank HERACLES General
Cement Company for providing the industrial cement
samples used in this work.
REFERENCES
[1] Bruggemann, H. and Brentup, L., ‘Relationship between cement
strength and the chemico-mineralogical parameters of the
clinker’, Translation ZKG, 1/90, 30-33.
[2] Aldridge, L. P., ‘Estimating strength from cement composition’,
Proceedings of the 7th International Congress on the Chemistry
of Cement, Vol VI, Paris, 1980, 83-86.
[3] Alexander, K., Taplin, J. and Wardlaw, J., ‘Correlation of
strength and hydration with composition of Portland cement’,
Proceedings of the 5th International Congress on the Chemistry
of Cement, Vol III, Tokyo, 1968, 152-165.
[4] Alexander, K., ‘The relationship between strength and the composi-
tion and fineness of cement’, Cem. Concr. Res. 2 (1972) 666-680.
[5] Popovics, S., ‘Calculations of strength development from the
compound composition of Portland cement’, Proceedings of the
7th International Congress on the Chemistry of Cement, Vol VI,
Paris, 1980, 47-51.
[6] Dragicevic, L. J., ‘Prognosis of the physical-mechanical character-
istics of Portland cement’, Proceedings of the 8th International
Congress on the Chemistry of Cement, Vol II, Rio de Janeiro,
1986, 95-101.
[7] Hargrave, R., Venkateswaran, D., Deshmukh, V. and Chatterjee,
A., ‘Quantification of OPC clinker microstructure - An
approach for prediction of cement strength’, Proceedings of the
8th International Congress on the Chemistry of Cement, Vol II,
Rio de Janeiro, 1986, 167-172.
[8] Popovics, S., ‘Model for the quantitative description of the kinetics of
hardening of Portland cement’, Cem. Concr. Res. 17 (1987) 821-838.
[9] Relis, M., Ledbetter, W. and Harris, P., ‘Prediction of mortar-
cude strength from cement characteristics’, Cem. Concr. Res. 18
(1988) 674-686.
[10] Philippou, Th., Marinos, J. and Kostakis, G., ‘On the strength
behavior and phase composition of low lime saturation clinkers’,
Ibid. 18 (1988) 804-811.
[11] Petrasinovic, L. and Duric, M., ‘Predicting the compressive
strength of Portland cement and optimizing its raw mixture
composition’, Ibid. 20 (1990) 484-492.
[12] Ftikos, Ch., ‘Study of the action of Santorin Earth during the
hydration of cements’, PhD Thesis (Athens, 1977).
[13] EXECUSTAT (PWS-Kent, Boston, 1991).
[14] Baragano Coronas, J., Vasquez de la Torre, P. Y., ‘Difficulties in the
manufacture of blended cements’, Translation ZKG, 11/86, 617-621.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 103-111

Lifetime of concrete dam models under constant loads

SCIENTIFIC REPORTS
F. Barpi 1, G. Ferrara2, L. Imperato3, S. Valente1
(1) Politecnico di Torino, Italy
(2) E.N.E.L.-C.R.I.S., Milano, Italy
(3) I.S.M.E.S., Bergamo, Italy.

Paper received: September 23, 1997; Paper accepted: February 18, 1998

A B S T R A C T R É S U M É

This paper presents the results of crack creep tests on
three dam models (240 cm high) made with three types of
concrete, with maximum aggregate size Φ equal to 3.15,
12 and 25 mm. In order to characterize the material a
series of tensile tests was performed for each kind of con-
crete. In this way the static parameters (Young’s modulus
E, ultimate tensile strength σu, compressive strength σc,
fracture energy ) and creep laws were obtained.
The tests on the dam models were performed in many
phases: in the last one, the load was kept constant in order
to examine the effects of concrete relaxation in terms of
Crack Opening Displacement and crack propagation.
Through the Cohesive Crack model it was possible
to simulate the experimental results on gravity dam
models by using the creep laws obtained from tensile
test. Numerical and experimental results (failure life-
time, load vs. C.M.O.D. (Crack Mouth Opening
Displacement), C.M.O.D. vs. time, load vs. time, crack
trajectories) were found to be in good agreement.
Ce travail présente les résultats d’essais de fluage dans la
zone endommagée sur trois modèles de barrage en béton (hauts de
240 cm) avec une dimension maximale des granulats de 3,15,
12 et 25 mm. Le matériau a été caractérisé par des essais de trac-
tion pour chaque type de béton. Ainsi ont été déterminés les
paramètres statiques (module de Young, résistance maximale à la
traction, résistance à la compression, énergie de rupture) et les lois
de fluage du béton.
Les essais sur les modèles de barrages ont été conduits en plu-
sieurs phases : dans la dernière, la charge a été gardée constante
afin d’examiner les effets de relaxation du matériau en termes
d’ouverture de la fissure et de sa propagation.
À travers le modèle de la fissure cohésive, il a été possible de
simuler les résultats expérimentaux des essais sur les modèles de
barrage en appliquant les lois de fluage obtenus dans les essais de
traction. Les résultats numériques et expérimentaux (durée de vie
à la fatigue, courbes de charge-ouverture de la fissure, ouverture
de la fissure-temps, trajectoires des fissures) se sont révélés en
bonne corrélation.

1. STATIC TESTS Table 1 – Mix proportion for concrete

Test Portland 325 Plasticizer Water Sand Gravel Gravel
Table 1 shows the mix propor- series Cement Φ < 3.15 mm 3.15 < Φ < 12 mm 12 < Φ < 25 mm
tion for the concrete used. The (kg/m3) (kg/m3) (kg/m3) (kg/m3) (kg/m3) (kg/m3)
specimens for static and creep tests 2788 300 3 190 915 915 0
were prepared together with the
three dam models. The specimens 2857 270 2.7 170 656 502 772
and the models were stored at 20° C 2859 400 4 240 1506 0 0
and 95% relative humidity. Mean
compressive strength was measured on six 15 × 15 × 15
cm cubes, Young’s modulus on two 15 × 15 × 60 cm
prisms. The results at 8 months are given in Table 2. Table 2 – Mean compressive strength
Stable bending tests were performed at 8 months and mean Young’s modulus
according to RILEM recommendations, on 10 × 10 × 84
Test Maximum Mean compressive Mean Young’s
cm beams with a 5 cm deep notch (testing span 80 cm), series aggregate size, Φ strength, σc modulus, E
in order to determine the fracture energy. Results are (mm) (MPa) (MPa)
summarized in Table 3.
2788 12 28 28000
Stable tensile tests were performed at 8 months on
20 cm high, 10 cm diameter cylinders with a 1 cm deep 2857 25 35 35200
notch. Results are summarized in Table 4. 2859 3.15 32 22100

1359-5997/99 © RILEM 103

Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Table 3 – Stable bending tests results

Test Maximum Maximum Fracture energie
series aggregate size, Φ load, Pmax
(mm) (N) (N/m)

2788 12
774
857
137.79
144.58 } 141.15

2857 25
843
951
139.29
106.69 } 121.49

2859 3.15
794
813
88.21
104.89 } 96.55

Table 4 – Stable tensile tests results

Test Maximum Maximum Fracture Tensile strength,
series aggregate size, Φ load, Pmax enery, σu
(mm) (N) (N/m) (GPa)

2788 12
12446
11172
116.86
102.99
2.51
2.24 } 2.375

2857 25
13377
13524
125.67
119.35
2.84
2.93 } 2.885

2859 3.15
8624
10388
–
48.28
1.78
2.12 } 1.950

The testing machines were Galdabini PM60 of the Fig. 1 – Tensile crack creep experimental setup.
electro-hydraulic and closed loop type, with a loading
capacity of 60 tons. Deformations were measured using
HBM W2 type LVDTs with a measuring range of ± 2
mm and accuracy of 0.2%. In tensile tests, C.O.D.
(Crack Opening Displacement) was measured as the mean
value of the readings obtained on three transducers,
glued at 120° to one another.
In the numerical simulation, the mean values shown
in Tables 3 ad 4 were used. The Poisson ratio, ν, was 0.1.

2. TENSILE CRACK CREEP TESTS

The creep law was determined by means of direct
tensile tests, performed on 20 cm high, 10 cm diameter
cylinders, with a 1 cm deep notch (Fig. 1). C.M.O.D. is
measured as the mean value of the readings obtained on
three transducers, glued at 120° to one another. If speci-
men size is properly chosen compared to the rotational
stiffness of the machine, it is possible to obtain a stable
tensile test, as shown for a similar test in [1]. Due to the
heterogeneity of the material, some acceptable scatter in
the load vs. the C.M.O.D. curve may occur. The age of
the concrete was 8 months.
Fig. 2 – Tensile crack creep tests procedure.
The following phases (Fig. 2) were carried out [2]:
– phase 1: load up to the maximum load Pmax,
– phase 2: follow the softening branch until the load – phase 4: reload up to PG corresponding to a certain
decreases to the prescribed value, PD, percentage (80%, 85%, 90%) of PD,
– phase 3: unload to PE (about 10% of the maximum – phase 5: the load is kept constant (± 5 N) until creep
load Pmax), rupture occurs.

104
Barpi, Ferrara, Imperato, Valente

During phases 1 and 2 a C.O.D. rate of 0.2 µm/s was 3. SUSTAINED LOAD LEVEL VERSUS FAILURE
imposed (C.O.D. control). Phases 3, 4 and 5 were exe- LIFETIME
cuted under load control. The results are summarized in
Tables 5, 6 and 7. Failure lifetime is plotted against sustained load level
Fig. 2 shows that the curve obtained during a static σG /σD for three tensile creep test series in Figs. 3-5. For
test (phases 1, 2 and related extrapolation) can be each series, failure lifetime tcr is expressed as a power
assumed as a valid envelope criterion for creep fracture, function of σG /σD [1, 5-6]:
as suggested in [2]. The same conclusions were drawn −n
σ 
from the observation of other types of concrete (see tcr = a  G  (1)
Table 1) and other sustained load levels PG /PD = 0.80,  σD 
0.85, 0.90 (where PG and PD are defined in Fig. 2). The By best fitting the curves in Figs. 3-5, it is possible to
complete series of nine diagrams is given in [3]. determine the a and n coefficients for each type of concrete:
This criterion implies that fracture energy is com- 3 σ 
−26
pletely independent of the time duration of the sustained tcr = 10  G  for series 2788 (2a)
load, as pointed out in [4] for three-month-old concrete 6.90  σD 
specimens. −19
4 σ 
tcr = 10  G  for series 2857 (2b)
16.72  σD 
−27
3 σ 
tcr = 10  G  for series 2859 (2c)
9.71  σD 
Table 5 – Tensile creep results for test series 2788
Pmax C.O.D. at PD PG /PD PG (C.O.D.)G (C.O.D.)H Failure These functions agree with the experimental
Specimen Pmax lifetime evidence that a small change in load level (for
(kgf) (µ m) (kgf) (%) (kgf) (µ m) (µ m) (s) example from 0.80 to 0.90) causes a change in
2788C 1136 7.1 800 90 720 11.5 15.7 1184 the failure lifetime of more than one order of
2788D 1221 8.3 800 90 720 16.3 24.5 4902 magnitude. This sensitivity was also found in [7]
analysing the growth rate of a crack in a visco-
2788E 1314 9.5 800 85 683 17.2 27.0 11916
elastic tension-softening material.
2788F 1125 6.5 800 85 680 10.0 15.4 4392 The best-fitting equations for direct tensile
2788G 1260 8.0 800 85 680 13.5 22.8 16590 tests show that, for σG /σD > 0.81, failure life-
2788H 1050 7.5 800 80 644 9.1 18.2 62985 time increases with increasing maximum
aggregate size, while the opposite occurs for
2788I 1062 8.4 800 80 640 10.3 17.4 33642
σG /σD < 0.81.
2788L 1235 8.8 800 80 640 14.8 24.8 64867

Table 6 – Tensile creep results for test series 2857 4. CREEP LAW
Pmax C.O.D. at PD PG /PD PG (C.O.D.)G (C.O.D.)H Failure Normalised creep C.O.D. (defined as w c/w cult
Specimen Pmax lifetime
(kgf) (µ m) (kgf) (%) (kgf) (µ m) (µ m) (s) = (w - wG )/(wH - wG ), see Fig. 2) vs. normalized
time (defined as t/tcr ) curves are shown in Figs. 6-8
2857E 900 8.0 800 90 723 12.1 21.6 4693
for each type of concrete. They display a three-
2857F 1383 10.8 800 90 723 18.6 24.7 2626 stage process, according to the change of creep
2857H 1029 9.1 800 85 683 12.3 20.9 30184 rate (defined as w• c/w cult; the dot means derivation
2857I 1160 9.0 800 80 644 16.0 27.4 30058
with respect to time). Creep rate decreases gradu-
ally in the primary stage, is almost constant in the
2857L 1407 9.3 800 80 641 15.5 26.4 35926 secondary stage, and increases rapidly until failure in
the tertiary stage. In terms of failure lifetime, the
Table 7 – Tensile creep results for test series 2859 primary, secondary and tertiary stages take up
about 5%, 75% and 20%, respectively. In other
Pmax C.O.D. at PD PG /PD PG (C.O.D.)G (C.O.D.)H Failure
Specimen Pmax lifetime
words, the secondary stage dominates the failure
(kgf) (µ m) (kgf) (%) (kgf) (µ m) (µ m) (s) lifetime in creep rupture [2, 5].
For each type of concrete, it is observed that
2859C 1116 8.9 640 90 579 12.0 17.9 3051
the difference between creep curves at different
2859G 858 8.4 640 90 575 9.7 12.1 877 levels is not greater than that between different
2859D 615 8.3 640 85 543 8.3 16.0 9781 tests at the same levels. Therefore, it is reason-
2859F 918 9.6 640 85 547 10.9 16.8 10589 able to assume that a unique relationship may exist
between normalised creep C.O.D. and nor-
2859E 1034 8.4 640 80 515 11.1 20.0 28206 malised time, independent of sustained load (or
2859I 649 7.5 640 80 515 7.4 12.7 56760 stress) level.

105
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Fig. 3 – Sustained load level vs. failure time curve in tensile crack
creep test series 2788.

Fig. 6 – Experimental and best-fitting creep curve in tensile

crack creep test series 2788.

Fig. 4 – Sustained load level vs. failure time curve in tensile crack
creep test series 2857.

Fig. 7 – Experimental and best-fitting creep curve in tensile

crack creep test series 2857.

Fig. 5 – Sustained load level vs. failure time curve in tensile crack
creep test series 2859.

Equation (3) is proposed to best f it mean creep

curves for series 2788, 2857 and 2859:

 c
d wc  5
 wult   c  c  c 
= Φ wc  = c1 + c2  wc  + c3 exp −50 wc  (3)
   wult   wult   wult 
d t 
 tcr  Fig. 8 – Experimental and best-fitting creep curve in tensile
The coefficients c1, c2, c3 are determined by a numerical crack creep test series 2859.
integration of Eq. (3) in order to best fit mean creep curves
for the three series (see Figs. 6-8). According to a phenom- and
enological approach, equation (3) is assumed as a constitutive
law for the process zone. From equation (3), it is possible to  c  wc 
• d wc  n d c 
obtain the creep C.O.D. rate as a function of the nor-
w = 1  wult  = 1  σG   wult  = 1  σG  Φ wc  =
n
c
malised creep C.O.D. instead of normalised time [8]:    
c
wult tcr  t  a  σD   t  a  σD   wult c 

d  d 
 c   tcr   tcr 
d wc  d w − wG  • (5)
 wult   wH − wG  dwc tcr n 5 
= = w c
= c tcr  σG    wc   c 
    c
dt wult wult
(4) =   c1 + c2  c  + c3 exp −50 wc  
1
d t  d t  a  σD    wult   wult  
 cr 
t  cr 
t  

106
Barpi, Ferrara, Imperato, Valente

By using equations (1-4), the following creep laws

are obtained (equations (6a-c)):
•
26
wc = 6.90 × 10 −3  σG 
 
c
wult  σD 
(6a)
  c 
5
 
0.40 + 25 w + 9 exp −50 wc   for series 2788
  c   c 
 wult   wult 
 
•
19
wc = 16.72 × 10 −4  σG 
 
c
wult  σD 
 5  (6b)
 wc   c 
0.35 + 32 w 
  c  + 15 exp −50 c   for series 2857
 wult   wult 
  Fig. 9 – Dam model.

•
27
wc = 9.71 × 10 −3  σG 
 
c
wult  σD 
  c 
5
  (6c)
0.45 + 12 w + 35 exp −50 wc   for series 2859
  c   c 
 wult   wult 
 
where:
(a) the first factor, depending on σG /σD, is related to
time until failure tcr through the best fitting functions
shown in Figs. 3-5 and described by equation (1),
(b) the second factor, depending on w c/w cult , is related
to three creep stages through the best fitting functions
shown in Figs. 6-8 and described by equation (3).

5. GRAVITY DAM MODELS

Experimental tests were carried out on three planar
gravity dam models (Fig. 9), as described in [2]. The
geometric scale ratio between model and prototype was
assumed to be Sl = 1:40 (2.40 m (dam model height)/ 96
m (real dam height)). The scale adopted for density was
Sρ = 1, while the scale for stresses was Sσ = 1 (same value
for both the model and the prototype). Having estab- Fig. 10 – Dam model scheme (dimensions in centimeters).
lished the scales of three independent quantities (length,
stress and density), the scales of the other quantities play-
ing a role in static problems are determined according to
Buckingham’s theorem: scheme, notch depth and position, the extension and the
Sσ
gravity center of each elementary portion are shown in
=1 (7) Fig. 10.
S ρ S g Sl
In order to check each vertical force applied, a spring-
Using equation (7), we obtain the scale for the vol- dynamometer is used. As shown in Fig. 11, each
ume forces, Sg: dynamometer is connected to a steel framework placed
Sσ under the foundation of the dam. In order to disturb the
Sg = = 1 = 40 (8) static behaviour of the model as little as possible, the verti-
Sρ Sl 1 × 1 : 40
cal forces in the fracture zone are applied through glued
For this reason, it was necessary to simulate a self- metallic plates.
weight increment (40 times) by means of a system of dis- Another procedure, based on centrifuge equipment is
crete vertical forces (see Fig. 9). The total volume of the described in [9]. This was used to obtain an artificial
model was subdivided into elementary portions and the increment in self-weight in smaller dam models (64 and
equivalent forces to be applied were estimated. The model 96 cm high).

107
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
Fig. 11 – Experimental setup.
Fig. 12 – Dam model testing procedure.
6. TESTING PROCEDURES
The hydrostatic thrust was generated by a 2000 kN
servo-controlled actuator and applied to the upstream side
of the model perpendicularly, according to the distribution
shown in Fig. 11. Five hydraulic jacks applied a vertical dis-
placement to a steel framework, so as to produce an equal
lengthening of all the interposed dynamometers.
Therefore the force applied to each of the loading rods was
proportional to the spring constant of the dynamometers.
C.M.O.D. (Crack Mouth Opening Displacement) was mea-
sured as the mean value of the readings obtained on two
transducers, glued to the opposite sides of the models.
Their base length was 8 cm.
The following steps (Fig. 12) were carried out:
– phase 1: the vertical load is applied, to simulate the
dead-weight increment required (Sg = 40). In order to
prevent concrete damage in compression, two plastic
sheets (0.1 mm thick each) and expansive mortar were
108
placed into the notch. Negative C.M.O.D. at this step is
due to the compression of the plastic sheets and to the
transducer base length of 8 cm. The vertical load is kept
constant during all the following steps,
– phase 2: load up to the maximum hydrostatic load Pmax
at a C.M.O.D. rate of 0.03 µm/s (C.M.O.D. control),
– phase 3: follow the softening branch until the hydro-
static load decreases to the prescr ibed value, P D
(C.M.O.D. control),
– phase 4: unload to PE (load control),
– phase 5: reload to PG < PD (load control),
– phase 6: the hydrostatic load is kept constant until
creep rupture occurs.
The shape of the load-C.M.O.D. diagram plotted for
dam models (Fig. 12) is different from the shape of the
same diagram obtained from tensile tests (Fig. 2). This
difference is due to the vertical load applied to the dam
(phase 1), which is constant during the following steps,
and tends to close the crack. During tensile tests the
effects of self-weight are negligible.
Experimental diagrams are compared (Figs. 13-22)
with the numerical results obtained by using the Cohesive
Crack model [10-12]. The generalisation to mixed-mode
condition is presented in [13]. The further extensions to
creep under constant load are presented in [3] and [8].
The main mechanical hypotheses are:
1) since the sustained load level is high (≥ 0.95), creep
outside the process zone was neglected compared to creep
in the process zone (as show in [14]),
2) on the basis of previous numerical experience, the tan-
gential stresses on process zone were neglected (see, for
instance, [9] and [15]),
3) the creep laws described by equations (6a-c) were used.
Experimental and numerical results related to cen-
trifuge tests on pre-notched gravity dam models show a
static envelope different from one obtained in direct ten-
sile tests (Fig. 2).
Fig. 13 – Load vs. Crack Mouth Opening Displacement for con-
crete 2788.
Barpi, Ferrara, Imperato, Valente

Fig. 17 – Crack Mouth Opening Displacement vs. time for con-

crete 2857.

Fig. 14 – Load vs. Crack Mouth Opening Displacement for con-

crete 2857.

Fig. 18 – Crack Mouth Opening Displacement vs. time for con-

crete 2859.

Fig. 15 – Load vs. Crack Mouth Opening Displacement for con-

crete 2859.

Fig. 19 – Load vs. time for concrete 2788.

Fig. 16 – Crack Mouth Opening Displacement vs. time for con-

crete 2788.

