Vol1No1 3
Vol1No1 3
1 (2005) 47-66 47
Abstract: When an arch is subjected to in-plane loading, it may suddenly deflect laterally and twist out of the plane
of loading and fail in a flexural-torsional buckling mode. This paper presents a static equilibrium approach for the
elastic flexural-torsional buckling of circular arches under uniform bending, or under uniform compression. Solutions
for the buckling moment and buckling load are obtained in closed form, and discrepancies among existing solutions
are clarified. It is found that it is reasonable to use material curvatures and twist, rather than spatial curvatures and
twist, for the flexural-torsional buckling analysis of arches. First order buckling deformations provide a sufficient
basis for the static equilibrium methods for the flexural-torsional buckling analysis of arches. Equilibrium of a
lengthwise differential element of an arch should be considered in the analysis, and so the couplings between the
lateral bending and torsional actions and resistances can be included in the differential equations of equilibrium. If the
equilibrium is considered only at the cross-section, this is equivalent to treating arches as analogous with straight
members, and so the coupling terms in the differential equations of equilibrium are lost.
1. INTRODUCTION
When an arch is subjected to in-plane loading, it may suddenly deflect laterally and twist out of the
plane of loading and fail in a flexural-torsional buckling mode. In general, finite element methods
can be used to obtain the flexural-torsional buckling load under general conditions of in-plane
loading and restraints. When a circular arch is subjected to in-plane uniform bending or to uniform
compression (Figures 1(a) and 1(b)), closed form solutions for the elastic flexural-torsional
buckling (Figs 1(c)(d)) can be obtained, and the buckling loads for an arch under uniform bending
or uniform compression are important reference loads for the design of a steel arch [1]. The elastic
flexural-torsional buckling of arches under uniform bending or under uniform compression has been
studied extensively by a number of researchers [2-9] using energy or virtual work approaches.
Energy and virtual approaches are suitable for systematic treatments. However, because energy and
virtual approaches usually deal with strains and stress resultants, second order nonlinear
strain-displacement relationships are needed [10]. Different methods are used in deriving the
strain-displacement relationships [2-9] and so different second order terms for strain-displacement
relationships can be obtained. As a result, the closed form solutions for flexural-torsional buckling
loads derived by different researchers may differ from each other.
Static equilibrium approaches for flexural-torsional buckling analysis deal with equilibrium
between external loads and internal resistances directly, and thus the second order
strain-displacement relationships are not needed. In addition, since the buckled equilibrium
configuration is infinitesimally close to the unbuckled equilibrium configuration, the curvatures and
the buckling actions related to the first order buckling deformations are sufficient for the buckling
analysis. Static equilibrium approaches have been used only by Timoshenko and Gere [2] for arches
under uniform bending, and by Vlasov [11] for arches under uniform bending or under uniform
compression. Timoshenko and Gere [2] considered equilibrium at the cross-section to establish the
differential equations of equilibrium in the same way as for straight beams. However, in an arch, the
lateral bending action and resistance are coupled with the torsional action and resistance, so that
48 Y.L. Pi, M.A. Bradford and Y.Y. Chen
neither the lateral bending action and resistance nor the torsion action and resistance alone are in
equilibrium. Equilibrium of a lengthwise differential element of an arch, rather than only at the
cross-section, should therefore be considered, and so the method of Timoshenko and Gere [2]
appears to miss the coupling terms in the differential equations of equilibrium for the buckling
deformations. Vlasov [11] substituted the spatial curvatures and twist into the differential equations
of equilibrium that were derived for the flexural-torsional buckling analysis of straight members. In
general, however, this method is not suitable for the flexural-torsional buckling of arches, and it has
been shown [4-9] that the method may lead to an incorrect flexural-torsional buckling load for
arches under uniform compression.
The purpose of this paper is to use a static equilibrium approach to study the elastic
flexural-torsional buckling of circular arches with doubly symmetric cross-section under uniform
bending or under uniform compression, to obtain closed form solutions for the buckling moment
and load, and to clarify the discrepancies that arise among existing solutions.
The following assumptions are made in this study: (1) the Euler-Bernolli bending theory and
Vlasov's torsion theory are used, so that the cross-section is assumed to remain rigid and local
buckling and/or effects of distortion of the cross-section are not considered; (2) the arches are
circular and of doubly symmetric uniform cross-section, so that the centroid and the shear centre of
the cross-section coincide; (3) the height D of the cross-section is much smaller than both the length
S and the radius R of initial curvature of the arch, i.e. D/S<<1 and D/R<<1, which imply that the
arches considered are slender; and (4) the strains are small.
A body attached curvilinear orthogonal axis system oxys is defined as follows. The axis os passes
through the locus of the centroids of the cross-section of the arch and the axes ox and oy coincide
with the principal axes of the cross-section, as shown in Figures 1-3. In the undeformed
configuration, the axis oy is towards to the centre of the arch. After deformation, the origin o
displaces u, v, w in the directions of the axes ox, oy, and os to o1 and the cross-sections (that are
assumed to remain rigid in their plane and so do not distort) rotate through an angle φ , and so the
body attached curvilinear orthogonal axis system oxys moves and rotates to a new position o1x1y1s1
as shown in Figure 2. In general, the centroidal axis os of a circular arch has an initial curvature
κ x 0 about the major principal axis ox (i.e. in the direction of the axis oy) as shown in Figure 1).
