Thin-Walled Structures 127 (2018) 373–388

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Coupled bending and torsional vibrations of non-uniform thin-walled beams T

by the transfer differential transform method and experiments
⁎ ⁎⁎
Jingwei Zhoua,b, Shurui Wenc, , Fengming Lia,c, , Yu Zhub
a
College of Mechanical Engineering, Beijing University of Technology, Beijing 100124, China
b
Xinjiang Goldwind Science & Technology Co., Ltd., Urumqi 830000, China
c
College of Aerospace and Civil Engineering, Harbin Engineering University, Harbin 150001, China

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

Keywords: This paper describes a new methodology capable of analyzing coupled bending and torsional vibrations in non-
Bending and torsional vibrations uniform thin-walled beams. The transfer matrix method (TMM) and the differential transform method (DTM) are
Non-uniform thin-walled beams combined to create the proposed transfer differential transform method (TDTM) for solving equations coupling
Transfer matrix method bending and torsional vibrations in non-uniform beams. Compared with the finite element method (FEM), the
Differential transform method
TDTM utilizes a type of changeable mode shape functions so that the number of meshed elements can be reduced
Modal test
greatly when the beam's geometric size is uniform. The equations of motion of the non-uniform thin-walled
beams are established using Hamilton's principle. The method considers both the warping and rotary effects of
the beam section. The natural frequencies and mode shapes of the bending and torsional components are ob-
tained using the TDTM. The accuracy of solutions can be controlled by the mesh density and the series expansion
order of the mode shape whose ranges are also discussed. For illustrative purposes, the natural frequencies and
the frequency response curves of a uniform beam and a non-uniform beam are studied respectively and are
validated by experiments. The effects of warping and the distance between the centroid of a given section and
the corresponding shear center on the vibration properties of the beam are also investigated.

1. Introduction
Non-uniform thin-walled beam can be widely found in blade
structures and many other engineering applications. If the blade has a
long span, such as in the case of a helicopter blade or a wind turbine
blade, it is reasonable to simplify the structure as a non-uniform beam.
Besides, in each blade section, the shear center may not be coincident
with its centroid, which results in a coupled bending and torsional ef-
fect. There have been several papers examining the dynamic features of
such a model. Most of them used the conventional methods while
dealing with the associated partial differential equations, such as the
finite element method (FEM), the dynamic stiffness matrix method and
the Ritz method. In this paper, the transfer differential transform
method (TDTM) is applied to solve the coupled partial differential
equations. The method greatly reduces the mesh number and improves
the computing efficiency by combining the advantages of the transfer
matrix method (TMM) and the differential transform method (DTM).
Problems arising from coupling of bending and torsion in open
section beams have been examined by several researchers. Mei [1] used
the FEM to study channel beam vibration problems and compared the
computing results with the experimental results of the first-order fre-
quency. Dokumaci [2] solved the equations of motion without con-
sidering the warping effects.
Banerjee [3] and Banerjee and Williams [4] extended Dokumaci's
theory, and developed the dynamic matrix methods for solving equa-
tions coupling bending and torsion, while taking the influences of shear
deformation and axial loads into account. However, the warping effects
were neglected. Lai et al. [5] studied the dynamic properties of a C-
shape composite channel beam. The material properties needed for the
formulation were obtained using the law of average and the warping
stiffness was considered. Ritz method was used to formulate the ac-
companying eigenvalue problem. Bishop et al. [6] demonstrated that
this approach would lead to considerable errors while solving the dy-
namic characteristics if the warping strain energy term is omitted.
Klausbruckner and Pryputniewicz [7] employed the FEM for studying
coupled bending and torsion. The warping stiffness was considered in
the equations of motion. The errors between the experimental and the
numerical results were analyzed.
The coupled bending and torsional vibrations in two directions of a
thin-walled beam were investigated by Tanaka and Bercin [8]. Li et al.

⁎
Corresponding author.
⁎⁎
Corresponding author at: College of Mechanical Engineering, Beijing University of Technology, Beijing 100124, China.
E-mail addresses: wenshurui@hrbeu.edu.cn (S. Wen), lifengming@hrbeu.edu.cn (F. Li).