109
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
Fig. 20 – Load vs. time for concrete 2857.
Fig. 21 – Load vs. time for concrete 2859.
For the softening phase, Figs. 13-15 show a smaller
slope, tending to zero. Experimental and numerical results
related to three-point bending tests under constant load [2]
show that creep rupture occurs approximately when the
load path reaches the static envelope. For these two rea-
sons, the constant load level PG /PD was chosen very high,
to be sure that creep collapse would occur.
During the experimental tests, collapse occurred too
early in two cases (concretes 2788 and 2859) over three.
It occurred at the end of the reloading phase so that no
constant load step exists. Nevertheless a comparison
between experimental and numerical results is done in
Figs. 13-21, with the hypotheses that the experimental
time starts at the beginning of phase 1 and the numerical
time starts at the beginning of phase 6 (constant load)
and during the previous phase no creep occurs.
Load vs. C.M.O.D. experimental and numerical dia-
grams (Figs. 13-15) are in good agreement; they show
that the creep rupture occurs before the static envelope is
110
Fig. 22 – Experimental and numerical crack trajectories for con-
crete 2788.
Fig. 23 – Experimental and numerical crack trajectories for con-
crete 2857.
Fig. 24 – Experimental and numerical crack trajectories for con-
crete 2859.
reached, and that the constant load phase is shorter than
in direct tensile tests (Fig. 2).
Fig. 20 shows that dam model with concrete 2857
presents a long constant load phase. In this case, the
numerical results are directly comparable to the experi-
mental ones and they are in good agreement, including
the failure lifetime (tcr = 136000 s).
Finally, Figs. 22-24 show the experimental and numer-
ical crack trajectories. The results are in good agreement.
7. CONCLUSIONS
The main conclusions are:
(a) Creep induced crack propagation in concrete struc-
tures can be analysed within the framework of the
Cohesive Crack model, a well established approach to
the fracture mechanics of quasi-brittle materials.
(b) The results of direct creep tensile tests (h = 20 cm)
seem to be useful in predicting pre-notched gravity dam
models failure lifetime (H = 12 h).
(c) Experimental and numerical load vs. C.M.O.D.
diagrams are in good agreement.
Barpi, Ferrara, Imperato, Valente
(d) The load vs. C.M.O.D. diagrams obtained from
static tests can be assumed as a valid envelope criterion
for creep fracture in mode I problems (direct tensile tests
and three-point bending tests). On the contrary, in the
mixed-mode problems analysed, creep fracture occurred
before the static envelope was reached.
REFERENCES
[1] Zhou, F.P., ‘Some Aspects of Tensile Fracture Behaviour and
Structural Response of Cementitious Materials’, Lund
University (Sweden), Report TVBM-1008 (1988).
[2] Carpinteri, A., Valente, S., Zhou, F.P., Ferrara, G. and Melchiorri,
G., ‘Crack propagation in concrete specimens subjected to sus-
tained loads’, in Wittmann, F.H. (ed.) ‘Fracture Mechanics of
Concrete Structures’ (Aedificatio, 1995) 1315-1328.
[3] Barpi, F. and Valente, S., ‘Time induced crack propagation in
concrete structures: cohesive crack model in mixed-mode condi-
tions’, Politecnico di Torino, Structural Engineering Department
(Research Report A845/95, 1996).
[4] Hansen, A.E., ‘Influence of sustained load on the fracture energy
and the fracture zone of concrete’, in van Mier J.G.M., Rots
J.G., Bakker A. (eds.) ‘RILEM/E.S.I.S. Conference on Fracture
Processes in Brittle Disordered Materials: Concrete, Rock,
Ceramics’ (E&FN SPON, 1991) 829-838.
[5] Reinhardt, H.W. and Cornellissen, H.W.A., ‘Sustained tensile
tests on concrete’, Baustoff 85 (1985) 162-167.
[6] Bažant, Z.P. and Gettu, R., ‘Rate effects and load relaxation in
111
static fracture of concrete’, ACI Materials Journal 89 (5) (1992)
456-468.
[7] Karihaloo, B., ‘Fracture Mechanics and Structural Concrete’
(Longman Scientific and Technical, 1995) 230-239.
[8] Barpi, F., Chillè, F., Imperato, L. and Valente, S., ‘Creep induced
cohesive crack propagation in mixed mode’, in Durban D. and
Pearson J.R.A. (eds.), ‘Non-Linear Singularities in Deformation
and Flow’, (Kluwer Academic Publishers, 1999) 153-166.
[9] Valente, S., Barpi, F., Ferrara, G. and Giuseppetti, G., ‘Numerical
simulation of centrifuge tests on pre-notched gravity dam models’,
in Bourdarot E., Mazars J., Saouma V. (eds.), ‘Workshop on Dam
Fracture and Damage’ (Balkema, 1994) 111-119.
[10] Barenblatt, G.I., ‘The formation of equilibrium cracks during
brittle fracture: general ideas and hypotheses. Axially-symmetric
cracks’, Journal of Applied Mathematics and Mechanics 23 (1959)
622-636.
[11] Dugdale, D.S., ‘Yielding of steel sheets containing slits’, Journal
of Mechanics and Physics of Solids 8 (1960) 100-104.
[12] Hillerborg, A., Modeer, M. and Petersson, P.E., ‘Analysis of
crack formation and crack growth in concrete by means of frac-
ture mechanics and finite elements’, Cement and Concrete Research
6 (1976) 773-782.
[13] Bocca, P., Carpinteri, A. and Valente, S., ‘Mixed-mode fracture
of concrete’, International Journal of Solids and Structures 27 (1991)
1139-1153.
[14] Bažant, Z.P. and Xiang, Y., ‘Crack growth and lifetime of con-
crete under long time loading’, Journal of Engineering Mechanics 4
(1997) 350-358.
[15] Carpinteri, A., Valente, S., Ferrara, G. and Imperato, L.,
‘Experimental and numerical fracture modeling of a gravity dam’,
in Bažant Z.P. (ed.) ‘Fracture Mechanics of Concrete Structures’
(Elsevier Applied Science, 1992) 351-360.
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 112-116
Shear stress contributions and failure mechanisms of
SCIENTIFIC REPORTS
high strength reinforced concrete beams
S. Sarkar, O. Adwan and B. Bose
School of Construction & Environment, University of Abertay Dundee, Bell Street, Dundee DD1 1HG, Scotland.
Paper received: November 19, 1997; Paper accepted: February 25, 1998
A B S T R A C T
The problem of shear capacity of reinforced concrete
members has engaged the attention of many researchers.
Despite this, it has not yet been understood fully
because of the fact that the behaviour of concrete struc-
tures under variable stresses is very complex. Moreover,
little information is available regarding both the shear
stress contribution of different elements and the failure
mode mechanisms of high strength reinforced concrete
(HSRC) members. In this paper an attempt has been
made to discuss in some detail these aspects in the con-
text of high strength reinforced concrete beams. In addi-
tion, new expressions have been developed to predict
the ultimate shear stress of high strength reinforced con-
crete beams.
1. INTRODUCTION
In recent years the use of high strength concrete
(HSC) has increased significantly. This has been due pri-
marily to the development in material technology that
has allowed the successful achievement of the required
high strength. These developments have led to the
recognition of the excellent engineering properties and
economic advantages of this material [1, 2]. However,
very scant information is available regarding the shear
stress contribution made by different components, and
the physical mechanism of diagonal shear cracking.
Due to the complexity of shear failure of concrete
structures and the inter-dependence of factors affecting
the structural behaviour, e.g. concrete compressive
strength, fcu, the tensile reinforcement ratio, ρ, and shear
span to depth ratio, a/d, recommendations of various
codes of practice for predicting the shear capacity of
reinforced concrete beams are mainly based upon exper-
imentally derived equations and are for normal strength
concrete members [3-5]. Consequently, the use of these
equations to predict the shear capacity of HSRC beams
may be inappropriate [6]. Therefore, new equations
were developed by the second author to predict the ulti-
mate shear stress of HSRC beams [7].
1359-5997/99 © RILEM
R É S U M É
Le problème de capacité de cisaillement des éléments en
béton armé a toujours retenu l’attention de nombreux cher-
cheurs. Malgré cela, on ne l’a pas encore complètement
résolu, parce que le comportement des structures en béton
soumises à différentes contraintes est très complexe. De
plus, il existe peu d’informations sur la contribution des
contraintes de cisaillement des différents éléments et les
mécanismes de rupture des poutres en béton armé à haute
ténacité. Dans cet article, nous entamons une discussion
détaillée de ces aspects, dans le contexte des poutres en
béton armé à haute ténacité qui subissent une défaillance
critique du cisaillement. Nous y ajoutons également de
nouvelles expressions développées pour prédire le taux de
cisaillement critique de ces poutres.
2. EXPERIMENTAL INVESTIGATION
Three concrete mix designs, with a single source of
Portland Cement, were used for this investigation. The
fine aggregate used was natural sand having a wet density
of 2.61. The coarse aggregate was porphyrite andesite
crushed stone with maximum size of 20 mm. Silica
fume was used to enhance the strength of the concrete
mix. The workability of the mix was improved by using
a naphthalene-based superplasticizer.
Shear critical reinforced concrete beams were made
with normal and high strength concrete. A total of six
reinforced concrete beams were cast to study the contribu-
tion of different shear elements and failure mode mecha-
nisms. This comprised three reinforced concrete beams
without preformed diagonal cracks (Group I) and three
reinforced concrete beams with preformed diagonal cracks
(Group II). The test variable was the concrete compressive
strength, fcu, which was varied from 40 to 110 MPa.
Beams with preformed cracks (Group II) were iden-
tical to the beams without preformed cracks (Group I) in
all respects, except that a smooth diagonal crack was
introduced in the beam by using a thin steel plate of
0.8 mm thickness which eliminated the aggregate inter-
lock. The width of the plate was the same as that of the
112
Sarkar, Adwan, Bose

The contribution of compression zone, νcz, to shear

was computed analytically from concrete strain readings
measured at different levels across the beam depth,
adopting Taylor’s semi-empirical approach, which is
explained below.
The equation used is:
y
δσ δM dy
υcz = ∫ δM δx
(1)
0
where νcz = shear stress at depth y from the compression
face and at a distance x from the support; σ = longitudi-
nal stress at a distance x from the support; M = moment
at a distance x from the support.
The calculation of νcz involves the following steps:
(i) Plot strain, ε, against bending moment, M, for each
gauge level and from this the slope δε of the plot at
every gauge level is determined. δM
(ii) Determine the total shear, δM , for the load stage
δx
under consideration. This has a constant value at any
load stage for all the gauge level.
(iii) Determine δσ δM by multiplying the product of
Fig. 1 – Specimen details of preformed crack beams. δM δx
expressions in step (i), (ii) and the modulus of elasticity
of concrete, Ec.
(iv) Integrate the expression in step (iii)
Table 1 – Concrete mix proportions for shear programme from the compression face down to each
Mix C S CA SF SP W W/C Nominal gauge level to determine the shear stress at
Strength that gauge level.
kg/m3 kg/m3 3
kg/m 3
L/m 3
L/m 3L/m (MPa)
1 528.88 663.4 1561.33 65.6 14.05 121.64 0.23 115 3.1.2 Shear carried by dowel action
2 528.88 663.4 1561.33 65.6 14.05 163.95 0.31 85 The compression zone shear stress, νcz,
3 371.26 814.37 1221.77 – – 234.22 0.63 40
was computed analytically from the longitu-
dinal strain readings taken in the compression
C = Cement, S = Sand, CA = Coarse aggregate, SF = Silica fume, SP = Superplasticizer,
W = Water.
zone of the beam. The dowel shear resis-
tance, νd, was postulated to be equal to the
numerical difference between the total shear
and the shear contribution due to compression zone
beams, i.e. 150 mm, and each beam had a preformed crack mechanism, i.e.:
in the short side of the beam’s span. The position of the
diagonal crack was determined from the diagonal failure νd = ν - νcz (2)
cracks observed in Group I beams with stirrups up to the As can be seen from Figs. 2-4, the dowel action pro-
loading point. Figs. 1 (a),(b) show specimen details and the vided the predominant shear resistance in all the beams
position of the preformed cracks. All beams with and compared with the compression zone contribution.
without preformed cracks were 1800 mm long, 150 mm
wide, 250 mm deep and of 225 mm effective depth. The 3.1.3 Shear carried by aggregate interlock
ratio ρ was 2.92% and the a/d ratio was 2. The summary of As described earlier the aggregate interlock mecha-
the concrete mix design is given in Table 1. nism was eliminated from all beams in Group II by pre-
forming smooth diagonal cracks. Consequently, the dif-
ference in shear between beams of Group I and beams of
3. RESULTS, ANALYSIS AND DISCUSSION Group II should represent the shear resistance by the
aggregate interlock, νa. The details of Group II beams
3.1 Estimation of shear carried by different are given in Fig. 1.
mechanisms
3.1.1 Shear carried by the compression zone 3.2 Contribution of shear mechanisms to
A semi-empirical method, originally developed by beam shear at ultimate load
Taylor [8], was employed to estimate the shear carried by
the compression zone, νcz, from strain readings taken at Fig. 5 illustrates the shear capacity of Groups I and II
different levels along the beam’s depth. beams at ultimate load. The shear contribution by aggre-

113
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Fig. 2 – Shear carried by various components (HSRW1-1, Group Fig. 5 – Contribution of shear mechanisms at ultimate state.
II Beam).

Table 2 – Ultimate shear capacity and element shear contribution

RC beams without RC beams with
Beam preformed diagonal preformed diagonal
crack group I crack group II
fcu νu fcu νu νcz νd
MPa MPa MPa MPa MPa MPa
HSRW1-1 111 7.37 122 4.38 1.24 3.14
HSRW1-2 95 6.29 92 3.57 0.94 2.63
NSRW1-3 37.5 3.46 40 2.29 0.44 1.85

gate interlock, νa, at any fcu should be the difference

Fig. 3 – Shear carried by various components (HSRW1-2, Group
II Beam).
between the curves representing Groups I and II. The
shear stress contributed by the compression zone, νcz, and
dowel action, νd, are also shown in Fig. 5. The contribu-
tions of various shear transfer mechanisms represented in
terms of the total shear capacity, νu, are given in Table 2.
3.2.1 Significance of aggregate interlock
From the analysis of the test data at low and high
concrete strengths it was inferred that the aggregate
interlock mechanism was a significant contributor to the
total shear carried by the beam, i.e. about 34% for beams
having 40 MPa concrete strength and about 42% for
beams having concrete strength of 110 MPa. The results
indicated clearly that the role of this mechanism at
higher concrete strengths is slightly enhanced. In addi-
tion, this mechanism had a predominant inf luence on
the ultimate load carried by the beam. In other words,
the contribution of this mechanism to the total shear
strength carried by the beam was around 42% for higher
concrete strength beams.

3.2.2 Significance of compression zone shear

Fig. 4 – Shear carried by various components (NSRW1-3, Group
II Beam).
The test results indicated clearly that the shear car-
ried by the compression zone, νcz, showed a trend of
slight increase with higher fcu (see Fig. 5 and Table 3).
However, from a practical point of view νcz remained
fairly constant at high fcu.

114
Sarkar, Adwan, Bose
Table 3 – Percentage contribution of shear
mechanisms to beam strength at ultimate load
Shear Percent of νu* contributed
component by each mechanism
fcu
interlock,
νa 34-40%+ 40 < fcu < 110 MPa
compression,
νcz 13-17 % + 40 < fcu < 110 MPa
dowel,
νd 53-43%+ 40 < fcu < 110 MPa
* νu was calculated from beams without preformed crack (Group I)
+ The first number represents the percentage contribution at lower of fcu and the
second at the upper limit of fcu.
3.2.3 Significance of dowel action
As a result of the sectional rotation about the compres-
sion zone, caused by flexure, shear displacement is created
along the critical horizontal and diagonal cracks. The verti-
cal displacement at the level of the flexural reinforcement
cannot develop unless either the flexural reinforcement or
the surrounding concrete deforms. Thus, a resistance is
induced at the level of the flexural reinforcement, which is
the dowel action. Moreover, it introduces tensile stresses in
the concrete surrounding the flexural reinforcement. As
this stress exceeds the tensile strength of concrete, splitting
occurs along the reinforcement.
The results indicated clearly that the shear carried by
dowel action, νd, is a predominant contributor to the
ultimate shear load carried by the beam. However, as the
concrete compressive strength, fcu, is increased from 40
to 110 MPa, the relative shear carried by dowel action
does not increase, i.e. about 53% of total shear for a
beam having 40 MPa concrete strength and about 43%
of total shear for a beam having 110 MPa concrete
strength (see Fig. 5). Table 3 presents the contribution of
different shear mechanisms at ultimate state for different
concrete strengths.
3.3 Proposed formulae for predicting ultimate
shear stress
The authors have endeavoured to establish suitable
design expressions to evaluate the ultimate shear stress of
HSRC beams without web reinforcement. Most of
design code provisions [9, 10] pertaining to the shear
stress were based on the assumption that the useful shear
capacity of a reinforced concrete member is exhausted
once diagonal cracking develops. However, this is a con-
servative assumption; the shear capacity of a reinforced
concrete beam can be considerably higher at failure.
Formulae similar to that proposed by Zsutty [11], but
applicable to HSRC beams were developed using multi-
ple regression analysis. These expressions were based on
a total of 130 HSRC beams tested by the authors and
others, which failed in shear. Readers interested in the
115
comparison between predicted and measured shear stress
should consult reference [7].
Basically, the available data were classified into two
categories, (i) beams possessing a/d ≤ 2 and (ii) beams
having a/d > 2. The general form of the ultimate shear
stress equation for beams without web reinforcement is:
νn = β (fc . ρ . d/a)n (3)
where, νn is the nominal shear stress β is a constant and
fc = cylinder strength = 0.85 fcu.
Using the above considerations the following equa-
tions were developed:
(i) For beams having a/d ≤ 2:
νn = 4.13 (fc . ρ . d/a)0.66 (4)
(ii) For beams having a/d > 2:
νn = 3.05 (fc . ρ . a/a)0.55 (5)
The above equations are valid for reinforced concrete
beams with concrete strength in the range 40 < fc <110
MPa. The correlation coefficients are 0.80 and 0.90 for
equations (4) and (5) respectively, which indicate a positive
correlation. It should be recognised that in the past no
shear equations were proposed for such a wide range of
concrete strength for predicting the ultimate shear stress of
HSRC beams without web reinforcement. The measured
and predicted shear stress values are shown in Fig. 6.
Fig. 6 – Comparison between predicted and measured shear
stresses.
4 CONCLUSIONS
On the basis of the results obtained and analysis made
concerning the shear stress contribution carried by differ-
ent mechanisms and the physical mechanism of the diago-
nal shear crack, the following conclusions can be drawn:
(i) In the absence of aggregate interlock mechanism,
dowel action was the predominant contr ibutor.
However, its contribution decreases with the increase of
concrete strength.
(ii) For beams having concrete strength in the range of
40 < fcu < 110 MPa the shear contribution of various
components at ultimate state was found to be:
νa in the range of 34 - 40%
νcz in the range of 13 - 17%
νd in the range of 53 - 43%
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
(iii) The ultimate shear capacities obtained using Zsutty’s
equations are in close agreement with the experimental
results.
(iv) Regression equations predict the ultimate shear
stress, νu, of HSRC beams which are in reasonable
agreement with test data.
ACKNOWLEDGEMENTS
The investigation into high strength reinforced con-
crete beams, of which this paper reports only a part, was
supported by grants from the Research Committee of
the University of Abertay Dundee. The authors are
greatly indebted to Dr. J. G. L. Munday and Dr. J. R.
Underwood of the University of Abertay Dundee for
advice and support.
REFERENCES
[1] Nawy, E.G., ‘Fundamentals of high strength performance con-
crete’, (Longman, London, 1996).
116
[2] Price, W.F. and Hynes, J.P., ‘In-Situ strength testing of high
strength concrete’, Magazine of concrete research 48 (16) (1996)
189-197.
[3] ACI Committee 363, ‘State of art report on high strength con-
crete’, ACI J. 81 (4) (1984) 364-411.
[4] Sarkar, S., Adwan, O.K. and Munday, J., ‘Shear capacity and duc-
tility of high strength reinforced concrete Beams’, Structures in
the new millennium, the Fourth International Kerensky
Conference, (1997), 405-411.
[5] Sarkar, S., Adwan. O. and Munday, J., ‘High strength concrete:
An investigation of the flexural behaviour of HSRC beams’, The
Structural Engineer 75 (7) (1997) 115-121.
[6] Kim, K. and Park, Y., ‘Shear strength of reinforced high strength
concrete beams without web reinforcement’ Magazine of concrete
research 46 (166) (1994) 7-16.
[7] Adwan,O.K., ‘Engineering properties and structural behaviour of
high strength reinforced concrete beams’, Ph.D. thesis,
University of Abertay Dundee, 1997.
[8] Taylor, H.P.T., ‘Further tests to determine shear stresses in rein-
forced concrete beams’, Technical report TR 42.438, (Cement
and Concrete Associations, London, 1970).
[9] British Standards Institution, BS 8110:1985, Structural use of
concrete, Part 1, Code of practice for design and construction.
[10] ACI Committee 318:1995, Building code requirements for
reinforced concrete, ACI, Detroit.
[11] Zsutty, T.C., ‘Beam shear strength prediction by analysis of
existing data’, ACI J. 65 (November 1968) 942-951.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 117-124

Évaluation expérimentale des caractéristiques modales

SCIENTIFIC REPORTS
et ajustement du modèle dynamique d’un bâtiment de
portiques
(Experimental evaluation of the modal characteristics of a framed building and adjustment of
the mathematical model)

C. Genatios1, G. Cascante2, M. Lafuente1, O. Luente1, P. Bellan3, M. Lorrain4

Article reçu : 1er décembre 1997 ; Article accepté : 11 mars 1998

R É S U M É A B S T R A C T

Cet article présente l’évaluation des propriétés dyna-
miques (fréquences, coefficients d’amortissement, et
formes modales) d’un bâtiment de portiques à cinq étages,
au moyen de plusieurs types d’essais (vibration libre, har-
monique et environnementale). Les résultats ont été simi-
laires, malgré les différences remarquables des amplitudes
de vibration, vingt fois plus faibles dans les essais de vibra-
tions environnementales que dans les essais de vibration
libre. On traite spécialement l’essai de vibration libre et
l’analyse des résultats dans le domaine de la fréquence.
L’ajustement du modèle dynamique est aussi effectué, dans
le but d’étudier l’influence de l’inclusion dans ce modèle
de la maçonnerie de remplissage, de l’interaction sol-struc-
ture, des escaliers, et des dalles. Le modèle proposé repro-
duit convenablement les résultats expérimentaux obtenus
pour les trois premiers modes.
This article presents the evaluation of the dynamic
properties (frequencies, damping ratios, and modal shapes)
of a five-story framed building by various experimental
procedures (free, harmonic and ambient vibration tests).
The results were very similar, even though ambient vibra-
tions tests deal with amplitudes 20 times lower than free
vibration. Special attention is given to free vibration in the
frequency domain. Mathematical model adjustment is also
carried out. A model including masonry, soil structure iter-
ation, stairs, and slab influences reproduced quite accu-
rately the experimental values of the dynamic properties of
the first three modes.

1. INTRODUCTION Il est, en effet, nécessaire de connaître d’une manière

À l’heure actuelle on dispose de méthodologies très
avancées pour la modélisation mathématique numérique
des structures. Leur utilisation pour l’évaluation des
caractéristiques dynamiques permet la réalisation d’ana-
lyses qui ont comme but la vérification des dimensions
et le renforcement éventuel des structures érigées dans
des zones à risque sismique.
Cependant, même si les techniques dont on dispose
sont très appropr iées, certaines des hypothèses
employées, certaines valeurs attribuées aux propriétés
mécaniques des matériaux, doivent être vérifiées et le cas
échéant révisées à partir d’essais in situ, ce qui pose le
problème du recalage des modélisations de calcul.
plus exacte les propriétés structurales des systèmes
employés pour améliorer les modèles de calcul, et, de ce
fait, mieux évaluer les durées de vie résiduelles de bâti-
ments pathologiques, augmenter la sécurité des construc-
tions par des réparations appropriées, rationaliser l’utilisa-
tion des matériaux, en particulier pour ce qui concerne les
structures situées en zones sismiques [1-9].
Dans cet article, nous présentons les résultats de l’éva-
luation des propriétés dynamiques d’un bâtiment à cinq
étages en béton armé, immeuble de logement social, de
type très fréquemment construit au Vénézuela. Ce bâti-
ment appartient à l’Institut National du Logement
(INAVI) du Venezuela et se trouve dans la ville de Guatire,
à 60 km approximativement à l’est de Caracas.

(1) Prof. Instituto de Materiales y Modelos Estructurales (IMME) Universidad Central de Venezuela, IMME-UCV, P.O. 50361 Caracas
1050A, Venezuela
(2) INTEVEP, Caracas, Venezuela
(3) Directeur du Service Informatique, Institut National des Sciences Appliquées (INSA) Toulouse, France
(4) Prof. Laboratoire d’Études Thermiques et Mécaniques, Département de Génie Civil, INSA, Av. de Rangueil, 31077 Toulouse Cedex, France

1359-5997/99 © RILEM 117

Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Après un premier paragraphe consacré à la

description des caractéristiques du bâtiment
étudié, nous présentons dans le deuxième para-
graphe les résultats de l’évaluation expérimen-
tale des propriétés dynamiques. La comparaison
avec les résultats de la modélisation numérique
initiale puis le recalage de cette modélisation et
sa validation font l’objet du troisième et dernier
paragraphe.

2. CARACTÉRISTIQUES DU
BÂTIMENT ÉTUDIÉ

La structure du bâtiment est composée de

portiques en béton armé, placés suivant deux
Fig. 1 – Bâtiment de portiques.
directions perpendiculaires. Pour leur dimen-
sionnement initial, on a utilisé la norme anti-
sismique provisoire vénézuélienne de 1967 et
le code ACI 318-77 avec un coeff icient
d’action sismique de 0,10 (10% de la gravité).
La Fig. 1 montre les dimensions générales de
ce bâtiment.
Il s’agît d’un bâtiment à cinq étages avec
quatre appartements par niveau. Son poids
approximatif est de 12 550 kN (hors charges
variables). Le béton a une résistance caractéris-
tique de f ′c = 21 MPa (210 daN/cm2) et les
armatures sont en acier FeE400 (fy = 420 MPa
(4 200 daN/cm2)). La maçonnerie est compo-
sée de briques creuses d’argile qui ont une
épaisseur de 15 cm. Le système de fondations
est composé d’une dalle de 50 cm d’épaisseur,
placée sur un sol mou d’origine sédimentaire.

3. ÉTUDE EXPÉRIMENTALE
Cette étude a été destinée à permettre
l’évaluation des propriétés dynamiques réelles
du bâtiment pour les comparer aux valeurs
prises a priori dans les modélisation numé-
riques et pour opérer un recalage ultérieur.

Fig.2 – Distribution des accéléromètres et du générateur des vibrations.

3.1 L’instrumentation et la procédure
d’essai
Des essais de vibration libre, vibration forcée harmo-
nique et vibration environnementale ont été effectués
sur le bâtiment.
Pour l’essai de vibration libre, le mécanisme d’excita-
tion a été conçu sur la base d’une déformation initiale [10],
obtenue grâce à l’application d’une charge horizontale
ayant des valeurs de 50 kN.
L’essai de vibration forcée harmonique a été mis en
œuvre au moyen d’un générateur de vibrations (généra-
teur de masses contrarotatives), pouvait produire des fré-
quences de 3,5 à 8,5 Hz [11].
L’essai de vibration environnementale a été conduit à
partir des mouvements de faible intensité communiqués
à la structure par son environnement. En général ces
vibrations sont dues au passage de véhicules et au vent de
faible intensité. Le temps de mesure utilisé a été de 15
minutes approximativement [6, 8, 9, 12, 13].
Pour le relevé des enregistrements, six accéléromètres de
haute sensibilité (Kinemetrics FBA-11), ayant une gamme
de fréquences de 0 à 50 Hz, ont été utilisés. Les signaux des
accéléromètres sont amplifiés et filtrés au moyen d’un
conditionneur de signaux (Kinemetrics SC-1), puis
envoyés à un magnétophone à huit bandes (HP-3968A)
ayant une bande dynamique approximative de 40 db. Un

