You are on page 1of 12

Received: 21 March 2018 Revised: 20 November 2018 Accepted: 3 January 2019

DOI: 10.1002/suco.201800090

TECHNICAL PAPER

Assessment of model error for reinforced concrete beams


with steel fibers in bending
Stephen J. Foster | Ahsan Parvez

Centre for Infrastructure Engineering and Safety,


School of Civil and Environmental Engineering, Numerous models have been developed to determine the strength contributions of
UNSW Sydney, New South Wales, Australia steel fibers in reinforced concrete beams in flexure; however, little work has been
Correspondence undertaken on determining their reliability, particularly when combined with use of
Stephen J. Foster, School of Civil and
inverse analysis procedures for determining the postcracking residual tensile
Environmental Engineering, The University of
New South Wales Sydney, NSW, Australia. strength provided by the fibers. Recently, leading codes have accepted design
Email: s.foster@unsw.edu.au guidelines for designing flexurally reinforced concrete structural elements that also
contain steel fibers using rational principles that are compatible with conventional
reinforced concrete design. To provide confidence to the designer, it is imperative
to verify the design models against experimental results and test their reliability.
Despite extensive studies, design models have not been validated against large
amount of test data. In this paper, 50 flexurally reinforced steel fiber reinforced
concrete beams from the literature are analyzed for flexural strength. Ultimate
moment capacities are predicted using the design procedures of NZS 3101–2–
2006, fib Model Code 2010, DAfStb Guideline-2015 and AS3600–2018. The
results of the analyses show that the models perform well, with the AS3600 being
the most conservative of the models assessed, and the NZS 3101–2–2006 the least
conservative. The model errors for each are determined, and model reliability is
discussed.

KEYWORDS

bending, database, flexure, model error, reliability, steel fiber reinforced concrete

1 | INTRODUCTION While research into the performance of structures fabri-


cated from steel fiber reinforced concrete (SFRC) has been
Structural performance of reinforced concrete (RC) members on-going for over four decades, SFRC remains an emerging
that contain both conventional flexural reinforcement and material in practice. New types of “structural fibers” have
steel fibers has been widely investigated, and this research become available in the marketplace, which rely on the
shows that steel fibers enhance the flexural and shear structural properties of the fiber, in addition to the anchorage
strength, as well as fatigue and service performance of struc- performance of the fibers in the concrete matrix. The addi-
tural concrete members. However, its application in practice tion of fibers enables hardened concrete to carry tensile
is limited by availability of national design standards and the forces after the concrete has cracked. Postcracking fracture
reluctance of engineers in adopting performance solutions, energy of SFRC enhances mechanical behavior of structural
or alternative pathways, provided by national construction members in service conditions and at ultimate limit state.2
codes (e.g., Ref. 1) in justification of their designs. When properly designed in terms of matrix composition,
fiber dosage, and workability, SFRC can reduce concrete
Discussion on this paper must be submitted within two months of the print
publication. The discussion will then be published in print, along with the tensile brittleness, increase ductility, durability, energy
authors' closure, if any, approximately nine months after the print publication. absorption, fatigue, and toughness.3–5

1010 © 2019 fib. International Federation for Structural Concrete wileyonlinelibrary.com/journal/suco Structural Concrete. 2019;20:1010–1021.
FOSTER AND PARVEZ 1011

Extensive research has been conducted into the shear 2 | F L E X U RA L AN AL YSI S OF RC - SFR C
strength of flexurally reinforced SFRC beams and slabs.6–14 MEMBERS
Research outcomes show that steel fibers can replace a sub-
stantial quantity of conventional shear reinforcement in RC Much effort has been undertaken over the past two decades
beams. In other studies, fibers have shown to improve in the design of RC-SFRC beams and slabs on the basis of
fatigue performance of RC beams significantly.15,16 mechanical models and physical understandings. Most
Experimental studies have also been conducted to evalu- models proposed in the literature for predicting the flexural
ate the flexural performance of reinforced concrete steel behavior of RC-SFRC members in bending adopt the usual
fiber reinforced concrete (RC-SFRC) beams. Ductile design sectional analysis assumptions of conventional RC: plane
of RC beams is generally related to the yielding of tensile sections remain plane, strain compatibility is assumed, and
steel at ultimate, giving sufficient warning before failure. force and moment equilibrium are attained. In most design
The bridging action of the fibers is related to pull-out resis- codes, the ultimate moment capacity of a RC-SFRC beam is
tance, resulting in diffused cracks in the members. Owing to calculated considering a rectangular stress distribution for
high tensile strength and high strain capacity of the RC- concrete in compression. The sectional analysis takes into
SFRC composite, large deformations in longitudinal steel account both the effect of fibers and concrete.
reinforcements are involved, and consequently, high ductil- An example of this approach is the25 model, illustrated
ity of the members is achieved.17 Studies highlight the fact in Figure 1, where D = depth of the section, dn = neutral
that the moment capacity of RC beams are enhanced by axis depth, b = width of the beam, Tf = tensile force by RC-
0
addition of steel fibers.4,12,18–23 These studies focused to SFRC, Ts = tensile force by reinforcing bars, f 1:5 = residual
develop a relation between increased flexural capacity with tensile strength of RC-SFRC at crack-opening–displacement
fiber dosage, fiber type, and strength. (COD) 1.5 mm and kg = member size factor. The compres-
In recent years, a worldwide research effort has offered sive rectangular stress block parameters defining the com-
knowledge on design procedures at a level that enables the pressive stresses are those of AS3600, where α2 is the height
consideration of steel fibers as primary reinforcement for of the stress bock, and γdn is its depth. In NZS 3101–229 and
structural applications. The latest accepted design Model Code 2010,28 the same parameters are defined as α1,
methods24–29 provide guidelines for designing RC-SFRC β1 and, η, γ, respectively. The definition of these parameters
structural elements using rational principles that are compati- is given in Table 1. For concrete in tension, the contribution
ble with conventional RC design. Most standards base their of steel fibers is established from residual tensile strength of
recommendations on the utilization of the residual tensile the RC-SFRC that is evaluated from inverse analysis of
strength obtained either from prism bending test or uniaxial prism bending tests. In the case of references,25,26,28 inverse
tensile test. analysis is determined from 3-point notched bending test
Despite extensive studies, the existing design models undertaken according30; this test has a 150 mm square sec-
have not been validated with a large amount of test data. tion, a 25 mm notch depth, and a 500 mm span. The
Design models for determining flexural strength of RC- New Zealand Standard29 uses 3-point notched RILEM test-
SFRC beams need to be evaluated against test data collected ing arrangement,31 which is similar to EN 14651 test. The
from a range studies. In this paper a database is established German DAfStb Guideline-2015 adopts a 4-point unnotched
for testing of design model approaches and determining the prism bending test on a span 600 mm. A further test that is
model error (ME) and coefficients of variation needed to commonly used in the literature is ASTM C1609,41 which is
establish the structural reliability of RC-SFRC members fail- an unnotched four point 150 × 150 mm prism tested on a
ing in flexure. span of 450 mm. A comparison of results of different test

