Engineering Structures 226 (2021) 111394

Contents lists available at ScienceDirect

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

Short communication

Refining the torsion design of fibered concrete beams reinforced with FRP
using multi-variable non-linear regression analysis for experimental results
Ahmed Deifalla
Associate Professor, Department of Structural Engineering and Construction Management, Future University in Egypt, (FUE), New Cairo City, Egypt

A R T I C L E I N F O A B S T R A C T

Keywords: There is very little guidance for practitioners regarding the torsion design of fiber-reinforced polymer (FRP) and
Concrete beam design hybrid reinforced concrete beams, especially those with fibered concrete (FC). The purpose of this study is to
Glass FRP (GFRP) improve the handful of methods used for predicting the torsion cracking and ultimate strength of reinforced
Carbon FRP (CFRP)
concrete (RC) beams with FRP reinforcements. An experimental database of 46 RC beams with FRP or hybrid
FRP
reinforcements and tested under torsion were compiled from seven different studies. Two proposed models (PM1
Torsion
Regression and PM2) based on the existing torsional model for FRP reinforced concrete beams and calibrated to fit the
SFC experimental data using multilinear non-linear regression. The cracking and ultimate torque predicted using the
PFC proposed models are more accurate compared with that calculated using selected ones existing in the literature.
CMC The PM1 is consistent with the existing design codes, yet more accurate compared to them. While the PM2 is non-
iterative yet capture the actual variation of the strength with the effective parameters.

differences exist between them, as follows [7,17–19]: (1) the torsion

cracks propagate in a spiral pattern while shear cracks propagate in a
1. Introduction parallel pattern, therefore, at the opposite faces of the beam, torsion
cracks spread in two opposite directions while shear cracks spread in the
With the fast-growing economy demand and consequently reduced same direction; (2) the strains due to shear forces are uniform and
safety factors, comes the refining of the existing design methods as the developed only across the two-dimension plane perpendicular to in the
ultimate goal of the research community, where many studies every day applied shear plane, while strains due to torsion moments are non-
are tackling [1]. Fiber reinforcement polymers (FRP) reinforcements are uniform and varies in the three dimensions; (3) in torsion, beam sides
being commonly used worldwide in many projects due to their high are subjected to bending and therefore the stress varies across the sec­
ability to resist corrosion [2–3]. Although providing the engineering tion both vertically and horizontally, while shear forces develop mostly
community with reliable design methods is an endeavor objective for uniform stresses; (4) the effective thickness of the equivalent hollow
researchers, most of the available guidelines and design codes include section that resists the applied torque varies with the applied load
detailed provisions only for shear and bending design of beams, except similar to the effective compression zone in the case of bending moment;
the Canadian Standards Association (CSA) [3]. The reasons behind such and (5) In flanged beams, the torsional moments are resisted by both the
a delay in including the torsion design provisions in design codes, is as web and the flange, while the shear forces are carried by the web only.
follows: 1) torsion is rare compared to shear and bending; 2) torsion can The most recent report by the ACI 445 [4] report addressed the
be neglected in many cases, if applied values are less than that of torsion behavior of members subjected to torsion including Plain Con­
cracking, and 3) the lack of investigation in this area compared to shear crete, RC, Prestressed Concrete (PC), and High Strength Concrete. It
and bending [3–8]. Since the complete understanding of the torsion discussed several approaches that can be implemented for torsion
design is paramount for the safety and economy of reinforced concrete analysis including the thin hollow tube analogy, elastic theory, and skew
(RC) members, thus, it is still under investigation [4–5,7–15]. bending theory as well as the space truss analogy were outlined. There is
Most of the research focus was directed towards the shear design of no agreement on the torsion design of conventional RC beams, where
FRP-reinforced concrete beams, thus, it is an advantage to study the many significantly different models exist [4]. In addition, the ACI-445
differences between shear and torsion behavior. Although beams resist report did not provide any guidance on either the torsion design of
both shear and torsion by forming a set of orthogonal internal diagonal FRP-reinforced beams or flanged cross sections (L-shaped or T-shaped).
forces one in tension and the other one in compression. However, many

E-mail address: [email protected].

https://doi.org/10.1016/j.engstruct.2020.111394
Received 8 November 2019; Received in revised form 7 September 2020; Accepted 29 September 2020
Available online 17 October 2020
0141-0296/© 2020 Elsevier Ltd. All rights reserved.
A. Deifalla Engineering Structures 226 (2021) 111394

Notation HD torsion design model proposed by Hassan and Deifalla

(2016)
Ac total area of concrete cross section (mm2) Long longitudinal
AL the longitudinal reinforcement resisting the torsion (mm2) pc perimeter of the concrete section (mm)
and the Σx2 y σcr term Ph perimeter of centerline of the outermost closed stirrup
Ao enclosed area inside the shear flow path (mm2) (mm)
Aoh enclosed area inside the centerline of the outermost closed PM1 first PM based on conventional torsion design
stirrup (mm2) PM2 second PM based on a power law equation
At the area of transversal reinforcement resisting the torsion Rft. reinforcements
(mm2) s The spacing between stirrups (mm)
b beam width (mm) SI the safety index, taken as the ratio between measured and
c clear cover to the closed stirrup (mm) calculated strength
CSA torsion design provisions of the CSA/S806 (2012) t the depth of the beam (mm)
d cross section depth (mm) T ultimate torsional strength of the cross section (N mm)
Ef Young’s Modulus of longitudinal reinforcement (GPa) Tcr cracking torsion of the cross section (N mm)
Eft Young’s modulus of transversal reinforcements in (GPa) Trans transversal
Esteel Young’s modulus of steel taken as 210 GPa tw wall thickness of idealized hollow section that is assumed
f’c concrete compressive strength (MPa) to be not greater than either the ratio of Aoh/ph or two times
fFt design strength of the FRP bar which shall be less than the minimum c (mm)
0.4fFu or 1200 for bent stirrups and fFu for bonded stirrups x the dimension of the short side of the cross section (mm)
(MPa) y the dimension of the long side of the cross section (mm)
fFu ultimate strength of the FRP bar which shall be less than θ the angle of inclination of the diagonal compressive stress
0.005Et in (MPa) with respect to the longitudinal axis of the element
Flc force carried by compressive longitudinal reinforcements σcr nominal concrete stress at the cracking torsion (MPa)
(N) σu nominal concrete stress at the ultimate torsion (MPa)
Flt Force carried by tensile longitudinal reinforcements (N) εl longitudinal strain at the mid-depth of section
fy yield stress of the steel reinforcements (MPa)

Fig. 1. Behavior of L-shaped beams under torsion: (a) isometric; (b) cracking models; and (c) ultimate models.