118
Genatios, Cascante, Lafuente, Luente, Bellan, Lorrain

enregistreur à quatre bandes (Soltec-5M28) a permis de torsion peut également produire des effets d’interfé-
l’observation des signaux in situ. Un analyseur dynamique rence modale.
de structures (HP-5423A) a été utilisé pour le traitement de Pour la direction E-O le quotient w2/w1 est égal à
ces signaux, avec une unité graphique (HP-7470A). Pour 3,25 tandis que pour la direction N-S il est de 3,50.
l’étude des signaux dans l’espace des temps, un enregistreur Tous les deux sont très proches du comportement de
oscillographe à lumière ultraviolet (Honeywell-1508B) a cisaillement pur de 3,0 qui est typique des bâtiments
été employé, avec un oscilloscope à deux canaux composés de portiques.
(Tectronix-468). La Fig. 2 montre la localisation des accélé- Le Tableau 2 donne les coefficients d’amortissement.
romètres et du générateur. Le premier mode N-S a une valeur un peu élevée d’amor-
tissement, laquelle peut être due à l’interférence modale
produite par le premier mode de torsion [9, 15-17]. Cette
3.2 Résultats de l’évaluation expérimentale interférence peut aussi affecter le premier mode de la
direction E-O, mais dans une moindre mesure, étant
Les enregistrements des réponses des essais sont pré- donné que les fréquences concernées sont un peu plus
sentés sur les Figs. 3 à 7. Les propriétés dynamiques ont séparées. Les coefficients d’amortissement ont été calculés
été évaluées en calculant les moyennes des résultats obte- au moyen des méthodologies suivantes : décrément loga-
nus pour chaque accéléromètre, étant donné que les rithmique (vibration libre), largeur de bande (vibration
coefficients de variation sont très faibles. harmonique, module de la transformée à la fréquence pour
Les valeurs des fréquences modales sont consignées la vibration libre et pour la vibration environnementale) [1,
dans le Tableau 1. On remarque que les résultats obtenus à 9, 14], méthode de la tangente (pour les enregistrements
partir des différents types d’essai sont très similaires, même d’accélération de l’essai de vibration harmonique) [1, 9,
si les niveaux d’accélération des vibrations environnemen- 14], méthode de la largeur de bande modifiée (pour les
tales sont vingt fois plus faibles que ceux de la vibration composantes réelle ou imaginaire de la transformée de
libre et huit fois plus faibles que ceux de la vibration har- Fourier de la vibration libre) [9, 10] et moindre carrés
monique. On peut observer une certaine tendance à obte- (pour vibration harmonique et environnementale, à partir
nir des valeurs plus faibles de fréquence pour les essais met- de l’ajustement de la fonction de transfert) [9, 12].
tant en jeu des forces plus importantes sur la structure. La Fig. 8 montre les formes modales dans les directions
Cela peut être associé à des effets de non-linéarité du com- E-O et N-S, et la première forme de torsion. Étant donné
portement des murs en maçonnerie. La proximité des fré- que le bâtiment présente une double symétrie, les modes
quences du premier mode N-S et celle du premier mode peuvent très bien être représentés par une seule coordon-
née latérale. Les résultats concernant la première forme
sont très proches pour les trois types d’essais et pour les
Tableau 1 – Fréquences bât. portiques deux directions. Pour le deuxième mode on trouve des
Fréquences (Hz) perturbations dues aux faibles amplitudes des spectres. Le
déplacement de la base représente 10% du déplacement du
φi Essai E-O N-S Torsion
= dernier niveau pour la direction E-O et 16% pour la direc-
Vib. libre 3,88 4,69 – tion N-S, montrant ainsi l’importance de l’interaction sol-
1 Vib. forcée 4,00 4,75 5,00 structure [9, 14]. Cette variation est due à une plus grande
Vib. envir. 4,00 5,00 5,00
raideur du bâtiment dans la direction N-S. Le premier
mode de torsion a été trouvé à partir des essais de vibration
Vib. libre 13,46 17,28 – forcée harmonique et de vibration environnementale
2
Vib. envir. 14,00 17,33 – puisque pour le cas de vibration libre, la force a été appli-
quée directement au centre de masse. L’interac-
Tableau 2 – Coefficients d’amortissement / bât. portiques
tion sol-structure s’avère être assez importante :
la rotation de la base équivaut à 10% du mou-
φi Direct. E-O Direct. N-S Torsion vement du dernier niveau.
= Essai Méthode
εi % σ* N* εi % σ* N* εi % σ* N*
Dec. logar. 2,5 0,2 4 4,1 0,7 4 1,8 0,7 4
Vib. 4 ÉTUDE NUMÉRIQUE ET
Mod. trans. 3,6 0,1 6 3,9 0,2 6 – – –
libre RECALAGE MATHÉMATIQUE
Comp. réel-im. 3,0 0,3 5 4,7 0,1 6 – – –
Lar. bande 5,3 0,0 6 5,9 0,1 6 4,8 0,4 2 4.1 Le logiciel
1 Vib.
Moin. carrés 4,3 0,3 6 4,9 0,2 6 4,1 0,0 2 L’analyse du bâtiment a été réalisée en plu-
forcée
Tangentes 3,1 0,3 4 5,2 0,6 5 – – – sieurs étapes. D’abord le bâtiment a été
Vib. Lar. bande 3,0 0,4 6 6,0 0,2 5 1,9 0,3 4 modélisé avec les techniques traditionnelles
d’analyse structurale. Ensuite, nous avons pris
envir. Moin. carrés 2,6 0,2 5 4,7 0,4 5 1,5 0,4 4
en compte des particularités du bâtiment au
Vib. Mod. trans. 8,9 – 1 5,3 0,3 5 – – – moyen des options disponibles dans le logiciel
2
libre Comp. réel-im. 3,5 0,7 4 4,8 0,4 5 – – – utilisé : ETABS (Three Dimensional Analysis

119
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

of Building Systems) développé sous

la direction du prof. E. Wilson de
l’Université de California, Berkeley
[18, 19].
Le ETABS est un logiciel d’ana-
lyse bi et tridimensionnel de bâti-
ments soumis à des charges sis-
miques, du vent et des charges
verticales. Il effectue l’analyse par
spectres ainsi que dans le temps. Le
bâtiment est analysé comme étant
composé par des portiques verticaux
unis par des dalles horizontales, tout
en garantissant la compatibilité de
déplacements (3 degrés de liberté par
étage). Des dispositions provenant de
divers codes sont disponibles pour la
considération des charges du vent et
sismiques, les combinaisons des
charges pouvant être appliquées auto-
matiquement. Divers algorithmes de
génération de maillage sont proposés.
Les éléments disponibles sont des élé-
ments de portique (poteaux et
poutres de 3 ou 6 degrés de liberté
Fig. 3 – Vibration libre (accélérations).
par nœud), mais aussi des éléments de
coque garantissant la compatibilité de
déplacements et des éléments pour la
modélisation des murs (“shear walls”)
et pour la considération de la maçon-
nerie. Il est possible aussi de tenir
compte de la f lexibilité des fonda-
tions (intération sol-structure) au
moyen de ressorts et au moyen d’un
étage fictif (“dummy level”). Égale-
ment, les effets P-Delta sont pris en
compte par une procédure de correc-
tion géométrique.

4.2 Modélisation initiale

Pour l’étude numérique du bâti-
ment en question, nous avons pris
en compte, lors de la modélisation
initiale, les déformations par
cisaillement des poutres et poteaux,
les déformations axiales des
colonnes, les dalles étant considé-
rées comme infiniment rigides dans
leur plan et inf iniment f lexibles
dans la direction perpendiculaire à
leur plan. La matrice de raideur est
condensée à trois degrés de liberté
Fig. 4 – Module de la transformée de Fourier des enregistrements de vibration libre
(accélérations).
par étage. L’interaction sol-structure
est initialement négligée, respectant
de cette façon la procédure tradi-
tionnelle d’analyse, considérant le
bâtiment comme encastré au niveau
de la fondation.

120
Genatios, Cascante, Lafuente, Luente, Bellan, Lorrain

Fig. 6 – Enregistrements des vibrations

Fig. 5 – Amplitudes de réponse harmonique (accélérations). environnementales (accélérations).

Fig. 7 – Spectres de puissance de la vibration environnementale.

121
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Fig. 8 – Formes modales bâtiments de portiques. Fig. 9 – Comparaison des modes expérimentaux et numériques.

122
Genatios, Cascante, Lafuente, Luente, Bellan, Lorrain
Les trois premières fréquences de vibration résultant
de cette modélisation initiale prennent les valeurs sui-
vantes :
w1(E-O) = 1,29 Hz. w1(N-S) = 2,44 Hz.
w1(torsion) = 1,68 Hz.
Les résultats obtenus par l’expérimentation sont bien
différents, puisqu’on avait enregistré :
w1(E-O) = 3,96 Hz. w1(N-S) = 4,81 Hz.
w1(torsion) = 5,00 Hz,
avec une bonne corrélation entre les diverses procédures
d’essai.
Des écarts aussi considérables imposent la correction
de la modélisation initiale par la technique du recalage et,
selon l’efficacité du recalage, de l’incorporation dans
l’analyse d’hypothèses complémentaires, plus fines et
plus en accord avec la réalité de la structure.
4.3 Le recalage
La procédure de recalage consiste à introduire cer-
taines modifications dans la modélisation initiale dans le
but de restituer plus fidèlement la réalité du comporte-
ment dynamique du bâtiment. Nous avons ainsi tenu
compte de :
– la présence des murs en maçonnerie,
– la contribution des dalles à la rigidité des poutres en
direction E-O,
– la présence des escaliers et leur contribution aux rai-
deurs des portiques C et D,
– la flexibilité des fondations, simulée au moyen d’un étage
fictif ajouté représentant l’interaction sol-structure [14].
La modélisation ainsi recalée fournit des résultats
analytiques très proches de ceux issus des essais.
Dans le but de quantifier l’influence de chacune des
variables mentionnées sur les propriétés dynamiques du
bâtiment, quatre modèles additionnels ont été considé-
rés. Le modèle 2 ne considère pas les escaliers; le modèle
3 ne considère pas la présence des murs en maçonnerie,
tandis que le modèle 4 ne considère pas la flexibilité de
la fondation. Finalement, un cinquième modèle est pro-
posé ne comportant pas d’escaliers ni de murs, et négli-
geant la f lexibilité de la fondation. Ce dernier modèle
cherche à reproduire les hypothèses d’usage courant dans
l’analyse structurale des bâtiments.
4.4 Comparaison des résultats analytiques et
expérimentaux
Les comparaisons que nous rapportons ci-après,
prennent en compte les moyennes des caractéristiques
modales évaluées expérimentalement.
a) fréquences modales
Le Tableau 3 présente les fréquences des premiers
modes dans les directions E-O, N-S et de torsion, four-
nies aussi bien par la voie expérimentale que par la voie
numérique. Ceci montre la bonne corrélation du pre-
mier modèle avec l’expérience.
123
Tableau 3 – Fréquences expérimentales et théoriques (Hz)
Premier mode Expert. 1 2 3 4 5
E-O 3,96 3,87 3,60 1,82 4,20 1,29
N-S 4,81 5,27 5,27 2,39 5,91 2,44
Torsion 5,00 4,52 4,51 1,68 4,70 1,68
Dans la direction E-O, la suppression des escaliers
(modèle 2) génère une réduction de la première fréquence
de 7%, c’est-à-dire une variation de 14% de la raideur
dans cette direction. La présence de la maçonnerie est très
importante, étant donné que son élimination du modèle
(modèle 3), entraîne une diminution des fréquences de
50%. Également, dans le modèle 3, la distribution des
modes est changée, le mode de torsion devenant le pre-
mier mode du fait de la présence des escaliers et de la nou-
velle distribution des rigidités que ceci entraîne. L’élimi-
nation de l’étage fictif (modèle 4) produit une diminution
de la fréquence de 12% dans la direction N-S et de 9%
dans la direction E-O, montrant ainsi que dans la direc-
tion N-S, la plus rigide, l’étage fictif est plus flexible que
dans la direction E-O. Le modèle 5, d’utilisation plus fré-
quente pour l’analyse sismique des structures, fournit des
résultats de fréquences plus basses : 65% plus faibles pour
la direction E-O et la direction de torsion, et 54% pour la
direction N-S.
b) formes modales
La comparaison des formes modales expérimentales
et numériques mène également à des résultats très satis-
faisants. Étant donné la symétrie du bâtiment, les formes
modales présentent des déplacements dans une seule
direction. Le modèle 1 donne les résultats les plus
proches des évaluations expérimentales (Fig. 9).
Pour le modèle 3 (sans murs en maçonnerie), on
observe une variation par rapport aux déplacements du
premier et du deuxième étages. Cela est dû au fait que la
diminution des raideurs de la structure rend plus impor-
tante la raideur de l’étage fictif vis-à-vis de la structure, et
les déplacements des premiers étages s’en trouvent limités.
L’élimination de l’étage fictif entraîne des variations consi-
dérables sur les formes modales, étant donné l’importance
de l’interaction sol-structure. L’élimination des murs en
maçonnerie sur le modèle de fondation rigide fournit des
formes modales tout à fait en accord avec le comportement
de cisaillement d’un bâtiment en portiques.
La première forme modale en torsion ne se voit guère
affectée par la présence de l’étage fictif, ce qui met en évi-
dence la raideur élevée de cet étage fictif en torsion. Ceci
est aussi dû au fait que le modèle 1 présente un niveau de
rotation plus faible que le bâtiment réel. Cette différence
peut être diminuée en changeant de position les éléments
qui fournissent la rigidité à l’étage fictif [14, 17].
5 CONCLUSIONS
Nous avons déterminé les principales propriétés
dynamiques d’un bâtiment en portiques à cinq étages et
en béton armé, à partir d’essais de vibration libre, de
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
vibration forcée harmonique et de vibration environne-
mentale. Les résultats des évaluations expérimentales ont
été comparés aux résultats fournis par des modèles
numériques de complexité croissante. Les conclusions
les plus importantes du travail sont résumées ci-après.
L’utilisation de l’analyse de vibration libre dans l’espace
de la fréquence, nous a permis de déterminer les proprié-
tés des modes supérieurs au premier. Une variante de la
méthode de la largeur de bande a été appliquée pour la
détermination des coefficients d’amortissement, à partir
des résultats des trois types d’essai. Dans le cas où les fré-
quences sont relativement proches, l’utilisation du module
de la transformée et de sa composante imaginaire, nous a
permis d’obtenir les meilleurs résultats pour l’évaluation
des coefficients d’amortissement [10].
Les propriétés de deux premiers modes dans chaque
direction, ainsi que celles du premier mode de torsion ont
été déterminées. La proximité des fréquences produit des
effets d’interférence modale faisant apparaître des valeurs
d’amortissement élevées (3% à 6%) [9, 13, 14, 16, 17, 20,
21], résultant entre autres de la forte interaction sol-struc-
ture [1, 14, 16]. Celle-ci est de l’ordre de 10% pour la
direction E-O et de 16% pour la direction N-S.
Les différents types d’essais produisent des résultats
comparables, à l’exception de l’essai de vibration libre,
lequel fournit de plus basses fréquences pour le bâtiment
en portiques. Cela s’explique par le fait que cet essai
comporte des amplitudes de force plus élevées, qui
entraînent une réponse non-linéaire du bâtiment.
L’essai de vibration environnementale est le plus facile
et rapide à réaliser, du fait qu’il ne requiert l’utilisation
d’aucun dispositif d’application de charge ; cependant son
utilisation demande une analyse plus compliquée des
signaux de réponse [9, 12, 13]. L’essai de vibration libre, et
son analyse dans l’espace de la fréquence, s’est montré le
plus fiable.
Les premiers trois modes présentent des résultats ana-
lytiques de fréquences et de formes modales très proches
des résultats expérimentaux. Le modèle numérique tient
compte de la présence des murs en maçonnerie, des esca-
liers, de la contribution des dalles à la rigidité des poutres
et de l’interaction sol structure. En raison des limitations
du modèle de flexibilité du terrain que nous avons mis en
œuvre, les deuxièmes modes n’ont pas pu être reproduits
simultanément. L’élimination des escaliers entraîne une
réduction des fréquences de 10%. L’élimination des murs
en maçonnerie réduit les fréquences de 60% et l’élimina-
tion simultanée des escaliers, des murs et de la fondation
flexible produit une diminution des fréquences d’environ
65%. L’élimination de l’étage fictif entraîne une augmen-
tation des fréquences de 12%.
RÉFÉRENCES BIBLIOGRAPHIQUES
[1] Hudson, D. ‘Dynamic testing of full-scale structures’, Chapter
VII in Earthquake Engineering, (Edited by R.L.Wiegel, Prentice
Hall, 1970).
[2] Bernardini, A., Gori, R. and Modena, C. ‘Structural identifica-
tion of a monument for its seismic behavior forecasting’, 9th
European Conference on Earthquake Engineering, (9ECEE),
124
(Moscow, Sept. 1990).
[3] Vestroni, F., Capecchi, D., Antonacci, E., Di Pasquale, G. and
Silvano, G. ‘Indagine sperimentale e modelli interpretativi del
comportamento dinamico di una casa in murata del ‘700’, Conf.
C.N.R.: (Gruppo Nazionale per la difesa dai terremoti, Italia,
1990).
[4] Kapsarov,H. ‘Dynamic properties of a 305 m high reinforced
concrete chimney based on ambient, forced and free vibration
experiments’, 9ECEE, (Moscow, Sept.1990).
[5] Naito, Y. and Ishibashi, T. ‘Identification of structural systems
from microtremors and accuracy factors’, 11 World Conference
on Earthquake Engineering (11WCEE), Acapulco, México,
[6] Felber, A. and Cantieni, R. ‘Advances in ambient vibration
Testing: Ganter bridge, Switzerland’, Journ. International
Association for Bridges and Structural Engineering 3 (1996) 187-191.
[7] Prion, H. and Rezai, K. ‘Damage detection by modal testing’,
11WCEE, México 1996, (Ed. Balkena, 1996).
[8] Skrinar, M. and Strukelj, A. ‘Eigenfrequency monitoring during
bridge erection’, Journal International Association for bridges and
Structural Engineering 3 (1996) 191-194.
[9] Genatios, C. ‘Contribution à l’évaluation des procédés expéri-
mentaux pour la détermination des propriétés dynamiques des
structures’, Thèse Dr. INSA Toulouse France, 1991.
[10] Genatios, C., Lafuente, M. and Lorrain, M. ‘On the use of free
vibration frequency techniques for dynamic properties evalua-
tion’, 10th World Conf. on Earthquake Eng. (10WCEE)
Madrid, 1992 (Edited Balkena, 1992).
[11] Genatios, C., Lafuente, M. and Lorrain, M. ‘Sur la technique de
vibration harmonique pour l’évaluation des propriétés dyna-
miques des structures’, Annales des ponts et chaussée 66 (2) (1993).
[12] Kinemetrics, Inc. ‘Salmon Creek dam ambient vibration measu-
rements’ (edited by Kinemetrics Pasadena, USA, 1982).
[13] Genatios, C., Lafuente, M. and Lorrain, M. ‘Sur la détermina-
tion des propriétés dynamiques des structures à partir de l’action
des vibrations environnantes’, Revue Française de Mécanique 1995
(3) 227-238.
[14] Bouwkamp, J.G., Kollegger, J.P. and Stephen, R.M. ‘Dynamic
properties of a twelve-story prefabricated panel building’, (edited
by University of California, Berkeley, repport no. UBC/EERC-
80/29, 1980).
[15] Mertens, M ‘The complex stiffness method to detect and iden-
tify non-linear behavior of SDOF systems’ J. Mechanical Systems
and Signal Processing 3 (1), (1989) 37-54.
[16] Afra, H., Argoul, P. and Bard, P. ‘Identification of building
structural behaviour from earthquake records’, 9th European
Conference on Earthquake Engineering, (9ECEE), (Moscow,
Sept.1990).
[17] Murià-Vila, D. and González, R. ‘Seismic response of a tall buil-
ding in México City’, 11WCEE, México 1996, (Ed. Balkena,
1996)
[18] Wilson, E., Hollings, J. and Dovey, H. ‘ETABS three dimensio-
nal analysis of building systems’ (edited by University of
California, Berkeley, repport no. UBC/EERC-75/13, 1975).
[19] Habibullah, A., ‘ETABS users manual’ (Edit by Computers and
Structures, 1918 University Av Berkeley, California, 94704
USA, Feb. 1989).
[20] Sawada, T., Hata, K. and Tsujihara, O., ‘A localized identifica-
tion of mdof structures by extended Kalman Filter considering
noise in input motion records’, 11WCEE, México 1996, (Ed.
Balkena, 1996).
[21] Rouanet, C., ‘Contribution à l’identification structurale : adé-
quation d’un modèle mathématique à des résultats d’essais’,
(Thèse Dr École Nat. Supérieure de l’Aéronaut. et l’Espace,
Toulouse, France, p. 206, 1987).
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 125-130
A model for cracking in fibre-reinforced concrete
structures
P. Rossi
LCPC, Division des Bétons et des Composites Cimentaires, 58 bd Lefèbvre, 75732 Paris Cedex 15.
Paper received: June 30, 1998; Paper accepted: July 17, 1998
A B S T R A C T
After a quick review concerning the probabilistic discrete
cracking model relative to non-steel fibre reinforced con-
crete (Rossi probabilistic discrete cracking model) and a crit-
ical analysis of its extension to steel fibre reinforced concrete,
a new Rossi probabilistic discrete cracking model applied to
steel fibre reinforced concrete is proposed. In this model,
cracking is modelled through contact elements which will
have a perfect elastic-plastic behavior with brittle fracture.
The plastic step as well as the postcracking energy are uncor-
related random variables. Numerical work still remains,
consisting in adapting the existing contact elements to the
case of perfect elastic-plastic behavior with brittle fracture, as
well as work on a suitable resolution algorithm.
1. INTRODUCTION
This article analyzes the limits of the probabilistic dis-
crete cracking model for fibre-reinforced concrete devel-
oped at the LCPC (French transport research laboratories)
and proposes possible improvements to the model.
2. ROSSI PROBABILISTIC DISCRETE
CRACKING MODEL
This model, which we shall be calling today the RPDC
(Rossi Probabilistic Discrete Cracking model), under
development at the LCPC since 1985, has been dealt with
in many publications throughout its development and evo-
lution (from the research tool to the present industrial
tool), among which may be mentioned [1-3]. We shall be
considering here only very briefly the physical assumptions
and their numerical translation which form the basis of the
probabilistic cracking model for concrete.
Concrete is regarded as a heterogeneous material
because of its composite nature, and physical phenomena
taking place during its setting and hardening. Its mechani-
Editorial Note
Dr. Pierre Rossi is a Senior Member and a member of the Board of Advisors. He is a member of RILEM TC 162-QFS.
This text is the English version of an article published in LPC’s Bulletin.
1359-5997/99 © RILEM
SCIENTIFIC REPORTS
R É S U M É
Après un rapide rappel concernant le modèle probabiliste de
fissuration discrète relatif aux bétons non fibrés, et une analyse
critique de son extension aux bétons fibrés, un nouveau modèle
probabiliste de fissuration discrète de Rossi appliqué aux bétons
fibrés est proposé. Dans ce modèle, la fissuration est modélisée au
travers d’éléments de contact qui auront un comportement élasto-
plastique parfait avec une rupture fragile. Le palier plastique
ainsi que l’énergie postfissuration y sont des variables aléatoires
non corrélées. Un travail numérique qui consiste à adapter les
éléments de contact existants au cas d’un comportement élasto-
plastique parfait avec rupture fragile, ainsi qu’un algorithme de
résolution adapté, restent maintenant à réaliser.