FIGURE 1 Stress block and forces in reinforced concrete steel fiber reinforced concrete (RC-SFRC) section
1012 FOSTER AND PARVEZ

TABLE 1 Rectangular stress block parameters

Stress block height Stress block depth


Standard parameter (α2 in Figure 1) parameter (γ in Figure 1)
0 0
AS3600-2018 0:85 − 0:0015f c ≥ 0:67 0:97 − 0:0025f c ≥ 0:67
0 0
NZS 3101–2: f c ≤ 55MPa: 0.85 f c ≤ 30 MPa: 0.85
0 0
2006 f c > 55 MPa: f c > 30 MPa:
 0   0 
0:85 − 0:004 f c − 55 ≥ 0:75 0:85 − 0:008 f c − 30 ≥ 0:65
0 0
fib Model f c ≤ 50 MPa: 1.0 f c ≤ 50 MPa: 0.8
0 0
Code 50 < f c ≤ 100 MPa: 50 < f c ≤ 100 MPa:
2010  0   0 
1:0 − f c − 50 =200 0:8 − f c − 50 =400

methods can be found in Conforti et al.32 One difference in


FIGURE 2 Definitions of key points on the applied force versus crack
the test setups is that the RILEM, EN, and German tests pro-
mouth opening displacement (CMOD) curve for flexural testing of prisms
vide sufficient rollers to remove frictional restraint from the
according to fib Model Code 201028
supports, whereas the ASTM method provides only pins in
the supporting arrangements. Data outputs and comparisons
from one test type to another may be compared using the
procedures described Amin et al.33 and Foster et al.34 and section (for un-notched tests hsp = D), FRj = load
outputs converted to a the common form, being the Force– recorded at CMODj (Figure 2), and η is a factor to con-
Crack Mouth Opening Displacement (CMOD) relationship vert the result of an unnotched test to a notch test, or vice
described in the fib Model Code 201028 and outlined briefly versa. Equation (2) applies for any prism test configura-
below. tion; for example, for a 3-point bending test a = L/2,
where L is the span, and the fib Model Code 2010 rela-
tionship is attained, whereas for the ASTM C160941
3 | D E T E R M I N A T I O N O F P O S T C RA CK I N G test a = L/3.
R ESIDU AL T EN S IL E S TR EN GTH FROM The parameter η in Equation (2) accounts for the influ-
I NV E R S E AN A L YS IS ence of a notch on determining the location and path of a
crack through the prism section, and its influence on the
The residual tensile strength is the most important aspect of measured strengths. This is described in Htut36 and Foster
SFRC. According to fib Model Code 2010,28 uniaxial tensile et al.37 Where the test specimen does not include a notch
testing is not advised for standard testing of new mixtures, emanating from the extreme tensile surface, the critical
because tensile tests requires specialized laboratory testing crack finds the path of least resistance, and failure occurs
facilities. Instead indirect test such as prism bending tests are
at sections where fiber distributions are statistically lower
recommended, in combination with an inverse analysis pro-
that the mean value. Thus, the equivalent fiber dosage at
cedure to determine the fundamental stress versus crack
the failure section in an unnotched test is less than the
opening (σ–w) law needed for design.
specimen's average dosage. By contrast for specimens that
During the prism bending test, the CMOD is either mea-
are notched, the location of the fracture line is predeter-
sured directly, or converted from displacement measurement
mined, and the fiber volume fraction at the failure
using the principle of rigid body rotations and calcu-
section will, on average, be equal to the supplied fiber dos-
lated as35:
age for the specimen. To convert the results of notched
4δ prism tests to those of unnotched Htut36 and Foster et al.37
CMOD = ðD − dn Þ, ð1Þ
L determined that η = 0.82, which represents the 75th percen-
where L is the span, δ is the mid-span deflection, D is the tile of the of the fiber volume fraction based on a SD of
depth of the section, and the neutral axis depth is taken as 27% of the mean for a normally distributed sample. Simi-
dn = 0.1D. larly, to convert results of an unnotched test to a notched
The residual flexural tensile strength, fRj, is evaluated one, η = 1/0.82 = 1.22.
from the Force–CMOD relationship described in As different codes models have different adaptations
Figure 2 as33: of the prism bending test that define the contribution of
  the steel fibers to a member's bending capacity, some care
3 ηF Rj a
f Rj = , j = 1−4 ð2Þ is needed when assessing ME from a test database where
bh2sp
the material characterization is determined from a varia-
tion of the designated standard test. This potentially intro-
where a = shear span, b = width of the prism, hsp = dis- duces a new source of ME and is discussed in Sections
tance between tip of the notch and top of cross 5.2 and 6.
FOSTER AND PARVEZ 1013