https://doi.org/10.1016/j.tws.2018.02.026
Received 7 November 2017; Received in revised form 14 February 2018; Accepted 23 February 2018
Available online 02 March 2018
0263-8231/ © 2018 Elsevier Ltd. All rights reserved.
J. Zhou et al.
[9,10] reviewed the works of former researchers and gave two illus-
trative examples of thin-walled beams with a single symmetric axis. The
natural frequencies and the mode shapes were calculated by the dy-
namic TMM. The effects of axial force, warping stiffness and shear
deformation were also included. Ozgumus and Kaya [11,12] employed
the DTM for solving a beam with linearly changing cross section. Ha-
milton's principle was applied to derive the energy expressions. The
high efficiency resulting from the DTM while solving the coupled dif-
ferential equations was demonstrated. However, the section of the
beam was assumed to be solid and the warping stiffness was not con-
sidered.
Based on the Vlasov's theory, Borbon et al. [13] established the
equations of motion of a non-symmetric thin-walled beam subjected to
an axial load. Shear deformation and rotatory inertia were considered.
Egidio et al. [14] studied the nonlinear warping effects of a thin-walled
beam with open cross section. Galerkin's discretization was performed
by using a suitable expansion of displacements based on the mode
shapes. The dynamic responses of the beam subjected to a harmonic
force were investigated. Li et al. [15,16] used the Ritz method to solve
the forced vibration problems arising in beams and shells and in-
vestigated the frequency response curves.
Until recent years, Ozgumus and Kaya [17] conducted an extended
analysis of a rotating Timoshenko beam with the bending-bending-
torsion coupling effect by using the DTM. Bastawrous and EI-Badawy
[18] used the assumed modes method to discretize the equations of
motion and studied the coupled bending and torsional vibrations of the
wind turbine blades. Rao and Jin [19] focused on uncertainties asso-
ciated with the problem of coupled bending and torsion in beams. They
compared the results obtained from a truncation-based interval analysis
method, a universal grey number-based approach and an interval-dis-
cretization method, respectively. Liu and Shu [20] investigated the
coupled bending-torsion vibration of a single delamination beam sub-
jected to axial loads and static end moments. The buckling loads and
critical moments were also calculated. Similarly, the analytical results
were compared with those obtained from some commercial software
and with the experimental data by Kashani et al. [21]. Axial loads and
end moments were considered while performing the beam stability
analysis. Li et al. [22] analyzed the stochastic responses of an axially
loaded composite beam with bending-torsion coupling considering the
effects of shear deformation and rotary inertia. Cooley and Parker [23]
derived the equations of motion for a spinning cantilever beam to
analyze the bending in the direction of rotation axis coupled with the
torsional motion through the gyroscopic terms. The natural frequencies
and vibration modes were also investigated. Li et al. [24] studied the
coupled bending-torsional vibrations of a symmetric laminated com-
posite beam. The dynamic stiffness matrix was formulated from the
exact analytical solutions. Bekir and Omer [25] investigated the buck-
ling problem of linear tapered micro-columns using the strain gradient
elasticity theory. The results were compared with the modified couple
stress theories. Eken and Kaya [26] applied the thin-walled beam
theory for an arbitrary cross-section composite beam. The influences of
geometrical aspects were studied. Garcea et al. [27] reviewed the
Generalized Eigenvectors and the Generalized Beam Theory. They in-
vestigated the cross-section deformation modes of thin-walled members
with deformable cross-section. Both the calculating efficiency and the
accuracy of two approaches were compared in detail. Latalski et al. [28]
discussed the dynamic characteristics of a rotating composite thin-
walled beam attached with a rigid hub. The differential equations of
motion featuring bending-twist elastic coupling were derived from
Hamilton's principle. A shift of resonance zones and vibration absorp-
tion were observed. After that, the strong non-linear torsional angle was
taken into account by Bourihane et al. [29]. They used the Asymptotic
Numerical Method to solve the non-linear equations. Pavazza et al.
[30,31] combined the Vlasov's theory and the Timoshenko's beam
bending theory together to derive a novel theory of bending of thin-
walled beams with influence of shear. The close-form analytical results
374
Thin-Walled Structures 127 (2018) 373–388
were obtained for three-dimensional expressions of normal and shear
stresses. Furthermore, the results were compared with the Euler-Ber-
noulli's beam theory and the FEM.
The discretizing process and the methods for solving the associated
partial differential equations have also been developed. The DTM was
used to solve the non-linear equations by Wang et al. [32]. Shabnam
and Reza [33] extended the application of the DTM to formulate a two-
dimensional model that could deal with thin plates with variable
thickness. Rajasekaran and Tochaei [34] made some improvements by
developing a new method called the differential transformation element
method and achieved a satisfactory solution. Nuttawit and Arisara [35]
used the DTM to simultaneously solve linear and non-linear vibration
responses of elastically restrained beams. Nuttawit and Jarruwat [36]
studied the vibration characteristics of a stepped beam made of func-
tionally graded materials via the DTM. Later, a vibration analysis of a
thick nano-beam using the DTM was presented by Farzad and Parisa
[37]. Mustapha and Hawwa [38] obtained the approximate solutions
for a functionally graded micro-scale beam using the DTM. Certain non-
classical boundary conditions were included in the analysis. Ravi and
Hariharan [39] investigated the vibration characteristics of a fluid
conveying single walled carbon nanotube using the DTM.
Although a number of researchers have investigated vibration pro-
blems associated with thin-walled channel beams, only a few have
validated their computing results against experimental findings. Earlier
literatures focused mostly on conventional methods based on the partial
differential equations. By contrast, in this work, the analytical results
are compared with the experimental ones. The TDTM is developed to
solve the vibration problems of a non-uniformed thin-walled beam. The
equations of motion of the non-uniform thin-walled beams are estab-
lished using Hamilton's principle. The coupled bending and torsional
vibration equations describing uniform and non-uniform beams are
solved respectively. Two illustrative examples are carried out to vali-
date the results of the TDTM by comparing with the experimental data
as well as the commercial codes.
2. Equation of motion of the axial symmetric thin-walled beam
The displacements of a reference point on the cross section of a thin-
walled beam are shown in Fig. 1. The coupled bending and torsional
deformations of the beam can be divided into two independent steps,
i.e. the pure torsion with respect to the shear center and the pure
bending based on the former torsional deformation. For the first step,
the cross section rotates around the shear center in the cross-section
view as shown in Fig. 1(a). For the second step, the position of the
reference point will be corrected after considering the bending effects in
a side view as shown in Fig. 1(b). The points s and c represent the
positions of the shear center and the centroid. The symbol e refers to the
distance between them. The point P indicates the position after the
deflection in the beam cross section while the initial position is ex-
pressed as P0. ψ and w denote the torsional angle and the deflection of
the beam. They are functions of the geometrical parameters and the
time. Although the warping displacement is not shown in Fig. 1, the
potential energy of this part can be directly added into the total po-
tential energy according to the Vlasov's theory [13]. The warping ki-
netic energy is not taken into account in this work for it is trivial.
The coordinates of the reference point before and after the de-
formations can be written as (x0, y0, z0) and (x1, y1, z1). The initial
position of the reference point is
x 0 = x, (1a)
y0 = y, (1b)
z 0 = z. (1c)
The deformed position can be expressed as
x1 = x − (z cos(ψ) + y sin(ψ))sin(w′), (2a)
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

Fig. 1. Coupled bending and torsional deformations of a reference point.

y1 = y cos(ψ) − z sin(ψ), (2b) 1 L

Ubending =
2
∫0 EIy (w″)2dx ,
(8)
z1 = w + z cos(ψ) + y sin(ψ), (2c)
where Iy denotes the moment of inertia which is given in Appendix A.
where w′ = ∂w / ∂x . Based on the small deformation assumption, Some higher order terms are small and they are omitted in the bending
ψ2
sin(ψ) ≈ ψ and cos(ψ) ≈ 1 − 2
. So, Eqs. (2a)-(2c) can be written as the potential energy.
following form: Similarly, the shear potential energy follows the following for-
mulation
ψ2
x1 = x − ⎛z (1 −
⎜ ) + yψ ⎞ w′, ⎟
1
⎝ 2 ⎠ (3a) Ushearing = ∭ G (γxy2 + γxz 2)dx dydz,
2 (9)
ψ2
y1 = y (1 − ) − zψ, where G is the shear modulus. Substituting Eqs. (6b) and (6c) into Eq.
2 (3b)
(9), the shear potential energy can be written as
ψ2 1 L
z1 = w + z (1 − ) + yψ. Ushearing = ∫0 GJ (ψ′)2dx ,
2 (3c) 2 (10)
According to Eqs. (1a)-(1c) and Eqs. (3a)-(3c), the displacements where J represents the torsional moment of inertia. By shifting the
can be written as principle axes to the shear center and calculating the principle moments
ψ2 of inertia Ix and Iy, then the torsional moment of inertia follows J = Ix
u1 = −⎛z (1 −
⎜ ) + yψ ⎞ w′, ⎟
+ Iy.
⎝ 2 ⎠ (4a)
For a thin-walled beam, the warping potential energy should not be
yψ2 ignored [6], and it is given by
u2 = − − zψ,
2 (4b) 1 L
Uwarping =
2
∫0 E1 Γ (ψ″)2dx ,
(11)
ψ2
u3 = w + z (1 − ) + yψ − z ,
2 (4c) where Γ is the warping constant, and E1 is the warping elastic modulus.
where u1, u2 and u3 are the displacement components in the x, y and z The relation between E1 and E can be expressed as E1 = E /(1 − μ2 ) , in
directions. According to the Green's principle [12] which μ is the Poisson's ratio.
To sum up all the potential energy terms, the total potential energy
1 1 1 U can be written as
E11 = dx u1 + (dx u1)2 + (dx u2)2 + (dx u3)2 ,
2 2 2 (5a)
U = Ubending + Ushearing + Uwarping. (12)
2E12 = dy u1 + dx u2 + dy u1 dx u1 + dy u2 dx u2 + dy u3 dx u3, (5b)
Expanding Eq. (12) results in
2E13 = d z u1 + dx u3 + d z u1 dx u1 + d z u2 dx u2 + d z u3 dx u3, (5c)
1 L

where E11, 2E12 and 2E13 are the three dominant axial strains [12].
U=
2
∫0 (EIy (w″)2 + GJ (ψ′)2 + E1 Γ (ψ″)2)dx .
(13)
Substituting Eqs. (4a)-(4c) into Eqs. (5a)-(5c) leads to
The kinetic energy is derived from the velocity vectors. Taking the
1 1 1 derivative of Eqs. (3a)-(3c) with respect to the time results in
εxx = (y 2 + z 2)(ψ′)2 + z 2 (w″)2 + z (ψ)2w″ − yψ (x ) w″ − zw″
2 2 2
1
1 2 x1̇ = w′zψψ̇ − w′yψ̇ − ẇ ′z + ẇ ′zψ2 − ẇ ′yψ,
+ (w′) , 2 (14a)
2 (6a)

γxy = −zψ′, (6b) x2̇ = −yψψ̇ − zψ,̇ (14b)