cal performance under traction may be regarded as brittle
elastic seen from a very small material description scale in
relation to the total volume considered. It may then be
characterized mechanically by its Young’s modulus and its
direct tensile strength. However, because of its heterogene-
ity, its properties are of a random nature and depend on the
scale of observation. On the scale of a structure, it is
assumed that deteriorations resulting from stresses produce
local tension, even in compressed zones [2]. This results in
the appearance of cracks ranging in size from a few microns
to a few tens of centimetres depending on the stresses gen-
erated by the loading conditions.
Numerical modelling of a structure by means of a
cracking model thus presupposes that the size of mesh
elements is small in relation to the gradients.
Within these elements, the behaviour of the concrete
is assumed to be elastic and isotropic. These elements are
linked together by special interface elements capable of
opening when they are subjected to sufficient tension or
shearing: a displacement discontinuity is thus created,
simulating a crack. These are elements of very small
thickness compared with the volume elements, which
open when the normal or tangential stress at their centre
125
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
Fig. 1 – Principle of probabilistic cracking model.
of gravity (middle Gauss point) reaches the local tensile
strength. The heterogeneity of the material is taken into
account numerically by a random distribution of the
Young’s moduli of the massive elements, and the tensile
and shearing strength of the contact elements (Fig. 1).
These properties follow distribution laws of the trun-
cated normal type which depend on the considered vol-
ume. An experimental study demonstrated that these
parameters depended essentially on the ratio of the vol-
umes of the largest grain of the concrete and of the spec-
imen, as well as on the compression strength determined
on the standard specimen [4]. Smoothing of experimen-
tal results has allowed the establishment of an analytical
expression of the variation of the average and standard
deviation of tensile strength as a function of the ratio of
the largest grain volume of the concrete to the total volume of the
two volume elements f lanking the contact elements, and
Young’s modulus as a function of the volume of the largest
grain of the concrete to the volume of the volume elements.
The probabilistic cracking model is thus based on the
hypothesis that the volume of a grid element corre-
sponds to the same volume of material.
3. PRESENT EXTENSION OF RPDC MODEL
TO FIBRE-REINFORCED CONCRETE
The modelling of steel-fibre-reinforced concrete
(SFRC) consists in introducing, after the opening of a
crack (i.e. of a contact element), a return force which is a
function of crack opening. From the numerical view-
point, this is achieved by imposing, at the Gauss points, a
normal stress which is a function of the normal displace-
ment of the open contact element [1]. The variation law
126
of the normal stress as a function of the normal displace-
ment (Fig. 2) is obtained by performing a uniaxial tensile
test on a notched specimen. The experimental proce-
dure relative to this test, the size of the specimens, their
number, how they are obtained, and the analysis of
results have already been published [5-7].
4. LIMITS AND IMPERFECTIONS OF THE
RPDC MODEL APPLIED TO STEEL-FIBRE-
REINFORCED CONCRETE
The limits or the imperfections of the RPDC model
applied to steel-fibre-reinforced concrete are such that it
does not have the same degrees of pertinence, perfor-
mance and credibility as in the case of non-fibre-rein-
forced concrete. For this reason, we chose to inventory
these limits in decreasing order of importance in our
opinion. It is evident that the inventory we propose
below is not exhaustive in any case:
– The model (and its associated numerical process-
ing) of the postcracking behaviour of SFRC is not objec-
tive (nor effective from the numerical viewpoint).
– The scale effects, or the scatter, relative to the post-
cracking behaviour of SFRC under tension (curve of
tensile stress-crack opening) are insufficiently taken into
account.
– The model of the reclosing of cracks is not very real-
istic from the physical viewpoint.
– Only the action of the fibre with regard to the nor-
mal displacements of the crack lips is taken into account
whereas, for certain types of loading, the action of the
fibre with respect to the tangential displacements of the
crack lips is not negligible (case of shearing force cracks,
for example).
– Modelling and numerical treatment of postcrack-
ing behaviour – We pointed out earlier that, after the
opening of a contact element (simulating the cracking of
the matrix), at the level of this contact element is imposed
a stress normal to this element which follows a mathemati-
cal relation (a polynomial relation, for example) deter-
mined from the curve presented in Fig. 2. From a numeri-
cal viewpoint, this is accomplished by imposing at the
Gauss point of the element a normal stress as a function of
the normal displacement, at this point, of the open contact
Fig. 2 – Example of law of evolution of normal stress as a func-
tion of the displacement normal to the contact element.
Rossi
element. During an iteration, the residue is the difference
between the tensile strength and the normal elastic stress
relative to a contact element. When the contact element
has just opened for the first time, the calculated elastic
stress is greater than or equal to the tensile strength
assigned to the element. In the RPDC model applied to
SFRC, the stress imposed as a function of the normal dis-
placement may be very low for a very high ft, or very high
for a very low ft. Consequently, when a large number of
contact elements open during an increment, and if there
are very significant residues, the convergence of the calcu-
lation may be very difficult or even impossible to obtain
owing to the oscillation phenomena. This is not related
only to the chosen resolution algorithm, which follows an
explicit pattern, but also to the very modelling of the post-
cracking behaviour which involves a one-to-one relation
between the normal stress and the normal displacement
without the evolution law proper.
– Scale effects in postcracking behaviour – We
pointed out that, after the opening of the contact element,
its normal stress follows a law which is a function of the
crack opening. In reality, i.e. during uniaxial tensile tests, a
certain scatter is observed in the postcracking behaviour,
this scatter increasing as the diameter of the specimen
decreases. On the other hand, no sufficiently solid experi-
mental information is available concerning average post-
cracking behaviour as a function of specimen diameter. As
concerns the RPDC model, the scatter of postcracking
behaviour is taken into account, roughly, using a distribu-
tion function of the postcracking law independent of the
volume of elements of the mesh. Thus, the average value
and the standard deviation are obtained by means of direct
tensile tests on a specimen having arbitrarily chosen
dimensions. There is thus a sort of contradiction or incon-
sistency inherent in the model. In fact, it is hardly relevant
to consider a random distribution of the tensile strengths of
the matrix as a function of the size of the elements, as the
random distribution of the postcracking law is indepen-
dent of this size. The gap constituted by the second point
offsets the advantage obtained.
– Modelling of crack reclosing – In the present
RPDC model, crack reclosing takes place without any
residual opening, i.e., for total reclosing, the normal
stress in the contact element becomes nil (Fig. 2). But,
the physical reality is quite different because, when load-
unload cycles are performed during a uniaxial tensile test
on SFRC, an increase in the residual opening of the
crack is observed during the cycles.
– Action of fibre with regard to model II displace-
ment of crack lips – As concerns certain types of load-
ing and structural geometry, mode II displacements of
crack lips (tangential displacements of crack lips) can play
an important part in the mechanisms causing damage to
the structure in question. The RPDC model currently
considers only the action of the fibre with respect to
cracks opening in mode I.
In light of all the present imperfections of the RPDC
model applied to SFRC, we propose a certain number of
modifications that could be made in the future.
127
5. TOWARDS A NEW RPDC MODEL APPLIED
TO SFRC
– Modelling of postcracking behaviour – It is first to
be noted that no systematic study of the scale effect on the
postcracking behaviour of SFRC under tension is available
in the literature. This important work remains to be car-
ried out in order to enable the RPDC model to be as
effective for fibre-reinforced concrete as for non-fibre-
reinforced concrete. To accomplish this, it would be nec-
essary to carry out, for different types of fibre (geometry
and dimensions) and for different types of industrial appli-
cations (the mechanical characterization of an SFRC
depends to a great extent on its industrial application, [6-
8], a large number of uniaxial tensile tests while varying the
compressive strength of the matrix (hence its compact-
ness), the percentage of fibre and, of course, the diameter
of the specimen. The aim of this work is to obtain, for a
given industrial application, postcracking distribution
functions as a function of the dimensions of the contact
element, the compressive strength of the matrix, and the
type and percentage of fibre. This is an essential phase if
the RPDC model is to be used industrially.
It is possible to consider the possible or probable
inf luence of specimen diameter on the postcracking
energy distribution function. For a given fibre (type and
dimensions), this distribution function will depend
mainly on the number of fibres in the cracked section,
and their orientation in relation to this section.
More precisely, the distribution function depends
mainly, or preferentially, on the number of fibres correctly
oriented, i.e. forming an angle as close as possible to 90°
with the cracked section. Thus, a small variation in the
number of well-oriented fibres will have a greater influ-
ence on scatter relative to postcracking energy than a great
variation in the number of poorly-oriented fibres.
From the theoretical viewpoint, as postcracking energy
is not related to the presence of a small mesh (Weibull the-
ory relative to the fracture of brittle materials) [9], but a
strong mesh (well-oriented fibres), the average postcrac-
king energy value should be influenced little by the area of
the cracked section, whereas the standard deviation on this
energy would be greatly influenced; in particular it should
increase as this area decreases.
Some experimental artefacts may occur in the determi-
nation of the postcracking energy distribution function
which, for SFRC and for a given industrial application, is
itself a function of the cracked section area. These artefacts
are related to the fact that, as mentioned earlier, the distrib-
ution functions relative to a given industrial application
must be obtained by means of uniaxial tensile tests on core
specimens (specimens taken within the considered struc-
ture). But, when a core sample is taken in a SFRC, it is
clear that the fibres located near the area of the core and
inclined in relation to the failure section of the specimen
are cut. Thus, the poorer the orientation the fibres near
the area, in the sense defined earlier, the greater the possi-
bility that they will be cut over a great length. This prob-
lem can have two consequences:
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
– The average postcracking behaviour obtained on the
core specimen may not be as good as the real behaviour in
the structure; the scatter on this behaviour should, for its
part, not be significantly affected because it is masked (cov-
ered) by that related to the variation in the number of well-
oriented fibres (see above).
– The artefact on the average postcracking behaviour
should increase as the diameter of the specimen decreases.
In fact, since the specific area of the specimen increases as
its diameter decreases, there is a greater possibility of
obtaining cut fibres as the diameter of the specimen
decreases.
Summarizing, taking into account what is stated
above, it is possible to assume that the distribution func-
tions of postcracking energy as a function of the cracked
section area will be correctly determined if the average
postcracking energy chosen is that determined with the
specimens having the largest diameter.
The new model to be developed should also take into
account, as done in the present model, the fact that,
physically, there is no correlation between the tensile
strength of SFRC and its postcracking behaviour. There
are two reasons for this:
– In a zone of SFRC in which there is a high fibre den-
sity and hence surely very good postcracking behaviour,
the matrix may be less compact than elsewhere and
hence stronger.
– In a SFRC zone in which the matrix is very compact,
hence very strong, fibre density may be very low and
thus the postcracking behaviour mediocre.
In addition, it is noted experimentally that the scatter
relative to the postcracking behaviour of a given SFRC
(and for a given industrial application) depends on the
non-negligible manner in which the crack opens [10, 11,
5] Consequently, in the proposed model, the postcrack-
ing energy distribution function must be considered to
be dependent on crack opening.
The latter point is very important in the orientation
that will be given to the model. In fact, we do not find it
very realistic, in the RPDC model, to consider a random
distribution of the postcracking energy that depends on
both the area of the contact elements and on the crack
opening, modelled by the normal displacement at the
level of these contact elements. The choice will thus be to fix
the random distribution of the postcracking energy. The numeri-
cal calculations to be carried out will then be considered to be
acceptable only for a given crack opening. Consequently, in the
approach that we propose, the necessary and sufficient
data relative to SFRC are respectively the average post-
cracking energy and the standard deviation relative to a
crack opening, and not the shape of the σN-w curve (σN:
normal postcracking stress, w: normal postcracking dis-
placement) at the level of the contact elements.
The question raised then is what shape should be
chosen for this curve?
To be consistent and homogeneous with the RPDC
model applied to non-SFRC, a perfectly plastic post-
cracking behaviour appears to us to be appropriate in the
scale of the contact elements. It is also proposed that the
stress level should be equal to the tensile strength ft if it is
128
Fig. 3 – Example of the proposed new law concerning the evolu-
tion of normal stress as a function of the displacement normal to
the contact element.
the tensile criterion which is reached, or the normal
stress reached at the instant of the shear cracking of the
matrix, if it is the shearing criterion which is activated.
It should be noted that in the model we propose there
is, quite naturally, additional data, in the form of the criti-
cal opening wc which is calculated in a usual manner
knowing the tensile strength of the matrix, and the post-
cracking energy relative to the contact element. When this
critical opening is reached, the contact element behaves
like an open contact element of non-SFRC.
In the preceding discussion, it is evident that we are
proposing a new RPDC model applied to SFRC in
which the cracking process is modelled by contact ele-
ments having a perfect elastic-plastic behaviour with
brittle fracture (Fig. 3).
Practically, the new approach we are proposing is
characterized by the following:
– For a given structure is chosen a crack opening scale
for which information is desired. This information may
be the orientation and spatial distribution of the cracks,
as well as the overall displacement and forces on the
structure. Thus, strictly, to have information on behav-
iour in service, and the ultimate behaviour of a given
structure, two types of calculations must be carried out
corresponding to two crack opening scales.
– Once this crack opening scale has been chosen, it is
necessary to calculate, from the experimental results
obtained in the uniaxial tensile tests, the average value
and the standard deviation of the energy under the
stress-crack opening curve, corresponding to this scale of
crack opening.
– Modelling of crack reclosure – In a conventional
perfect elastic-plastic model, crack closing results in a
residual opening equal to the plastic opening. This
reflects better the physical reality than does the present
model (mentioned earlier) even if this calculated residual
opening is then overestimated in relation to the experi-
ment. If precise experimental information is available
Rossi
Fig. 4 – Example of test that it would be of interest to develop in
order to access experimental shear stress/mode II crack opening
for SFRCs.
concerning, for a given SFRC, the evolution of the
residual crack opening as a function of the crack open-
ing, it would then be possible to refine the model by
proposing that Young’s modulus should be a function of
the plastic displacement wp. This type of elastic-plastic
model with plastic damage was already proposed in the
past by Ulm [12] as concerns continuous models. It must
be adapted for the case of contact elements.
– Modelling of mode II crack lip displacement –
Modelling with a perfect elastic-plastic model under brittle
fracture does not pose any additional theoretical problems
in relation to mode I modelling of crack opening. The cru-
cial problem lies in the lack of experimental data in the lit-
erature. As the inclusion of tangential displacements does
not appear to us to be a primary aspect of the model, and as
we wish to conserve the simplicity of this model, it appears
acceptable to us to uncouple mode I opening and mode II
displacements at the level of the contact elements. In other
words, this means that we are assuming, experimentally
and surely without any grounds, that crack opening within
a SFRC has no influence on the evolution of tangential
displacements of the lips of this crack.
If crack lip displacements take place in mode II, the
step stress will be equal to the shearing resistance τc of
the matrix if the shearing criterion is activated, or to the
shearing stress reached if the tensile fracture criterion is
activated.
A new test is to be developed to obtain the informa-
tion required for modelling mode II crack opening.
Given the immense experimental work to be completed
to obtain statistical data (mode II crack lip displacements
exhibit the same randomness in the model as the open-
ing of the cracks) on many SFRCs currently used, it is
indispensable that the test in question should be simple.
We shall consider that it would be of interest to investi-
gate in detail and to develop a test derived from the test
presented very schematically in Fig. 4.
129
Remark:
The fact that brittle fracture at the level of an element
appears in a random manner (random distribution of
critical tangential displacements or openings) leads to a
SFRC structural behaviour that can reduce hardness
although a perfect elastic-plastic behaviour is adopted at
the level of the contact elements.
In the following, we are indicating the basic relations
serving as the theoretical framework for the model [13].
Postcracking behaviour under uniaxial tension
∗ wp < wc
∂ψ ∂ψ
σN = − ζ=−
∂w p ∂χ
where ψ is the free energy of the elementary surface sys-
tem, ζ is the strain hardening force, and χ the strain
hardening variable.
– Evolution law
dw p = dλ
(
∂g σ N , ζ ) dχ = dλ
(
∂h σ N , ζ )
∂σ N ∂ζ
where dλ is the plastic multiplier, g = g(σN, ζ) the plastic
potential, and h = h(σN, ζ) the strain hardening potential.
∗ wp = wc
σN = 0
Postcracking behaviour under shear
We use the same relations as above, replacing σN by τ
which is the shearing stress, and wP by vP which is the
tangential displacement.
6. CONCLUSIONS
After the critical analysis of the present RPDC (Rossi
Probabilistic Discrete Cracking) model applied to
SFRCs, we propose the development of a new RPDC
model for SFRCs. In this model, cracking is modelled
through contact elements which will have a perfect elas-
tic-plastic behaviour with brittle fracture. The plastic
step as well as the postcracking energy are uncorrelated
random variables. Having set the foundations of the
model, numerical work still remains which consists in
adapting the existing contact elements to the case of per-
fect elastic-plastic behaviour with brittle fracture, as well
as work on a suitable resolution algorithm.
ACKNOWLEDGEMENTS
The author thanks Franz Ulm for the fruitful discus-
sions he was able to have with him and which were use-
ful in the preparation of the present article.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

REFERENCES [7] Rossi, P., ‘Les bétons de fibres métalliques’, (Presses de l’École
Nationale des Ponts et Chaussées editor, 1998, in french).
[8] Rossi, P. (1994) ‘Steel-fibre-reinforced concretes (SFRC): an
[1] Rossi, P. and Wu, X., ‘A probabilistic model for material behav-
exemple of French research’, ACI Material Journal 91 (3) (1994)
iour analysis and appraisement of the concrete structures’,
273-279.
Magazine of Concrete Research 44 (161) (1992) 271-280.
[9] Weibull, W., ‘A statistical theory of the strength of the materials’,
[2] Rossi, P., Ulm, F. and Hachi, F., ‘Compressive behaviour of con-
Ingenioessvetenskaps Akad 151 (1939) 1-55.
crete: physical mechanisms and modelling’, ASCE Engineering
[10] Harrouche, N., ‘Formulation et comportement à la fissuration
Mechanics 122 (11) (1996) 1038-1043.
des bétons de fibres métalliques’, Thèse de l’Université Paris VI
[3] Rossi, P. and Ulm, F., ‘Size effects in biaxial tensile-compressive
(1989) (in french).
behaviour of concrete: physical mechanisms and modelling’,
[11] Nogues, M., ‘Étude des propriétés mécaniques de différents
Mater. Struct. 30 (198) (1997) 210-216.
bétons de fibres’, Rapport du DEA ‘Mécanique des Solides’ de
[4] Rossi, P., Wu, X., Le Maou, F. and Belloc, A., ‘Scale effect on
Université Paris VI-ENPC (1992) (in french).
concrete in tension’, Ibid. 27 (1994) 437-444.
[12] Ulm, F., ‘Modélisation élastoplastique avec endommagement du
[5] Casanova, P., ‘Bétons renforcés de fibres métalliques : du
béton de structure. Application aux calculs statiques et dynamiques
matériau’, Thèse de l’École Nationale des Ponts et Chaussées
des structures en béton armé et béton précontraint’, Thèse de
(1995) (in french).
l’École Nationale des Ponts et Chaussées, (1994) (in french).
[6] AFREM-BFM, ‘Recommandations AFREM sur les méthodes de
[13] Coussy, O., ‘Mechanics of Porous Continua’, (John Willey edi-
dimensionnement, les essais de caractérisation, de convenance
tor, 1995).
etd de contrôle - Éléments de structure fonctionnant comme des
poutres’, sous la direction de P. Rossi (1995) (in french).

130
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 131-139
Rheological properties of SEBS, EVA and EBA polymer
modified bitumens
X. Lu, U. Isacsson and J. Ekblad
Division of Highway Engineering, Royal Institute of Technology, S - 100 44, Stockholm, Sweden
Paper received: February 2, 1998; Paper accepted: June 26, 1998
A B S T R A C T
The rheological properties of various polymer modi-
fied bitumens were studied. Three bitumens from two
different sources were mixed with styrene-ethylene-
butylene-styrene (SEBS), ethylene vinyl acetate (EVA)
and ethylene butyl acrylate (EBA) copolymers at differ-
ent polymer contents. The rheological properties of the
modified binders were investigated by means of dynamic
mechanical analysis and creep test (bending beam
rheometer). The results indicated that polymer modifi-
cation increased binder elastic responses and dynamic
moduli at intermediate and high temperatures, and
reduced binder complex and stiffness moduli at low
temperatures. Polymer modification also reduced tem-
perature susceptibility, glass transition temperature as
well as limiting stiffness temperature. The degree of the
improvement generally increased with polymer content,
but varied with bitumen source/grade and polymer type.
1. INTRODUCTION
Rheology is the science which studies the deforma-
tion and flow of materials, whether in liquid, melted, or
solid form, in terms of the materials’ elasticity and vis-
cosity [1]. Bitumen is a viscoelastic material and may
exhibit either elastic or viscous behaviour, or a combina-
tion of these, depending on temperature and time over
which the bitumen is observed. At sufficiently low tem-
peratures and/or high rates of strain (frequencies), bitu-
men behaves essentially as an elastic solid. As tempera-
ture increases and/or loading rate reduces, the viscous
property of bitumen becomes more obvious. At suffi-
ciently high temperatures and/or long loading time,
bitumen is essentially a Newtonian liquid, and can be
described by a shear rate independent viscosity value.
The rheological properties of bituminous binders may
significantly influence the performance of asphalt mix-
Editorial Note
Prof. Ulf Isacsson is a RILEM Senior Member, he works at the Royal Institute of Technology, a RILEM Titular Member.
1359-5997/99 © RILEM
SCIENTIFIC REPORTS
R É S U M É
Les propriétés rhéologiques de divers bitumes modifiés
ont été étudiées. Trois bitumes de deux origines différentes
ont été mélangés avec les copolymères styrène-éthylène-
butylène-styrène (SEBS), éthylène-vinyl-acétate (EVA) et
éthylène-butyl-acrylate (EBA), en utilisant différents
pourcentages de polymère. Les propriétés rhéologiques des
différents liants modifiés ont été examinées au moyen
d’essais dynamiques et d’essais de fluage (bending beam
rheometer). Les résultats montrent que l’ajout de polymère
augmente la réponse élastique du liant et le module dyna-
mique aux températures moyennes et hautes, et diminue le
module complexe et le module de rigidité à basse tempéra-
ture. L’ajout de polymère diminue la susceptibilité ther-
mique, la température de transition vitreuse et la tempéra-
ture de « rigidité limite ». Le degré d’amélioration augmente
généralement avec le pourcentage de polymère, mais il varie
avec l’origine et la nature du bitume, et le type de polymère.
ture during mixing, compaction and in service. For this
reason, various types of rheological testing are often
included in binder specifications.
The rheological properties of bitumens are determined
by molecular interactions (molecular forces), which in turn
depend on chemical compositions [2-6]. In principle,
desired rheological properties of bitumen may be achieved
by changing chemical composition using an additive or
chemical reaction modification [7]. The addition of poly-
mers is a versatile way of modifying the rheology of bitu-
mens [8-11]. Many types of polymers are available; how-
ever, there is no universal polymer and selection should be
made according to the specific needs.
This paper presents a rheological investigation of var-
ious polymer modif ied bitumens by using dynamic
mechanical analysis and creep test (bending beam
rheometer). The effects of base bitumen and polymer
type and content on rheological properties are studied.
131
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
2. MATERIALS AND TEST PROCEDURES
2.1 Materials
Three bitumens (one B85 and two B180) from two
sources (Venezuela and Mexico) were used in this study.
They are denoted A, B and C, respectively. The physical
properties and chemical characteristics of the base bitu-
mens have been reported elsewhere [12].
The polymers investigated are styrene-ethylene-
butylene-styrene (SEBS), ethylene vinyl acetate (EVA)
and ethylene butyl acrylate (EBA) copolymers, which are
representative of synthetic elastomers (SEBS) and ther-
moplastics (EVA, EBA). The SEBS polymer contains
29% by mass styrene. The two EVA copolymers used are
coded as EVA1 and EVA2. Melt indices (MI) of EVA1
and EVA2 are 6 and 158, and their vinyl acetate (VA)
contents are 28 and 18% by mass, respectively.
The polymer modified bitumens were prepared using a
low shear mixer at 180°C and a speed of 125 rpm. The
mixing time was two hours. The process for preparing the
modified binders exposed the bitumen to high tempera-
ture and air for an extended time, which led to hardening
of the bitumen. For the accurate evaluation of polymer
effects, the base bitumens were also subjected to the same
treatment as the polymer-bitumen blends.
2.2 Test methods
Dynamic mechanical analysis (DMA): DMA with fre-
quency sweeps (from 0.1 to 100 rad/s) at 25 and 60°C and
temperature sweeps (from -30 to 135°C) at 1 rad/s were
performed using a rheometer. Parallel plates, diameter 8
and 25 mm, were used in a temperature range of -30 to
50°C and 40 to 135°C, respectively. The test procedure
has been described previously [13].
Low-temperature creep test: Creep tests were carried out
at four different temperatures (-35, -25, -15 and -10°C)
using a bending beam rheometer. The rheometer was
developed in the SHRP (Strategic Highway Research
Program) [14]. In tests, the binder was heated to fluid con-
dition, poured into the mold and then allowed to cool at
room temperature for about 90 minutes. The sample was
then cooled to approximately -5°C for 1 minute and
demolded. After demolding, the sample beam (125 mm
long, 12.5 mm wide and 6.25 mm thick) was submerged
in a constant-temperature bath and kept at each test tem-
perature (starting at -35°C) for 30 minutes. A constant
load of 100 g was then applied to the rectangular beam of
the binder, which was supported at both ends by stainless
steel half-rounds (102 mm apart), and the deflection of
center point was measured continuously. Creep stiffness
(S) and creep rate (m) of the binders were determined at
several loading times ranging from 8 to 240 seconds.
132
3. RESULTS AND DISCUSSION
3.1 Viscoelastic properties
In DMA, a sinusoidally var ying shear strain is
imposed on the sample and the amplitude of the result-
ing shear stress is determined. The ratio of the peak
stress to the peak strain is defined as the complex modu-
lus (G∗), which is a measure of the overall resistance to
deformation of a material. The in-phase and out-of-
phase components of G∗ are defined as the storage mod-
ulus (G′) and the loss modulus (G″), respectively. G′ is
proportional to the stress in-phase with the strain and
provides information on the elastic responses of a mater-
ial, while G″ is proportional to the stress out-of-phase
with the strain and is associated with viscous effects. The
phase difference between the stress and strain in an oscil-
latory deformation is defined as phase angle (δ). This
parameter is a measure of the viscoelastic character of the
material. A purely viscous liquid and an ideal elastic solid
demonstrate δ of 90° and 0°, respectively. The viscoelas-
tic parameters of bitumens are functions of temperature
and frequency, which may be modified by the addition
of polymers.
In Figs. 1 and 2, the temperature and frequency
dependences of G′ and δ are illustrated. As can be seen,
polymer modif ication increases bitumen elastic
responses at intermediate (0 - 40°C) and high tempera-
tures (≥ 40°C), as evidenced by increased G ′ and
decreased δ. The temperature susceptibility (Fig. 1) and
frequency susceptibility (Fig. 2) of G′ are also reduced
by polymer modification.The improvement increases
with polymer content, but varies widely with the type of
polymers and the base bitumens. For a given polymer
content, the increase in the elastic responses is observed
to be highest for the modified bitumens with SEBS,
while it is the lowest for those with EVA, as indicated in
Figs. 1 and 2, as well as in Table 1. The increased elastic
responses are mainly due to polymer networks (cross-
links and entanglements) and/or polymer crystallinity.
When sufficient polymer networks are formed in the
modified binders, e.g. in a continuous polymer phase,
four regions, glassy, transition, plateau, and terminal or
flow, are observed in G′, and correspondingly, a maxi-
mum and a minimum are observed in δ (Fig. 1).
The four regions can be associated qualitatively with
different kinds of molecular responses. In the glassy
region, the binder possesses the highest modulus and is
hard and brittle, with little or no intermolecular move-
ment possible. In the transition region, the binder is
resilient, and there is sufficient thermal energy to allow
for localized molecular rotation and slipping. In this
region, the modulus falls off rapidly as secondary inter-
molecular forces are overcome. In the plateau region, the
modulus changes little with temperature, and there are
still significant molecular entanglements, while local sec-
tions are very mobile. In the f low region, molecular
interactions have less energy than the thermal energy
applied and significant slippage occurs.
Lu, Isacsson, Ekblad

Fig. 1 – Storage modulus and phase angle at 1 rad/s as a function of temperature.

Fig. 2 – Storage modulus and phase angle at 60°C as a function of frequency.

133
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Table 1 – Storage modulus and phase angle at 60°C and different frequencies Different degrees of prominence may
be observed in the four regions, depending
Binder Storage Modulus G′ @ 60°C, Pa Phase angle δ @ 60°C, Deg
on the chemical and physical nature (e.g.
1 rad/s 10 rad/s 100 rad/s 1 rad/s 10 rad/s 100 rad/s molecular weight, microstructure and net-
Bitumen A 4 128 1970 88.8 86.2 83.1 work density) of the binders. The density
Bitumen A + 6% SEBS 1420 5420 19100 53.2 52.6 57.1 of polymer networks is associated with sev-
Bitumen A + 6% EVA1 46 833 6640 82.8 72.3 63.5 eral factors such as the chemical composi-
Bitumen A + 6% EVA2 115 666 4400 70.1 71.5 70.0 tions of the bitumen and the characteristics
Bitumen A + 6% EBA 628 2340 1050 53.9 60.3 62.5 and content of the polymer. It is also influ-
Bitumen B 1 24 488 89.4 88.0 85.5 enced by the temperature. For EVA and
Bitumen B + 3% SEBS 5 137 2290 87.9 83.4 75.1 EBA modif ied binders, the crystalline
Bitumen B + 6% SEBS 454 1780 8370 55.2 60.9 61.4 portions (packed polyethylene segments) of
Bitumen B + 9% SEBS 2090 5840 19700 41.6 48.2 51.4 the polymers may melt and the polymer
Bitumen B + 3% EVA1 4 71 993 86.6 83.7 78.8 networks weaken at temperatures higher
Bitumen B + 6% EVA1 13 305 3020 85.2 76.1 66.7 than about 60°C (EVA) or 80°C (EBA),
Bitumen B + 9% EVA1 73 952 6900 79.2 69.2 58.9 leading to a pronounced decrease in G ′
Bitumen B + 3% EVA2 8 64 758 83.4 83.9 81.3 and a sharp increase in δ (Fig. 1). In the
Bitumen B + 6% EVA2 14 164 1680 82.6 79.3 74.5 case of SEBS modified binders, the dra-
Bitumen B + 9% EVA2 37 375 3160 79.4 75.1 69.7 matic changes in G ′ (decrease) and δ
Bitumen B + 3% EBA 37 209 1470 72.9 76.2 76.1 (increase) are observed at temperatures
Bitumen B + 6% EBA 240 916 4360 55.9 62.8 65.1 higher than about 100°C, where the physi-
Bitumen B + 9% EBA 841 2570 9430 44.9 51.6 54.2 cal cross-links of polystyrene diminish
Bitumen C 1 17 344 89.4 88.5 86.7 rapidly.
Bitumen C + 6% SEBS 673 1900 7060 43.1 55.0 62.1 Polymer modification of bitumen rhe-
Bitumen C + 6% EVA1 18 286 2910 83.4 77.1 68.9 ology is also identified in Fig. 3, where
Bitumen C + 6% EVA2 126 526 2830 61.6 68.8 71.2 BLACK diagrams (G∗ as a function of δ)
Bitumen C + 6% EBA 447 1290 4940 44.9 56.8 64.2 are illustrated. These diagrams are gener-

Fig. 3 – Complex modulus as a function of phase angle.