4 | DE SIGN CODE APP ROAC H ES TO and where the characteristic tensile strength of the concrete
C ALC UL AT E FL EXU RA L S T RE NGTH OF is determined by taking the coefficient of variation of the
R C- S F R C BE AM S EN 14651 notch prism tests as 0.25. In Equation (7) Actu =
area of concrete in tension (mm2) at ultimate (for a rectangu-
4.1 | AS3600 (2018) lar section Actu = [D – dn] b) and Ao=15,600 mm2 is a refer-
ence value. In treating the small scale materials tests in this
AS360025 determines the postcracking residual direct tensile manner, the lower coefficients of variation that occur at the
strength (fw) of SFRC based on prism bending tests con- structural level, due to the larger failure surfaces, are appro-
ducted according to European Standard EN 14651. Compar- priately treated.27 Where mean materials strengths are
ing prism bending test results with matched direct tensile used kg = 1.0.
tests, Amin et al.33 demonstrated that the uniaxial tensile
strength can be accurately described by:
4.2 | NZS 3101–2 (2006)
f w = 0:4f R2 + 1:2ð f R4 − f R2 Þ ξðwÞ, ð3Þ The New Zealand Standard adopts the RILEM31 model for
w ðD − d n Þ 1 determining the ultimate resistance of a cross section. In this
ξ ðw Þ =   − , ð4Þ
3 hsp − d n 4 approach the stresses in the SFRC are derived from the
stress–strain diagram shown in Figure 3, where:
where w is the uniaxial tensile crack width, D is the depth of
the prism, hsp is the distance from the extreme compressive σ 1 = 0:7f fctm, fl ð1:6 − d Þ ðd in metersÞ, ð8Þ
fiber to the top of the cut notch, and dn is the depth to the
neutral axis at the point of interest. For the test configuration σ 2 = 0:45f R1 K h at ε2 = ε1 + 0:01% with ε1 = σ 1 =E c ,
of EN 14651 and RILEM Guidelines, D = 150 mm, hsp= ð9Þ
125 mm, and taking, conservatively, dn = 0.3hsp Equa-
σ 3 = 0:37f R4 K h at ε3 = 2:5%: ð10Þ
tion (3) gives the equation25:
  qffiffiffiffi
0 0 0 0
f 1:5 = k 3Db 0:4f R4 − 0:07f R2 ≤ k3Db 0:36 f c , ð5Þ The term Kh in Equations (9) and (10) is an empirically
determined member size factor29:
1
k 3Db = ≤ 1, ð6Þ K h = 1:0 − 0:6ðD − 125Þ=475;
1 + 0:19lf =b
for 125 ≤ D ≤ 600 ðD in mmÞ, ð11aÞ
where lf is the length of the longest fibers, b is the width of
the prism (b = 150 mm), and the term k3Db takes into K h = 0:4; for D > 600: ð11bÞ
account favorable alignment of fibers due to the wall
The need for an empirical coefficient for flexure demon-
effect.34 The limit on the right hand side of Equation (5) is
strates the lack of the predictive ability of early inverse anal-
required due to a breakdown in the inverse analysis proce-
ysis methods for determining the fibers postcracking tensile
dure where the fiber strengthened concrete in the prism
contribution.
(including the wall effect) becomes strain hardening, leading
to multiple cracking.
The tensile stress block in the RC-SFRC in bending is 4.3 | fib model code 201028
0
taken as, 0:9k g f 1:5 , where: The fib Model Code 201028 provides two simplified stress-
k g = 1 + 0:0067Actu =A0 ≤ 1:6 ð7Þ crack opening constitutive laws: a plastic rigid behavior, or a
linear postcracking behavior. The rigid-plastic model, the
simpler of the two approaches, identifies the postcracking
tensile strength of the concrete at the strength limit state, fFtu.
and is determined as:
f R3
f Ftu = ð12Þ
3
where fR3 = residual flexural tensile strengths corresponding
to a CMOD of 2.5 mm (see Figure 2).

4.4 | DAfStb guideline (2015)


DAfStb Guideline quantifies the ability of SFRC to transfer
FIGURE 3 Stress–strain diagram for steel fiber reinforced concrete tensile loads across a cracked section by means of perfor-
(SFRC) (NZS 3101–2, 2006) mance classes (L). Performance classes L1, for serviceability,
1014 FOSTER AND PARVEZ

and L2, for ultimate, correspond to prism bending test deflec- TABLE 2 Identified RC-SFRC beams having a flexural failure
tions of 0.5 and 3.5 mm, respectively. Prism bending
 0  Reference Specimens test method
Basic values for residual tensile strengths f cto are
Dinh et al.12 B18-1a, B18-1b, B18-7a, B18-7b, ASTM C1609
obtained from the characteristics value of residual flexural
 0  B27-3a, B27-3b, B27-5, B27-6
strength f cflk as follows: Meda et al.19 2Ф16-B-30, 2Ф16-B-60, 4Ф16-B-30 UNI 11039
de Montaignac R150RF/F60-1.0, R150RF/F35-1.0, EN 14651
et al.4 R300RF/F60-1.0, R300RF/F35-1.0,
f fcto, L1 = f fcflk, L1 :βL1 ð13Þ R600RF/F60-1.0, R600RF/F35-1.0,
R600RF/F35-1.25
f fcto, L2 = f fcflk, L2 :βL2 ð14Þ Conforti et al.18 W750 FRC25-1, W750 FRC25-2, W750 EN 14651
FRC35-1, W750 FRC35-2, W1000
f fcto, u = f fcflk, L2 :βu ð15Þ FRC25-1, W1000 FRC25-2, W1000
FRC35-1, W1000 FRC35-2
For L2/L1>0.7, βL1= 0.4, and βL2= 0.25; other- Minelli20 H500 FRC50, H500 FRC75 EN 14651