γxz = yψ′, (6c) x3̇ = ẇ − zψψ̇ + yψ,̇ (14c)

where εxx represents the axial normal strain, and γxy and γxz are the where f ̇ = ∂f / ∂t . Due to the definition of the kinetic energy T, it can be
shear strains. The bending potential energy is written as written as
1 1
Ubending =
2
∭ Eεxx 2dx dydz, (7)
T=
2
∭ ρ (x1̇ 2 + x2̇ 2 + x3̇ 2)dx dydz, (15)

where E is the Young's modulus. Taking integration along the whole where ρ is the material density. Taking integration on the whole vo-
beam yields lume of the beam element yields

375
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

1 L 2 d4 d2
T=
2
∫0 (ρIy ẇ ′2 + Ia ψ̇ + ρAẇ 2 + 2ρAewψ ̇ x,
̇ )d
(16) EIy 4
W (x ) + ρIy ⎛ 2 W (x ) ⎞ p2 − ρAW (x ) p2 − ρAeΨ (x ) p2 = 0,
⎜ ⎟

dx ⎝ dx ⎠
in which Ia is known as the rotary inertia, and A is the cross-section area (21a)
of the beam. For the shear center being selected as the origin of the
d4 d2
coordinate system, the expressions of Ia, A and Ae are listed in Appendix E1 Γ Ψ (x ) − GJ Ψ (x ) − Ia Ψ (x ) p2 − ρAeW (x ) p2 = 0.
dx 4 dx 2 (21b)
A.
The equation of motion and the associated boundary conditions can
be obtained using Hamilton's principle [15,16] 3.2. Dimensionless transformation
t2 t2
∫t1
δ(U − T )dt + ∫t1
δDdt = 0,
(17) In order to obtain simpler forms of the equations, the following
dimensionless parameters are introduced:
where D represents the work done by the external forces applied on the
x W Ψ
structure. For a free vibration problem, this term is zero. x = , w = , ψ = ,
L L L (22a)
Substituting Eqs. (13) and (16) into Eq. (17) and performing the
variation operation in terms w and ψ, the equations of motion of the EIy
beam can be obtained as follows: p2 = ,
L4 ρAwp2 (22b)
∂4 ∂4 ∂2 ∂2 where x , w and ψ denote the dimensionless coordinate and displace-
EIy w (x , t ) − ρIy w (x , t ) + ρA w (x , t ) + ρAe ψ (x , t ) = 0,
∂x 4 ∂t 2∂x 2 ∂t 2 ∂t 2 ments, p is the dimensionless circular frequency, and L represents the
(18a) length of the beam.
Then, Eqs. (21a) and (21b) can be written as the following di-
∂4 ∂4 ∂2 ∂2
E1 Γ ψ (x , t ) − GJ 2 ψ (x , t ) + Ia 2 ψ (x , t ) + ρAe 2 w (x , t ) = 0. mensionless form:
∂x 4 ∂x ∂t ∂t
(18b) d4 Iy d2 1 e
4
w (x ) + 2 2 ⎛ 2 w (x ) ⎞ − 2 w (x ) − 2 ψ (x ) = 0,
⎜ ⎟

dx L Ap ⎝ dx ⎠ p p (23a)
The parameter e performs as a switch of the bending and torsional
coupling effect. If e keeps at zero, which implies that the shear center is d4 L2GJ d2 EIa Iy eIy
coincident with the centroid, Eqs. (18a) and (18b) will be naturally ψ (x ) − ψ (x ) − ψ (x ) − w (x ) = 0.
dx 4 E1 Γ dx 2 E1 ΓρAp 2 Γp 2 (23b)
decoupled so that they can be solved separately. If the rotary effect of a
beam element is not included, Eq. (18a) will be changed into the form
in [6]. If the warping effect is ignored, Eq. (18b) will become into the 3.3. DTM's solution
form in [4].
And the geometric and force boundary conditions at the fixed end The differential transformation method (DTM) is based on the
(x = 0) and the free end (x = l) of the beam can be obtained as Taylor series expansion, and the differential equations are converted
into a set of algebraic equations, in which the corresponding coeffi-
d
w (0, t ) = 0, w (0, t ) = 0, cients are periodic. According to the DTM, the equations of motion and
dx (19a)
the boundary conditions can be transformed into a series of finite order
d expressions with periodic coefficients [11,12]. For the single degree of
ψ (0, t ) = 0, ψ (0, t ) = 0, freedom system, the polynomial curves approach the analytical mode
dx (19b)
shapes when taking higher order of the expansion. The natural fre-
d2 d3 quencies can be obtained by solving the corresponding characteristic
EIy w (l, t ) = 0, EIy 3 w (l, t ) = 0,
dx 2 dx (19c) equation.
For a continuously differentiable function f(x), its Taylor series ex-
d3 d d2 pansion is given as
E1 Γ ψ (l, t ) − GJ ψ (l, t ) = 0, E1 Γ 2 ψ (l, t ) = 0,
dx 3 dx dx (19d) ∞
(x − x 0) k dk
where
d
t ) denotes the warping ratio of cross section of the beam.
ψ (x ,
f (x ) = ∑ k! dx k
f (x )x = x 0 .
dx k=0 (24)
The two equations in Eq. (19c) represent the bending moment and
1 dk
shear force. In Eq. (19d), according to the Vlasov's theory, the total Defining an operator F [k ] = f (x )x = x 0 , Eq. (24) can be written
k ! dx k
torque at the free end of the beam equals to the difference between the as
restrained torque and the St. Venant torque. Additionally, the bi-mo- ∞
ment remains at zero when the warping happens. f (x ) = ∑ (x − x 0)kF [k ].
k=0 (25)
3. Free vibration analysis Consider a finite-order series expansion, Eq. (25) can be written as
m
3.1. Harmonic motion f (x ) = ∑ (x − x 0)kF [k ],
k=0 (26)
For the harmonic motion, the transverse displacement and the tor-
where m is the order of the series expansion. The DTM transformation
sional angle can be expressed as
regulations are listed in Table B.1 of Appendix B. Here, once the ex-
w (x , t ) = W (x )eipt , (20a) pansion factors F[k] are solved, Eq. (26) can be written in a form of
polynomial function to approach the analytical mode shapes. The ac-
ψ (x , t ) = Ψ(x )eipt , (20b) curacy of this approach is subject to the order of expansion.
The differential transformation operators in the bending and tor-
where W(x) and Ψ(x) are the vibration modal shapes, and p is the 1 dk
sional directions are defined as P and Q, where P [k ] = k ! dx k
w (x )x = x 0
circular frequency.
1 dk
Substituting Eqs. (20a) and (20b) into Eqs. (18a) and (18b), one can and Q [k ] = (x )x = x 0 . According to Table B.1 of Appendix B, the
k ! dx k
ψ
get expansion form for these two operators can be expressed as