q
M M
Q
Q=qR θ s θ
R
x
o1 u o Prebuckled position s1
x
1
s
φ Buckled position
y1 y
(c) Flexural-torsional buckling (d) Lateral restraints (plan)
A basis vector p s in the tangent direction of the axis os, and basis vectors p x and p y in the
directions of the axes ox and oy (Figures 2-3) are used as the fixed reference basis. The basis
vectors p x , p y , and p s do not change with the deformation, but their directions change from
point to point along the arch axis os. In the deformed configuration, a basis vector q s is defined in
the tangent direction of the axis o1s1, and basis vectors q x and q y are defined in the direction of
the principal axes o1x1, o1y1 of the rotated cross-section at o1 as shown in Figures 2-3. The basis
vectors q x , q y , q s are attached to the arch and move with the arch during the deformation with the
vector q s being normal to the cross-section at all times.
x
o
y
P
o M y
x ds
1 x1
Q ex o1 dQex
s1 Q ex
y1 s1
dM y
M y
s
(a) Laterial buckling of column (b) Alternative method
(a) Lateral buckling of column
(b) Alternative method
Figure 3. Lateral buckling of a column under uniform compression
The rotation from the basis vectors px, py, ps to the basis vectors qx, qy, qs (Figure 2) can be
expressed using a rotation matrix R as [12-13]
q i = Rp i (i = x, y, s ) (1)
Static equilibrium methods deal with the equilibrium of external loads and internal resistances
directly, and it is well known that the classical flexural-torsional buckling is characterized by the
fact that, as the load passes through its critical stage, the structure passes from its unbuckled
equilibrium configuration to an infinitesimally close buckled equilibrium configuration, Therefore,
50 Y.L. Pi, M.A. Bradford and Y.Y. Chen
second order terms are not needed in the flexural-torsional buckling analysis for arches by static
equilibrium approaches, and so a small-rotation matrix R is sufficient for undertaking the
buckling analysis, which is given by
⎡ 1 − φ u′⎤
R=⎢ φ⎢ 1 v~′⎥⎥ (2)
~
⎢⎣− u ′ − v ′ 1 ⎥⎦
where v~ ′ = v ′ − wκ x 0 , ( )′ = d( ) / ds , and the initial curvature κ x 0 of the centroidal axis os of a
circular is given by κ x 0 = −1 / R for the arches with the upward rise as shown in Figure 1.
It can be obtained from calculus of vectors [14] that the spatial curvatures referred to the
undeformed configuration and the material curvatures referred to the deformed configuration can be
obtained from the rotation matrix R as
dR T dR
Ω= R + RK 0 R T and K = R T + RT K 0R (3)
ds ds
where ( )T is the transpose of a matrix, K0 is the matrix of the initial curvatures and given by
⎡0 0 0 ⎤
K 0 = ⎢⎢0 0 − κ x 0 ⎥⎥ (4)
⎢⎣0 κ x 0 0 ⎦⎥
while Ω and K are the matrices of spatial and material curvatures and given by
⎡ 0 − ωs ω y ⎤ ⎡ 0 −κs κ y ⎤
⎢ ⎥ ⎢ ⎥
Ω = ⎢ ωs 0 − ω x ⎥ and K = ⎢ κ s 0 −κx ⎥ (5)
⎢− ω y ω x 0 ⎥⎦ ⎢− κ y κ x 0 ⎥⎦
⎣ ⎣
in which ω x and ω y are the spatial curvatures about the positive direction of the axes ox and oy,
and ωs is the spatial twist about the positive direction of the os axis in the undeformed
configuration; while κ x and κ y are the material curvatures about the positive direction of the
axes o1x1 and o1y1, and κ s is the material twist about the positive direction of the o1s1 axis in the
deformed configuration.
Substituting Eqs. (2) and (4) into Eq. (3) and ignoring second and higher order terms leads to the
spatial curvatures ω x and ω y , and the spatial twist ωs as
1
ω x ≈ −v~′′ + κ x 0 = −v~′′ − (7)
R
φ
ω y ≈ u ′′ + φκ x 0 = u′′ − (8)
R
u′
ωs ≈ φ ′ − u′κ x 0 = φ′ + (9)
R
1
κ x ≈ −v~′′ + κ x 0 = −v~′′ − (10)
R
An Equilibrium Approach for Flexural-Torsional Buckling of Elastic Steel Arches 51
φ
κ y ≈ u′′ − φκ x 0 = u′′ + (11)
R
u′
κ s ≈ φ ′ + u′κ x 0 = φ′ − (12)
R
where ( )″ ≡ d 2 ( ) / ds 2 .
In studies of the flexural-torsional buckling of an arch, the spatial curvatures and twist given by
Eqs. (7)-(9) have been used by a number of researchers [2-3,5,11] while the material curvatures
given by Eqs. (10)-(12) have been used by other researchers [4,6-9]. It is noteworthy that the
equilibrium for flexural-torsional buckling of an arch should be referred to the buckled
configuration. Hence, it is reasonable to use the material curvatures and twist rather than the spatial
curvatures and twist for the flexural-torsional buckling analysis of arches.