wise, βu=0.37. Ning et al.23 BS-A-SF30, BS-B-SF30, BS-A-SF50, RILEM


BS-B-SF50 TC162-TDF
The calculated residual tensile strengths are obtained
Biolzi et al.21 T40S-NF, T40S-SF, T75S-NF, T75S-SF, EN 14651
from the basic values of residual tensile strengths for: T90S-NF, T90S-SF, T40 M-NF,
T40 M-SF, T75 M-NF, T75 M-SF,
f fctR, L1 = αcf K fF :K fG :f fcto, L1 ð16Þ T90 M-NF, T90 M-SF, T40 L-NF,
T40 L-SF, T75 L-NF, T75 L-SF,
T90 L-NF, T90 L-SF
f fctR, L2 = αcf K fF :K fG :f fcto, L2 ð17Þ
RC-SFRC: reinforced concrete steel fiber reinforced concrete.
f fctR, u = αcf K fF :K fG :f fcto, u ð18Þ

where K fF is a fiber orientation factor that equals 0.5 where


the failure line forms in the direction of casting and equals reinforcement ratio, (ρl = 1.6 to 2.7%). Residual flexural
1.0 where approximately normal to the direction of casting. strength was determined using the method of ASTM
For beams cast horizontally and where the failure crack C1609.41 Of the 28 beams tested, eight beams are identified
forms normal to longitudinal direction of the members, as failing in flexure.
K fF = 1:0; K fG , member size factor, = 1:0 + 0:5Afct ≤ 1:70; Meda et al.19 tested seven beams in four-point loading to
investigate the fiber contribution to the flexural behavior of
Afct is the area of the tension zone of the cracked concrete
RC-SFRC members. All samples were 4.0 m long with a
cross section at ultimate, in m2. The parameter αcf in
span of 3.6 m and a depth of 300 mm. Two different rein-
Equations 16–18 allows for long-term effects on the residual
forcement ratios were provided, 0.75% and 1.5 (2–16 mm
tensile strength of RC-SFRC. As the tests collected for the
and 4–16 mm diameter bars, respectively, with fsy
model assessment, below, are short term only, this parameter
534 MPa). The same concrete matrix was used for all sam-
is taken as unity in the consideration of model performance.
ples with fcm 45.5 MPa. Hook-ended steel fibers were added
to four beams in two different volume fractions, Vf of 0.38%
5 | M O D E L AS S E S S M E N T and 0.76%. The fibers were 50 mm long and had a diameter
of 1.0 mm. The fracture properties of SFRC were deter-
mined from four point bending tests on 150 × 150 ×
5.1 | Experimental test data
600 mm notched beams (span = 450 mm) in accordance
A search of the literature found 50 tests of for RC-SFRC with the Italian Standard UNI 11039.42 Three of the four
members that failed in flexure where corresponding prism beams failed in flexure and are considered for the model
bending tests had also been performed for material character- analysis.
ization. The studies are summarized in Table 2. de Montaignac et al.4 conducted an experimental pro-
Dinh et al.12 tested 28 beams to investigate the shear gram on 19 rectangular and 2 T-beams of different depths
behavior of RC-SFRC beams. The experimental program (150, 300, 500, and 600 mm).The web widths of the beams
consisted of two series of tests: Series B18, with an overall were 400–500 mm; the flange width of the T-beams were
beam depth of 455 mm, and Series B27, with a beam depth 1,000 mm. Five SFRC mixes were used containing two
of 685 mm. There were eight pairs of beams for Series B18 types of hook-ended fibers: RC80/60 fibers (60 × 0.75 mm)
and four pairs plus four single beams for Series B27. The and RC65/35 fibers (35 × 0.55 mm). The concrete strengths
target concrete compressive strength for all of the beams were fcm = 58, 63, 47, 47, and 56 MPa. Three fiber dosages
was 41 MPa. The studied parameters included beam depth were selected: 0.75, 1.0, and 1.25% by volume. Longitudinal
(455 or 685 mm), fiber length (30 or 60 mm), fiber aspect reinforcement ratio was 0.33% and fsy = 450 MPa. Residual
ratio (55 or 80), fiber strength (1,100 or 2,300 MPa), fiber flexural strength of SFRC mixes were evaluated from prisms
volume fraction, (Vf = 0.75 and 1.5%), and longitudinal bending test according to EN 14651. Seven RC-SFRC
FOSTER AND PARVEZ 1015

