ACI STRUCTURAL JOURNAL TECHNICAL PAPER

Title no. 104-S11

Experimental Investigation of Composite

Ultra-High-Performance Fiber-Reinforced
Concrete and Conventional Concrete Members
by Katrin Habel, Emmanuel Denarié, and Eugen Brühwiler

Composite ultra-high-performance fiber-reinforced concrete

(UHPFRC) and conventional reinforced concrete structural members
are investigated to assess the rehabilitation potential for existing
concrete structures. The composite structural response is
determined by testing 12 full-sized flexural beams, loading the
UHPFRC layer in tension. The results demonstrate that the
exceptional material properties of UHPFRC significantly improve the
composite member structural response, including the ultimate
force, stiffness, and cracking behavior. An analytical model is
developed to predict the composite UHPFRC and conventional
reinforced concrete structural response, and is employed to further
analyze the experimental test results.

Keywords: fiber-reinforced concrete; rehabilitation; stress-strain.

INTRODUCTION Fig. 1—Uniaxial tensile behavior: comparing UHPFRC,

Ultra-high-performance fiber-reinforced concrete (UHPFRC) conventional SFRC and conventional concrete.5
is an advanced cementitious material consisting of a dense, high-
strength matrix containing a large number of evenly should not significantly improve a structural element’s long-
embedded steel fibers. UHPFRC has exceptional mechanical term durability as compared to normal strength concrete.10-12
and transport properties including a very high tensile strength, Composite UHPFRC and conventional reinforced concrete
strain hardening, and a density leading to a very low perme- (RC) beams have been tested to ultimate flexure by Denarié
ability making it ideal for the rehabilitation and modification et al.13 In these test, the composite beams were comprised of
of existing structures.1-4 Regarding its tensile behavior, an UHPFRC overlay to replace the standard tensile rein-
UHPFRC belongs to the group of high performance fiber- forcing bar in an RC beam and exhibited an ultimate force
reinforced cementitious composites (HPFRCC), but offers the comparable to the standard RC beams. The composite beams
additional advantage of a very dense low-permeable matrix. exhibited an increased stiffness until the ultimate force was
Figure 1 shows the uniaxial tensile test results for UHPFRC, reached and, thereafter, a pronounced softening behavior,
conventional steel fiber-reinforced concrete (SFRC), and while the RC beams showed a slight hardening behavior
conventional concrete. The UHPFRC exhibits a significantly commonly observed in concrete flexural members.
increased tensile strength and strain-hardening behavior as UHPFRC has been used as bonded strips applied to the
tensile face to rehabilitate and improve existing reinforced
compared to other cementitious materials.5
concrete beams.14 Tested in four-point bending, the rehabil-
Thin HPFRCC overlays, between 10 and 50 mm (0.4 and itated composite beams behaved monolithically until frac-
2.0 in.), have been used to rehabilitate deteriorated cementitious ture. The composite beam ultimate force was equal to or
structures such as cracked pavements and bridge decks. The higher than the reference concrete member, but experienced a
overlay tensile strain-hardening behavior improves the original softening phase after reaching the ultimate force.
structure’s deformation and energy dissipation capacity. Thus, In summary, UHPFRC are promising materials for the
HPFRCC overlays are able to bridge existing cracks in the rehabilitation of existing concrete elements. The advantage
concrete substrate. The cracks in the HPFRCC layer remain offered by UHPFRC, as compared to HPFRCC, is their low
small, between 30 and 50 μm (1.1 × 10–3 and 2.0 × 10–3 in.), permeability that prevents the ingress of detrimental
and are closely distributed.6 Field studies with slurry infiltrated substances and should therefore significantly improve the
fiber concrete (SIFCON)7 have reported promising results durability of composite members.
including no damage or debonding in a thin 25 mm (1.0 in.) The objective of this study is to experimentally document
SIFCON overlay after 9 years of service. It has been argued and to develop an analytical modeling tool to predict and
that the small crack widths considerably diminish the
penetration of detrimental substances and thus diminish the ACI Structural Journal, V. 104, No. 1, January-February 2007.
member’s reinforcement deterioration.6,8,9 The detrimental MS No. S-2006-126 received March 22, 2006, and reviewed under Institute publication
policies. Copyright © 2007, American Concrete Institute. All rights reserved, including
substances, however, can still penetrate through the the making of copies unless permission is obtained from the copyright proprietors. Pertinent
discussion including author’s closure, if any, will be published in the November-
HPFRCC porous matrix and, therefore, the use of HPFRCC December 2007 ACI Structural Journal if the discussion is received by July 1, 2007.