376
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

1
P (0) = c1 = 0, P (1) = c2 = 0, P (2) = c3, P (3) = c4, P (4) = l1 c3,
12
(30a)

1
Q (0) = c5 = 0, Q (1) = c6 = 0, Q (2) = c7, Q (3) = c8, Q (4) = r2 c7.
12
(30b)

By means of the geometric boundary conditions of the cantilevered

Fig. 2. Cross section of a non-uniform thin-walled beam. beam, one obtains c1 = c2 = c3 = c4 = 0. The boundary conditions in
Eqs. (29a)-(29d) can be rewritten in form of the matrix equation as-
l1 (k + 2)(k + 1) P (k + 2) + l2 P (k ) + l3 Q (k )
sociated with the force boundary conditions as
P (k + 4) = , k = 1, 2, ... m − 4,
(k + 4)(k + 3)(k + 2)(k + 1) a a34 a37 a38 c3
⎛ 33 ⎞ ⎛ ⎞ ⎛0⎞
(27a) a43 a44 a47 a48 c4 0
⎜a a74 a77 a78 ⎟ ⎜ c7 ⎟ = ⎜ 0 ⎟,
⎜ 73 ⎟⎜ ⎟ ⎜ ⎟
r1 P (k ) + r2 (k + 2)(k + 1) Q (k + 2) + r3 Q (k ) ⎝ a83 a84 a87 a88 ⎠ ⎝ c8 ⎠ ⎝ 0 ⎠ (31)
Q (k + 4) = , k = 1, 2, ...
(k + 4)(k + 3)(k + 2)(k + 1)
(27b) where the elements aij (i, j = 3, 4, 7, 8) are determined by the corre-
m − 4.
sponding polynomial expressions in Eqs. (29c) and (29d), and they are
The above two equations are characterized by periodicity. Here, l1, functions of the series expansion order m. For a nontrivial solution of
l2, l3, r1, r2 and r3 are given as the matrix equation, the determinant of the coefficient matrix must be
zero
Iy eIy
l1 = − , r1 = ,
L2p 2 A Γp 2 (28a) a33 a34 a37 a38
a43 a44 a47 a48
1 L2GJ a73 a74 a77 a78 = 0.
l2 = 2
, r2 = , a83 a84 a87 a88 (32)
p EΓ (28b)

EIa Iy The natural frequencies can be solved from Eq. (32), and the mode
e
l3 = , r3 = . shapes can also be represented by the polynomial functions when
p2 E1 Γp 2 ρA (28c)
substituting the eigenvalues into Eq. (24).
The differential transformation boundary conditions are given as

P (0) = 0, P (1) = 0, (29a) 3.4. TDTM in dealing with a non-uniform beam

Q (0) = 0, Q (1) = 0,
m m
∑ k (k − 1) P (k ) = 0, ∑ k (k − 1)(k − 2) P (k ) = 0,
k=0 k=0
m m
∑ [E1 Γk (k − 1)(k − 2) − GJk ] Q (k ) = 0, ∑ E1 Γk (k − 1) Q (k ) = 0.
k=0 k=0
The DTM transformations for the boundary conditions in Eq. (29)
are listed in Table B.2 of Appendix B.
Assuming that c1, c2, c3, c4, c5, c6, c7 and c8 denote the eight un-
known parameters P(0), P(1), P(2), P(3), Q(0), Q(1), Q(2) and Q(3), the
terms in the first loop of the expressions are described as
(29b) Combining the TMM and the DTM, the transfer differential trans-
formation method (TDTM) is developed to solve the beam with variable
section. In order to obtain the equations of motion, the ideas of discrete
(29c) structures are applied. Assuming each beam element has a constant
structural property, and material parameters such as the mass, the in-
ertia, the torsional stiffness, and the warping stiffness keep constant in
an element. Based on the continuity conditions of each pair of adjacent
beam segments, the relations between the first and the last elements can
(29d)
be established. The natural frequencies and mode shapes can be ob-
tained through the eigenvalues. The process is just as that of the uni-
form beam described in the above section.
As shown in Fig. 2, the geometric properties of the cross section are
the linear interpolation of those of the two neighbor sections.
The discrete equations of Eqs. (23a) and (23b) can be written as
2
45

4
2

Unit: mm 25

(a) (b)
Fig. 3. (a) Three-dimensional (3D) plot of the uniform channel beam, (b) cross section plot of the uniform channel beam.

377
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

Fig. 4. (a) Three-dimensional (3D) plot of the non-uniform channel beam, (b) cross section plot of the non-uniform channel beam.

Uniform channel beam

Acceleration
sensors
Data
Acquisition
system

(a) (b)
Fig. 5. (a) Locations of the sensors installation on the beam, (b) the modal test process of the uniform beam.

d4 d2 a a12 a13 a14 a15 a16 a17 a18 c

(EIy )i wi (x ) + (ρIy )i ⎛ 2 wi (x ) ⎞ p2 − (ρA)i wi (x ) p2 − (ρA)i ei ψi (x ) p2 = 0, ⎛ 11 ⎞ ⎛ 1⎞
dx 4
⎜ ⎟
a21 a22 a23 a24 a25 a26 a27 a28 c2
d
⎝ x ⎠ ⎜a ⎟ ⎜
31 a32 a33 a34 a35 a36 a37 a38 c3 ⎟
(33a) ⎜a a42 a43 a44 a45 a46 a47 a48 ⎟ ⎜c ⎟
41 4
[Α]i = ⎜ a a52 a53 a54 a55 a56 a57 a58 ⎟, [V]i = ⎜ c5 ⎟.
51
⎜ a61 a62 a63 a64 a65 a66 a67 a68 ⎟ ⎜ c6 ⎟
d4 d2 ⎜ a71
(E1 Γ )i ψ (x ) − (GJ )i 2 ψi (x ) − (Ia)i ψi (x ) p2 − (ρA)i ei wi (x ) p2 = 0, a72 a73 a74 a75 a76 a77 a78 ⎟ ⎜ c7 ⎟
dx 4 i dx ⎜ ⎟ ⎜ ⎟
⎝ a81 a82 a83 a84 a85 a86 a87 a88 ⎠ ⎝ c8 ⎠ (35)
(33b)
Similar to the case of the uniform beam, c1, c2, c3 and c4 are the
where i is the number of the element.
unknowns in the bending direction, and c5, c6, c7 and c8 are in the
Considering each element is attached to its adjacent element and
torsional direction. These eight parameters and the corresponding
they share an exact continuous condition, the relationship between two
matrix indicate the structural properties of a single beam segment. With
elements is determined by their own parameter matrices. The re-
the four geometric continuous relations (i.e. the transverse displace-
lationship between elements i and i + 1 can be obtained by the TMM as
ment, the rotational angle, the torsional displacement and the warping
[Α]i [V]i = [Α]i + 1 [V]i + 1 , (34) ratio) and the four generalized forces (i.e. the bending moment, the
shearing force, the total torque and the bi-moment), the dimension of
where the coefficient matrix [A]i and the vector [V]i of the ith element the matrix [A]i is 8 × 8.
are given by Eq. (34) can be transformed as

378
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

0.98m
Acceleration Non-uniform
sensors channel beam
0.75m

Force
hammer Platform

(a) (b)
Fig. 6. (a) Locations of the sensors installation on the beam, (b) the modal test process of the non-uniform beam.

Table 1 [V]i + 1 = [Tr]i [V]i . (37)

Comparison of the natural frequencies of a uniform channel thin-walled beam obtained
by the present TDTM, the FEM and the experiments. Based on the TMM, the vector [V]N of the last element can be ex-
pressed by [V]1 of the first element as
Frequency Natural frequencies(Hz)
order [V]N = [Tr]N − 1 [Tr]N − 2 ... [Tr]1 [V]1 , (38)
Exp. ANSYS ANSYS Present Present Present
Shell181 Solid186 e=0 warping warping
where N is the total element number. Letting [Z] = [Tr]N − 1 [Tr]N − 2 ... [Tr]1,
ignored included Eq. (38) is rewritten as

1 37.247 36.875 37.711 45.520 32.889 37.949

[V]N = [Z][V]1 , (39)
2 75.648 77.134 77.362 49.273 67.334 75.688
where [Z] is the product of the N− 1 transfer matrices. It shows the
3 150.828 149.37 154.103 160.311 112.865 157.052
4 330.045 327.31 330.070 308.104 182.102 334.345 relation between the first and the last parameter vectors. Through this
5 414.031 417.50 410.780 342.566 257.903 412.671 approach, the eight parameters c1, c2, c3, c4, c5, c6, c7 and c8 can be
6 591.171 586.77 587.202 609.079 330.963 609.788 mapped into the last element. Thereby, the total number of the para-
meters is reduced to eight.
Multiplying [A]N on both sides of Eq. (39), one can obtain
[A]N [V]N = [A]N [Z][V]1 . (40)
1st mode
2nd mode On the right side of Eq. (40), similar to the situation appeared in
3rd mode Eqs. (30a) and (30b), four parameters, c1(1), c2(1), c5(1) and c6(1) , in vector
Relative error (%)