rectangular beams failed in flexure are considered for model inverse analysis procedure, and different material testing
analysis. arrangements for the testing of prisms, each requires the
Conforti et al.18 tested 15 wide-shallow beams under determination of unique values of FRj, given in Equation (2),
four point loading. The beams were made with fiber volume or an equivalent materials test result. Similarly, in the struc-
fractions of 0, 0.32, and 0.45%, respectively. For each fiber tural tests that make up the database, different coexisting
content two beams with different widths, 750 and prism tests were undertaken to establish the materials charac-
1,000 mm, were cast. The depth of all beams was 250 mm. terization properties; the most common test being the EN
The concrete strength, fcm, was 38 MPa, 1.0% of the longitu- 14651/RILEM and the less common the ASTM C160941
dinal reinforcement was used and fsy = 530 MPa. The fibers and the Italian UNI 1103942 tests. The procedure outlined in
were 50 mm long end-hooked steel fibers at volume frac- Section 3 is used to determine the materials parameters FR1–
tions of 0.32 and 0.45%. Residual flexural strength of SFRC FR4 described by each of the standard tests and for use in the
mixes were evaluated from prisms bending test according to design models.
EN 14651. Eight RC-SFRC beams failed in flexure and are A number of models have been developed that use the
considered for model analysis. values of fR1–fR4 to determine the σ − w law for SFRC in
Minelli et al.20 tested nine beams under three-point bend- tension by inverse analysis.33,34,38 Different code models
ing. The beams were made with 50 mm long end-hooked
have adopted different approaches to determine the contribu-
steel fibers at fiber volume fractions, Vf of 0, 0.64, and 1.0%.
tion of SFRC composite in structural members, each using a
The concrete strength was 32 MPa. For each group, three
variation on the prism bending test for materials characteri-
beams of different depths were tested: 500, 1,000, and
zation. This introduces a source of ME and is discussed in
1,500 mm. All beams had the same width of 250 mm. The
Section 6.
longitudinal reinforcement ratio ρl was 1.0%. The residual
flexural strength of the SFRC mixes was evaluated from
prisms bending test according to EN 14651. Two of RC- 5.3 | Validation of design models
SFRC beams having depth of 500 mm failed in flexure, with The bending moment capacities of RC-SFRC beams pre-
the remainder failing in shear. sented in Table 2 were calculated using the design models of
Ning et al.23 tested seven full-scale steel-fiber-reinforced AS3600,25 NZS 3101–2,29 fib Model Code 2010,28 and
self-consolidating concrete beams. All beams had the same DAfStb Guidelines.27 The resistance (R) of flexural RC
dimensions: width 200 mm, depth 300 mm, and span members is:
2,100 mm. The major test variables were fiber volume frac-
tion (Vf = 0, 0.38, and 0.64) and longitudinal reinforcement R = ME × Rn ð19Þ
ratiosρl (ρl=0.76, 0.96, and 1.18). The concrete strength was where Rn is the nominal resistance determined by the design
53.6 MPa for all specimens. The flexural tensile behavior of model and ME is the ME. The predictive capacity is:
SFRC is determined by performing 3-point prism bending tests
on 150 × 150 × 550 mm notched beams (span = 500 mm, M u, model = T f × zf + T s × zs ð20Þ
depth of notch = 25 mm), according to the RILEM.31 Four where Tf and zf are the internal tensile force and lever arm
beams failed in flexure and are included in the model analysis. for the fibers component, and Ts and zs are the internal ten-
Biolzi et al.21 tested 36 RC-SFRC beams, with or with- sile force and lever arm for the tensile reinforcement, respec-
out conventional shear reinforcement, under a four-point tively; actual or mean values are used for materials
bending to investigate the effectiveness of steel fibers on properties, the not design values (kg = 1.0; K fG = 1:0).
shear and flexural strengths. The specimens had a width of The ME is defined as:
150 mm, depth of 300 mm and spans of 1,820–2,860 mm.
The experimental program considered three types of con- Experimental Capacity
Model Error =
crete with strengths of fcm = 49.6, 73.6, and 84 MPa; two Capacity Calculated from Predictive Model
16 mm diameter reinforcing bars were used for the as tensile rein- ð21Þ
forcement with fsy = 522 MPa. End hooked steel fibers of and are compared in Figures 4–7 for the different design
30 mm length were used at a volumetric ratio, Vf, of 0.65 and standards. The figures compare the influence specimen depth
0.90%. Residual flexural strength of SFRC mixes were evaluated 0
(D), concrete strength (f c ), relative contribution of fibers to
on prisms according to EN 14651. Eighteen RC-SFRC beams
section capacity and influence of volume of longitudinal bar
failed in flexure and are included in the model assessment.
reinforcement ratio (ρl).
Figure 8 shows the cumulative distribution functions
5.2 | Procedure for the determination of residual plotted on normal probability paper for the ME. The vertical
tensile strength axis (y) denotes the ME, and the horizontal axis (x) repre-
For each of the different models, the residual tensile strength sents the distance from the mean as a function of SD. The
needs to be established; while each model uses a different bias factor is the value at zero on the horizontal (average of
1016 FOSTER AND PARVEZ

the MEs), and the slope gives the SD. It is observed that the Standard, based on the earlier RILEM model, has the lowest
lower tail of the distribution is approximately linear, indicat- bias (least conservative). For the25 model for RC-SFRC
ing a normal distribution. beams, the model bias is consistent with that for RC beams
Equation (21) includes variability of test procedures and without fibers (for the AS3600 model39 determined bias fac-
specimen variability, and represents more than just the accu- tor = 1.06 and CoVm = 0.05 for beams without fibers).
racy of the model. The variability of the model error COVME It is observed that the test to model prediction outcomes
is determined as39: is higher for the ASTM results than that of the EN 14651/
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi RILEM results. It is seen, however, the ASTM results are
COVME = COV2m − COV2test − COV2spec ð22Þ from just eight tests from one series.12 When the tests are
examined more closely, it is seen that the Dinh et al. speci-
where COVm is the coefficient of variation of the mea-
mens have high volumes of bar reinforcement (1.56–2.06%)
sured and predicted strengths, COVtest is the coefficient of
and the contribution of the fibers to the ultimate moment is
variation in the measured test loads, and COVspec is the
relatively small (between 4.4 and 7.4% of the total – see Fig-
uncertainty of strengths of the test specimen and in control
specimens and variations in actual specimen dimensions ures 4c–7c). Thus this higher ME is not attributable to the
from those measured and so on. For the testing of RC fibers or inverse analysis procedure but to the reinforcement
beams, Ellingwood et al.40 gives COVtest = 0.02 and component, in particular strain hardening of the reinforce-
COVspec = 0.04. The resulting model statistical parameters ment post yield.
are given in Table 3. Comparing Figures 4a, 6a, and 7a it is seen that for the
Australian, fib and German models strength is independent
of section depth (D). This is not the case, however, for the
New Zealand31 model (Figure 5a), where a higher degree of
6 | DISCUSSION OF RESULTS
conservatism is observed for increasing section depth. This
is due to the empirical depth factor incorporated into the
6.1 | Observations on ME
design model and given by Equation (11).
Of the four models assessed, it is observed that the model of Figures 4b–7b and Figures 4d–7b show only minor
Ref. 25 has the highest model bias factor with a low coeffi- influences of concrete strength and longitudinal reinforce-
cient of variation (most conservative). The New Zealand ment ratio on ME outcomes. However, comparing

