10ing) Huang, X 2020

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Engineering Structures 209 (2020) 109953

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Reliability-based design of FRP flexural strengthened reinforced concrete T


beams: Guidelines assessment and calibration
Xiaoxu Huanga, Yingwu Zhoua, , Feng Xinga, Yufei Wub, Lili Suia, Ningxu Hana

a
College of Civil Engineering, Shenzhen University, Shenzhen 518060, China
b
School of Engineering, RMIT University, Australia

ARTICLE INFO ABSTRACT

Keywords: In the past few decades, many guidelines and standards for FRP strengthened reinforced concrete (RC) structures
Reliability assessment have been presented based on the theory of reliability-based limit state with the load and resistance factor design
FRP (LRFD) approach. The LRFD method consists of two formats: the resistance reduction factor format and the
Reinforced concrete material partial safety factor format. In this paper, several well-recognized design guidelines with these two
Design guidelines
formats for FRP flexural strengthened RC beams have been assessed based on model error and reliability index.
Calibration
First, the model error of each design guideline was evaluated based on a collected test database. Then, the
reliability index was obtained by the importance sampling (IS) method, in which the optimum density function is
determined by the first-order reliability method (FORM). Finally, the calibration of resistance reduction factor or
FRP partial safety factor was performed with different target reliability indexes. The results show that: (1) The
model error of each failure mode is different for each design guideline. For flexural failure, China’s GB 50608
2010 has the lowest model error. For intermediate crack-induced (IC) debonding, UK’s TR 55 2012 shows the
highest model error. For end debonding, Italy’s CNR-DT 200 R1/2012 is less conservative. (2) It is difficult to
meet the pre-given reliability level for some failure modes in several design guidelines using the originally
suggested design parameters. For example, GB 50608 2010, fib T5.1 2019 and JSCE 2001. (3) The calibrated
resistance reduction factors or the FRP partial safety factors are provided to satisfy the target reliability level.
However, it is meaningless and uneconomical to adopt partial safety factors that are too large or reduction
factors that are too small to fit into a design. Other options can be chosen such as applying additional anchorages
to avoid end debonding, increasing the FRP bonding area, and reducing the FRP thickness to avoid the prob-
ability of IC debonding failure.

1. Introduction The former generally comes from the inherent variability or random-
ness in nature including material properties (e.g., strength of concrete,
In the past few years, fiber reinforced polymer (FRP) composites steel, and FRP), geometries (e.g., cross-section of RC beam, FRP width),
have been deemed as a promising engineering material in the re- and load effects [9–12]. The aleatory type has been referred to as data-
habilitation and strengthening of civil infrastructures [1–3]. FRP com- based uncertainty and can be measured and assessed through random
posites possess numbers of superiorities including high strength and variables based on statistical observations. The epistemic uncertainty is
stiffness-to-weight ratio, excellent corrosion resistance, and simpler more prone to inaccuracy due to human error when predicting reality,
transportation [4,5]. Common strengthening techniques for RC struc- that is, the assumptions or approximations adopted in analytical
tures with FRP composites include externally bonded (EB) type and the models. This type of uncertainty is termed as knowledged-based un-
near surface mounted (NSM) type [6,7]. The former is considered in certainty and can be represented by model error to qualify subjective
this paper. judgments (e.g., model error for the ultimate strength of FRP flexural
In practical engineering, uncertainties are unavoidable. The key strengthened RC beams) [13]. To accounts for these two types of un-
task is to handle uncertainties from a proper and rational basis. Two certainties, the study of FRP strengthened RC structures has been ex-
types of uncertainties are involved in the capacities of FRP-strength- tended from the deterministic domain to a random domain in safety
ened RC structures [8]: aleatory uncertainty and epistemic uncertainty. assessment; thus, a corresponding reliability analysis is necessary


Corresponding author.
E-mail address: ywzhou@szu.edu.cn (Y. Zhou).

https://doi.org/10.1016/j.engstruct.2019.109953
Received 25 March 2019; Received in revised form 16 October 2019; Accepted 15 November 2019
Available online 22 November 2019
0141-0296/ © 2019 Elsevier Ltd. All rights reserved.
X. Huang, et al. Engineering Structures 209 (2020) 109953

Nomenclature ε’s strain of compression steel reinforcement


fcd design concrete compressive strength
α1, β1, η, λ, k1, k2, ψ, δG, ω parameters defining a rectangular stress f'c, fcu concrete compression strength
block in the RC cross-section ψf additional reduction factor for FRP
a, a’ distance from tensile and compression steel to concrete ρs, ρf area fraction of steel and FRP
compression zone edge α live-to-dead load effect ratio
ϕ, ϕc, ϕs, ϕf strength reduction factor for the whole resistance, KR model error
concrete, steel, and FRP fy stress in steel
εfe,m1, εfe,m2 effective strain in FRP when concrete crush and IC εfdd debonding strain
debonding occurs ΓFk fracture energy
As, A’s area of the tensile and compression steel reinforcement βw width coefficient
Ld distance from FRP cut-off to the nearest applied load ft concrete tensile strength
df , d height and effective depth of concrete cross section x height of the neutral axis
γc, γs, γf partial safety factor for concrete, steel, and FRP βT target reliability index
b, bf width of rectangular cross-section and FRP Gf interfacial fracture energy
Af, tf, Ef area, thickness, and Young’s modulus of FRP fctm concrete tensile strength
fs, ffe stress of the tensile steel reinforcement and FRP εfk characteristic strain at failure
Tk,max maximum ultimate bond force for debonding ηa environmental conversion factor

[2,14,15]. A reliability-based design is of vital importance if en- (ULS) while 1.0 is given for the serviceability limit state (SLS). For the
gineering structures are to achieve a more rational balance between European International Federation for Structural Concrete fib—féderation
safety and life-cycle costs [12,16–19]. Combined with the reliability- internationale du béton—fib 2019 [22], γf varies with the type of fiber. In
based design philosophy, various guidelines and standards (e.g., ACI the Italian Research Council’s guideline CNR-DT 200 R1/2012 [34], the
440.2R 2017 [20], GB 50608 2010 [21], fib T5.1 2019 [22] and TR 55 value of γf differs for different limit states and different failure modes.
2012 [23]) have been released to facilitate the reliable and economic For the British Concrete Society Technical Report 55 (TR-55) [23]
design of a FRP strengthening system for RC structures. guideline, γf relies on the fiber type and method of manufacturing.
The load and resistance factor design (LRFD) approach developed Table 1 lists the recommended resistance reduction factor or FRP par-
by Ellingwood et al. [13] in 1980 provides a valuable tool to establish a tial safety factor for some guidelines. The value of these design factors
reliability-based design framework for FRP-strengthened RC structures. varies with FRP types, manufacturing methods, and reinforcement
In this method, a reduced resistance (strength) of a structural member types. However, part of these design factors was obtained experimen-
exceeds the magnified load effects [24]. The reduced resistance can be tally or modified based on the traditional design method for RC struc-
expressed in two different formats: the resistance reduction factor tures [35]. The rationality of these design factors has to be further
format and partial safety factor format. The general design equations verified.
using these two formats for FRP strengthened RC members are written
as Eq. (1.1) and Eq. (1.2), respectively. Table 1
The design factor of FRP strengthened RC structures with references.
Rn (fck, fsk, f fk , …) Qi Qi (1.1)
Reference Design factor Application
Rd (fck c , fsk s , f fk f , …) Qi Qi (1.2) ACI 440.2R 2017 [20] ϕ = 0.65–0.9, ϕf = 0.85 EB FRP RC beams
GB 50608 2010 [21] γf = 1.4 for prefabricated FRP EB FRP RC beams
In Eq. (1.1), the whole nominal strength (Rn) is multiplied by a γf = 1.25 for other FRP
resistance reduction factor (ϕ) to reflect the design resistance. In Eq. fib T5.1 2019 [22] γf = 1.35 for CFRP EB FRP RC beams
(1.2), the nominal strength of concrete, steel and FRP (fck, fsk and ffk) is γf = 1.45 for AFRP
divided by the material partial safety factors (γc, γs and γf) to ensure the CNR-DT 200 R1/2012 γf = 1.1–1.25 FRP rupture EB FRP RC beams
[25] γf = 1.2–1.5 FRP debonding
safety margin for the design resistance (Rd). The design load effect is γf = 1.0 for serviceability
expressed as a linear combination of several nominal load effects (Qi) TR 55 2012 [23] γf = γE × γε×(γm)2 EB FRP RC beams
multiplied by corresponding partial safety factors (γQi). γE = 1.1 for CFRP, 1.6 for GFRP
In the existing design guidelines, ACI 440.2R 2017 [20] follows the γε = 1.25 for CFRP, 1.85 for GFRP
γm = 1.2 for wet lay-up, 1.05 for
concept of resistance reduction factor while GB 50608 2010 [21] and
pultruded
fib T5.1 2019 [22] are the representatives for the partial safety factor JSCE 2001 [26] γf = 1.2–1.3 for safety EB FRP RC beams
format. Notably, an additional reduction factor (ϕf) is recommended in γf = 1.0 for serviceability
the American Concrete Institute (ACI) guidelines to consider the un- Plevris et al. [27] ϕ = 0.85, ψf = 0.95 (βT = 3.0) EB CFRP RC beams
certainty inherent in FRP systems [20]. This additional factor is similar ϕ = 0.8, ϕf = 1
Okeil et al. [28] ϕ = 0.85 (βT = 3.5) EB CFRP bridge
in some ways to the partial safety factor format. The resistance reduc-
girders
tion factor and partial safety factors are determined based on prob- Pham and Al-Mahaidi ϕ = 0.6 for flexural, IC debonding EB CFRP bridge
ability theory by considering a target reliability level. As listed in [16] ϕ = 0.5 for end debonding girders
Table 1, in ACI 440.2R 2017 [20], the resistance reduction factor ϕ (βT = 3.25)
Wang [29] ϕf = 0.75 (βT = 3.0) EB FRP RC beams
ranges from 0.65 to 0.9 by considering the strain of tensile steel; this is
Shi [30] ϕf = 0.75 for ACI IC debonding EB FRP RC
consistent with ACI-318 [33]. An additional factor for FRP ϕf is sug- (βT = 3.5) members
gested as 0.85 by considering the target reliability index of 3.5. In the Monti and Santini γf = 1.33 EB FRP RC
Chinese Technical Code for Infrastructure Application of FRP Composites [31] members
GB 50608 2010 [21], γf is equal to 1.4 and 1.25 for prefabricated FRP He and Li [32] γf = 1.25 FRP bar in RC
beams
and other FRP, respectively. For the Japan Society of Civil Engineers
Zeng et al. [15] γf = 1.65 for concrete crushing EB FRP RC beams
concrete guideline JSCE 2001 [26], the value of γf is given for different (βT = 3.8)
limit states. For example, 1.2–1.3 is given for the ultimate limit state

