Available online at www.sciencedirect.

com

Procedia Engineering 14 (2011) 2986–2993

The Twelfth East Asia-Pacific Conference on Structural Engineering and Construction

Flow-Induced Vibration of Pipeline on Elastic Support

a
R. T. FAAL, bD. DERAKHSHAN
FACULTY OF ENGINEERING, ZANJAN UNIVERSITY, ZANJAN, IRAN

Abstract

We study the self-excited flexural vibration of a pipe due to internal flow which is entirely supported on elastic
foundation. We consider a finite length pipe with clamped ends made of isotropic material consist of an internal flow
with constant velocity. The Euler-Bernoulli model of beam is considered to analysis the vibratory behavior of pipe.
The governing equation of motion is then derived which is partial differential equation in terms of derivative of
transverse displacement of pipe with respect to time and axial distance. The full term governing equation of motion
contains the second derivative of transverse displacement with respect to time and axial distance which has been
neglected in earlier work is considered here and solved analytically. The natural frequencies of coupled pipe-fluid
system are then obtained and the effects of the neglected term on the natural frequencies of system are studied. The
effects of the stiffness of elastic foundation, velocity and density of inner fluid, inner diameter of pipe with constant
thickness, elasticity modulus of the pipe and finally pipe length are studied. The stability analysis of the vibration is
also accomplished.

© 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.

Selection and/or peer-review under responsibility of [name organizer]
Keywords: Flow-induced Vibration, elastic foundation, flexural vibration

1 INTRODUCTION

Concern for the flow-induced vibration of the tube has become a serious consideration in the design of
shell-and-tube equipment. All tubes vibrate in all flow conditions! However, we are concerned with the
vibrations that cause significant damage to the pipe. Structural failure due to flow-induced vibration is a
common problem of heat exchangers which affects on the reliability and performance of them. Also flow-
induced vibrations can damage the tubes in evaporators. On the other hand because of increasing urban

a
Corresponding author: Email: faal92@yahoo.com
b
Presenter: faal92@yahoo.com

1877–7058 © 2011 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2011.07.376
 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993 2987

population, under soft ground piping is an increasingly common activity to manage oil or water
transportation facilities in many large cities around the world. In this regards determination of flow-
induced lateral vibration of buried pipelines due to an internal flow become a subject of considerable
importance in this field. Literature is replete with studies related to flow-induced vibrations in the last
four decades. The first attempt in flow-induced vibration study was done by (Ashley and Haviland 1950).
The practical vibration problems of heat exchanger tubes in power plants and nuclear reactors were the
subject of many studies as other aspects of this field. The finite elements models and also other numerical
methods are frequently used for these studies. For want of space, we review the studies which are only
based on the analytical methods. Most of the earlier studies are related to investigations about inner
transient flow with time-varying velocity, which are reviewed at first here.
Coupled pipe-dynamics equations for the axial, radial, and lateral vibrations of the pipeline as well as
for the transients of unsteady internal fluid pressure and velocity were the subject of studies by (Lee and
Kim 1999) and (Lee et al. 1995). Afterward, (Gorman et al. 2000) included the effects of radial shell
vibration and initial axial tension on Lee’s pipe-dynamics model.
A spectral element model is developed by (Lee and Park 2006) for the uniform straight pipelines with
an internal unsteady fluid. Four coupled pipe-dynamics equations are derived in terms of the transverse
displacement, the axial displacement, the fluid pressure and velocity and then linearized them. The fast
Fourier transform FFT is used to investigate the structural dynamic characteristics and the internal fluid
transient properties.
The wave characteristics, divergence stability and dynamics of the viscoelastic pipelines conveying
internal flow are examined by (Lee et al. 2009) using the spectral element method. Small perturbations
with respect to the steady state values of inner fluid velocity and pressure is considered to make the
governing equations to be linear. In the spectral element model the governing differential equation of
motion is transformed into the frequency-domain by using the discrete Fourier transformation theory. The
internal flow velocity at which the divergence instability occurs is derived in an analytical form.
In second part of review we consider steady-state flow-induced vibration problem in pipe conveying
an internal flow with constant velocity. The self-excited flexural vibration of a pipe due to an internal
flow with constant velocity was investigated by (Biswas and Ahmed 2001). In the above-mentioned paper
the second order derivatives of transverse displacement with respect to time and axial distance (product
term) is neglected in the governing equation. The critical value of fluid velocity as a stability boundary is
evaluated in terms of the system parameters such as, pipe material density and rigidity and fluid density,
and so on. The optimal flow velocity for minimum vibration was obtained.
In this study the full term governing equation of motion is solved analytically i.e. the first derivative of
transverse displacement with respect to time and axial distance has not been neglected and the
weaknesses of the earlier work is studied. To derive this governing equation, an infinitesimal element of
inner fluid of pipe is chosen which is imposed to horizontal movement with constant velocity and lateral
movement due to pipe flexural vibration. Therefore the transverse acceleration of this element may be
evaluated using concept of total derivative or material derivative of transverse displacement of the pipe
which contains time derivative and convective inertia term. Consequently, the distributed force per unit
length of the pipe is the summation of inertia force of the fluid, the inertia force of the pipe and also the
distributed force of the elastic foundation. Applying the boundary conditions, the natural frequencies of
coupled pipe-fluid system are then obtained. Analysis of the effects of stiffness of the elastic foundation
and also the complete governing equation on the natural frequencies of the coupled system and also
analytic solution of problem may be the novelty of the present work.
2988 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993