(a) (b)

(c) (d)

FIGURE 4 Comparison of experimental-to-predicted moment capacity for AS3600 (2018) model for reinforced concrete steel fiber reinforced concrete (RC-
SFRC) beams against: (a) overall depth, (b) concrete strength, (c) ratio of the moment contributed by fibers to ultimate moment capacity, and (d) longitudinal
reinforcement ratio
FOSTER AND PARVEZ 1017

(a) (b)

(c) (d)

FIGURE 5 Comparison of experimental-to-predicted moment capacity for NZS 3101–2 (2006) model for reinforced concrete steel fiber reinforced concrete
(RC-SFRC) beams against: (a) overall depth, (b) concrete strength, (c) ratio of the moment contributed by fibers to ultimate moment capacity, and
(d) longitudinal reinforcement ratio

(a) (b)

(c) (d)

FIGURE 6 Comparison of experimental-to-predicted moment capacity for fib Model Code 201028 model for reinforced concrete steel fiber reinforced
concrete (RC-SFRC) beams against: (a) overall depth, (b) concrete strength, (c) ratio of the moment contributed by fibers to ultimate moment capacity, and
(d) longitudinal reinforcement ratio
1018 FOSTER AND PARVEZ

(a) (b)

(c) (d)

FIGURE 7 Comparison of experimental-to-predicted moment capacity for DAfStb Guidelines27 model for reinforced concrete steel fiber reinforced concrete
(RC-SFRC) beams against: (a) overall depth, (b) concrete strength, (c) ratio of the moment contributed by fibers to ultimate moment capacity, and
(d) longitudinal reinforcement ratio

(a) (b)

(c) (d)

FIGURE 8 Normal probability plot showing model error (ME) for resistance of reinforced concrete steel fiber reinforced concrete (RC-SFRC) beams (lower
half data)
FOSTER AND PARVEZ 1019

TABLE 3 Statistical parameters for model error for RC-SFRC beams in flexure

Full data set Lower half statistical parameters


Design model Mean CoV Bias factor CoVm CoVME
All data (50 tests)
AS360025 1.107 0.118 1.079 0.073 0.058
29
NZS 3101–2 1.029 0.146 0.973 0.073 0.057
fib Model Code 201028 1.040 0.127 1.013 0.084 0.071
DAfStb Guideline27 1.062 0.118 1.044 0.080 0.066
Beams with paired EN 14651 or RILEM prism bending test data (39 tests)
AS360025 1.080 0.076 1.073 0.063 0.044
NZS 3101–229 0.988 0.098 0.960 0.068 0.052
fib Model Code 201028 1.006 0.082 1.007 0.082 0.069
DAfStb Guideline27 1.032 0.078 1.039 0.081 0.068

RC-SFRC: reinforced concrete steel fiber reinforced concrete; CoV: coefficient of variation.

Figures 5c, 6c, and 7c shows that the trend in the ME deter- contribution to the ultimate moment, y o is the y-axis
mined using the New Zealand, fib and German approaches intercept and m is the slope of the function. Using only
reduces with increasing fibers contribution. This is not seen the EN 14651/RILEM test data (39 tests), to avoid any
for the Australian model (Figure 4c). The trend for each undue influence of the inverse analysis procedure of the
dataset can be expressed as: outcomes, the following tend lines are determined for
  each model:
M u, exp =M u, model = yo + m M uf =M u ð23Þ
Australian Standard : yo = 1:075;m = + 0:101 ð24aÞ
where M u,exp and M u,model are the ultimate moments
determined from the experimental test and model, New Zealand Standard : yo = 1:052;m = − 0:291 ð24bÞ
respectively, M uf /M u is the ratio of the fibers
fib 2010 Model Code : yo = 1:040;m = − 0:166 ð24cÞ

German Guideline : yo = 1:118;m = − 0:348 ð24dÞ


The Australian model shows a small positive slope as
the fibers contribution to the capacity increases; this is attrib-
uted to the cap on Equation (5) that properly accounts for
the strain hardening limit of the cementitious matrix when
determining the residual tensile strength of SFRC by inverse
analysis. The New Zealand, fib 2010 Model Code and Ger-
man Guideline have reasonably strong negative slopes. This
may be indicative of higher variability of the fibers compo-
nent, compared with that provided by the bar reinforcement,
or it may indicate a flaw in the inverse analysis procedure
where the postcracking residual tensile strength provided by
fibers is over-estimated for higher fiber volumes. Further
research is needed to determine which of these effects is
influential.
The influence of the inverse analysis model is further
examined in Figure 9 where the ratio of the moment contri-
bution of the fibers to the code predicted ultimate is plotted
for the 39 specimens where the EN 14651/RILEM test was
used for the determination of the residual tensile strength for
design. It is seen that for a given SFRC specification the
New Zealand Standard consistently gives a higher fibers
component to the ultimate moment, while the Australian
Standards model gives the lowest. This is consistent with the
observations on the computed MEs with the Australian
FIGURE 9 Ratio of the moment contributed by fibers to the ultimate model providing higher mean test to predicted ratios that the
moment capacity for specimens with EN 14651/RILEM control test data others.
1020 FOSTER AND PARVEZ