134
Lu, Isacsson, Ekblad
Fig. 4 – G∗/sinδ as a function of frequency (at 60°C) and temperature (at 1 rad/s).
ated with temperature sweeps (from -30 to 135°C) at
1 rad/s. As can be seen, for the base bitumens and modi-
fied binders containing 3% by mass of polymer, G∗
decreases smoothly as the increasing temperature changes
the behaviour of the binder from essentially solid elastic
(δ → 0°) to completely liquid viscous (δ = 90°). A differ-
ence between the base bitumens and the modified binders
with 3% by mass of polymer is observed at δ higher than
about 60°, where G∗ is mainly contributed by its viscous
part G″. This indicates that, at a low polymer content (3%
by mass in this study), the nature of the polymers shows
only a secondary effect on the rheological properties of
the modified binders, and the behaviour of the modified
binders remains close to that of the base bitumens.
However, after modification with a sufficiently high poly-
mer content (≥ 6% by mass), the binders change funda-
mentally in their rheological behaviour. As indicated in
Fig. 3, for all the modified binders containing 6 or 9%
by mass of polymer, a substantial decrease in δ (a sub-
135
stantial increase in the elastic response) is
observed as G∗ decreases with increasing
temperature. In these cases, the rheologi-
cal response of the modified binders is
mainly imposed by the viscoelastic prop-
erties of the polymers, those of the bitu-
mens waning considerably.
These observations imply that the
rheological properties of polymer modi-
fied binders are probably governed by
their continuous phase. Fluorescence
photomicrographs indicated that the
modif ied binders with a low polymer
content displayed a continuous bitumen
phase, while those containing a high
polymer content (≥ 6% by mass) showed
a continuous polymer phase [15]. The
rheological changes are also influenced by
characteristics of the bitumens and the
polymers. For the modified binders with
SEBS, G∗ deviates from that of the base
bitumens at δ as low as 10° (correspond-
ing to binder glassy state at low tempera-
tures). However, the curves of the modi-
f ied binders with EVA and EBA are
almost the same as those of the base bitu-
mens, even at the point where δ = 50°
and G∗ = 10 MPa. These differences sug-
gest that SEBS may improve bitumen
rheology over a wide temperature range,
while EVA and EBA show their effect
mainly at high temperatures.
Due to the increased elastic compo-
nent of the modified binders, the SHRP
r utting parameter, G∗/sinδ [16], is
improved. However, care should be taken
in using this parameter for evaluating the
contribution of a modified binder to rut-
ting performance. The reason is that
G∗/sinδ of a polymer modified binder is
more dependent on testing conditions than a conven-
tional bitumen, and for different modified binders, the
plots of G∗/sinδ as a function of frequency or tempera-
ture may cross. This phenomenon is shown in Fig. 4.
By extrapolating the value of G∗ for δ = 0 in Fig. 3,
the glassy modulus Gg∗ of the binders may be obtained.
As can be seen, the values of Gg∗ are close to 1 GPa and
are insignificantly influenced by polymer modification.
The 1 GPa reflects the rigidity of the carbon hydrogen
bonds as the bitumens reach their minimum thermody-
namic equilibrium volume [17]. However, the influence
of polymer modification on binder low-temperature
properties may be demonstrated by DMA glass transi-
tion temperature (Tg), which is defined as the tempera-
ture at peak G″. As shown in Table 2, the addition of a
high content (≥ 6% by mass) SEBS results in a signifi-
cant reduction (up to 14°C) in Tg of the binders. In con-
trast, relatively small changes (a few °C) in Tg are caused
by EVA and EBA modification. Table 2 also indicates
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Table 2 – Glass transition temperature and corresponding dynamic that, at Tg, the rheological parameters (e.g. G∗
moduli and phase angle obtained using DMA at 1 rad/s and δ) vary considerably. However, the differ-
Binder Tg from Peak G″ peak G′ at Tg, G* at Tg, δ at Tg,
ences are smaller compared with those at high
G″, °C value, MPa MPa MPa Deg temperatures. This is probably because, in the
glassy state, only very short segments of the
Bitumen A -16.3 70.6 205 217 19.0 molecules are invloved in molecular motion,
Bitumen A + 6% SEBS -30.6 72.0 321 329 12.6 while the motion of large segments is frozen-
Bitumen A + 6% EVA1 -19.6 85.8 247 262 19.2 in, leading to inhibition of polymer effects.
Bitumen A + 6% EVA2 -19.7 74.8 253 264 16.4
Bitumen A + 6% EBA -19.6 71.0 219 230 18.0
Bitumen B -22.1 75.8 252 264 16.7 3.2 Low-temperature creep responses
Bitumen B + 3% SEBS -20.5 58.6 179 208 16.0
Bitumen B + 6% SEBS -32.7 84.8 311 323 15.2 For a flexible pavement, one of the failure
Bitumen B + 9% SEBS -36.6 86.1 361 371 13.4 modes is low-temperature cracking. Low-
Bitumen B + 3% EVA1 -21.7 73.2 224 235 18.1 temperature cracking is caused by thermal
Bitumen B + 6% EVA1 -23.6 75.6 251 262 16.8 and/or load-induced tensile stresses when
Bitumen B + 9% EVA1 -23.5 74.0 224 236 18.3 these exceed the tensile strength of the pave-
Bitumen B + 3% EVA2 -21.5 72.9 231 242 17.6 ment material. To reduce the risk of cracking,
Bitumen B + 6% EVA2 -23.6 81.0 270 282 16.7 the binder should have a low stiffness and a
Bitumen B + 9% EVA2 -24.5 73.1 256 267 16.5 high ability of stress relaxation at the lowest
Bitumen B + 3% EBA -21.6 73.9 227 238 18.1 pavement temperature [17]. In the SHRP
Bitumen B + 6% EBA -23.5 75.7 245 256 17.2 binder specification, a limited creep stiffness
Bitumen B + 9% EBA -25.5 72.2 252 262 16.0 (S) and logarithmic creep rate (m-value, which
Bitumen C -24.0 67.0 255 264 14.7 is related to binder stress relaxation ability)
Bitumen C + 6% SEBS -30.7 60.0 286 292 11.9 have been used as performance-based criteria
Bitumen C + 6% EVA1 -23.6 55.5 199 207 15.6 at low temperatures [18]. The measurements
Bitumen C + 6% EVA2 -23.6 61.3 223 231 15.4 of low-temperature creep responses are con-
Bitumen C + 6% EBA -25.4 57.5 229 236 14.1 ducted at temperatures 10°C above the mini-
mum pavement design temperature.
In this study, the SHRP bending beam rheometer
was employed to determine binder low-temperature
creep responses at six different loading times (8, 15, 30,
60, 120 and 240 sec) and four different temperatures
(-35, -25, -15 and -10°C). Typical examples of creep
response are shown in Fig. 5, in which the influence of
polymer modification is clearly illustrated. In Table 3,
the creep stiffness and m-value obtained at a loading
time of 60 sec and four different temperatures are pre-
sented. As can be seen, the polymer modified binders
generally display a lower creep stiffness than the corre-
sponding base bitumen, especially at temperatures lower
than -15°C. The improvement increases with polymer
content, but varies with the base bitumen and polymer
type. At -35°C and for a given polymer content, the
modif ied binders containing SEBS show a higher
reduction in creep stiffness than those with EVA and
EBA. However, at the other three low temperatures,
varying changes are observed in creep stiffness. Some
modified binders (e.g. Bitumens A and B modified with
SEBS) even show an increased creep stiffness at -10 and
-15°C. Moreover, the use of a softer bitumen seems
favourable for obtaining higher improvement in creep
stiffness (cf. B85 and B180). As regards effect of the
polymer modification on binder stress relaxation ability,
both increased and reduced m-values are observed,
depending on the testing temperature and loading time.

Fig. 5 – Creep response of the base and

SEBS modified bitumen B at -25°C.

136
Lu, Isacsson, Ekblad

Table 3 – Creep stiffness and m-value of the binders at a loading time Consequently, no definite conclusions
of 60 s and different temperatures could be drawn on this matter.
Binder Creep Stiffness [MPa] m-value
As illustrated in Fig. 6, creep stiff-
ness statistically (risk level 5%) corre-
-10°C -15°C -25°C -35°C -10°C -15°C -25°C -35°C lates with complex modulus. The lin-
Bitumen A 46 185 789 1590 0.58 0.45 0.26 0.09 ear relationship becomes better as
Bitumen A + 6% SEBS 80 195 684 1140 0.45 0.38 0.20 0.11 loading conditions are similar for creep
Bitumen A + 6% EVA1 46 132 624 1560 0.54 0.47 0.30 0.13 test and dynamic mechanical analysis
Bitumen A + 6% EVA2 72 180 790 2000 0.47 0.42 0.25 0.12 (frequency 1 rad/s corresponds to a
Bitumen A + 6% EBA 66 161 656 1620 0.45 0.41 0.26 0.13 loading time of approximate 6.3 s).
Bitumen B 12 54 487 1560 1.00 0.61 0.35 0.15 To avoid low temperature cracking,
Bitumen B + 3% SEBS 17 63 423 1180 0.72 0.54 0.33 0.15 the SHRP binder specif ication [18]
Bitumen B + 6% SEBS 27 80 329 774 0.52 0.41 0.28 0.19 requires that creep stiffness must not
Bitumen B + 9% SEBS 27 60 230 575 0.45 0.38 0.25 0.21 exceed 300 MPa at a loading time of 60
Bitumen B + 3% EVA1 12 40 350 1360 0.97 0.64 0.41 0.17 seconds. A resulting parameter for the
Bitumen B + 6% EVA1 12 36 326 1200 0.77 0.61 0.42 0.19 evaluation of low temperature properties
Bitumen B + 9% EVA1 13 31 302 1210 0.67 0.58 0.46 0.20 is the temperature at 300 MPa creep
Bitumen B + 3% EVA2 13 50 371 1190 0.86 0.56 0.40 0.16 stiffness, or the limiting stiffness temper-
Bitumen B + 6% EVA2 14 48 366 1240 0.67 0.54 0.38 0.18 ature. As indicated in Fig. 7, polymer
Bitumen B + 9% EVA2 10 27 265 1140 0.70 0.59 0.42 0.19 modification reduces the limiting stiff-
Bitumen B + 3% EBA 13 45 383 1250 0.82 0.58 0.37 0.18 ness temperature. The modification is
Bitumen B + 6% EBA 13 38 298 1160 0.59 0.52 0.38 0.19 influenced by polymer type/content and
Bitumen B + 9% EBA 10 27 234 1020 0.58 0.54 0.42 0.21 bitumen source/grade. This temperature
Bitumen C 33 99 437 931 0.59 0.49 0.32 0.19 may statistically (risk level 5%) correlate
Bitumen C + 6% SEBS 43 97 333 765 0.52 0.42 0.28 0.18 with DMA glass transition temperature
Bitumen C + 6% EVA1 38 89 357 958 0.51 0.47 0.31 0.19 (R = 0.50). The low correlation coeffi-
Bitumen C + 6% EVA2 37 94 411 1040 0.51 0.44 0.32 0.19 cient is due to the four modified binders
Bitumen C + 6% EBA 24 60 167 875 0.50 0.35 0.16 0.20 containing a high content (6 and 9% by
mass) of SEBS, which deviate from the
others in a significant way. The correlation becomes
much better (R = 0.91) if those modified binders are
excluded from the statistical analysis, as illustrated in Fig.
8. This indicates that, for some polymers at sufficiently
high contents, the rheological properties of the modified
binders measured is strongly dependent on the type of
test procedure.

4. CONCLUSIONS
The rheological properties of various polymer modi-
f ied binders have been investigated using dynamic
mechanical analysis and creep test (bending beam
rheometer). On the basis of the data and interpretations
presented in this study, the following conclusions may be
drawn.
Polymer modification increases the elastic response
and dynamic modulus of bitumens and reduces their
temperature susceptibility at intermediate (0 - 40°C)
and high (≥ 40°C) temperatures. The degree of the
modification generally increases with polymer content.
Significant improvement is observed as the content of
polymer is sufficiently high (6% by mass in this study).
Other factors affecting the rheological behaviour of the
binder are the characteristics of the base bitumen and
the polymer. For a given base bitumen, the effectiveness
of the four polymers tested may be ranked as SEBS,

Fig. 6 – Relationship between creep

stiffness and complex modulus.

137
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

EBA, EVA2 and EVA1, SEBS being the most effective.

The improved dynamic rheological properties should be
favourable for resistance to rutting at high temperatures.
Polymer modification also improves the low-tempera-
ture properties of bitumens, as indicated by decreases in
creep stiffness, glass transition and limiting stiffness tem-
peratures. The relative improvement varies with the base
bitumen, polymer type and polymer content. Of the four
polymers studied, the SEBS polymer seems the most
effective in improving bitumen low-temperature parame-
ters (except for the limiting stiffness temperature), similar
to its greatest inf luence on bitumen high-temperature
rheology. Unlike the situation at high temperatures,
insignificant differences in the low-temperature parame-
ters are observed between the EVA and EBA modified
binders. For a given polymer, the B180 bitumens may
result in a relatively higher reduction in creep stiffness
than the B85 bitumen. The influences of polymer modi-
fication may also vary with testing conditions (tempera-
ture and loading time). The improved low temperature
parameters should be favorable with respect to asphalt
cracking performance in practice.

ACKNOWLEDGEMENTS
The f inancial support provided by the Swedish
Transport and Communications Research Board (KFB)
is gratefully acknowledged. The authors also wish to
express their appreciation to Britt Wideman and Clarissa
Villalobos, who provided laboratory assistance during
the testing program.
Fig. 7 – Influence of bitumen source/grade and polymer
type/content on the temperature at 300 MPa creep stiffness.

REFERENCES
[1] Barnes, H. A., Hutton, J. F. and Walters, K., ‘An Introduction to
Rheology’ (Elsevier Science Publishers B. V., 1989).
[2] Dukatz, Jr., E. L., Anderson, D. A. and Rosenberger, J. L.,
‘Relationship between asphalt flow properties and asphalt com-
position’, Proceedings of the Association of Asphalt Paving
Technologists, 53 (1984) 160-185.
[3] Petersen, J. C., Robertson, R. E., Branthaver, J. F., Harnsberger,
P. M., Duvall, J. J., Kim, S. S., Anderson, D. A., Christiansen,
D. W. and Bahia, H. U., ‘Binder Characterization and
Evaluation’, Vol. 1, SHRP-A-367, National Research Council,
Washington, D. C., 1994.
[4] Corbett, L. W., ‘Relationship between composition and physical
properties of asphalt’, Proceedings of the Association of Asphalt Paving
Technologists, 39 (1970) 481.
[5] Thenoux, G., Bell, C. A. and Wilson, J. E., ‘Evaluation of physi-
cal and fractional properties of asphalt and their interrelationship’,
TRR 1171, TRB, National Research Council, Washington, D.
C., 1988, 82-97.
[6] Halstead, W. J., ‘Relation of asphalt chemistry to physical proper-
Fig. 8 – Temperature at 300 MPa creep stiffness as a function of ties and specifications’, Proceedings of the Association of Asphalt
DMA glass transition temperature. Paving Technologists, 54 (1985) 91-117.

138
Lu, Isacsson, Ekblad
[7] Isacsson, U. and Lu, X., ‘Testing and appraisal of polymer modi-
fied road bitumens - State-of-the-art’, Mater. Struct. 28 (1995)
139-159.
[8] Brûlé, B., Ramond, G. and Such, C., ‘Relationships between
composition, structure, and properties of road asphalts: State of
research at the French Public Works Central Laboratory’, TRR
1096, TRB, National Research Council, Washington, D. C.
(1986) 22-34.
[9] Collins, J. H., Bouldin, M. G., Gelles, R. and Berker, A.,
‘Improved performance of paving asphalts by polymer modifica-
tion’, Proceedings of the Association of Asphalt Paving Technologists 60
(1991) 43-79.
[10] King, G. N., King, H. W., Harders, O., Arand, W. and Planche,
P.-P., ‘Influence of asphalt grade and polymer concentration on
the low temperature performance of polymer modified asphalt’,
Ibid. 62 (1993) 1-22.
[11] Valkering, C. P., Vonk, W. C. and Whiteoak, C. D., ‘Improved
asphalt properties using SBS modified modified bitumens’, Shell
Bitumen Review 66 (May 1992) 9-11.
[12] Lu, X. and Isacsson, U., ‘Compatibility and storage stability of
styrene-butadiene-styrene copolymer modified bitumens’, Mater.
Struct. 30 (1997) 618-626.
[13] Lu, X. and Isacsson, U., ‘Rheological characterization of
styrene-butadiene-styrene copolymer modified bitumens’,
139
Construction and Building Materials 11 (1997) 23-32.
[14] Bahia, H. U., Anderson, D. A. and Christensen, D. W., ‘The
bending beam rheometer: A simple device for measuring low-
temperature rheology of asphalt binders’, Proceedings of the
Association of Asphalt Paving Technologists 61 (1992) 117-153.
[15] Isacsson, U., Lu, X. and Ekblad, J., ‘Characterization of Polymer
Modified Bitumens Containing SEBS, EVA and EBA’, Journal of
Materials Science, in press.
[16] Petersen, J. C., Robertson, R. E., Branthaver, J. F.,
Harnsberger, P. M., Duvall, J. J., Kim, S. S., Anderson, D. A.,
Christiansen, D. W., Bahia, H. U., Dongre, R., Antle, C. E.,
Sharma, M. G., Button, J. W. and Glover, C. J., ‘Binder
Characterization and Evaluation, Vol. 4: Test Methods’, SHRP-
A-370, National Research Council, Washington, D. C., 1994.
[17] Bahia, H. U. and Anderson, D. A., ‘The new proposed rheolog-
ical properties of asphalt binders: why are they required and how
do they compare to conventional properties’, in ‘Physical
Properties of Asphalt Cement Binders’, ASTM STP 1241, John
C. Hardin, Ed., (American Society for Testing and Materials,
Philadelphia, 1995) 1-27.
[18] Anderson, D. A. and Kennedy, T. W., ‘Development of SHRP
binder specification’, Proceedings of the Association of Asphalt Paving
Technologists 62 (1993) 481-507.
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 140-143
Assessment of glass fibre reinforced cement by in-situ
TECHNICAL REPORTS
SEM bending test
P. Trtik and P. J. M. Bartos
Advanced Concrete and Masonry Centre, University of Paisley, High St, Paisley, PA1 2BE, UK
Paper received: June 16, 1998; Paper accepted: July 17, 1998
A B S T R A C T
The purpose, development and the first results of a
new test method, an SEM in-situ 3-point flexural test are
presented. A large SEM chamber housing a eucentric ten-
sile testing rig has been redesigned and fitted with a new
apparatus enabling microfracture processes in bending to
be simultaneously scanned and the corresponding
load/deflection and other data and parameters accurately
recorded. The new test facility provides an opportunity to
obtain new quantitative and qualitative data considerably
more relevant to the fracture processes concerned than
those obtained when previous SEM in-situ tests are used. It
is now possible to examine specimens of a more realistic
size and including different types of reinforcement in
bending, a very common and practical mode of loading.
The new test method has been initially applied to an inves-
tigation of fracture mechanism in glass fibre reinforced
cement (GRC) where the results have shown major differ-
ences between flexural performance of aged and unaged
GRC. The in-situ SEM flexural test offers an effective new
tool for the investigation of the complex micro-fracture
mechanics of fibre reinforced composites as represented by
cement reinforced with bundles of glass fibres.
1. INTRODUCTION
As shown in the previous studies carried out at
University of Paisley e.g. [1, 2], a new mode of failure of
bundled reinforcement in cementitious composites,
specifically in GRC, had been discovered. The “tele-
scopic” mode of failure of bundled reinforcement
proved to be highly desirable for practical GRC and
opened new avenues for development of composites
reinforced with bundles of fibres. Within the framework
of this investigation, a new test method, an in-situ 3-
point bending test, was developed, which provided a
useful tool for assessment of micromechanisms of frac-
ture in cementitious composites.
Editorial Note
Prof. P. J. M. Bartos is a Senior Member and the Chairman of the RILEM 145-WSM: Workability of special concrete mixes. He is also a member of
RILEM TC 162-TDF: Test and design methods for steel fibre reinforced concrete and RILEM TC 174-SCC: Self-compacting concrete.
1359-5997/99 © RILEM
R É S U M É
Cette étude présente le développement et les premiers
résultats d’une nouvelle méthode d’essai par un microscope
électronique au scanner utilisant 3 points de flexion. L’équi-
pement de la chambre large d’essai de traction par flexion a
été modifié et ajusté avec un nouvel appareil favorisant le
processus de la microrupture. Ce dernier permet simultané-
ment de scanner et d’enregistrer la charge-déformation ainsi
que d’autres paramètres avec une bonne précision. L’équipe-
ment du nouvel essai permet d’obtenir d’avantages de nou-
velles données relatives au processus de rupture qu’avec
l’ancien microscope électronique au scanner. Il est mainte-
nant possible d’examiner les échantillons ayant des dimen-
sions réalistes et incluant différents types de renforcement et
de modes communs et pratiques de chargement. La nouvelle
méthode d’essai a été utilisée dans l’étude du mécanisme de
la rupture des composites ciment-fibres de verre (CCV) où
les résultats ont montré des différences majeures entre les per-
formances de la flexion des échantillons en vieillissement
normal et accéléré. L’essai de flexion in-situ du microscope
électronique avec scanner offre un outil efficace pour l’étude
de la mécanique de la microrupture des composites, tels les
composites ciment-fibres de verre.

In-situ SEM tests have been used for obtaining infor-
mation about micromechanisms of fracture in cement
based composites since the early 1980’s. Major pioneer-
ing work was carried out by Mindess & Diamond and
Lovell [3-5], who introduced both wedge opening load-
ing and compression test methods. It is also necessary to
mention the work of Tait and Garrett, who developed
and utilised double torsion method [6]. The micro-
mechanisms of bundled reinforcement in cementitious
composites observed via the in-situ SEM test methods
were described by Bentur & Diamond [7]. At the
University of Paisley the in-situ SEM tests have been
widely used by Swift [8], Duris [9] and Zhu [2], in all
cases in the form of either tensile or fibre pull-out tests.
140
Trtik, Bartos
The advantages and disadvantages of the test methods
mentioned above and their comparisons with the test
method presented are discussed later in this paper.
2. EXPERIMENTAL
2.1. Test arrangement
The tests were carried out inside the special chamber
previously developed at the University of Paisley. This
chamber replaces the ordinary specimen chamber on the
CamScan SEM. The chamber is fitted with load cell of
maximum load capacity of 500 N. The test stage allows
the eucentric movements of grips with a controlled veloc-
ity from 2 µm/sec to 20 µm/sec. A newly designed rig
enabled the authors to perform 3-point bending test. A
new data capture system was fitted to the chamber provid-
ing output in a digital form. During the testing, the value
of the instantaneous load was captured in intervals of one
second. Typical set-up of the testing chamber prepared for
the bending test is shown on Figs 1 and 2.
2.2 Specimen production
Specimens which have been used within the project
were prisms 110 × 20 × 9 mm in size containing one layer
of parallel glass fibre strands. The fibre reinforcement was
positioned 3 mm above the bottom edge of the specimens.
Twelve strands were spaced equally across the 20 mm
width of the prism in such a manner that the inner strands
(10 in number) were fully embedded in the matrix whilst
two outer strands were placed on the side of the prism thus
being partially embedded only. This particular arrange-
ment of strands was chosen to suit the test method. The
partially-embedded strands were visible on the surface of
the specimen which allowed the authors to observe the
bridging effect of fibre bundle during the crack propaga-
tion. Alkali resistant NEG-AR fibres were used in this
project. The overall fibre content achieved was about 0.3%
by volume and the fibres represented a case of a continu-
ous reinforcement. A matrix of a constant composition
was used for all the specimens tested. In order to avoid the
Fig. 1 – SEM in-situ test chamber fitted with the rig for three
point bending test.
141
necessity to compact the matrix by vibration a cement
slurry which had been developed for slurry infiltrated fibre
concrete [10] was used in this project. The mix propor-
tions were: 1000 g of Portland Cement 42.5R, 1000 g of
fine aggregate (passing 600 microns sieve), 550 ml of tap
water, 15 ml of superplasticizer, 5 g of underwater concrete
admixture. A mould for the specimen preparation which
was specially designed for previous research at the
University of Paisley has been utilised for this project. In
order to ensure the accurate positioning of the strands
within the specimen a detachable frame with notches and a
double sided tape were attached to the bottom plate of the
mould. After the bundles of fibres had been assembled into
the mould the matrix was carefully poured in. The speci-
mens were covered by a polyethylene sheet and they were
demoulded after 48 hours. Fibres protruding from the
specimens were cut off. The specimens were marked for
identification and placed in 20°C water for the initial cur-
ing. After five days all specimens were removed from the
water. The specimens selected for the accelerated ageing
were transferred into a hot water bath (60°C) and they
were kept there for 21 days. The aged specimens were
notched. The aim was to produce identical notches in each
specimen. Difficulties occurred due to the requirement of
the highest possible sharpness of the tip of the notch. First,
the base of the notch was carefully cut by a fine hand saw,
then the notch was “engraved” by means of a sharp steel
tool. The tip of the notch was “finished” by means of a
map pin. The result was a V-shaped notch about 1.5 mil-
limetres deep. Typical shape of the V-shaped notch is
shown in the Fig. 3. For a successful SEM observation, the
specimens had to be dried in the oven for three days at the
temperature of 75°C. The specimen preparation proce-
dures were completed by a gold coating which ensured a
much better image in the scanning electron microscope.
3. RESULTS
The results of bending tests confirmed the authors’
expectations concerning the behaviour of OPC-matrix
GRC in bending. Typical load-def lection diagrams of
Fig. 2 – Drawing of the testing rig for 3-point bending test: gen-
eral arrangements and dimensions.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

both aged and unaged specimen are shown in the Fig. 4.

The results of all three-point bending tests were sta-
tistically evaluated and are shown in Table 1.
The SEM photographs presented on Figs. 5 and 6
show the difference observed in the failure modes of the
specimens subjected to the accelerated ageing process
and those subjected to normal curing.