2
X. Huang, et al. Engineering Structures 209 (2020) 109953

Numerous studies have been focused on the calibration of these 2. Overview of main design guidelines for FRP-strengthened RC
design factors, as shown in the latter part of Table 1. The first attempt to beams
use LRFD for FRP-strengthened RC structures was performed by Plevris
et al. [27]. Based on the Monte Carlo simulation (MCS), the resistance According to the design equation utilized for LRFD, the design
was regarded as a normal random variable. Within the ACI guideline, a standards and guidelines for FRP-strengthened RC structures can be
general resistance reduction factor ϕ (ϕ = 0.8) was calibrated to satisfy divided into two categories. Category 1: Guidelines with a resistance
a reliability index of 3.0 by considering flexural failure modes. Okeil factor design format, i.e., Eq. (1.1). USA’s ACI 440.2R 2017 [20] as well
et al. [28] recommended that ϕ be equal to 0.85 (consistent with ACI as Canada’s CSA S806-02 [39] are selected. Category 2: Guidelines with
440.2R 2017 [20]) for FRP strengthened RC bridge girders in the FRP a partial safety factor design format (i.e., Eq. (1.2)). China’s GB50608
rupture failure mode. Pham and Al-Mahaidi [16] presented a resistance 2010 [21], Europe’s fib T5.1 2019 [22], Italy’s CNR-DT 200 R1/2012
reduction factor without consideration of the randomness of the load [25], UK’s TR 55 2012 [23], and Japan’s JSCE 2001 [26] are chosen.
effect: ϕ = 0.6 for flexural failure and intermediate crack-induced (IC) Normally, four failure modes within the ultimate limit state (ULS)
debonding, and ϕ = 0.5 for end debonding. Shi et al. [30] re- [40] are described in FRP flexural strengthened RC beams: concrete
commended ϕf = 0.75 for IC debonding with βT = 3.5 for the ACI crushing, FRP rupture, IC debonding and end debonding [6]. Concrete
guidelines. In the aspect of partial safety factor design, Monti and crushing occurs when the concrete compressive strain equals the ulti-
Santini [31] suggested that γf is equal to 1.33 according to the prob- mate value. FRP rupture occurs once the FRP effective strain equals the
ability distribution of γf. Coelho et al. [36] defined a calibrated γf that ultimate strain. Debonding failure exists once the force in the FRP ex-
varied with guidelines and failure modes, e.g., 1.1 and 1.2 for FRP ceeds the capacity provided by the substrate. The details of these failure
rupture and debonding, respectively. Here γf was based on the Italian modes are in [6]. In this paper, the description of each design guideline
guideline CNR-DT 200 R1/2012 [25]. The γf was 1.18 in the Canadian is based on a rectangular section. Besides the plane-section assumption,
Highway Bridge Design Code [37]. By considering compressive mem- the tensile strength of concrete is also ignored to simplify the calcula-
brane action (CMA), Zeng et al. [15] presented a value of 1.65 for γf in tion. To maintain consistency with the rules in each guideline, the RC
the scope of the European guideline (fib TG 9.3 2001) [38]. He and Li beam considered in this paper are externally bonded with wet lay-up
[32] suggested γf is equal to 1.25 for RC beams strengthened by FRP CFRP as shown in the following reliability analysis, and the corre-
bars within the Chinese guideline GB 50608 2010. sponding parameters recommended by each guideline are summarized
From the literature review, it can be found that: (1) The reliability in Table 2.
analysis was carried out with one or two types of failure modes for FRP
strengthened RC beams. This could have possibly caused an in-
comprehensive guide in the design process. (2) The systematic com- 2.1. ACI 440.2R 2017
parison and safety assessment of different design guidelines is not en-
ough. The calibration of design factors in these guidelines has to be In the American guideline ACI 440.2R 2017 [20], the nominal
further identified. Targeting these issues, a comparative study of re- flexural bending moment M can be calculated from conventional sec-
liability-based limit state design within different guidelines was carried tion analysis, as shown in Fig. 1(c).
out for FRP-strengthened RC beams, including mode error evaluation,
reliability index calculation, and design factor calibration for different M = As fs (d 1x 2) + f Af f fe (df 1x 2) (2)
failure modes. The remaining sections of this paper are organized as
follows: In Section 2, seven well recognized design guidelines (ACI The height of the neutral axis x in Eq. (2) can be obtained with a
440.2R 2017 [20], CSA S806-02 2002 [39], GB 50608 2010 [21], fib trial-and-error procedure as per Eq. (3).
T5.1 2019 [22], CNR-DT 200 R1/2012 [25], TR 55 2012 [23], and
x = (As fs + Af f fe ) ( 1 f c 1 b) (3)
JSCE 2001 [26]) for FRP strengthened RC structures are reviewed. In
Section 3, statistical features of relevant design variables are presented,
in which the model error of each design guideline is evaluated based on The terms α1 and β1 are defining parameters of a rectangular
a database dedicated to references. In Section 4, a reliability analysis is equivalent concrete stress block. This is in line with the Whitney stress
performed for each failure mode, and the corresponding resistance re- block where α1 = 0.85 and β1 is selected from ACI 318 [33], i.e., Eq. (4)
duction factor or FRP partial safety factor is calibrated within each is employed in the ACI guideline since the Whitney stress block can give
design guideline. Section 5 summarizes findings as conclusions. reasonably accurate results for failure modes, including concrete
crushing, FRP rupture, and debonding [41]. The additional reduction
factor for FRP ϕf is recommended as 0.85 in the ACI model based on
reliability analysis.