3. UNIFORM COMPRESSION
In order to compare the differences between the flexural-torsional buckling of arches and the
buckling of columns under uniform compression, the buckling of a column under uniform
compression is briefly summarized in the following. In the buckled configuration, the axial
compressive force P has the components: a lateral force Qex in the direction of the axis o1x1, a lateral
force Qey in the direction of the axis o1y1, and an axial force Qes in the direction of the axis o1s1and
they are given by
⎧Qex ⎫ ⎧ 0 ⎫ ⎧− Pu ′⎫
⎪ ⎪ T⎪ ⎪ ⎪ ⎪
⎨Qey ⎬ = R ⎨ 0 ⎬ = ⎨ − Pv′ ⎬ (13)
⎪Q ⎪ ⎪P ⎪ ⎪ P ⎪
⎩ es ⎭ ⎩ ⎭ ⎩ ⎭
where the rotation matrix R is the same as that given by Eq. (2) with v~′ = v′ .
The elastic lateral buckling load of a column about its minor principal axis oy can be determined by
considering lateral equilibrium in the buckled configuration as shown in Figure 3. Two methods can
be used to obtain the buckling load. Most researchers [2,10,15] have used the method shown in
Figure 3(a) and have obtained the differential equation of equilibrium for the lateral buckling as
Pu + EI y u ′′ = 0 (14)
where E is Young's modulus of elasticity and Iy is the second moment of area of the cross-section
about its minor principal axis oy.
Alternatively, the differential equation of equilibrium for lateral buckling can be obtained by
considering the lateral equilibrium of a differential element of length ds in the buckled
configuration shown in Figure 3(b). For lateral bending equilibrium,
dM y + Qex ds = 0 (15)
52 Y.L. Pi, M.A. Bradford and Y.Y. Chen
where the bending moment My about the axis o1y1 is equal to the internal bending resistance of the
cross-section Miy and so given by [15] M y = M iy = EI y u′′ ; and Qex is the lateral component of
the axial load P in the buckled configuration given by Eq. (13). Hence, Eq. (15) becomes
Pu ′ + EI y u′′′ = 0 (16)
The solution that satisfies the kinematic boundary conditions u0 = u S = 0 can be assumed to be
given by
u nπs
= sin (17)
um S
where um is the maximum lateral displacement during buckling, s is the coordinate in the column
axis os as shown in Figure 3, S is the length of the column and n is the number of the buckled half
waves. Substituting Eq. (17) into Eq. (14) or (16) leads to the n-th mode elastic flexural buckling
load of a column about its minor principal axis of the cross-section [2,15] given by
(nπ ) 2 EI y
Pyn = . (18)
S2
The axial compressive force Qes in the direction of the axis o1s1 may produce an external torque
about the axis o1s1 and so the column may buckle in a torsional mode. Consider the differential
element of length ds and differential strip dAds as shown in Figure 4, where dA is the
cross-sectional area of the differential strip. The relative twist angle between both ends can be
represented by dφ (Figure 4(b)). The force acting on the differential area dA due to the axial
compressive force Qes is equal to d qes = (Qes/A)dA, where A is the area of the cross-section. The
distance of the differential area dA from the shear centre of the cross-section is given by
ρ = x 2 + y 2 (Figure 4(a)) and the relative displacement of both ends of the differential strip due
to the relative twist angle dφ is equal to ρdφ . Hence, the force dqes has a component
dq = dqes ( ρdφ / ds ) (Figure 4(b)) which exerts a torque dmes about the shear centre given
by dmes = dq × ρ = dqes ρ 2 (dφ / ds ) , and so the total external torque Mes produced by the axial
compressive force Qes is
dφ Q dφ
M es = ∫ dmes = ∫ dqes ρ 2 = ∫ es ρ 2 dA = r02 Pφ ′ (19)
A A
ds A A ds
∫ ρ 2 dA ∫ (x + y 2 )dA
2
Ix + Iy
r0 = A
= A
= (20)
A A A
and in which Ix is the second moment of area of the cross-section about its major principal axis ox.
where G is the elastic shear modulus, and J and Iw re the torsion and warping constants of the
cross-section. The axial force Qes = P is given by Eq. (13). During torsional buckling, the resultant
of the external and internal torques vanishes and so
which is the differential equation of equilibrium for the torsional buckling of a column under
uniform compression [2,15].
The solution that satisfies the kinematic boundary conditions φ0 = φS = 0 can be assumed to be
given by
φ nπs
= sin (23)
φm S
where φm is the maximum twist angle of the cross-section during buckling. Substituting Eq. (23)
into Eq. (22) leads to the n-th mode elastic torsional buckling load of a column about its shear
centre axis [2,15]given by
1⎛ (nπ ) 2 EI w ⎞
Psn = 2 ⎜⎜ GJ + ⎟⎟ (24)
r0 ⎝ S2 ⎠
ds
ds
dA
dA
o dq
d
x s d dqes
o s
x
When a pin-ended arch with a in-plane radial load q acting at the centroid of the cross-section,
uniformly distributed around the arch axis, the arch is subjected primarily to a uniform compressive
force Q = qR as shown in Figure 1(a). In this study, it is assumed that the directions of the radial
loads do not change and remain parallel to their initial directions during deformation. Under this
action, the arch may buckle out of the plane of loading and fail in a flexural-torsional mode.