4. DISCUSSION
SEM in-situ bending test offers a viable alternative to
previously used in-situ test methods. Since the specimens
are subjected to bending and the load cell is connected
directly to the data capture system it is possible to deter-
mine the load on the specimen at any moment during the
testing. Another advantage of the test method presented
Fig. 3 – Detail of the V-shaped notch in the centre of a test speci-
men.

Fig. 5 – Typical failure mode of the bundle of fibres in cement

matrix subjected to accelerated ageing process. Note the short
pullout length and complete fracture of all fibres in the bundle.

Fig. 4 –Typical load-deflection diagrams of specimens subjected

to A) accelerated ageing, B) normal ageing procedures.

Table 1 – Results of 3-point bending tests

Maximum Deflection at Load at Deflection at
load max. load LOP LOP
[N] [mm] [N] [mm]
Normal average 88.60 1.91 48.72 0.57
ageing standard
procedure 4.74 0.35 6.69 0.15
deviation
Accelerated average 69.70 0.28 N/A N/A
Fig. 6 – Typical failure mode of the bundle of fibres in cement
ageing standard 7.37 0.02 N/A N/A matrix subjected to normal ageing process. Note the bridging
procedure deviation
effect of inner filaments of the bundle of fibres across very large
LOP = Limit of proportionality. crack opening.

142
Trtik, Bartos
lies in a much more precise determination of the point
where a loading crack starts its propagation. The notches
employed in the previous studies were all cut by a saw.
The tip of such a notch had been usually at least 0.5 mm
wide. The tip of the V-shaped notch used in this project
is only tens of microns wide. The onset of the crack
propagation is therefore identified much more precisely.
Since the notches were produced after the ageing proce-
dures had finished, no apparent shrinkage cracks propa-
gating from the tip of the notch were observed.
A drawback of the current arrangement of the in-situ
SEM bending test is in the necessity of an adequate dry-
ing of the specimens before testing. Here, the use of a
“wet cell” SEM which was not available to the authors
would be beneficial. The “wet cell” SEM permits testing
and scanning of moist specimens.
5. CONCLUSIONS
At this stage, major outcomes of this investigation
can be summarised as follows:
(i) A new test method which provides an opportunity for
both qualitative and quantitative observations of microfrac-
ture mechanisms of composites has been developed.
(ii) In comparison with previously used methods, the
SEM in-situ bending test provide the means for an exam-
ination of specimens of a more realistic size and rein-
forcement ratio. As the test method simulates well the
standard bending test, the specimens are subjected to a
common, practical type of loading. There are no diffi-
culties with mounting of the specimens in the test stage
within the SEM chamber.
(iii) The SEM in-situ bending test offers new avenues for
additional investigation, such as measurements of crack
mouth opening/displacement, observations of the crack
path and its propagation in relation to the load or dis-
placement applied.
143
(iv) The results obtained by means of the SEM in-situ
three point bending test confirmed the dissimilarities in
performance of aged and unaged GRC. The failure
modes of the bundles of fibres subjected to bending dif-
fered greatly according to the ageing procedure.
REFERENCES
[1] Bartos, P. J. M. and Zhu, W., ‘Assessment of interfacial
microstructure and bond properties in aged GRC using a novel
micro-indentation method’, Cement & Concrete Research 27
(1997) 1701-1712.
[2] Zhu, W., ‘Effect of ageing on durability and micro-fracture
mechanisms of fibre reinforced cement composites’, PhD Thesis,
University of Paisley, Paisley, Scotland, UK (1995) 273 pp.
[3] Diamond, S., Bentur, A. and Mindess, S., ‘Cracking processes in
steel fibre reinforced cement paste’, Cement & Concrete Research
15 (1985) 331-342
[4] Lovell, J., Diamond, S. and Mindess, S., ‘Use of Robinson
backscatter detector and “wet cell” for examination of wet
cement paste and mortar specimens under load’, Ibid. 13 (1983)
107-113
[5] Mindess, S. and Diamond, S., ‘A preliminary study of crack per-
formance in mortar’, Ibid. 10 (1980) 509-519
[6] Tait, R. B. and Garrett, G. G., ‘In-situ double torsion fracture
studies of cement mortar and cement paste inside a scanning
electron microscope’, Ibid. 16 (1996) 143-155
[7] Bentur, A. and Diamond, S., ‘Fracture of glass fiber reinforced
cement’, Ibid. 14 (1984) 31-42.
[8] Swift, D. S., ‘Bond in cement based composites reinforced with
bundles of fibres. PhD Thesis, Paisley College of Technology,
Paisley, Scotland, UK (1987) 293 pp.
[9] Duris, M., ‘Micromechanics of fracture of inclined fibres in the
cement-based composite’, PhD Thesis University of Paisley,
Paisley, Scotland, UK (1993) 252 pp.
[10] Marrs, D. L. and Bartos, P. J. M., ‘Development and testing of
self-compacting low strength slurries for SIFCON’, in
‘Production Methods and Workability of Concrete’, P.J.M.
Bartos et. al. Eds., (E & FN Spon, Chapman &Hall, London,
UK, June 1996) 199-208.
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 144-150

Fracture assessment of a high strength structural steel

TECHNICAL REPORTS

on the basis of Eurocode 3

A. Valiente1, M. Elices1, J. Ruiz1, J. A. Alvarez2, F. Gutiérrez-Solana2, J. González3,
F. J. Belzunce4 and C. Rodríguez5

Paper received: June 26, 1998; Paper accepted: December 1, 1998

A B S T R A C T R É S U M É

This work examines the design method of Eurocode 3
against brittle failure with regard to its application to a
high strength structural steel S460N EN 10113 type. The
fracture behaviour of the steel is experimentally deter-
mined by means of fracture tests, considering also the
effects of temperature and plate thickness. A direct verifi-
cation of the method of Eurocode 3 is made by contrast-
ing the assumptions on which it is based with the actual
behaviour of the steel. The failure criterion and the
mechanical properties characterizing the resistance to brit-
tle and ductile fracture are the basis of the comparison.
La méthode de calcul de l’Eurocode 3 en vue de se garantir du
risque de rupture fragile est examinée dans cet article, par rapport
à l’application à un acier à haute résistance pour construction
métallique (l’acier S460N de la norme EN 10113). Les carac-
téristiques mécaniques de la rupture de l’acier, en particulier
l’influence de la température et de l’épaisseur de laminage, ont été
déterminées par des essais de rupture sur éprouvettes préfissurées.
La méthode de l’Eurocode 3 a été vérifiée de manière directe en
comparant les fondements des règles de calcul au comportement
mesuré de l’acier. La comparaison est basée sur le critère de rup-
ture mixte (fragile/ductile) adopté par l’Eurocode et sur les valeurs
de tenacité établies à partir des valeurs de la résilience.


1. INTRODUCTION
The preservation and control of the structural integrity
is becoming a basic condition of structural design, even
though the theoretical basis for this purpose cannot be pro-
vided by classical stress analysis disregarding the existence
of f laws, and must be complemented on the basis of
Fracture Mechanics. The steel structures of civil engineer-
ing have been brought into this trend by Eurocode 3 [1],
which is probably the first attempt to refine a whole code
for design of steel structures with safety requirements
derived from Fracture Mechanics, although there are valu-
able precedents for this such as the fracture toughness spec-
ifications developed by AASHTO for steel bridges [2, 3].
The steels used in civil engineering are intentionally
very ductile materials, so Eurocode 3 has adopted the cri-
terion of the R6 method, option 2 [4] as the failure condi-
tion for structural integrity requirements. This criterion
may be roughly described as an engineering interpolation
between the two well-known failure conditions for the
extreme cases of brittle fracture and plastic collapse. Its
requirements take into account the effects of temperature,
thickness and loading rate and are formulated in accor-
dance with the rest of Eurocode 3, as a general design rule
(Annex C) for all the steels admitted by Eurocode 3.
Besides the failure criterion, an empirical correlation
between fracture toughness and the Charpy V-notch
energy is assumed in Annex C of Eurocode 3. The experi-
mental basis of this correlation is provided by research
work previously carried out at IRSID [5] to analyze the
influence of temperature on the fracture toughness of very
thick steel plates used for structural purposes in different
engineering fields. Fracture toughness was measured as a
function of temperature by using specimens of large thick-

(1) Dpto. de Ciencia de Materiales. E. T. S. Ingenieros de Caminos. Universidad Politécnica de Madrid (Spain).
(2) Ciencia e Ingeniería de los Materiales. E. T. S. Ingenieros de Caminos. Universidad de Cantabria (Spain).
(3) Dpto. de Ingeniería Minera y Metalúrgica. E. T. S. Ingenieros Industriales. Universidad del País Vasco (Spain).
(4) Dpto. de Ciencia de Materiales. E. T. S. Ingenieros Industriales. Universidad de Oviedo (Spain).
(5) Dpto. de Construcción. E. T. S. Ingenieros Industriales. Universidad de Oviedo (Spain).

Editorial Note
Prof. M. Elices is a Senior Member and a member of the Board of Advisors. He is a member of RILEM TC 147-FMB: Fracture mechanics applications to
anchorage and bond, TC 148-SSC: Test methods for the strain softening response of concrete and TC QFS: Size effect and scaling of quasibrittle fracture.

1359-5997/99 © RILEM 144

Valiente, Elices, Ruiz, Alvarez, Gutiérrez-Solana, González, Belzunce, Rodríguez
ness, and an excellent correlation was found when com-
paring the transition temperatures exhibited by fracture
toughness and by the Charpy V-notch energy. A conven-
tional definition of these two transition temperatures
yielded the same linear relationship in all the tested steels,
as well as for similar steels tested by other authors. With a
safety correction, this empirical relationship is incorpo-
rated into Annex C as a basic fracture feature of the typi-
fied structural steels.
This paper compares the fracture behaviour assumed by
Eurocode 3 for design purposes with that measured on a
standardized high strength structural steel by a series of
fracture tests with a high constraint at the crack tip. To the
authors’ knowledge, such experimental work aimed at
supporting the methodology of Eurocode 3 is scarce, and
deserves to be known by design engineers and researchers.
Given the critical influence on fracture that Eurocode 3
attributes to thickness and temperature, a wide range of
these variables is explored in the fracture tests.
Steels of tensile strength improved with no special pro-
tection of fracture toughness are often sensitive to brittle
failure. Among the structural steels, this risk is higher for
those of medium toughness and maximum tensile
strength, so a weldable fine grain structural steel with a
nominal yield strength and Charpy transition temperature
respectively of 460 MPa and -30°C was chosen for the
comparison. These steels [6] are included in the definitive
version of Annex C [1], in addition to the plain carbon hot
rolled steels [7] of the provisional version [8]. The compar-
ison is focused on the failure criterion and on the fracture
resistance values assumed in the methodology of Annex C,
since the validity of these assumptions is scarcely supported
by experimental data from the steels of Eurocode 3.
Indeed, the failure assessment diagrams of the method R6
were generalized from finite element calculations [9] and
the steels tested at IRSID included only one structural steel
for civil engineering [5].
2. THE FAILURE CRITERION OF EUROCODE 3
Annex C of Eurocode 3 is essentially a calculation pro-
cedure whose theoretical basis is given in [10]. However,
since the failure criterion of Eurocode 3 is the major
instrument to asses the fracture behaviour of the investi-
gated steel, their main features are summarized here.
The starting point of the criterion of Eurocode 3 is the
Option 2 failure assessment diagram of the R6 method for
flawed structural components [4]. These diagrams can be
viewed as engineering approaches by means of which the
well-founded failure criteria for brittle fracture and plastic
collapse are extended to cover the whole ductile to brittle
transition regime. The extended criterion comes from an
empirical generalization of single results and is formulated
as a relation between two functions of load, material and
geometry, each of which, when equalled to a constant
value, provides the failure criterion for one of the two fail-
ure mechanisms. For ideally plastic materials, with no
strain-hardening capacity, the failure assessment diagram of
the R6 method, Option 2 is:
145
P <1
PY
−1 / 2
K < 1 + P 2  (1)
Kc  
2PY2 
where P is the applied load, K is the stress intensity fac-
tor due to the applied load (primary contribution) and
the thermal and internal stresses (secondary contribu-
tion), and PY and K c are the values of P and K that
respectively produce plastic collapse and brittle fracture
of the flawed component.
Since most structural steels exhibit a plateau in the
stress-strain tensile curve immediately after yielding,
Eurocode 3 does not take into account the strain hard-
ening capacity shown by these steels after the plateau,
and adopts the failure criterion given by equation (1).
Therefore, what Eurocode 3 assumes as a safety criterion
against mixed failure between brittle fracture and plastic
collapse may be described by saying that a flawed com-
ponent cannot sustain any applied load above that of
plastic collapse or any stress intensity factor above a frac-
tion of that of brittle fracture, this fraction being the
function of the ratio between the applied load and the
plastic collapse load given by equation (1) and Fig. 1.
Fig. 1 - Failure criterion of Eurocode 3.
The material properties present in the failure criterion
are the yielding and fracture resistance, represented respec-
tively by the yield stress assumed to determine the plastic
collapse load PY and by the material constant Kc. Since
structural steels are considered by Eurocode 3 as non-hard-
ening materials, PY must be calculated from the 0.2% yield
strength fY as the yield stress. The remaining variables that
determine PY are geometrical, so fY is the only mechanical
property through which the material influences the plastic
collapse load. For a given structural steel, fY depends on
temperature, but also on plate thickness.
The fracture resistance Kc recommended for design
purposes by Eurocode 3 for hot rolled structural steels
(plain carbon and weldable fine grain steels) should be
properly considered as a “design” fracture resistance. This
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
design property is given by Eurocode 3 as a function of
thickness, temperature, strain rate, yield strength and
toughness (as determined by a characteristic temperature
measured in Charpy tests rather than in fracture tests). As
already mentioned, the relationships that provide Kc are
based on experimental data and empirical correlations,
coming from European steels as to the effect of tempera-
ture and thickness [5], and generalized from American
steels as regards strain rate effects [11]. The axiomatic defi-
nition of the design fracture resistance Kc implicit in
Eurocode 3 is given by the following assumptions:
a) For each temperature and strain rate there exists a
minimum thickness Bm above which the toughness Kc
achieves a value KIc (plain strain fracture toughness)
independent of thickness. The minimum thickness and
the plain strain fracture toughness are related by:
2
K 
Bm = 2.5 Ic 
 fY 
fY being the 0.2% yield strength.
b) The plain strain fracture toughness KIc depends on
temperature T and strain rate ε⋅ according to the expres-
sion:
K Ic = K0eα(T −T0 ) (3)
where K0 = 100 MPam1/2, -
α = 0.01(°C) and T0 is a
1
temperature whose value in °C is that given by:
T0 = 25°C + 1.4T27 + (83 − 0.08 fY )ε˙ 0.17
T 27 being the test temperature (in °C) at which the
Charpy V-notch energy has a minimum value of 27
Joules for the steel, the 0.2% yield strength fY being
expressed in MPa and the strain rate ε⋅ in s-1.
c) For thicknesses under the minimum B m, the
design fracture toughness Kc is the decreasing function
of thickness B given by:
0.4
B 
0.4  K  
2
Kc = K Ic  m  = K Ic  2.5  Ic   = 1.4427
 B  B  fY  
 
3. MATERIALS AND TESTS
A S460N EN 10113-2 type structural steel was used
for comparison with the assumptions of Eurocode 3.
This weldable fine grain steel is a normalized/normalized
hot rolled product and was supplied in four different
plate thicknesses 10, 15, 30 and 50 mm. The mechanical
test programme included tensile, Charpy and fracture
tests, complemented with chemical and microstructural
analysis.
Chemical composition
Table 1 gives the chemical composition of the steel as
determined from the final product. The results meet the
requirements of EN 10113-2 for normalized/normalized
rolled steels and confirm the presence of alloying ele-
ments (Ni, Al, V, Nb) used for grain refinement and
transformation control.
Microstructure
The microstructure of the four plates showed alter-
nating bands of ferrite and perlite (Fig. 2) with the fer-
ritic grain size ranging from 7 µm for the thinnest plate
to 15 mm for the thickest one. In addition to the general
ferritic-perlitic microstructure, a bainitic layer was
detected at the middle plane of the plates. Microhardness
measurements showed that this layer was more brittle
than the remaining material, the difference being an
increasing function of the plate thickness.
(2)
Fig. 2 – Microstructure of the plate 30 mm thick with alternating
perlitic and ferritic layers (× 400).
(4)
Tensile tests
Tensile tests were performed on the four plates at three
temperatures: –30°C, 0°C and 50°C. At least three
threaded cylindrical specimens, of 6 mm diameter, were
tested for each thickness plate and temperature. They were
machined longitudinally oriented in the rolling direction
and centered in the plate thickness. A strain gage exten-
someter attached to the specimen allowed its elongation to
K Ic1.8 be measured up to fracture, thus determining the stress-
(5)
B0.4 fY0.8 strain curve of the steel at the test temperature. In all the
cases, regardless of temperature, yielding occurs at essen-
tially constant stress, after which the steel strain hardens so
that stress increases with strain up to necking. The most
significant mechanical properties (0.2% yield strength,
tensile strength, maximum load elongation and ultimate
elongation) determined from the tensile tests and the
resulting experimental scatter are shown in Fig. 3 as a
function of plate thickness and temperature. The specifica-
tions of EN 10113-2 for S460N steel of these properties at
room temperature are fulfilled by the four steel plates.
In the method of Eurocode 3, the only relevant ten-
sile property is the 0.2% yield strength fY. To take into
account the effect of the plate thickness on fY, the fol-
lowing linear relationship between f Y and the plate
Table 1 – Chemical composition of the steel
C Si Mn P S Al Ni Nb V
0.18 0.41 1.54 0.015 0.002 0.071 0.47 0.057 0.14
146
Valiente, Elices, Ruiz, Alvarez, Gutiérrez-Solana, González, Belzunce, Rodríguez
Fig. 3 - Tensile properties as a function of temperature and plate thickness (●:
–30°C, ▲: 20°C, ■: 50°C).
thickness B is assumed in Annex C:
fY = fY 0 (1 − 1 B )
940 B0
where B0 is a reference thickness (1 mm) and fY0 is the
nominal 0.2% yield strength of the steel grade, namely,
460 MPa for steel S460N of EN 10113-2. This relation-
ship seems to be a linear fit of the minimum values of fY
prescribed by this standard and EN 10025 for each steel
quality as a function of the plate thickness B. Equation
(6) is plotted in Fig. 3 for the 460 grade and, as expected
from its origin, it is a conservative prediction of the yield
strength, even at the highest test temperature.
Charpy tests
V- notch Charpy tests were carried out to determine
the energy-temperature transition curve from -196°C to
100°C for each plate thickness at no fewer than 20 points.
The specimens were machined with L-T orientation cen-
tred in the plate thickness. A hyperbolic tangent curve fit-
ted to the experimental points was used to determine the
transition temperature T27 as defined in the previous sec-
tion from Annex C of Eurocode 3 (Fig. 4).
Fig. 4 - Charpy energy– temperature curves for the four plate
thicknesses.
The four experimentally determined
values of T27 are equal to or lower than the
maximum value of -30°C prescribed by
Annex C for 460 N steel of EN 10113.
Their plot against plate thickness (Fig. 5)
shows a fast increase of T27 with the plate
thickness B that becomes slower as thick-
ness increases.
Fracture tests
Fracture tests based on E813 ASTM
Standard [12] were performed for the steel
plates with thicknesses of 15, 30 and 50
mm at -30°C, 20°C and 50°C. Compact
tensile specimens were used in all the tests,
with the width and thickness indicated in
Table 2 as a function of the plate thickness.
Fatigue precracking was carried out
according to the standard up to a crack size
(6)
Fig. 5 - Charpy transition temperature as a function of plate
thickness
around 0.6 times the specimen width. After precracking
the specimens were side-grooved up to a total reduction of
20% of the gross thickness. Two specimens were tested for
each temperature and plate thickness. The crack growth
produced throughout the test was determined from the
elastic compliance measured at the load line by performing
periodically partial unloadings.
Two different types of macroscopic failure mecha-
nism were observed in the fracture tests: unstable crack
extension and uninterrupted ductile tearing. The fatigue
precrack, as well as the crack extension due to ductile
tearing when this occurred, were measured on the frac-
Table 2–Dimensions of the compact tensile specimens
Plate Thickness Specimen Width Specimen Thickness
15 mm 60 mm 15 mm
30 mm 60 mm 30 mm
50 mm 90 mm 45 mm
147
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
ture surfaces by the procedure of the standard. For unin-
terrupted ductile tearing the failure load was identified
with the load at crack growth initiation, as determined
from the J Integral curve versus crack extension accord-
ing to the standard. For unstable crack extension, the
instability load was taken as the failure load.
The level of yielding previous to failure (as shown by
the load–crack opening displacement register) and the
macroscopic failure mechanism show three modes of fail-
ure from an engineering point of view: brittle failure
(unstable crack extension with no macroscopically signifi-
cant previous plastic deformation), ductile failure (uninter-
rupted ductile tearing) and transition failure (unstable
crack extension with macroscopically significant previous
plastic deformation, preceded or not by ductile tearing).
Fig. 6 shows three registers obtained for the 50 mm thick
plate, one at each test temperature. The engineering mode
of failure for each plate thickness and temperature is given
in Table 3. When the pair of specimens tested in the same
test conditions failed in a different mode, the two failure
modes are indicated.
Fig. 6 - Load-COD curves recorded at the fracture tests for the
50 mm thick plate.
Table 3 – Engineering failure mode produced in the
fracture tests (B: Brittle; D: Ductile; T: Transition)
Plate Thickness Temperature 15 mm 30 mm 50 mm
-30°C T T/B B
20°C D T T so this criterion provides a wide safety margin through a
50°C D D D/T
Since the purpose of the tests was a different one, no
attempt was made to use the results for measuring the
fracture toughness of the steels by means of the J inte-
gral. As far as the plain strain fracture toughness KIc is
concerned, not in any test the specimen thickness was
enough to maintain a state of plain strain up to fracture,
so no valid measurement of this property was obtained.
148
4. FRACTURE ASSESSMENT OF THE S460N
EN 10113-2 STEEL
A first assessment of the fracture behaviour of the
steel, made with the failure criterion of Eurocode 3 as a
reference, would consist of comparing the failure condi-
tion given by equation (1) with the experimental failure
loads of the tested specimens. To this end, the compact
tensile specimen would provide not only a particular
geometrical configuration to explore a possible inf lu-
ence of size, thickness, and crack size on the failure cri-
terion, but also to produce a high stress triaxiality at the
crack tip. In addition to this, the possible inf luence of
temperature and plate thickness would also be explored,
since a wide range of these variables was covered in the
test programme.
To plot the experimental points from the fracture
tests in the failure assessment diagram of Fig. 1, the fol-
lowing values were used:
– The failure load P of each compact specimen as
described in the previous section.
– The elastic stress intensity factor K at failure deter-
mined according to the standard from the above failure
load and the measured crack size.
– The fracture resistance Kc determined from equa-
tions (5, 3, 4) with the gross thickness B of the speci-
men, the strain rate ε⋅ applied in the tests (10-4s-1), the
transition temperature T27 of Fig. 5, and the 0.2% yield
stress fY calculated from equation (6) for the thickness of
the plate and the nominal yield stress fY0 = 460 MPa of
the steel.
– The plastic collapse load PY of the compact tensile
specimen as computed from the expression (see, for
instance, [13]):
 2 
PY = BN WfY  2[1 + a 2 ] − 1 − a  (7)
 W W
where BN is the net thickness of the specimen, W is the
specimen width, a is the crack size, and fY is again the
0.2% yield stress of the steel as derived from equation (6).
The failure data corresponding to the 18 fracture tests
are plotted in the failure assessment diagram of Eurocode
3 as shown in Fig. 7. All these failures occur well beyond
the limit defined by the failure criterion of Eurocode 3,
conservative prediction of the failure load, as should be
expected on account of its generality.
In fact Fig. 7 shows that the ratio P/PY holds good far
from unit even for low temperatures and large plate thick-
nesses, when brittle failure occurs and the ratio K/Kc
approaches unit. This conservatism might be due to the
use of the 0.2% yield strength to compute the plastic col-
lapse load, since proceeding in this way the strain harden-
ing capacity of the steel is neglected. Indeed, values even
higher than the tensile strength should be used for predict-
ing the maximum loads sustained by the specimens in the
tests as a plastic collapse load.
Valiente, Elices, Ruiz, Alvarez, Gutiérrez-Solana, González, Belzunce, Rodríguez

Fig. 7 - Experimental failure results compared with the failure

criterion of Eurocode 3.
Plate 15 mm thick:-30°C (●), 20°C (), 50°C (❍)
Plate 30 mm thick:-30°C (▲), 20°C (), 50°C ()

With the exception of the thickest plate (50 mm) at

the lowest temperature (-30°C), the failure of the speci-
mens was not brittle, so a fracture criterion based on
Linear Elastic Fracture Mechanics, i. e., K = Kc, cannot
be applied. Consequently, only in that test condition is it
possible to ascertain whether the design fracture resis-
tance Kc of Eurocode 3 has the meaning of a critical
stress intensity factor for brittle fracture.
Nevertheless, the remaining tests provide informa-
tion concerning this question if the influence of temper-
ature and plate thickness on the fracture tests is exam-
ined as in Fig. 8, where the values of the elastic stress
intensity factor at failure are plotted against temperature
for each plate thickness. Likewise, the design fracture
resistance of Eurocode 3, Kc, has been drawn as a func-
tion of temperature from equations (5, 3, 4, 2, 6) partic-
ularized for the thicknesses and transition temperatures
T27 of the tested plates.
For the three plate thicknesses, the trend shown by the
experimental points as temperature decreases may be
described by an extrapolation curve (the dotted curves of
Fig. 8), which roughly meets the curve of the fracture resis-
tance values of Eurocode 3 at the transition from ductile to
brittle failure. In fact, for the largest plate thickness this is
not an extrapolation, but the reality since the two curves
join. So Fig. 8 provides a positive semiquantitative verifica-
tion of the values Kc that Eurocode 3 assigns to the tested
steel plates: according to the failure criterion of equation (1)
and Fig 1, as plastic collapse interferes with fracture, the
experimental values of the elastic stress intensity factor at
failure are lower than the design fracture resistance, but the
two values coincide as the interference vanishes. This shows
that the design fracture resistance of Eurocode 3 has the
meaning of a critical stress intensity factor when brittle frac-
ture occurs, but one has to be careful not to use it with this
aim, because it provides no safety margin for non brittle
failure as confirmed by Fig. 8. The reason is that the value
Kc of Eurocode 3 is not an upper bound of the elastic stress
intensity factor under which no failure can occur, but a
Fig. 8 - Elastic stress intensity factor at failure ( ❍: experimental
values; : equation (5)).
material constant measuring the fracture resistance through
the elastic-plastic failure criterion of equation (1), and
should not be used out of this context.
The curve Kc(T) adopted for design in Eurocode 3 is
the empirical one of reference [5] shifted to the right by
25°C, since the linear correlation of reference [5] was
replaced in Annex C by equation (4), a parallel straight
line with an offset of 25°C [10]. For the tested steel, the
safety margin gained with the offset seems to be almost
exhausted, since the two Kc values from brittle fracture
found experimentally are on the shifted curve instead of
being on that of reference [5].
5. CONCLUSIONS
The fracture behaviour of a high strength structural
steel S460N EN 10113 type and the influence of tempera-
ture and plate thickness were determined by direct mea-
surements made on compact tensile specimens. As regards
the application of the design method of Eurocode 3 against
brittle failure to this steel, the results allow the following
conclusions to be drawn for the explored ranges of temper-
ature, thickness and strain rate:
1. The procedure of Eurocode 3 for design against