Table 2
Parameters recommended in each design guideline.
Categories Parameters Guidelines

ACI CSA GB fib CNR TR JSCE

Concrete strain εcu 0.003 0.0035 0.0033 0.0035 0.0035 0.0035 0.0035
Concrete stress block α1 0.85 Eq. (8.1) Eq. (10) 0.8/Eq. (17.1) Eq. (23.1) Eq. (29.1) Eq. (33.1)
β1 Eq. (4) Eq. (8.2) 0.8 0.4/Eq. (17.2) Eq. (23.2) Eq. (29.2) Eq. (33.2)
Partial safety factor γc – ϕc = 0.6 1.4 1.5 1.5 1.5 1.3
γs – ϕs = 0.85 1.1 1.15 1.15 1.05 1.0
γf – ϕf = 0.75 1.4 1.35 1.1 1.98 1.15
Resistance reduction factor ϕf 0.85 – – – – – –
ϕ Eq. (6) – – – – – –
Load factor γD 1.2 1.25 1.2 1.35 1.35 1.4 1.0
γL 1.6 1.5 1.4 1.5 1.5 1.6 1.1
Target reliability βT 3.5 3.8 3.2/3.71 3.8 3.8 3.8 3.8

1
3.2 is for brittle failure and 3.7 is for ductile failure.

3
X. Huang, et al. Engineering Structures 209 (2020) 109953

Fig. 1. Internal strain and stress distribution for a rectangular cross-section. (a) RC section; (b) strain distribution; (c-f) equivalent concrete stress distribution: c- ACI
440.2R 2017, GB 50608 2010, JSCE 2001, CSA S806-02; d- fib T5.1 2019; e - CNR-DT 200 R1/2012; f - TR 55 2012.

0.85, f c 28 MPa M = fc bx (d x 2) + Af f , md (df d) (9.1)


1 =
0.85 0.007(f 28), f > 28 MPa (4)
c c fc bx = fy As + Af f , md (9.2)
Based on the equation proposed by Teng et al. [42], a modified IC The steel stress fy is evaluated by its stress-strain behavior. In the
debonding strain is given in the ACI model: above equations, the reduction parameter for the equivalent stress ω is
= 0.41 f (nEf t f ) 0.9 evaluated by:
fd c fu (5)
0.5 + 0.5 fe, m1, >
The strength reduction factor ϕ is based on the steel strain εs. =
fe, m2 fe, m1 fe, m2
0.5 + 0.5f fd (Ef fe, m1), Ef fe, m1 > f fd (10)
0.90, s 0.005
= 0.65 + 0.25( s sy ) (0.005 sy ), sy < s < 0.005 The design FRP tensile stress σf,md in the ultimate limit state is de-
0.65, s < sy (6) termined by:

The ACI guideline has been deemed capable of avoiding or at least f , md = min{f fd , Ef fe, m1, Ef fe, m2} (11)
controlling the end debonding failure mode by using general measures, where ffd is the design FRP tensile strength, εfe,m1 is the effective strain
such as providing sufficient overlap or using anchorage. However, no in FRP when concrete reaches to the ultimate strain (εcu = 0.0033) and
detailed stress equation for end debonding is given in ACI; thus, the end can be obtained by the following equations:
debonding will not be considered as pertaining to the ACI guideline in
the following model assessment. fc bx = f y As + Ef fe, m1 Af (12.1)

2.2. CSA S806-02


x = 0.8df cu ( cu + fe, m1) (12.2)
Based on Lu et al. [43], the IC debonding effective strain for FRP
The design internal force equilibrium equations of the cross-section, εfe,m2 is written as:
according to the CSA S806-02 [39], are expressed as:
fe, m2 = (1.1 Ef t f 0.2 Ld) w ft (13)
M= s As f y (d 1x 2) + f Af Ef f (h 1x 2) (7.1)
with concrete tensile strength ft and width coefficient βw estimated as:
c 1 fc 1 bx = s As f y + f Af Ef f (7.2)
ft = 0.53 f c , w = (2.25 bf b) (1.25 + bf b) (14)
with parameters α1 and β1 in Fig. 1(c) expressed as:
Ld is the distance from the nearest load to FRP edges. The Chinese
1 = 0.85 0.0015fc 0.67 (8.1) guideline GB 50608-2010 [21] states that anchors should be applied for
the end part between FRP and concrete to avoid the end debonding
1 = 0.97 0.0025fc 0.67 (8.2) failure mode. Thus, end debonding failure mode will not be considered
In Canada’s CSA S806-02 [39], several additional assumptions are in the following material when mentioned.
required, such as (1) adequate anchorage and development length is
ensured for FRP reinforcement, and (2) perfect bonding exists between 2.4. fib T5.1 2019
concrete and steel reinforcement and between concrete and FRP re-
inforcement. Therefore, it mentioned that the FRP debonding failure In the fib T5.1 2019 [22], as depicted in Fig. 1(d), the design ca-
mode will not occur and can be ignored by using the specialized an- pacity M is estimated by:
chorage technique. In this guideline, only concrete crushing and FRP M = As fyd (d G x) + Af Ef f (d f G x) + As Es s ( Gx a) (15)
rupture are considered in the model assessment.
And the neutral axis depth x can be calculated with a trial-and-error
2.3. GB 50608 2010 procedure based on:
0.85 fc bx = As f yd + Af Ef f (16)
According to the Chinese GB 50608 2010 [21], the internal force
equilibrium equations of the cross section are expressed as: The steel in tension is assumed to be yield first before other failure

4
X. Huang, et al. Engineering Structures 209 (2020) 109953

modes occur (i.e., fy = fyd). A reduction factor 0.85 is utilized to con- f fdd,2 = kcr 2Ef Fk tf ( f ,d c) (26.2)
sider the reduced compressive strength under long-term loading. The
selection of ψ and δG varies with failure modes. For instance, ψ = 0.8 where kcr = 3.0.
and δG = 0.4 are recommended for the concrete crushing failure mode.
While for FRP rupture or debonding failure mode, ψ and δG in Fig. 1(d) 2.6. TR 55 2012
can be derived as follows based on integration:
In British guideline TR 55 2012 [23], the design ultimate bending
1000 c (0.5 1000 c 12), c 0.002 moment M as shown in Fig. 1(f) can be derived according to BS 8110-
=
1 2 (3000 c ), 0.002 c 0.0035 (17.1) 1:1997 [46] and Teng [47]:
8 1000 c M = k1 fcu c bx (h 2 k2 x ) + s As (h 2 d) + f Af (h 2 df ) (27)
4(6 1000 c )
, c 0.002
G = 1000 c (3000 c 4) + 2 in which fcu is the concrete compression strength, and the neutral axis
, 0.002 0.0035
2000 c (3000 c 2) c
(17.2) height x is estimated according to the Eq. (28):
For IC debonding, the FRP stress between two adjacent cracks sa- k1 fcu c bx + s As + f Af = 0 (28)
tisfies:
The stress in steel σs is determined based on the stress-strain curve.
fd max fd (18) In this model, an accurate stress distribution for compressive concrete is
estimated through integration. The corresponding parameters k1 and k2
in which Δσfd is the increase in tensile stress between two cracks. The
in Fig. 1(f) are derived as [47]:
detailed equations can be found in the fib guideline (fib TG 9.3 2001
[38]). For end debonding, the effective FRP stress is defined by [44]: 0.67( c co
2
3 co2 ), 0
c c co
k1 =
= c1 Ef fck fctm t f
0.67(1 c 3 co), co c 0.0035 (29.1)
f c (19)
with concrete tensile strength fctm written as:
k2 =
( 1
3 12 co
c
) (1 c
3 co ), 0 c co

fctm = 0.3(fck )2 3 = 0.3(fc 8)2 3 (20)


( ) ( ),
2
c c co co
+ c co c 0.0035
2 12 co 3 3
(29.2)

2.5. CNR-DT 200 R1/2012 The maximum ultimate bond force for debonding Tk,max is given by:

Tk,max = 0.5 w bf Efd t f fctm (30)


In Italian guideline CNR-DT 200 R1/2012 [25], the flexural capa-
city M is expressed as: with width coefficient βw and concrete tensile strength fctm expressed as:
M = bxfcd (d x ) + As s (d a ) + Af f a (21.1) w = 1.06 (2 bf b) (1 + bf 400) > 1.0, fctm = 0.18fcu 2 3
(31)
with the position x of the neutral axis in Fig. 1(e) expressed as: British TR 55 2012 [23] concludes that increasing the FRP bonding
0 = bxfcd + As s As fyd Af f (21.2) area and reducing the FRP thickness reduce the likelihood of separation
failure modes. These are the recommendations for debonding design.
The effective strain for FRP is determined by: Besides, adequate FRP end anchorage must be provided. Thus, the
= min{ analysis of the end debonding failure mode for TR 55 2012 will not be
fd a fk f , fdd } (22)
considered in the following reliability analysis.
where fcd is the design concrete compressive strength, ηa is the en- Different from other design guidelines, the FRP partial safety factor
vironmental conversion factor, εfk is the characteristic strain at failure in TR 55 is the product of three parts: the factor for FRP Young’s
and εfdd is the debonding strain. The stress in steel is determined based modulus γE, the factor for FRP ultimate strain γε, and the factor for
on the strain curve. No recommendation on the concrete stress block (η, manufacture method γm, as shown in Table 1. For example, γf takes the
λ) has been provided in the CNR model. Therefore the parameters used value of 1.98 (γf = γE × γε × (γm)2 = 1.1 × 1.25 × 1.22) for wet lay-
in the study have referred to the Eurocode 2 [45]. The following as- up CFRP. It can be found that, the γf in TR 55 is larger than the γf in
sumptions for the concrete stress block in Fig. 1(e) have been adopted: other guidelines due to the product form.

