Journal of Sound and Vibration 505 (2021) 116157

Contents lists available at ScienceDirect

Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsv

An approximate method for pipes conveying fluid with strong

boundaries
Xiao-Ye Mao a, Song Shu b, Xin Fan b,c, Hu Ding a,∗, Li-Qun Chen a
a
Shanghai Key Laboratory of Mechanics in Energy Engineering, Shanghai Institute of Applied Mathematics and Mechanics, School of
Mechanics and Engineering Science, Shanghai University, Shanghai, 200444, China
b
State-owned Machinery Factory in Wuhu, Wuhu 241007, China
c
University of Science and Technology of China, Anhui 230026, China

a r t i c l e i n f o a b s t r a c t

Article history: An approximate method is proposed for the strong nonlinear and non-homogenous bound-
Received 24 December 2020 ary value problem of a pipe conveying fluid for the first time. Usually, the boundary value
Revised 11 March 2021
is satisfied transcendentally in the truncation processing acting on the partial differen-
Accepted 20 April 2021
tial governing equation. However, the nonlinear and non-homogenous boundary disables
Available online 23 April 2021
it. To overcome this problem, the method of modal correction together with the modal
Keywords: projection is proposed. This method treats nonlinear and non-homogenous boundaries as
gyroscopic system generalized governing equations. Since then, nonlinear and non-homogenous terms in the
axially moving beam boundary could be discussed fully based on the harmonic balance method (HBM). The
pipe conveying fluid discussion on natural frequencies suggests the standard of the convergence of the modal
approximate method projection. The harmonic convergence can be judged by the solution with more harmon-
nonlinear boundary ics. By treating those coefficients of harmonics as time-varying parameters, state equations
non-homogenous boundary will be produced. Based on them, both of the stability of the approximate solution and
pseudo arc-length method
the type of bifurcation could be judged. Besides, coefficients of each order harmonic on
different modal projections or spatial corrections could reveal the detailed information of
the response, such as the harmonic caused by the nonlinearity or the power distribution
on different modal projection. By comparing with the Dirac operator and the multiscale
method, the advantage of the proposed method on dealing with strong boundaries is veri-
fied. Nonlinear and non-homogenous boundaries are not the trap for the gyroscopic system
anymore.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction

The pipe conveying fluid is widespread in many fields. It plays a key role in the petroleum transportation, the fuel feeding
for all kinds of engines, the chemical processing and so on. Theoretically, it is a typical gyroscopic system subjected to the
Coriolis force in the lateral direction and the investigation on it bloomed after word-war II [1,2].
At the beginning, studies focused on simply-supported pipes [1,2]. The cantilever tube was first investigated by Long
[3]. Different from a ordinary beam, the stability of the equilibrium could change with the fluid speed. Except the buckled

∗
Corresponding author. H. Ding, Shanghai Institute of Applied Mathematics and Mechanics, 149 Yan Chang Road, Shanghai 20 0 072, China, Tel: +86 21
56331451
E-mail address: [email protected] (H. Ding).

https://doi.org/10.1016/j.jsv.2021.116157
0022-460X/© 2021 Elsevier Ltd. All rights reserved.
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

instability, oscillatory instability was found on the cantilever tube [4–7]. In this area, Païdoussis and his co-workers did
outstanding studies [8–10]. They also introduced nonlinearities into the investigation [11]. Following these pioneering works,
the investigation on the pipe conveying fluid has developed for about 70 years. Among these jobs, the parametric vibration
takes an important position. Païdoussis and Sundararajan investigated it as early as 1975 [12]. In 1986, Ariaratnam and
Namachchivaya discussed the dynamic stability of a pipe conveying pulsating fluid [13]. Parametric resonance can also be
found in [14,15]. Meanwhile, the resonance subjected to an external excitation [16] is also discussed. For complex vibrations,
the internal resonance [17–20] and the 3D vibration [21,22] also took the researchers’ attention. Nowadays, the investigation
on the pipe conveying fluid has been developed to the nano-scale [23].
However, most of the previous investigations were based on the ideal boundary condition. They neglected the nonlinear
and the non-homogenous activities in the boundary. This is because it is hard to get a spatial discretization before the
approximate or the numerical integration is carried out. For the statics model, the nonlinear boundary value problem is
not a trap anymore. The homotopy analysis method was developed and widely used [24]. Besides, the modified Adomian
decomposition method [25], the wavelet method [26] and the generalized differential transform method [27] also made
their contribution to the nonlinear boundary value problem. Although these methods are developed for the statics problem,
some ideas can be employed in the analysis on dynamics. For example, the series expansion of the modified Adomian
decomposition method is meaningful in the processing of spatial coordinates. Recently, Mao and Ding proposed the modal
correction based multi-scales method and the model correction based harmonic balance method for nonlinear boundary
problems of flexible beams [28-31]. They employed the series expansion to modify the modal expansion. In this way, the
solution could satisfy the nonlinear boundary while the natural characters still can be revealed. The boundary operator
proposed by Guo and Rega is another kind of method for the nonlinear and non-homogenous boundary [32]. However,
they found that the boundary operator had deficiency while comparing with the method by Mao and Ding. The boundary
operator is suitable for the weak boundary. Anyhow, all their works have found that the boundary has important effects to
the dynamic of a continuum without Coriolis force. The similar thing must be real for the pipes conveying fluid. Therefore,
it is necessary to develop the approximate method to the gyroscopic system.
Except the boundary value problem, how to obtain an accurate response of a system suffering from a strong nonlinear-
ity is another key problem. As known, some classical perturbation methods such as Poincaré–Lindstedt, multiple scale, or
Krylov–Bogoliubov–Mitropolski method are suitable for the weak linear system [33]. Hence, the harmonic balance method
(HBM) is a good choice for its outstanding advantage. HBM is a frequency-domain method based on the Fourier series. It
is obvious to improve the accuracy of the solution by employing a higher-order expansion. However, this advantage will be
a disadvantage for the number of coefficient equations is increasing exponentially. Consequently, lots of efforts are paid on
reducing the calculation, such as the incremental harmonic balance method (IHBM) [34,35], the adaptive harmonic balance
method (AHBM) [36], the multi-level residue HBM [37] and the global residue HBM [38]. Until recently, the efficiency of
HBM is still improving. In 2017, Hamed Haddad et al. proposed a HBM based method who can solve multi-dimensional non-
linear systems using linear equation [39]. In the same year, Lee demonstrated a modified multi-level residue HBM [40] which
can generate the higher-level nonlinear solutions that are neglected by [37]. Zhu and Wang also showed a modified IHBM
based on the fast Fourier transformation (FFT) to enhance the efficiency of IHBM further more [41]. Besides, the HBM com-
bined with the IHBM [42], the reduced space HBM [43], the two-step pseudo-response based adaptive HBM [44] and the
Galerkin averaging-IHBM [45] also show their high efficiency in calculation. Except the efficiency, some HBMs are devel-
oped to improve the capacity of HBM. For example, the multiple HBM was developed to study the aperiodic vibration [46].
The Galerkin HBM can overcome Gibbs phenomenon [47]. The multi-term HBM was proposed to investigate gear pairs [48].
The constrained HBM can follow the Hopf bifurcation of a self-excitation [49]. The homotopy HBM by Fung et al. [50] and
the generalized HBM by Luo [51] can follow very complicated bifurcations. The coupled static/dynamic HBM can obtain the
static and the dynamical displacement at the same time [52]. The mixed shooting-HBM combines the best of both shooting
method and HBM [53] and it is suitable for the dry friction. Huang and Zhu developed a multi-time scales IHBM for quasi-
periodic response [54]. The mentioned works indicate that HBM has an exuberant vitality in dealing with different kinds of
nonlinear problems.
In the present work, an approximate method based on the modal correction together with the harmonic balance method
is proposed. A general statement of the method will be introduced first. After that, the convergence of the modal projection
and also of the harmonic expansion will be discussed. Then, a few numerical examples will be given to demonstrate the
capacity of the proposed method. The advantage of the proposed method for the strong boundary will be also verified by
other methods. At last, some conclusions are deduced based on the discussion.