In flanged beams, stress concentration occurs at the intersect between Because it is highly dependent on the cross-section shape, relative di­
the flange and the web. In addition, thicker flanges reduce the warping mensions of the parts, and reinforcement detailing. Moreover, L-shaped
of the cross section under torsion, and thus it is more favorable in torsion beams are unsymmetrical, thus, adds more complexity to the behavior
to have flanges. However, the behavior of flanged beams under torsion is under torsion as shown in Fig. 1a [7,9–12,15]. For flanged concrete
more complicated when compared with rectangular cross sections. beams, there is no exact solution for the cracking torque. However, there

2
A. Deifalla Engineering Structures 226 (2021) 111394

Table 1
Summary of the details and reinforcements of the experimental database.
Ref. Beam fc’ fFu Dimensions (mm) Conc. type Long. Rfts Trans. Rfts. Tcr (kN. Tu (kN.
MPa MPa m) m)

[20] B5 25 400 Rectangle 200x100 NC Steel (3φ10) & GFRP (2 Steel 8@200 mm 1.8 2.7
φ10)
B6 25 400 NC Steel (3φ10) Steel 8@200 mm & GFRP 1.1 1.7
8@200 mm
B7 25 400 NC Steel (3φ10) & GFRP (2 Steel 8@200 mm & GFRP 1.58 2.2
φ10) 8@200 mm
B8 25 400 NC GFRP (3φ10) GFRP 8@200 mm 0.86 1.2
B9 25 400 NC GFRP (3φ10) GFRP 8@200 mm 1.09 1.5
B10 25 400 NC GFRP (3φ10) GFRP 8@100 mm 0.58 1.1
B11 25 400 NC GFRP (2φ10) GFRP 8@100 mm 0.96 1.3
B12 25 400 NC GFRP (3φ10) GFRP 8@200 mm 1.14 1.6
B13 25 400 NC GFRP (3φ10) GFRP 8@200 mm 1.02 1.3
B14 25 400 NC GFRP (3φ10) GFRP 8@200 mm 1.31 1.8
B15 25 400 NC GFRP (3φ10) GFRP 8@200 mm 1.0 1.4
B16 25 400 NC GFRP (2φ10) Steel 8@100 mm 1.33 1.9
B17 25 1500 NC CFRP (3φ10) Steel 8@200 mm 1.91 1.9
B18 25 1500 NC CFRP (2φ10) Steel 8@200 mm 1.28 1.9

[21] B1 30 360 Rectangle 200 × 100 SCC GFRP (4φ10) GFRP 8@100 mm 1.4 2.3
B2 45 360 SCC GFRP (4φ10) GFRP 8@100 mm 2.4 3.4
B3 90 360 HSCC GFRP (4φ10) GFRP 8@100 mm 3.0 4.5
B4 45 360 SCSFC (0.75%) GFRP (4φ10) GFRP 8@100 mm 3.0 3.8
B5 90 360 HSCSFC GFRP (4φ10) GFRP 8@100 mm 3.0 5.0
(0.75%)
B6 90 360 HSCSFC GFRP (4φ10) GFRP 8@100 mm 4.0 5.1
(1.5%)

[22] BC120 38.5 1562 Rectangle 600 × 250 NC CFRP (8φ13) CFRP 9.5@120 mm 30.45 62.9
BC180 38.5 1562 NC CFRP (8φ13) CFRP 9.5@180 mm 29.87 49.4
BC240 38.5 1562 NC CFRP (8φ13) CFRP 9.5@240 mm 27.35 39.4
BC300 38.5 1562 NC CFRP (8φ13) CFRP 9.5@300 mm 28.65 35.7

[23] LB2 25 360 L-shaped 350 × 150/150 × NC GFRP (2φ12) Steel 6@143 mm 2.1 8.4
LB3 25 400 400 NC GFRP (2φ12) GFRP 6@143 mm 2.1 10.0
LB4 25 400 NC GFRP (2φ12) GFRP 8@143 mm 2.1 14.0
LB5 25 400 NC GFRP (2φ12) GFRP 10 @143 mm 2.1 20.0

[23] BG120 41.47 948 Rectangle 600 × 250 NC CFRP (7φ19) GFRP 9.5@120 mm 27.46 52.7
BG180 41.47 948 NC CFRP (7φ19) GFRP 9.5@180 mm 26.19 41.8
BG240 38.5 948 NC CFRP (7φ19) GFRP 9.5@240 mm 26.14 34.2
BG300 41.47 948 NC CFRP (7φ19) GFRP 9.5@300 mm 25.98 29.9

[25] BG-W 41.47 948 Rectangle 600 × 250 NC CFRP (8φ13) – 23.11 25.0
BG- 41.47 948 NC CFRP (8φ13) GFRP 9.5@120 mm 27.46 52.7
120
BG-60 39.25 948 NC CFRP (8φ13) GFRP 9.5@60 mm 27.76 56.9
BC-W 38.5 1562 NC CFRP (8φ13) – 28.62 34.1
BC120 38.5 1562 NC CFRP (8φ13) CFRP 9.5@120 mm 30.45 62.9
BC60 39.25 1562 NC CFRP (8φ13) CFRP 9.5@60 mm 30.14 69.3

[26] S-1 42.5 700 Rectangle 150 × 200 PFC GFRP (4φ14) GFRP 8@50 mm 2.88 5.5
S-2 42.5 700 PFC GFRP (8φ14) GFRP 8@50 mm 2.78 5.8
F-1 40.2 700 PFC GFRP (4φ14) GFRP 8@50 mm 2.88 6.7
F-2 40.2 700 PFC GFRP (8φ14) GFRP 8@50 mm 3.02 7.0
F-3 40.2 700 CMC GFRP (4φ14) GFRP 8@50 mm 2.78 5.7
F-4 40.2 700 CMC GFRP (8φ14) GFRP 8@50 mm 2.55 6.4
E-1 32.8 700 CMC GFRP (4φ14) GFRP 8@50 mm 2.28 8.2
E-2 32.8 700 CMC GFRP (8φ14) GFRP 8@50 mm 2.2 8.7

are a few approximations either Hollow Tube Analogy or Bach’s For­ [16–19,24]. Very few investigations for the design and analysis of FRP
mula as shown in Fig. 1b. In addition, the hollow tube analogy space reinforced beams under torsion were conducted by Deifalla and co-
truss analogy is used to predict the ultimate torsion strength as shown in workers [17–19]. A simplified torsional strength model was proposed
Fig. 1c. for the case of GFRP reinforced concrete beams, where, the difference
The recent FRP guidelines ACI 440 [2] did not offer any guidance on between the young’s modulus of FRP and Steel reinforcements was
torsion design of FRP-reinforced concrete beams. Most research efforts included in the calculation of the angle of inclination of the diagonal
focused on the area of FRP-reinforced beams under flexure and shear, strut, which provided more accurate predictions [17]. In another study,
however, very limited works have addressed torsion behavior of FRP- a full torsion behavior model was developed and proposed for the case of
reinforced concrete beams. In 2012, the Canadian Standards Associa­ FRP reinforced beams [18]. Later, another study for the torsion design of
tion issued a standard for the design and construction of building both GFRP and Carbon FRP reinforced beams confirmed the significance
components with FRP, CSA [3], which will be referred to as CSA. It is the of including the FRP stiffness in calculating the diagonal strut angle of
first and only code that provided torsion provisions for the design of inclination in the space truss (θ) as well as a significant difference be­
concrete members reinforced with FRP. A hand full of studies tackled the tween the cracking torque of flanged beams and rectangular beams was
torsion design and analysis of FRP-reinforced concrete beams found [19]. However, it was indicated that more experimental results