6.2 | Assessment of inverse analysis procedure on ME 7 | C O NC L U S IO N S


As mentioned in Section 3, the different prism bending tests
This paper presents the ultimate moment capacity predic-
add a source of uncertainty that requires consideration in a
tions of 50 RC-SFRC beams from literature according to the
reliability assessment. On the one hand, in the determination
design models adopted in four different codes. The flexural
of the ME, the different procedures necessarily increase the
strength of the beams is calculated by adding the contribu-
coefficient of variation of the results, leading to conserva-
tion from concrete, conventional tensile reinforcement and
tism in the assessment. On the other hand, the mean is also
steel fibers. The tensile stress versus crack opening displace-
influenced, which may provide a conservative or non-
ment relationship (σ-w) for the SFRC, needed to calculate
conservative error, depending on the type of prism bending
the fibers component to the flexural resistance, is obtained
test associated with a particular design model. An assess-
from standardized prism bending tests, combined with an
ment of uncertainly in the cross-application and analysis of
inverse analysis method.
the inverse models, and its influence on ME, can be ascer-
The results of the analyses show that all the design
tained by separating each combination of inverse analysis
models provide good correlation to the experimental tests,
results from each prism bending test with each design model.
with an appropriate degree of conservatism. Analysis of the
As noted above, the most common test is the EN/RILEM
results show that the model bias factors for the different
test accounting for 39 of the 50 tests found in the literature,
codes varies between 0.97 and 1.08, and ME CoVs are
recognizing that the EN test and its predecessor the RILEM
between 0.057 and 0.071. The recent Australian Standard is
test are similar.
4–10% more conservative that the other standards. Impor-
Table 3 presents the statistical results for all test speci-
tantly, the mean of model error maintains at a constant value
mens as well as for those that are conducted with the EN
for increasing fibers contribution to the ultimate flexural sec-
14651/RILEM prism bending test for characterization of the
tional capacity. The New Zealand standard has the lowest
fibers contribution to tension. While the averages are higher
model bias, but also a low CoV of model error at 0.057. The
for the full data set than those with the ASTM and Italian
fib 2010 Model Code has the highest CoV of model error at
materials characterization tests removed, it is seen that the
0.071, with a model bias of 1.01.
important lower-half statistics are not significantly influ-
enced; the higher averages for the full data are due to the
OR CID
Dinh et al. results being in upper-half data. It is notable
though that the ME coefficient of variation (CoVME) is sig- Stephen J. Foster https://orcid.org/0000-0003-1162-0929
nificantly lower for the AS 3600 model with small data set Ahsan Parvez https://orcid.org/0000-0002-8712-1184
(CoVME = 0.044). This is again attributable to the cap on
the fibers tension model in Equation (5). REFERENCES
1. NCC. National Construction Code. Australia: The Australian Building
Codes Board, 2016.
2. Massicotte B, Bischoff PH. Fibre reinforced concrete: A structural perspec-
6.3 | Assessment of strength reduction factors tive. Proceeding of the 5th Rilem Symposium on Fibre-Reinforced Concrete,
Lyon, France. Rilem Pro 15; 2000, p. 193–202.
For the AS 360025 model, the mean of 1.079 and coefficient 3. Bentur A, Mindness S. Fiber reinforced cementitious composites.
of variation of the model (CoVm) of 0.073, the 5th percentile New York, NY: Taylor & Francis, 2007.
is determined as 0.96 and the 1st percentile as 0.91, 4. de Montaignac R, Massicotte B, Charron JP. Design of SFRC structural ele-
ments: Flexural behaviour prediction. Mater Struct. 2012;45(4):623–636.
respectively. This indicates the adopted strength reduction 5. Olivito RS, Zuccarello FA. An experimental study on the tensile strength of
factor of 0.85 is likely sufficient when undertaking a full the steel fibre reinforced concrete. Compos Part B. 2010;41:246–255.
reliability analysis, whereas for the New Zealand Standard 6. Adebar P, Mindess S, Pierre DS, Olend B. Shear tests of fiber concrete
the 1st percentile value is 0.80 and a lower value for ϕ is beams without stirrups. ACI Struct J. 1997;94(1):68–76.
7. Aoude H, Belghiti M, Cook WD, Mitchell D. Response of steel fiber-
required. reinforced concrete beams with and without stirrups. ACI Struct J. 2012;109
The fib and German standards use a partial safety factor (3):359–367.
approach with the reinforcing steel 0.87 (1/1.15); the fibers 8. Amin A, Foster SJ. Shear strength of steel fibre reinforced concrete beams
with stirrups. Eng Struct. 2016;111:323–332.
component factors are 0.67 (1/1.5) and 0.80 (1/1.25) for the 9. Casanova P, Rossi P, Schaller I. Can steel fibres replace transverse reinforce-
fib 2010 Model Code and German guidelines, respec- ments in reinforced concrete beams? ACI Mater J. 1997;94(5):341–354.
tively. The differences between the reinforcing bar and 10. Cucchiara C, Mendola LL, Papia M. Effectiveness of stirrups and steel fibres
as shear reinforcement. Cement Concrete Comp. 2004;26:777–786.
fibers component accounts for the differences in the vari- 11. Dancygier AN, Savir Z. Effects of steel fibres on shear behaviour of high
ability of tension determined, and hence moment, for the strength reinforced concrete beams. Adv Struct Eng. 2011;14(5):745–761.
steel and fibers components. This is clearly evident in 12. Dinh HH, Parra-Montesinos GJ, Wight JK. Shear behaviour of steel fiber
reinforced concrete beams without stirrup reinforcement. ACI Struct J. 2010;
Figures 6c and 7c where it is seen that the model to test
107(5):597–606.
ratio is reducing with increasing fiber contribution to 13. Minelli F, Plizzari GA. On the effectiveness of steel fibers as shear reinforce-
strength. ment. ACI Struct J. 2013;110(3):379–389.
FOSTER AND PARVEZ 1021