2. Method Introduction

As demonstrated in Fig. 1, a pipe conveying fluid is restrained by clips at two ends. For most of pipes are slender, the
Euler-Bernoulli model is considered. Together with the generalized Hamilton’s principle, the governing equation of motion
of the pipe conveying fluid can be produced.
The kinetic energy of the pipe is
 L
1
Tp = ρp A p U , T 2 d X (1)
2 0

2
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 1. Model of a pipe conveying fluid

where L is the distance between the two supports. ρ p is the density of the pipe and Ap is the cross section area. The trans-
verse displacement of the pipe is denoted by U(X,T). The comma in front of T represents partial derivation of T. Meanwhile,
the kinetic energy of the fluid is
 L  L
1 1
Tf = ρf A f  2 d X + ρf A f U , T 2 d X (2)
2 0 2 0

where ρ f is the density of the fluid and Af is the cross section area.  is the speed of the fluid inside the pipe.
The potential energy of the pipe reads
 L
1 1
 2  L
1
Up = Ep Ap U ,X 2 dX + E IpU ,XX 2 dX (3)
0 2 2 0 2
The virtual work of the axial force P0 along the pipe caused by the installation error is
 L  L
 
δ QP0 = (dX )2 + (dU )2 − dX
P0 (dS − dX ) = P
0 0
 L   L 
 
= P0 (dX )2 + (U ,X dX )2 − dX = P0 1 + U,2X − 1 dX (4)
0 0

The virtual work caused by the pressure of the fluid is

 L  L
 
δ Qf = Af pf ( dS − dX ) = Af pf (dS )2 + (dU )2 − dX
0 0
 L
  L  
= Af pf (dX )2 + (U ,X dX )2 − dX = Af pf 1 + U,2X − 1 dX (5)
0 0

The virtual work of the excitation can be expressed as

 L
δ QF = F UdX (6)
0

Although the boundary is complicated, but it can be expressed by the restraint force and restraint torque. Consequently, the
virtual work of the restraint activities is
δ QB = FBL δ ( U |X=0 ) + FBR δ ( U |X=L )
+ MBL δ ( U ,X |X=0 ) + MBR δ ( U ,X |X=L ) (7)
where the subscripts BL and BR mark the location of these activities, which means BL stands for the left while BR repre-
sents the right. With the help of the generalized Hamilton’s principle, the governing equation of the pipe conveying fluid is
deduced as
3Ep Ap 2
(ρp Ap +ρf Af )U ,T T + 2ρp Ap U ,XT + ρp Ap  2 − P0 − pAf − U,X U ,XX + Ep IpU ,X X X X = F (8)
2
with the corresponding boundary
Ep IpUL ,XX − MBL = 0, Ep IpUR ,XX + MBR = 0,
Ep IpUL ,X X X +FBL = 0; Ep IpUR ,X X X − FBR = 0 (9)
Considering the Kelvin’s material derivative, which means
dε
σ =Ep ε + p (10)
dT
, the material damping of the pipe is introduced. As an equivalent rule, Eq. (10) could be simplified as
d
Ep ∼ Ep + p (11)
dT
3
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Substitution of Eq. (11) in the governing equation Eq. (8) yields

(ρp Ap +ρf Af )U ,T T + 2ρp Ap U ,XT + Ep IpU ,X X X X + p IpU ,X X X X T
3Ep Ap 2
+ ρp Ap  2 − P0 − pAf − U,X −3p ApU ,X U ,XT U ,XX = F (12)
2
In order to describe the processing of the proposed method more cleanly in visual, the governing equation will be nondi-
mensionalized according to following rules.

U X T Ep Ip ρp A p
u= ,x = ,t = 2 , v = L ,
L L L ρp A p + ρf A f Ep Ip

ρp A p Ap L2 (P0 +Af p)L2 p Ip

Mr = ,κ = , P0 = ,ζ = 2 (13)
ρp A p + ρf A f Ip Ep Ip L (ρp Ap + ρf Af )Ep
Based on these rules, the governing equation will be
3κ 2
u,tt + 2Mr vu,xt + v2 − P0 − u,x 2 − 3ζ κ 2 u,x u,xt u,xx + u,xxxx + ζ u,xxxxt = f (14)
2
with following boundaries
uL ,xx − M˜ BL = 0, uR ,xx + M ˜ BR = 0,
˜ ˜
uL ,xxx +FBL = 0, uR ,xxx − FBR = 0 (15)
To describe restraint activities in a brief style, the boundary given in Eq. (15) is replaced by:
L0 + L 1 = 0 (16)
In Eq. (16), L0 stands for the linear and homogenous part of the boundary while the second one is on behalf of those
nonlinear or non-homogenous ones. Based on L0 , modal shapes can be determined via the complex mode method and
they are marked as i . It should be mentioned here that modal shapes of a gyroscopic system are composed by normal
modes of a non-traveling Euler beam in transverse vibration with L0 . This is one of the biggest difference between the
static continuum and the gyroscopic continuum. Based on the modal shape, the solution satisfying L0 can be given as:

n m  
u˜ = i aˆi,0 + aˆi, j cos ( j t ) + bˆ i, j sin ( j t ) (17)
i=1 j=1

where aˆ and bˆ are coefficients of corresponding harmonics. As mentioned above, the modal shape i will read
r

i = ci,k ϕk (18)
k=1

where ϕ is the mode of a non-traveling Euler beam in transverse vibration with the boundary L0 . Substitution of Eq. (18) in
Eq. (17) yields

n  m    
u˜ = ϕi ai,0 t˜ + ai, j t˜ cos ( j t ) + bi, j t˜ sin ( j t ) (19)
i=1 j=1

where t˜ denotes the slow state time. The coefficients a and b here are different with those aˆ and bˆ in Eq. (17), as they imply
the modal coefficients c in Eq. (18). According to the classical HBM rule, a and b will be treated as constants. However, the
slow state time here is meaningful for the stability analysis [55]. Furthermore, Eq. (19) can be simplified as
n
u˜ = qi ϕi (20)
i=1