ACI Structural Journal/January-February 2007 93

 analytical model is proposed for predicting the structural
Katrin Habel is a Postdoctoral Fellow at the University of Toronto, Toronto, Ontario,
Canada. She received her diploma in civil engineering from the University of response of UHPFRC-RC and similar composite elements.
Karlsruhe, Karlsruhe, Germany, and her doctoral degree from the Swiss Federal Institute
of Technology, Lausanne, Switzerland. Her research interests include the time-dependant
behavior and the static and dynamic response of structural members incorporating
EXPERIMENTAL PROGRAM
ultra-high-performance fiber-reinforced concrete and other cementitious materials. Overview
The behavior of composite UHPFRC-RC structural members
Emmanuel Denarié is a Senior Research Engineer at the Laboratory of Maintenance was investigated on twelve 5400 mm (17 ft, 8-1/2 in.) long
and Safety at the Swiss Federal Institute of Technology. He received his diploma in
civil engineering and his doctoral degree in civil engineering from the Swiss Federal flexural beams. The beams were composed of RC substrates
Institute of Technology. His research interests include concrete technology and ultra- and UHPFRC layers, as shown in Fig. 2. The RC substrate
high-performance fiber-reinforced concrete (UHPFRC).
represented an existing structural element and had two layers
Eugen Brühwiler is Professor at the Laboratory of Maintenance and Safety at the of longitudinal reinforcing bars and 8 mm (0.32 in.) diameter
Swiss Federal Institute of Technology. He received his diploma and doctoral degrees stirrups spaced at 200 mm (8 in.). The future UHPFRC-RC
in civil engineering from the Swiss Federal Institute of Technology in Zürich and
Lausanne, respectively. His research interests include the rehabilitation, improvement,
contact surface of the RC substrate was roughened with a high-
and safety of existing concrete structures. pressure water jet prior to casting the UHPFRC layer.
The composite beams were subjected to four-point bending
tests (Fig. 2) at an UHPFRC age of 90 days and an RC age of
further analyze the composite UHPFRC-RC beam structural more than 240 days. Prior to the fracture tests, the beam
response and behavior. The study is aimed to propose an structural behavior was studied in long-term tests under two
original method for efficient rehabilitation and strengthening different static systems and loading modes.16 These different
of beams and slabs of existing RC structures. A first on-site loading modes introduced residual stresses of approximately
application has been successfully performed on a road bridge.15 3 MPa (0.44 ksi) in the beams. Further details on this testing
can be found in Habel5 and Habel et al.16
RESEARCH SIGNIFICANCE This experimental program’s parameters were the UHPFRC
The exceptional properties of UHPFRC, including uniaxial layer thickness hU and the presence of steel reinforcing bars
tensile strain-hardening, high strength, and very low perme- in the UHPFRC layer As,U. The additional UHPFRC steel
ability make UHPFRC a promising material for application reinforcing bar reinforcement ratio corresponded to 2 Vol.-% of
in the rehabilitation field. However, their composite struc- the UHPFRC cross section. An overview of the experimental
tural behavior with conventional reinforced concrete is not parameters is given in Table 1. (The unusually large rein-
yet known. This paper presents full-scale composite UHPFRC- forcement in the compression chord of the beams was
RC beam experimental test results that validate the potential chosen to avoid macrocrack formation during the long-term
of UHPFRC for rehabilitation applications. Additionally, an creep tests prior to the fracture tests.)

Materials
The composition of the UHPFRC used in this study is
given in Table 2. It corresponds to the one developed by Rossi.3
The UHPFRC was neither heat- nor pressure-treated. The
normal strength concrete (NSC) had a water-cement ratio (w/c)
of 0.40. The mechanical properties of the two materials,
determined experimentally, are given in Table 3.
The UHPFRC tensile behavior was characterized by a
linear stress rise until the initial cracking strength fUt,1st
followed by strain hardening until the ultimate tensile
strength fUt,max at a strain of εUt,max = 2.8% (Fig. 3). The
strain hardening is expressed by the hardening magnitude
εU,hard. Then, softening occurred until complete fracture at a
crack width of w = 5 mm (0.20 in.).5
The reinforcing bars and stirrups used corresponded to the
European code requirements and had a yield strength of
500 MPa (72.5 ksi).

Test setup
The four-point bending tests were displacement-controlled
by imposing a displacement at one of the two hydraulic jacks
Fig. 2—Composite UHPFRC-RC beam test.

Table 1—Composite beam experimental parameters Table 2—UHPFRC composition

Constituent Type Weight, kg/m3(lb/yd3)
hu, mm As,U, mm2 As,ct, mm2 As,cc, mm2
Cement CEM I 52.5 N 1050 (1770)
Name No. (in.) (in.2) (in.2) (in.2)
Sand <0.5 mm (0.02 in.) 730 (1230)
NR3 3 30 (1-1/4) — 339 (0.53) 1005 (1.56)
NR5 3 50 (2) — 339 (0.53) 1005 (1.56) Silica fume Specific surface: 12 m2/g 275 (465)
R5 3 50 (2) 314 (0.48) 339 (0.53) 1005 (1.56) Steel fibers Straight (10 mm, 0.2 mm) 470 (790)
R10 3 100 (4) 616 (0.95) 339 (0.53) 1005 (1.56) High-range water-
Polycarboxylate 35 (60)
reducing admixture
Note: Beams without reinforcement bars in UHPFRC layers are called NR beams,
beams with reinforcement bars in UHPFRC layer are called R beams. Total water — 190 (320)