1.0 fck 50MPa


= 2.7. JSCE 2001
1.0 (fck 50) 200, 50 < fck 90MPa (23.1)
Based on the fracture mechanics, the debonding design in Japanse
0.8 fck 50MPa guideline JSCE 2001 [26] is derived. And the maximum value for the
=
0.8 (fck 50) 400, 50 < fck 90MPa (23.2) difference of tensile stress within the continuous fiber sheet satisfies:

In guideline CNR-DT 200 R1/2012 [25], fracture mechanics are f 2Gf Ef (nf t f ) (32)
adopted to describe the debonding failure mode. The specific fracture
energy ΓFk of the FRP-concrete interface is given by: where Gf = 0.5 N/mm as the interfacial fracture energy. The details can
be found in [48].
Fk = 0.03 w fck fctm (24) If debonding does not occur, the flexural design capacity may be
determined using the same method for RC members [49]. The para-
with
meters which define the equivalent stress block as shown in Fig. 1(c)
w = (2 bf b) (1 + bf 400) 1 (25) are recommended as:

The ultimate design strength for end debonding and IC debonding 1 =1 0.003fck 0.85 (33.1)
are derived by Eq. (26.1) and Eq. (26.2), respectively:
1 = 0.52 + 80 cu (33.2)
f fdd = 2Ef tf ( c) (26.1)
Fk f with the ultimate concrete compressive strain expressed as:

5
X. Huang, et al. Engineering Structures 209 (2020) 109953

cu = (155 f ck ) 30000 0.0035 (34) The model error follows a normal distribution for ACI 440.2R 2017 and
fib T5.1 2019, and a lognormal distribution for GB and JSCE, a weibull
distribution for CNR, and an extreme value distribution for UK’s TR 55
3. Statistical characteristics of random variables 2012. The PDF and CDF of model error for the ACI 440.2R 2017
standard are plotted in Fig. 4 as an example.
3.1. Model uncertainties
3.1.3. End debonding
Model uncertainties may result from several approximations or as- For the end debonding failure mode, the authors performed the
sumptions when predicting the ultimate strength with the prediction estimation of model error and the corresponding K-S test. The results
model. This type of epistemic uncertainty can be reduced with im- are shown in Table 7. It can be inferred that, both the fib T5.1 2019 and
proved knowledge of the prediction model. Model errors were analyzed CNR-DT 200 R1/2012 guidelines have a large mean value of model
for various design models with two specific purposes: to provide design error indicating a more conservative prediction for end debonding. A
guideline comparison and to utilize the epistemic uncertainty quanti- weibull distribution was selected for the model error in the European fib
fication. The value of the model error is denoted as the ratio of the T5.1 2019 standard while a normal distribution was chosen for the CNR
experimental result Mexp to the prediction value Mpre, i.e., guideline in the following reliability analysis. The PDF and CDF of
KR = Mexp Mpre (35) model error for the fib T5.1 2019 guideline are plotted in Fig. 5 as an
example.
The random information of KR is necessary for the assessment of
prediction models and the reliability index within each design guide- 3.2. Uncertainties of resistance design variables
line. Therefore, in order to obtain this random information, a screened
test database that contains 624 sets of test data for FRP-strengthened According to the formulation of resistance, the design variables of
RC beams was established [14]. The number of test data for each failure resistance mainly come from four parts, i.e., properties of geometry,
type is listed in Table 3. Part of the experimental FRP strengthened RC concrete, steel, and FRP. Thus, the uncertainties of these design vari-
beam was equipped with anchors, such as U-jacketing in the middle ables are required for the following reliability analysis. This kind of
part or end part. The following assessment for model error is based on data-based uncertainty was determined according to published refer-
this database. It is pointed out that the guidelines with the same failure ences, as listed in Table 8. The area fraction of steel and FRP (i.e., ρs and
mode can be compared in the form of the statistical characteristic for ρf) are considered as deterministic since the variations are small. The
the corresponding model error. term “Bias” refers to the ratio of mean value to nominal value or
characteristic one.
3.1.1. Flexural failure: Concrete crushing and FRP rupture
Based on the database, the model uncertainty for flexural failure 3.3. Load uncertainties
and its stochastic characteristics within each design guideline are pre-
sented in Fig. 2. The average values of KR are close to 1.0 with the Consider the general linear combination of live load and dead load,
coefficient of variation (COV) close to 0.2. Since there exists little im- the design load Sd is expressed as follows:
provement for experimental flexural capacity when additional anchors
are applied, the mean value of KR in the case with anchor is a little
Sd = D Dn + L Ln (36)
larger than the one in the case without anchors for all the guidelines. where Dn and Ln are the standard values for the dead load and live load
The Chinese code [21] shows the least conservative prediction of effect, respectively. The load partial factors for the dead and live load
flexural capacity with the smallest mean value of KR, while the British effects are denoted as γD and γL, respectively. The suggested γD and γL
code [50] is the most conservative prediction model. within each guideline are listed in Table 2. A more complicated design
To identify the probability distribution for the model error based on case, such as the inclusion of wind loads and earthquake effects, will
the experimental data and the expected one, hypothesis testing tech- not be considered in this study. The statistical information on loads is
niques can be utilized with several possible distributions in practical given in Table 9.
engineering including normal, lognormal, weibull or extreme value Based on the limit state in Eq. (1), the value for Dn and Ln can be
distributions. Taking the ACI guideline as an example, Fig. 3 depicts the derived as:
probability distribution function (PDF) and cumulative distribution
function (CDF) of KR with the assumed distributions. The Kolmogorov-
Dn = Rd ( D + L· ), Ln = Rd ( D + L· ) (37)
Smirnov (K-S) testing approach was performed with a significance level where Rd is the design value for resistance (i.e., considering partial
of 0.05 to identify the distribution of KR. The asymptotic p-values safety factors or resistance reduction factors), and α = Ln/Dn is the
within each guideline are listed in Table 4. A higher p-value means that nominal live-to-dead load ratio.
the assumed distribution is closer to the actual one; thus, all the model
error for flexural failure can be taken as obeying the lognormal dis- 4. Reliability analysis of FRP strengthened RC beams
tribution.
4.1. Design cases
3.1.2. IC debonding
The model uncertainty for IC debonding failure mode within each To conduct reliability analysis and design factors calibration, a
design guideline is identified as listed in Table 5. The mean values of KR
in the case with anchors is a little larger than the values in the case Table 3
without anchors, since the ultimate flexural capacity is increased with Number of test data for each failure mode.
anchors. It can also be inferred that, applying additional anchors is a
Failure mode Data type
useful approach to increasing the safety margin of FRP-strengthened RC
beams. Except for fib T5.1 2019, TR 55 2012 and JSCE 2001 standards, Total With anchor Without anchor
the average values of KR for others are close to 1.0. China’s GB 50608
Flexural 226 116 110
2010 model is the least conservative model for IC debonding prediction.
IC debonding 240 63 177
Similarly, the K-S testing was performed, and results are shown in End debonding 158 3 155
Table 6, including the asymptotic p‑values for each distribution type.

6
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) ACI 440.2R 2017 (b) CSA S806-02

(c) GB 50608 2010 (d) fib T5.1 2019

(e) CNR-DT 200 R1/2012 (f) TR 55 2012

(g) JSCE 2001


Fig. 2. Model uncertainties for the flexural failure mode.

series of design cases representing the whole design space should be permutation, a total number of Nm = 1296 design cases have been
established first. Important parameters have been summarized and constructed. Furthermore, the live-to-dead load ratio α ranges from 0.5
several representative nominal values have been selected for each to 2.5 with an interval of 0.5.
parameter, as shown in Table 10. Following the principle of

7
X. Huang, et al. Engineering Structures 209 (2020) 109953

Fig. 3. Probabilistic distribution of the flexural failure model uncertainty in ACI.