Now the solution here just satisfies the linear and homogenous boundary. In order to make the solution satisfy the nonlinear
and non-homogenous boundary, an amended solution p is introduced [28,31]. Thus, the whole solution is combined by
u = u˜ + p (21)
Substituting Eq. (21) into Eq. (16), making use of Eq. (20), the amended solution satisfies
L0 ( p) + L1 (u˜ + p) = 0 (22)
In Eq. (22), the second part, or the nonlinear and non-homogenous boundary namely, will be replaced by four coefficients
as an Euler beam has four boundaries, which means
L0 ( p)|le f t = α0 andα1 ; L0 ( p)|right = β0 andβ1 (23)

4
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

They are treated as generalized coordinates like q in Eq. (20). Now the key here is the solution of p. Based on Eq. (23), a
general form of the amended p is proposed.
1 1
p(x, t ) = αi (t )φi (x ) + βi (t )ψi (x ) (24)
i=0 i=0

where the special function φ and ψ can be integrated from Eq. (23) or determined by a series expansion. In Refs. [28-31],
the Legendre polynomial is proposed to get the spatial expression conveniently. Substituting the solution Eq. (24) into the
boundary Eq. (22), the boundary will evolve to four ODEs. They can be treated as supplementary governing equations. The
four coefficients can take the same form as those of q based on the generalized HBM [55].
 m    
α0 = α0,0 t˜ + α0, j1 t˜ cos ( j t ) + α0, j2 t˜ sin ( j t ) ;
j=1

 m    
α1 = α1,0 t˜ + α1, j1 t˜ cos ( j t ) + α1, j2 t˜ sin ( j t ) ;
j=1

 m    
β0 = β0,0 t˜ + β0, j1 t˜ cos ( j t ) + β0, j2 t˜ sin ( j t ) ;
j=1

 m    
β1 = β1,0 t˜ + β1, j1 t˜ cos ( j t ) + β1, j2 t˜ sin ( j t ) (25)
j=1

Now substituting the solution Eq. (21) into the governing equation Eq. (14) and the boundary Eq. (16), making use of
Eq. (19) and Eq. (25), there comes five equations with harmonics. The first one is the governing equation of the continuum.
It can be projected to the modal shape space via the Galerkin method to produce a set of ordinary differential equations
(ODEs). The other four are boundaries and they are just ODEs. Finally, the nonlinear and non-homogenous boundary value
problem of gyroscopic systems becomes a normal problem of ODEs.
Substituting Eq. (21) into the governing equation, making use of the Galerkin method and those boundaries, there comes
⎡ ⎤ ⎡ ⎤
ä a˙
⎢ b̈ ⎥⎢ b˙ ⎥
D2 ⎣ ⎦ + D1 ⎣ ⎦ + D = 0 (26)
α̈ α˙
β̈ β˙
where D2 stands for the coefficient matrix of the second order differential operators to the time scale. D1 is on behalf of the
coefficient matrix of the first order differential operators to the time scale. Meanwhile, D is a column vector and stands for
all other terms. Consequently, the stability of the response can be judged based on this state function. As only the steady-
state response is considered, the second and the first order differential operators are equal to 0. Namely, Eq. (26) becomes

D=0 (27)
Now, all those undetermined coefficients for the period response can be solved out from it. After that, the combined re-
sponse curve comes out. Sometimes, the nonlinearity causes jumping phenomena to the response curve. It makes the
Newton’s method fail to extend the curve, for that the Jacobian matrix here is singular. To overcome this problem, the
pseudo-arc-length method is proposed. Usually, the response curve function Eq. (27) is expressed in the form of
 
D = F a , b , α, β , =0 (28)
Now, treat the external frequency as a variable like those harmonic coefficients, a [(n+4)(2m+1)] × [(n+4)(2m+1)+1] Jaco-
bian matrix to all these variables is produced. Hence, the unit tangent vector reads
J
τ= (29)
J 
where
Ji = (−1 )i+1 det (Jacobiandel ete col umn i ) (30)
Based on the unit tangent vector, the next point in the response curve could be predicted like
y p = y0 + τ (y0 )s (31)
where y stands for all those generalized variables, including the external frequency. s is the given arc-length. After that, the
real solution could be deduced by the Newton’s method:
 −1   
F (y ) F yic−1
yic = yic−1 − T p (32)
J (y p ) 0

5
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157


In this paper, the solution is considered to be convergent while |Fi (yc )| ≤ 10−16 . The given arc-length is 0.1 at the begin-
ning of the curve continuation. After that, it is adjusted according to the number of iterations. For example, if the number
of iterations is smaller than 4, the arc-length is set to 0.1. If the number is between 4 to 9, the arc-length is 0.5number . Oth-
erwise, the arc-length is 0.001. Specially, to track the jumping area successfully, it is suggested that the arc-length should be
changed to 10% of the last one while the iterations number overs 20, until a convergent solution is obtained. Consequently,
the arc-length always evolves to track the response curve efficiently. The stability of the solution could be determined by
Eq. (26), just employing the classical stability theory of ODE. Time derivatives of those coefficients are just zero.

3. Convergence of the method

The convergence is the key for an approximate method. As the proposed method is produced to deal with the partial
differential equation, the convergence of it will be discussed via both the spatial projection and the harmonic expansion. The
spatial projection indicates the convergence of x while the harmonic expansion indicates the convergence of t respectively.

