You are on page 1of 22

Engineering Structures 256 (2022) 114046

Contents lists available at ScienceDirect

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

An innovated theory and closed form solutions for the elastic lateral
torsional buckling analysis of steel beams/columns strengthened with
symmetrically balanced GFRP laminates
Phe Van Pham
Research and Application Center for Technology in Civil Engineering and Department of Civil Engineering, University of Transport and Communications, #3 Cau Giay,
Hanoi, Vietnam

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

Keywords: An innovated theory based on the principle of total stationary buckling energy is successfully developed for the
Elastic buckling resistance elastic lateral torsional buckling analysis of steel beams/columns strengthened with symmetrically balanced
Orthotropic GFRP strengthening GFRP laminates. Two closed form solutions and four eigenvalue solutions of the buckling resistances are then
Steel beam columns
developed based on the theory and based on postulated buckling displacement functions. The present theory
Closed form solution
captures the partial interaction between the steel member and the GFRP laminates, stacking sequences and
orthotropic properties of GFRP laminae, shear deformations in the GFRP laminates, and local and global warping
deformations. The elastic buckling resistances predicted by the present solutions are well validated against those
of three dimensional finite element analyses, as presented in three examples and two parametric studies of the
present study. The present solutions are fast and convenient to predict the elastic buckling resistances of GFRP-
strengthened beams/columns. Based on the parametric studies conducted, it is observed that the effects of GFRP
lamina stacking sequences (with different fiber orientation angles), GFRP laminate thicknesses, GFRP moduli of
elasticity, GFRP shear moduli, and adhesive shear moduli on the elastic buckling resistances are significant.

1. Introduction and motivation GFRP-strengthened steel members necessitates the development of


closed form solutions those are both simple and accurate enough
1.1. General (compared to the 3D FEA solutions).

Glass Fiber Reinforced Polymer (GFRP) laminates are often stacked 1.2. Review of studies on steel members strengthened with GFRP
by a series of orthotropic GFRP laminae to form relatively thick plates, laminates
those can carry tension, compression, shear forces [1–3]. Based on a
high strength-to-weight ratio and a good corrosion resistance, GFRP Behavior investigations of steel members strengthened with GFRP
laminates are increasingly being used as a retrofit material for steel laminates were reported in a number studies. Analytical theories/solu­
beams/columns [4–8]. The laminates can be bonded to steel members tions and one-dimensional finite element formulations for the analyses
by using thin layers of adhesives with typically low moduli of elasticity of elastic behaviors (e.g., stresses, strains, displacements) of GFRP-
(compared to those of steel and GFRP materials), thus providing a partial strengthened beams were developed in studies [11–14]. Closed form
interaction between the steel members and the laminates. Besides, shear solutions for the interfacial shear and peeling stresses highly concen­
deformations and lamina stacking sequences of GFRP laminates also trated near the bond end (or GFRP laminate ends) were presented in a
have significant effects on the torsional behaviors of composite struc­ recent study [15]. Experimental investigations and analytical and nu­
tures [1–3,9–10]. Such effects in the GFRP laminate-strengthened steel merical solutions for the ultimate capacities of the laterally braced
members can be reliably captured by using three dimensional finite GFRP-strengthened steel beams were reported in studies [4–6,16]. Nu­
element analyses (3D FEA). However, the 3D FEA solutions typically merical investigations based on a commercial FEA software for local and
involve a significant effort to correctly model the structures, to run an­ global buckling behaviors of steel beams strengthened with GFRP lam­
alyses and to process output results. As such, an effective design of the inates were conducted in studies [7,8,17–19]. Pham et al. [19] also

E-mail address: phe.phamvan@utc.edu.vn.

https://doi.org/10.1016/j.engstruct.2022.114046
Received 1 June 2021; Received in revised form 5 February 2022; Accepted 20 February 2022
Available online 4 March 2022
0141-0296/© 2022 Elsevier Ltd. All rights reserved.
P. Van Pham Engineering Structures 256 (2022) 114046

developed a finite element (FE) formulation for the prediction of the interaction). Also, the neglection of the shear deformations in the FRP
lateral torsional buckling of a steel beam strengthened with an isotropic laminate as postulated in study [27] for the FRP-strengthened beams
GFRP laminate bonded to the beam compression flange. The study of under torques led to a significant underprediction for the system re­
Pham et al. [19] had some restrictions in application, in which the GFRP sponses (e.g., twisting angles, lateral displacements, longitudinal normal
laminate was only considered as an isotropic material. In fact, such a stresses and buckling resistances), as recently discussed in studies
laminate is normally stacked by orthotropic GFRP laminae and thus it [11,29]. A similar treatment was also observed in study [30] that
may also have typical orthotropic properties. Also, the theory and the FE developed a buckling solution for two layer Timoshenko beams with
formulation developed in [19] were relatively complicated to apply for interlayer slip. Besides, the solution of study [30] was developed for
practical designs, as they capture shear deformation fields due to beams with rectangular cross-sections and with isotropic materials and
bending and warping in all three materials. Further obstacles of the FE thus it was also not applicable to the present steel W beams/ columns
formulation in study [19] will be later discussed in the present study. strengthened with orthotropic GFRP materials. Recently, Phe et al. [19]
Within this context, the present study develops an more simply inno­ developed a FE formulation for the prediction of lateral torsional
vated theory and novel simple closed form solutions for the elastic buckling of a wide flange steel beam bonded with a GFRP plate through
lateral torsional buckling resistances of steel beams/columns strength­ an adhesive layer. The formulation accounted for the effects of partial
ened with orthotropic GFRP laminates. interaction between the steel beam and GFRP plate, shear deformations
in the steel and those in the GFRP plate. However, the GFRP plate was
assumed as an isotropic material and thus the material postulation in
1.3. Review of GFRP mechanical properties considered in lateral torsional
study [19] was unable to capture the orthotropic properties of GFRP
buckling problems
laminates. As indicated in [29], the effect of shear deformations in the
GFRP laminate was important in GFRP-strengthened steel beams under
GFRP laminates generally have different orthotropic properties
torques, but that in the steel beams might be negligible due to the fact
based on lamina properties and stacking sequences (with different fiber
that shear deformation effects due to bending and twist were low for the
orientation angles) [1–3,20–23]. Correia et al [1] and Silva et al [2]
steel beams. The inclusion of the shear deformations in steel beam in
presented experimental and numerical studies on the first order, buck­
study [19] led to a relatively complicated FE formulation that was not
ling and post buckling behaviors of orthotropic GFRP pultruded beams.
convenient to design calculations. Furthermore, the pre-buckling anal­
Xin et al [3] presented a systematic evaluation study on material be­
ysis conducted in Phe et al. [19] included pre-buckling internal forces in
haviors of pultruded GFRP laminates and they emphasized that the ef­
the GFRP laminate and those in the adhesive layer. However, the in­
fects of GFRP laminate manufacturing process (e.g., lamina properties,
ternal forces were order of magnitude smaller than those in the steel
fiber orientations) on the buckling responses of pultruded GFRP mem­
beam [19,29]. Their contributions for the buckling resistances were low,
bers were significant. Lee et al. [20] developed a FE formulation for the
but their inclusions into the theory of study [19] created a complicated
prediction of lateral torsional buckling resistance of laminated FRP
FE formulation.
beams with I-section and with material coupling for an arbitrary lami­
Based on the context, the present study is going to develop an
nate stacking sequence. Cortinez and Piovan [21] and Machado and
innovated theory for the prediction of elastic lateral torsional buckling
Cortinez [22] adopted Ritz variational method to develop closed form
resistances of steel beam/columns strengthened with one or two GFRP
solutions for lateral buckling of thin walled composite bisymmetric
laminates by using adhesive layers. The theory capture orthotropic
beams with symmetrically balanced FRP laminates and with taking the
properties of GFRP laminates with different fiber orientations and
effect of shear deformations. Qiao et al. [23] presented a combined
lamina properties (as not done in previous studies
analytical and experimental study of the flexural torsional buckling of
[11,19,24–26,29–30]), partial interaction between the steel and the
symmetrically balanced FRP cantilever I beams. Studies [1–3,20–23]
GFRP laminate (as not done in studies [24,27–28]), shear deformations
indicated that FRP laminates stacked by 450 / − 450 fiber orientations
in the GFRP laminate and local and global warping deformations (those
provided the highest, while those with 00 /900 provided the lowest,
are different to the kinematic assumptions of study [19]). To avoid mesh
elastic buckling resistances for the FRP laminates.
discretization errors common in numerical solutions, the present study is
going to develop simple novel closed form solutions and eigenvalue
1.4. Review of theories for the analysis of lateral-torsional buckling of solutions for the prediction of the elastic lateral torsional buckling re­
composite systems sistances of GFRP-strengthened beams/columns.

To predict the lateral torsional buckling resistance of multilayer 2. Description of the problem and scope of the present study
beams/columns, several analytical solutions were developed based on
different kinematic treatments [e.g., 24–29, 19]. Girhammar and Pan A wide flange beam/column with a span L and with boundary con­
[24] developed a buckling solution for two-layer members. However, ditions B.C.1 and B.C.2 is subjected to a longitudinally concentrated load
the effect of shear deformations and partial interaction between layers ( )
Pz , a transversely concentrated load Py yp , zp and a transversely
were neglected in their study. Challamel and Girhammar [25] proposed
a solution for lateral torsional buckling of layered composite beams that distributed load qy (z, y)(Fig. 1a). The beam flanges are bonded with
captured the partial interaction but again neglected the shear de­ GFRP laminated plates 1 and 2 by using adhesive layers 1 and 2. The
formations. Schnabl and Planinc [26] developed a non-shear deformable steel section has a depth h, a flange width b, a flange thickness tf , and a
theory for the inelastic buckling analysis of two-layer composite col­ web thickness tw (Fig. 1b). The width and the thickness of GFRP laminate
umns with non-linear interface compliance. Kabir and Seif [27] devel­ 1 are b1 and tg1 , respectively, while those of the adhesive layer 1 are b1
oped an analytical solution for the lateral torsional buckling of steel I and ta1 , those of the GFRP 2 are b2 and tg2 , and those of the adhesive
beams retrofitted with FRP sheets without considering the effect of shear layer 2 are b2 and ta2 .
deformations. Shojaee et al. [28] developed an iso-geometric finite Possible failure modes of the GFRP-strengthened beams typically
element method based on non-uniform rational B-splines basis functions include local buckling, elastic lateral torsional buckling, inelastic lateral
for the buckling analysis of laminated composite plates. In general, the torsional buckling, material flexural strength based on yielding or
kinematic treatments proposed in the above solutions [i.e., 25–28] may plastification, GFRP plate tension/compression failure, adhesive shear
not be applicable to steel beams/ columns strengthened with orthotropic failure, adhesive peeling failure, adhesive delamination failure at the
FRP laminates by using thin adhesive layers, in which the interaction steel/adhesive or GFRP/adhesive interfaces, delamination within the
between the steel and the laminate is not fully achieved (i.e., partial GFRP plate, cohesive failure within the adhesive layer, adhesive

2
P. Van Pham Engineering Structures 256 (2022) 114046

(a) Elevation view (b) Cross-section A-A


Fig. 1. A wide flange beam/column strengthened with two GFRP laminates under axial and transverse loads.

softening due to thermal effects, fatigue failure and deflection limit in the GFRP laminates and adhesive layers are neglected, (3) Local and
states. The bearing capacity of such composite members may be the global warping deformations are captured in the steel and GFRP lami­
lowest resistance based on the possible failure modes. In the scope of the nates. (4) The effect of web distortions is neglected. (5) Interaction be­
present study, only the solutions of the elastic lateral torsional buckling tween the steel member and the GFRP laminates are partial, in which
mode are developed for the GFRP-strengthened beams/columns while displacements in the adhesive layer are linearly interpolated from those
those of other failure modes are not considered. As they were or will be of the steel and the laminates. (6) Perfect bond is assumed. And (7) the
considered in other studies [e.g., 4–8, 11–19, 27, 29]. effects of residual stresses and initial geometric imperfections are not
considered in the present study.
3. Kinematic assumptions, local and governing displacements,
and material assumptions 3.2. Local and governing displacement fields

3.1. Kinematic assumptions Three global coordinate systems OXYZ, Og1 XYg1 Z, and Og2 XYg2 Z are
selected for the steel beam, GFRP laminate 1 and GFRP laminate 2,
The following kinematic assumptions are applied to the present respectively (Fig. 2a). In which O,Og1 and Og2 are the sectional centroids
theory: (1) Shear deformations due to bending and twist in the GFRP (also, they are shear centers), X is the lateral axis, Y, Yg1 and Yg2 are the
laminates are captured, but those in the steel member are neglected, (2) transverse axes, and Z is the longitudinal axis of the steel, GFRP laminate
Pre-buckling internal forces in the steel member are captured but those 1 and GFRP laminate 2. To capture local warping deformations and

Fig. 2. Local and global coordinates and displacements (a) Local and global coordinate systems, (b) Local displacement fields, (c) Governing prebuckling dis­
placements (a part of elevation view), and (d) Governing buckling displacement fields (cross-section view).