14. Sahoo DR, Kumar N. Monotonic behaviour of large scale SFRC beams 37. Foster SJ, Htut TNS, Ng TS. High Performance Fibre Reinforced Concrete:
without stirrups. Eng Struct. 2015;92:46–54. Fundamental Behaviour and Modelling. Proceedings of 8th International
15. Parvez A, Foster SJ. Fatigue behaviour of steel-fiber-reinforced concrete Conference on Fracture Mechanics Concrete and Concrete Structures
beams. J Struct Eng ASCE. 2015;141(4):04014117-1-8. (FraMCoS-8); 2013 March 10–14; Toledo, Spain; 2013, p. 69–78.
16. Parvez A, Foster SJ. Fatigue of steel fibre-reinforced concrete prestressed 38. de Oliveira e Sousa JLA, Gettu R. Determining the tensile stress- crack
railway sleeper. Eng Struct. 2017;141:241–250. opening curve of concrete by inverse analysis. J Eng Mech ASCE. 2006;
17. Campione G, Mangiavillano ML. Fibrous reinforced concrete beams in flex- 132:141–148.
ure: Experimental investigation, analytical modelling and design consider- 39. Foster SJ, Stewart MG, Sirivivatnanon V, et al. UNICIV report no. R-464:
ations. Eng Struct. 2008;30:2970–2980. A re-evaluation of the safety and reliability indices for reinforced concrete
18. Conforti A, Minelli F, Plizzari GA. Wide-shallow beams with and without structures designed to AS3600. Sydney, Australia: UNSW Australia, 2015.
steel fibres: A peculiar behaviour in shear and flexure. Compos: Part B. 40. Ellingwood B, Galambos TV, MacGregor JG, Cornell CA. Development of
2013;51:282–290. a probability based load criterion for American national standard A58. Washington
19. Meda A, Minelli F, Plizzari GA. Flexural behaviour of RC beams in fibre D.C.: National Bureau of Standards Special Publication 577, U.S. Government
reinforced concrete. Compos Part B Eng. 2012;43(8):2930–2937. Printing Office, 1980.
20. Minelli F, Conforti A, Cuenca E, Plizzari G. Are steel fibres able to mitigate 41. ASTM C1609. Standard test method for flexural performance of fiber-
or eliminate size effect in shear? Mater Struct. 2014;47:459–473. reinforced concrete (using beam with third-point loading). West Consho-
21. Biolzi L, Cattaneo S. Response of steel fiber reinforced high strength con- hocken, PA: ASTM International, 2012;p. 9.
crete beams: Experiments and code predictions. Cement Concrete Comp. 42. UNI 11039. Steel Fiber Reinforced Concrete – Part I: Definitions, Classifica-
2017;77:1–13. tion Specification and Conformity–Part II: Test Method for Measuring First
22. Mahmood SMF, Agarwal A, Foster SJ, Valipour H. Flexural performance of Crack Strength and Ductility Indexes. Rome: Italian Board for Standardiza-
steel fibre reinforced concrete beams designed for moment redistribution. tion, 2003.
Eng Struct. 2018;177:695–706.
23. Ning X, Ding Y, Zhang F, Zhang Y. Experimental study and prediction
A U T H O R S ' B I OG R A P H I E S
model for flexural behaviour of reinforced SCC beam containing steel fibers.
Construct Build Mater. 2015;93:644–653.
24. ACI 318. Building code requirements for structural concrete and commen- Stephen J. Foster, Professor
tary. Farmington Hills, MI: American Concrete Institute, 2008. Centre for Infrastructure Engineering
25. AS 3600. Australian Standard, concrete structures. Sydney, Australia: Stan-
dards Australia, 2018. and Safety, School of Civil and
26. AS 5100.5. Australian Standard, Bridge design part 5: Concrete. Sydney: Environmental Engineering, UNSW
Standards Australia, 2017. Sydney, Australia
27. DAfStb Guideline. Deutscher Ausschuss für Stahlbeton (German Committee
for Structural Concrete), commentary on the DAfStb guideline “steel fibre
s.foster@unsw.edu.au
reinforced concrete”. Berlin: Beuth, 2015;p. 90.
28. fib Model Code 2010. fib Model Code for concrete structures 2010. Ernst &
Sohn, ISBN: 978–3–433-03061-5; 2013.
29. NZS 3101-2. Concrete structures standard part 1–The design of concrete Ahsan Parvez
structures. Wellington: Standards New Zealand, 2006. Centre for Infrastructure Engineering
30. 14651 EN. Test method for metallic fibre concrete- measuring the flexural
tensile strength (limit of proportionality [LOP], residual). Brussels: and Safety, School of Civil and
European Committee for Standardization, 2007;p. 17. Environmental Engineering, UNSW
31. RILEM Final Recommendation TC-162-TDF. Test and design methods for Sydney, Australia
steel fibre reinforced concrete; σ-ε design method. Mater Struct. 2003;36:
560–567.
m.parvez@unsw.edu.au
32. Conforti A, Minelli F, Plizzari GA, Tiberti G. Comparing test methods for
the mechanical characterization of fiber reinforced concrete. Struct Concr.
2018;19:656–669.
33. Amin A, Foster SJ, Muttoni A. Derivation of the σ- w relationship for SFRC
from prism bending tests. Struct Concr. 2015;16(1):93–105.
34. Foster SJ, Agarwal A, Amin A. Design of steel fibre reinforced concrete
beams for shear using inverse analysis for determination of residual tensile
strength. Struct Concr. 2018;19:129–140.
35. Vandewalle L, Dupont D. Bending test and interpretation, test and design
How to cite this article: Foster SJ, Parvez A.
methods for steel fibre reinforced concrete–Background and experiences. Assessment of model error for reinforced concrete
Proceedings, RILEM TC 162-TDF workshop. Schnutgen B, Vandewalle L, beams with steel fibers in bending. Structural Con-
editors; 2003, p. 1–13.
crete. 2019;20:1010–1021. https://doi.org/10.1002/
36. Htut TNS. Fracture processes in steel fibre reinforced concrete”. PhD disser-
tation, School of Civil & Environmental Engineering. Australia: The Univer- suco.201800090
sity of New South Wales; 2010.

You might also like