3.1. Convergence of the modal projection

Theoretically, the modal projection will be convergent while n (the highest order of the mode involved in the modal
projection) in Eq. (20) is big enough. However, considering the good balance between the computation efficiency and the
accuracy, there is an appropriate n. It can be determined by the natural frequency of the derived system. The detailed
processing of the convergence of the modal projection is introduced as follows.
Neglecting those nonlinearities in Eq. (14), replacing f by bsin( t), the typical linear governing equation of a pipe con-
veying fluid is produced.
 
u,tt + 2Mr vu,xt + v2 − P0 u,xx + u,xxxx + ζ u,xxxxt − b sin ( t ) = 0 (33)
Those nonlinearities are neglected for demonstrating the effort of the proposed method acting on the nonlinearity in the
boundary. Assuming that the motion of two ends in the transverse direction is so weak that can be ignored, then the pipe
subjects to the boundary condition given in Eq. (34).
uL = 0, uR = 0,
uL ,xx − r3 uL ,x 3 − duL ,xt = 0,
uR ,xx + r3 uR ,x 3 + duR ,xt = 0 (34)
where those restraint torques are expressed by the nonlinear torque and the linear damping torque. In general, the pipe
is simply supported. Based on it, the rotational cubic nonlinearity and the rotational linear damping are added. As intro-
duced in Section 2, the modal projection needs the natural mode of a non-traveling Euler beam in transverse vibration with
corresponding linear boundaries. Since then, the mode used here is
ϕi = sin (iπ x ) (35)
The corrected solution p is deduced based on the Legendre polynomial, which is shown as below:
 7 5 3 5
  
p= − x − x4 + 5x3 + 1 α0 + − x − x4 + 5x3 β0
2 2 2 2
 1 1 2 7 3 5 4
 1 1 5 4

+ − x+ x − x + x α1 + x − x3 + x β1 (36)
8 2 12 24 24 4 24
Now, the expressions of u˜ and p are obtained. Neglecting nonlinearities and the external excitation in the original system,
there comes
 
u˜,tt + 2Mr vu˜,xt + v2 − P0 u˜,xx + u˜,xxxx + ζ u˜,xxxxt
 
+ p,tt + 2Mr v p,xt + v2 − P0 p,xx + p,xxxx + ζ p,xxxxt = 0 (37)

p|x=0 = 0, p|x=1 = 0,
( p,xx − du˜,xt − dp,xt )|x=0 = 0, (38)
( p,xx + du˜,xt + dp,xt )|x=1 = 0
Making use of Eqs. (20) and (36), employing the Galerkin method, there comes the generalized ordinary differential govern-
ing equations like
     
q̈ q˙ q
M i +C i +K i =0 (39)
p̈ p˙ p

where p=[α 0 , α 1 , β 0 , β 1 ]T . By setting Mr =0.8, v=2, P0 =-2.5 and ζ =0.001, natural frequencies are demonstrated with the
change of boundary damping d under different order truncations. In Fig. 2, real parts of the first-two natural frequencies are
presented. Obviously, the natural frequency needs at least one higher order projection to get a reasonably accurate value.
This suggests that the convergence of the modal expansion can be verified by higher-order results.

6
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 2. Discussion of the convergence of the modal projection via natural frequencies: (a) the first-order natural frequency changing with v; (b) the second-
order natural frequency changing with v

3.2. Convergence of the harmonic expansion

After the convergence of the modal projection is discussed, the partial-differential governing equation can be discretized
into a set of ordinary-differential-equations. At this step, the order of the harmonic expansion is the key for the accuracy of
the analytical solution. The convergence of the harmonic expansion is proposed to be judged by the solution of one higher
order expansion.
By setting Mr =0.8, v=2, P0 =-2.5, ζ =0.001, d=0.01 and r3 =104 , coefficients of the 3rd -order harmonic expansion and the
5th -order harmonic expansion are compared in the following figures. The modal projection of them are both set to the third
order. Hence, only those harmonic coefficients on the first-two modes are demonstrated in Fig. 3. It is clear that coefficients
produced by the 3rd -order expansion are very close to those of the 5th -order expansion. This clarifies that the 3rd -order
expansion is already convergent for the primary resonance under current parameters.

4. Strong boundary and complex excitations

Now both of the convergence of the spatial projection and the convergence of the time response have been proved.
As the boundary is treated as the generalized governing equation, it will be taken into the investigation fully. Besides, the
harmonic balance method is employed to deal with the time response, the proposed method can deal with strong nonlinear
boundaries. In this section, some numerical examples will be demonstrated to verify the accuracy of the proposed method.

4.1. Complex excitation with non-homogenous and nonlinear boundaries

Section 3 has introduced the convergence criterion of the analytical method. To validate the accuracy of the convergence,
the differential-quadrature-element-method (DQEM) is introduced. It can satisfy any kinds of boundary conditions of an
Euler beam model perfectly. Usually, the computation could be convergent while the node points exceeds 5 for a straight
uniform beam [56,57]. In the current work, 13 nodes are employed together with the Dolph-Chebyshev distribution to en-
hance the convergence [31]. In Fig. 4, the response of the middle point under a uniform excitation with the frequency 6 rad/s
is demonstrated based on this simulation method. The amplitude in the steady-state region, which is given in Fig. 4 (b), is
used to verify the analytical method.
Based on the simulation, the accuracy of the proposed method can be verified in a wide frequency region, which is given
in Fig. 5. The modal projection here reaches to the 3rd order. In Fig. 5(a), the fork and the circle are simulations via DQEM.
The solid line is on behalf of the amplitude-frequency response curve of the 3rd -order expansion. The dashed line stands for
that of the 5th -order expansion. The tiny error between them is demonstrated in the right sub-figure. It clarifies that the 3rd
order expansion already achieves a good convergence. The left sub-figure shows the accuracy of the proposed method in the
resonant region. The jumping phenomenon occurs here. In Fig. 5(b) and (c), the stability of the jumping point is discussed
based on the 3rd order expansion and the 5th order expansion respectively. They obtain the same conclusion.
As the typical example, Fig. 5 has provided the response of the middle point. In Fig. 6, the responses at the 1/4 point
and the 3/4 point are demonstrated. They stand for that the nonlinearity in the boundary arouses the nonlinear response
along the whole pipe. Besides, the comparison between the 3rd order expansion and the 5th order expansion states that the
3rd order expansion is convergent for the response along the whole pipe.
It is clear that the proposed method can deal with the uniform harmonic excitation under nonlinear and non-
homogenous boundaries convergently and accurately. It can also deal with a more complex excitation. As mentioned in
[58], the forced vibration combined with the parametric resonance could produce a twin resonant peak. To validate the ca-
pacity of the proposed method, the complicated parametric excitation here contains two components. They are the pulsating