3
P. Van Pham Engineering Structures 256 (2022) 114046

partial interactions between materials, local coordinates Css ns z, resultants in GFRP laminate i and they are respectively obtained by
Cg1 sg1 ng1 z, Ca1 sa1 na1 z, Ca2 sa2 na2 z and Cg2 sg2 ng2 z are assigned to the integrating stresses σgi,zz ,τgi,xz , ngi σgi,zz , ngi τgi,xz from ngi = − tgi /2 to ngi =
thickness contours of the steel, GFRP laminate 1, adhesive layer 1, ad­ tgi /2, where σ gi,zz is the longitudinal normal stress field while τgi,xz is the
hesive layer 2 and GFRP laminate 2, respectively (Fig. 2a). In which inplane shear stress field in the GFRP laminate i, (i = 1, 2). Plate stiff­
origins C, Cg1 , Ca1 , Ca2 , Cg2 lies on the contours of the five material sec­ nesses Agi,11 , Agi,66 , Dgi,11 , Dgi,66 in Eqs. have been evaluated and presented
tions, s is the curvilinear contour coordinate measured from Origin O, n in Appendix A based on composite theories [e.g., 9–10].
is the normal-to-tangent axis of the contour measured from the contour
line, and z is the longitudinal axis. 4. Formulation of the general solution
Local displacement fields along the local coordinate systems (i.e., at
a Point with coordinates (s, n, z)) are depicted in Fig. 2b. In the steel, the 4.1. Expression of displacement fields in term of governing displacements
tangential displacement field is denoted as us , the normal-to-tangent
displacement field is denoted as vs while the longitudinal displace­ The displacement fields us , vs , ws at a Point with coordinate (ss , ns , z)
ment field is ws . In the GFRP laminate i, (i = 1, 2), the tangential in the steel beam can be expressed based on a non-shear deformable
displacement field is denoted as ugi , the normal-to-tangent displacement beam theory [e.g., 29] as
field is denoted as vgi and the longitudinal displacement field is wgi
(Fig. 2b). In the adhesive layer i, (i = 1, 2), the tangential displacement

us (ss , ns , z) = sinα(ss )Vp (z) + cosα(ss )Ub (z) + [r(ss ) + ns ]θzb (z),
vs (ss , ns , z) = − cosα(ss )Vp (z) + sinα(ss )Ub (z) − q(ss )θzb (z), (2)
′ ′ ′
ws (ss , ns , z) = Wp (z) + [ns cosα(ss ) − y(ss ) ]V p (z) − [ns sinα(ss ) + x(ss ) ]U 1b (z) + [ns q(ss ) − ω(ss ) ]θ zb (z)

field is denoted as uai , the normal-to-tangent displacement field is in which the governing displacement fields of steel Vp (z),
denoted as vai and the longitudinal displacement field is wai (Fig. 2b). Wp (z),Ub (z),θzb (z) have been defined in Fig. 2c,d.
The local displacements account for both pre-buckling and buckling q(ss ) = x(ss )cosα(ss ) + y(s)sinα(ss ),r(ss ) = x(ss )sinα(ss ) − y(s)cosα(ss ),
deformations. where α(ss ) is an angle between the positive directions of the s-axis and
Prebuckling governing displacement fields along three global coor­ X-axis (Fig. 2a) and it is taken positive in the clockwise direction from
∫s
dinate systems in the steel, GFRP laminates 1 and 2 are depicted in the X-axis. ω(ss ) = 0s r(ss )dss is the sectorial coordinate of the Point with
Fig. 2c, in which the longitudinal and transverse displacements of the coordinate (ss , ns , z). The displacement fields ug1 , vg1 , wg1 at a Point with
steel section centroid are denoted as Wp (z) and Vp (z), respectively, while ( )
coordinate sg1 , ng1 , z in the GFRP laminate 1 are expressed based on a
those of the GFRP plate i, (i = 1, 2), are Wip (z) and Vp (z). Governing
shear deformable beam theory [11,19] as
buckling displacement fields along the global coordinate systems are ( )
described in Fig. 2d, in which the lateral displacement of the steel sec­ u(g1 sg1 , ng1 ,)z = U1b (z) + (ng1 θ)zb (z),
tion centroid is symbolled as Ub (z), that of the GFRP laminate i, (i = 1, (
v s , n , z = − Vp (z) − x sg1 θzb (z),
) g1 g1 g1 ( ) ( ) ′ (3)
2), is Uib (z). The beam and the laminates are assumed to have the same

wg1 sg1 , ng1 , z = W1p (z) + ng1 V p (z) − x sg1 θy1b (z) + ng1 x sg1 θ zb (z)
twisting angle θzb (z). To account for shear deformations in the laminates
Where the governing displacement fields W1p (z) and U1b (z) have
under lateral bending, bending slopes θy1b (z) of the GFRP laminate 1 and
been defined in Fig. 2c,d. A similar expression for the displacement
θy2b (z) of the GFRP laminate 2 are proposed in the present study, here ( )
fields ug2 , vg2 , wg2 at a Point with coordinate sg2 , ng2 , z in the FRP plate 2
θy1b (z) may be different to ∂U1b (z)/dz while θy2b (z) may be different to
can be obtained as
∂U2b (z)/∂z.
( )
u(g2 sg2 , ng2 ,)z = U2b (z) + (ng2 θ)zb (z),
(
v s , n , z = − Vp (z) − x sg2 θzb (z),
) g2 g2 g2 ( ) ( ) ′ (4)
3.3. Material assumptions

wg2 sg2 , ng2 , z = W2p (z) + ng2 V p (z) − x sg2 θy2b (z) + ng2 x sg2 θ zb (z)

The governing displacements W2p (z) and U2b (z) have been defined in
The steel beam and the adhesive layers are assumed as linearly
Fig. 2c,d. The displacement fields ua1 , va1 , wa1 at a Point with coordinate
elastic isotropic materials, in which the steel beam has a modulus of
(sa1 , na1 , z) in the adhesive layer 1 are linearly interpolated from the
elasticity Es , a shear modulus Gs and a Poisson’s ratio of μ, while ad­
displacements at the bottom steel flange and those at the bottom of the
hesive layer i (i = 1, 2) has a modulus of elasticity Eai , a shear modulus
FRP laminate 1, i.e.,
Gai and a Poisson’s ratio of μai . GFRP laminates 1 and 2 are also assumed
( ) ( ( ) (
as linearly elastic materials and they are symmetrically balanced lami­ 1 na1 tf ) 1 na1 tg1 )
ua1 (sa1 , na1 , z) = − u s ss , , z + + ug1 sg1 , − ,z ,
nates stacked by n orthotropic laminae, of which the kth lamina (k = 1,2, 2 ta1 2 2 ta1 2
..., n) has a longitudinal modulus of elasticity Ek,z , a lateral modulus of (
1 na1
) (
tf )
(
1 na1
) (
tg1 )
elasticity Ek,s , a shear modulus Gk,sz , and Poisson’s ratio vk,zs ,vk,sz (Ap­ va1 (sa1 , na1 , z) = − v s ss , , z + + vg1 sg1 , − ,z , (5)
2 ta1 2 2 ta1 2
pendix A). Constitutive equations of the GFRP laminates 1 and 2 may be ( ) ( ( ) (
1 na1 tf ) 1 na1 tg1 )
expressed as [1–3,9–10,22]. wa1 (sa1 , na1 , z) = − ws ss , , z + + wg1 sg1 , − ,z
2 ta1 2 2 ta1 2
⎧ ⎫ ⎧ ⎫

⎪ Ngi,zz ∗ ⎪⎪ ⎡ ⎤⎪⎪ εgi,zz ∗ ⎪⎪ From Eqs. (2), by setting ns = tf /2, From Eqs. (3), by setting ng1 =

⎪ ⎪
⎪ A 0 0 0 ⎪
⎪ ⎪

⎨N ∗⎪
⎪ ⎬ ⎢ gi,11 ⎨γ ∗ ⎪

0 A 0 0 ⎥ ⎬ − tg1 /2, then substituting into Eqs. (5), one has
(1)
gi,xz ⎢ gi,66 ⎥ gi,xz
∗ = ⎣ 0 ⎦ ∗ , (i = 1, 2)

⎪ M ⎪ 0 D 0 ⎪ κ ⎪
⎪ gi,zz ⎪
⎪ ⎪

gi,11 ⎪
⎪ gi,zz ⎪
⎪ ⎪

⎩ Mgi,xz ∗ ⎪
⎪ ⎭ 0 0 0 Dgi,66 ⎪⎩ κgi,xz ∗ ⎪⎭

in which Ngi,zz ∗ , Ngi,xz ∗ , Mgi,zz ∗ , Mgi,xz ∗ , (i = 1, 2) are shell stress

4
P. Van Pham Engineering Structures 256 (2022) 114046

ua1 (sa1 , na1 , z) = c1 (n)Ub (z) + c2 (n)U1b (z) + c4 (n)θzb (z),


va1 (sa1 , na1 , z) = − Vp (z) − x(s)θzb (z), (6)
′ ′ ′ ′
wa1 (sa1 , na1 , z) = c1 (n)Wp (z) + c2 (n)W1p (z) + c4 (n)V p (z) − c1 (n)x(s)U b (z) − c2 (n)x(s)U 1b (z) + c3 (n)x(s)θ zb (z)

In which c1 (n) = 1/2 − na1 /ta1 , c2 (n) = 1/2 + na1 /ta1 , c3 (n) = 4.2. Strain fields in material layers
( )/ ( )
− hb + tf − tg1 4 + hb − tf − tg1 na1 /2ta1 , and c4 (n) =
( )/ ( )
h − tg1 4 − h + tg1 na1 /2ta1 . 4.2.1. General strains
The displacement fields ua2 , va2 , wa2 at a Point with coordinate (sa2 , General longitudinal normal strains and inplane shear strains in the
na2 , z) in the adhesive layer 2 are linearly interpolated from the dis­ steel beam, GFRP laminate 1 and GFRP laminate 2 corresponding to the
placements bottom steel fiber and those at the top of the FRP laminate 2, state of buckling deformation are assumed as [19,21,29]:
i.e., [( ) ( )2 ]
∂wi 1 ∂vi 2 ∂ui
( ) ( ( ) ( εi = + + ,
1 na2 tf ) 1 na2 tg2 ) ∂z 2 ∂z ∂z
ua2 (sa2 , na2 , z) = + us ss , − , z + − ug2 sg2 , , z , (9)
2 ta2 2 2 ta2 2 ∂ui ∂wi ∂ui ∂ui ∂wi ∂wi ∂vi ∂vi
( ) ( ( ) ( γi,sz = + + + +
1 na2 tf ) 1 na2 tg2 ) ∂z ∂si ∂z ∂si ∂si ∂z ∂si ∂z
va2 (sa2 , na2 , z) = + vs ss , − , z + − vg2 sg2 , , z , (7)
2 ta2 2 2 ta2 2 in which subscript i =s for steel, i = g1 for GFRP laminate 1 and i =
( ) ( ) ( ) ( )
1 na2 tf 1 na2 tg2 g2 for GFRP laminate 2. From Eqs. (2), by substituting into Eqs. (9), the
wa2 (sa2 , na2 , z) = + ws s s , − , z + − wg2 sg2 , , z
2 ta2 2 2 ta2 2 strains in the steel beam can be expressed in terms of the governing
displacement fields as
From Eqs. (2), by setting ns = − tf /2, From Eqs. (3), by setting ng2 =

εs = W ’ p (z) + [ns cosα(ss ) − y(ss )]V ˝ p (z) − [ns sinα(ss ) + x(ss )]U ˝ b (z) + [ns q(ss ) − ω(ss )]θ˝ zb (z)
1 1 1 (10)
2 ’2
+ Ub’2 (z) + [ns cosα(ss ) − y(ss )]U ’ b (z)θ’ zb (z) + q2 (ss )θ’2
zb (z) + [r(ss ) + ns ] θzb (z)
2 2 2

tg2 /2, then substituting them into Eqs. (7), one has new expressions of
the adhesive displacements in terms of the governing displacements as and

′ ′ ′ ′ ′ ′
γs,sz = − cosα(ss )W p (z)U b (z) + [ns − r(ss ) ]W p (z)θ zb (z) + sinα(ss )V p (z) + cosα(ss )V p (z)θzb (z)
′ ′
− cosα(ss )[ns cosα(ss ) − y(ss ) ]V ′′ p (z)U b (z) + [ns cosα(ss ) − y(ss ) ][ns − r(ss ) ]V ′′ p (z)θ zb (z)
′ ′
+2ns θ zb (z) − sinα(ss )U b (z)θzb (z) + cosα(ss )[ns sinα(ss ) + x(ss ) ]U b (z)U ′′ b (z)

(11)
′ ′ ′′
+q(ss )θ zb (z)θzb (z) − cosα(ss )[ns q(ss ) − ω(ss ) ]U b (z)θ zb (z)
′ ′
− [ns sinα(ss ) + x(ss ) ][ns − r(ss ) ]U ′′ b (z)θ zb (z) + [ns q(ss ) − ω(ss ) ][ns − r(ss ) ]θ zb (z)θ′′ zb (z)

ua2 (sa2 , na2 , z) = c5 (na2 )Ub (z) + c6 (na2 )U2b (z) + c8 (na2 )θzb (z), Similarly, from Eqs. (3), by substituting into Eqs. (9), the strains in
va2 (sa2 , na2 , z) = − Vp (z) − x(sa2 )θzb (z), the GFRP laminate 1 can be obtained as
′ (8)
wa2 (sa2 , na2 , z) = c5 (na2 )Wp (z) + c6 (na2 )W2p (z) + c8 (na2 )V p (z)
′ ′ εg1 = εg1,zz ∗ + ng1 κg1,zz ∗ , γ g1,sz = γg1,sz ∗ + ng1 κg1,sz ∗ (12)
− c5 (na2 )x(sa2 )U b (z) − c6 (na2 )x(sa2 )θy2b (z) + c7 (n4 )x(sa2 )θ zb (z)
in which
In which c5 (na2 ) = 1/2 + na2 /ta2 , c6 (na2 ) = 1/2 − na2 /ta2 , c7 (na2 ) =
( )/ ( ) ( )/
h + tg2 4 + h − tg2 na2 /2ta2 , c8 (na2 ) = − h − tg2 4 −
( )
h + tg2 na2 /2ta2 .