Similarly to the case of a column, in the buckled configuration of an arch, the axial compressive
force Q in the direction of the axis os has an axial compressive component Qes = Q in the direction
of the axis o1s1, and a lateral component Qex = −Qu′ in the direction of the axis o1x1. The force Qes
54 Y.L. Pi, M.A. Bradford and Y.Y. Chen
also produces an external torsional moment action Mes about the axis o1s1 of the arch and is given
by [7,9]
⎛ u′ ⎞
M es = r02Qesκ s = r02Q⎜ φ ′ − ⎟ (25)
⎝ R⎠
while the total internal torsional resistance Mis can be obtained by combining the uniform and
warping torsional resistances as [9,11]
⎛ u′ ⎞ ⎛ u′′′ ⎞
M is = −GJ ⎜ φ ′ − ⎟ + EI w ⎜ φ ′′′ − ⎟ . (26)
⎝ R⎠ ⎝ R⎠
In a column, the torsional action is in equilibrium with the torsional resistance as shown in the
previous section. However, in an arch, the torsional action couples with the lateral bending action
while the torsional resistance also couples with the lateral bending resistance. The resultant
torsional moment at the cross-section is then given by
⎛ u′ ⎞ ⎛ u′ ⎞ ⎛ u′′′ ⎞
M s = M es + M is = r02Q⎜ φ ′ − ⎟ − GJ ⎜ φ ′ − ⎟ + EI w ⎜ φ ′′′ − ⎟ (27)
⎝ R⎠ ⎝ R⎠ ⎝ R⎠
which does not vanish.
The resultant lateral bending moment at the cross-section is equal to the lateral bending resistance
Miy of the cross-section about the axis o1y1 and given by
⎛ φ⎞
M y = M iy = − EI yκ y = − EI y ⎜ u ′′ + ⎟ . (28)
⎝ R⎠
dθ M
( M y + dM y ) − M y + 2M s sin + Qex ds = dM y + s ds + Qex ds = 0 (29)
2 R
which leads to the differential equation of equilibrium for the lateral deformations as
dM y Ms 1
+ + Qex = 0 . (30)
ds s R
For the torsional moment equilibrium, it can be shown from Figure 5 that
dθ ds
( M s + dM s ) − M s − 2M y sin = dM s − M y =0 (31)
2 R
which leads to the differential equation of equilibrium for the torsional deformations
dM s M y
− =0. (32)
ds R
An Equilibrium Approach for Flexural-Torsional Buckling of Elastic Steel Arches 55
My Qex
Ms ds o1
x1 M s +dMs
Qex+dQex M y+d M y s1
y1
R
R
dθ
It can be seen that the alternative method for the lateral buckling analysis for columns discussed in
the previous section can be used for the flexural-torsional buckling analysis for arches. However, if
the first method for the lateral buckling analysis of columns is employed for the flexural-torsional
buckling of arches, the lateral bending action M ey = −Qu at the cross-section that is produced by
the axial compressive force Q will directly be used for the lateral bending equilibrium. Because of
the coupling between the lateral bending and torsional actions, the lateral bending action Mey has a
spurious component that will falsely contribute to the torsional action. Hence, to avoid the spurious
torsional action, the alternative method is preferred and the lateral component Qex of the axial
compressive force Q is used directly for the lateral bending equilibrium analysis of arches under
uniform compression.
Substituting Qex = −Qu′ and the expressions for Ms and My given by Eqs (27)-(28) into Eqs (30)
and (32) leads to the differential equations of equilibrium given by
′ ′
⎡ ⎛ φ ⎞⎤ ⎛ ′ u′ ⎞ 1 ⎡ ⎛ u ′′ ⎞ 1 ⎤ ⎡ 2⎛ u′ ⎞ 1 ⎤
⎢ y⎜
EI u ′′ + ⎟⎥ + GJ ⎜ φ − ⎟ − ⎢ EI w⎜ φ ′′ − ⎟ ⎥ + Q ⎢u ′ − r0 ⎜ φ ′ − ⎟ ⎥ = 0 (33)
⎣ ⎝ R ⎠⎦ ⎝ R⎠R ⎣ ⎝ R ⎠ R⎦ ⎣ ⎝ R ⎠ R⎦
′ ″ ′
⎛ φ⎞1 ⎡ ⎛ u′ ⎞⎤ ⎡ ⎛ u ′′ ⎞⎤ ⎡ 2 ⎛ u ′ ⎞⎤
EI y ⎜ u ′′ + ⎟ − ⎢GJ ⎜ φ ′ − ⎟⎥ + ⎢ EI w ⎜ φ ′′ − ⎟⎥ + ⎢r0 Q⎜ φ ′ − ⎟⎥ = 0 (34)
⎝ R⎠R ⎣ ⎝ R ⎠⎦ ⎣ ⎝ R ⎠⎦ ⎣ ⎝ R ⎠⎦
The n-th mode buckled shapes of a pin-ended arch (Figure 1(c)) can be assumed to be given by
u φ nπs
= = sin (35)
u m φm S
56 Y.L. Pi, M.A. Bradford and Y.Y. Chen
which satisfy the boundary conditions at both ends of the arch u0 = u S = φ0 = φS = 0 . Substituting
Eq. (35) into Eqs (33) and (34) leads to the homogeneous equations
⎡ ⎛ Pyn ⎞ Q ⎤
k11 = ⎢1 + an2bn2 − ⎜⎜1 + an2bn2 ⎟ ⎥ Pyn (37)
⎣⎢ ⎝ Psn ⎟⎠ Pyn ⎥⎦
⎛a Pyn Q ⎞
k12 = k 21 = −⎜ n + anbn − anbn ⎟ M ysn (38)
⎜b P P ⎟
⎝ n sn yn ⎠
⎛ a 2 Pyn Q ⎞ 2
k 22 = ⎜1 + n2 − ⎟r0 Psn (39)
⎜ b P P ⎟
⎝ n sn yn ⎠
S nπM ysn
an = and bn = (40)
nπR Psn S
and Mysn is the n-th mode elastic flexural-torsional buckling moment of the corresponding simply
supported beam of length S under uniform bending [2,15] given by
Eq. (36) has non-trivial solutions for um and φm when the determinant of its coefficient matrix
vanishes, i.e.