Table 4
K-S test results for flexural failure (p-value).
Distribution ACI CSA GB fib CNR TR JSCE

Normal 0.0886 0.0482 0.1640 0.0285 0.0173 0.1179 0.0409


Lognormal 0.6418 0.4432 0.7795 0.4118 0.2508 0.7914 0.4082
Weibull 0.0311 0.0131 0.0703 0.0124 0.0041 0.0714 0.0104
Extreme value I 0.0009 0.0002 0.0012 0.0001 0.0003 0.0047 0.0001

Table 5
Model uncertainties for IC debonding.
Guidelines Estimations Guidelines Estimations

Mean COV Data type Mean COV Data type

ACI 440.2R 2017 1.013 0.170 all data fib T5.1 2019 1.367 0.219 all data
1.026 0.184 with anchor 1.403 0.221 with anchor
1.007 0.165 without anchor 1.354 0.230 without anchor
CNR-DT 200 R1/2012 1.142 0.166 all data TR55 2012 1.416 0.231 all data
1.168 0.165 with anchor 1.448 0.222 with anchor
1.133 0.169 without anchor 1.403 0.234 without anchor
GB 50608 2010 0.944 0.152 all data JSCE 2001 1.421 0.246 all data
0.979 0.166 with anchor 1.438 0.221 with anchor
0.931 0.144 without anchor 1.414 0.255 without anchor

Table 6 extremely time-consuming. Besides, a large error will occur when di-
K-S test results for IC debonding failure (p-value). rectly utilizing the MCS-FORM method in reliability analysis due to the
Distribution ACI GB fib CNR TR JSCE
linear approximation with Taylor expansion to limit state function in
the FORM. The accuracy of the method is measured by the error of β,
Normal 0.9144 0.8362 0.7094 0.3848 0.2620 0.4838 i.e., the relative error of β obtained by the method and the MCS.
Lognormal 0.8574 0.9828 0.5886 0.0264 0.0452 0.8225 Considering the one design point in Table 10 for the concrete crushing
Weibull 0.5686 0.4921 0.3576 0.4543 0.5736 0.2052
Extreme value I 0.2254 0.2485 0.1137 0.1387 0.9341 0.0553
failure mode with anchor, the comparison of reliability results among
MCS, MCS-FORM, and IS are plotted in Fig. 6. Take the ACI guideline as
an example and two different resistance reduction factors are studied:
4.2. Reliability analysis procedure case 1 is for ϕ = 0.5, and case 2 is for ϕ = 1.0. It can be found that, the
largest error of β produced by MCS-FORM is 0.903% for case 1 and 2.35%
For the FRP-strengthened RC beams, the limit state function can be for Case 1. And the largest error of Pf is 15.53% for case 1 and 10% for
expressed in a mathematical form: Case 2. In the IS method, the optimal density function for each random
variable is selected to a normal distribution with a standard deviation of
g = kR R D L (38) original PDF, and the mean equals to the most probable point (MPP)
determined by FORM. As shown in Fig. 6, the largest error of β produced
Failure occurs when g < 0. Several methods have been adopted for by IS is 0.3% for Case 1 and 0.41% for Case 2. Moreover, the largest error
the reliability evaluation, including the MCS, MCS-FORM [15] and of Pf is 4.94% for Case 1 and 1.84% for Case 2. To perform the reliability
importance sampling (IS) [14]. FRP-strengthened RC beams are ex- analysis in an accurate and efficient manner, the IS method [14] was
pected to have a low level of failure probabilities (i.e., high reli- adopted in the study presented in Section 4.3.
abilities), indicating that the crude MCS for reliability evaluation can be

8
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) PDF (b)CDF


Fig. 4. Probabilistic distribution of the IC debonding model uncertainty in ACI.

Table 7 0.9 and the variation of ϕf is focused. The equation of γf in the TR 55


Model uncertainties and K-S test result for end debonding failure mode. guideline is composed of three parts (γf = γE × γε × (γm)2). In this
Guideline Estimation p-value
study, γf is regarded as a whole variable and the variation is analyzed.
Besides, since end debonding can be avoided or controlled by effective
Mean COV Normal Lognormal Weibull Extreme anchorages, reliability analysis for end debonding is only based on the
value I data without anchor in the following part. While for other three modes,
all the data is utilized.
fib T5.1 2019 1.463 0.255 0.2932 0.0039 0.4696 0.1072
CNR-DT 200 R1/ 1.381 0.264 0.3235 0.0085 0.1857 0.0084
2012 4.3.1. Concrete crushing
The reliability index of concrete crushing for each design guideline
is plotted in Fig. 7. For the ACI 440 and CSA S806-02, the reliability
4.3. Evaluation of reliability index
index decreases with the increase of ϕf in the whole range. In these
guidelines, the larger the ϕf is, the smaller the safety margin and lower
With the statistical properties of model error, predicted resistance,
reliability index is. For the remaining guidelines, the variation of γf has
and load effects, reliability analysis for each failure mode was con-
no effect on the reliability index since the design resistance does not
ducted with the importance sampling (IS) method, based on the com-
involve FRP yield strength or the FRP partial safety factor for concrete
prehensive design space established previously. In order to examine the
crushing.
reliability for each guideline and give reasonable suggestions if pos-
The effect of live-to-dead load ratio α on the reliability index is
sible, a series of FRP partial safety factors γf or resistance reduction
different for different guidelines. For ACI 440.2R 2017, β decreases as α
factors ϕ are tested in Section 4.3.1. In the ACI guideline, ϕ is fixed at
increases for different ϕf. Meanwhile, β gradually increases with the

(a) PDF for fib T5.1 2019 (b) CDF for fib T5.1 2019
Fig. 5. Probabilistic distribution of the end debonding failure model uncertainty in fib.

9
X. Huang, et al. Engineering Structures 209 (2020) 109953

Table 8 increases of α for GB 50608 2010 and JSCE 2001; however, for the CSA
The statistical characteristic of design variables for resistance. S806-02, fib T5.1 2019, CNR-DT 200 R1/2012, and TR 55 2012
Category Variable Bias COV Distribution Source guidelines, β first increases and then decreases with the increases of α.
Among guidelines with the γf format, GB 50608 2010 shows the lowest
Geometry Width b (mm) 1 0.02 Normal [51] average reliability estimation while fib T5.1 2019 shows the largest
Height df (mm) 1 0.02 Normal [52]
average reliability estimation.
Effective height d (mm) 1 0.03 Normal [28]
Effective length L (mm) 1 0.03 Normal [14]
Concrete Compressive strength fc (MPa) 1.25 0.15 Normal [14] 4.3.2. FRP rupture
Steel Area As (mm2) 1 0.03 Normal [52] The reliability index of FRP rupture for each design guideline is
Young’s modulus Ey (MPa) 1 0.024 Normal [16] plotted in Fig. 8. It can be found that, (1) for ACI 440.2R 2017 and CSA
Steel yield strength fy (MPa) 1.1 0.075 Normal [53]
FRP FRP thickness t (mm) 1 0.05 Lognormal [44]
S806-02, β decreases with the increases of ϕf. For other guidelines, β
Young’s modulus Ef (MPa) 1 0.04 Normal [14] increases with the increases of γf. For the FRP rupture failure mode, the
tensile strength ff (MPa) 1.15 0.1 Weibull [14] FRP yield strength is utilized, and γf will affect the design resistance and
loads effect within the guidelines designed by γf. (2) The effect of α on
the reliability index is different. For ACI 440.2R 2017, β decreases as α
Table 9 increases for different values of ϕf; β increases distinctly with the in-
The statistical characteristic of loads. creases of α for GB 50608 2010, CSA S806 and JSCE 2001. For fib T5.1
Guidelines Load Bias COV Distribution Source 2019, CNR-DT 200 R1/2012, and TR 55 2012, β first increases and then
decreases with the increases of α. (3) Among guidelines with the γf
GB 50608 2010 D 1.06 0.074 Normal [54]
format, GB shows the lowest average reliability estimation, while TR 55
L 0.644 0.233 Extreme type I [54]
ACI 440.2R 2017 D 1.05 0.10 Normal [55] 2012 shows the largest average reliability estimation due to a larger
L 1.00 0.25 Extreme type I [55] safety margin caused by a larger γf in TR guideline. The average β in
fib T5.1 2019 D 1.00 0.05 Normal [5656] ACI is close to the one in CSA.
CSA S806-02, TR 55 2012 L 0.6 0.35 Extreme type I
JSCE 2001, CNR-DT 200
R1/2012
4.3.3. IC debonding
The reliability index of IC debonding for design guidelines is plotted
in Fig. 9. For ACI 440.2R 2017, the reliability index decreases with the
Table 10 increase of ϕf. For GB 50608 2010, fib T5.1 2019, TR 55 2012, and JSCE
Summary of design variables. 2001, the variation of γf has almost no effect on the reliability index.
For CNR-DT 200 R1/2012, the reliability index increases with the in-
Categories Parameters Values
creases of γf since γf is involved in the design equation of the IC de-
Geometry Width b (mm) 100 300 500 bonding strain. The reliability index increases with α for GB, fib and
Height h (mm) 2×b CNR. Among guidelines with γf format, TR 55 2012 shows the largest
Effective height d (mm) 0.9 × h
reliability estimation for IC debonding.
Effective span l0 (mm) 9×h 18 × h
Concrete Concrete strength fc (MPa) 13.4 20.1 32.4
Steel Steel yield strength fy (MPa) 335 400 4.3.4. End debonding
Steel area fraction ρs (%) 0.3 1.5 Many experiments have shown that the end debonding failure mode
FRP strength ffu (MPa) 1000 2500 4000
can be effectively avoided by utilizing appropriate anchors. Thus, sev-
FRP strain εfu 0.01 0.02
Area fraction ρf (%) 0.05 0.1 1
eral guidelines have not mentioned the detailed design equation for end
Area Af (mm2) b × h × ρf debonding. For fib T5.1 2019 and CNR-DT 200 R1/2012, the reliability
thickness tf (mm) Af b f index of end debonding is plotted in Fig. 10. Different from fib, the
reliability index for CNR increases as γf increases since γf is involved in
the design equation of the end debonding strain. Besides, the reliability
index increases with the increases of α. In general, the reliability index