( ) 1 1 ( ) ( )
ε*g1,zz = W ’ 1p (z) − x sg1 θ’ y1b (z) + U1b
’2
(z) + x2 sg1 θ’2 zb (z), κ*g1,zz = V ˝ p (z) + x sg1 θ˝ zb (z) + U ’ 1b (z)θ’ zb (z),
2 2
( ) ( )
γ*g1,sz = U ’ 1b (z) − θy1b (z) − W ’ 1p (z)θy1b (z) + V ’ p (z)θzb (z) + x sg1 θ’ y1b (z)θy1b (z) + x sg1 θzb (z)θ’ zb (z), (13)
( ) ( )
κ*g1,sz = W ’ 1p (z)θ’ zb (z) − V ˝ 1p (z)θy1b (z) + 2θ’ zb (z) − x sg1 θ’ y1b (z)θ’ zb (z) − x sg1 θy1b (z)θ˝ zb (z)

5
P. Van Pham Engineering Structures 256 (2022) 114046

Also, from Eqs. (4), by substituting into Eqs. (9), the strains in the [ ]
1 1 h− tg2
GFRP laminate 2 are obtained as
′ ′
γ a2,zn =− x(sa2 )U b (z)+ x(sa2 )θy2b (z)+ − 1 x(sa2 )θ zb (z),
ta2 ta2 2ta2
εg2 = εg2,zz ∗ + ng2 κg2,zz ∗ , γ g2,sz = γg2,sz ∗ + ng2 κg2,sz ∗ (14) 1 1 h+tg2 ′
γa2,sn = Ub (z)− U2b (z)− θzb (z)− θzb (z), γa2,sz =[c7 (n)+c8 (n)]θ zb (z)
in which ta2 ta2 2ta2
(18)

( ) 1 1 ( ) ( )
ε*g2,zz = W ’ 2p (z) − x sg2 θ’ y2b (z) + U2b
’2
(z) + x2 sg2 θ’2
zb (z), κ*g2,zz = V ˝ p (z) + x sg2 θ˝ zb (z) + U ’ 2b (z)θ’ zb (z),
2 2
( ) ( )
γ *g2,sz = U ’ 2b (z) − θy2b (z) − W ’ 2p (z)θy2b (z) + V ’ p (z)θzb (z) + x sg2 θ’ y2b (z)θy2b (z) + x sg2 θzb (z)θ’ zb (z), (15)
( ) ( )
κ*g2,sz = W ’ 2p (z)θ’ zb (z) − V ˝ p (z)θy2b (z) + 2θ’ zb (z) − x sg2 θ’ y2b (z)θ’ zb (z) − x sg2 θy2b (z)θ˝ zb (z)

4.2.2. Pre-buckling strains


Three shear stress components of the adhesive layers can be assumed
Based on Assumption 2 in Section 3.1, the pre-buckling governing
as
displacements for the steel beam are assumed as Wp (z) and Vp (z)

∂vi ∂wi ∂vi ∂vi ∂wi ∂wi ∂ui ∂ui


+ γ i,zn =
+ + + ,
∂z ∂ni ∂z ∂ni ∂z ∂ni ∂z ∂ni
(16)
∂vi ∂ui ∂vi ∂vi ∂ui ∂ui ∂wi ∂wi ∂ui ∂wi ∂ui ∂ui ∂wi ∂wi ∂vi ∂vi
γi,sn = + + + + , γ i,sz = + + + + ,
∂si ∂ni ∂si ∂ni ∂si ∂ni ∂si ∂ni ∂z ∂si ∂z ∂si ∂z ∂si ∂z ∂si

(Fig. 2c). They can be related to the beam resultant forces


in which subscript i =a1 is denoted for the adhesive layer 1 while i = Np (z),Qp (z),Mp (z) as Wp (z) = Np (z)/Es As , Vp′′ (z) = Mp (z)/Es Ixx,s , Vp (z) =
′ ′

a2 is for the adhesive layer 2. From Eqs. (6), by substituting into Eqs. Qp (z)/Gs As , where sectional properties As and Ixx,s of the steel beam
(16) and neglecting pre-buckling and nonlinear terms, the shear strains have been defined in Appendix A. From Eqs. (10), pre-buckling strains
in the adhesive layer 1 can be simplified as can be obtained in terms of the resultant forces as
[ ]
1 ′ 1 hb − tf − tg1 ′
Np (z) Mp (z) Qp (z)
γa1,zn = + x(sa1 )U b (z) −
ta1 ta1
x(sa1 )θy1b (z) +
2ta1
− 1 x(sa1 )θ zb (z), εs p = + [ns cosα(ss ) − y(ss ) ] , γs p = sinα(ss ) (19)
Es As Es Is,xx G s As
1 1 h + tg1 ′
γa1,sn = − Ub (z) + U2b (z) − θzb (z), γa1,sz = [c3 (n) + c4 (n)]θ zb (z)
ta1 ta1 2ta1 4.2.3. First variation of linear buckling strains
(17) From Eqs. (10)-(18), by neglecting pre-buckling and high-order
terms, the linear buckling strains can be obtained. Then, by taking the
Similarly, from Eqs. (8), by substituting into Eqs. (16) and neglecting
first variation of the linear buckling strains with respect to the governing
pre-buckling and nonlinear terms, the shear strains in the adhesive layer
displacement fields and neglecting high-order terms, one obtains
2 can be simply expressed as

δεs bL = − [ns sinα(ss ) + x(ss ) ]δU ’’ b (z) + [ns q(ss ) − ω(ss ) ]δθ’’ zb (z), δγs,sz bL = 2ns δθ’ zb (z),
( ) ( )
δεg1,zz bL = − x sg1 δθ’ y1b (z), δκg1,zz bL = x sg1 δθ’’ zb (z), δγg1,sz bL = δU ’ 1b (z) − δθy1b (z),
[ ]
1 1 hb − tf − tg1
δκg1,sz bL = 2δθ’ zb (z), δγa1,zn bL = + x(sa1 )δU ’ b (z) − x(sa1 )δθy1b (z) + − 1 δθ’ zb (z),
ta1 ta1 2ta1
[ ]
1 1 h + tg1
δγa1,sn bL = − δUb (z) + δU1b (z) − 1 + δθzb (z), δγ a1,sz bL = 2c4 (na1 )δθ’ zb (z),
ta1 ta1 2ta1
( ) ( )
δεg2,zz bL = − x sg2 δθ’ y2b (z), δκg2,zz bL = +x sg2 δθ’’ zb (z), δγg2,sz bL = δU ’ 2b (z) − δθy2b (z),
[ ]
1 1 h − tg2
δκg2,sz bL = +2δθ’ zb (z), δγa2,zn bL = − x(sa2 )δU ’ b (z) + x(sa2 )δθy2b (z) + − 1 x(sa2 )δθ’ zb (z), (20)
ta2 ta2 2ta2
1 1 h + tg2
δγa2,sn bL = δUb (z) − δU2b (z) − δθzb (z) − δθzb (z), δγa2,sz bL = [c7 (na2 ) + c8 (na2 ) ]δθ’ zb (z)
ta2 ta2 2ta2

6
P. Van Pham Engineering Structures 256 (2022) 114046

4.2.4. Second variation of nonlinear buckling strains expression of the second variation of the strain energy in terms of the
Also, from Eqs. (10)-(18), by neglecting pre-buckling and linear governing displacement fields can be obtained. After integrating over
terms, the nonlinear buckling strains can be obtained. Then, by taking cross-sectional areas, each term of second variation can be obtained as
the second variation of the nonlinear buckling strains with respect to the ∫ L∫ ∫L [ ]{ ’’ }
( )2 Es Is,yy 0 δU b (z)
governing displacement fields of the steel beam and neglecting high- Es δεs bL dAs dz= 〈δU ’’ b (z) δθ’’ zb (z) 〉 dz
’’
order terms, one obtains 0 As 0 0 Es Is,ωω δθ zb (z)
(25)

{ }
δ2 εs bN = (δU ’ b )2 + 2[ns cosα(ss ) − y(ss ) ]δU ’ b (z)δθ’ zb (z) + q2 (ss ) + [r(ss ) + ns ]2 (δθ’ zb )2 ,
(21)
δ2 γs,sz bN = − 2sinα(ss )δU ’ b (z)δθzb (z)

4.3. Second variation of the total buckling potential energy ∫ L ∫


[ ]
∫ L [ ]
Is,xx + Is,yy ’ 2
Es εs p δ2 εs bN dAs dz = Np (z) (δU ’ b )2 + (δθ zb ) dz
As
The total potential energy π of the system is contributed by the total 0 As 0
∫L
buckling strain energy U and total load potential loss V. The variation of +2 ’ ’
Mp (z)δU b δθ zb dz
the second variation of the total potential energy should vanish at the 0

onset of buckling, i.e., (26)


( ) [ ] ∫ ∫ ∫ ∫
1 1( 2 ) L ( )2 L [ ]
δ δ2 π = δ δ U + δ2 V =0 (22) Gs γs bL dAs dz + Gs γs p δ2 γs bN dAs dz
2 2
(27)
0 A1 0 A1
∫ ∫
where the total buckling strain energy U is contributed by
L L
’ ’ ’
= δθ zb (z)(Gs Js )δθ zb (z)dz − 2 Qp (z)δU b (z)δθzb (z)dz
0 0

∫ L∫ ∫ L∫ ∫ L∫ b1 ∫ L∫ b1
U= εs σ s dAs dz + γ s,sz τs,sz dAs dz + εg1,zz ∗ Ng1,zz ∗ dAg1 dz + κg1,zz ∗ Mg1,zz ∗ dAg1 dz
0 As 0 As 0 0 0 0
∫ L∫ b1 ∫ L ∫ b1 ∫ L∫
( )
+ γg1,sz ∗ Ng1,sz ∗ dAg1 dz + κg1,sz ∗ Mg1,sz ∗ dAg1 dz + γa1,zn τa1,zn + γa1,sn τa1,sn + γa1,sz τa1,sz dAa2 dz
(23)
0 0 0 0 0 Aa1
∫ L∫ b2 ∫ L∫ b2 ∫ L∫ b2 ∫ L∫ b2
∗ ∗ ∗ ∗ ∗ ∗ ∗ ∗
+ εg2,zz Ng2,zz dAg2 dz + κg2,zz Mg2,zz dAg1 dz + γ g2,sz Ng2,sz dAg2 dz + κg2,sz Mg2,sz dAg2 dz
0 0 0 0 0 0 0 0
∫ L∫
( )
+ γa2,zn τa2,zn + γ a2,sn τa2,sn + γ a2,sz τa2,sz dAa2 dz
0 Aa2

Based on linearly elastic material assumptions, stresses in Eqs. (23) ∫ L ∫ b1 ( )2 ∫ L ( / )


can be related to corresponding strains based on Hooke’s law [19,22]. Ag1,11 δεg1,zz bL dxdz = δθ’ y1b (z) Ag1,11 b1 12 δθ’ y1b (z)dz (28)
Then, by taking the second variation of the total buckling strain energy 0 0 0

[19], one obtains

{∫ L ∫ [( ] ∫ L∫ [( ]
1 2 1 )2 )2
δU= Es δεs bL + εs p δ2 εs bN dAs dz + Gs δγs bL + γ s p δ2 γs bN dAs dz
2 2 0 As 0 As
∫ L ∫ b1 ( )2 ∫ L ∫ b1 ( )2 ∫ L ∫ b1 ( )2 ∫ L ∫ b1 ( )2
+ Ag1,11 δεg1,zz bL dxdz + Dg1,11 δκg1,zz bL dxdz + Ag1,66 δγg1,sz bL dxdz + Dg1,66 δκg1,sz bL dxdz
0 0 0 0 0 0 0 0
∫ L∫ ( )2 ∫ L∫ ( )2 ∫ L∫
( )
+ Ga1 δγa1,zn bL dAa1 dz + Ga1 δγa1,sn bL dAa1 dz + Ga1 δγa1sz bL 2
dAa1 dz (24)
0 Aa1 0 Aa1 0 Aa1
∫ L∫ b2 ( )2 ∫ L∫ b2 ( )2 ∫ L∫ b2 ( )2 ∫ L∫ b2 ( )2
+ Ag2,11 δεg2,zz bL dxdz + Dg2,11 δκg2,zz bL dxdz + Ag2,66 δγg2,sz bL dxdz + Dg2,66 δκ5sz bL dxdz
0 0 0 0 0 0 0 0
∫ L∫ ( )2 ∫ L∫ ( )2 ∫ L∫ ( )2 }
+ Ga2 δγa2,zn bL dAa2 dz + Ga2 δγa2,sn bL dAa2 dz + Ga2 δγa2,sz bL dAa2 dz
0 Aa2 0 Aa2 0 Aa2

From Eqs. (20), (21) and (19), by substituting into Eqs. (24), an

7
P. Van Pham Engineering Structures 256 (2022) 114046

∫ L ∫ b1 ( )2 ∫ L ( / ) ∫ ∫ ∫
Dg1,11 δκg1,zz bL dxdz = δθ’’ zb (z) Dg1,11 b1 12 δθ’’ zb (z)dz (29) L ( )2 L

0 0 0 Ga2 γa2,sz bL dAa2 dz = δθ’ zb (z).(Ga2 Ja2 ).δθ’ zb (z)dz (41)


0 Aa2 0
∫ ∫ ( )2
in which sectional properties Is,yy , aa1,o , aa2,o , Is,xx , χ a1,yy , χ a2,yy , Is,
L b1
Ag1,66 δγg1sz bL dxdz
0 0
[ ]{ } (30) ωω, χ a1,a , χ a2,a , Js , Aa1 , Aa2 , As , Ia1,yy , Ia2,yy , Ja1 ,da1,o ,da2,o , Ja2 , ha1,o , ha2,o are
∫ L
1 − 1 δU ’ 1b (z) defined in Appendix A, while material properties Es ,Gs , Gai ,Agi,11 , Agi,66 ,
= 〈 δU ’ 1b (z) δθy1b (z) 〉Ag1,66 b dz
0 − 1 1 δθy1b (z) Dgi,11 ,Dgi,66 (i = 1,2) are defined in Section 3.3 of the present study. The
∫ ∫ ( )2 ∫ second variation of the total load potential loss caused by a transverse
L b1 L ( ) ( )
Dg1,66 δκg1,sz bL dxdz = δθ’ zb (z) 4b1 Dg1,66 δθ’ zb (z)dz (31) point load Py (zP , yP ) and a transverse distributed load qy zq , yq can be
0 0 0
determined as
{ ∫L }
1 2 1 [ ( ) ]2 ( ) [ ( ) ]2
∫ L∫ ( )2 δV= − Py (zP , yP ).yP . δθzb zp − qy zq , yq .yq . δθzb zq dz
2 2
Ga1 δγa1,zn bL dAa1 dz 0
0 Aa1 (42)
⎤⎧ ’ ⎡ ⎫
∫ L
4 2ha1,o − 4 ⎨ δU b (z) ⎪
⎪ ⎬
= ’ ’ ⎢ 2 ⎥ ’
〈δU b (z) δθ zb (z) δθy1b (z) 〉χ a1,yy ⎣ 2ha1,o ha1,o − 2ha1,o ⎦ δθ zb (z) dz 4.4. Eigenvalue problems and closed form solutions developed in the
0
− 4 − 2ha1,o 4


δθy1b (z)

⎭ present study
(32)
A general closed form solution of Eqs. (22) may be complicated
because they include coupled differential equations of 6 governing
∫ L∫ displacement fields (i.e., Ub , U1b (z), U2b (z), θy1b (z), θy2b (z), θzb (z)) and
( )2
Ga1 δγ a1,sn bL dAa1 dz their derivatives. Therefore, the present study is going to develop simple
0 Aa1 closed form solutions and eigenvalue solution for 6 typical GFRP-
⎡ ⎤⎧ ⎫
∫ L
4 2da1,o − 4 ⎨ δUb (z) ⎪
⎪ ⎬ strengthened beam/columns with given boundary, loading, and GFRP
=