149
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999
brittle failure is safe for the tested steel, which indicates
that the extension of the method to the high strength
structural steels of EN 10113 is successful.
2. The failure criterion assumed by Eurocode 3 pro-
vides a reasonable safety margin for predicting the failure
load of a cracked member, even under the conditions
most favourable to brittle failure: the lowest tempera-
ture, the largest thickness, and a high stress triaxiality at
the crack tip.
3. The design fracture resistance of Eurocode 3 should
be used only in conjunction with the failure diagram of
Fig. 1. In particular, this fracture resistance should not be
identified with fracture toughness for design based on
Linear Elastic Fracture Mechanics, since it would produce
unsafe results as is shown by this research.
ACKNOWLEDGMENTS
This work was supported by the Spanish Department
for Public Works (Ministerio de Fomento) under
Contract 30.270/93.3. The contribution of ACERALIA
(Largos) in supplying the steel for the experiments is also
gratefully acknowledged by the authors. The authors A.
V., M. E. and J. R. also thank CICYT for additional
support from PETRI 95-0091-OP.0210
REFERENCES
[1] European Committee for Standardization (CEN), ‘European
Prestandard ENV 1993-1-1/A1: Eurocode 3: Design of Steel
Structures’, European Committee for Standardization, Brussels,
1994.
150
[2] American Association of State Highway and Transportation
Officials, ‘AASHTO Standard Specifications for Highway
Bridges, Section 7’, The Association General Offices,
Washington, 1977.
[3] Barsom, J. M., ‘Development of the AASHTO Fracture-
Toughness Requirements for Bridge Steels’, Engineering Fracture
Mechanics 7 (3) (1975) 605-618.
[4] Milne, I., Ainsworth, R. A., Dowling, A. R. and Stewart, A. T.,
‘Assessment of the Integrity of Structures Containing Defects’
Report R/H/R6–rev 3, Central Electricity Generating Board,
London, 1984.
[5] Marandet, B. and Sanz, G., ‘Étude par la mécanique de la rupture
de la ténacité d’aciers à résistance moyenne fournis en forte épais-
seur’, Revue de Metallurgie (April 1976) 359-383.
[6] European Committee for Standardization (CEN), ‘EN 10113:
Hot Rolled Products in Weldable Fine Grain Structural Steels’,
European Committee for Standardization, Brussels, 1993.
[7] European Committee for Standardization (CEN), ‘EN 10025
+EN 10025/A1: Hot Rolled Products of Non-Alloy Steels’,
European Committee for Standardization, Brussels, 1993.
[8] The Commission of the European Communities, ‘Eurocode 3:
Design of Steel Structures’, The Commission of the European
Communities, Brussels, 1990.
[9] Chell, G. C. ‘A Procedure for Incorporating Thermal and
Residual Stress into The Concept of A Failure Assessment
Diagram’, ASTM STP 668, (ASTM, Philadelphia, 1979) 581-
605.
[10] Brozzetti, J., Sedlacek, G. and Hensen, W. ‘Fondements des
règles de l’Eurocode 3 en vue de se garantir du risque de rupture
fragile’, Construction Metallique 1 (1991) 15-28.
[11] Barsom, J. M. and Rolfe, S. T. ‘Correlations between KIc and
Charpy V-Notch Test Results in the Transition-Temperature
Range’, ASTM STP 466, (ASTM, Philadelphia, 1970) 281-382.
[12] American Society for Testing and Materials, ‘ASTM E813
Standard: Standard Test Method for JIc, A Measure of Fracture
Toughness’, ASTM, Philadelphia, 1988.
[13] Kanninen, M. F. and Popelar, C. H. ‘Advanced Fracture
Mechanics’, Oxford and Clarendon Press, New York, 1985.
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 151-157
Caractérisation des bétons de sable routiers compactés :
application au cas du Sénégal
(Characterisation of compacted road sandcretes: Study applied to Senegal)
I. K. Cisse1, M. Laquerbe1, A. Gaye2, M. Diene3
(1) Laboratoire de Géomécanique, Thermique, Matériaux ( G.T.Ma.), I.N.S.A.,Génie Civil, Complexe Scientifique de Rennes - Beaulieu,
35043 Rennes Cedex, France
(2) École Nationale des Ponts et Chaussées, 28 rue des Saints Pères, 75007 Paris
(3) Société Africaine de Géotechnique, 113 avenue Lamine Gueye, B.P. 10945, Dakar (Sénégal)
Article reçu : 9 juillet 1997 ; Article accepté : 20 mai 1998
A B S T R A C T
En substitution aux matériaux hydrocarbonés dont
l’importation grève la balance commerciale déjà défici-
taire au Sénégal, les bétons de sable peuvent être une
alternative de taille n’utilisant que des matériaux locaux,
excédentaires (sable des dunes) et résiduels (sables filléri-
sés de concassage).
Les études ont porté sur quatre bétons de sable com-
pactés utilisant comme ajouts, des sables fillérisés 0/3,
résidus de concassage de calcaire, de grès, de silexite et
de basalte.
Les caractéristiques physiques et les performances
mécaniques des bétons confectionnés sont largement ras-
surantes pour justifier le choix des bétons de sable compac-
tés en structures de chaussées. En effet, en comparaison
avec des chaussées souples, des économies financières
variant entre 54% et 110% peuvent être réalisées suivant le
type d’ajout et la classe de trafic considérés.
1. INTRODUCTION
L’accentuation du déficit de la balance commerciale du
Sénégal, causée entre autres par l’importation des maté-
riaux hydrocarbonés tels le bitume, a incité les maîtres
d’ouvrages, maîtres d’œuvre, organismes techniques à
rechercher des palliatifs, notamment en faisant appel aux
techniques permettant de valoriser les matériaux dispo-
nibles. Dans ce contexte, il est aisé d’entrevoir l’intérêt tant
économique qu’écologique que pourrait présenter l’utilisa-
tion des matériaux excédentaires (sables des dunes) ou rési-
duels (sables fillérisés de concassage).
En effet, les sables des dunes occupent 70% du terri-
toire national [1] et procurent des performances méca-
niques similaires à celles des sables de plage [2] interdits
d’exploitation par décret préfectoral, pour des raisons de
protection de l’environnement . Quant aux sables filléri-
sés 0/3 (de basalte, de grès, de silexite et de calcaire), ce
sont des matériaux résiduels inutilisés, issus du concas-
sage des roches pour granulats à béton classique.
L’objet de cette étude qui s’inscrit dans le cadre de la
valorisation des matériaux locaux, est de mettre en évi-
1359-5997/99 © RILEM 151
LOCAL MATERIALS
R É S U M É
The import of bituminous binders to Senegal is putting a
strain on the balance of trade which is already in deficit. It is
therefore essential to find substitutes for these materials. The use
of concretes made from fine aggregates exclusively is an alternative
to materials using bituminous binders; this type of concrete can be
made using local materials, naturally occuring sands from dunes
and residual sands from the crushing process.
Studies have been carried out on four compacted concretes
made from fine aggregates exclusively using 0-3 mm filled sands
produced from the crushing of limestone, sandstone, cherts and
basalts.
The physical characteristics and mechanical performance of
these materials are very encouraging for justifying the choice of
sandcrete in road construction. A comparison with flexible pave-
ments has shown that an economic advantage varying between
54% and 110% can be obtained, depending on the type of addi-
tion and the road classification.
dence les performances techniques et la compétitivité
financière des bétons de sable compactés en structure
routière.
2. ORIGINE ET IDENTIFICATION DES
MATÉRIAUX UTILISÉS
2.1 Le sable des dunes
Il s’agit du sable des dunes rouges qui existe en quantité
« inépuisable ». Dans le cadre de cette étude, les prélève-
ments ont été effectués dans un quartier de la proche ban-
lieue de Thiès (Fig. 1) et plus précisément à Ballabey. Le
sable quartzeux à 99% est recouvert d’une mince pellicule
ferri-argileuse qui lui confère sa couleur rouge. Il est égale-
ment constitué de grains en majorité émoussés – luisants
(95%), ce qui atteste de son origine éolienne, avec 5% de
ronds – mâts. Il renferme cependant quelques matières
organiques qui en amoindrissent la propreté. Les résultats
de l’analyse granulométrique (Tableau 1) révèlent l’extrême
finesse des particules (Mf = 1,31) avec une granulométrie
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Fig. 1 – Régions de production des granulats concassés au Sénégal (source : Plan Minéral du Sénégal, volume 1).

Tableau 1 – Résultats des essais d’identification des sables Il est de couleur blanche, attestant de
la forte teneur en carbonate de cal-
Matériaux Cu Mf γ s absolue Abs Sp γ s réelle ES
(t/m3) (%) (cm2/g) (t/m3) (%) cium, avec une structure compacte.
Quant au silexite, il provient de
Sable des dunes 2,58 1,31 2,65 0,84 4365 2,55 52 la Compagnie Sénégalaise des
Sable fillérisé de calcaire 5,00 2,67 2,69 3,56 4628 2 ,16 Phosphates de Taïba, à 30 km à vol
Sable fillérisé de grès 4,67 2,03 2,70 1,43 4789 2,19 d’oiseau de Thiès (Fig. 1). C’est un
rejet industriel non recyclé. Il se
Sable fillérisé de silexite 7,69 2,49 2,73 1,47 3008 2,21
présente sous forme de blocs à cœur
Sable fillérisé de basalte 9,17 2,66 2,98 1,63 1704 2,29 sombre et à cortex blanc. C’est le
cœur que l’on retrouve le plus sou-
continue (Cu > 2). Par ailleurs, la valeur de l’équivalent de vent sous forme de résidus. La roche est composée en
sable (ES = 52) est indicative de la présence d’une certaine moyenne pondérale de 91% de silex, 6% d’indurés
teneur en fines qui peut être favorable aux bétons compac- phosphatés et 3% de fines d’éléments argileux [6]. Les
tés, composés entres autres de graves 0/D [3]. caractéristiques physiques, pondérales et granulomé-
triques résumées au Tableau 1 montrent que les sables

2.2 Les sables fillérisés Tableau 2 – Résultats des essais d’analyse chimique
(SO.CO.CIM. Sénégal)
Le basalte prélevé à Diack situé à 30 km de Thiès Oxydes Calcaire Grès Silexite Basalte
(Fig. 1) est à grains fins, de couleur gris-noir. L’étude
SiO2 10,07 89,30 69,05 49,33
minéralogique montre qu’il est constitué de phénocris-
taux de pyroxène et d’olivine dans une mésostase riche Al2O3 1,00 5,06 1,33 13,48
en verre et en microlites de plagioclase. Fe2O3 1,20 2,43 3,43 12,04
Le grès qui provient de Paki Toglou à 20 km de CaO 74,50 0,25 16,64 9,69
Thiès (Fig. 1) est de couleur rosâtre, à ciment siliceux [4]
et à grains fins constitués essentiellement de quartz [5]. MgO 1,58 0,12 0,90 8,64
Le calcaire est prélevé à Bargny, à 30 km de Thiès (Fig. 1). Perte au feu 3,40 1,80 2,40 1,33

152
Cisse, Laquerbe, Gaye, Diene

fillérisés présentent une teneur en fillers variant entre

4, 54 × 45, 7 × 25 × 5
2% et 7%, participant de manière significative à la com- E = MHnN = ≈ 27, 50 kg/cm 3
pacité, avec une granulométrie continue (Cu > 7), favo- V 943, 70
rable à l’obtention d’une bonne résistance mécanique. Le Le nombre de coups de dames sera pour le moule de
Tableau 2 représentant les résultats d’analyse chimique diamètre 11cm, égal à :
révèle la forte teneur en silice du basalte, du grès et du
27, 50 × 2090, 73
silexite. Ces silices peuvent suivant leur état de cristalli- N = EV = = 55, 4 ≈ 55 coups
sation, contribuer à améliorer la résistance mécanique. Mh 4, 54 × 45, 7 × 5
Pour le compactage proprement dit, les masses de
sables sont telles que :
2.3 Le ciment
Sd
=2 (1)
Il s’agit d’un CPJ-CEM II 32,5, fabriqué par la Sf
Société de Commercialisation du Ciment avec :
(SO.CO.CIM.), située à 40 km de Thiès. Ces caractéris- Sd = sable de base = sable des dunes
tiques principales sont résumées au Tableau 3. Il est à Sf = ajout = sable fillérisé.
signaler que la tendance actuelle, pour des raisons écono- Ainsi, pour une masse totale de 900 Kg, on aura
miques et techniques (vitesse de prise faible donnant une 300 Kg de sable fillérisé et 600 Kg de sable des dunes.
plus grande souplesse au niveau de l’exécution du chan-
tier), est maintenant d’utiliser des ciments à forte teneur
en ajouts (cendres volantes, laitiers, pouzzolanes) [3]. 3.2 Détermination de l’optimum Proctor
Donc, nous sommes bien en conformité avec la norme
en utilisant un CPJ - CEM 32,5 dont la classe réelle est Nous avons fait varier les teneurs en eau de compac-
d’ailleurs de 42,5 (Tableau 3). Cette résistance relative- tage de 4 à 12%, par pas de 2%. Les résultats obtenus
ment élevée s’expliquerait par la valeur de la surface spé- (Wopt, γdmax) sont consignés dans le Tableau 4. L’examen
cifique Blaine très élevée (4 600 cm2/g). de ce tableau appelle de notre part quelques commentaires :
1) On remarque que les teneurs en eau optimales
Tableau 3 – Valeurs caractéristiques du ciment utilisé mesurées après passage à l’étuve sont plus faibles que celles
introduites lors de la préparation des mélanges. Ceci est
Masse spécifique (g/cm3) 3,15 dû à la non restitution de l’eau liée physiquement aux
Surface spécifique (cm2/g) 4060 hydrates et nécessaire à l’hydratation du ciment [7], relati-
2 jours 4,0 vement importante et à celle absorbée par les granulats
(sable des dunes et sables fillérisés) beaucoup moins
Résistance à la traction (MPa) 7 jours 6,7
importante. En tout état de cause, les teneurs en eau opti-
28 jours 8,9 males au P. M. sont conformes à celles des bétons routiers
2 jours 14,4 classiques compactés (4 % ≤ Wopt ≤ 7%) [3] puisqu’elles
Résistance à la compression (MPa) 14 jours 26,0 varient entre 4% et 9%.
2) On remarque que l’ordre de classement des densi-
28 jours 42,5 tés maximales est le même que celui des densités des
ajouts. Ceci s’explique par le fait que la densité maximale
dépend certes de la teneur en eau mais aussi, de la den-
sité absolue des grains (γs).
3. DÉTERMINATION DE L’OPTIMUM
PROCTOR DES BÉTONS DE SABLE Tableau 4 – Résultats des essais de compactage en Proctor modifié
COMPACTÉS
Calcaire Grès Silexite Basalte
Paramètres de compactage fillérisé fillérisé fillérisé fillérisé
3.1 Méthodologie adoptée 0/3 0/3 0/3 0/3
Pourcentage d’eau ajouté (%) 8 10 10 10
L’objet étant de faire apparaître l’effet du compactage, le
dosage en ciment choisi a priori, sera plus faible que celui Teneur en eau optimale Wopt (%) 3,97 9,44 6 5,76
des bétons routiers classiques (coulés sur place), nous le Densité sèche maximale γ dmax (t/m3) 1,88 1,87 1,94 1,98
prendrons égal à 200 kg/m3 au plus . La méthode de
dosage adoptée et exposée au paragraphe 4, conduira effec-
tivement à un dosage inférieure à 200 kg/m3. Puisqu’il
s’agit par ailleurs de matériaux pour couche de base, on 4. FORMULATION DES BÉTONS DE SABLE
appliquera au moule de diamètre 11 cm et de hauteur COMPACTÉS
22 cm (volume de 2 090,73 cm3), l’énergie équivalente à
celle du Proctor Modifié (P. M.), appliquée au moule stan- La formulation est basée sur les résultats d’essais de
dard de diamètre de 10,16 cm et de hauteur 11,64 cm compactage. Aussi, pour un dosage en ciment de 200 kg,
(volume de 943,70 cm3). L’énergie à appliquer sera alors : on respecte la relation (1), avec une teneur en eau égale

153
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

Tableau 5 – Variation de la résistance mécanique avec cela devrait avoir une conséquence sur le comportement
la composition des bétons à compacter (cas du béton mécanique.
à ajout de sable fillérisé de calcaire)
Ciment (Kg) 200 200 200 200 200 200 5.2 Caractéristiques mécaniques
Sable des dunes (Kg) 1100 1200 1300 1400 1500 1450
Sable fillérisé (Kg) 550 600 650 700 750 725
Dans l’optique d’un dimensionnement des chaussées
suivant le catalogue SETRA-LCPC [8], nous avons
Eau (l) 148 160 172 184 196 190 effectué des essais de résistance à la traction par fendage à
R c 28j (MPa) 3,4 3,4 3,5 3,9 3,8 3,9 7 et 28 jours. Les résultats obtenus sont résumés aux
R t 28j (MPa) 1,37 1,53 1,69 1,72 1,73 1,73 Tableaux 8a et 8b suivant que le mode de mûrissement a
lieu à l’air ou à l’eau. Les commentaires suivants peuvent
Compacité du béton 0,899 0,919 0,903 0,931 0,954 0,952
être faits :
E/C 0,74 0,80 0,86 0,92 0,96 0,95 1) Les coefficients d’âge Rt28/Rt7 indiquent une aug-
mentation de la résistance quelque soit le
mode de cure. Dans le cas de la cure à l’air,
Tableau 6 – Dosage des bétons de sable à compacter, suivant le type d’ajout cela est lié au fait qu’on a très peu de perte
Ajout Ciment Sable des Sable fillérisé Eau Ciment/Sable d’eau par évaporation comme on l’a remar-
(Kg) dunes (Kg) (Kg) (litres) total (%) qué au Tableau 7. En effet, un phénomène
Calcaire 0/3 200 1450 725 190 9,2 inverse a été observé par CISSE [4] sur des
bétons de sable non compactés, donc relati-
Grès 0/3 200 1500 750 245 8,9
vement moins fermes. Par ailleurs, la faible
Silexite 0/3 200 1725 862,5 278,75 7,7 perte d’eau au bout de 28 jours nous fait
Basalte 0/3 200 1750 875 282,5 7,6 penser que la cure à l’air n’est pas préjudi-
ciable aux bétons confectionnés.

non pas à l’optimum mais, à celle conduisant à Wopt. Tableau 7 – Variations pondérales des éprouvettes
Autrement dit, pour le béton à ajout de sable fillérisé de cal- de béton de sable compacté (BSC)
caire par exemple, on ajoutera 8% et non pas 3,97% d’eau. BSC BSC BSC BSC
Après avoir déterminé l’optimum de compactage (Wopt), et Paramètres pondéraux
calcaire grès silexite basalte
à l’instar de la méthode expérimentale de dosage de l’A.C.I.
Masse volumique (Kg/m3) 2012 2064 2131 2178
(American Concrete Institute), on cherche maintenant la
meilleure formulation c’est-à-dire celle respectant la teneur Masse à 0 jour (g) 4207 4316 4246 4554
en eau optimale et la quantité de ciment de 200 Kg mais Masse à 7 jours cure à l’air 4097 4170 4101 4392
avec le maximum de granulats (sables) tout en fournissant la (M7j) (g) cure à l’eau 4272 4250 4272 4461
meilleure résistance mécanique. Ainsi, pour le béton à ajout
de sable fillérisé de calcaire par exemple, on aboutit aux Masse à 28 jours cure à l’air 4067 4128 4069 4353
résultats du Tableau 5 qui permettent de retenir comme (M28j) (g) cure à l’eau 4301 4279 4312 4573
formulation optimale celle qui est mentionnée au Tableau cure à l’air 0,99 0,99 0,99 0,99
6. L’analyse de ce tableau permet de constater – et ceci est M7j/M28j
cure à l’eau 1,01 1,01 1,01 1,03
lié à l’esprit de la méthode – que le volume du mélange
retenu peut dépasser 1 m3. Toutefois, on vérifie également
que les teneurs en ciment sont dans les limites de celles d’un Tableau 8a – Résistances à la traction par fendage des
béton routier compacté de graves puisqu’elles sont bien bétons mûris à l’air (BSC = Béton de Sable Compacté)
comprises entre 7% et 14%.
Résistance BSC calcaire BSC grès BSC silexite BSC basalte
(MPa) 0/3 0/3 0/3

5. CARACTÉRISATION DES BÉTONS DE Rtf 7j 1,40 1,11 1,82 1,82

SABLE COMPACTÉS Rtf 28j 1,79 1,78 2,06 2,22
Rtf 28j /Rtf 7j 1,27 1,6 1,13 1,22
5.1 Caractéristiques physiques
Quelques paramètres pondéraux des bétons confec- Tableau 8b – Résistances à la traction par fendage des
tionnés sur la base de la méthode de formulation men- bétons mûris à l’eau (BSC = Béton de Sable Compacté)
tionnée ci-dessus sont donnés au Tableau 7. On y constate Résistance BSC calcaire BSC grès BSC silexite BSC basalte
que les bétons de sable compactés sont plus légers que les (MPa) 0/3 0/3 0/3
bétons ordinaires puisque les masses volumiques sont Rtf 7j 1,48 1,43 2,15 2,03
comprises entre 2 012 Kg/m3 et 2 178 kg/m3, ils seront
alors plus manipulables. Par ailleurs, les variations des Rtf 28j 2,11 1,91 2,63 2,30
masses (perte ou gain) sont faibles quelque soit l’ajout et Rtf 28j /Rtf 7j 1,43 1,34 1,22 1,13

154
Cisse, Laquerbe, Gaye, Diene

2) Les résistances mesurées sur les bétons Tableau 9 – Différentes classes de trafics définies par le catalogue
mûris à l’eau sont plus importantes que celles SETRA/LCPC
mesurées sur les bétons mûris à l’air. Cela Classes
semble tout à fait normal. En effet, la présence de trafic
T6 T5 T4 T3- T3+ T2 T1 T0
d’eau dans le béton favorise le processus
Nombre 0 10 25 50 100 150 300 750
d’hydratation. C’est ainsi qu’il faut que la cure de Poids à 10 à 25 à 50 à 100 à 150 à 300 à 750 à 2000
à l’eau soit systématique et que le dimension- Lourds/j
nement des couches de chaussées se fasse sui-
vant la résistance mesurée après
une cure à l’eau. Tableau 10 – Différentes classes de portances définies par le catalogue SETRA/LCPC
3) L’ordre de classement des Module de
valeurs de résistance s’explique par Indice portant
Portance Examen visuel (essieu de 13 t) déformation à la
CBR
plusieurs facteurs dont la densité plaque EV2 (MPa)
sèche maximale, la pouzzolanicité P0 Circulation impossible, sol inapte, très déformable CBR ≤ 3 EV2 ≤ 15
des ajouts, la nature minéralo-
P1 Ornières derrière l’essieu de 13 t, déformable 3 < CBR ≤ 6 15 < EV2 ≤ 30
gique, la forme et la texture des
granulats comme cela a été évoqué P2 Déformable 6 < CBR ≤ 10 30 < EV2 ≤ 50
Pas d’ornières
par CISSE [4]. Il serait hasardeux P3 derrière l’essieu Peu déformable 10 < CBR ≤ 20 50 < EV2 ≤ 120
de vouloir justifier cet ordre par P4
de 13 t
Très peu déformable 20 < CBR ≤ 50 120 < EV2 ≤ 200
un ou deux paramètres seuls.
Cependant, dans le cas des bétons
de sable à ajout de sable fillérisé de basalte, il est fort pro- tance P3, la dalle de béton compacté est réalisée directe-
bable qu’une bonne partie de la résistance provienne de ment sur le sol convenablement nivelé et compacté [10]
la densité du basalte et de l’activité de ses fillers qui ont dans le cas de trafic faible (T5). Dans le cas de trafic fort
un caractère pouzzolanique similaire à celui des cendres (T3), on interposera une couche de béton maigre dosée
volantes. D’ailleurs, l’ajout de basalte broyé aux bétons entre 150 Kg/m3 et 300 Kg/m3, soit 200 Kg/m3. Dans
compactés ordinaires a déjà été signalé [3]. On remarque ces conditions, l’épaisseur de la couche de roulement
même que le coefficient d’âge est plus faible à l’eau serait de 20 cm pour le trafic T5 [3] alors que pour le tra-
(1,13) qu’à l’air (1,22). C’est la preuve que la méthode de fic T3, l’épaisseur des couches de roulement et de fonda-
cure a moins contribué à faire accroître la résistance que tion seraient respectivement de 22 cm et 12 cm [10].
la densité et la réactivité pouzzolanique . S’agissant de la couche de fondation, un extrait de la note
d’information SETRA/décembre 1985 [3] stipule qu’elle
doit être de 15 cm au moins dans le cas des bétons com-
6. DIMENSIONNEMENT DES DALLES DE pactés. Ce béton maigre dosé à 200 Kg/m 3, donc de
BÉTONS DE SABLE COMPACTÉS classe de résistance 2, devrait avoir une résistance à la
traction par fendage supérieure ou égale à 1,7 MPa,
Le dimensionnement des chaussées en béton de sable comme cela est stipulé pour les bétons de sable maigres
est conduit selon la démarche définie par SETRA/LCPC pour couche de fondation [9]. Lorsque les résistances à la
[8]. Il se fait en fonction de la portance de la plate-forme traction par fendage sont différentes de celles considérées
et de la classe de trafic [9]. Les différentes classes de tra- dans les hypothèses du catalogue, comme c’est le cas
fics définies sont données au Tableau 9 où T0, T1, T2 et pour nos bétons, on détermine les épaisseurs correspon-
T3 sont classées fortes tandis que les autres sont considé- dantes de la manière suivante :
rées comme faibles. Dans l’optique d’une comparaison On sait que :
avec les chaussées en matériaux hydrocarbonés, nous fai-
sons les calculs pour deux classes de trafic : T3 (fort) et Rtflexion = Mv
I
T5 (faible). Les hypothèses de calcul sont les suivantes :
a) Voiries à fort trafic (> 50 poids lourds/jour) : avec : Rtflexion = résistance à la traction par flexion
– période de service : 20 ans, M = moment fléchissant correspondant à la rupture
– taux de croissance du trafic : 7% par an, v = distance entre la fibre neutre et le centre de gravité
– résistance à la traction par fendage : 3,3 MPa. de la section = h/2
b) Voiries à faible trafic (< 50 poids lourds/jour) I = inertie de la section = bh3/12
– période de service : 20 ans,
– taux de croissance du trafic : 4% par an, Rtflexion = M ⋅ 12 ⋅h = 6⋅M
– résistance à la traction par fendage : 2,7 MPa. b ⋅ h ⋅ 2 b ⋅ h2
3