7
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 3. Discussion of the convergence of the harmonic expansion via higher order results: (a) the coefficients of the 1st -order harmonic on the 1st mode;
(b) coefficients of the 3rd -order harmonic on the 1st mode; (c) the coefficients of the 1st -order harmonic on the 2nd mode; (d) coefficients of the 3rd -order
harmonic on the 2nd mode; (e) the coefficients of the 1st -order harmonic on the left boundary coordinate β 0 ; (f) coefficients of the 3rd -order harmonic
on the left boundary coordinate β 0 ; (g) the coefficients of the 1st -order harmonic on the right boundary coordinate β 1 ; (h) coefficients of the 3rd -order
harmonic on the right boundary coordinate β 1
8
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 4. Time domain response of the middle point with the excitation frequency 6 rad/s produced based on DQEM: (a) time domain response of the middle
point; (b) steady-state response of the middle point

Fig. 5. Response of the middle point subjected to a forced vibration in a wide frequency region: (a) convergence and accuracy of the response of the
middle point; (b) stability of the response based on the 3rd order expansion; (c) stability of the response based on the 5th order expansion

of the fluid speed and the pulsating of the tension. Thus, the governing equation here is:
 2

u,tt + 2Mr [v0 + v1 sin (2 t )]u,xt + [v0 + v1 sin (2 t )] − [P0 + P1 sin (2 t )] u,xx
(40)
+u,xxxx + ζ u,xxxxt − b sin ( t ) = 0

where v1 is the pulsating speed and P1 is the pulsating tension. With parameters given in Fig. 7, the response curve is ob-
tained. Those forks here denote the simulation while the solid line stands for the result via the 3rd order expansion. The 5th
order expansion result is presented by the dashed line. In general, the 3rd order expansion already has a reasonable accu-
racy. It is clear that the proposed method has the capacity of dealing with the complex excitation based on an appropriate
modal projection together with the appropriate harmonic expansion.

9
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 6. Responses along the pipe: (a) response at the 1/4 point; (b) response at the 3/4 point

Fig. 7. Primary resonance subjected to a forced vibration combined with a parametric resonance

4.2. Strong damping in the boundary

As can be observed from Eqs. (38) and (39), the boundary damping is involved in the frequency equation. Based on it,
Fig. 8 shows the influence of the boundary damping to the natural frequency. The natural frequency changes little at first.
But after a critical damping, the natural frequency changes rapidly to another constant. The first-two figures in Figure 8 show
the first-two natural frequencies via different modal projections. The frequency after the critical damping needs a projection
with at least two higher orders to obtain a convergent value. The last two figures demonstrate the natural frequency chang-
ing with both of the fluid speed and the boundary damping in 3-D maps. It is clear that the natural frequency is always
constant in the sub-critical damping region.
In Fig. 9, free vibrations with two damping coefficients in the subcritical region are compared via the DQEM. Time domain
responses are compared in Fig. 9(a) while Fig. 9(b) demonstrates the fast Fourier transform (FFT) of them. Obviously, the
vibration declines more rapidly with the bigger damping while the frequency of it keeps the same as that of the smaller
damping. It is meaningful for the vibration control.

10
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 8. Natural frequencies changing with the boundary damping: (a) the 1st -order natural frequency via different orders of modal projection; (b) the
2nd -order natural frequency via different orders of modal projection; (c) the 1st -order natural frequency changing with both of v and d; (d) the 2nd -order
natural frequency changing with both of v and d

Fig. 9. Verification of natural frequencies in the subcritical region: (a) time domain response of the middle point; (b) FFT of the response of the middle
point

In the super-critical damping region, the proposed method still has a reasonable accuracy. Fig. 10 shows the response
curve subjected to a strong damping in the super-critical region based on the 3rd order projection. The solid line stands for
the analytical result while the forks denote those simulations. It is clear that the analytical method obtains a reasonable
accuracy in a wide frequency region.

4.3. Strong nonlinearity in the boundary

The proposed method has shown its capacity of dealing with the complex excitation and the strong damped boundary.
In fact, as the proposed method has the advantage of the harmonic balance method, it also can deal with the strong non-

11
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 10. Verification of the proposed method in the supercritical damping region

Fig. 11. Super-harmonic resonance of the forced vibration via the proposed method: (a) response curve of three sample points; (b) stability analysis and
the accuracy of the analytical result

linear boundary. Fig. 11 shows the primary resonance subjected to the strong nonlinear boundary, which is demonstrated
in Fig. 11(a). The analytical response curve is produced based on the 3rd order projection together with the 5th order ex-
pansion. What can be observed from this figure is that the response curve is upturned while the response curve in Fig. 5 is
downturned. Besides, the strong nonlinearity arouses the obvious super-harmonic resonance in the low-frequency region. In
Fig. 11(b), the accuracy of the analytical result is clarified by the simulation of the middle point. At the two jumping points,
the saddle-node bifurcation occurs.

5. Comparison with other methods

The foregoing works have shown the convergence and the capacity of the proposed method for the pipe conveying fluid
with nonlinear and non-homogenous boundaries. To outstand its advantage of dealing with the strong boundary, it will be
compared with the Dirac operator and the multiscale method in this section. The comparison with the Dirac operator is
to outstand the advantage of the proposed method in the spatial projection. The comparison with the multiscale method
outstands the advantage of the convergence of the time-varying response.

5.1. Comparison with the Dirac operator based on HBM

Those nonlinear and non-homogenous terms in the boundary can be introduced into the governing equation as con-
centrated activities by the Dirac operator [59-61]. It is a convenient way of dealing with these uncommon boundary value
problems. The biggest difference between the Dirac operator and the proposed method is that the Dirac just employs the
modal projection, or namely the modal expansion. The converted form of the system given in Section 3 via Dirac operator
is
 1    
u,tt + 2Mr vu,xt + v2 − P0 u,xx + u,xxxx + ζ u,xxxxt − b sin ( t ) ϕi dx
0
 1     
+ δ (x − 0 ) r3 uL ,x 3 + duL ,xt +δ (x − 1 ) r3 uR ,x 3 + duR ,xt ϕi ,x dx = 0 (41)
0

12
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

Fig. 12. Advantage of the proposed method comparing with the Dirac operator: (a) response subjected to relatively weak parameters; (b) response sub-
jected to the strongly damped boundary; (c) response subjected to the strong nonlinear boundary

where ϕ i is given by Eq. (35) based on the vestigial boundary as following

uL = 0, uR = 0, uL ,xx = 0, uR ,xx = 0 (42)
Fig. 12 shows the results via the proposed method, the Dirac operator and the DQEM. All the calculations are carried
out based on the same modal projection and the same harmonic expansion. It is clear that the result via Dirac operator in
Fig. 12(a) has a reasonable accuracy based on those parameters in subsection 4.1. However, while the Dirac operator faces
with that strongly damped boundary in subsection 4.2 and the strong nonlinear boundary in subsection 4.3, it will lose the
accuracy obviously. The corresponding comparisons are given in Fig. 12(b) and (c) respectively. Especially, the comparison in
Fig. 12(b) indicates that the Dirac operator has the qualitative imperfection while facing with the strong damped boundary.
This clarifies that the modal correction of the proposed method ensures the convergence of the spatial expansion on the
other hand.
Although the Dirac operator has imperfections in dealing with strong boundaries, it still has advantages in the com-
putational efficiency. As can be observed in Eq. (36), four generalized variables are introduced by the proposed method.
Expanded by HBM, more coefficients will be involved into the calculation.