〈δUb (z) δθzb (z) δU1b (z) 〉χ a1,a ⎣ 2da1,o da1,o 2

− 2da1,o ⎦ δθzb (z) dz strengthening conditions (Table 1) as examples. More solutions for other
0
− 4 − 2da1,o 4


δU1b (z)

⎭ strengthening schemes can be similarly developed.
(33)
4.5. Closed form solutions for the elastic lateral torsional buckling
∫ L ∫ ( )2 ∫ L resistances of Cases 1 and 2
Ga1 γa1,sz bL dAa1 dz = δθ’ zb (z)(Ga1 Ja1 )δθ’ zb (z)dz (34)
4.5.1. Case 1: Simply supported GFRP-strengthened beams under a uniform
0 Aa1 0

∫ L ∫ b( )2 ∫L
( / ) bending M
Ag2,11 δεg2,zz bL dxdz = δθ’ y2b (z) Ag2,11 b2 12 δθ’ y2b (z)dz (35) A general wide flange beam strengthened with two GFRP laminates
(i.e., GFRP laminates 1 and 2) has been defined in Fig. 1a. In Case 1
0 0 0

∫ L ∫ b2 ( )2 ∫ L ( / ) (Table 1), GFRP laminate 1 is bonded to the compression flange while


Dg2,11 δκg2,zz bL dxdz = δθ’’ zb (z) Dg2,11 b2 12 δθ’’ zb (z)dz (36) GFRP laminate 2 is removed (by setting U2b (z) = 0 and θy2b (z) = 0). The
governing bucking displacement fields of the present GFRP-
0 0 0

∫ L ∫ b2 ( )2 strengthened beam maybe assumed as


Ag2,66 δγg2,sz bL dxdz
0 0
(37) δUb (z) = δuo sin(αz), δθzb (z) = δθzo sin(αz), δU1b (z)
∫ L [
1 − 1
]{
δU ’ 2b (z)
} (43)
= 〈 δU ’
δθy2b (z) 〉Ag2,66 b dz = δu1o sin(αz), δθy1b (z) = δθy1 o cos(αz)
2b (z)
0 − 1 1 δθy2b (z)
in which α = π/L. The assumed displacements satisfy the boundary
∫ L ∫ b2 ( )2 ∫ L ( ) conditions of a simply supported beam (i.e., Ub (z0 ) = 0, U′′b (z0 ) = 0,
Dg2,66 δκg2,sz bL dxdz = δθ’ zb (z) 4b2 Dg2,66 δθ’ zb (z)dz (38) θzb (z0 ) = 0, θ′′zb (z0 ) = 0, U1b (z0 ) = 0, U′′1b (z0 ) = 0, θy1b (L/2) = 0 where
0 0 0
z0 = 0 or L). Based on the present loading condition (i.e., uniform
bending moment), pre-buckling forces can be determined as Np (z) =
∫ L∫ ( )2 Qp (z) = 0, Mp (z) = M. From Eqs. (43) and the given pre-buckling loads,
Ga2 δγa2,zn bL dAa2 dz by substituting into Eqs. (25)-(34), an expression of (1/2)δ2 U(i.e., Eqs.
〈 〉
(24)) in term of δuo δθzo δu1o δθy1 o can be obtained. Then, by
0 Aa2
⎡ ⎤⎧ ’ ⎫
∫ 4 − 2ha2,o − 4 ⎨ δU b (z) ⎪

L
⎢ ⎥
⎬ substituting the generated result of (1/2)δ2 U into Eqs. (22) and taking its
= 〈δU ’ b (z) δθ’ zb (z) δθy2b (z) 〉χ a2,yy ⎣ − 2ha2,o ha2,o 2 2ha2,o ⎦ δθ’ zb (z) dz 〈 〉
0 ⎪
⎩ ⎪
⎭ first variation with respect to vector δuo δθzo δu1o δθy1 o , one
− 4 2ha2,o 4 δθy2b (z) obtains
(39) ⎡ ⎤⎧ ⎫
β1 − M 0 β7 ⎪ ⎪ δuo ⎪ ⎪
⎢ − M β2 β6 0 ⎥ δθzo ⎬⎨

⎣ 0

⎦ =0 (44)
∫ L∫ β6 β3 β4 ⎪ ⎪ δu1o ⎪ ⎪
( )2 ⎩ ⎭
β7 0 β4 β5 δθy1 o
Ga2 δγ a2,sn bL dAa2 dz
in which factors βi , (i = 1, 7), of Eqs. (44) have been defined in Ap­
0 Aa2
⎤⎧ ⎡ ⎫
∫ 4 − 2da2,o − 4 ⎪ ⎨ δUb (z) ⎪
L
⎢ ⎥
⎬ pendix A. For a non-trivial solution of Eqs. (44), one obtains an eigen­
= 〈δUb (z) δθzb (z) δU2b (z) 〉χ a2,a ⎣ − 2da2,o da2,o 2 2da2,o ⎦ δθzb (z) dz value problem of the buckling moment Mcr as
0 ⎪
⎩ ⎪

− 4 2da2,o 4 δU2b (z)
(40)

8
P. Van Pham Engineering Structures 256 (2022) 114046

Table 1
Typical structures, closed form solutions and eigenvalue problems developed in the present study.
Cases Description by figure Boundary conditions GFRP laminates Loading conditions Eigen-value Closed form
problem solution

1 Simply-Supported A GFRP laminate bonded to the Uniform bending moment Yes Yes
Beams compression flange

2 Midspan point load Yes Yes

3 Linear bending moment Yes –

4 Arbitrary point load Yes –

5 Simply-Supported One or Two GFRP laminates Linear bending moment Yes –


Beam-columns bonded to the flanges + axial force

6 Cantilever Beam- Yes –


columns

⃒ ⃒
⃒ β1 − Mcr 0 β7 ⃒⃒ ⎡ ⎤⎧ ⎫
⃒ β1 − ψ P(Py ) 0 β7 ⎪ δuo ⎪
⃒− Mcr β2 β6 0 ⃒⃒ ⎪ ⎪
⃒ =0 (45) ⎢ − ψ P Py β2 − 2yP ψ P Py β6 0 ⎥

⎥ δθzo


⃒ 0 β6 β3 β4 ⃒⃒ ⎢
⎣ 0 =0 (48)
⃒ β6 β3 β4 ⎦⎪ δu 1o ⎪
β7 0 β4 β ⃒
5

⎩ ⎪

β7 0 β4 β5 δθy1 o
From Eq. (45), by solving for Mcr , a closed form solution of the elastic
By applying a similar procedure as done in Section 4.4.1, a closed
lateral torsional buckling resistance can be obtained as
form solution of the elastic lateral torsional buckling resistance can be
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
̅
( ∗ )2 obtained as
Mcr = − βb ∗ + βb − β c ∗ (46)
[ ̅ ]/
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( ∗∗ )2
in which Py,cr = − βb ∗∗ + βb − βc ψp (49)
/( )
βb ∗ = β4 β6 β7 β 3 β5 − β4 , in which
( )/( ) (47)
βc ∗ = β2 β3 β7 − β6 β7 − β1 β2 β3 β5 + β1 β 2 β4 + β1 β5 β6 β3 β5 − β4

[ ( ) ]/( ) /( )
βb ∗∗ = β4 β6 β7 + yoP β3 β7 + β1 β4 2 − β1 β3 β5 β3 β5 − β4 2 , yoP = yP α2 ψ P L ,
∗ ( )/( ) [ / /( )] (50)
βc = β 2 β3 β7 − β6 β7 − β1 β2 β3 β5 + β1 β2 β4 2 + β1 β5 β6 β3 β5 − β4 2 , ψ P = 1 8 + 1 2π2 L

(
where denominator β3 β5 − β4 can be simplified as b1 Ag1,11 b1 A where βi , (i = 1, 7) have been defined in Appendix A.
) ( )/
g1,66 + 4χ a1,a /α +4χ a1,yy b1 Ag1,66 /α2 +4χ a1,a b1 Ag1,66 + 4χ a1,yy α4 that
2

never vanishes.
4.6. Eigenvalue problems for the elastic lateral torsional buckling
resistances of Cases 3 to 6
4.5.2. Case 2: Simply supported GFRP-strengthened beams under a
midspan point load Py (yP , zP )
4.6.1. Cases 3 and 4: Simply supported GFRP-strengthened beams under
The governing bucking displacement fields in Case 2 are assumed to
ununiform moments or an arbitrary point load
take the form as presented in Eqs. (43). Meanwhile, the pre-buckling
In Case 3, ununiform bending moments M and κM are applied to the
loads are evaluated as Np (z) = 0,Qp (z) = Py /2, Mp (z) = Py z/2 when 0⩽
beam end sections (Table 1). In Case 4, a point load Py (yP , zP ) = P is
z⩽L/2 while Np (z) = 0, Qp (z) = − Py /2, Mp (z) = Py (L − z)/2 when
applied at a transverse coordinate yP and at a longitudinal coordinate zP
L/2⩽z⩽L. From the thus obtained pre-buckling loads and Eqs. (43), by
in the steel beam (Table 1). For both Cases 3 and 4, the governing
substituting into Eqs. (25)-(34) and Eq. (42), an expression of (1/2)δ2 U buckling displacement fields satisfying boundary conditions of a simply
〈 〉
(i.e., Eqs. (24)) in term of δuo δθzo δu1o δθy1 o can be obtained. supported beam may be assumed to take the form
Also, from Eqs. (43), by substituting in Eq. (42), the effect of load height
δUb (z) = 〈S(z) 〉1×m {δUo }m×1 , δθzb (z) = 〈S(z) 〉1×m {δθ } ,
{ oz }m×1
position can be evaluated as (1/2)δ2 V = − (δθzo )2 yP Py , where yP has (51)
δU1b (z) = 〈S(z) 〉1×m {Uo1 }m×1 , δθy1b (z) = 〈C(z) 〉1×m δθoy1 m×1
been defined in Fig. 1a. Then, by substituting the generated results of (1/
2)δ2 U and (1/2)δ2 V into Eqs. (22) and by taking its first variation with in which
〈 〉
respect to vector δuo δθzo δu1o δθy1 o , one obtains 〈S(z) 〉1×m = 〈 sin(αz) sin(2αz) … sin(iαz) … sin(mαz) 〉,
(52)
〈C(z) 〉1×m = 〈 cos(αz) cos(2αz) … cos(iαz) … cos(mαz) 〉

while the amplitudes of the sin, cos functions are defined as

9
P. Van Pham Engineering Structures 256 (2022) 114046

〈δUo 〉1×m = 〈 δuo,1 δuo,2 … δuo,i … δuo,m 〉, 〈δθoz 〉1×m = 〈 δθoz,1 δθoz,2 … δθoz,i … δθoz,m 〉,
〈 〉 (53)
〈δUo1 〉1×m = 〈 δuo1,1 δuo1,2 … δuo1,i … δuo1,m 〉, δθoy1 1×m = 〈 δθoy1,1 δθoy1,2 … δθoy1,i … δθoy1,m 〉

From Eqs. (51), by substituting into Eqs. (25) to (42), then by


( )
substituting the generated result of (1/2)δ2 U and (1/2)δ2 V into Eqs. (22) with EIωω = Es Is,ωω + Dg1,11 b1 /12 , GJ = Gs Js + 4b1 Dg1,66 +
and by taking its first variation with respect to vector 〈Δ〉1×4m = Ga1 Ja1 + ha1,o 2 χ a1,yy , and
〈 〈 〉 〉
〈δUo 〉1×m 〈δθoz 〉1×m 〈δUo1 〉1×m δθoy1 1×m , an eigenvalue prob­

∫ L ∫ L ∫ L
′ ′
Bm×m = 〈S(z) 〉m×1 〈S(z) 〉1×m dz, Bm×m = 〈S (z) 〉m×1 〈S (z) 〉1×m dz, Bm×m = 〈S′′(z) 〉m×1 〈S′′(z) 〉1×m dz,

0
L
0
∫ L
0
∫ L
(56)
′ ′ ′
Fm×m = 〈C(z) 〉m×1 〈C(z) 〉1×m dz, Fm×m = 〈C (z) 〉m×1 〈C (z) 〉1×m dz, T m×m = 〈S (z) 〉m×1 〈C(z) 〉1×m dz
0 0 0

lem can be derived for Cases 3 and 4 as while matrix K g in Case 3 (ununiform bending moments) is obtained
as

(57)

[ ]
[ ] in which the pre-buckling loads in Eqs. (57) are Np (z) = 0, Qp (z) =
[K e ]4m×4m − K g 4m×4m {Δ}4m×1 = 0 (54)
[(κ − 1)/L ], Mp (z) = [(κ − 1)/L ]z + 1.
in which matrix K e for both Cases 3 and 4 is expressed as Also, matrix K g in Case 4 (an arbitrary point load) is obtained as

(55)

10
P. Van Pham Engineering Structures 256 (2022) 114046

(58)

where the pre-buckling loads in Eqs. (58) are Np (z) = 0, Qp (z) = (L − where
zP )/L, Mp (z) = (L − zP )z/L when 0⩽z⩽zP and Np (z) = 0, Mp (z) =
k(1,1) = α4 Es Is,yy + 4α2 χ a1,yy + 4χ a1,a + 4α2 χ a2,yy + 4χ a2,a , k(1,2)
(L − z)zP /L, Qp (z) = − zP /L when zP ⩽z⩽L.
From Eqs. (54), by applying a condition of non-trivial solution, one = 2α2 χ a1,yy ha1,o + 2χ a1,a da1,o − 2α2 χ a2,yy ha2,o − 2χ a2,a da2,o ,
obtains the elastic lateral torsional buckling load M in Case 3 and P in
Case 4. k(2,2) =EIωω α4 + GJ α2 + α2 χ a1,yy ha1,o 2 + χ a1,a da1,o 2 + α2 χ a2,yy ha2,o 2
+ χ a2,a da2,o 2 , k(3,3) = α2 Ag1,66 b1 + 4χ a1,a
4.6.2. Case 5: Simply supported beam-columns under an axial force and
/
uniform bending moments k(4,4) = α2 Ag1,11 b3 12 + Ag1,66 b1 + 4χ a1,yy , k(5,5) = α2 Ag2,66 b + 4χ a2,a , k(6,6)
In this case, a simply supported steel beam-column is strengthened /
= α2 Ag2,11 b2 12 + Ag2,66 b + 4χ a2,yy
with two GFRP laminates (as defined in Table 1). The steel member is
subjected to an axial force Pz = P applied at the centroid of the movable where EIωω = Es Is,ωω +Dg1,11 b1 /12 +Dg2,11 b2 /12, and GJ = Gs Js +
support and a uniform bending moment M. The relationship between M Ga1 Ja1 + Ga2 Ja2 + 4b1 Dg1,66 + 4b2 Dg2,66 .
and P is assumed as M = ηP, where η is a load relation factor. The And matrix K g of Eqs. (60) is defined as
governing buckling displacements satisfying boundary conditions may
be assumed as
δUb (z) = δuo sin(αz), δθzb (z) = δθzo sin(αz), δU1b (z) = δu1o sin(αz),
δθy1b (z) = δθy1 o cos(αz), δU2b (z) = δu2o sin(αz), δθy2b (z) = δθy2 o cos(αz)
(59) (62)