94 ACI Structural Journal/January-February 2007

located 300 mm (11.8 in.) from the beam ends, as shown in which approached a more fully restrained beam behavior.
Fig. 2. Both jacks were identical models and were connected The loading condition differences could not be clearly distin-
to the same hydraulic system and therefore applied the same guished for the NR5 beams and the residual stress differences
force F. The controlled jack displacement rate was 0.4 mm/ among the three beams were too small to observe significant
minute (0.016 in./minute). The deformations were measured differences in the structural response. Further details on the
with embedded optical fiber deformation sensors (Fig. 2, long-term tests can be found in Habel et al.16
ODS) and linear variable displacement transducers Localized macrocrack widths were recorded with Ω-gauges
(LVDT) (Fig. 2, f1 to f7). Crack formation was visually docu- attached to the tensile side of the beams (Fig. 5(a)). During
mented. Localized macrocrack widths were followed with Ω- the force rise, distributed macrocracks formed in the UHPFRC
gauges placed on the beam tensile face and were used to until a midspan deflection of approximately 4 mm (0.16 in.)
measure an average deformation over a 100 mm (3.92 in.) (l/600). This corresponded to an apparent UHPFRC strain-
length. (Ω-gauges are Ω-shaped steel pieces equipped with hardening of nearly 1‰, documented by the optical defor-
strain gauges to measure the displacement between two points mation sensor (ODS in Fig. 2) embedded in the UHPFRC
on the specimen surface, for example, adjacent to a crack.) layer. At the ultimate force, the distributed macrocracks
were spaced every 100 mm (3.9 in.) (Fig. 5(b)) and thus were
TEST RESULTS not influenced by the stirrups in the RC substrate. Then, the
Beams without reinforcing bars in UHPFRC layer width of one or two localized macrocracks, from the set of
(NR beams) distributed macrocracks in the UHPFRC layer, rapidly
The composite conventional RC substrate and UHPFRC increased. The post-failure crack pattern, for an example
layer without reinforcing bars beams, denoted with the NR5 beam, is shown in Fig. 5(b).
prefix NR, force-deflection curves are shown in Fig. 4. For Interface cracks started to form at a midspan deflection of
comparison, the RC substrate force-deflection curve is also 6 mm (0.23 in.). All interface cracks were initiated by a
given and was determined with a cross-sectional model, bending crack and no interface debonding occurred due to
considering the tensile and compression reinforcing bars and interface zone shear stresses prior to localized transverse
a parabola-shaped concrete compression zone. The RC beam cracking. The interface cracks then developed into a
crack spacing was assumed to be 200 mm (8 in.) in accor- debonding crack, and the composite beam structural
dance with the stirrup spacing. response deviated from monolithic behavior.
The NR composite beam structural behavior is characterized
by a disproportionately high increase in stiffness for midspan
deflections (deflection = f1 – 0.5 × [f6 + f7]) less than 20 mm
(0.79 in.) (l/120) as compared to the original RC section (Fig. 4)
and a positive correlation between the UHPFRC overlay
thickness and the stiffness increase was observed. The ultimate
NR beam force was close to the original RC section calculated
ultimate force. This is explained by localized macrocrack
formation in the UHPFRC at deflections beyond approximately
4 mm (0.16 in.). Moreover, the NR5 beam stiffness was
higher than the NR3 beam stiffness for deflections less than
10 mm (0.39 in.).
The NR5 beam test results showed good correlation, while
the NR3 beam results experienced more variation. This
scatter can be explained by different degrees of residual
stresses and microcracking in the UHPFRC layer prior to the
fracture tests.16 These different residual stress degrees can
be seen in the NR3 beam fracture test results, a composite
UHPFRC-RC beam with a 30 mm (1.2 in.) UHPFRC thick-
Fig. 4—Force-midspan deflection curves: Beams NR3 and
ness, which showed different maximum loads and different
NR5.
initial stiffnesses (Fig. 4). Furthermore, Beam NR3.1 had the
lowest residual stresses, followed by Beams NR3.2 and NR3.3,
Table 3—Mechanical properties at
28 days (mean values)
UHPFRC NSC
fUc = 168 MPa (24.4 ksi) fcc = 51 MPa (7.4 ksi)
Compressive
strength (cylinder 11/22 cm (cylinder 16/32 cm
[(4-1/4)/(8-3/4) in.]) [(6-1/4)/(12-1/2) in.])
EU = 48 GPa (6962 ksi) Ec = 38 GPa (5511 ksi)
Modulus of
elasticity (cylinder 11/22 cm (cylinder 16/32 cm
[(4-1/4)/(8-3/4) in.]) [(6-1/4)/(12-1/2) in.])
fUt,1st = 9.1 MPa (1.3 ksi)
First cracking
(notched plates 5 x 20 cm —
strength
[2 x 8 in.])
fUt,max = 11.0 MPa (1.6 ksi) fct = 3.4 MPa (0.5 ksi)
Tensile
strength (notched plates 5 x 20 cm (notched plates 5 x 20 cm
[2 x 8 in.]) [2 x 8 in.])
Fig. 3—Modeling UHPFRC tensile behavior: (a) continuum Hardening εU,hard = 2‰ —
phase; and (b) discrete crack model. magnitude

ACI Structural Journal/January-February 2007 95

Beams with reinforcing bars in UHPFRC layer Closely distributed macrocracks, spaced every 30 mm
(R beams) (1.2 in.) formed preceding the maximum force (Fig. 7(b)).
Six composite UHPFRC-RC beams were tested with The apparent UHPFRC strain-hardening was deduced from
reinforcing bars embedded in the UHPFRC layer (Fig. 6) and optical fiber deformation sensors embedded in the
are denoted with the prefix R. The addition of reinforcing bars UHPFRC layer and approximately corresponded to a 3‰
in the UHPFRC layer increased the ultimate force by two deformation—three times larger than for the NR beams
times for R5 beams (hU = 50 mm [2.0 in.]) and by five times (Fig. 7(a)). This strain-hardening performance was due to
for R10 beams (hU = 100 mm [3.9 in.]) as compared to the the tension-stiffening effect introduced by the reinforcing
respective NR beams (Fig. 6). A comparison with a comparable bar embedded in the UHPFRC layer (As,U) that produced a
conventional RC beam showed a significant increase in more homogenous deformation distribution. Then, up to four
stiffness, as already observed for the NR beams, and an localized macrocracks remained active at a midspan deflection
increase in the ultimate force. The ultimate force was higher of approximately 15 mm (0.59 in.) (l/160), as documented
in the R10 beams than in the R5 beams due to the larger with the Ω-gauges (Fig. 7(a)).
member depth, the larger number of reinforcing bars, and the The first interface cracks formed at a deflection of 17 mm
larger UHPFRC layer thickness. (0.67 in.). Thereafter, the interface crack opening remained
constant at approximately 0.5 mm (0.02 in.) beyond a deflection
of 20 mm (0.78 in.). These interface cracks were small when
compared to the NR beam debonding cracks. In combination
with visual observations during testing, it can be deduced
that final beam fracture occurred in a vertical flexure crack.
This phenomenon is explained by the tensile stress trans-
ferred across the active macrocracks by the reinforcing bar
embedded in the UHPFRC layer. Consequently, less stress
had to be transferred through the interface. All interface
cracks were initiated by flexural cracks that had intersected
the interface. No other interface cracks were observed.