3
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 2. Experimental database distribution for the following parameters (a) fc′ , (b) y/x, (c) s/d, (d) Trans. Rft. Material, (e) Long. Rft. Material, (f) Concrete Type.

will refine and generalize the proposed design provisions [19]. In a [7,12,17–19,23,27–30]. The current study aims to refine the existing
study by Mohamed and Benmokrane [24], the CSA torsion provisions torsion design models of FRP reinforced concrete beams. The available
were found to be conservative except for the well-confined RC beams. experimentally tested FRP-reinforced concrete beams, while subjected
However, these conclusions were based on assuming θ equal to 45◦ and to torsion were gathered. A total of 46 concrete beams reinforced with
only four tested beams. Razaqpur [16] evaluated the accuracy of the FRP or steel-hybrid reinforcements were compiled from seven different
CSA for the torsion design of FRP RC/PC members. It was found that CSA studies [20–26]. Available design codes and models in the literature
reasonably predicts the torsion strength. In addition, the ACI-318 [34] were compiled. Two models were selected and calibrated using multi­
and AASHTO LRFD [31] torsion cracking formula is reasonable [16]. variable nonlinear regression to fit the experimental data. The proposed
Over the last 3 decades, a very limited number of studies tackled the and available models were used to calculate the cracking and torsion
torsion of FRP reinforced concrete beams, because this issue is complex strength. Which is compared with that observed experimentally. The
and requires further investigations. With the availability of more torsion strength predicted using the proposed models is closer to that
experimental data, the available models can be refined and updated experimentally measured, compared with that calculated using selected
[20–26]. The current study is part of an extensive project for investi­ design code and model available in the literature.
gating various types of reinforced concrete under combined loading

4
A. Deifalla Engineering Structures 226 (2021) 111394

2. Experimental database profile reinforcement used to resist torsion is expressed below such that:
2Ao ϕFfFt At
An experimental database was gathered from seven different T= cotθ ≤ 0.20fc’ (1.7Aoh tw ) (2)
s
experimental, which are investigating concrete beams reinforced with
FRP and/or steel reinforcements investigations [20–26]. Table 1 lists the The expression represents a simplification of actual behavior,
properties and the experimental results of over forty tested beams. Fig. 2 considering the calculated torsional strength is controlled by the
(a–f) show the distribution of concrete compressive strength, cross sec­ strength of the transverse reinforcement, independent of concrete
tion aspect ratio, the ratio between the stirrup spacing and cross section strength and longitudinal reinforcement. The strut angle of inclination
depth, type of material used for transversal and longitudinal re­ with respect to the beam’s longitudinal axis θ shall be calculated as:
inforcements, and the concrete type, respectively, among tested beams. θ = 30 + 7, 000εL (3)
As shown in Fig. 2, the compressive strength of tested beams varied from
25 MPa to 90 MPa. The aspect ratio (y/x) of cross section of tested beams where εL = 0.225 AoTP
Ef AL and θ shall not be taken as greater than 60 nor
h ◦
ranged between 1.3 and 2.4. The stirrup spacing to depth (s/d) ratio of
less than 30◦ . In addition, the CSA provides for proportioning the lon­
tested beams varied from 0.25 to 1.0, as shown in Fig. 2(c). Also, Fig. 2(d
gitudinal reinforcement on the flexural tension and compression sides so
& e) shows the type of longitudinal and transversal reinforcement used
that the factored resistance of the reinforcement in all sections, taking
in the tested beam specimens in the experimental database, respectively,
into account the stress that can develop in these longitudinal re­
it is clear that the majority reinforcement used in the tested beams was
inforcements, shall be greater than or equal to the force Flt , as follows:
GFRP reinforcement. Fig. 2(f) presents that 76% of tested beams were
( )
constructed using a normal concrete mix, while only 24% were con­ 0.45Tph
structed using a special concrete mix including Steel Fiber Concrete Flt = Flc = 1.3 (4)
2Ao
(SFC), Polypropylene Fiber Concrete (PFC), and Engineered Cementi­
tious Composite Concrete (CMC), and Self-Compacted Concrete (SCC).
3.2. Hassan and Deifalla [19]
Using adhesively bonded stirrups, increased the allowable design stress
of stirrups as it eliminated the strength reduction due to the bent cor­
3.2.1. Cracking torque
ners, thus, stirrups failed at a much larger strain compared to the bent
Due to the limited number of experimentally tested FRP RC beams
stirrups.
under torsion, Hassan and Deifalla [19] proposed the use of the CSA-
A23.4 [32] cracking formula, while adapted the PCI [33] cracking for­
3. Available design methods in the literature mula for the flanged beams as follows.

Recent years have seen valuable research work to investigate the

⎧ ( )√̅̅̅̅’
⎨ 0.38 A2c /pc f c for rectangular beams
performance of FRP as primary reinforcement for concrete structures Tcr = √̅̅̅̅ ∑ (5)
⎩
under axial, shear, and flexural loads [2–3,30]. However, the torsional 0.067 f ’c x2 y for flanged beams
behavior of concrete members reinforced with FRP bars and stirrups has
not been fully investigated. Yet, due to a lack of research, North
3.2.2. Ultimate torque
American codes and design guidelines ACI 440 [1], AASHTO LRFD [31],
Modifications for the CSA [2] were proposed by Hassan and Deifalla
CSA-A23.4 [32], PCI [33], and ACI-318 [34], except CSA [2], do not
[19] as follows, where Eq. (3) for inclination angle of the strut was
include torsion provisions for design of RC members with FRP rein­
modified, such that:
forcement. For cracking and ultimate torsion, the following formulas ( )
were selected: CSA [2] and Hassan and Deifalla [19]. θ = 29 + 7000εL Eft /Esteel (6)

where the longitudinal strain is such that:

3.1. CSA [2] ⎡ ⎤
0.2% for bent steel stirrups
TPh
3.1.1. Cracking torque εL = 0.225 ⎣
≤ 0.4% for GFR bent stirrups ⎦ (7)
Ao Ef AL
0.8% for GFRP bomded stirrups
CSA only offer ultimate torsion design provisions for FRP-reinforced
concrete beams, thus the cracking torque formula of the conventional And Eqs. (2), (4) were kept the same as they are based on hollow tube
RC design CSA A23.4 [32] will be implemented. Thus, the cracking space truss analogy.
torsional moment (Tcr) is computed, as follows:
( )√̅̅̅̅’ 4. Proposed models
Tcr = 0.38 A2c /pc fc (1)
4.1. Introduction
3.1.2. Ultimate torsion
CSA [2] is the most recently issued Canadian code on the design and Pioneering work by Belarbi and Hsu [35] showed that the true strain
construction of building components with FRPs. Several modifications distribution is non-uniform and varies depending on the examined cross
were applied on the torsion provisions of the conventional steel rein­ section (i.e. at stirrup or at concrete crack), the transversal reinforce­
forced concrete design CSA A23.4 [32] in order to account for the FRP ment ratio as well as the concrete crack angle. Based on these experi­
reinforcements as follows: 1) the mechanical properties of FRP rein­ mental observations, Rahal [8] was the first to propose power law
forcement such as the modulus of elasticity and the ultimate FRP equation for effective parameters in order to capture such highly
strength were implemented; 2) the angle of inclination of the nonuniform behavior. In addition, pioneering work by Mitchell and
compression strut were taken 45 in order to simplify the calculation of Collins 1974 [36] and Hsu and Mo 1985 [37] showed that the concrete
◦

the longitudinal forces; 3) the diagonal compressive stress were reduced strut in the walls of beams subjected to torsion are under bending with
to 80% of that with the steel; 4) the upper limit of the longitudinal strain strain varying in the horizontal direction. Thus, the true torsion
of 0.3% was removed. In CSA [2], the ultimate torsional strength (T) is behavior of RC beams is complicated and non-uniform in both the ver­
provided based on the strength capacities of torsional FRP stirrups. The tical and horizontal direction [35–37]. Moreover, the idealized strut
hollow-tube, space-truss analogy used in CSA [2] provisions to relate the angle of inclination in the space truss model cannot be arbitrary
required torsional strength T to the area of one leg of closed transverse assigned, however, it is variable and function in longitudinal strain,

5
A. Deifalla Engineering Structures 226 (2021) 111394

f A
reinforcement ratio ( ys t + fFusAt ) and longitudinal reinforcement forces
(fFu Al ) [3,9,17–19]. In order to capture such a behavior, the develop­
ment of a physically sound model, which is calibrated using experi­
mental results. Using multi-variable nonlinear regression analysis is the
best method for calibrating physically sound mechanical models. Thus,
modified models were suggested and examined using statistical mea­
sures. Two models will be proposed, the first is based on the conven­
tional methods and calibrated based on the experimental data, while the
second is based on a power equation for all the effective parameters
calibrated with the experimental data.

4.2. First proposed model (PM1)

The PM1 is based on the conventional hollow tube analogy for

cracking torque combined with the variable angle space truss model for
ultimate torque [3–4,17–19]. While multivariable non-linear regression
is implemented in order to refine the design equations to fit the exper­
imental database, the following design provisions are proposed:

4.2.1. Cracking torque

The PM1 model was proposed based on the hollow tube analogy for
cracking torque, thus, only the cracking equation constant was best
fitted as follows:
⎧ ( )√̅̅̅̅’
⎨ 0.35 A2c /pc f c for rectangular beams for rectangular beams
Tcr = ( 2 )√̅̅̅̅’
⎩
0.10 Ac /pc f c for flanged beams
(8)
It is worth noting that the proposed equation is very close to the ACI-
318 [33], which is developed for conventional RC beams under torsion
Fig. 3. Iterative algorithm for ultimate torque calculation using CSA, HD,
for rectangular beams, confirming the findings by Razapur et al. [16].
and PM1.
While for the case of Flanged beams, a new different equation was
proposed, which confirms the findings by many researchers that torsion
young’s modulus of the FRP reinforcements [19]. Last but not least, behavior is affected significantly by cross section shape [7,9–12,17–19].
cracking torsion strength is affected by many variables including area of
cross section (Ac ), perimeter of cross section (pc ), and the concrete 4.2.2. Ultimate torque
compression stress (fc’ ) [3,9,17–19]. While the ultimate torque is For ultimate torque, the hollow tube analogy and the variable angle
affected by the following parameters: area of cross section (Ac ), perim­ space truss were adopted, thus Eqs (2), (4, (7) were kept the same as they
eter of cross section (pc ), the concrete compression stress (fc’ ), transversal are based on a physically sound mechanical model, while Eqs (6) for the
angle of the inclination of the compression strut developed by Hassan

Fig. 4. Regression analysis for the model PM1 (a) cracking torque and (b) ultimate torque.

6
A. Deifalla Engineering Structures 226 (2021) 111394

Table 2 4.4. Comparison between the proposed models and selected models
Regression statistics.
Parameter Eq. (10) Eq. (11) The safety of available and proposed models was investigated by
calculating the safety index, which is calculated as the ratio between
Multiple R 0.96 0.983
R Square 0.92 0.967 measured strength and that calculated using various models (SI). In the
Adjusted R Square 0.92 0.964 current study, the Demerit Points Classification (DPC) proposed by
Standard Error 0.37 0.258 Collins and used in many investigations [6,27,38–39], where the model
Observations 46 46 predictions are categorized depending on the SI values into seven cat­
egories ranging from extremely dangerous (SI < 0.5), dangerous (SI =
and Deifalla [16–19] was adapted. With the availability of a larger 0.5–0.65), low safety (SI = 0.65–0.85), approximate safety (SI =
database, a slightly modified equation for strut angle of inclination is 0.85–1.30), conservative (SI = 1.30–2.0), to extremely conservative (SI
proposed using nonlinear regression, instead of Eq. (6), as follows. > 2.0). The number of beams failing in each category as well as the DPC
total score was calculated for each model and presented in Table 5. It is
θ = 29 + 23εL Eft (9) clear that all models have almost the same score, with the proposed
Fig. 3 shows the flow chart for the iterative technique used to model slightly better. Thus, it is fair to say that the balance between
calculate the ultimate torsion strength using Eqs (2), (4), (7), (9). Fig. 4 safety and economy of the proposed models is the same and even slightly
(a–b) show the experimentally measured strength versus the calculated better than the existing one. In addition, statistics of computed SI for
cracking and ultimate strength, respectively, using Eq (8) and Eqs (2), both cracking and ultimate torque using all models have been calcu­
(4), (7), (9). Where the regression coefficient (R2) was 0.99 and 0.897 for lated, which included the following parameters including mean, coef­
the calculated cracking and ultimate strength, respectively, compared to ficient of variation, lower 95% as well as correlation coefficient of the SI.
the measured ones. The average is used to evaluate the accuracy of the model, the closer the
SI to unity, the more accurate are the model predictions. The consistency
of the model predictions is evaluated based on the value of the coeffi­
cient of variation, the smaller this value, the better consistency is the
4.3. Second proposed model (PM2)
model predictions. In addition, the lower limit with level of confidence
95%, which evaluate the conservativeness of the model, required in
In a previous study by Deifalla [17] regarding the Glass FRP RC
order to be included in design codes. Last but not least, the correlation
beams, adapted the power-law equation for torsion, which was first
proposed by Rahal [8] and based on experimental findings by Belarbi
and Hsu [35]. However, with the increase in the experimental testing of Table 4
FRP RC beams under torsion, it is worth revisiting this concept using Coefficient regression table.
multi-variable non-linear regression. In the second proposed model Eq. Parameter Coeff. Std. t P-value Lower Upper
(PM2), the concept was adapted and implemented for both the cracking Error 95% 95%