The pre-buckling loads of the problem can be evaluated as Np (z) = P,


Mp (z) = M, Qp (z) = 0. From the present pre-buckling loads and Eqs.
(59), by substituting into Eqs. (25) through (42), then by substituting the
generated result of (1/2)δ2 U and (1/2)δ2 V into Eqs. (22) and by taking From Eqs. (60), by applying a condition of non-trivial solution, one
its first variation with respect to vector obtains the elastic buckling load P.
〈 〉
δuo δθzo δu1o δθy1 o δu2o δθy2 o , one obtains an eigenvalue
problem as 4.6.3. Cases 6: Cantilever GFRP-strengthened beam columns under an axial
[ [ ] ] point load and a uniform bending moment
[K e ]6×6 − K g 6×6 {Δ}6×1 = 0 (60) A uniform bending moment M is applied to the steel beam, in which
the relation between M and P is assumed as M = ηP, where η is a load
in which
relation factor. The governing buckling displacements of the system for
both Cases maybe assumed as

(61)

δUb (z) = 〈S(z) 〉1×m{{δUo }}m×1 , δθzb (z) = 〈S(z) 〉1×m {δθoz }m×1 , δU1b (z) = 〈S(z) 〉1×m {δU
{ o1 }}m×1 , (63)
δθy1b (z) = 〈S(z) 〉1×m δθoy1 m×1 , δU2b (z) = 〈S(z) 〉1×m {δUo2 }m×1 , δθy2b (z) = 〈S(z) 〉1×m δθoy2 m×1

11
P. Van Pham Engineering Structures 256 (2022) 114046


its variation with respect to vector 〈δUo 〉1×m 〈δθoz 〉1×m
where shape functions 〈S(z) 〉1×n satisfying boundary conditions of a 〈 〉 〈 〉 〉
〈δUo1 〉1×m δθoy1 1×m 〈δUo2 〉1×m δθoy2 1×m , one obtains an eigenvalue
cantilever beam-column are assumed as problem as

〈S(z) 〉1×m = 〈 1 − cos(αz) | 1 − cos(3αz) | … | 1 − cos((2i − 1)αz ) | … | 1 − cos((2m − 1)αz ) 〉 (64)

[ ]
in which α = π/2L. In Eqs. (63), the amplitudes of sin, cos functions [ ]
[K e ]6m×6m − Kg {Δ}6m×1 = 0 (66)
are defined as 6m×6m

in which

〈δUo 〉1×m = 〈 δuo,1 δuo,2 … δuo,i … δuo,m 〉, 〈δθoz 〉1×m = 〈 δθoz,1 δθoz,2 … δθoz,i … δθoz,m 〉,
〈 〉
〈δUo1 〉1×m = 〈 δuo1,1 δuo1,2 … δuo1,i … δuo1,m 〉, δθoy1 1×m = 〈 δθoy1,1 δθoy1,2 … δθoy1,i … δθoy1,m 〉, (65)
〈 〉
〈δUo2 〉1×m = 〈 δuo2,1 δuo2,2 … δuo2,i … δuo2,m 〉, δθoy2 1×m = 〈 δθoy2,1 δθoy2,2 … δθoy2,i … δθoy2,m 〉

(67)

The pre-buckling loads are of the given problem are Np (z) = P, where
Mp (z) = M, Qp (z) = 0. From the given pre-buckling loads and Eqs. (63),

( ) ( )
k1,1 = Es Is,yy B + 4 χ a1,yy + χ a2,yy B + 4 χ a1,a + χ a2,a B,
( ) ( )
k1,2 = 2 ha1,o χ a1,yy − ha2,o χ a2,yy B + 2 da1,o χ a1,a − da2,o χ a2,a B,
( ) ( )
k2,2 = EIωω B + GJB + ha1,o 2 χ a1,yy + ha2,o 2 χ a2,yy B + da1,o 2 χ a1,a + da2,o 2 χ a2,a B,
( ) ( / ) ( )
k3,3 = Ag1,66 b1 B + 4χ g1,a B, k4,4 = Ag1,11 b1 12 F + Ag1,66 b1 + 4χ a1,yy F,
( ) ( / ) ( )
k5,5 = Ag2,66 b B + 4χ a2,a B, k6,6 = Ag2,11 b2 12 F + Ag2,66 b2 + 4χ a2,yy F

by substituting into Eqs. (25) through (42), then by substituting the and matrix Kg of Eqs. (66) is evaluated as
generated result of (1/2)δ2 U and (1/2)δ2 V into Eqs. (22) and by taking

(68)

12
P. Van Pham Engineering Structures 256 (2022) 114046

In Eqs. (67) and (68), EIωω = Es Is,ωω +Dg1,11 b1 /12 +Dg2,11 b2 /12, GJ = 0, − 10, − 20, − 30, − 45, − 60, − 70, − 80,900 degrees (Table 3). The
Gs Js + Ga1 Ja1 + Ga2 Ja2 + 4b1 Dg1,66 + 4b2 Dg2,66 , and tenth scenario assumes that the GFRP laminae are arranged as
(0/90/90/0/0/90/90/0)s . The eleventh scenario postulates that the

∫ L ∫ L ∫ L
′ ′
Bm×m = {S(z) }m×1 〈S(z) 〉1×m dz, Bm×m = {S (z) }m×1 〈S (z) 〉1×m dz, Bm×m = {S′′(z) }m×1 〈S′′(z) 〉1×m dz,
0
∫ L
0
∫ L ∫
0
L
(69)
′ ′ ′
Fm×m = {S(z) }m×1 〈S(z) 〉1×m dz, Fm×m = {S (z) }m×1 〈S (z) 〉1×m dz, T m×m = {S (z) }m×1 〈S(z) 〉1×m dz
0 0 0

GFRP laminate is isotropic with an modulus of elasticity Eg1,z =


From Eqs. (66), by applying a condition of non-trivial solution, one 36.87 GPa and a Poisson’s ratio μg1,sz = 0.33 (denoted as “Isotropic
obtains the elastic buckling load P. material” in Table 3). In the twelfth scenario, only the bare steel beam is
considered. For all scenarios, the steel beam is subjected to uniform
5. Validations and examples positive bending moment M applied at two beam ends (i.e., Case 1 in
Table 1). It is required to validate the buckling moments predicted by
5.1. Example 1: Verification of the closed form solution in Case 1 (i.e., the present closed form solution developed in Eqs. (46) against those
Eqs. (46)) predicted by the 3-dimensional finite element analyses (3D FEA solu­
tion) conducted in ABAQUS.
Description of the problem: A simply supported steel beam with a span Description of the 3D FEA solution: A general numerical model of the
of L = 5 m is considered in the present example. The steel cross-section GFRP-strengthened beam is created in ABAQUS by using input data (i.e.,
is W250x45 that has a depth of h = 266 mm, a flange width of b = inp. files). The wide flange steel beam, adhesive layers 1 and 2, and
148 mm, a flange thickness of tf = 13 mm and a web thickness of tw = GFRP laminates 1 and 2 are meshed by using 8-node C3D8R brick ele­
7.6 mm. The compression flange of the beam (i.e., the top flange) is ments. Reduced integrations are used in order to avoid volumetric
strengthened with a 20 mm- thick GFRP laminate (i.e., Case 1 in locking and each element has an integration point located at its centroid.
Table 1). The GFRP is assumed as a symmetrically balanced laminate of The element has 8 nodes and each node has three translations. In the
16 equal-thick orthotropic laminae. The mechanical properties of each present study, it is assumed that the mesh model is controlled by 9 in­
lamina are taken as GF600 material [3], of which the longitudinal dependent parameters from n1 to n9 (Fig. 3a-d), in which n1 is the
modulus of elasticity is Ek,z = 36.87 GPa, the lateral and transverse number of elements across the overhang parts of the flanges, n2 is the
moduli of elasticity are Ek,s = Ek,n = 10.64 GPa, the shear moduli are number of elements across the thickness of GFRP laminate 2, n3 is the
Gk,sn = Gk,zn = 3.27GPa, Gk,sz = 3.93GPa, and Poisson’s ratio are μk,sn = number of elements across the thickness of adhesive layer 2, n4 is the
μk,zn = 0.28, μk,sz = 0.33. Twelve scenarios of the lamina orientation number of elements across the flange thicknesses, n5 is the number of
elements across the clear web depth, n6 is the number of elements across
angles are assumed for the GFRP laminate in the present example in
the web thickness, n7 is the number of elements across the thickness of
order to investigate the effect of fiber orientation angles on the elastic
adhesive layer 1, n8 is the number of elements across the thickness of
buckling resistances. The first nine scenarios assume that the GFRP
GFRP laminate 1, and n9 is the number of elements along the beam span.
laminae are stacked at orientation angles (i/j/i/j/i/j/i/j)s where i can be
Fiber angle orientations of GFRP laminae are inputted into the 3D FEA
a value of 0, 10, 20, 30, 45, 60, 70, 80, 900 degrees, and j can be

Fig. 3. Mesh parameters of the GFRP-strengthened beam in the 3D FEA model (a,b) cross-section view, (c) Elements i and j of two adjoining parts sharing the same
four nodes on interface, (d) three-dimensional view.

13
P. Van Pham Engineering Structures 256 (2022) 114046

solutions based on keyword *Solid section. subsequent step is written to conduct buckling analyses based on
Five steel web stiffeners (Fig. 3d) may be added to the model to keyword *Buckle.
reduce the effect of web distortion when the system is globally buckled A mesh study is conducted based on different numbers of elements (i.
[17–18]. Dimensions of the web stiffeners may be required more strictly e., n1 , from n4 to n9 ) to investigate the convergence of the elastic
based on design specifications. However, in the scope of the present buckling moments in the 3D FEA solution. Such a study is presented in
study, it is assumed that the lateral dimension of the stiffener equals to Table 2 for Scenario 5 (i.e., GFRP laminates stacked by 45/ − 45◦ lamina
the lateral overhanging length of the flange while the transverse angles). Six meshes are investigated, in which the numbers of elements
dimension of the stiffener equals to the clear web depth. Each half of a are increased for Meshes from 1 to 6. Stiffeners are removed in Meshes 1
stiffener is modelled by n1 elements in the lateral direction, and by n5 and 2 while they are applied to Meshes 3 to 6 with different thicknesses.
elements in the vertical direction, and by one element in the longitudinal The effect of web distortions can be obviously observed in Mesh 2
direction. At the interface between two adjoining parts A and B (e.g., an (without using web stiffeners) while it is almost eliminated in Mesh 4
adhesive layer and the steel beam, an adhesive layer and a GFRP lami­ (with using five web stiffeners), as depicted in Fig. 5. The buckling
nate, or a stiffener and the steel beam), two elements i and j of the moment 186.3 kN.m based on the finest mesh (i.e., Mesh 6) is taken as a
adjoining parts sharing the same four nodes on the part interface reference solution. The buckling moments based on Meshes 1, 2 and 3
(Fig. 3c). are relatively different to, while those based on Meshes 4 and 5 are close
Boundary conditions (i.e., movable and pinned supports of the sim­ to, the reference solution (Table 2). This reflects that the buckling mo­
ply supported steel beam) in the 3D FEA model are presented in Fig. 4a- ments are almost converged based on Meshes 4, 5, 6. However, the
b, while loading conditions (i.e., end moments) are applied by using 4 buckling analysis times of Meshes 4 to 6 are relatively long, as shown in
axial loads 1.0 kN applied at 4 centroids of the top and bottom flanges at Table 2. Mesh 4 consumes the least time among Meshes 4, 5, 6. Mean­
the beam ends (Fig. 4c). In the present problem (i.e., a wide flange steel while, the buckling moment based on Mesh 4 is only 0.5% different to
beam is bonded with GFRP laminate 1 through adhesive layer 1), an that of Mesh 6. Based on the mesh study, the results based on Mesh 4 are
intermediate step with a keyword *Model Change is applied to remove considered as the buckling moments of the present 3D FEA solution.
adhesive layer 2 and GFRP laminate 2 from the general model. Then, a Description of the present solution: The buckling moment predicted by

Fig. 4. Boundary and loading conditions in the 3D FEA model (a) Boundary conditions at the axially movable support, (b) boundary conditions at the pinned support,
and (c) loading conditions.

Table 2
Mesh study of the 3D FEA solution.
Mesh n1 n4 n5 n6 n7 n8 n9 Stiffener thickness Analysis time (minutes)(2) Buckling moment (kN.m) Difference(1)%

1 4 1 10 2 1 2 100 0.0 0.9 204.9 9.5


2 8 2 20 4 2 4 250 0.0 ̴ 4.2 176.0 − 5.9
3 8 2 20 4 2 4 250 20 4.2 189.5 1.3
4 10 4 30 4 4 8 300 16.7 12.8 188.0 0.5
5 12 4 30 4 4 8 400 12.5 26.0 187.6 0.3
6 12 4 30 4 4 8 500 10.0 39.0 187.0 0.0

(1) % difference = [Moment based on Mesh i – Moment based on Mesh 6]*100% / Moment based on Mesh 6, (i = 1,2,…,6). (2) ABAQUS Finite Element Analyses are
run in a computer with two Intel (Santa Clara, California) Xeon processors with an E5-24300 and a central processing unit with a speed of 3.20 and 64.0 GB of RAM.