BENDING ANALYTICAL MODEL

General
It is proposed to determine the composite UHPFRC-RC
flexural behavior with an analytical cross-sectional model.

Fig. 5—(a) Ω-gauge experimental results for NR5 beams;

and (b) typical post-failure crack pattern of NR5 beams
(cracks are redrawn for clarity).

Fig. 7—(a) Ω-gauge experimental results for R5 beams; and

(b) typical post-failure crack pattern of R5 beam (cracks are
Fig. 6—Force-deflection curves: Beams R5 and R10. redrawn for clarity).

96 ACI Structural Journal/January-February 2007

The moment-curvature relationship is computed for a cross assessment. Force-deflection curves are derived by considering
section with a tensile UHPFRC layer (Fig. 8) in accordance geometry and the structural element’s static system.
with the tested members. The analysis approach is based on
the bending beam theory and is an extension of the commonly DISCUSSION OF STRUCTURAL RESPONSE
employed cross-sectional model for the determination of a Method
reinforced concrete element load bearing capacity by The composite UHPFRC-RC beam testing was simulated
considering the UHPFRC and concrete tensile behaviors. with the analytical model. The vertical forces applied in the
four-point bending test were calculated by transforming the
Hypotheses analytically computed moments (F = M/1.2 m [3.9 ft]). The
The model for bending is based on the following hypotheses: midspan deflection was determined by assuming a constant
1. Plane sections remain plane (Bernoulli’s hypothesis)— curvature over the entire beam midspan. This assumption
the bond between cementitious materials and reinforcing was accurate prior to the formation of localized macrocracks.
bars remains intact, no debonding occurs at the composite Thereafter, the experimental results showed localization of
interface and the structural element behaves monolithically; deformation into one or several localized macrocracks, and
2. The reinforcing bar and cementitious material stress- this constant curvature assumption induced small deflection
strain behaviors are described by material laws; and calculation errors. This curvature simplification was made
3. Equilibrium of forces and moments is achieved in the because there was insufficient experimental evidence to
cross section. determine the plastic hinge lengths that formed at the localized
macrocracks. The transition from hardening to softening is
Description indicated in Fig. 11 and 12 by the transition from solid to
The UHPFRC tensile behavior is modeled in two parts: for dotted lines.
strains smaller than εUt,max, the material is considered as a The conventional concrete and the steel reinforcing bar
bilinear continuum (Fig. 3(a)) and beyond εUt,max, a fictitious material laws were directly obtained from material tests.5
crack formation with an opening w is assumed following the The UHPFRC tensile properties were determined by
bilinear softening law (Fig. 3(b)). The crack opening w is uniaxial tensile tests performed on notched specimens.5
transformed into strain ε by using the reference length LR However, the question of whether the uniaxial tensile test
formula as detailed in Fig. 3(b).17 results could be directly converted into a material law,
Conventional concrete compression is modeled with a particularly because the ultimate tensile tests were performed
parabolic distribution (Fig. 9(b))18 and the tensile behavior is on structural elements and may have to be interpreted
modeled with a linear-elastic stress rise and a bilinear softening accordingly, had to be investigated.
diagram, as proposed by Hillerborg19 (Fig. 9(a)). Two approaches were chosen to ascertain the UHPFRC
A bilinear material law is symmetrically assumed for the tensile behavior:
tensile and compression steel reinforcing bars (Fig. 10). The • Approach I: the uniaxial tensile test results were used
model can also consider different material properties for the as input parameters for the analytical model and analyt-
UHPFRC and the concrete substrate reinforcing bars. ical results were compared to the beam experimental
The flexural UHPFRC-RC strain and stress state is calculated results; and
with the analytical model given in Fig. 8. The linear strain • Approach II: the beam experimental results were fitted
distribution and individual material laws facilitate the curvature by inverse analysis (by adapting the UHPFRC layer
κ and the cross-sectional stress calculations. tensile properties) and the determined UHPFRC tensile
The normal force equilibrium is determined through iteration
by varying the cross-sectional strain distribution over the depth
for a given maximum tensile strain εU,l. The moment is
calculated from this equilibrium. The curvature variation
allows the cross-sectional moment-curvature relationship

Fig. 9—Concrete material laws: (a) tension (ss = stirrup

spacing = 200 mm [7.87 in.]); and (b) compression.

Fig. 8—Analytical model definitions. Fig. 10—Steel reinforcement bar stress-strain relationship.