which leads to the equation for the elastic flexural-torsional buckling load of a pin-ended arch under
uniform compression as
2
⎛ Q ⎞ ⎡⎛ an2 ⎞ Psn Q⎤
⎜ ⎟ − ⎜1 + ⎟
⎜ P ⎟ ⎢⎢⎜ b 2 ⎟ P
(
+ 1 − an2 )
2
(
⎥ + 1 − an
Pyn ⎥⎦
2
)
2 Psn
Pyn
=0 (43)
⎝ yn ⎠ ⎣⎝ n ⎠ yn
which is the same as that of Pi and Bradford [9] using an energy method.
It can be seen from Eqs (33), (34) and (43) that if a pin-ended arch under uniform compression
buckles out-of-plane, it will buckle in a flexural-torsional mode. This is different from a column
under uniform compression which may buckle only in a flexural mode or in a torsional mode.
An Equilibrium Approach for Flexural-Torsional Buckling of Elastic Steel Arches 57
However, if the torsional deformations of a pin-ended arch under uniform compression are fully
prevented ( φ = φ ′ = φ ′′ = φ ′′′ = 0 ), the arch may buckle in a flexural mode and the differential
equation of equilibrium for flexural buckling of the arch can be obtained from Eq. (33) by
substituting φ = φ ′ = φ ′′ = φ ′′′ = 0 as
u′ u ′′′ ⎛ u′ ⎞
( EI y u′′)′ − GJ + EI w 2 + Q⎜ u′ + r02 2 ⎟ = 0 . (44)
R 2
R ⎝ R ⎠
which is equivalent to Eq. (14) or (16) for the flexural buckling of a column under uniform
compression.
By substituting the n-th mode buckling shapes given by Eq. (17) into Eqs (44), the flexural buckling
load of an arch under uniform compression can be obtained as
R2 ⎛ r02 ⎞
Q yn = 2 2 ⎜⎜ Pyn + 2 Psn ⎟⎟ (46)
R + r0 ⎝ R ⎠
which is higher than the n-th mode flexural buckling load Pyn of a column under uniform
compression [2,15] given by Eq. (18). When R → ∞ , Eq. (46) becomes the flexural buckling load
of a column of length S
Q yn = Psn . (47)
If the lateral deformations of a pin-ended arch under uniform compression are fully prevented
( u = u ′ = u′′ = u′′′ = 0 ), the arch may buckle in a torsional mode and the differential equation of
equilibrium for torsional buckling of the arch can be obtained from Eq. (34) by substituting
u = u ′ = u′′ = u′′′ = 0 as
φ
EI y − [GJφ ′]′ + [ EI wφ ′′]′′ + [r02Qφ ' ]′ = 0 . (48)
R 2
which is equivalent to Eq. (22) for the torsional buckling of a column under uniform compression.
By substituting the n-th buckling shapes given by Eq. (23) into Eq. (48), the torsional buckling load
of an arch under uniform compression can be obtained as
EI y an2
Qsn = + Psn (50)
r02
58 Y.L. Pi, M.A. Bradford and Y.Y. Chen
which is higher than the n-th mode torsional buckling load of a column under uniform compression
[2,15] given by Eq. (24). When R → ∞ , an → 0 and Eq. (50) becomes the torsional buckling
load of a column of length S
3.5 Comparisons
When the radius of gyration of the cross-section r0 is very small such as for a narrow rectangular
section, the effects of the external torsional moment Mes given by Eq. (25) produced by the axial
compressive force become very small compared with those of the external lateral force Qex given by
Eq. (13). In this case, the effects of the torsional moment Mes, i.e. the terms in the differential
equations of equilibrium (33) and (34) containing the factor r02 can be ignored, and so the
flexural-torsional buckling load reduces to
Q (1 − an2 ) 2
= (52)
Pyn 1 + an2 / bn2
which is the same as that obtained by Timoshenko and Gere [2] for arches with a narrow
rectangular section, and by Papangelis and Trahair [4], when energy methods were used.