[V]1 are zero. The force boundary conditions of the last element are
expressed as the following form:

⎛ b33 b34 b37 b38 ⎞ c3

⎛ ⎞ ⎛0⎞
⎜ b43 b44 b47 b48 ⎟ c4 0
⎜ c ⎟ = ⎜ ⎟,
⎜ b73 b74 b77 b78 ⎟ ⎜ 7 ⎟ ⎜ 0⎟
⎜ ⎟
⎝ b83 b84 b87 b88 ⎠ N ⎝ c8 ⎠1 ⎝ 0 ⎠ (41)
where bij (i, j = 3, 4, 7, 8) are the elements in matrix [A]N [Z] which is
originally a 8 × 8 matrix. Because [V]1 = (0, 0, c3(1), c4(1), 0, 0, c7(1), c8(1) )T ,
Eq. (41) is derived by reducing the dimension of the previous matrix. To
The order of expansion of the TDTM obtain a non-trivial solution, the determinant of the parameter matrix
Fig. 7. The relative error with respect to the expansion orders of the TDTM. should be zero, i.e.

b33 b34 b37 b38

[V]i + 1 = [Α]−i +11 [A]i [V]i . (36) b43 b44 b47 b48
= 0.
b73 b74 b77 b78
Defining [Tr]i = [Α]−i +11 [A]i ,
which is known as the transfer matrix b83 b84 b87 b88 (42)
and it indicates the parametric relations between the beam segments i
and i + 1. Eq. (36) is written as The natural frequencies and the corresponding mode shapes will be

379
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

Fig. 8. The first six order bending mode shapes of the uniform beam calculated by the TDTM.

Fig. 9. The first six order torsional mode shapes of the uniform beam calculated by the TDTM.

obtained. This process is similar to that of the DTM for a uniform beam. walled beam under a harmonic excitation force, the displacements in
The Newton-Raphson interaction method is adopted in finding roots. Eqs. (20a) and (20b) should be rewritten in the following forms:
The TDTM gives an effective way in dealing with the non-uniform beam ∞
structures. The idea of discretization from FEM is applied in the TDTM. w (x , t ) = ∑ Wn (x ) rn (t ) = W T (x ) γ (t ),
n=1 (43a)
4. Forced vibration analysis
∞

In this section, the ordinary differential equations of motion of the

ψ (x , t ) = ∑ Ψn (x ) sn (t ) = Ψ T (x ) ξ (t ),
n=1 (43b)
beam structure will be established and the frequency response curves
will be calculated by the mode expansion method [18,19]. In order to where W(x) = [W1(x), W2(x),…, Wn(x)] and Ψ(x) = [Ψ1(x), Ψ2(x),…,
T

investigate the responses of a bending and torsional coupled thin- Ψn(x)]T are the mode shapes obtained by Eq. (26), n denotes the order

380
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

are also given in Appendix C.

Present method The virtual work is expressed as
Experiment l
δD = ∫0 (qt (x )δw + qm (x )δψ)dx , (46)
where qt (x) and qm (x) are the transverse force distribution and the
Responses (g/N)

torque distribution. For a certain cross section of the beam at x = x0, if

the transverse force is not applied at the shear center, it will also
generate a related torque. The relation between the two loads can be
described as qm(x0) = qt(x0)R, in which R is the arm of force with re-
spect to the shear center of the beam section.
Substituting Eqs. (43a) and (43b) into Eq. (46), the virtual work is
expressed as
δD = Ft δγ + Fm δξ, (47)
where Ft and Fm are the generalized forces which are given in Appendix
C.
Substituting Eqs. (44), (45) and (47) into Eq. (17), the equations of
Frequency (Hz) motion can be expressed as

Fig. 10. Bending FRFs of the uniform thin-walled beam. ⎛⎜ M1 + M3 M 4 ⎞⎟ ⎛ γ̈ ⎞ + ⎜⎛ K1 0 ⎞ ⎛ γ ⎞ ⎛ Ft ⎟⎞.

⎟⎜ ⎟ = ⎜
⎜ ⎟
M T
M ξ ̈ 0 K2 + K3 ξ F
⎝ 4 2 ⎠⎝ ⎠ ⎝ ⎠⎝ ⎠ ⎝ m⎠ (48)
Table 2
Letting
Comparison of the natural frequencies of a non-uniform channel thin-walled beam ob-
tained by the present TDTM, the FEM and the experiments.
M1 + M3 M 4 ⎞ ⎛ K1 0
M = ⎜⎛ T ⎟, K = ⎜
⎞⎟,
Frequency Natural frequencies(Hz) M 4 M2 0 K2 + K3
⎝ ⎠ ⎝ ⎠
order
Exp. ANSYS ANSYS Present Present X = (γT , ξ T) T , Ω = (Ft T , Fm T) T , (49)
Shell181 Solid186 N = 50 N = 200
then Eq. (48) can be integrated as
1 63.101 63.674 63.110 62.548 62.583
2 130.032 124.850 128.146 132.665 132.629 MẌ + KX = Ω. (50)
3 253.967 241.100 250.711 254.112 254.126
4 454.142 447.740 430.063 450.356 450.650 The total distributed loads are assumed to be harmonic and written
5 538.232 552.410 551.726 577.735 577.911 as
6 797.682 763.970 764.848 808.508 808.993
qt (x , t ) = qt (x )sin(ωt ), (51a)

qm (x , t ) = qm (x )sin(ωt ). (51b)
The steady state solution of Eq. (50) can be written as
Xi (t ) = Ci sin(ωt ), i = 1, 2, ... ,n, (52)
Relative error (%)

1st mode where Ci is the participating factor to be determined, and n is the order
2nd mode of modes to be selected.
3rd mode Substituting Eq. (52) into Eq. (50) and eliminating sin(ωt) at both
4th mode sides of the equation yields
(K − ω2 M) C = Ω, (53)
T
where C = [C1, C2,…, Cn] means the vector of the participating fac-
tors. By solving Eq. (53) and by means of Eq. (43a) and Eq. (43b), the
steady state responses of a bending and torsional coupled thin-walled
beam are finally obtained.
The element number
Fig. 11. The relative error with respect to the total element number. 5. Experimental and numerical results

In order to validate the numerical results, two sets of modal tests

of the modes, and γ(t) = [γ1(t), γ2(t),…, γ(t)]T and ξ(t) = [ξ1(t), ξ2(t),…,
have been carried out. In the experiments, the two cantilevered beam
ξn(t)]T are the modal coordinates.
models are: a uniform channel thin-walled beam and a non-uniform
Substituting Eqs. (43a) and (43b) into Eqs. (13) and (16), the po-
channel thin-walled beam.
tential and the kinetic energies can be written in terms of the modal
coordinates:
5.1. Modal tests of a uniform and a non-uniform beams
1 1 1
U = γTK1γ + ξ TK2ξ + ξ TK3ξ,
2 2 2 (44) The geometries of the three-dimensional (3D) view and cross section
of the two beam models are shown in Fig. 3 and Fig. 4. The total length
1 T 1 T 1 1 1 T
T= γ˙ M1γ˙ + ξ˙ M2 ξ˙ + γ˙ TM3 γ˙ + γ˙ TM 4 ξ˙ + ξ˙ M 4 γ˙ , for both the beams is 1 m.
2 2 2 2 2 (45)
The structures are excited by impact forces using an electrical force
where K1, K2 and K3 are the modal stiffness matrices listed in the hammer as displayed in Figs. 5 and 6. For the uniform beam, two ac-
Appendix C, and M1, …, M4 represent the modal mass matrices which celeration sensors are installed at a horizontal plane in two vertical

381
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

The 3rd natural frequency (Hz) Accurate result

Element number
Series expansion order Element number Series expansion order

(a) (b)
Fig. 12. (a) The 3D plot of the 3rd natural frequency with respect to the series expansion order and the element number, (b) The 2D plot of the 3rd natural frequency with respect to the
series expansion order and the element number.