and ultimate torsion of FRP RC beams. The nonlinear regression is used (10) Intercept 0.19 0.78 − 2 0.04 6E− 4 0.84
in order to be able to capture nonuniform stress and strain distribution X Variable 1.15 0.17 7 1.34E− 08 0.82 1.49
1
across the cross sections and stirrups [8,18,35]. For cracking torque,
X Variable 0.9 0.043 20 9.62E− 24 0.79 0.963
nonlinear regression was employed with the inclusion of all effective 2
parameters area of cross section (Ac ), perimeter of cross section (pc ), and
(11) Intercept 0.16 0.65 − 3 0.008 7.7E− 4 0.32
the concrete compression stress (fc’ ), thus, after some simplification, the X Variable 0.67 0.14 5 1.97E− 05 0.39 0.95
following cracking torque formulas were developed, such that: 1
⎧ X Variable 1.3 0.06 22 2.95E− 24 1.2 1.4
⎨ 0.19( A2 /p )0.9 ( f ’ )1,15 for rectangular beams 2
(10)
c c c
Tcr = X Variable 0.10 0.028 4 0.0006 0.05 0.16
⎩ 0.09( A2 /p )0.9 ( f ’ )1,15 for flanged beams
c c c 3
X Variable 0.12 0.083 1.5 0.045 0.05 0.29
While for ultimate torque, nonlinear regression was employed with 4
the inclusion of all effective parameters: area of cross section (Ac ),
perimeter of cross section (pc ), the concrete compression stress (fc’ ),
f A
transversal reinforcement ratio ( ys t + fFusAt ) and longitudinal reinforce­ Table 5
ment forces (fFu Al ), thus, after some simplification, the following ulti­ Evaluation of various models using DPC.
mate torque formula was developed, such that: Category SI DP CSA HD PM1 PM2
( )0.1 Extremely dangerous <0.5 10 0 0 0 0
( )0.67 fy At fFu At
T = 0.16 fc’ (Ac )1.3 + (fFu Al )0.12 ≤ 0.34Aoh tw fc’ (11) Dangerous 0-5-0.65 5 2 4 5 5
s s Low safety 0.65–0.85 2 9 12 14 20
Approximate safety 0.85–1.3 0 42 48 54 56
Output results from regression analysis are shown in Tables 2–4, Conservative 1.3–2 1 23 17 14 11
which included regression statistics, ANOVA, and coefficient regression Extremely conservative >2 2 16 11 5 0
table, respectively. Where the regression coefficient (R2) was 0.92 and Total score 83 83 77 76
0.964 for Eqs. (10)–(11), respectively.

Table 3
ANOVA.
Parameter Eq. (10) Eq. (11)

Degree of freedom Sum of square Mean square F Degree of freedom Sum of square Mean square F

Regression 2 74 37 267 4 80.7 20.2 303

Residual 43 6 0.14 – 41 2.7 0.067 –
Total 45 80 – – 45 83.5 – –

7
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 5. The cracking torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

coefficient, where it is a statistical measure of the strength of the rela­ proposed model, respectively, which is higher than the factor of safety
tionship between two variables. A correlation coefficient value close to considered by design codes. In addition, SI values for proposed models
±1.0 shows a perfect correlation. A correlation coefficient value close to are mostly between the category of low safety to the category of con­
0.0 shows weak or no linear relationship between the two variables. servative. While CSA and HD models have several SI values within the
For cracking torque, Fig. 5(a–d) show the SI using all models. Table 6 category of extremely conservative. Coefficient of correlation is almost
shows the SI for cracking torque of all beams. The proposed models are perfect for all models. Moreover, the strut angle of inclination from the
more accurate, consistent and reasonably conservative with average experimental testing results and that calculated using all available
almost unity and coefficient of variation 24% as well as a lower limit models is presented in Table 7.
with confidence level 95% value of 0.95 and 0.88 for the first and second
proposed mode, respectively, which is higher than the factor of safety 5. Effect of various parameters on the torsion design
considered by design codes. In addition, SI values for proposed models
are in mostly in the category of low safety to conservative. While CSA In order to investigate the effect of various design parameters on the
and HD models have several SI values within the category of extremely level of safety and consistency of the available models, the calculated SI
conservative. Coefficient of correlation is almost perfect for all models. values using all models are plotted against the values of the effective
For ultimate torque, Fig. 6(a–d) show the SI using all models. Table 7 parameters. Effective parameters included concrete compressive
shows the SI and θ for ultimate torque of all beams. The proposed models strength, longitudinal and transversal reinforcement index, cross-
are more accurate, consistent and reasonably conservative with average section area to perimeter ratio, and depth. The ideal model behavior is
almost unity and coefficient of variation 24% as well as a lower limit indicated on figures as a flat line solid at a constant value of unity, and a
with confidence level 95% value of 0.95 and 0.88 for the first and second best-fit dotted line is plotted for each model. The closer the dotted line to

8
A. Deifalla Engineering Structures 226 (2021) 111394

Table 6 respect to the concrete compressive strength, while CSA is the most
SI for cracking torsion moment for all tested beam. scattered. In addition, all models underestimate the strength.
Study Beam CSA HD PM1 PM2 PM1-simplified

[20] B5 0.70 0.70 0.66 0.55 1.54 5.2. Hollow tube thickness
B6 1.15 1.15 1.07 0.91 0.94
B7 0.80 0.80 0.75 0.63 1.35 The hollow tube analogy is the base of the torsion design worldwide,
B8 1.47 1.47 1.37 1.16 0.74 where the core strength is neglected. The thickness of the hollow tube is
B9 1.16 1.16 1.08 0.91 0.93
an important parameter that influences the torsional strength of beams.
B10 2.18 2.18 2.04 1.72 0.50
B11 1.32 1.32 1.23 1.04 0.82 For cracking torque, Fig. 9(a–d) show the SI for cracking torque using all
B12 1.11 1.11 1.04 0.87 0.98 models versus the Hollow tube thickness. All models are reasonably
B13 1.24 1.24 1.16 0.98 0.87 consistent with respect to the hollow tube thickness, except for the CSA.
B14 0.97 0.97 0.90 0.76 1.12
For ultimate torque, Fig. 10(a–d) show the SI for ultimate torque using
B15 1.27 1.27 1.18 1.00 0.86
B16 0.95 0.95 0.89 0.75 1.14 all models versus the Hollow tube thickness. All models are reasonably
B17 0.66 0.66 0.62 0.52 1.64 consistent with respect to the hollow tube thickness, except for the CSA.
B18 0.99 0.99 0.92 0.78 1.10