Le catalogue définit également cinq classes de por-
tance du sol de plate-forme indiquées au Tableau 10. La
classe est désignée à partir du trafic moyen journalier
annuel « MJA », estimé à la mise en service et exprimé en
poids lourds (PL/j) de charge utile supérieure ou égale à
5 tonnes. Dans l’hypothèse d’une plate-forme de por-
On sait par ailleurs que les résistances à la traction par
fendage et celle en traction par flexion sont liées par la
relation suivante :
Rtfendage = α Rtflexion avec α = 0,6
Pour deux sections de même largeur b, subissant le

155
Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999

même moment M, la relation entre les épaisseurs est base par rapport au revêtement ;
obtenue en écrivant le rapport des résistances respec- – pour les chaussées en béton de sable, on envisage
tives, soit : une mince pellicule d’enduit bitumineux gravillonné
d’adhérence de 2 cm pour les circulations rapides ;
Rtf 1 6 M 1bh22 h22 – la distance de transport des sables des dunes est sup-
= =
Rtf 2 bh12 6 M2 h12 posée égale à 30 km en moyenne ;
– la distance de transport des sables fillérisés varie des
Finalement, on a : lieux d’extraction ;
– la main d’œuvre des chaussées en béton de sable est
Rtf 1 ⋅ h12 prise égale à 10% du coût d’investissement (la valeur
h2 = (2) réelle se trouvant entre 8% et 10%) ;
Rtf 2
– la couche de fondation est identique pour les deux
Par application de cette formule à nos différents types de chaussées en cas de trafic fort ;
bétons, nous aboutissons aux résultats indiqués aux – les coûts calculés (en francs CFA) ne prennent pas
Tableaux 11a et 11b. La modélisation des structures avec le en compte les frais éventuels d’entretien ultérieur qui
programme Alizé III du LCPC a été faite par Gaye [11] et sont quasi nuls pour les chaussées en béton et qui doivent
a fait aboutir à des épaisseurs de dalles variant entre 23 cm être programmés et budgétisés pour les chaussées en
et 28 cm. Enfin, pour l’organisation du chantier, toutes les matériaux hydrocarbonés ;
règles de l’art établies pour la fabrication, le transport, la – les paramètres non quantifiables tels que la sécurité,
mise en œuvre (répandage, régalage), le compactage, la le confort sont supposés identiques ; une étude appro-
cure, le compactage des bords, les dispositions construc- fondie permettrait de les estimer.
tives (joints), la protection de surface, etc., doivent être
respectées en vue d’un bon comportement en service. Ces
règles ont fait l’objet de recommandations SETRA-
LCPC et sont disponibles [3].
Tableau 12 – Prix des matériaux utilisés, au Sénégal
Matériaux Prix (F CFA)
Tableau 11a – Dimensionnement des dalles de béton
en trafic T3 Latérite (m3) 12 000
Ajout Calcaire Grès Silexite Basalte Transport au Km de 1 m3 47,63
0,3 0/3 0/3 0/3
Enrobés , MO comprise (t) 52 000
Rtf 28j (MPa) 2,11 1,91 2,63 2,30
Bitume fluidifié 0/1 (t) 320 000
Épaisseur 28 29 25 26
dalle (cm) Bitume gravillonné 8/16 (m3) 30 000
Bitume gravillonné 3/8 (m3) 30 000
Bitume gravillonné 0/3 (m3) 18 000
Tableau 11b – Dimensionnement des dalles de béton
en trafic T5 Ciment (t) 48 500
Ajout Calcaire Grès Silexite Basalte Eau (m3) 597
0,3 0/3 0/3 0/3
Sable des dunes (t) 662
Rtf 28j (MPa) 2,11 1,91 2,63 2,30
Calcaire 0/3 (t) 667
Épaisseur 20 21 18 19
dalle (cm) Grès 0/3 (t) –
Silexite 0/3 (t) –
Basalte 0/3 (t) 1818
7. ANALYSE ÉCONOMIQUE DE COÛT
Compte tenu de l’objectif de cette étude, il est néces- Tableau 13 – Quantités de matériaux utilisés
saire de vérifier par une analyse économique si les épais- pour chaussée souple, au Sénégal
seurs de dalle trouvées sont surabondantes ou non par Classe
Nature couche
Épaisseur Volume
rapport aux épaisseurs correspondantes en matériaux de trafic (cm) (litres/m2)
hydrocarbonés (enrobés denses). Base (enrobés) 20 –
T3
Roulement (enrobés) 4 –

7.1 Hypothèses de calcul Fondation (latérite) 30 –

Liaison (bitume fluidifié) 1
– La largeur des chaussées est prise égale à 6 m pour T5 8/16 – 12
les chaussées à trafic faible et 7 m pour celles à trafic fort Surface
(bitume 3/8 – 7
pour lesquelles on ajoutera une surlargeur de 40 cm gravillonné)
(prise en compte de l’effet de bord) pour la couche de 3/8 – 6

156
Cisse, Laquerbe, Gaye, Diene
7.2 Chaussées en enrobés denses
Les données relatives aux pr ix des matér iaux
(Tableau 12), aux épaisseurs des couches et volumes des
matériaux hydrocarbonés (Tableau 13) sont fournies res-
pectivement par l’entreprise Jean Lefebvre et le Bureau
de Gestion des Données Routières (1995) du Ministère
de l’Équipement et des Transports Terrestres. Les calculs
permettent d’aboutir pour 1 kilomètre de route, au coût
de 54 290 079 F CFA pour un trafic T3 et 45 636 000 F
CFA pour un trafic T5.
7.3 Chaussée en béton de sable compacté
Les prix des ingrédients indiqués au Tableau 12 sont
recueillis auprès des carrières et de l’usine productrice du
ciment. Tous calculs faits, on aboutit aux coûts au kilo-
mètre, mentionnés au Tableau 14, suivant le type d’ajout
utilisé. Nous donnons enfin au Tableau 15, le rapport
des coûts au kilomètre entre chaussées en béton de sable
compacté et chaussées en matériaux hydrocarbonés. Il
apparaît nettement que les chaussées en béton de sable
sont plus avantageuses à performances techniques équi-
valentes. En fort trafic, on peut réaliser au moins 1,54
fois plus de chaussées en béton alors qu’en faible trafic,
ce rapport minimum passe à 1,98. Il faut signaler que la
surlargeur non circulée de 40 cm a été surdimensionnée
puisque 25 cm seulement auraient suffi d’après le cata-
logue SETRA/LCPC et on aurait pu par conséquent
augmenter alors le rapport de coût.
Tableau 14 – Prix au kilomètre d’une chaussée en
béton de sable compacté suivant la classe de trafic
Classes Calcaire Grès Silexite Basalte
de trafic 0,3 0/3 0/3 0/3
T3 34 114 680 34 552 924 31 340 551 35 291 993
T5 21 961 396 21 676 338 21 235 544 22 960 268
Tableau 15 – Rapport de coût entre chaussée en béton de
sable compacté et chaussée en matériaux hydrocarbonés
Classes Calcaire Grès Silexite Basalte
de trafic 0,3 0/3 0/3 0/3
T3 1,59 1,57 1,73 1,54
T5 2,08 2,10 2,15 1,98
8. CONCLUSION
L’intérêt technico-économique des bétons de sable est
prouvé par l’étude comparative de coût ; ceci est très
important pour un pays comme le Sénégal, dans le
contexte actuel de dévaluation du franc CFA et de rareté
des produits hydrocarbonés. À cela, il faut ajouter que la
réalisation des chaussées en béton de sable nécessite une
main d’œuvre moins qualifiée et des équipements moins
sophistiqués tandis que celles en enrobés denses exige du
157
personnel technique qualifié et des engins indispensables
très onéreux. De manière générale, les bétons de sable rou-
tiers coûtent moins cher que les matériaux hydrocarbonés
puisqu’une étude comparative entre bétons de sable coulés
sur place et matériaux hydrocarbonés montre que les rap-
ports de prix varient entre 1,73 et 1,99 pour trafics faibles
et entre 1,70 et 2,05 pour trafics forts [4]. Aussi, la compa-
raison entre béton de sable coulé et béton de sable com-
pacté montre que les premiers sont plus compétitifs pour
les forts trafics alors que les seconds le sont pour des trafics
plus faibles. Assurément, le Sénégal et tous les pays bénéfi-
ciant des mêmes contextes géologiques et géotechniques
peuvent trouver dans les bétons de sable un moyen de pré-
server leur environnement naturel tout en augmentant la
valeur ajoutée dans la réalisation de leur réseau routier.
RÉFÉRENCES
[1] Diop, B., ‘Recherche d’un sable de substitution au sable normal
de Leucate’, Thèse de doctorat de géotechnique, I.N.S.A. de
Rennes, (1995).
[2] Laquerbe, M., Cisse, I. et Ahouansou, G., ‘Pour une utilisation
rationnelle des graveleux latéritiques et des sables des dunes
comme granulats à béton : application au cas du Sénégal’, Mater.
Struct. 28 (1995) 604-610.
[3] CIM béton, ‘Les bétons compactés’, Documentation technique,
Extrait de Routes N° 24 (décembre 1987).
[4] Cissé, I., ‘Contribution à la valorisation des matériaux locaux au
Sénégal : application aux bétons de sable’, Thèse de doctorat de
génie civil, I.N.S.A. de Rennes (1996).
[5] Gilbert, A., ‘Étude comparative des bétons à base de granulats
usuels en pays tropicaux (calcaire, basalte, grès, graveleux latéri-
tiques) avec sable de plage et sable des dunes’, Mémoire de fin
d’études d’ingénieur en génie civil, E.N.S.U.T., Université
Cheikh Anta Diop (Sénégal) (1993).
[6] Gaye, M. G., ‘Caractérisation et performance d’un granulat de
type nouveau en couche de chaussée : les silexites (produits
secondaires dans la chaîne de production de la C.S.T.P.)’,
Mémoire de fin d’études d’ingénieur géologue, I.S.T., Université
Cheikh Anta Diop (1995).
[7] Dupain, R. et al., ‘Caractérisation des matériaux de génie civil par
les essais de laboratoire : granulats, sols, ciments et bétons’, (Édi-
tions Casteilla, 1995).
[8] SETRA/LCPC, ‘Conception et dimensionnement des structures
de chaussées’, Guide technique, (1994).
[9] CIM béton, ‘Les bétons de sable en structures de chaussées’,
Documentation technique, Extrait de Routes, N° 53 (octobre
1995).
[10] CIM béton, ‘Conception et exécution du béton désactivé :
Voiries et aménagements urbains’, Collection technique, avril
1995.
[11] Gaye, A., ‘Détermination de la teneur en eau optimale dans
l’optique d’un béton de sable routier compacté’, Mémoire de fin
d’études d’ingénieur de génie civil, E.P.T., Université Cheikh
Anta Diop (1996).
 Materials and Structures/Matériaux et Constructions, Vol. 32, March 1999, pp 158-159

Materials and Technologies for Sustainable

RILEM BULLETIN

Construction
Reported by Christer Sjöström, Royal Institute of Technology, Sweden

INTRODUCTION
Sustainable development implies the use of
natural and physical resources that enables
people and communities to provide for their
immediate social, economic and cultural
wellbeing without compromising the ability
of future generations to meet their needs.
The construction industry must evidently
address certain consequential issues in the
process of achieving sustainable develop-
ment; it consumes considerable natural and
physical resources and as such has a signifi-
cant impact on the environment. Current
building design, engineering methods, man-
ufacturing technologies, and construction
techniques need to be altered to accommo-
date requirements for sustainable construc- Prof. A. Lucchini, Politecnico di Milano; Dr. Z. Lounis, NRCC; Dr. M. A. Lacasse,
tion. To help resolve the complexity of the NRCC.
problem, systematic approaches are
needed to address regulatory restrictions, SUMMARY were discussed, in particular those deal-
lack of innovation and of effective manufac- ing with cladding or products used to seal
turing technology and as well, the integra-
OF THE SYMPOSIUM or protect materials.
tion of construction technologies. Although systematic methods for service
This symposium has highlighted research Service life and durability life prediction have been proposed and
undertaken in the development of basic In the area of performance, durability and some of these implemented for specific
approaches for and dissemination of infor- service life of materials, components, materials, overall concepts that link in-ser-
mation on the design, maintenance, reuse assemblies or structures, contributions on vice conditions to laboratory evaluations,
or renewal and assessment of materials the use of innovative materials for the modelling and field-testing are still being
and technologies for sustainable construc- renewal of building systems and structures developed. Indeed, new approaches are
tion. Contributions have been prepared were offered and design for durability. As being sough to relate ‘real-life’ performance
within four areas including: Performance, well, papers related to predictive models for to that obtained from testing, evaluation
Durability and Service Life; Information service life are given and attention is and modelling. A series of papers focus on
Technologies; Life Cycle Analysis and brought to contributions on standards being this topic, in particular those dealing with
Maintenance Management, and; Environ- developed in the area of service life plan- stochastic methods and probabilistic
mental Technologies and Processes. ning and sustainable construction. approaches to service life design. The sig-
Provided below is a summary of the sympo- The majority of contributions fell in the nificance of durability or performance pre-
sium and brief reviews of some of the more area of materials; particular emphasis diction to the development of performance
significant contributions within the specific was placed on cementitious materials based codes is reviewed and the interrela-
areas mentioned above. and the durability and performance of tionship between product performance and
concrete. Recycled concrete and aggre- service life is examined.
gates and the use of industrial waste by- Standard development in the construction
products incorporated in concrete were sector is important in that it potentially
also prominent. Polymer based materials offers practitioners the means to implement

Editorial Note
This joint CIB/RILEM Symposium was one of the events organised during the CIB World Building Congress 1998 in Gävle, Sweden 7-12 June. As
the President of CIB, Prof. C. Sjöström was in charge of the organisation of this congress. The Executive Secretariat of CIB 1998 was the Division of
Materials Technology, Centre of Built Environment, Royal Institute of Technology, which is also a RILEM Titular Member.

1359-5997/99 © RILEM 158

consensus approaches to resolving design
and planning issues that otherwise would
be left in the research domain. In the area f
service life planning of buildings and con-
struction, considerable efforts have recently
been made to provide useful standards for
assessing and auditing the performance of
constructed facilities. These are reported in
the current proceedings as well as a pro-
posed standard on sustainable building.
The proposed standard would encompass
additional information required to integrate
approaches to service life planning to
include issues related to the environment.
Information technologies
Paramount in making informed decisions
about environmental issues related to sus-
tainable construction is the manner in
which information is collected, formatted
and structured. A series of contributions
focused on this area and standard methods
were suggested on structuring data for con-
ducting life cycle analysis, assessing the
design life of buildings or performing envi-
ronmental assessment of materials. There
is also an increased awareness of the
Internet as a useful tool for providing rele-
vant product information; in this instance
data formatting is, necessarily, paramount
for making information readily available.
Examples were provided on the use of infor-
mation technologies for, life cycle analysis
and assessment decision making and,
environmental technologies and processes.
In particular, software systems for manag-
ing the maintenance process as applied to
buildings or those offering life cycle simula-
tion were featured.
Special emphasis has focused on IT tools
for assessing the conservation state of his-
torically significant buildings as for example
provided in the Wood-Assess project. The
main objectives of this joint European
research project was to develop and vali-
date methods and technologies for the
proper assessment of the conservation
state of wood buildings and for mapping
and assessing environmental risk factors on
a macro, local and micro-scale. The work
consisted of developing three basic items:
assessment protocols for a field inspection
system; continuous monitoring device
(WETCORR) for the collection of tempera-
ture and moisture data on the surface and
within the wood; and, a system for mapping
the ‘rotting’ index for wood, based on exist-
ing climatic data, standards for assessing
the level of performance of the cladding
and, WETCORR measurements obtained in
the field. A PC-based geographical assess-
ment system was developed that allows for
maintenance management of historical
buildings made of wood (GISWood). A
schematic view of the integration of infor-
mation and information elements in pro-
vided in the accompanying figure.
Life cycle analysis
A series of contributions were made in this
session that offered a general overview of
the possible effects sustainable construc-
tion. For example, towards understanding
socio-economic changes and their implica-
tions on sustainable construction. Issues
related to building for environmental and
economic sustainability were also
addressed. The significance of mass and
energy flows in existing buildings and its
purported significance on sustainable
development were reviewed, as were the
energy implications of the transportation of
building materials.
Of particular interest were those contribu-
tions offering broad oversight towards rec-
onciling design requirements and environ-
mental needs. Integrating existing methods
of design and planning is of singular impor-
tance if sustainable construction is to move
from the research to practice. In this con-
text, efforts focused on the integration of life
cycle design of materials and structures
were particularly relevant and revealing.
The life cycle of a product or assembly
includes the manufacturing, construction,
use and maintenance, repair and renewal
as well as the demolition, recycling and dis-
posal of these items. In the context of sus-
tainable construction, planners and design-
ers are being asked to consider the effects
and implications of each of these stages
and their effect on the environment. Hence
integrated solutions must also incorporate
the means to select, assess and provide
insight into the consequence of planning
and design options in regards to energy
159
usage, resource depletion, and the creation
of environmental hazards over the life cycle
of a built facility (e.g. air, water and soil
contaminants). Presentations offered in this
and related sessions and have provided
useful examples of the integrated approach
in regards to product component and build-
ing design, as well as urban renewal.
Environmental technologies and
processes
Contributions to this session described pro-
jects involved in providing for either collect-
ing or disseminating data on the ramifica-
tion of selecting different building products
on the environment. Environmental assess-
ment methods, systems and protocols
advanced by different countries were also
addressed. In this session, significant con-
tributions were offered in the area of envi-
ronmental ‘impact’ of various materials and
components. More specifically, life cycle
assessment methods, analytical compar-
isons and models were described that were
used to assess the consequence of specific
building design choices on the environ-
ment. As well, the environmental ‘impact’
of buildings in regards to their use of energy
was presented. This session also touched
on technologies related to the recycling of
materials from waste products, expressly for
brick masonry and cementitious materials.
SUMMARY
The main themes on which the sympo-
sium focused helped address some of the
issues related to achieving sustainable
construction. Evidently the collectively of
the work suggests that this area is in the
incipient stages and minor but nonethe-
less significant contributions have been
made towards resolving these complex
problems. Maintaining and sustaining
the built environment through information
integration, assessment and analysis
techniques and the use of specifications
as well as service life models for building
materials, components and assemblies is
likely to be an involved yet compelling
challenge in the decades to come.
 INSTRUCTIONS TO AUTHORS
According to adjustments in the editorial policy for the journal Materials
and Structures, henceforth, papers submitted to the journal are
grouped into two main categories: Scientific Reports and
Technical Reports.
Scientific Reports are original scientific papers reporting the results
of research in the field of building materials and structures. The
review process and existing publication schedule for such papers
will remain the same, and these papers will be published in the
order of their acceptation. Scientific reports should not exceed 20
typewritten pages in length, illustrations included.
Technical Reports concern technical applications based on scientific
research to problems in the field of materials and structures, and
include the former categories of “Testing Methods and
Equipment”, “Applied Technical Papers”, “Technical Notes”.
Technical Reports must be scientifically-based practical papers of
immediate use to the reader. They must be short and to the point.
In order to bring useful information to readers in a timely manner,
the review and publication process of Technical Reports benefit
from priority treatment. The length limitation for such papers is 6-
8 typewritten pages, illustrations included.
A third category, Local Materials, also exists for submitted papers.
These papers report research, test methods or applications
concerning local materials.
Submissions
Securely protected manuscripts should be sent in triplicate to: RILEM
Publications, E.N.S., 61, av. du Président Wilson, F-94235
Cachan Cedex, France. Manuscripts and illustrations will not be
returned.
The manuscript
Electronic manuscript: If a paper is accepted, the authors shall provide a
disk containing the final text. Guidelines will be supplied by the
Editorial Secretariat.
Language and style: All submissions should be in English or French, typed
double-spaced on one side of the paper (type size 12; left, right, top
and bottom margins at least 2.5 cm). Papers are also required to be
numbered on the bottom of the page in order to ease the work of our
Scientific Committee. However specialised or technical the subject,
papers should be written in a brief and comprehensible manner.
Poorly written papers or those containing numerous grammar and
spelling errors will not be considered.
Presentation: Papers should include a bilingual abstract (English/French)
not exceeding 200 words. The main text of the paper should be
divided into numbered sections and subsections. Tables should be
headed with short titles. References should be numbered consecu-
tively in the text and listed at the end of the paper. Illustrations
should also be numbered consecutively as referred to in the text,
and captioned as Fig. 1, Fig. 2, etc. Illustrations will be reproduced
at column width (83 mm) and should be submitted in that format.
Units and Abbreviations: SI units must be used. Other units (inches,
etc.) may be placed in parentheses after the SI units, if desired. All
abbreviations should be spelt out in full on first appearance.
Copyright
Submission of an article to Materials and Structures is taken to imply that
it represents original, unpublished work, not under consideration
for publication elsewhere. By submitting a manuscript, the author
transfers copyright to RILEM.
For further information and detailed guidelines for preparing your
manuscript, please contact the Editorial Secretariat, RILEM
Publications, at the address above.
160
INSTRUCTIONS AUX AUTEURS
Conformément aux ajustements de politique éditoriale, les articles
soumis au journal Matériaux et Constructions sont désormais classés
dans deux catégories principales : Rapports Scientifiques et
Rapports Techniques.
Les Rapports Scientifiques présentent les résultats de recherches sur
les matériaux et les ouvrages. La procédure de lecture et la
programmation de la publication resteront les mêmes, et ces
articles seront publiés dans l’ordre de leur acceptation. Les rapports
scientifiques ne devront pas dépasser 20 pages dactylographiées
(illustrations comprises).
Les Rapports Techniques présentent des applications techniques basées
sur la recherche relative à des problèmes pratiques. Ils incluent les
anciennes catégories « Méthodes expérimentales et appareillage »,
« Technologie appliquée » et « Note technique ». Les Rapports
Techniques doivent avoir une base scientifique et fournir des
informations d’utilité immédiate. Afin de porter rapidement ces
informations à la connaissance des lecteurs, les procédures
d’évaluation et de publication des Rapports Techniques bénéficieront
d’un traitement prioritaire. Les rapports techniques ne devront pas
dépasser 6-8 pages dactylographiées (illustrations comprises).
Une troisième catégorie, Matériaux Locaux, existe également pour
les articles soumis. Ces articles présentent la recherche, des modes
opératoires ou des applications concernant des matériaux locaux.
Envoi
Les manuscrits, bien protégés, doivent être envoyés en 3 exemplaires à :
RILEM Publications, E.N.S., 61, av. du Président Wilson, F-94235
Cachan Cedex, France. Les manuscrits et leurs illustrations ne seront
pas renvoyés aux auteurs.
Manuscrit
Manuscrit électronique : Si un article est accepté, l’auteur fournira une
disquette contenant le texte définitif. Des directives seront
fournies par le Secrétariat de Rédaction.
Langage et disposition : Tous les articles soumis seront rédigés en anglais ou
en français, dactylographiés du coté recto du papier en interligne
double (taille des caractères 12 ; marges haut, bas, gauche et droite 2,5
cm). Les pages seront numérotées en bas de page afin de faciliter le
travail de notre comité scientifique. Les auteurs seront aussi concis
que possible sans nuire à la clarté et à l’intelligibilité. Un article mal
rédigé ou qui comporte de nombreuses erreurs de grammaire et
d’orthographe ne sera pas pris en considération.
Présentation : Les articles comprendront un résumé bilingue
(français/anglais) du contenu (pas plus de 200 mots). Les corps du
texte sera divisé en sections et sous-sections. Numéroter les
tableaux et les faire précéder d’un titre court. Les références seront
numérotées à la suite dans le texte et rassemblées à la fin de
l’article. Numéroter à la suite les illustrations et y faire référence
dans le texte et les légendes en tant que Fig. 1, Fig. 2, etc. Les
figures seront réduites à la largeur d’une colonne (83 mm) et
devront être soumises dans ce format.
Unités et abréviations : Utiliser les unités SI. Il est toléré de mentionner
entre parenthèses, après les valeurs exprimées en unités SI, les
unités non conventionnelles. Toute abréviation doit être exprimée
clairement la première fois qu’elle est utilisée.
Copyright
Soumettre un article à Matériaux et Constructions implique qu’il s’agit
d’un travail original, non publié, qui n’est à l’examen dans aucune
autre rédaction. En soumettant leur article, les auteurs
transmettent le copyright à la RILEM.
Pour obtenir un complément d’informations ou des directives pour la
préparation d’un manuscrit, veuillez contacter le Secrétariat de
Rédaction, RILEM Publications, à l’adresse ci-dessus.