Table 3 directions and located at a certain altitude of the beam as shown in

The 3rd natural frequency with respect to the series expansion order and the element Fig. 5(a). The total number of testing points is 20 which also equals to
number.
the number of the moving excitation points. For there are two direc-
The 3rd natural frequency (Hz) tions at each testing point, 40 frequency response functions (FRF) are
obtained. The FRFs are taken average, and the natural frequencies and
Element number Series expansion order the mode shapes are identified by the poly-reference least squares
methods in the complex frequency domain. In the tests, the data ac-
8 12 16 20 24 28
quisition device is LMS SCM-05.
5 273.653 250.920 250.486 250.484 250.484 250.484 For the non-uniform beam, Fig. 6(a) and (b) show the distributions
10 266.092 253.675 253.621 253.621 253.621 253.621 of the four acceleration sensors. They are located at 0.75 m and 0.98 m
15 262.195 254.211 254.203 254.203 254.203 254.203 from the fixed end of the non-uniform beam. The modal parameters can
20 260.198 254.401 254.399 254.399 254.399 254.399
be identified from the measured FRFs. In particular, double-impacts are
25 259.008 254.486 254.485 254.485 254.485 254.485
30 258.224 254.529 254.529 254.529 254.529 254.529 commonly found at the free end of the beam which will make the FRFs
fuzzy and result in noises. To weaken this effect, two sets of sensors are

1st 2nd 3rd

Position Position Position

4th 5th 6th

Position Position Position

Fig. 13. The first six order bending mode shapes of the non-uniform beam calculated by the TDTM.

382
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

1st 2nd 3rd

Position Position Position

4th 5th 6th

Position Position Position

Fig. 14. The first six order torsional mode shapes of the non-uniform beam calculated by the TDTM.

Fig. 15. (a) The 1st experimental mode shape, (b) the 1st mode shape by the FEM.

Fig. 16. (a) The 2nd experimental mode shape, (b) the 2nd mode shape by the FEM.

383
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

Fig. 17. (a) The 3rd experimental mode shape, (b) the 3rd mode shape by the FEM.

Fig. 18. (a) The 4th experimental mode shape, (b) the 4th mode shape by the FEM.

Fig. 19. (a) The 5th experimental mode shape, (b) the 5th mode shape by the FEM.

used to guarantee good quality of the FRFs, and the average value of 80
measured FRFs is used for the identification of the modal parameters.
Due to the small axial displacements, the axial vibrations of the two
beam models are neglected.
5.2. Numerical calculations by the TDTM
For the uniform channel thin-walled beam, the material and geo-
metric properties are listed as follows:
E = 1.95 × 1011N⋅m−2 , μ = 0.3, G = 7.5 × 1010N⋅m−2 ,
ρ = 7900kg⋅m−1, E1 = 2.143 × 1011N⋅m−2, A = 1.78 × 10−4m−2, e
= 1.46 × 10−2m,
Ia = 8.208 × 10−4kg⋅m, Iy = 5.59 × 10−8m4 , J = 2.408 × 10−10m4 , Γ
= 3.238 × 10−12m6.
The natural frequencies obtained by the experiments and the nu-
merical methods are listed in Table 1. The influences of the parameter e
and the warping stiffness are studied. The commercial software ANSYS
is applied to obtain the FEM results for the comparison. The number of
3D elements meshed in the ANSYS is 14,400. Whereas, in the TDTM,
only one beam element is needed for the uniform beam, and the ac-
curacy is controlled by the order of modal expansion.
The influence of the series expansion order on the computing ac-
curacy is analyzed. In this study, on the condition of the series