5.2. Comparison with the multiscale method based on modal correction

The last subsection has shown the advantage of the proposed method in dealing with the spatial expansion. As it is
based on HBM, it has the advantage in dealing with the strong nonlinear boundary. The multiscale method is also a widely
used method for the nonlinear problem. Taking the system demonstrated by Eqs. (33) and (34) as the example, the rescaled
systems are the one on T0 scale
 
D20 u0 + 2Mr vD0 u0 ,x + v2 − P0 u0 ,xx + u0 ,xxxx = 0 (43)
with the boundary
uL0 = 0, uR0 = 0,
(44)
uL0 ,xx = 0, uR0 ,xx = 0
and the one on T1 scale
 
D21 u1 + 2Mr vD1 u1 ,x + v2 − P0 u1 ,xx + u1 ,xxxx = b sin ( t ) − ζ D0 u0 ,xxxx − 2D0 D1 u0 − 2Mr vD1 u0 ,x (45)

13
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

with the boundary

uL1 = 0, uR1 = 0,
uL1 ,xx = r3 uL0 ,x 3 + D0 duL0 ,x , (46)
uR1 ,xx = −r3 uR0 ,x 3 − D0 duR0 ,x
At this step, the original system is converted into a non-homogenous but linear system. Based on the modal correction, an
accurate solution will be obtained. What should be mentioned is that, the so called accurate solution is just real for the
rescaled system. However, the solution of the initial system is
u = u0 + ε u1 (47)
where ε is a symbol that indicates u1 is tiny. The accuracy of the multiscale method is based on this assumption. In other
words, the solution will lose accuracy while u1 is considerable. This can be verified by substituting the solution Eq. (47) into
one of the nonlinear boundary as a typical example, which means the solution must satisfy
uL0 ,xx +ε uL1 ,xx = r3 (uL0 ,x +ε uL1 ,x )3 +d (uL0 ,xt +ε uL1 ,x ) (48)
However, based on Eqs. (44) and (46), the solution just satisfies
uL0 ,xx +ε uL1 ,xx = r3 uL0 ,x 3 +duL0 ,xt (49)
Hence, the multiscale method has a reasonable accuracy while the boundary is not too strong. Otherwise, a complicated
iteration must be introduced into the processing of the multiscale method to improve the accuracy [29]. Besides, the rescaled
rules of the multiscale method are different while facing with different resonances. For example, if the super-harmonic
resonance is concerned, the excitation must occur in Eq. (43). But if the resonance on the natural mode is concerned, it
must occur in Eq. (45).
On the other hand, the multiscale method for the nonlinear boundary value problem also has advantages. The first one
is that the solvability condition could reveal the resonance condition clearly. The second one is that the multiscale method
captures the soul of the resonance skillfully. Based on it, the bifurcation and the stability of the system can be studied con-
veniently. By contrast, the proposed method is a little crude as it always studies the dynamics by solving out all coefficients.

6. Conclusions

The work introduced above showed an approximate method for the gyroscopic system with nonlinear and non-
homogenous boundaries. To get a reasonably accurate result, both of the spatial convergence and the harmonic convergence
were discussed. Based on the work above, some conclusions are given as follows:
1. As the response functions have no concern with the natural frequency, the proposed method can follow a multi-
resonance curve in a wide region. The outstanding advantage of the proposed method is that it can follow the strong
boundary.
2. The modal expansion can use modes of a non-traveling Euler beam in transverse vibration with the linear and homoge-
nous part of the initial boundary condition. To obtain a convergent modal expansion, at least one more mode should
be considered into the modal projection. The convergence could be determined by the natural frequency which is con-
cerned.
3. The nonlinear and non-homogenous boundary condition is treated as a set of generalized governing conditions. Thus,
just like the harmonic balance method (HBM), the convergence of the harmonic can be judged by the result with more
harmonics. A good balance between the efficiency and the accuracy must be considered.
Besides, the result in Appendix shows the capacity of the proposed method in dealing with strong damping in the gov-
erning equation, or namely, a strong viscoelastic pipe.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have
appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Xiao-Ye Mao: Investigation, Writing – original draft. Song Shu: Writing – review & editing. Xin Fan: Writing – review &
editing. Hu Ding: Conceptualization, Supervision. Li-Qun Chen: Validation.

Acknowledgments

The authors gratefully acknowledge the support of the National Natural Science Foundation of China (Nos. 11872159,
12002195), the National Science Fund for Distinguished Young Scholars (No. 12025204), the Program of Shanghai Municipal
Education Commission (Grant No. 2019-01-07-0 0-09-E0 0 018) and the Pujiang Project of Shanghai Science and Technology
Commission (Grant No. 20PJ14040 0 0).

14
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

APPENDIX (STRONG DAMPING IN GOVERNING EQUATION)

Subsection 4.2 has discussed the accuracy of the current method facing with the strong damped boundary. Another
interesting question is that whether the current method is still accurate while facing with the strong damped structure.
As known, the strong damping will produce a considerable dynamic stiffness. Consequently, one can forecast that the pipe
will almost be a rigid body while vibrating, especially in the high frequency region. Similar with Subsection 4.2, the natural
frequency changing with the viscoelastic damping in the governing equation is demonstrated in Fig. 13. It is obvious that
the real part of the complex natural frequency becomes zero after a critical value. Higher order projections will provide
more accurate results.

Fig. 13. Influence of the structural damping to the natural frequency: (a) real part of the complex frequency; (b) imaginary part of the complex frequency

Fig. 14. Response of the middle point of the over damped pipe: (a) time-domain response under the excitation frequency 10 via the simulation; (b)
time-domain response under the excitation frequency 20 via the simulation; (c) comparison between the simulation and the proposed method

15
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

To illustrate the capacity of the current method while facing with the strong viscoelastic damping, a very strong value is
set as 1, which is in the over damped region. Fig. 14(a) and (b) show the time-domain response by DQEM under different
excitation frequencies. The initial value is given in each subfigure. It is clear that the response here is very different with
that in Fig. 4. Although initial values are all zero as those in Fig. 4, the response here still shows a degenerative center line.
The main purpose here is the discussion of the capacity of the current method under strong viscoelastic damping.
Fig. 14(c) shows the comparison between the current method and the simulation. The solid line stands for the result given
by the current method while forks are simulations. The star locating at the 0 frequency is the static displacement suffering
from the uniform constant force, or namely the harmonic excitation has an infinitely long period. There comes a conclusion
that the current method is still accurate even when the pipe is over damped.