(a) Case 1: φ = 0.5 (b) Case 2: φ = 1.0


Fig. 6. Method comparison for concrete crushing in ACI.

10
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) ACI 440.2R 2017 (b) CSA S806-02

(c) GB 50608 2010 (d) fib T5.1 2019

(e) CNR-DT 200 R1/2012 (f) TR 55 2012

(g) JSCE 2001


Fig. 7. Reliability index of concrete crushing for each design guideline.

11
X. Huang, et al. Engineering Structures 209 (2020) 109953

Fig. 8. Reliability index of FRP rupture for each design guideline.

12
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) ACI 440.2R 2017 (b) GB 50608 2010

(c) fib T5.1 2019 (d) CNR-DT 200 R1/2012

(e) TR 55 2012 (f) JSCE 2001


Fig. 9. Reliability index of IC debonding for each design guideline.

for end debonding is much lower than the one for other failure modes. 4.4. Calibration of the design factors
This implies that the end debonding failure mode has the highest
probability to occur compared with other failure modes. Based on the original ϕf or γf suggested in each design guideline, the
average reliability index for each failure mode is listed in Table 11. The
value in the bracket is the target reliability index βT within the corre-
sponding guideline. The determination of βT is a complicated problem

13
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) fib T5.1 2019 (b) CNR-DT 200 R1/2012


Fig. 10. Reliability index of end debonding for each design guideline.

Table 11
Average reliability index for each design guideline.
Failure modes ACI (3.51) CSA (3.81) GB (3.21, 3.71) fib (3.81) CNR (3.81) TR (3.81) JSCE (3.81)
ϕf = 0.85 ϕf = 0.75 γf = 1.40 γf = 1.35 γf = 1.10 γf = 1.98 γf = 1.15

Concrete crushing 3.42 3.58 2.99 3.36 3.25 3.34 2.95


FRP rupture 3.02 3.19 3.03 3.20 2.99 3.87 2.80
IC debonding 3.21 N/A2 3.31 3.48 3.22 3.56 3.27
End debonding N/A2 N/A2 N/A2 2.90 2.81 N/A2 N/A2

1
Target reliability index.
2
Detailed measures were given to prevent debonding failure modes instead of detailed equations.

(a) ACI 440.2R 2017 (b) CSA S806-02


Fig. 11. LSA analysis for concrete crushing.

that should consider both construction cost and consequences of failure, In this work, to remain consistent with existing design codes, load
as well as the service life of structures (JCSS 2001 [57]). For a 50-year factors were drawn from these codes, and only the resistance reduction
design service life, different target reliability indexes are usually re- factors or FRP partial safety factors should be calibrated. The least-
commended for different design guidelines. For instance, 3.5 for the squares average (LSA) technique was utilized to quantify the deviation
American ACI guideline [20], 3.2 and 3.7 for the Chinese guideline of the reliability index from βT in the calibration process:
[21], and 3.8 for the European fib guideline [38]. In this study, 3.8 is
nc
also used for CSA S806-02 [39], Italy’s CNR-DT 200 R1/2012 [25] and 1 2
H= ( T)
Japan’s JSCE 2001 [26]. It can be found that, the reliability index in nc i=1
i
(39)
many cases is lower than the corresponding target ones. Within the
partial safety factor design format, China’s GB 50608 2010 is the most The calibrated design factor is the value of ϕf or γf leading to the
conservative one while UK’s TR 55 2012 is the least conservative in minimum H. Eq. (39) is an optimization problem, and the optimal de-
reliability assessment. To provide some useful suggestions on these sign factor is obtained herein by the exhaustion method instead of other
design factors, a calibration process is necessary. optimization methods. In the following calibration, a wide range of γf or

14
X. Huang, et al. Engineering Structures 209 (2020) 109953

Fig. 12. LSA analysis for FRP rupture.

15
X. Huang, et al. Engineering Structures 209 (2020) 109953

(a) ACI 440.2R 2017 (b) CNR-DT 200 R1/2012


Fig. 13. LSA analysis for IC debonding.

factor. The behaviors of H as a function of γf or ϕf with these five dif-


ferent target reliability indexes are plotted in Figs. 11–14. It can be
observed that, H first decreases and then increases with the increase of
ϕf or γf.
For the concrete crushing failure mode, only the calibration of the
resistance reduction factor in ACI 440.2R 2017 and CSA S806-02 is
available. With the increase in the target reliability index, the re-
commended resistance reduction factor decreases. The value of 1.0 is
enough to satisfy βT = 3.0 and 3.2 for the ACI and CSA guidelines.
For the FRP rupture failure mode, the calibrated ϕf for the FRP
rupture in ACI 440.2R 2017 and CSA S806-02 is smaller than the one
for concrete crushing. For example, ϕf = 0.8 and 0.75 is recommended
for βT = 3.2 within ACI and CSA, respectively. For GB 50608 2010, fib
T5.1 2019 and CNR-DT 200 R1/2012, the behaviors of H are similar. A
larger γf is required for a higher βT. However, it is not economical to
Fig. 14. LSA analysis for end debonding in CNR. design if γf is too large, such as when γf is calibrated as 2.1 and 2.45 for
βT = 3.8 within GB and JSCE, respectively.
For the IC debonding failure mode, the calibrated ϕf in ACI is a little
ϕf is selected, and the target reliability index includes 3.0, 3.2, 3.5, 3.7
larger than the one for the FRP rupture. In CNR-DT 200 R1/2012, even
and 3.8 [33,55]. The value of ϕf ranges from 0 to 1, while the range of γf
γf = 1 can ensure that the target reliability index is equal to 3.0.
is different for each guideline to show the optimal calibrated design
However, for βT > 3.5, the recommended γf is larger than 2, which is

Table 12
Calibrated resistance reduction factor or FRP partial safety factor.
Failure modes βT ACI CSA GB fib CNR TR JSCE
ϕf ϕf γf γf γf γf γf