A number of researchers [3-9] have used energy or virtual work methods to obtain closed form
solutions for the flexural-torsional buckling load of arches under uniform compression. Solutions
for the first mode flexural-torsional buckling load of arches under uniform compression given by
Eq. (43) are compared with the solutions by other researchers [3-9,11] in Figure 6 for arches with
an Australian steel cross-section 250UB37 [16] (A = 4750 mm2, Ix = 55.7×106 mm4, Iy = 5.66 ×
106 mm4, J = 158 × 103 mm4, Iw = 85.2 × 109 mm6, E = 200,000 MPa, υ = 0.3) and with a length S
= 2000 mm. It can be seen that there are some differences between the results. In particular, the
result of Yoo [3] has a substantial discrepancy with the other results [4-9]. Although generally the
energy or virtual work methods used by most researchers [4-9] were correct, the second order terms
in the nonlinear strains differed from each other. In addition, the treatments for the external
torsional moment were also somewhat different from each other. These may be the sources of small
discrepancies among the solutions of [4-9]. Yoo [3] substituted curvature expressions for an arch
directly into the strain energy and potential energy expressions that were derived for straight
members, and this is probably the source of the large differences between Yoo's result and that of
others. It is worth pointing out that Vlasov [11] in his study assumed that uniform compression in
arches was produced by hydrostatic loads, although his result has been benchmarked by a number
of researchers [3-9] as being the case for the radial loads discussed in this study. Vlasov [11]
substituted the curvature κ y and twist κ s into the differential equations of equilibrium that were
derived for straight members, and obtained the differential equations of equilibrium as
1 1
EI yκ ′y′ + GJκ s′ − EI wκ s′′′ + Qκ y = 0 (53)
R R
for the lateral deformations, and
1
EI yκ y − GJκ s′ + EI wκ s′′′+ r02Qκ s′ = 0 (54)
R
An Equilibrium Approach for Flexural-Torsional Buckling of Elastic Steel Arches 59
for the torsional deformations. It can be seen from Eqs. (53) and (54) that the coupling between the
internallateral bending resistance and the internal torsional resistance was considered. However, the
coupling term contributed by the external torsional moment Mes to the lateral bending deformations
was missed in Eq. (53). In addition, direct substitution of curvature κ y leads to a lateral force term
Q κ y in Eq. (53). By substituting κ y given by Eq. (11), Qκ y = Q(u ′′ + φ / R) , which contains a
lateral force Qφ / R . In the present study and in most other studies [2-9], it is assumed that the
directions of the radial uniformly distributed loads do not change but always remain parallel to the
initial directions during buckling and so the lateral force Qφ / R does not exist for this load case.
However, the directions of hydrostatic loads change during buckling. Hence, the lateral force
Qφ / R does exist and may be understood as the contribution of the horizontal component of
hydrostatic loads to the lateral resistance. It is therefore argued that the lateral force Qφ / R is the
major source of the large differences of the result of Vlasov [11] from those of the other researchers
[4-9].
Yoo 3
0.8
Vlasov11
0.6
0.4
0.2
0
0 20 40 60 80 100 120 140 160 180
Included angle Θ (degree)
Figure 6. Comparisons of results for flexural-torsional buckling under uniform compression
It can also be seen from Figure 6 that the flexural-torsional buckling load of a pin-ended arch under
uniform compression is lower than both the flexural buckling load and the torsional buckling load
of a corresponding column.
4. UNIFORM BENDING
When an arch simply supported in-plane and pin-ended laterally is subjected to in-plane equal and
opposite bending moments M at both ends as shown in Figure 1(b), the arch is under uniform
bending. Under this action, the arch may also buckle out of the plane of loading and fail in a
flexural-torsional mode. Because the ends of the arch are free to move together and apart in the
plane of loading for this special case, the arch can also be considered as a beam curved in elevation.
In the buckled configuration, the in-plane uniform bending moment M has the components: an
in-plane bending moment Mex about the axis o1x1, a lateral bending moment Mey about the axis o1y1,
and a torsional moment Mes about the axis o1s1 and given by
60 Y.L. Pi, M.A. Bradford and Y.Y. Chen
⎧M ex ⎫ ⎧M ⎫ ⎧ M ⎫
⎪ ⎪ T⎪ ⎪ ⎪ ⎪
⎨M ey ⎬ = R ⎨ 0 ⎬ = ⎨− Mφ ⎬ (55)
⎪M ⎪ ⎪ 0 ⎪ ⎪ Mu ′ ⎪
⎩ es ⎭ ⎩ ⎭ ⎩ ⎭
Timoshenko and Gere [2] studied the flexural-torsional buckling of a circular arch with a narrow
rectangular cross-section under uniform bending. The warping moment resistance of a narrow
rectangular cross-section is very small and so was ignored in their study. Timoshenko and Gere [2]
considered equilibrium in the buckled configuration as shown in Figure 7. If the warping moment
resistance is considered, the solution of Timoshenko and Gere [2] can be generalized in the
following. For lateral equilibrium, the sum of the external bending action Mey and the internal
bending resistance Miy vanishes, i.e.
M ey + M iy = 0 . (56)
Substituting Mey given by Eq. (55) and Miy given by Eq. (28) into Eq. (56) leads to the differential
equation of equilibrium for lateral deformations given by
⎛ φ⎞
Mφ + EI y ⎜ u ′′ + ⎟ = 0 . (57)
⎝ R⎠
For torsional equilibrium, the sum of the external torsional action Mes and the internal torsional
resistance Mis as vanishes, i.e.
M es + M is = 0 . (58)
Substituting Mes given by Eq. (55) and Mis given by Eq. (26) into Eq. (58) leads to the differential
equation of equilibrium for torsional deformations as
′
⎛ u′ ⎞ ⎡ ⎛ u ′′ ⎞⎤
Mu ′ − GJ ⎜ φ ′ − ⎟ + ⎢ EI w ⎜ φ ′′ − ⎟⎥ = 0 (59)
⎝ R⎠ ⎣ ⎝ R ⎠⎦
x o
o1
x1
M
M
s
y1 y
It can be seen that Timoshenko and Gere [2] treated the flexural-torsional buckling of an arch under
uniform bending in the same way as for the flexural-torsional buckling of a beam under uniform
bending. However, because of the curved shape, the lateral bending action necessarily couples with
the torsion action and the lateral bending resistance necessarily couples with the torsion resistance,
and the coupling terms were missed in the differential equations of equilibrium (57) and (59).