References

[1] H. Ashley, G. Haviland, Bending vibrations of a pipe line containing flowing fluid, J. Appl. Mech.-T. ASME 17 (1950) 229–232.
[2] V.P. Feodos’Ev, Vibrations and stability of a pipe when liquid flows through it, Inzhenernyi Sbornik 10 (1951) 169–170.
[3] R.H. Long, Experimental and theoretical study of transverse vibration of a tube containing flowing fluid, J. Appl. Mech.-T. ASME 22 (1955) 65–68.
[4] T.B. Benjamin, Dynamics of a system of articulated pipes conveying fluid .1. theory, in: Proceedings of the Royal Society of London Series a-Mathemat-
ical and Physical Sciences, 261, 1961 457-+.
[5] T.B. Benjamin, Dynamics of a system of articulated pipes conveying fluid .2. experiments, Proceedings of the Royal Society of London Series a-Mathe-
matical and Physical Sciences 261 (1961) 487-+.
[6] R.W. Gregory, M.P. Païdoussis, Unstable oscillation of tubular cantilevers conveying fluid II. Experiments, Proc. R. Soc. Lond. A 293 (1966) 528–542.
[7] R.W. Gregory, M.P. Païdoussis, Unstable oscillation of tubular cantilevers conveying fluid I. Theory, Proc. R. Soc. Lond. A 293 (1966) 528–542.
[8] M.P. Païdoussis, N.T. Issid, Dynamic stability of pipes conveying fluid, J. Sound. Vib. 33 (1974) 267–294.
[9] M.P. Païdoussis, G.X. Li, Pipes conveying fluid-a model dynamic problem, J. Fluid Struct. 7 (1993) 137–204.
[10] G.X. Li, M.P. Païdoussis, Stability, double degeneracy and chaos in cantilevered pipes conveying fluid, Int. J. Nonlin. Mech. 29 (1994) 83–107.
[11] M.P. Païdoussis, C. Semler, Nonlinear dynamics of a fluid-conveying cantilevered pipe with an intermediate spring support, J. Fluid Struct. 7 (1993)
269–298.
[12] M.P. Païdoussis, C. Sundararajan, Parametric and combination resonances of a pipe conveying pulsating fluid, J. Appl. Mech.-T. ASME 42 (1975) 780–784.
[13] S.T. Ariaratnam, N.S. Namachchivaya, Dynamic stability of pipes conveying pulsating fluid, J. Sound. Vib. 107 (1986) 215–230.
[14] H.R. Oz, H. Boyaci, Transverse vibrations of tensioned pipes conveying fluid with time-dependent velocity, J. Sound. Vib. 236 (20 0 0) 259–276.
[15] L.N. Panda, R.C. Kar, Nonlinear dynamics of a pipe conveying pulsating fluid with combination, principal parametric and internal resonances, J. Sound.
Vib. 309 (2008) 375–406.
[16] G.H. Chang, Y. Modarres-Sadeghi, Flow-induced oscillations of a cantilevered pipe conveying fluid with base excitation, J. Sound. Vib. 333 (2014)
4265–4280.
[17] L.N. Panda, R.C. Kar, Nonlinear dynamics of a pipe conveying pulsating fluid with parametric and internal resonances, Nonlinear Dynam 49 (2007)
9–30.
[18] X.Y. Mao, H. Ding, L.Q. Chen, Steady-state response of a fluid-conveying pipe with 3:1 internal resonance in supercritical regime, Nonlinear Dynam 86
(2016) 795–809.
[19] J. Xu, Q.-b. Yang, Flow-induced internal resonances and mode exchange in horizontal cantilevered pipe conveying fluid (II), Appl. Math. Mech.-Engl.
27 (2006) 943–951.
[20] J. Xu, Q.-b. Yang, Flow-induced internal resonances and mode exchange in horizontal cantilevered pipe conveying fluid (I), Appl. Math. Mech.-Engl. 27
(2006) 935–941.
[21] M.P. Païdoussis, C. Semler, M. Wadham-Gagnon, S. Saaid, Dynamics of cantilevered pipes conveying fluid. Part 2: Dynamics of the system with inter-
mediate spring support, J. Fluid Struct. 23 (2007) 569–587.
[22] M.H. Ghayesh, M.P. Païdoussis, Y. Modarres-Sadeghi, Three-dimensional dynamics of a fluid-conveying cantilevered pipe fitted with an additional
spring-support and an end-mass, J. Sound. Vib. 330 (2011) 2869–2899.
[23] K. Hu, Y.K. Wang, H.L. Dai, L. Wang, Q. Qian, Nonlinear and chaotic vibrations of cantilevered micropipes conveying fluid based on modified couple
stress theory, Int. J. Eng. Sci. 105 (2016) 93–107.
[24] S. Abbasbandy, E. Shivanian, Prediction of multiplicity of solutions of nonlinear boundary value problems: Novel application of homotopy analysis
method, Commun. Nonlinear. Sci. 15 (2010) 3830–3846.
[25] Y.Q. Hasan, L.M. Zhu, Solving singular boundary value problems of higher-order ordinary differential equations by modified Adomian decomposition
method, Commun. Nonlinear. Sci. 14 (2009) 2592–2596.
[26] X.J. Liu, Y.H. Zhou, X.M. Wang, J.Z. Wang, A wavelet method for solving a class of nonlinear boundary value problems, Commun. Nonlinear. Sci. 18
(2013) 1939–1948.
[27] A. Di Matteo, A. Pirrotta, Generalized differential transform method for nonlinear boundary value problem of fractional order, Commun. Nonlinear. Sci.
29 (2015) 88–101.
[28] X.-Y. Mao, J.-Q. Sun, H. Ding, L.-Q. Chen, An approximate method for one-dimensional structures with strong nonlinear and nonhomogenous boundary
conditions, J. Sound. Vib. 469 (2020) 115128.
[29] X.-Y. Mao, H. Ding, L.-Q. Chen, Passive Isolation by Nonlinear Boundaries for Flexible Structures, J. Vib. Acoust. 141 (2019) 051003.
[30] X.Y. Mao, H. Ding, L.Q. Chen, Nonlinear Torsional Vibration Absorber for Flexible Structures, J. Appl. Mech.-T. ASME 86 (2019) 021006.
[31] X.Y. Mao, H. Ding, L.Q. Chen, Vibration of Flexible Structures Under Nonlinear Boundary Conditions, J. Appl. Mech.-T. ASME 84 (2017) 111006.
[32] T.D. Guo, G. Rega, Solvability conditions in multi-scale dynamic analysis of one-dimensional structures with non-homogeneous boundaries: A general
operator formulation, Int. J. Nonlin. Mech. 115 (2019) 68–75.
[33] L. Xu, M.W. Lu, Q. Cao, Nonlinear vibrations of dynamical systems with a general form of piecewise-linear viscous damping by incremental harmonic
balance method, Physics Letters A 301 (2002) 65–73.
[34] S. Lau, Y. Cheung, Amplitude Incremental Variational Principle for Nonlinear Vibration of Elastic Systems, J. Appl. Mech.-T. ASME (1981) 48.
[35] K.Y. Sze, S.H. Chen, J.L. Huang, The incremental harmonic balance method for nonlinear vibration of axially moving beams, J. Sound. Vib. 281 (2005)
611–626.
[36] V. Jaumouille, J.J. Sinou, B. Petitjean, An adaptive harmonic balance method for predicting the nonlinear dynamic responses of mechanical system-
s-Application to bolted structures, J. Sound. Vib. 329 (2010) 4048–4067.
[37] A. Leung, Z. Guo, Feed forward residue harmonic balance method for a quadratic nonlinear oscillator, Int. J. Bifurcat. Chaos. 21 (2011) 1783–1794.
[38] M. Akbarzade, A. Farshidianfar, Nonlinear transversely vibrating beams by the improved energy balance method and the global residue harmonic
balance method, Appl. Math. Model. 45 (2017) 393–404.
[39] H.H. Khodaparast, H. Madinei, M.I. Friswell, S. Adhikari, S. Coggon, J.E. Cooper, An extended harmonic balance method based on incremental nonlinear
control parameters, Mech. Syst. Signal Pr. 85 (2017) 716–729.
[40] M.S. Rahman, Y.-Y. Lee, New modified multi-level residue harmonic balance method for solving nonlinearly vibrating double-beam problem, J. Sound.
Vib. 406 (2017) 295–327.