ACI Structural Journal/January-February 2007 97

 properties were compared against the uniaxial tensile In Approach I, that is, UHPFRC behavior calculated from
tests results. uniaxial tensile test inputs, the analytical model overestimated
The composite beam structural response discussion is the NR5 beam force by a maximum of 30%. Localized
based on the experimental results and the interpretation by macrocracks were observed at a deflection of 4 mm (0.16 in.)
the analytical model. during the experimental testing, while the model predicted
macrocrack formation at 13.5 mm (0.53 in.). The NR3 beam
NR beams force was within the experimental scatter, however, the
Figure 11 shows the comparison between the test results analytically obtained macrocrack localization deflection was
for beams without reinforcing bars in the UHPFRC layer also significantly overestimated.
(NR3 and NR5) and the analytically determined UHPFRC Considering Approach II, the experimental beam test results
tensile behavior. Only the UHPFRC hardening phase is were fitted by adapting the UHPFRC tensile behavior. The
adapted; the softening phase has not been changed. resulting strength values fUt,1st and fUt,max were approximately
33% smaller than the mean experimental values. The tensile
behavior approximation required a secant modulus that was 50%
lower than the secant modulus observed in the uniaxial tensile
tests. This resulted in an underestimation of the initial beam
stiffness and indicated UHPFRC tensile non-linearity behavior
prior to the deformation εUt,1st. Additionally, the hardening slope
was steeper than observed in the uniaxial test results. The fUt,1st
scatter for the different beams was 1 MPa (0.15 ksi) and the
strain εUt,max ranged from 0.75 to 1‰. An inverse correlation
between the strength values fUt,1st and fUt,max and the UHPFRC
layer thickness was observed—the thicker the UHPFRC layer,
the lower the UHPFRC strength values.
The significant difference between the uniaxial UHPFRC
tensile behavior and the tensile behavior determined by inverse
analysis from the beam tests may have several sources:
• Most importantly, the apparent UHPFRC layer’s lowered
secant modulus and increased hardening modulus
EU,hard, obtained in the inverse analysis, may have been
effected by residual stresses and microcracks introduced
in the UHPFRC layer by the composite element’s time-
Fig. 11—(a) Analytical modeling: NR3 beams; (b) analytical dependent behavior during the 90 days prior to the
modeling: NR5 beams; and (c) UHPFRC tensile behavior used fracture test. This would also explain the UHPFRC tensile
for modeling (UHPFRC hardening phase [ε ≤ εUt,max] and non-linearity behavior for strains smaller than εUt,1st;
softening phase [ε > εUt,max] contributions are respectively
drawn in solid and dotted lines in Fig. 11(a) and (b)).

Fig. 12—(a) Analytical modeling: R5 and R10 beams;

(b) UHPFRC tensile behavior used for modeling (UHPFRC
hardening phase [ε ≤ εUt,max] and softening phase [ε > Fig. 13—(a) Composite beam structural response phases
εUt,max] are respectively drawn in solid and dotted lines in (for example, Beam R10); and (b) cracking pattern during
Fig. 12(a) and (b)). Stages II and IV: Beams NR and R.

98 ACI Structural Journal/January-February 2007

• The uniaxial tensile test propensity to overestimate the strain in the UHPFRC layer at the formation of localized
UHPFRC layer tensile properties may have been due to macrocracks εUt,max was equal to or higher than the strain
the concentrated fracture process inherent in the indicated by the tensile tests. The R5 and R10 force-deflection
notched uniaxial tensile test. By introducing a notch, curves were correctly modeled with the modified input values
fracture was induced at a specific location rather than at with a variance less than 5% and the formation of localized
the weakest section within an entire element, as was the macrocracks was predicted within the experimental scatter.
case in the four-point bending tests; The difference between the UHPFRC tensile behavior
• The inverse correlation between the UHPFRC layer obtained from the uniaxial tensile tests and determined by
thickness and UHPFRC tensile strength, as determined by inverse analyzing the beam results is explained by the same
inverse analysis, may be caused by slight fiber segregation mechanisms as in the NR beams. The difference in the
in the UHPFRC layer. This thesis is supported by the hardening magnitude, however, seems to be smaller for the
observation of fewer fibers near the UHPFR top surface R beams. The deformation at the maximum tensile strength
and the strength reduction being more pronounced in (at macrocrack formation) was εUt,max = 2.8‰ in the
thicker UHPFRC layers; and uniaxial tensile tests and εUt,max = 2.8 and 3.4‰ by inverse
• The analytical model underestimated the original stiffness. analysis for the R5 and R10 beam tests, respectively. The
This result may have been a product of the model’s residual stresses in the R beams were comparable to the NR
consideration of only a bilinear stress distribution during beams; however, the UHPFRC layer reinforcing bar tension