In the present study, the lateral and torsional equilibrium of a differential element of length ds in the
buckled configuration is considered as shown in Figure 8. The resultant lateral bending moment My
at the cross-section is equal to the sum of the lateral bending action Mey given by Eq. (55) and
internal bending resistance Miy given by Eq. (28), and given by
⎛ φ⎞
M y = M ey + M iy = − Mφ − EI y ⎜ u ′′ + ⎟ (60)
⎝ R⎠
The resultant torsional moment Ms at the cross-section is given by combining the external torsional
moment Mes given by Eq. (55) and internal torsional resistance Mis given by Eq. (26) as
⎛ u′ ⎞ ⎛ u ′′′ ⎞
M s = M es + M is = Mu ′ − GJ ⎜ φ ′ − ⎟ + EI w ⎜ φ ′′′ − ⎟ (61)
⎝ R⎠ ⎝ R⎠
For the lateral moment equilibrium, it can be shown from Figure 8 that
dθ ds
( M y + dM y ) − M y + 2 M s sin = dM y + M s =0 (62)
2 R
from which the differential equation of equilibrium for lateral deformations becomes
dM y Ms
+ =0. (63)
ds R
For the torsional moment equilibrium, it can be shown from Figure 8 that
dθ ds
( M s + dM s ) − M s − 2 M y sin = dM s − M y =0 (64)
2 R
from which the differential equation of equilibrium for torsional deformations becomes
dM s M y
− =0. (65)
ds R
62 Y.L. Pi, M.A. Bradford and Y.Y. Chen
My
Ms
ds
o1
M s +dMs
x1 M y +d M y s 1
y1
R
dθ
By substituting Eqs (60) and (61) into Eqs (63) and (65), the differential equations of equilibrium
(63) and (65) become
′ ′
⎡ ⎛ ′′ φ ⎞⎤ ⎛ u′ ⎞ 1 ⎡ ⎛ u ′′ ⎞⎤ 1 ′ Mu ′
⎢ EI y ⎜ u + R ⎟⎥ + GJ ⎜ φ ′ − R ⎟ R − ⎢ EI w ⎜ φ ′′ − R ⎟⎥ R + (Mφ ) − R = 0 (66)
⎣ ⎝ ⎠⎦ ⎝ ⎠ ⎣ ⎝ ⎠⎦
′ ″
⎛ ′′ φ ⎞ 1 ⎡ ⎛ ′ u′ ⎞⎤ ⎡ ⎛ u ′′ ⎞⎤ Mφ
EI y ⎜ u + ⎟ − ⎢GJ ⎜ φ − ⎟⎥ + ⎢ EI w ⎜ φ ′′ − ⎟⎥ + Mu ′′ + =0 (67)
⎝ R⎠ R ⎣ ⎝ R ⎠⎦ ⎣ ⎝ R ⎠⎦ R
for the torsional deformations. The n-th buckled shapes of the arch are also assumed to be given by
Eq. (35). Substituting Eq. (35) into Eqs (66) and (67) yields the homogeneous equations (36). In this
case, the coefficients k11 , k12 , k 21 , and k 22 are given by
⎛ M ⎞⎟
k11 = ⎜1 + an2bn2 + anbn P , (68)
⎜ M ⎟ yn
⎝ ysn ⎠
⎛a M ⎞⎟
k12 = k 21 = −⎜ n + anbn + M , (69)
⎜b M ⎟ ysn
⎝ n ysn ⎠
⎛ a 2
a M ⎞⎟ 2
k 22 = ⎜1 + n2 + n r P . (70)
⎜ b b M ⎟ 0 sn
⎝ n n ysn ⎠
In the same way as for arches under uniform compression, Eq. (36) has non-trivial solutions for um
and φm when the determinant of its coefficient matrix vanishes, i.e.
which leads to the equation for the elastic flexural-torsional buckling moment of an arch under
uniform bending as
2
⎛ M ⎞ ⎛ ⎞
⎜ ⎟ + ⎜ anbn + an ⎟ M − (1 − an2 ) = 0 (72)
⎜M ⎟ ⎜ bn ⎟⎠ M ysn
⎝ ysn ⎠ ⎝
which is consistent with those of Vlasov [1] and of Rajasekaran and Padmanabhan [6].
4.4 Comparisons
The solution for the first mode elastic flexural-torsional buckling moment of in-plane simply
supported and laterally pin-ended circular arches under uniform bending given by Eq. (71) is
compared with the solutions obtained by other researchers [3-9,11] in Figure 9 for arches with an
Australian steel cross-section 250UB37 [16] and a length S = 2000 mm. The solutions of Papangelis
and Trahair [4], Yang and Kuo [5], Trahair [7], Kang and Yoo [8] and Pi and Bradford [9] have
small discrepancies from each other. The solution of Yoo [3] is significantly different from the
other solutions[4-9]. Again, most of solutions [4-9] were correctly based on energy or virtual work
methods, but different second order nonlinear strains lead to small discrepancies amongst them.
Yoo [3] again used an analogy of arches with straight members, and so his solution resulted had
large discrepancies with other solutions 4-9].
3
This paper, Vlasov11
Rajasekaran and Padmanabhan6
2.5
Dimensionless buckling moment M/M ys
0.5
0
−150 −100 −50 0 50 100 150
Included angle Θ (Degrees)
Figure 9. Comparisons of results for flexural-torsional buckling under uniform bending
Vlasov [11] substituted the curvature κ y and twist κ s into differential equations of equilibrium
that were derived for the flexural-torsional buckling analysis of straight members, and considered
the coupling between the internal lateral bending resistance and the internal torsional resistance.