Table 3
Buckling moment comparisons between the present study and the 3D FEA solution in Example 1.
Scenario GFRP fiber orientation (degree) Present study(kN.m) 3D FEA(kN.m) % difference
(1) (2) |(1)-(2)|0.100% / (2)

1 (0/0/0/0/0/0/0/0)s 160.8 166.1 3.2


2 (10/ − 10/10/ − 10/10/ − 10/10/ − 10)s 165.8 159.1 4.2
3 (20/ − 20/20/ − 20/20/ − 20/20/ − 20)s 177.5 170.9 3.8
4 (30/ − 30/30/ − 30/30/ − 30/30/ − 30)s 188.9 180.9 4.4
5 (45/ − 45/45/ − 45/45/ − 45/45/ − 45)s 196.0 188.0 4.3
6 (60/ − 60/60/ − 60/60/ − 60/60/ − 60)s 187.3 187.5 0.1
7 (70/ − 70/70/ − 70/70/ − 70/70/ − 70)s 175.0 180.5 3.0
8 (80/ − 80/80/ − 80/80/ − 80/80/ − 80)s 163.1 170.0 4.0
9 (90/90/90/90/90/90/90/90)s 158.1 152.7 3.5
10 (0/90/90/0/0/90/90/0)s 159.5 162.4 1.8
11 Isotropic material 218.1 208.7 4.5
12 Bare steel beam (not strengthened) 125.3 126.0 0.6

14
P. Van Pham Engineering Structures 256 (2022) 114046

the present study and the 3D FEA solution are within 4.4%. This in­
dicates that the buckling moments predicted by the present study are in
excellent agreements with those of the 3D FEA solutions. Also, this im­
plies that the buckling shape functions postulated in the present solution
(i.e., Eqs. (43)) may approximately capture the buckling deformation of
the 3D FEA solution. The low differences between the two solutions may
be originated by several simplified kinematic assumptions made in the
present study. Although five web stiffeners have been added to the 3D
FEA models, they may not completely eliminate the effect of web
distortion when the system is buckled. The effect is captured by the 3D
FEA solution but it is neglected in the present solution.
The difference between the peak and the smallest buckling moments
of Scenarios from 1 to 9 based on the 3D FEA solution (i.e., 188.0 and
152.7 kN.m, respectively) is 23.2%. Meanwhile, that based on the pre­
Fig. 5. Steel beam cross-sections near beam supports in the elastic lateral sent solution (i.e., 196.0 and 158.1 kN.m, respectively) is 24.0%. Both
torsional buckling mode based on (a) Mesh 2 and (b) Mesh 4 of the 3D FEA differences are relatively large and they reflect that the buckling mo­
mesh study. ments are significantly influenced by the GFRP fiber orientation
arrangements.
the present study is based on a closed form solution (i.e., Eqs. (46)), in In Scenario # 10, the GFRP fiber orientation angles are arranged as
which factors βb , βc are defined in Eqs. (47) and they can be evaluated (0/90/90/0/0/90/90/0)s (Table 3) (i.e., not in the form of a ± stacking
from the material and mechanical prosperities defined in Appendix A. sequence). The buckling moment based on the present solution is 159.5
The implementation of the closed form solution maybe relatively simple, kN.m while that based on 3D FEA solution is 162.4 kN.m, a difference of
short and fast compared to that of the 3D FEA solution. An analysis of the only 1.8%. Therefore, the results based on the two solutions are in a very
closed form solution implemented in MATAB consumes less than 1 s. No close agreement. It is also observed that the buckling moments are
mesh study is required for the present closed form solution. within a limit between the buckling moment of Scenario 9 and that of
Scenario 1.
Comparisons of the buckling moment predictions: Fig. 6 and Table 3
The buckling moment based on the present study is 218.1 kN.m for
present the comparison of buckling moments against different GFRP
Scenario #11 (of which the GFRP laminate is treated as an isotropic
fiber orientations from Scenarios 1 through 9 as obtained from the
material (Table 3)), while that is 160.8 kN.m for Scenario #1 (of which
present closed form solution and the 3D FEA solution. The 3D FEA so­
the GFRP laminate is stacked by only longitudinal fibers), corresponding
lution is taken as a reference solution to validate the prediction of the
to a difference of 26.3%. Based on the 3D FEA solution, the buckling
present study. It is observed in Fig. 6 that both buckling moments of the
moments and the difference are 188.1 kN.m, 166.1 kN.m, and 20.4%,
present study and the 3D FEA solution have a concave form, in which the
respectively. Both GFRP laminates in Scenarios 11 and 1 have the same
peak values of the curves corresponds to Scenario 5 with the GFRP fiber
longitudinal modulus of elasticity Ek,z = 36.87 GPa and Poisson’s ratio
orientation arrangements of 450 / − 450 , while the smallest values
μk,sz = 0.33, but Scenario 1 assumes the GFRP laminate as a orthotropic
correspond to Scenario 9 with the fiber arrangements of 900 . Based on
material while Scenario 11 assumes it as an isotropic material. There­
Eqs. (1), (30), (31) and Appendix A, it can be explained that Scenario 5
fore, it can be observed that the buckling resistance of GFRP-
with the fiber orientation arrangements 450 / − 450 maximize Q66 and strengthened steel beams significantly depends on the orthotropic or
the shear and torsional rigidities (i.e., Ag1,66 b1 and b1 Dg1,66 , as presented isotropic properties of the GFRP laminate. Further assessments of the
in Eqs. (30) and (31)). Such an observation in the present study is property dependences will be investigated in the parametric studies of
consistent with previous solutions [1–3,20–23]. In Table 3, the peak the present study.
value based on the 3D FEA solution is 188.0 kN.m while that based on Based on the buckling moments predicted by the present study as
the present study is 196.0 kN.m, corresponding to a difference of only presented in Table 3, the effectiveness of GFRP strengthening can be
4.3%. Meanwhile, the smallest value (i.e., Scenario 9) based on the 3D evaluated. For example, the buckling moment of the bare steel beam is
FEA solution is 152.7 kN.m while that based on the present study is 125.3 kN.m, while that of the GFRP-strengthened beam in Scenario 1 is
158.1 kN.m, corresponding to a difference of 3.5%. For other Scenarios 160.8 kN.m, corresponding to an increase of 22.1%. Also, that of the
from 1 to 4 and from 6 to 8, it is observed that the differences between GFRP-strengthened beam in Scenario 5 is 196.0 kN.m, an increase of
36.1%.

5.2. Example 2: Validation of the closed form solution in Case 2 (i.e., Eqs.
(49))

Description of the problem: A simply supported GFRP-strengthened


steel beam subjected to a midspan point load Py (yP , L/2) is considered.
The steel beam has a span of L = 4.0 m and a cross-section W150 × 13
(h = 148 mm, b = 100 mm, tf = 4.9 mm and tw = 4.3 mm). The top
flange of the steel beam is strengthened with a 10 mm-thick GFRP
laminate by using a 1 mm-thick adhesive layer. Elastic modulus and
Poisson’s ratio of the steel are Es = 200 GPa and μ = 0.3, respectively,
while those of the adhesive are Ea1 = 3.182 GPa and μ = 0.3. It is
assumed that the GFRP laminate is a symmetrically balanced laminate of
16 equal-thick orthotropic laminae. The properties of each lamina are
Fig. 6. Comparison of buckling moments against different GFRP fiber orien­ taken as GF800 material [3], of which the longitudinal modulus of
tations of Scenarios 1 through 9 as predicted by the present study and the 3D elasticity is Ek,z = 45.95 GPa, the lateral and transverse moduli of
FEA solution. elasticity are Ek,s = Ek,n = 14.56 GPa, the shear stiffnesses are Gk,sn =

15
P. Van Pham Engineering Structures 256 (2022) 114046

Gk,zn = 4.50 GPa, Gk,sz = 5.51 GPa, and the Poisson’s ratio are μk,sn = study. A steel beam with a span of L = 5.0 m and with a prismatic cross-
μk,zn = 0.25, μk,sz = 0.30. The effect of load positions is considered in section W250x45 is considered. Elastic modulus and Poisson’s ratio of
the present example, in which the transverse coordinate yP of load Py (yP , the steel are Es = 200 GPa and μ = 0.3, respectively, while those of the
L/2) is set as − hb /2, 0, and hb /2 corresponding to the load positions at adhesive are Ea1 = 3.182 GPa and μ = 0.3. In Cases 3 and 4, a 20 mm-
the centroid of the steel top flange, that of steel section, and that of the thick GFRP laminate is strengthened to compressed flange of the steel
steel bottom flange, respectively. For each position of the load, four beam, while in Cases 5 and 6, the beam is bonded with two 20 mm- thick
different scenarios of the GFRP fiber orientation angles (i.e.,00 , ± 450 , GFRP laminates (Table 5). Two types of GFRP materials are considered
in Cases 3 to 6. The first type is an isotropic material (with Ek,z =
900 and isotropic arrangements) are assumed (Table 4), in which the
isotropic treatment assumes a longitudinal modulus of elasticity of 36.87 GPa and μk,sz = 0.33) while the second one is an orthotropic GFRP
Ek,z = 45.95 GPa and a Poisson’s ratio of μk,sz = 0.30. It is required to material assembled by 16 GF600 laminae stacked at angles of
validate the buckling load as predicted by the present closed form so­ (0/90/90/0/0/90/90/0)s . The GFRP laminates are bonded to the steel
lution (i.e., Eqs. (49)) against those of the 3D FEA solution in ABAQUS. beam by using 1 mm- thick adhesive layers. It is required to validate the
The mesh study conducted for the 3D FEA solution in the present buckling loads predicted by the present eigenvalue problems against
example is in a manner similar to that conducted in Example 1. those obtained from the 3D FEA solutions under ABAQUS. The mesh
Comparisons of the buckling moment predictions: Table 4 presents the study conducted for the 3D FEA solution in the present example is in a
comparisons of the buckling loads predicted by the present study and manner similar to that conducted in Example 1.
those of the 3D FEA solution. As observed, the highest difference be­ Validation of the eigenvalue problem developed for Case 3: It is assumed
tween the two solutions is 3.5% for the scenario of isotropic GFRP plate that end moment ratios κ are taken as 0 or − 0.5 to validate the eigen­
andyP = 0. While the lowest difference between the two solutions is value problem (as depicted in Table 1). The buckling moment of the
0.3% for the scenario of orthotropic GFRP plate with fiber orientation present solution is evaluated through Eqs. (54). Because the governing
buckling displacements are assumed as a summation of m trigonometric
angle of 900 andyP = hb /2. As observed, the buckling loads based on the
terms of sin functions (i.e., Eqs. (51), (52)), a convergent study has been
present study are in an excellent agreement with those based on the 3D
conducted for the present solution and it is found the buckling loads are
FEA solutions.
converged when m⩾9(terms). In Table 5, these buckling moments are
Similar to the comments obtained in Example 1, it is observed in the
presented next to those of the 3D FEA solutions. Whenκ = 0, the buckling
present example that the highest buckling loads are of the systems
moment of the beam strengthened with the isotropic GFRP laminate
strengthened with GFRP laminates stacked by 450 / − 450 lamina angles,
based on the present solution is 388.2 kN.m, while that based on the 3D
while the lowest values are of the systems strengthened with GFRP
FEA solution is 377.6 kN.m, corresponding to a difference of only 2.7%.
laminates stacked by 900 lamina angles (Table 4). Also, the buckling
Meanwhile, the buckling resistance of the system with the orthotropic
loads of the systems with the orthotropic GFRP laminates are consid­
GFRP laminate based on the present study is 291.5 kN.m and that based
erably lower than those with isotropic plates. For example, when load is
on the 3D FEA solution is 301.0 kN.m, a difference of 3.2%. It is
applied at the centroid of the steel beam (i.e., yP =0), the buckling load
observed that the buckling moments predicted by the present study and
based on the present study for the system with 450 / − 450 GFRP laminae
those by the 3D FEA solutions are in close agreements. Very similar
is 34.0 kN, while that with 00 GFRP laminae is 27.1 kN, that with 900 comments are applied to the buckling loads of the systems under end
GFRP laminae is 26.3 kN. These values are 11.5, 29.4 and 31.5%, moment ratio κ= − 0.5.
respectively lower than the buckling load of the system strengthened
Validation of the eigenvalue problem developed for Case 4: A point load
with the isotropic GFRP laminate (i.e., 38.4 kN).
Py (yP , zP ) is applied at coordinate zP = 2.0 m while the transverse co­
The effect of load positions on the buckling loads of the GFRP-
strengthened beams can be also observed in Table 4. For example of ordinates yP are taken as − hb /2, 0, or hb /2(as depicted in Table 1). The
the system strengthened with 450 / − 450 GFRP laminae, the buckling solution based on the present study is based on Eqs. (54). The buckling
load corresponding to yP =hb /2 (bottom flange) is 41.1 kN, while that loads obtained from the present study and those of the 3D FEA solutions
corresponding to yP =− hb /2 (top flange) is 28.1 kN, corresponding to a are compared. As presented in Table 5, it is observed that the differences
difference of 31.6%. between the two solutions are from 1.5 to 5.3%, indicating that the
buckling loads predicted by the present study are close to those of the 3D
FEA solutions. Also, the effect of the load positions on the buckling loads
5.3. Example 3: Validations of the eigenvalue problems developed for of the GFRP-strengthened beams are significant. For example, the
Cases 3, 4, 5 and 6 buckling loads of the system with the orthotropic GFRP laminate are
141.2 kN when the point load is applied at yP =− hb /2, that is 235.1 kN
Description of the problem: This example is conducted to validate the when the point load is applied at yP =hb /2, corresponding to a difference
eigenvalue problems developed for Cases 3, 4, 5, and 6 in the present

Table 4
Buckling load comparisons between the present study and 3D FEA solutions in Example 2.
Load position yP GFRP orientation scenarios (degree) Present study (kN) 3D FEA (kN) % difference
(2) |(1)-(2)|*100% / (2)
(1)

− hb /2 (0/0/0/0/0/0/0/0)s 21.1 21.3 0.9


(45/ − 45/45/ − 45/45/ − 45/45/ − 45)s 28.1 27.5 2.1
(90/90/90/90/90/90/90/90)s 20.7 21.1 1.9
Isotropic 32.1 31.3 2.5
0 (0/0/0/0/0/0/0/0)s 27.1 27.7 2.2
(45/ − 45/45/ − 45/45/ − 45/45/ − 45)s 34.0 33.2 2.4
(90/90/90/90/90/90/90/90)s 26.3 26.7 1.5
Isotropic 38.4 37.1 3.5
hb /2 (0/0/0/0/0/0/0/0)s 35.0 35.8 2.2
(45/ − 45/45/ − 45/45/ − 45/45/ − 45)s 41.1 39.9 2.9
(90/90/90/90/90/90/90/90)s 33.5 33.4 0.3
Isotropic 46.0 44.7 2.9