384
J. Zhou et al.
Fig. 20. (a) The 6th experimental mode shape, (b) the 6th mode shape by the FEM.
Responses (g/N)
Present method
Experiment
Frequency (Hz)
Fig. 21. Bending FRFs of the non-uniform thin-walled beam.
expansion order m = 50 in Eqs. (29c)-(29d), the corresponding results
are treated as the convergence solutions of the TDTM. According to the
calculations of the first three modes, the relative errors between the
results of different series expansion orders and the convergence solu-
tions are obtained as shown in Fig. 7. It can be seen that the first two
modes converge faster than the third mode with the increase of the
series expansion order. In this paper, 0.5% is set as the threshold value.
If the relative errors are smaller than 0.5%, we think the modal fre-
quency converges. For the calculations of the first three modes in Fig. 7,
the series expansion order should be higher than 25 to obtain the sa-
tisfied results. Obviously, if the order of the concerned mode increases,
the series expansion order of the TDTM should also increase. For ex-
ample, if the first six modes are considered, the series expansion order
should be set at least 30. However, the increment of series expansion
orders will not lead to the increase of the matrix dimension. Thus, the
calculating efficiency of the present method can be significantly im-
proved compared with other methods.
The mode shapes of the first six orders are calculated by the TDTM
as shown in Fig. 8 and Fig. 9. It can be seen that the amplitudes of the
vibration in the bending and the torsional directions occur simulta-
neously, which indicates that there exist the bending-twist coupling
effects in the thin-walled uniform beam. Besides, the number of the
modal nodes reflects the complexity of the mode shape. From the mode
shapes, one can observe that the number of the modal nodes no longer
increases with the increase of the mode order, which is different from
the general rule of a solid beam. In some particular modes, such as the
fifth mode of the torsional degree of freedom, the number of modal
385
Thin-Walled Structures 127 (2018) 373–388
nodes decreases compared with the fourth mode. This is due to the
structural properties of the thin-walled beam. The complexity of the
mode shape does not necessarily result in a higher natural frequency. In
thin-walled beams, some modes appear to have high potential energy
but are less complex in the mode shapes. The similar conclusion can be
found in Refs. [10,22].
The FRFs in the bending direction of the uniform beam obtained by
the present and the experimental methods are plotted in Fig. 10. The
FRFs present a series of response amplitudes of the forced vibration
with different excitation frequencies. In Eqs. (43a) and (43b), the first
six mode shapes are used for the calculations of the potential energy U
and the kinetic energy T in Eqs. (44) and (45). In the experiments, the
excited point and the response point being selected are located at
0.97 m and 0.75 m from the bottom of the beam, respectively.
By comparing the amplitudes and the natural frequencies calculated
by the present method with those of the experiment, it can be con-
cluded that the present method is correct and effective, and the para-
metric settings are satisfactory. The errors between the numerical cal-
culations and the experiments mainly result from the structural
damping, the location deviations of the applied forces and the sensi-
tivities of the sensors.
For the non-uniform beam, the contours of the beam section vary
linearly along the beam length. Each structural property of the beam
cross section is interpolated and calculated individually. Geometries at
both ends of the beam are given in Fig. 4. The material properties being
used are
E = 1.95 × 1011N⋅m−2, G = 7.5 × 1010N⋅m−2,
μ = 0.3, E1 = 2.143 × 1011N⋅m−2, ρ = 7900kg⋅m−1.
Table 2 shows the natural frequencies obtained by the experiment,
the FEM and the TDTM for the non-uniform beam. The results of dif-
ferent meshed numbers of elements for the TDTM are compared. The
meshed number in the FEM is 12,800. It is observed that there is a good
agreement between the experiment, the FEM and the TDTM results. The
natural frequencies become convergent as the meshed number of ele-
ments increases in the TDTM.
Similar to the uniform beam, the relative error with respect to the
element number used in the TDTM is calculated as shown in Fig. 11. It
is obvious that the accuracy of calculation is relatively low if the ele-
ment number is small. In this paper, the results are considered as the
convergence solutions when the total element number N = 200. If the
element number is 20, the relative errors of the first four modes are
smaller than 0.5%. For the element number 30, the relative errors are
smaller than 0.2%. During the calculations, the series expansion order
keeps at 20. In fact, the overall accuracy is affected both by the total
element number and the series expansion order, which will be discussed
as follows.
In order to examine the influences of the series expansion order and
the element number on the natural frequencies, the 3rd order natural
J. Zhou et al.
frequency is calculated as shown in Fig. 12. The range of the series
expansion order is from 8 to 28, while the range of the element number
is from 5 to 30. It is clear that once the element number is fixed, the
natural frequency will approach to the accurate result with the increase
of the series expansion order. In another direction, if the series ex-
pansion order remains the same, such as 20, refining the mesh can lead
to the results with a higher precision as listed in Table 3.
The first six mode shapes for a non-uniform channel beam calcu-
lated by the TDTM are plotted in Fig. 13 and Fig. 14. The mode shapes
identified through the test and calculated by the FEM are shown in
Figs. 15–20. It can be observed from these figures that the mode shapes
obtained by the test and by the FEM are quite correlated. The bending
and torsional vibrations will happen synchronously, which indicates
that the motions of the two degrees of freedom are coupled together.
The warping effect is also observed from the out-plane displacements of
the beam sections. For the higher order modes, the distortion de-
formations contribute more, which leads to the increasing errors.
The FRFs for the non-uniform beam obtained by the present nu-
merical method and the experiment are presented in Fig. 21. The ex-
cited point is at the height of 0.75 m while the measured point is at
0.97 m from the bottom of the beam. The first fourth natural fre-
quencies can be clearly observed in the figure and a good agreement is
achieved between the two FRFs, which further verifies the validity of
the present developed methodology.
6. Discussions and conclusions
In this study, the TDTM is introduced as a new methodology to solve
the non-uniform beams and validated by the experiments and the FEM.
The basic idea of TDTM derives from the characteristics of the DTM and
the TMM. The TDTM also utilizes the ideas drawn from the FEM, and it
gives the FEM a changeable mode shape function. This combination
provides the TDTM with the flexibility of controlling the accuracy of
solutions using both the number of the meshed element and the series
expansion order of the mode shape. For the uniform part of the struc-
ture, the DTM can greatly reduce the size of the model for achieving a
certain level of computing accuracy. For the non-uniform part of the
structure, the TMM helps capture the structural complexity in detail
and inherits the merits of the FEM. Consequently, this combination
significantly improves the computing efficiency in solving large scale
complex non-uniform structures.
Appendix A
See Table A1
Table A.1
The area integrals for the expressions in Eqs. (8)–(16).
∬A dy dz = A
∬A y dy dz = Ae
Appendix B
See Tables B1 and B2
Table B.1
The DTM transformation regulations.
Original function
f (x ) = g (x ) ± h (x )
f (x ) = λg (x )
dng (x )
f (x ) =
dnx
Thin-Walled Structures 127 (2018) 373–388
The structural governing equations featuring the bending and tor-
sional vibrations have been established through the virtual displace-
ments of each beam section. The equations of motion are obtained by
Hamilton's principle. Both warping and rotary inertia are considered.
The principles of the DTM and the TDTM have been introduced re-
spectively. Based on the Taylor's series expansion, the assumed mode
shapes approach the analytical ones as the order of expansion increases.
Specifically, the influences of the number of the meshed element and
the series expansion order of the mode shape are discussed.
Furthermore, two illustrative validation experiments have been pre-
sented. Comparisons among the numerical calculations, the test and the
FEM results show that the accuracy of the TDTM is satisfactory.
Through the investigations, the main conclusions are drawn as follows:
(1) For the uniform beam, since the calculating accuracy can be solely
controlled by the DTM, no meshing operation is needed. It is sug-
gested to set the series expansion order above 25 to obtain a good
accuracy for the first three modes.
(2) In the case of a non-uniform beam, the effect of the series expansion
order is smaller than that of the mesh density. Refining the mesh
can improve the calculating accuracy significantly since the total
element number contributes to the modeling of the structure
greatly. Thus, the computing efficiency of the TDTM can be en-
hanced by reducing the series expansion orders with a mesh density
satisfying the precision requirement.
(3) Ignoring the warping effect will lead to decreased natural fre-
quencies. The distance between the shear center and the centroid of
the beam section causes the bending and torsional coupling effect,
which might generate a significant error if this distance is omitted.
(4) The bending and torsional mode shapes of the uniform and non-
uniform beams prove the existence of the bending-twist coupling
vibration of the structures. Besides, for a thin-walled structure, the
number of the modal nodes does not increase linearly with the in-
crease of the mode order, which is different from the solid beams.
Funding
This study was funded by the National Natural Science Foundation
of China (Grant Nos. 11761131006, 11572007 and 11172084).Conflict
of Interest
The authors declare that they have no conflict of interest.
∬A y 2 dy dz = Iz ∬A z2dy dz = Iy
∬A ρ (y 2 + z2) dy dz = Ia ∬A z dy dz = ∬A yz dy dz = 0
Transformed function
F [k ] = G (k ) ± H (k )
F [k ] = λG [k ]
(k + n) !
F [k ] = G [k + n]
k!
386
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

Table B.2
The DTM transformations for boundary conditions (B.C.) in Eq. (29).

x=0 x=1
Original B.C. Transformed B.C. Original B.C. Transformed B.C.

f (0) = 0 F (0) = 0 f (1) = 0 ∞

∑k = 0 F (k ) = 0
df F (1) = 0 df ∞
(0) =0 (1) =0 ∑k = 0 kF (k ) = 0
dx dx
∞
d2f F (2) = 0 d2f ∑k = 0 k (k − 1) F (k ) = 0
(0) = 0 (1) = 0
dx 2 dx 2
∞
d3f F (3) = 0 d3f ∑k = 0 k (k − 1)(k − 2) F (k ) = 0
(0) = 0 (1) = 0
dx 3 dx 3

Appendix C

See Table C1

Table C.1
The expressions of the modal stiffness, modal mass and forcing matrices in Eqs. (44), (45) and (47).

Modal stiffness matrices Modal mass matrices Forcing matrices

l
K1 = ∫0 EI (x ) W (x ) WT (x )dx l dW (x ) dWT (x ) l
Ft = ∫0 qt (x ) WT (x )dx
M1 = ∫0 ρI (x ) dx
dx dx
T l l
K2 = ∫0l GJ (x ) dΨdx(x ) dΨdx(x ) dx M2 = ∫0 Ia (x ) Ψ (x ) ΨT (x )dx Fm = ∫0 qm (x ) ΨT (x )dx
2 2 T l
K3 = ∫0l E1 Γ (x ) d Ψ2(x ) d Ψ 2(x ) dx M3 = ∫0 ρA (x ) W (x ) WT (x )dx
dx dx
l
M 4 = ∫0 2ρAe (x ) W (x ) ΨT (x )dx