Because of the symmetric nature of the external lateral bending moment and resistance, and the
torsional moment and resistance under uniform bending, Vlasov's method for the flexural-torsional
buckling of arches under uniform bending is analogous to the present analysis. When the warping
resistance is ignored, the solution of Eq. (71) reduces to that of Timoshenko and Gere [2]. However,
as noted, since Timoshenko and Gere [2] considered equilibrium in the same way as for straight
beams only at the cross-section, some coupling terms were lost in the differential equations of
equilibrium.
64 Y.L. Pi, M.A. Bradford and Y.Y. Chen
It can also be seen from Figure 9 that the flexural-torsional buckling moment of an arch under
positive uniform bending as shown in Figure 1(b) is lower than that of a corresponding beam, while
the flexural-torsional buckling moment of an arch under negative uniform bending (the end
moments are opposite to those of Figure 1(b)) is higher than that of a corresponding beam.
5. CONCLUSIONS
This paper has used a rational static equilibrium approach to investigate the elastic
flexural-torsional buckling of circular arches under uniform compression or under uniform bending.
Solutions for the buckling moment and the buckling load have been obtained in closed form, and
discrepancies among the existing solutions have been clarified.
It has been found that it is reasonable to use material curvatures and twist, rather than spatial
curvatures and twist, for the flexural-torsional buckling analysis of arches. First order buckling
deformations are sufficient for static equilibrium methods in the flexural-torsional buckling analysis
of arches. When a rational static equilibrium approach is used for the flexural-torsional buckling
analysis, equilibrium of a differential element should be considered and so couplings between the
lateral bending action and resistance and the torsional action and resistance can be included in the
differential equations of equilibrium. If the equilibrium is considered only at the cross-section, the
coupling terms in the differential equations of equilibrium will be lost. A static equilibrium
approach of substituting the lateral curvature and twist into the differential equations of equilibrium
for the flexural-torsional buckling analysis of straight members is an analogy of arches with straight
members, which should not be used for the flexural-torsional buckling analysis for arches since it
may lead to incorrect results.
The flexural-torsional buckling load of a pin-ended arch under uniform compression is lower than
both the flexural buckling load and the torsional buckling load of a corresponding column. A
pin-ended arch under uniform compression may buckle in a flexural mode if torsional deformations
are fully prevented, or may buckle in a torsional mode if lateral deformations are fully prevented.
They are higher than their counterparts of a corresponding column. The flexural-torsional buckling
moment of an arch under positive uniform bending is lower than that of a corresponding beam,
while the flexural-torsional buckling moment of an arch under negative uniform bending is higher
than that of a corresponding beam.
ACKNOWLEDGEMENT
This work has been supported by a Federation Fellowship, and a Discovery Project awarded to the
second author, and a Discovery Project awarded to the first author, by the Australian Research
Council.
REFERENCES
[1] Pi YL, Trahair NS, Out-of-plane inelastic buckling and strength of steel arches. Journal of
Structural Engineering, ASCE 1998;124(2):174-183.
An Equilibrium Approach for Flexural-Torsional Buckling of Elastic Steel Arches 65
[2] Timoshenko SP, Gere JM. Theory of elastic stability, 2nd Edition New York, USA:
McGraw-Hill, 1961.
[3] Yoo CH. Flexural-torsional stability of curved beams. Journal of the Engineering Mechanics
Division, ASCE 1982;108(EM6):1351-1369.
[4] Papangelis JP, Trahair NS. Flexural-torsional buckling of arches. Journal of Structural
Engineering, ASCE 1987;113(4):889-906.
[5] Yang YB, Kuo SR. Effects of curvature on stability of curved beams. Journal of Structural
Engineering, ASCE 1987;113(6):821-841.
[7] Trahair NS. Flexural-torsional buckling of structures, London, UK: E & FN Spon, 1993.
[8] Kang YJ, Yoo CH. Thin-walled curved beams. II: Analytical solutions for buckling of arches.
Journal of Engineering Mechanics, ASCE 1994;120(10):2102-2125.
[9] Pi YL, Bradford MA. Elastic flexural-torsional buckling of continuously restrained arches.
International Journal of Solids and Structures 2002;128(6):719-727.
[10] Pi YL, Bradford MA. Effects of approximations in analysis of beams of open thin-walled
cross-section: 1. Flexural-torsional stability. International Journal for Numerical Methods in
Engineering 2001; 51(7):757-772.
[11] Vlasov VZ. Thin-walled elastic beams, 2nd Edition. Jerusalem, Israel: Israel Program for
Scientific Translation, 1961.
[12] Pi YL, Trahair NS. Nonlinear inelastic analysis of steel beam-columns. I: Theory. Journal
of Structural Engineering, ASCE 1994;120(7):2041-2061.
[13] Pi YL, Bradford MA, Uy B. Nonlinear analysis of members curved in space with warping
and Wagner effects. International Journal of Solids and Structures 2005; 42(11-12):
3147-3169.
[15] Trahair NS, Bradford MA, Nethercot DA. The behaviour and design of steel structures to
BS5950, Third Edition - British, London, UK: E & FN Spon, 2001.
[16] BHP. Hot rolled and structural steel products, 2000 Edition, Melbourne, Australia: BHP Co.
Pty Ltd, 2000.
[17] Pi YL, Papangelis JP, Trahair NS. Prebuckling deformations and flexural-torsional buckling
of arches. Journal of Structural Engineering, ASCE 1995;121(9):1313-1322.
66 Y.L. Pi, M.A. Bradford and Y.Y. Chen
[Blank Page]