the homogenous phase (prior to the formation of localized stiffening effect in the R beams produced an increase in the
macrocracks) even though the bilinear model was apparent hardening magnitude. This tension stiffening effect
deemed sufficiently accurate to model the uniaxial appeared to compensate for the given configuration’s
UHPFRC tensile behavior. residual stress-strain hardening loss.
The differences between the uniaxial tensile test results The UHPFRC layer reinforcing bars As,U can be seen as
and the tensile properties determined by inverse analysis macroscale “reinforcing fibers” as compared to the small
from the beam tests ascertain that the uniaxial tensile material 10 mm (0.39 in.) long steel fibers embedded in the
test results cannot be directly applied to structural elements. UHPFRC. The steel fibers act on the material level bridging
This direct application denial was mainly due to residual
microcracks and limiting microcrack propagation. When
stresses induced in the composite UHPFRC-RC beams and
macrocracks formed, the steel fibers continued to bridge the
because the uniaxial tensile tests must also be considered as
cracks until the crack width became too large and the fibers
structural tests. Translation of uniaxial tensile test results to
were gradually pulled out. Pull-out was completed when the
tensile performance in structural elements must thus be
crack width equaled approximately one-half of the fiber
investigated further.
length, 5 mm (0.20 in.), for the used UHPFRC, and, as a
Pre-peak responses of interface cracks and localized
result, no more force was transferred across the crack by the
debonding were observed in the NR beams at deflections
steel fibers. At this point, the reinforcing bar then bridged the
larger than 6 mm (0.23 in.). After the formation of these
tensile force over the large macrocracks. In contrast to the
cracks, the analytical model’s second hypothesis (composite
steel fibers, the reinforcing bar was not pulled out, but
member monolithic behavior) was no longer fully respected.
yielded and finally broke, leading to complete failure of the
At this stage, however, the localized flexural crack widths
structural element. During force rise, the reinforcing bar also
were already larger than 0.5 mm (0.02 in.) and the UHPFRC
contributed to the deformation distribution. Thus, closely
had already entered the softening phase. Occurrence of
debonding was considered in the analytical model through distributed macrocracks formed and macrocrack localization
the fitted reference length (LR) parameter, which transformed occurred at a higher deflection (l/160) in the R beams as
the crack width into deformation. The reference length is an compared to the NR beams (l/600). The increase in the
empirical parameter and a product of the given material and apparent hardening magnitude underlines that the reinforcing
structural properties. The reference length is not yet bar beneficial effect not only applies to the ultimate resistance
precisely known for composite UHFPRC-RC members and, but also to the composite UHPFRC-RC element crack
therefore, in this study, the analytical results were fitted to the formation because the UHPFRC hardening behavior was
experimental data to obtain a reference length of LR = 500 mm apparently regulated.
(19.6 in.) for the NR3 beams and to LR = 600 mm (23.6 in.) The reference lengths were fitted to the experimental results
for the NR5 beams.5 Additional research is needed to more producing LR = 1500 mm (59 in.) and LR = 2000 mm (118 in.),
accurately determine the reference length. respectively, for beams R5 and R10. These reference lengths
are significantly larger than for the NR beams. This indicates
R beams that, beyond the formation of localized macrocracks, the
The Approach I results, (UHPFRC layer behavior input deformations were localized over a shorter length in the R
deduced from uniaxial tensile tests) in the R5 and R10 beams, beams than in the NR beams and, therefore, the R beam
detail the mean UHPFRC tensile behavior values overestimated plastic hinge lengths were shorter.
the applied force by a 7% difference at a deflection of 15 mm The R beams, in some cases, exhibited softening behavior
(0.59 in.) (Fig. 12). The formation of localized macrocracks beyond the maximum force. This is explained by the
at deflections between 13 and 17 mm (0.51 and 0.67 in.) was pronounced UHPFRC tensile behavior, for the force was still
also predicted by the analytical model. transferred through the UHPFRC layer as the reinforcing
The apparent UHPFRC tensile behavior determined with bars As,U yielded. The decreased stress transfer across the
Approach II (fitting the beam test results by adapting the UHPFRC localized macrocracks was larger than the steel
UHPFRC tensile behavior) shows the beam UHPFRC reinforcing bar hardening stress increase. The occurrence of
apparent secant modulus, strength, and hardening modulus softening or hardening in the beams depended on the
EU,hard were, respectively, reduced by 50, 33, and 33%. The UHPFRC layer thickness. For the tested beams, this softening