[21] B1 0.99 0.99 0.92 0.88 1.10 5.3. Depth of cross-section

B2 0.71 0.71 0.66 0.82 1.53
B3 0.80 0.80 0.75 1.46 1.36 The depth has a significant effect on the torsion design of beams,
B4 0.57 0.57 0.53 0.65 1.92
B5 0.80 0.80 0.75 1.46 1.36
which is recognized by most design codes [3,34]. For cracking torque,
B6 0.60 0.60 0.56 1.09 1.81 Fig. 11(a–d) show the SI for cracking torque using all models versus the
depth. For ultimate torque, Fig. 12(a–d) show the SI for ultimate torque
[22] BC120 1.02 1.02 0.95 0.74 1.06
BC180 1.04 1.04 0.97 0.75 1.04 using all models versus the depth. PM2 is the most consistent while the
BC240 1.14 1.14 1.06 0.82 0.95 CSA is the least consistent model.
BC300 1.09 1.09 1.01 0.78 1.00

[23] LB2 3.91 2.33 1.03 1.02 0.28 5.4. Longitudinal reinforcement index (Ef Al /Ac )
LB3 3.91 2.33 1.03 1.02 0.28
LB4 3.91 2.33 1.03 1.02 0.28
The longitudinal reinforcement significantly contributes to the ulti­
LB5 3.91 2.33 1.03 1.02 0.28
mate torsion resistance of the cross-section through resisting the hori­
[24] BG120 1.18 1.18 1.10 0.89 0.92 zontal component of the diagonal tension forces developed due to
BG180 1.24 1.24 1.15 0.93 0.88
torsion. For ultimate torque, Fig. 13(a–d) show the SI for ultimate torque
BG240 1.19 1.19 1.11 0.86 0.91
BG300 1.25 1.25 1.16 0.94 0.87 using all models versus the Longitudinal reinforcement index. All
models are reasonably consistent with respect to the longitudinal rein­
[25] BG-W 1.40 1.40 1.31 1.06 0.77
BG-120 1.18 1.18 1.10 0.89 0.92
forcement ratio; however, the SI of the CSA is directly proportional with
BG-60 1.14 1.14 1.06 0.83 0.96 to the values of longitudinal reinforcement.
BC-W 1.09 1.09 1.02 0.79 1.00
BC120 1.02 1.02 0.95 0.74 1.06
BC60 1.05 1.05 0.97 0.76 1.04
5.5. Transversal renforcement index (Eft Abst )

[26] S-1 1.11 1.11 1.03 1.13 0.98

The transversal reinforcement significantly contributes to the ulti­
S-2 1.15 1.15 1.07 1.17 0.95
F-1 1.08 1.08 1.00 1.06 1.01 mate torsion resistance of the cross-section through resisting the vertical
F-2 1.03 1.03 0.96 1.01 1.06 component of the diagonal tension forces developed due to torsion. For
F-3 1.11 1.11 1.04 1.10 0.97 ultimate torque, Fig. 14(a–d) show the SI for ultimate torque using all
F-4 1.21 1.21 1.13 1.20 0.89
models versus the transversal reinforcement ratio. CSA and HD are less
E-1 1.23 1.23 1.14 1.06 0.88
E-2 1.27 1.27 1.19 1.10 0.85
conservative for high transversal reinforcement index. While PM1 and
PM2 are consistent.
Mean 1.33 1.20 1.01 0.95 0.99
Coefficient of Variation 63% 37% 24% 24% 35%
Lower 95% 1.09 1.07 0.95 0.88 0.89 5.6. Cracking angle
Correlation coefficient 0.99 0.99 0.99 0.99 0.99
The cracking angle report in the literature or extracted from the
cracking pattern reported in the literature and the calculated strut angle
the solid line indicates consistent model predictions with regard to the
of inclination are listed in Table 7. Although ideally, they should be
investigated parameter.
equal, however, this is not the case. The calculated angle was calculated
to fit the experimentally observed behavior using nonlinear regression.
More experimental testing is needed, especially with a pre-set cracking
5.1. Concrete strength
pattern to further investigate such a relationship for FRP RC beams,
similar to the pioneering work by Belarbi and Hsu [35]. Similar
The concrete strength is an important parameter that significantly
conclusion were arrived at by other researchers in shear of FRP-
affects the torsion strength. It indirectly affects the cracking torque as
√̅̅̅̅ reinforced concrete beams [40].
the tensile strength related to it as f ’c and directly affects the capacity of
the compression strut, thus setting the maximum limit of the ultimate 6. Further simplification for practical application
torsion. For the cracking torque, Fig. 7(a–d) show the SI for cracking
torque using all models versus the concrete compressive strength. It is For simplicity, a more practical unified equation (PM1-SIMPLIFIED)
clear that all models are fairly consistent with experimentally observed for both rectangular and flanged FRP RC beams will be examined later in
behavior. For ultimate torque, Fig. 8(a–d) show the SI for ultimate tor­ this study. The cracking torque formula for the PM1 model was
que using all models versus the concrete compressive strength. It is clear simplified to one equation irrelevant of the cross-section shape, which is
that ultimate torsion predication using PM1 is the least scattered with similar to the CSA-A23.4 [32], PCI [33], and ACI-318 [34], cracking

9
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 6. The ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

10
A. Deifalla Engineering Structures 226 (2021) 111394

Table 7
SI and θ for ultimate torsion moment using different methods for all tested beam.
Study Beam Measured CSA HD PM1 PM2

θ (deg.) SI θ (deg.) SI θ (deg.) SI θ (deg.) SI

[20] B5 40 1.74 31 1.74 32 1.73 31 1.71

B6 40 0.76 33 0.74 35 0.71 33 0.98
B7 40 0.91 32 0.89 33 0.86 32 1.24
B8 40 1.65 37 1.20 32 1.15 31 0.72
B9 40 1.79 34 1.40 31 1.35 30 0.86
B10 40 0.92 43 0.62 34 0.66 32 0.62
B11 40 1.26 47 1.10 34 1.17 32 0.73
B12 40 1.99 34 1.55 31 1.49 30 0.96
B13 40 0.91 45 0.76 33 0.80 32 0.76
B14 40 1.03 41 0.89 32 0.70 32 1.02
B15 40 1.28 41 1.28 42 1.17 39 0.86
B16 40 1.31 53 1.31 54 1.14 49 1.08
B17 41 1.19 31 1.19 32 1.18 31 1.16
B18 42 1.42 36 1.42 37 1.35 34 1.16