References
[1] C. Mei, Coupled virations of thin-walled beams of open section using the finite
elements method, Int. J. Mech. Sci. 12 (1970) 883–891.
[2] E. Dokumaci, An exact solution for coupled bending and torsion vibration of uni-
form beams having single cross-sectional symmetry, J. Sound Vib. 119 (3) (1987)
443–449.
[3] J.R. Banerjee, Coupled bending-torsional dynamic stiffness matrix for beam ele-
ments, Int. J. Numer. Methods Eng. 28 (1989) 1283–1298.
[4] J.R. Banerjee, F.W. Williams, Coupled bending-torsional dynamic stiffness matrix
for Timoshenko beam elements, Comput. Struct. 42 (1992) 301–310.
[5] H.Y. Lai, B. Fallahi, R. Gupta, S. Bo, Dynamic characterization of thin-walled la-
minated channel beams with warping, Comput. Struct. 48 (1993) 1137–1151.
[6] R.E.D. Bishop, S.M. Cannon, S. Miao, On coupled bending and torsional vibration of
uniform beams, J. Sound Vib. 131 (3) (1989) (475–464).
[7] M.J. Klausbruckner, R.J. Pryputniewicz, Theoretical and experimental study of
coupled vibrations of channel beams, J. Sound Vib. 183 (2) (1995) 239–252.
[8] M. Tanaka, A.N. Bercin, Finite element modelling of the coupled bending and
torsional free vibration of uniform beams with an arbitrary cross-section, Appl.
Math. Model. 21 (1997) 339–344.
[9] J. Li, R.Y. Shen, H.X. Hua, X.D. Jin, Coupled bending and torsional vibration of
axially loaded thin-walled Timoshenko beams, Int. J. Mech. Sci. 46 (2004)
299–320.
[10] J. Li, W.Y. Li, R.Y. Shen, H.X. Hua, Coupled bending and torsional vibration of
nonsymmetrical axially loaded thin-walled Bernoulli-Euler beams, Mech. Res.
Commun. 31 (2004) 697–711.
[11] O.O. Ozgumus, M.O. Kaya, Vibration analysis of a rotating tapered Timoshenko
beam using DTM, Meccanica 45 (2010) 33–42.
[12] O.O. Ozgumus, M.O. Kaya, Energy expression and free vibration analysis of a ro-
tating double tapered Timoshenko beam featuring bending-torsion coupling, Int. J.
Eng. Sci. 45 (2007) 562–586.
[13] F.D. Borbon, A. Mirasso, D. Ambrosini, A beam element for coupled torsional-flexral
vibration of doubly unsymmetrical thin walled beams axially loaded, Comput.
Struct. 89 (2011) 1406–1416.
[14] A.D. Egidio, A. Contento, F. Vestroni, The role of nonlinear torsional contributions
on the stability of flexural-torsional oscillations of open-cross section beams, J.
Sound Vib. 358 (2015) 236–250.
[15] F.M. Li, K. Kishimoto, Y.S. Wang, Z.B. Chen, W.H. Huang, Vibration control of
beams with active constrained layer damping, Smart Mater. Struct. 17 (6) (2008)
065036.
[16] F.M. Li, K. Kishimoto, W.H. Huang, The calculations of natural frequencies and
forced vibration responses of conical shell using the Rayleigh-Ritz method, Mech.
Res. Commun. 36 (2009) 595–602.
[17] O.O. Ozgumus, M.O. Kaya, Energy expressions and free vibration analysis of a ro-
tating Timoshenko beam featuring bending–bending-torsion coupling, Arch. Appl.
Mech. 83 (2013) 97–108.
[18] M.V. Bastawrous, A.A. EI-Badawy, A study on coupled bending and torsional vi-
brations of wind turbine blades, Adv. Mater. Res. 622–623 (2013) 1236–1242.
[19] S.S. Rao, H.L. Jin, Analysis of coupled bending-torsional vibration of beams in the
presence of uncertainties, J. Vib. Acoust. 136 (2014) 051004.
[20] Y. Liu, D.W. Shu, Coupled bending-torsion vibration of a homogeneous beam with a
single delamination subjected to axial loads and static end moments, Acta Mech.
Sin. 30 (4) (2014) 607–614.
[21] M.T.T. Kashani, S. Jayasinghe, S.M. Hashemi, On the flexural-torsional vibration
and stability of beams subjected to axial load and end moment, Shock Vib. (2014)
153532.
[22] J. Li, C.X. Shi, X.S. Kong, Stochastic response of an axially loaded composite
Timoshenko beam exhibiting bending-torsion coupling, Arch. Appl. Mech. 84
(2014) 109–122.
[23] C.G. Cooley, R.G. Parker, Vibration of spinning cantilever beams with an attached
rigid body undergoing bending-bending-torsional-axial motions, J. Appl. Mech. 81
(2014) 051002.
[24] J. Li, S. Wang, X.B. Li, X.S. Kong, Modeling the coupled bending–torsional vibra-
tions of symmetric laminated composite beams, Arch. Appl. Mech. 85 (2015)
991–1007.
[25] A. Bekir, C. Omer, Buckling analysis of linearly tapered micro-columns based on
strain gradient elasticity, Struct. Eng. Mech. (2013) 195–205.
[26] S. Eken, M.O. Kaya, Flexural-torsional coupled vibration of anisotropic thin-walled
beams with biconvex cross-section, Thin-Walled Struct. (2015) 372–383.
[27] G. Garcea, R. Gonçalves, A. Bilotta, D. Manta, R. Bebiano, L. Leonetti, D. Magisano,
D. Camotim, Deformation modes of thin-walled members: a comparison between
the method of generalized eigenvectors and generalized beam theory, ThinWalled
Struct. (2016) 192–212.
[28] J. Latalski, J. Warminski, G. Rega, Bending-twisting vibrations of a rotating hub-
thin-walled composite beam system, Math. Mech. Solids (2016) 1–23.
[29] O. Bourihane, A. Ed-dinari, B. Braikat, M. Jamal, F. Mohri, N. Damil, Stability
analysis of thin-walled beams with open section subject to arbitrary loads, Thin
Walled Struct. (2016), pp. 156–171.
[30] R. Pavazza, A. Matoković, Bending of thin-walled beams of open section with in-
fluence of shear, part I: Theory, Thin-Walled Struct. (2017) 357–368.
[31] R. Pavazza, A. Matoković, M. Vukasović, Bending of thin-walled beams of open
section with influence of shear, Part Ⅱ: application, ThinWalled Struct. (2017)
369–386.
[32] C.C. Wang, C.N. Huang, H.T. Yau, S.X. Hsu, Bifurcation and chaos analysis of
atomic force microscope system, Microsyst. Technol. 19 (2013) 1795–1805.
[33] J.S. Shabnam, A. Reza, Free vibration analysis of variable thickness thin plates by
two-dimensional differential transform method, Acta Mech. 224 (2013) 1643–1658.
[34] S. Rajasekaran, E.N. Tochaei, Free vibration analysis of axially functionally graded
tapered Timoshenko beams using differential transformation element method and
differential quadrature element method of lowest-order, Meccanica 49 (2014)
995–1009.
[35] W. Nuttawit, C. Arisara, On the linear and nonlinear vibration responses of elasti-
cally end restrained beams using DTM, Mech. Based Des. Struct. Mach. 42 (2014)

387
J. Zhou et al. Thin-Walled Structures 127 (2018) 373–388

135–150.
[36] W. Nuttawit, C. Jarruwat, Vibration characteristics of stepped beams made of FGM
using differential transformation method, Meccanica 50 (2015) 1089–1101.
[37] E. Farzad, N. Parisa, A nonlocal Timoshenko beam theory for vibration analysis of
thick nanobeams using differential transform method, J. Theor. Appl. Mech. 53 (4)
(2015) 1041–1052.
[38] K.B. Mustapha, M.A. Hawwa, Eigenanalyses of functionally graded micro-scale
beams entrapped in an axially-directed magnetic field with elastic restraints, Int. J.
Struct. Stab. Dyn. 16 (2016) 1550022.
[39] K.B. Ravi, S.S. Hariharan, Investigation on vibration characteristics of fluid con-
veying single walled carbon nanotube via DTM, Aust. J. Mech. Eng. (2017)
2204–2253.

388