ACI Structural Journal/January-February 2007 99

behavior was observed in the R10 beams, but not in the showed distributed macrocracks (width < 50 μm [2 × 10–3 in.])
R5 beams. before developing into the post-failure cracking pattern
For the R beams, the interface cracks remained small and (Fig. 5(b)). The R beams showed closely distributed macro-
did not alter the structural behavior. The tested beams all cracks before the formation of localized macrocracks at post-
failed in bending under monolithic cross-sectional behavior. peak (Fig. 7(b)). The R beam experimental results indicate
Thus, the hypotheses of the analytical model are respected. an average distributed macrocrack spacing of 200 and 30 mm
(7.8 and 1.2 in.) at a deflection of l/600 and l/160, respectively,
Description of structural response just before the formation of localized macrocracks.
The experimental test and analytical model results for the The composite beam deformations, at the point of localized
beam structural response without reinforcing bars in the macrocrack formation, were dependent on the UHPFRC
UHPFRC layer (NR) can be described in five stages (Fig. 13(a)). tensile behavior. The decisive parameter was the hardening
Stage 1—At low forces, the beam behaved quasi-linear- magnitude εUt,max for a given beam configuration. Therefore,
elastically and no microcracks developed in the UHPFRC. as the hardening magnitude εUt,max increased, the macrocrack
Stage 2—Microcracks formed and developed into closely- spacing reduced and the beam deformations at the point of
spaced small-width macrocracks (< 50 μm [2.0 × 10–3 in.]). The localized macrocrack formation increased. The formation of
UHPFRC deformations can be simulated as homogeneously localized macrocracks in the RC beams is governed by the
distributed over the beam (Fig. 13(b)). tensile behavior of concrete, which, as compared to
Stage 3—Localized macrocracks developed from the UHPFRC (concrete: GF = 120 J/m2 [0.7 lbf/in.], UHPFRC:
closely-spaced macrocracks at a midspan deflection of l/600 GF = 20,200 J/m2 [116.4 lbf/in.]5), exhibits a very low defor-
with spacing larger than 100 mm (3.9 in.) for NR beams and mation capacity. The first concrete macrocracks occurred at
at a deflection of l/160 with spacing larger than 30 mm (1.2 in.) small deflections as the concrete tensile strength was
for the R beams. When the localized macrocracks inter- reached, corresponding to the end of the RC bending beam
sected the interface, interface cracks developed at the level theory State I. All cracks in the RC beams were localized
of the concrete substrate tensile reinforcing bars located macrocracks and opened gradually with increasing defor-
near the interface. mation. In contrast, the UHPFRC macrocracks of small
Stage 4—The force transmitted through the macrocracked openings were finely distributed under service conditions.
UHPFRC layer decreased due to the UHPFRC softening The crack formation differences also explain the
behavior, however, the overall force in the beam increased, for composite UHPFRC-RC element’s high stiffness under
additional force was gradually transferred through the inter- service conditions, for the UHPFRC distributed macrocracks
face and into the concrete layer tensile reinforcing bars As,ct , in strain-hardening domain did not produce a sudden
which had not yet yielded. For the NR beams, the interface decrease in stiffness as would be the case in the RC beams.
crack developed into localized debonding. The debonding
After tensile reinforcing bar yielding, the underreinforced
crack length and width increased as the localized flexural
RC beams exhibited approximately equidistant-spaced
macrocrack width increased. Secondary bending macroc-
similar-sized macrocracks, coinciding with the stirrup
racks then formed in the concrete layer and the concrete
spacing. The beam structural response was governed by the
layer tensile reinforcing bars started to yield forming plastic
tensile reinforcing bar and the compression concrete response.
hinges (Fig. 13(b)).
Stage 5—The composite beam failed by fracture of the In contrast, the composite UHPFRC-RC beams showed
tensile reinforcing bars followed by crushing of the concrete. closely distributed macrocracks; however, plastic deformations
Local debonding only occurred in the NR beams from were concentrated in few, or even just one, localized
Stage 4 on because more force had to be transferred across macrocracks. The coexistence of these two macrocrack
the interface when a localized macrocrack in the UHPFRC types was caused by the UHPFRC tensile hardening domain
layer opened than for the R beams. The slight interface and the material inhomogeneities. The UHPFRC resistance
cracking observed in the latter did not alter the composite gradually diminished at the localized macrocracks as
structural response. Several macrocracks formed in the NR UHPFRC softening occurred. Simultaneously, the UHPFRC
beam concrete layer at the debonded zone, extending the plastic at other locations was still within the hardening domain and
hinge length and increasing the beam deformation capacity. exhibited only distributed macrocracks. Therefore, reinforcing
Debonding always occurred at the tensile reinforcing bar bar yielding was localized at a few plastic hinges as could be
As,ct level and not at the UHPFRC and concrete contact observed from the tested structural members subjected to a
surface confirming the hydrojetted contact surface and intra- constant bending moment. As a result, the overall composite
material bond quality. This weak horizontal plane in the UHPFRC-RC beam deflection in pure flexure may be
concrete section was due to the presence of the reinforcing slightly smaller than the RC beams in pure flexure.
bars. In the case of the composite beams, this plane was near Underreinforced RC beams exhibited hardening behavior
the contact surface and the debonding crack weakened the until fracture of the reinforcing bars, whereas the composite
bond between reinforcing bars and concrete. The reinforcing UHPFRC-RC beams, in some cases, exhibited softening
bars were still anchored in sound concrete beyond the after the ultimate force. The occurrence of softening or
debonded zones and worked as external reinforcement in the hardening in the composite UHPFRC-RC elements, whether
debonded zone. the force decreases or increases following tensile reinforcing
bar yielding, depends on the UHPFRC tensile properties, the
Comparison of composite UHPFRC-RC elements UHPFRC layer thickness, and the reinforcing bar and
with RC elements UHPFRC interaction. For the investigated composite beams,
Under service conditions, conventional RC beams exhibit hardening occurred in the NR3, NR5, and R5 beams and soft-
bending macrocracks spaced at approximately 200 mm (7.8 in.), ening occurred in the R10 beams. The R10 beam softening
coinciding with the stirrups. The NR beam UHPFRC layer behavior was attributed to the large UHPFRC thickness.