Concrete crushing 3.0 1.00 (3.33) 1.00 (3.47) 1.001 (3.00) 1.001 (3.36) 1.001 (3.25) 1.001 (3.34) 1.001 (2.95)
3.2 1.00 (3.33) 1.00 (3.47) 1.001 (3.00) 1.001 (3.36) 1.001 (3.25) 1.001 (3.34) 1.001 (2.95)
3.5 0.85 (3.42) 0.85 (3.53) 1.001 (3.00) 1.001 (3.36) 1.001 (3.25) 1.001 (3.34) 1.001 (2.95)
3.7 0.70 (3.50) 0.65 (3.64) 1.001 (3.00) 1.001 (3.36) 1.001 (3.25) 1.001 (3.34) 1.001 (2.95)
3.8 0.65 (3.53) 0.60 (3.67) 1.001 (3.00) 1.001 (3.36) 1.001 (3.25) 1.001 (3.34) 1.001 (2.95)
FRP rupture 3.0 0.90 (2.93) 0.85 (3.01) 1.35 (2.96) 1.15 (3.01) 1.10 (3.00) 1.50 (3.61) 1.40 (2.99)
3.2 0.80 (3.11) 0.75 (3.19) 1.50 (3.12) 1.35 (3.20) 1.25 (3.16) 1.50 (3.61) 1.65 (3.14)
3.5 0.65 (3.40) 0.60 (3.49) 1.85 (3.43) 1.65 (3.46) 1.50 (3.39) 1.50 (3.61) 2.10 (3.35)
3.7 0.55 (3.61) 0.55 (3.60) 2.05 (3.58) 1.90 (3.58) 1.75 (3.56) 1.60 (3.67) 2.25 (3.44)
3.8 0.50 (3.71) 0.50 (3.72) 2.10 (3.62) 2.05 (3.68) 1.90 (3.66) 1.70 (3.74) 2.45 (3.47)
IC debonding 3.0 0.95 (2.92) N/A 1.001 (3.31) 1.001 (3.48) 1.00 (3.02) 1.001 (3.56) 1.001 (3.27)
3.2 0.85 (3.10) N/A 1.001 (3.31) 1.001 (3.48) 1.20 (3.22) 1.001 (3.56) 1.001 (3.27)
3.5 0.70 (3.40) N/A 1.001 (3.31) 1.001 (3.48) 1.70 (3.48) 1.001 (3.56) 1.001 (3.27)
3.7 0.65 (3.50) N/A 1.001 (3.31) 1.001 (3.48) 2.25 (3.61) 1.001 (3.56) 1.001 (3.27)
3.8 0.60 (3.60) N/A 1.001 (3.31) 1.001 (3.48) 2.45 (3.66) 1.001 (3.56) 1.001 (3.27)
End debonding 3.0 N/A N/A N/A 1.001 (2.90) 1.80 (2.92) N/A N/A
3.2 N/A N/A N/A 1.001 (2.90) 4.00 (3.02) N/A N/A
3.5 N/A N/A N/A 1.001 (2.90) 4.00 (3.02) N/A N/A
3.7 N/A N/A N/A 1.001 (2.90) 4.00 (3.02) N/A N/A
3.8 N/A N/A N/A 1.001 (2.90) 4.00 (3.02) N/A N/A

1
The variation of γf has no effect on reliability indexes, and the smallest γf (i.e., γf = 1.0) is recommended, since a larger γf is meaningless and uneconomical for
design. The value in the brackets is the corresponding reliability index.

16
X. Huang, et al. Engineering Structures 209 (2020) 109953

not economical for designing with this model. To satisfy the pre-set reliability index, the calibrated design factor of
The calibration of the design factor for end debonding is only valid each design guideline are recommended in detail in Table 12. Ex-
for the CNR model due to the design end debonding equation. Since for cept for the FRP rupture in the JSCE guideline, the reliability in-
large βT, the increasing rate of β is so slow that the H function is difficult dexes with the calibrated values are close to the target reliability
to converge; thus, only the trends for βT = 3.0 and 3.2 are depicted in indexes for each failure mode in other guidelines. However, it is
Fig. 14. Results revealed the recommended γf for βT = 3.0 is equal to meaningless and uneconomical to adopt overly large partial safety
1.75. However, even when γf is equal to 4.0, it is not able to ensure that factors, especially for higher target reliability indexes. For example,
βT = 3.2. 4.0 for the end debonding in the CNR guideline, and 2.45 for the
The calibrated factors for each failure mode within each design FRP rupture in the JSCE guideline. In these cases, other options can
guideline are listed in Table 12. For each design guideline, the cali- be chosen such as applying additional anchorages to avoid end de-
brated factor and the corresponding reliability index (i.e., the value in bonding, increasing the FRP bonding area and reducing the FRP
the bracket) are given. (1) In general, within each guideline, the cali- thickness, which can reduce the probability of an IC debonding
brated ϕf or γf is different for each different failure mode. For a larger failure model. FRP rupture can be avoided by increasing the amount
βT, a larger safety margin is needed, and a smaller ϕf or a larger γf is of FRP appropriately.
recommended. (2) Based on previous analysis, the variation in design The effect of the live-to-dead load ratio α, resistance reduction factor
factors does not always affect reliability indexes, such as the concrete ϕf, or FRP partial safety factor γf on a reliability index is different for
crushing failure mode within GB 50608 2010, fib T5.1 2019, CNR-DT each design guideline. Generally, the reliability index decreases
200 R1/2012, TR 55 2012 and JSCE 2001 standards. In these cases, the gradually with the increase of ϕf and the decrease of γf. Besides, the
smallest design factor is recommended since a larger factor is mean- variation of γf is only valid for the FRP rupture failure mode. Owing
ingless and uneconomical in design. (3) For some cases, the re- to the formula of the debonding strain in the CNR-DT 200 R1/2012
commended design factor is too large, especially with a large βT. It is guideline, the reliability index increases as γf increases for de-
not economical and alternative measures should be adopted. For ex- bonding failure modes.
ample, additional anchorages can be applied to the FRP strengthened
RC beams to effectively avoid end debonding failure mode, as described Declaration of Competing Interest
in some guidelines; increasing the FRP bonding area and reducing the
FRP thickness can reduce the probability of the IC debonding failure The authors declare that they have no known competing financial
model. FRP rupture can be avoided by increasing the amount of FRP interests or personal relationships that could have appeared to influ-
appropriately. ence the work reported in this paper.

5. Conclusions Acknowledgments

In this paper, several representative design guidelines were sys- Funding: This work was supported by the National Natural Science
tematically studied and assessed with regard to the model error and Foundation of China [Grant numbers 51622808, 51778371, 51578337,
reliability index. They were ACI 440.2R 2017, CSA S806-02, GB 50608 and 51378338] and the Shenzhen Basic Research Project [Grant
2010, fib T5.1 2019, CNR-DT 200 R1/2012, TR 55 2012, and JSCE number GJHZ20170314105503899].
2001. They were chosen for their recommendations on handling FRP
strengthened RC beams controlled by concrete crushing, FRP rupture, References
IC debonding and end debonding. The model error of each design
guideline was assessed based on a database of references pertaining to [1] Shahnewaz M, Machial R, Alam MS, Rteil A. Optimized shear design equation for
this study. Then, the reliability index was calculated through the im- slender concrete beams reinforced with FRP bars and stirrups using Genetic
Algorithm and reliability analysis. Eng Struct 2016;107:151–65.
portance sampling (IS) method, in which the optimum density function [2] Nasrollahzadeh K, Aghamohammadi R. Reliability analysis of shear strength pro-
was determined by the first-order reliability method (FORM) in an visions for FRP-reinforced concrete beams. Eng Struct 2018;176:785–800.
iteration form. Finally, the calibration of the resistance reduction factor [3] Teng JG, Yao J. Plate end debonding in FRP-plated RC beams—II: Strength model.
Eng Struct 2007;29:2472–86.
or FRP partial safety factor was performed for different failure modes [4] Li P, Wu Y-F, Zhou Y, Xing F. Cyclic stress-strain model for FRP-confined concrete
with different target reliability indexes based on the LSA analysis, and considering post-peak softening. Compos Struct 2018;201:902–15.
new recommendations were provided for each design guideline. [5] Zhou Y, Guo M, Sui L, Xing F, Hu B, Huang Z, et al. Shear strength components of
adjustable hybrid bonded CFRP shear-strengthened RC beams. Compos B Eng
Therefore, based on the results presented in this study, several con-
2019;163:36–51.
clusions can be drawn: [6] Teng J, Chen J-F, Smith ST, Lam L. FRP: strengthened RC structures. Front Phys
2002:266.
[7] Bilotta A, Faella C, Martinelli E, Nigro E. Design by testing procedure for inter-
The model uncertainty of each failure mode is different for each
mediate debonding in EBR FRP strengthened RC beams. Eng Struct
standard’s design guideline. Based on the database, the Chinese GB 2013;46:147–54.
50608 2010 guideline seems to have the lowest model error for [8] Frangopol D. Life-cycle performance, management, and optimisation of structural
flexural failure prediction among the seven design guidelines. For IC systems under uncertainty: accomplishments and challenges1. Struct Infrastruct Eng
2011;7:389–413.
debonding failure, Europe’s fib T5.1 2019, UK’s TR 55 2012 and [9] Atadero R, Lee L, Karbhari VM. Consideration of material variability in reliability
Japan’s JSCE 2001guidelines show the highest model error in- analysis of FRP strengthened bridge decks. Compos Struct 2005;70:430–43.
dicating the most conservative, while China’s GB 50608 2010 [10] Ribeiro S, Diniz S. Reliability-based design recommendations for FRP-reinforced
concrete beams. Eng Struct 2013;52:273–83.
guideline is the least conservative prediction model with the lowest [11] Bojórquez J, Ruiz SE, Ellingwood B, Reyes-Salazar A, Bojórquez E. Reliability-based
model error. For end debonding failure, both fib T5.1 2019 and CNR optimal load factors for seismic design of buildings. Eng Struct 2017;151:527–39.
DT 200 R1/2012 are conservative guidelines. [12] Kim JH, Lee HS. Reliability assessment of reinforced concrete rectangular columns
subjected to biaxial bending using the load contour method. Eng Struct
Based on the test database, it is difficult to satisfy the corresponding 2017;150:636–45.
reliability target for some failure modes in several design guidelines [13] Ellingwood B. Development of a probability based load criterion for American
with the originally suggested design parameters (ϕf or γf) in a relia- National Standard A58: Building code requirements for minimum design loads in
buildings and other structures. US Department of Commerce, National Bureau of
bility assessment. The discrepancy between the reliability index and
Standards; 1980.
the target one is small in the cases of concrete crushing failure mode in [14] Huang X, Sui L, Xing F, Zhou Y, Wu Y. Reliability assessment for flexural FRP-
ACI, FRP rupture failure mode in TR. The discrepancy in the remaining strengthened reinforced concrete beams based on importance sampling. Compos B
Eng 2019;156:378–98.
cases is so large that the target reliability level cannot be satisfied.