16
P. Van Pham Engineering Structures 256 (2022) 114046

Table 5
Buckling load comparisons between the present study and the 3D FEA solution in Example 5.
Cases Description of the problem Isotropic treatment Orthotropic treatment

Configuration Given condition PS 3D FEA % diff.* PS 3D FEA % diff.*


(1) (2) (1–2)/(2) (3) (4) (4–3)/4

3 κ =0 388.2 377.6 2.7 291.5 301.0 3.2


κ = − 0.5 528.6 511.5 3.2 398.4 416.4 4.3

4 yP = − hb /2 203.8 193.0 5.3 141.2 143.3 1.5


yP = 0 249.0 236.7 4.9 182.8 186.6 2.0
yP = hb /2 303.1 292.8 3.3 235.1 245.8 4.4
5 η =0 712.5 712.9 0.1 658.0 698.6 5.8
η = hb 528.0 520.9 1.3 415.9 437.4 4.9

6 η =0 194.4 181 6.9 186.7 178.2 4.6


η = hb 180.0 167 7.2 152.3 153.5 0.8

Note: 1. Unit of M is kN.m, and that of P is kN; 2. * % diff. = Percentage difference

of 40%. the buckling shapes are observed as a globally lateral buckling mode, in
Validation of the eigenvalue problem developed for Case 5: A compres­ which torsional angles almost vanish (i.e., δθzb (z) ≈ 0) (Fig. 8). This
sion force Pz = P and a uniform bending moment M = ηP are applied to indicates that the present study naturally capture the globally lateral
the beam ends (as depicted in Table 1), where η is taken as 0 or hb . The buckling behavior of the system. Based on the thus observed charac­
present solution is based on Eqs. (60). Buckling loads P of the present teristic, a closed form solution for the lateral buckling load P in Case 5
solution are evaluated and they are presented next to those of the 3D with η = 0 can be developed from Eqs. (60), (61) and (62) as
FEA solutions in Table 5. As observed from the table, the differences of

( / / )( )
α2 Ag1,11 b3 12 + α2 Ag2,11 b2 12 4χ a1,yy + 4χ a2,yy + Ag1,66 b1 + Ag2,66 b
P = α2 Es Is,yy + ( / / ) ( ) (70)
α2 Ag1,11 b3 12 + α2 Ag2,11 b2 12 + 4χ a1,yy + 4χ a2,yy + Ag1,66 b1 + Ag2,66 b

buckling loads between the two solutions are ranged from 0.1 to 5.8%. In which the mechanical/ material properties on the right hand side
Thus, the buckling loads predicted by the present study are again in of Eq. (70) have been defined in Appendix A.
excellent agreements with those of the 3D FEA solutions. The
strengthening effectiveness of the present GFRP-strengthened beam- 6. Further validations and parametric studies
columns can be estimated based on the table. For example, the buckling
load of the bare beam (not strengthened) can be evaluated as 306 kN The simply supported steel beam strengthened with a GFRP laminate
when η = hb , while that is 415.9 kN when η = hb for the steel beam stacked based on Scenarios 1, 5 and 11 as presented in Example 1 are
strengthened with the orthotropic GFRP laminate, corresponding to a taken as reference structures in the present parametric studies. Thick­
buckling load increase by 26.4%. nesses and material properties of the GFRP laminate and the adhesive
Validation of the eigenvalue problem developed for Case 6: A cantilever layer are varied to investigate the buckling moment changes. The
beam-column strengthened with two GFRP laminates is subjected to a buckling loads presented in the parametric studies are based on both the
compression force P and a uniform bending moment M = ηP(as present closed form solution (i.e., Eqs. (46)) and the 3D FEA solution.
depicted in Table 1), where η is taken as 0 or hb . The solution based on When one parameter is varied, other parameters are kept unchanged as
the present study is based on Eqs. (66) in which buckling loads are provided in Example 1.
converged when m = 9 (terms) is taken. The buckling loads P predicted
by the present study are provided next to those of the 3D FEA solutions 6.1. The effects of GFRP laminate thickness and lamina properties on the
in Table 5. As observed, the differences of the buckling loads between elastic buckling moments
the two solutions are ranged from 1.7 to 7.2%. As a result, the buckling
loads predicted by the present study are as approximately good as those Four investigations are conducted in the present parametric study:
of the 3D FEA solution. It is supposed that the overestimation of the (1) The first investigation is the effect of the GFRP laminate thickness,
present study may be originated from the postulated buckling shape that is varied from 2.5 to 30 mm. For a given thickness of the orthotropic
functions and the neglection of web distortion effect in the steel beam GFRP laminate in Scenarios 1 and 5 (i.e., with fiber angles of 00 and
[5,7,8,19]. 450 / − 450 , respectively), it is assumed that the thickness is stacked by
Comparison of mode shapes predicted by the present study and that by the using 1.25 mm-thick GF600 laminae. This means that a 2.5 mm – thick
3D FEA solution: Fig. 7a,b present a comparison of mode shapes (i.e., GFRP laminate is created by stacking two 1.25 mm-thick GF600
δUb (z), δθzb (z), δU1b (z), δθy1b (z), δU2b (z), δθy2b (z)) as obtained by the laminae, while a 30 mm-thick GFRP laminate is stacked by 24 GF600
present study and those of the 3D FEA solution. These mode shapes laminae. (2) The second investigation is the effect of the longitudinal
correspond to the buckling loads of the steel beam strengthened with the modulus of elasticity of GFRP laminae, that is varied from 5.0 to 55 GPa.
isotropic GFRP laminate in Case 5 with η = 0 in Table 5. It is observed (3) The third investigation is the effect of the lateral modulus of elas­
from the figure that the mode shapes obtained from the present study ticity of GFRP laminae, that is varied from 0.0 to 30 GPa. And (4) the
and those from the 3D FEA solution are in excellent agreements. Also, fourth investigation is the effect of the shear modulus of GFRP laminae,
that is varied from 1.0 to 6.0 GPa.

17
P. Van Pham Engineering Structures 256 (2022) 114046

Fig. 7. Buckling mode shapes of Case 5 with isotropic GFRP laminate and with η = 0.

Fig. 8. Buckling mode shapes of Case 5 with isotropic GFRP laminate and with η = hb.

Fig. 9a-d respectively present the relationship between the elastic modulus. For the effect of the GFRP lamina shear modulus (Fig. 9d), an
buckling moment against the GFRP laminate thickness, the longitudinal opposite trend is observed that the buckling moments of Scenario 1
modulus of elasticity, the lateral modulus of elasticity and the shear significantly increase when the shear modulus is varied from 1.0 to 6.0
modulus, as obtained from the present study and the 3D FEA solutions. It GPa, but those of Scenario 5 are insensitive to the shear moduli. Ex­
is observed that the buckling moments predicted by the present study planations for the observations in Fig. 9b,c,d are presented in the
agree with those of the 3D FEA solutions. following.
The effect of the GFRP laminate thickness on the system buckling As discussed in Example 1, the buckling moments of the GFRP-
moment is presented in Fig. 9a. When the thickness is increased from 2.5 strengthened beams significantly depend on the GFRP laminate shear
to 30 mm, the buckling moment is increased from 128.4 to 203.2 kN.m and torsional rigidities, those are mainly contributed by factor Q66
for Scenario 1, from 129.8 to 269.1 kN.m for Scenario 5, and from 130.9 (defined in Appendix A). When fiber angles θ = 00 (i.e., Scenario 1), one
to 311.4 kN.m for Scenario 11, corresponding to the moment increases obtains Q66 = Q66 = Gk,sz . This indicates that for Scenario 1, factor Q66
of 36.8%, 51.8% and 58.0%, respectively. This indicates that the only depends on the GFRP lamina shear modulus Gk,sz , while it is
buckling moment of the GFRP-strengthened beam is sensitive to the
insensitive to the longitudinal and lateral moduli of the GFRP laminae.
GFRP laminate thicknesses.
As a result, the buckling moments of Scenario 1 significantly depend on
The effect of the GFRP lamina longitudinal modulus of elasticity is
the GFRP lamina shear modulus (Fig. 9d), but they are insensitive to the
presented in Fig. 9b. When the modulus is varied from 5.0 to 55 GPa, the
GFRP longitudinal and lateral moduli (Fig. 9b,c). In contrast, when fiber
buckling moment is increased from 157.5 to 162.5 kN.m for Scenario 1,
angles θ = ±450 (i.e., Scenario 5), one has Q66 = (Q11 + Q22 − 2Q12 )/4.
from 150.1 to 217.0 kN.m for Scenario 5, and from 143.1 to 250.3 kN.m
Because Q11 , Q22 , Q12 depend on the longitudinal and lateral moduli of
for Scenario 11, corresponding to the moment increases of 3.1%, 30.8%
elasticity of GFRP laminae (i.e., Ek,z , Ek,s in Appendix A). This indicates
and 42.8%, respectively. It is interesting to observe that the effects of the
longitudinal modulus on the buckling load are high for Scenario 5 and that factor Q66 only depends on the lateral and longitudinal moduli,
11, but it is low for Scenario 1. For the effect of the GFRP lamina lateral while it is insensitive to the GFRP shear modulus for Scenario 5. As a
modulus of elasticity (Fig. 9c), it is observed that the buckling moments result, the buckling moments of Scenario 5 significantly depend on the
of Scenario 5 are considerably increased when the modulus is varied GFRP lamina lateral and longitudinal moduli (Fig. 9b,c), but they are
from 1.0 to 30 GPa, but those of Scenarios 1 and 11 are insensitive to the insensitive to the GFRP shear modulus (Fig. 9d).

18
P. Van Pham Engineering Structures 256 (2022) 114046

Fig. 9. Effects of the (a) thickness of the GFRP laminate, (b) GFRP lamina longitudinal modulus of elasticity, (c) GFRP lamina lateral modulus, and (d) GFRP lamina
shear modulus on the buckling moment.

6.2. The effects of shear modulus and thickness of the adhesive layer on kN.m for Scenario 1, from 125 to 189.6 kN.m for Scenario 5, and from
the elastic buckling moments 125 to 208.1 kN.m for Scenario 11, corresponding to moment increases
of 21.2%, 34.1% and 39.9%, respectively. In contrast, when the adhe­
Two investigations are conducted in the present parametric study: sive shear modulus is increased from 0.5 to 2.0 GPa, the moments are
(1) the first investigation is the effect of the adhesive shear modulus, that slightly increased from 158.7 to 162.1 kN.m for Scenario 1, from 189.6
is varied from 0 to 2.0 GPa, and (2) the second investigation is the effect to 198.5 kN.m for Scenario 5, and from 208.1 to 221.7 kN.m for Scenario
of the adhesive layer thickness, that is varied from 0.0 to 4.0 mm. 11, corresponding to the moment increases of only 2.1%, 4.5% and
Fig. 10a presents the relationship of the elastic buckling moments 6.1%, respectively. This indicates that the buckling moments strongly
and the adhesive shear modulus, while Fig. 10b presents the relationship depend on the shear moduli those are lower than 0.5 GPa. However,
of the elastic buckling moments and the adhesive layer thicknesses, as they may not be sensitive to the adhesive shear moduli those are greater
obtained from the present study and the 3D FEA solution. In general, it is than 0.5 GPa.
observed that the buckling moments predicted by the present study In Fig. 10b, when the thickness of the adhesive layer is increased
agree with those of the 3D FEA solutions. from 0.1 to 4.0 mm, the buckling moments are increased from 157.6 to
In Fig. 10a, when the adhesive shear modulus is increased from 0 to 171.0 kN.m for Scenario 1, from 193.8 to 203.8 kN.m for Scenario 5, and
0.5 GPa, the buckling moments are quickly changed from 125 to 158.7 from 217.3 to 222.5 kN.m for Scenario 11, corresponding to small

Fig. 10. Effects of the (a) adhesive shear modulus, (b) adhesive layer thickness on the elastic buckling moment.

19
P. Van Pham Engineering Structures 256 (2022) 114046

increases of the buckling moments as 7.8%, 4.9% and 2.3%, strengthened with an isotropic GFRP laminate, as discussed in
respectively. Examples 1–3.
(4) The elastic buckling resistances of GFRP-strengthened beams/
7. Conclusions columns were significantly influenced by load height positions
(as presented in Example 2 and Example 3), by GFRP laminate
An innovated theory, two novel closed form solutions and four thicknesses, and by adhesive shear moduli (as observed in Para­
eigenvalue problems for the elastic lateral torsional buckling analysis of metric studies). Meanwhile, the elastic buckling resistances of
steel beams/columns strengthened with symmetrically balanced GFRP steel members strengthened with orthotropic GFRP laminates
laminates have been successfully developed and validated in the present stacked by fiber angles θ = 00 are depended on the GFRP shear
study. The theory was developed based on a principle of total stationary modulus, but they were insensitive to the GFRP longitudinal and
buckling energy. The closed form solutions and the eigenvalue problems lateral moduli. On the other hand, the buckling resistances of
were developed based on the innovated theory and based on postulated steel members strengthened with orthotropic GFRP laminates
buckling displacement functions. The present theory/solutions captured stacked by fiber angles ±45 are significantly depended on the
the partial interaction between the GFRP laminate and the steel beam, GFRP lateral and longitudinal moduli, but they were insensitive
different stacking sequences and orthotropic properties of the GFRP to the GFRP shear modulus. Further investigations for the effects
laminate, orthotropic properties of the GFRP laminates, shear de­ of the adhesive layer thicknesses, the GFRP longitudinal and
formations in the GFRP laminate, and local and global warping de­ lateral moduli of elasticity, and GFRP shear moduli were also
formations. Through the Examples and the Parameter studies conducted quantified in the present study.
in the present study, main conclusions are achieved as follows: (5) Based on the present study, the gain of elastic lateral torsional
buckling resistances of steel beams/columns strengthened with
(1) The closed form solutions and the eigenvalue solutions developed GFRP laminates can be quantified. When compared to the buck­
in the present study were successfully validated against numeri­ ling resistances of bare beams, those of the corresponding beams
cal solutions based on 3D FEA analyses. The present solutions are strengthened with GFRP laminates in Example 1 gained from
simple and fast for the prediction of the elastic lateral torsional 22.1 to 36.1%. The increases were from 36.3 to 58.7% for the
buckling resistances of GFRP-strengthened beam/columns and simply supported GFRP-strengthened beams subjected to point
thus they can be an alternative solution in lieu of the corre­ loads in Example 2, and from 15.6 to 42.1% for the GFRP-
sponding numerical solutions. To capture inelastic buckling re­ strengthened beam-columns in Example 3.
sistances, the elastic buckling solutions of the present study
would need to be adjusted by reliability-based factors to account CRediT authorship contribution statement
for initial mechanical and geometric imperfection conditions.
However, such adjustments are not considered in the present Phe Van Pham: Conceptualization, Data curation, Formal analysis,
study. Funding acquisition, Investigation, Methodology, Project administra­
(2) The buckling resistances of GFRP-strengthened beams were tion, Resources, Software, Supervision, Validation, Visualization, Roles/
significantly influenced by GFRP fiber orientation arrangements. Writing – original draft, Writing – review & editing.
As discussed in Examples 1–3 and Parametric studies, GFRP
laminates with fiber orientation angles of ±45 degrees maxi­
mized the laminate shear and torsional rigidities and thus maxi­ Declaration of Competing Interest
mized the elastic torsional buckling moments of the GFRP-
strengthened beams. Also, the steel beams strengthened with The authors declare that they have no known competing financial
GFRP laminates stacked by longitudinal and/or lateral fibers (i. interests or personal relationships that could have appeared to influence
e.,00 and/or 900 ) has lower buckling resistances. the work reported in this paper.
(3) The elastic buckling resistances of GFRP-strengthened steel
beams/columns are significantly depended on the isotropic or Acknowledgements
orthotropic property of the GFRP material. Given identical lon­
gitudinal modulus of elasticity and Poisson’s ratio for the GFRP This research is funded by Vietnam National Foundation for Science
laminate, the buckling resistance of a steel beam strengthened and Technology Development (NAFOSTED) under grant number
with an orthotropic GFRP laminate were typically lower than that 107.02-2019.12.