100 ACI Structural Journal/January-February 2007

These results support the use of rather thin (maximum of 50 mm fsy,U = UHPFRC reinforcing bar yield strength
[2 in.]) UHPFRC layers. fUc = UHPFRC compressive strength
fUt,1st = UHPFRC initial cracking strength
fUt,max = UHPFRC tensile strength
CONCLUSIONS GF = specific fracture energy
The composite UHPFRC-RC element structural response hU = UHPFRC layer thickness
experimental study and the analytical modeling results LR = reference length
l = composite member span
illuminate the following conclusions: M = moment
1. Applying a UHPFRC layer to form a composite wUt = UHPFRC crack width
UHPFRC-RC element increases the service condition stiffness, εint = composite member interface strain
minimizes deformations for given imposed forces, reduces εU,hard = UHPFRC hardening magnitude
crack widths and crack spacing, and delays the formation of εU,l = UHPFRC composite member extreme fiber strain
εUt,1st = UHPFRC strain at initial cracking strength
localized macrocracks as compared to the original conven- εUt,max = UHPFRC strain at tensile strength
tionally reinforced concrete element. This improved σs,ct = UHPFRC reinforcing bar stress
performance is attributed to the UHPFRC layer high tensile σs,U = concrete tensile reinforcing bar stress
strength and strain-hardening properties;
2. Composite UHPFRC-RC elements behaved monolithically REFERENCES
under service conditions. Interface cracks developed only 1. Bache, H. H., “Introduction to Compact Reinforced Composite,” Nordic
Concrete Research, No. 6, 1987, pp. 19-33.
once localized flexural macrocracks had propagated through 2. Dugat, J.; Roux, N.; and Bernier, G., “Mechanical Properties of Reactive
the UHPFRC layer and intersected the interface zone. In Powder Concretes,” Materials and Structures, V. 29, No. 188, 1996,
general, this occurred near the ultimate force, far beyond pp. 233-240.
normal service loading. The interface cracks developed into 3. Rossi, P., “Development of New Cement Composite Material for
Construction,” Innovations and Developments in Concrete Materials and
localized debonding cracks for composite elements without Construction, R. K. Dhir, P. C. Hewlett, and L. J. Csetenyi, eds., Dundee,
reinforcing bars in the UHPFRC layer (NR beams). Interface Scotland, Sept. 2002, pp. 17-29.
cracks remained sufficiently small and did not cause 4. Charron, J.-P.; Denarié, E.; and Brühwiler, E., “Permeability of UHP-
FRC under High Stresses,” Advances in Concrete Through Science and
UHPFRC layer debonding in elements with reinforcing bars Engineering, RILEM, Evanston, Ill., 2004, 12 pp. (CD-ROM)
in the UHPFRC layer; 5. Habel, K., “Structural Behaviour of Elements Combining Ultra-High
3. Composite UHPFRC-RC element stiffness and resistance Performance Fibre Reinforced Concretes (UHPFRC) and Reinforced
was further increased when reinforcing bar was embedded in Concrete,” PhD thesis No. 3036, Swiss Federal Institute of Technology,
Lausanne, Switzerland, 2004, 196 pp.
the UHPFRC layer. A 2 Vol.-% of reinforcing bars were 6. Krstulovic-Opara, N., and Toutanji, H., “Infrastructural Repair and
embedded in the UHPFRC layer and increased the Retrofit with HPFRCCs,” High Performance Fiber Reinforced Concrete
composite element’s apparent hardening magnitude by three Composites 2 (HPFRCC 2), A. E. Naaman and H. W. Reinhardt, eds.,
RILEM, Ann Arbor, Mich., 1996, pp. 423-442.
times and significantly delayed the formation of localized 7. Schneider, B., “Development of SIFCON through Applications,” High
macrocracks. In the UHPFRC softening domain, the force Performance Fiber Reinforced Cement Composites (HPFRCC), H. W. Rein-
transfer through the reinforcing bar enhanced the composite hardt and A. E. Naaman, eds., RILEM, Stuttgart, Germany, 1992, pp. 177-194.
element structural response by preventing debonding; 8. Li, V. C.; Horii, H.; Kabele, P.; Kanda, T.; and Lim, Y. M., “Repair
and Retrofit with Engineered Cementitious Composites,” Engineering
4. The composite UHPFRC-RC element structural response Fracture Mechanics, V. 65, No. 2-3, 2000, pp. 317-334.
under flexure was predicted by an analytical model. The 9. Krstulovic-Opara, N.; Dogan, E.; Uang, C.-M.; and Haghayeghi,
analytical model provided an efficient tool for rapidly deter- A. R., “Flexural Behavior of Composite R.C.-Slurry Infiltrated Mat
mining the composite UHPFRC-RC element structural Concrete (SIMCON) Members,” ACI Structural Journal, V. 94, No. 5,
Sept.-Oct. 1997, pp. 502-512.
response. The model was validated with the composite beam 10. Kosa, K.; Naaman, A. E.; and Hansen W., “Durability of Fiber Reinforced
fracture test results; and Mortar,” ACI Materials Journal, V. 88, No. 3, May-June 1991, pp. 310-319.
5. The apparent UHPFRC beam tensile properties, calculated 11. Lemberg, M., “Dichtschichten aus hochfestem Faserbeton,” Deutscher
Ausschuss für Stahlbeton, (DAfStb), Beuth Verlag GmbH, No. 465, Berlin,
by inverse analysis, were significantly lower than the notched Germany, 1996, pp. 1-163. (in German)
UHPFRC uniaxial tensile test properties. The UHPFRC 12. Li, V. C., and Weimann, M. B., “Hygral Behavior of Engineered
tensile behavior differences in the composite beams and in Cementitious Composites (ECC),” International Journal for Restoration of
the uniaxial tensile tests are mostly attributed to residual Buildings and Monuments, V. 9, No. 5, 2003, pp. 513-534.
13. Denarié, E.; Habel, K.; and Brühwiler, E., “Structural Behavior of
stresses induced in the composite structural member during Hybrid Elements with Advanced Cementitious Materials (HPFRCC),” High
long-term tests conducted before the fracture tests. Performance Fiber Reinforced Cement Composites (HPFRCC-4), A. E. Naa-
man and H. W. Reinhardt, eds., RILEM, Ann Arbor, Mich., 2003, pp. 301-312.
14. Alaee, F. J., and Karihaloo, B., “Retrofitting of Reinforced Concrete
ACKNOWLEDGMENTS Beams with CARDIFRC,” Journal of Composites for Construction, ASCE,
This project has been financially supported by the Swiss Federal Office for V. 7, No. 3, 2003, pp. 174-186.
Education and Science (OFES) in the context of the European project 15. Habel, K.; Denarié, E.; and Brühwiler, E., “Time Dependent Behavior of
Sustainable and Advanced Materials for Road Infrastructures (SAMARIS). Elements Combining Ultra-High Performance Fiber Reinforced Concretes
The authors would like to thank J. Birdsall for his English corrections. (UHPFRC) and Reinforced Concrete,” Materials and Structures, V. 39,
No. 289, June 2006, pp. 557-569.
NOTATION 16. Brühwiler, E., and Denarié, E., “Advanced Rehabilitation of Concrete
As,cc = concrete substrate compression reinforcement Bridges using Ultra-High Performance Fibre Reinforced Concrete,”
Structural Concrete in Switzerland, fib-CH, The Second fib-Congress,
As,ct = concrete substrate tensile reinforcement
Naples, Italy, June 2006, pp. 172-175.
As,U = UHPFRC layer reinforcement 17. Association Française du Génie Civil, “Bétons fibrés à ultra-hautes
Ec = concrete secant modulus performances,” Association Française du Génie Civil (AFGC), SETRA—
EU = UHPFRC secant modulus Service d'études techniques des routes et autoroutes, France, Jan. 2002, 152 pp.
EU,hard = UHPFRC hardening modulus 18. Schläfli, M., “Ermüdung von Brückenfahrbahnplatten aus Stahlbeton,”
F = jack force PhD thesis, No. 1998, Swiss Federal Institute of Technology, Lausanne,
f1...f7 = beam deflection measure by LVDT sensors Switzerland, 1999, 113 pp. (in German).
fcc = concrete compressive strength 19. Hillerborg, A., “Analysis of a Single Crack,” Fracture Mechanics of
fct = concrete tensile strength Concrete, F. H.Wittmann, ed., Elsevier Science Publishers B.V., Amsterdam,
fsy,c = concrete tensile reinforcing bar yield strength Netherlands, 1983, pp. 223-249.

ACI Structural Journal/January-February 2007 101