17
X. Huang, et al. Engineering Structures 209 (2020) 109953

[15] Zeng Y, Botte W, Caspeele R. Reliability analysis of FRP strengthened RC beams systems for strengthening existing structures. CNR-DT200; 2004.
considering compressive membrane action. Constr Build Mater 2018;169:473–88. [35] Zheng H, Guang L. Reliability-based calibration of material partial factor of fiber
[16] Pham HB, Al-Mahaidi R. Reliablity analysis of bridge beams retrofitted with fibre reinforced polymer (FRP) rod. Eng Mech 2008;9:036.
reinforced polymers. Compos Struct 2008;82:177–84. [36] Coelho M, Neves L, Sena-Cruz J. Designing NSM FRP systems in concrete using
[17] Nowak A, Collins KR. Reliability of structures. 2nd ed. 2013. partial safety factors. Compos B Eng 2017;139:12–23.
[18] Gu Xianglin, Zhou Binbin, Zhang Weiping, Jiang Chao. Corrosion non-uniformity of [37] CSA. Canadian highway bridge design code (CHBDC); 2000.
steel bars and reliability of corroded RC beams. Eng Struct 2018;167:188–202. [38] FIB. Externally bonded FRP reinforcement for RC structures. fédération inter-
[19] Ghasemi SH, Nowak AS. Target reliability for bridges with consideration of ultimate nationale du béton Bulletin. 2001;14:51–8.
limit state. Eng Struct 2017;152:226–37. [39] Association CS. Design and construction of building components with fibre-re-
[20] 440 AC. Guide for the design and construction of externally bonded FRP systems for inforced polymers: Canadian Standards Association; 2002.
strengthening concrete structures. American Concrete Institute; 2017. [40] CEN E. Eurocode–Basis of Structural Design, 2002. IBN-BIN; 1990.
[21] GB-50608. Technical code for infrastructure application of FRP composites. China [41] ACI A. 440.2 R-08. Guide for the design and construction of externally bonded FRP
Planning Press China; 2010. systems for strengthening concrete structures ACI Committee. 440.
[22] du Béton FI. Externally Applied FRP reinforcement for concrete structures. Fib [42] Teng JG, Smith ST, Yao J, Chen JF. Intermediate crack-induced debonding in RC
Bulletin; 2019. beams and slabs. Constr Build Mater 2003;17:447–62.
[23] Party CSW. Design guidance for strengthening concrete structures using fibre [43] Lu XZ, Teng JG, Ye LP, Jiang JJ. Intermediate crack debonding in FRP-strengthened
composite materials: Concrete Society; 2012. RC beams: FE analysis and strength model. J Compos Constr 2007;11:161–74.
[24] Assi IM. Reliability-Based Load Criteria for Structural Design: Load Factors and [44] Atadero RA, Karbhari VM. Calibration of resistance factors for reliability based
Load Combinations. Thr University of Jordan; 2005. design of externally-bonded FRP composites. Compos B Eng 2008;39:665–79.
[25] CNR D. 200/R1 (2012) Guide for the design and construction of externally bonded [45] De Normalisation CE. Eurocode 2: Design of concrete structures—Part 1-1: General
FRP systems for strengthening existing structures. Advisory committee on technical rules and rules for buildings. Brussels, Belgium. 2004.
recommendation for construction of national research council, Rome, Italy. [46] 8110 B. Structural use of concrete Part 1, Code of practice for design and con-
[26] Engineers JSoC. Recommendations for upgrading of concrete structures with use of struction. British Standard Institution, London. 1997.
continuous fiber sheets. JSCE Concr Eng Ser 2001;41:31–4. [47] Teng JG. FRP strengthened RC structures. China Architecture & Building Press;
[27] Plevris N, Triantafillou TC, Veneziano D. Reliability of RC members strengthened 2005.
with CFRP laminates. J Struct Eng 1995;121:1037–44. [48] Said H, Wu Z. Evaluating and proposing models of predicting IC debonding failure.
[28] Okeil AM, El-Tawil S, Shahawy M. Flexural reliability of reinforced concrete bridge J Compos Constr 2008;12:284–99.
girders strengthened with carbon fiber-reinforced polymer laminates. J Bridge Eng [49] JSCE. Standard Specifications for Concrete Structures – Design; 2007.
2002;7:290–9. [50] 55 CSTR. Design guidance for strengthening concrete structures using fibre com-
[29] Wang N, Ellingwood BR, Zureick AH. Reliability-based evaluation of flexural posite materials. The Concrete Society, Crowthorne, UK; 2004.
members strengthened with externally bonded fiber-reinforced polymer compo- [51] Lu R, Luo Y, Conte JP. Reliability evaluation of reinforced concrete beams. Struct
sites. J Struct Eng 2010;136:1151–60. Saf 1994;14:277–98.
[30] Shi J, Wu Z, Wang X, Noori M. Reliability analysis of intermediate crack-induced [52] Yang Y. Reliability analysis of RC concrete beams strengthened with CFRP sheets. J
debonding failure in FRP-strengthened concrete members. Struct Infrastruct Eng Build Struct 2008;29:88–91.
2014;11:1651–71. [53] Desayi P, Rao KB. Reliability of reinforced concrete beams in limit state of
[31] Monti G, Santini S. Reliability-based calibration of partial safety coefficients for cracking— failure rate analysis approach. Mater Struct 1989;22:269–79.
fiber-reinforced plastic. J Compos Constr 2002;6:162–7. [54] GB50068. Unified standard for reliability design of building structures. Beijing,
[32] Zheng H, Guang L. Reliability-based calibration of material partial factor of fiber China: China Building Industry Press; 2001 [in Chinese].
reinforced polymer (FRP) rod. Eng Mech 2008;25:212–23. [55] Galambos TV, Ellingwood B, Macgregor JG, Cornell CA. Probability based load
[33] ASCE. Building Code Requirements for Structural Concrete (ACI 318-95) and criteria: assessment of current design practice. J Struct Div ASCE 1982;108:959–77.
Commentary (ACI 318R-95) by ACI Committee 318. J Arch Eng 2011;2. [56] JCSS. Probabilisticmodel Code, Part 2: Loadmodels; 2000.
[34] Council NR. Guide for the design and construction of externally bonded FRP [57] JCSS. Probabilistic Model Code, Part 1 Basis of Design; 2000.

18

You might also like