16
X.-Y. Mao, S. Shu, X. Fan et al. Journal of Sound and Vibration 505 (2021) 116157

[41] X.F. Wang, W.D. Zhu, Dynamic Analysis of an Automotive Belt-Drive System With a Noncircular Sprocket by a Modified Incremental Harmonic Balance
Method, J. Vib. Acoust. (2017) 139.
[42] J. Niu, Y. Shen, S. Yang, S. Li, Higher-order approximate steady-state solutions for strongly nonlinear systems by the improved incremental harmonic
balance method, J. Vib. Control 24 (2018) 3744–3757.
[43] H. Liao, The application of reduced space harmonic balance method for the nonlinear vibration problem in rotor dynamics, Mech. Based Des. Struc.
47 (2019) 154–174.
[44] O. Sert, E. Cigeroglu, A novel two-step pseudo-response based adaptive harmonic balance method for dynamic analysis of nonlinear structures, Mech.
Syst. Signal Pr. 130 (2019) 610–631.
[45] R. Ju, W. Fan, W. Zhu, An Efficient Galerkin Averaging-Incremental Harmonic Balance Method Based on the Fast Fourier Transform and Tensor Con-
traction, J. Vib. Acoust. (2020) 142.
[46] Y.B. Kim, S.K. Choi, A multiple harmonic balance method for the internal resonant vibration of a non-linear Jeffcott rotor, J. Sound. Vib. 208 (1997)
745–761.
[47] W.J. Kim, N.C. Perkins, Harmonic balance/Galerkin method for non-smooth dynamic systems, J. Sound. Vib. 261 (2003) 213–224.
[48] A. Al-shyyab, A. Kahraman, Non-linear dynamic analysis of a multi-mesh gear train using multi-term harmonic balance method: sub-harmonic mo-
tions, J. Sound. Vib. 279 (2005) 417–451.
[49] N. Coudeyras, J.J. Sinou, S. Nacivet, A new treatment for predicting the self-excited vibrations of nonlinear systems with frictional interfaces: The
Constrained Harmonic Balance Method, with application to disc brake squeal, J. Sound. Vib. 319 (2009) 1175–1199.
[50] A.Y.T. Leung, Z. Guo, T.C. Fung, The multi-parameter homotopy harmonic balance method for steady state problems, Int. J. Comput. Math. 87 (2010)
1158–1177.
[51] A.C.J. Luo, J. Huang, Analytical dynamics of period-m flows and chaos in nonlinear systems, Int. J. Bifurcat. Chaos. (2012) 22.
[52] S. Zucca, C.M. Firrone, Nonlinear dynamics of mechanical systems with friction contacts: Coupled static and dynamic Multi-Harmonic Balance Method
and multiple solutions, J. Sound. Vib. 333 (2014) 916–926.
[53] F. Schreyer, R.I. Leine, A mixed shooting - harmonic balance method for unilaterally constrained mechanical systems, Archive of Mechanical Engineer-
ing 63 (2016) 297–314.
[54] J.L. Huang, W.D. Zhu, A New Incremental Harmonic Balance Method With Two Time Scales for Quasi-Periodic Motions of an Axially Moving Beam
With Internal Resonance Under Single-Tone External Excitation, J. Vib. Acoust. (2017) 139.
[55] A.C.J. Luo, J.Z. Huang, Approximate solutions of periodic motions in nonlinear systems via a generalized harmonic balance, J. Vib. Control 18 (2012)
1661–1674.
[56] X.W. Wang, Y.L. Wang, Free vibration analysis of multiple-stepped beams by the differential quadrature element method, Appl. Math. Comput. 219
(2013) 5802–5810.
[57] Y.L. Wang, X.W. Wang, Y. Zhou, Static and free vibration analyses of rectangular plates by the new version of the differential quadrature element
method, Int. J. Numer. Meth. Eng. 59 (2004) 1207–1226.
[58] X.Y. Mao, H. Ding, L.Q. Chen, Dynamics of a super-critically axially moving beam with parametric and forced resonance, Nonlinear Dynam 89 (2017)
1475–1487.
[59] H. Ding, Z.-Q. Lu, L.-Q. Chen, Nonlinear isolation of transverse vibration of pre-pressure beams, J. Sound. Vib. 442 (2019) 738–751.
[60] H. Ding, J. Ji, L.-Q. Chen, Nonlinear vibration isolation for fluid-conveying pipes using quasi-zero stiffness characteristics, Mech. Syst. Signal Pr. 121
(2019) 675–688.
[61] H. Ding, L.-Q. Chen, Nonlinear vibration of a slightly curved beam with quasi-zero-stiffness isolators, Nonlinear Dynam 95 (2019) 2367–2382.

17