Appendix A. Evaluation of the material and sectional properties for GFRP-strengthened steel beams

A.1. Evaluation of the GFRP laminate stiffnesses postulated in Eqs. (1)

Plane stress-reduced stiffnesses of the kth GFRP lamina may be evaluated as [10]
/ / /
Qk,11 = Ek,z (1 − vzs vsz ), Qk,12 = vzs Ek,s (1 − vzs vsz ), Qk,22 = Ek,s (1 − vzs vsz ), Qk,66 = Gk,sz

In which parameters Ek,z , Ek,s , Gk,sz , and vk,zs ,vk,sz have been defined in Section 3.3. Based on a stacking sequence as depicted in Fig. A.1, transformed
lamina stiffnesses can be evaluated as

20
P. Van Pham Engineering Structures 256 (2022) 114046

Fig. A1. Transverse coordinates and lamina arrangement of a GFRP laminate.

( )
Qk,11 = Qk,11 cos4 γ k + 2 Qk,12 + 2Qk,66 sin2 γ k cos2 γ k + Qk,22 sin4 γk ,
( ) ( )
Qk,12 = Qk,11 + Qk,22 − 4Qk,66 sin2 γ k cos2 γ k + Qk,12 sin4 γ k + cos4 γ k ,
( )
Qk,22 = Qk,11 sin4 γk + 2 Qk,12 + 2Qk,66 sin2 γk cos2 γk + Qk,22 cos4 γk ,
( ) ( ) (A.2)
Qk,16 = Qk,11 − Qk,12 − 2Qk,66 sinγ k cos3 γk + Qk,12 − Qk,22 + 2Qk,66 sin3 γk cosγk ,
( ) 3 ( )
Qk,26 = Qk,11 − Qk,12 − 2Qk,66 sin γk cosγk + Qk,12 − Qk,22 + 2Qk,66 sinγ k cos3 γk ,
( ) ( )
Qk,66 = Qk,11 + Qk,22 − 2Qk,12 − 2Qk,66 sin2 γk cos2 γk + Qk,66 sin4 γk + cos4 γk

where γ k is the fiber orientation angle of the kth lamina.


Then, extensional and bending stiffnesses of the GFRP laminate can be evaluated as
/

n ∑
n
( )
Agi,ej = Qk,ej (yk+1 − yk ), Dgi,ej = Qk,ej yk+1 3 − yk 3 3 (A.3)
k=1 k=1

where e, j = 1, 2, ..., 6. Based on Eqs. (A.3), plate stiffnesses Agi,11 , Agi,66 , Dgi,11 , Dgi,66 in Eqs. (1) can be derived as:
/
Agi,11 = Agi,11 − Agi,12 2 Agi,22 ,
/
Agi,66 = Agi,66 − Agi,26 2 Agi,22 ,
/
Dgi,11 = Dgi,11 − Dgi,12 2 Dgi,22 ,
/
Dgi,66 = Dgi,66 − Dgi,26 2 Dgi,22

For isotropic GFRP laminates (i.e., it is assumed that Ek,z = Ek,s = Egi , Gk,sz = Ggi ), the laminate stiffnesses may be simplified as
/
Agi,11 bgi 3 12 = Egi Iyy,gi ,
Agi,66 bgi = Ggi Agi ,
4bgi Dgi,66
/ = Ggi Jgi ,
Dgi,11 bgi 3 12 = Egi Iωω,gi

in which Iyy,gi , Agi , Jgi , Iωω,gi are the moment of inertia about Y-axis, area, St. Venant torsional constant, and warping constant of the laminate cross-
section, respectively. Thickness bgi is b1 for GFRP laminate 1, and that is b2 for GFRP laminate 2 (as defined in Fig. 1b).

A.2. Evaluation of sectional properties in Eqs. (25) to (41)

( ) / / / /
Is,yy = h − 2tf tw 3 12 + 2tf b3 12, aa1,o = 1 4ta1 2 , aa2,o = 1 4ta2 2 ,
[ / ] ( ) /
3
Is,xx = 2 btf 3 12 + btf (hb /2)2 + tw h − 2tf 12, χ a1,yy = Ga1 Ia1,yy aa1,o , χ a2,yy = Ga2 Ia2,yy aa2,o ,
/ / ( )3 3 /
3 3 3
Is,ωω = hb b tf 24 + 2b tf 144 + h − 2tf tw 144, χ a1,a = Ga1 Aa1 aa1,o , χ a2,a = Ga2 Aa2 aa2,o ,
/ ( ) /
Js = 2btf 3 3 + h − 2tf tw 3 3, Aa1 = ta1 b1 , Aa2 = ta2 b2 , (A.6)
( ) / /
As = 2btf + h − 2tf tw , Ia1,yy = ta1 b1 3 12, Ia2,yy = ta2 b2 3 12,
[( )/ ]2 [( )/ ]2 /
Ja1 = tf − tg1 2 Aa1 + tf + tg1 2ta1 b1 ta1 3 3, da1,o = h + 2ta1 + tg1 , da2,o = h + 2ta2 + tg2 ,
[ / ]2 [ / ]2 /
Ja2 = tg2 2 Aa2 + tg2 ta2 b2 ta2 3 3, ha1,o = hb − tf − tg1 − 2ta1 , ha2,o = h − tg2 − 2ta2

Symbols on the right hand sides of equations in Eqs. (A.6) are the sectional dimensions defined in Fig. 1b and adhesive shear moduli defined in
Section 3.3.

21
P. Van Pham Engineering Structures 256 (2022) 114046

A.3. Evaluation of factors βi (i = 1,7) of Eqs. (44), (47) and (50)

( / )
β1 = Es Is,yy + b1 3 Ag1,11 12 α2 ,
( / ) /
β2 = Es Is,ωω + b1 3 Dg1,11 12 α2 + Gs Js + 4b1 Dg1,66 + Ga1 Ja1 + χ a1,yy ha1,o 2 + χ a1,a da1,o 2 α2 ,
/ 2
β3 = b1 Ag1,66 + 4χ a1,a α ,
( / )( / ) /
β4 = b1 Ag1,66 + 4χ a1,a α2 χ a1,yy ha1,o α χ a1,a da1,o + b1 Ag1,66 α, (A.7)
( / 2 )( / )2 /
β5 = b1 Ag1,66 + 4χ a1,a α χ a1,yy ha1,o α χ a1,a da1,o + Ag1,11 b1 3 12
( / ) ( )/
+2b1 Ag1,66 χ a1,yy ha1,o χ a1,a da1,o + b1 Ag1,66 + 4χ a1,yy α2 ,
/ /
β6 = 2χ a1,a da1,o α2 , β7 = b1 3 Ag1,11 α 12

in which α = π/L. Terms βi , (i = 1, 7) of Eqs. (A.7) can be evaluated from the material properties defined in Section 3.3 and from the mechanical
properties defined in Sections A.1 and A.2 above.

Appendix B. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.engstruct.2022.114046.

References [15] Phe PV. Solutions of the interfacial shear and normal stresses in plate flexural-
strengthened beams based on different complementary strain energy assumptions.
Eng Struct 2020;229:111567.
[1] Correia JR, Branco FA, Silva NMF, Camotim D, Silvestre N, First-order, buckling
[16] Van Pham P, Mohareb M, Fam A. Numerical and analytical investigation for
and post-buckling behavior of GFRP pultruded beams. Part 1: Experimental study,
ultimate capacity of steel beams strengthened with GFRP plates. J Eng Struct 2021;
Comp Str 89;2011:2052–64.
243:112668. https://doi.org/10.1016/j.engstruct.2021.112668.
[2] Silva NMF, Camotim D, Silvestre N, Correia JR, Branco FA. First order, buckling
[17] Phe PV, Huy NX. Moment resistances of wide flange beams with initial
and post buckling behaviour of GFRP pultruded beams. Part 2: Numerical
imperfection and residual stresses. J. Mater Eng Struct 2020;7:651–8.
simulation. Comp Str 2011;89:2065–78.
[18] Phe PV. Enhancement of moment resistance of steel beams with initial
[3] Xin H, Liu Y, Mosallam AS, He J, Du A. Evaluation on material behaviors of
imperfections and residual stresses by using stiffeners and GFRP plates. J Mater
pultruded glass fiber reinforced polymer (GFRP) laminates. Comp Str 2017;182:
Eng Struct 2020;7:659–67.
283–300.
[19] Phe PV, Mohareb M, Fam A. Lateral torsional buckling analysis of steel beams
[4] El Damatty A, Abushagur M, Youssef MA. Experimental and analytical
strengthened with GFRP plate. J. Thin-Walled Struct 2018;131:55–75.
investigation of steel beams rehabilitated using GFRP sheets. Steel Comp Str 2005;
[20] Lee J, Kim S-E, Hong K. Lateral buckling of I-section composite beams. Eng Str
3:421–38.
2002;24(7):955–64.
[5] Accord NB, Earls CJ. Use of fiber-reinforced polymer composite elements to
[21] Cortinez VH, Piovan MT. Vibration and buckling of composite thin-walled beams
enhance structural steel member ductility. J Comp Constr ASCE 2006;10(4):
with shear deformability. J Sound Vib 2002;258:701–23.
337–44.
[22] Machado SP, Cortinez VH. Lateral buckling of thin walled composite bisymmetric
[6] Youssef MA. Analytical prediction of the linear and nonlinear behaviour of steel
beams with prebuckling and shear deformation. Eng Struct 2005;27:1185–96.
beams rehabilitated using FRP sheets. Eng Struct 2006;28(6):903–11.
[23] Qiao P, Zou G, Davalos JF. Flexural torsional buckling of fiber reinforced plastic
[7] Siddique MAA, El Damatty AA. Improvement of local buckling behavior of steel
composite cantilever I beams. Compos Struct 2003;60:205–17.
beams through bonding GFRP plates. Compos Struct 2013;96:44–56.
[24] Girhammar UA, Pan DH. Exact static analysis of partially composite beams and
[8] Harries KA, Peck AJ, Abraham EJ. Enhancing stability of structural steel sections
beam-columns. Int J Mech Sci 2007;49(2):239–55.
using FRP. Thin-Walled Struct 2009;47. 2092-1101.
[25] Challamel N, Girhammar UA. Lateral-torsional buckling of vertically layered
[9] Jones RM. Mechanics of composite materials. 2nd ed. Philadelphia: Taylor &
composite beams with interlayer slip under uniform moment. Eng Struct 2012;34:
Francis; 1998.
505–13.
[10] Reddy JN. Mechanics of laminated composite plates and shells – theory and
[26] Schnabl S, Planinc I. Inelastic buckling of two-layer composite columns with
analysis. 2nd ed. Boca Raton, Florida: CRC Press; 2004. p. 33431.
nonlinear interface compliance. Int J Mech Sci 2011;53:1077–83.
[11] Phe PV, Mohareb M. A shear deformable theory for the analysis of steel beams
[27] Kabir MZ, Seif AE. Lateral-torsional buckling of retrofitted steel I-beams using FRP
reinforced with GFRP plates. J Thin-Walled Struct 2014;85:165–82.
sheets. Int J Sci Tech Trans A Civ Eng 2010;17:262–72.
[12] Phe PV, Mohareb M. Finite-element formulations for the spatial static response of
[28] Shojaee S, Valizadeh N, Izadpanah E, Bui T, Vu TV. Free vibration and buckling
steel beams bonded to a GFRP Plate. J. Eng. Mechanics ASCE 2015;141:04014143.
analysis of laminated composite plates using the NURBS-based isogeometric finite
[13] Phe PV, Mohareb M, Fam A. Analysis of steel beams strengthened with GFRP plates
element method. Compos Struct 2012;94:1677–93.
including pre-existing loading effects. J Struct Eng ASCE 2017;143:04017163.
[29] Phe PV, Stress-deformation theories for the analysis of steel beams reinforced with
[14] Pham PV, Mohareb M, Fam A. A finite element formulation for the analysis of
GFRP plates. Master’s Science Thesis, University of Ottawa, Ottawa, Canada; 2013.
multilayered beams based on the principle of stationary complementary strain
[30] Grognec PL, Nguyen QH, Hjiaj M. Exact buckling solution for two-layer
energy. Eng Str 2018;167:287–307.
Timoshenko beams with interlayer slip. Int J Solids Struct 2012;49:143–50.

22

You might also like