[21] B1 40 1.57 37 1.28 32 1.22 31 1.02

B2 40 1.91 39 1.50 33 1.43 31 1.11
B3 40 1.16 51 2.50 32 2.39 31 0.96
B4 40 0.85 54 1.68 33 1.60 31 1.24
B5 40 1.03 56 1.89 33 2.02 31 1.03
B6 40 1.30 51 2.82 32 2.69 31 1.08

[22] BC120 45 1.37 42 1.19 39 1.08 37 1.34

BC180 45 1.44 39 1.29 37 1.18 35 1.10
BC240 45 1.43 37 1.30 35 1.21 33 0.90
BC300 45 1.55 36 1.42 34 1.33 33 0.83

[23] LB2 30 2.02 53 2.02 54 1.75 49 0.83

LB3 25 4.01 43 2.80 34 1.17 35 0.98
LB4 30 3.92 49 3.92 34 1.42 36 1.30
LB5 35 4.39 55 5.61 34 2.03 36 1.77

[24] BG120 45 2.02 40 1.33 33 1.26 32 0.97

BG180 45 2.17 37 1.52 32 1.45 31 0.80
BG240 45 2.23 35 1.62 32 1.55 31 0.71
BG300 45 2.35 34 1.75 31 1.68 30 0.60

[25] BG-W 45 0.77 42 0.77 33 0.83 31 1.02

BG-120 45 2.02 40 1.33 33 1.26 32 0.97
BG-60 45 1.36 46 0.80 36 0.74 34 1.01
BC-W 45 1.09 41 1.09 37 1.17 34 1.52
BC120 45 1.51 45 1.29 42 1.16 38 1.34
BC60 45 1.11 53 0.90 48 0.78 44 1.35

[26] S-1 45 1.30 52 0.82 37 0.59 36 0.87

S-2 45 1.17 47 0.63 37 0.63 34 0.88
F-1 45 1.59 52 0.99 38 0.76 36 1.10
F-2 45 1.41 47 0.80 37 0.80 34 1.09
F-3 45 2.07 44 1.10 36 1.02 34 1.00
F-4 45 2.05 40 1.17 34 1.10 33 1.07
E-1 45 1.96 52 1.17 39 1.15 35 1.54
E-2 45 1.75 47 1.22 36 1.22 34 1.55

Mean 1.33 1.65 1.47 1.24

Coefficient of Variation 63% 47% 60% 36%
Lower 95% 1.09 1.43 1.21 1.11
Correlation coefficient 0.94 0.95 0.95 0.95

torque formula, such that: 7. Conclusions

( )√̅̅̅̅’
Tcr = 0.35 A2c /pc fc (12) Two models for the torsion design of FRP-reinforced concrete beam
were proposed as follows: (1) the first proposed model, which is based
Table 6, it is clear that this simplification is practical and accurate on the conventional hollow tube analogy and variable angle space truss,
compared to experimentally measured cracking strength, however, was calibrated to fit the experimental data using multivariable nonlinear
inconsistent with a higher coefficient of variation value of 35% regression analysis. (2) The second proposed model, which is based on a
compared to 24% for PM1. In addition, SI for the cracking torque for the power-law equation for all the effective parameters, was calibrated to fit
L-shaped beams is less than 0.5, which is in the extremely dangerous the experimental data using multivariable nonlinear analysis. The
category. Thus, it is recommended to use Eq. (12), neglect the effect of cracking and ultimate torsion strength predicted by the proposed models
the flange. and the existing models. The main findings of this investigation can be
summarized as follows:

11
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 7. The effect of fc’ on the cracking torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

12
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 8. The effect of fc’ on the ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

13
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 9. The effect of tw on the cracking torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

14
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 10. The effect of tw on the ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

15
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 11. The effect of d on the cracking torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

16
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 12. The effect of d on the ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

17
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 13. The effect of EF Al/Ac on the ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

18
A. Deifalla Engineering Structures 226 (2021) 111394

Fig. 14. The effect of EFt At/bs on the ultimate torque SI for all tested beams using the following methods: (a) CSA, (b) HD, (c) PM 1, (d) PM 2.

• The first proposed model is a code-like format and consistent with Data availability
the conventional hollow tube analogy and variable angle space truss
model, yet it predicted the cracking and ultimate torsion strength All data produced during the development of this paper is provided
better than the existing design codes and selected models compared in the paper.
to that experimentally measured from the available in the literature.
• The second proposed model is non-iterative and based on a power- Funding body
law equation for all effective parameters, yet it predicted the
cracking and ultimate torsion strength better than the existing design This research did not receive any specific grant from funding
codes and selected models compared to that experimentally agencies in the public, commercial, or not-for- profit sectors.
measured from the available in the literature.
• Using the ACI-318 cracking torque formula for predicting the CRediT authorship contribution statement
cracking torque of FRP-reinforced concrete beams is conservative.
However, for the case of flanged FRP -reinforced concrete beams, it is Ahmed Deifalla: Conceptualization, Methodology, Validation,
recommended to neglect the flange contribution. Writing - review & editing, Resources, Writing - original draft, Meth­
odology, Supervision.

19
A. Deifalla Engineering Structures 226 (2021) 111394

Declaration of Competing Interest [17] Deifalla A, Khalil MS, Abd Elrahman A. Simplified model for the torsional strength
of concrete beams with GFRP stirrups. Compos Constr, ASCE 2014; 19(1) 10.1061/
(ASCE) CC.1943-5614.0000498, 04014032.
The authors declare that they have no known competing financial [18] Deifalla A. Torsional behavior of rectangular and flanged concrete beams with FRP
interests or personal relationships that could have appeared to influence reinforcements. J Struct Eng 2015;141(12):04015068. https://doi.org/10.1061/
the work reported in this paper. (ASCE)ST.1943-541X.0001322.
[19] Hassan MM, Deifalla A. Evaluating the new CAN/CSA-S806-12 torsion provisions
for concrete beams with FRP reinforcements. Mater Struct 2016. https://doi.org/
Acknowledgements 10.1617/s11527-015-0680-9.
[20] Shehab HKHS, El-Awady M, Husain M, Mandour S. Behavior of concrete beams
reinforced by FRP bars under torsion. Proceedings of the 13th ICSGE, Cairo, Egypt;
The author would like to acknowledge the continuous support of the 2009, 6 pp.
Future University in Egypt (FUE) in the course of this research study. [21] Ragab KS, Eisa AS. Torsion behavior of steel fibered high strength self compacting
concrete beams reinforced by GFRB bars. World Acad Sci, Eng Technol Int J Civ,
Archit Sci Eng 2013; 7 (9), pp. 331–341.
Appendix A. Supplementary material [22] Mohamed HM, Chaallal O, Benmokrane B. Torsional moment capacity and failure
mode mechanisms of concrete beams reinforced with carbon FRP bars and stirrups.
Supplementary data to this article can be found online at https://doi. J Compos Constr 2015;19(2):04014049. https://doi.org/10.1061/(ASCE)CC.1943-
5614.0000515.
org/10.1016/j.engstruct.2020.111394. [23] Deifalla A, Hamed M, Saleh Amin, Ali Tarek. Exploring GFRP bars as reinforcement
for rectangular and L-shaped beams subjected to significant torsion: An
References experimental study. Eng Struct 2014;59:776–86.
[24] Mohamed HM, Benmokrane B. Torsion behavior of concrete beams reinforced with
GFRP bars and stirrups. ACI, Struct J 2015; 112(5), 543–552 pp., doi: 10.14359/
[1] Kuchma D, Wei S, Sanders D, Belarbi A, Novak L. The development of the one-way
51687824.
shear design provisions of ACI 318-19. ACI Struct J 2019; 116(4 July), doi:
[25] Mohamed HM, Benmokrane B. Reinforced concrete beams with and without FRP
10.14359/51716739.
Web reinforcement under pure torsion. J Bridge Eng 2016;21(3):04015070.
[2] ACI-440. Guide for the design and construction of structural concrete reinforced
https://doi.org/10.1061/(ASCE)BE.1943-5592.0000839.
with FRP Bars. ACI 440.1R-15, ACI Committee 440. Farmington Hills, MI:
[26] Zhou J, Shen W, Wang S. Experimental study on torsional behavior of FRC and ECC
American Concrete Institute; 2015, 300 pp.
beams reinforced with GFRP bars. Constr Build Mater 2017;152:74–81.
[3] CSA. Design and construction of buildings components with fiber-reinforced
[27] Deifalla A. Strength and ductility of lightweight reinforced concrete slabs under
polymers. Canadian Standards Association (CSA), CSA S806-12, Toronto; 2012.
punching shear. Structures 2020;27:2329–45.
[4] ACI-445. Report on torsion in structural concrete. Reported by Joint ACI-ASCE
[28] Deifalla A. Design of lightweight concrete slabs under two-way shear without shear
Committee 445” ACI 445.1R-12, ASCE-ACI Committee 445 on shear and torsion,
reinforcements: a comparative study and a new model. Eng Struct, Elsevier 2020;
April 2013, ISBN-13; 2013: 978-0-87031-810-8, ISBN: 0-87031-810-1, 80 pp.
222(2020):111076. https://doi.org/10.1016/j.engstruct.2020.111076.
[5] Hsu, Thomas TC. Torsion of reinforced concrete. New York: Van Nostrand Reinhold
[29] Deifalla A, Awad A, AbdElrahman AA, Seleem H. Investigating the behavior of
Co., Inc.; 1984. 516 p.
lightweight foamed concrete T-beams under torsion, shear, and flexure. Eng Struct
[6] Deifalla A. Torsion Design of Lightweight Concrete Beams without or with Fibers: a
2020;219(2020):110741. https://doi.org/10.1016/j.engstruct.2020.110741.
comparative study and a refined cracking torque formula. Structures 2020;28
[30] ElMeligy O, El-Nemr AM, Deifalla A. Reevaluating the Modified Shear Provision of
(2020):786–802. https://doi.org/10.1016/j.istruc.2020.09.004.
CAN/CSA S806-12 for Concrete Beams Reinforced with FRP Stirrups. AEI 2017.
[7] Deifalla A, Ghobarah A. Behavior and analysis of inverted T-shaped RC beams
doi: 10.1061/9780784480502.027.
under shear and torsion. Engineering Structures, Elsevier 2014; 62: 776–786.
[31] AASHTO LRFD. “Bridge design guide specifications for GFRP-reinforced concrete
[8] Rahal KN. Torsional strength of normal and high strength reinforced concrete
bridge decks and traffic railings.” 1st Ed., American Association of State Highway
beams. Eng Struct 2013;56:2206–16.
and Transportation Officials, AASHTO, Washington, D.C.; 2009.
[9] Salama AE, Kassem ME, Mahmoud AA. Torsional behavior of T- shaped reinforced
[32] CSA-A23.3-04. Design of concrete structures for buildings. Canadian Standards
concrete beams with large web openings. J Build Eng 2018;18:84–94.
Association, Rexdale, Ontario, Canada; 2004.
[10] Kuan A, Bruun EPG, Bentz EC, Collins MP. Nonlinear sectional analysis of
[33] PCI. PCI Design Handbook.” PCI, Illinois; 2006, 736.
reinforced concrete beams and shells subjected to pure torsion. Comput Struct
[34] ACI-318. Building code requirements for reinforced concrete. American Concrete
2019;222:118–32.
Institute, Detroit, Michigan; 2014.
[11] Chalioris CE, Karayannis CG. Effectiveness of the use of steel fibres on the torsional
[35] Belarbi A, Hsu T. Stirrup stresses in reinforced concrete beams. ACI Struct J 1990;
behaviour of flanged concrete beams. Cement Concr Compos 2009; 31(5):
September-October, 87-S53, 530–538.
331–341.
[36] Mitchell D, Collins MP. Diagonal compression field theory—a rational model for
[12] Deifalla A, Awad A, Seleem H, Abdelrahman A. Experimental and numerical
structural concrete in pure torsion. ACI J 1974; 71(8): Aug., pp. 396–408.
investigation of the behavior of LWFC L-girders under combined torsion. Structures
[37] Hsu TTC, Mo YL. Softening of concrete in torsional members—theory and tests. ACI
2020;26:362–77.
J 1985; 82(3): May-June, pp. 290–303.
[13] Ilkhani MH, Naderpour H, Kheyroddin A. A proposed novel approach for torsional
[38] Collins MP. Evaluation of shear design procedures for concrete structures, A Report
strength prediction of RC beams. J Build Eng 2019;25:100810. https://doi.org/
Prepared for the CSA Technical Committee on Reinforced Concrete Design.
10.1016/j.jobe.2019.100810.
Canada. Canadian Standards Association; 2001.
[14] Kotsovos MD. Reinforced concrete shear design: shortcomings and remedy. ACI
[39] Kassem W. Non-linear analysis of shear-critical reinforced concrete beams using
Struct J July-August 2017; ACI-114-S86. 1055-1066.
the softened membrane model. Struct Concr 2015;16(4):524–36.
[15] Bernardo L. Modeling the full behavior of reinforced concrete flanged beams under
[40] Razaqpur A, Spadea S. Shear strength of FRP reinforced concrete members with
torsion. Appl Sci 2019; 9 (2370), 16 pp, DOI:10.3390/app9132730.
stirrups. J Compos Constr 2015; 19(1): 04014025.
[16] Razaqpur AG, Bencardino F, Rizzuti L, Spadea G. FRP reinforced/prestressed
concrete members: A torsional design model. Compos B Eng 2015;79:144–55.

20