2 DERIVATION OF GOVERNING EQUATION OF PIPELINE ON ELASTIC SUPPORT

We consider an elastic straight pipe with length " and two clamped ends which is totally bonded on
elastic support with the stiffness per unit length k . The bending rigidity and the inner surface of pipe
are EI and A, respectively. The pipe contains an incompressible fluid with density U f and axial
velocity U . Each element of inner fluid undergoes two movements consist of the horizontal movement
with the velocity U and the vertical displacement Y ( X , t ) (where X is the axial distance from one end)
resulting of the pipe flexural vibration, therefore the acceleration of this element may be easily written as
w w
a( X , t ) (U  )2Y ( X , t ) (1)
wX wt
where U w   wX  w   wt is the total derivative or material derivative operator. Hence the flow
induced vertical force on the unit length of pipe is F ( X , t ) U f Aa ( X , t ). Considering the Euler-Bernoulli
beam theory; the governing equation of flexural vibration of pipe in view of Eq. (1) and
relation F ( X , t ) U f Aa ( X , t ) is attained as

w 4Y w 2Y w w (2)
EI m p  U f A(U  ) 2 Y  kY
wX 4 wt 2
wX wt
where m p is the mass per unit length of the pipe. The above equation is simplified as

w 2Y wY w 2Y w 4Y
(m p  U f A)  2 U f AU  U f AU 2  EI  kY 0 (3)
wt 2
wXwt wX 2
wX 4
The transformation x X / ", y Y / " and W t / a are used to normalize the above equation in
convenient form (dimensionless form) where a " 2 m p / EI is a parameter with the dimension of the
time and all others are dimensionless. Making use of above transformation, the Eq. (3) may be rewritten
as follows

w2 y wy w2 y w4 y
 2 (1  E )u  (1  E )u 2 2  E 4  bEy 0 (4)
wW 2 wxwW wx wx
where b k" 4 / EI , u aU / " and E m p /( m p  U f A) are dimensionless parameters. The ends
boundary conditions imply that
y( 0,IJ )
0 ,y(",IJ ) 0
wy wy (5)
( 0,IJ ) 0, (",IJ ) 0
wx wx
There is a conventional harmonic solution of Eq. (3) in the form of Y ( X , t ) Y0 ( X )e i (Zt  rX ) which is
usual solution in the wave motion problems which Z is the circular frequency and is r a real constant.
Consequently the normalized solution of Eq. (4) may be considered in the form of y ( x, W ) y 0 ( x)e i (DW Gx )
where D aZ and G r". Substituting this solution form into Eq. (4) and setting two the real and
imaginary parts of the ensuing equation equal to zero results in

E y0( 4) ( x)  [(1  E )u 2  6EG 2 ] y0cc ( x)  [ EG 4  G 2 (1  E )u 2  D 2  bE  2DG u (1  E )] y0 ( x) 0

(6)
2EGy0ccc( x)  [u (1  E )(D  Gu )  2G 3 E ] y0c ( x) 0
 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993 2989

The characteristic equations of above linear ordinary differential equations in terms of s are

E s 4  [(1  E )u 2  6 EG 2 ]s 2  [ EG 4  G 2 (1  E )u 2  D 2  bE  2DG u (1  E )] 0
(7)
2 EGs 3  [u (1  E )(D  Gu )  2G 3 E ]s 0

The condition for the existence of the solution for the system of ordinary differential equations (6) is
the existence of identical roots for two algebraic equations (7). From the second equation of (7) we
obtain s 0 and s 2 G 2  u (1  E )(D  Gu ) / 2EG , where s 0 leads to the constant y 0 ( x) which doesn’t
satisfy the boundary conditions (5). Substituting s 2 G 2  u (1  E )(D  Gu ) / 2EG into the first equation
of (7) leads to

4G 6  4KG 4  (K 2  N )G 2  K 2J 2 0 (8)

where K u 2 ( E  1) / 2 E , J D / u and N (D 2 / E )  b. The above sixth order of algebraic equation

may be changed to a cubic equation. Making use of the available analytical solution of the cubic equation
(Zwillinger 2003), the roots of Eq. (8) are attained as
K
(G1, 2 ) 2 3
0.5(  q  ' )  3 0.5(  q  ' ) 
3
2Si 4Si
K
(G 3, 4 ) 2 e 3 3 0.5( q  ' ) e 3 3 0.5( q  ' )  (9)
3
4Si 2Si
K
(G 5,6 ) 2 e 3 3 0.5(q  ' )  e 3 3 0.5(q  ' ) 
3
3 2
   
where ' 4 p  q and p N  K 2 3 12 and q  K 3 9  NK 12  K 2J 2 4 . If discriminant ', is
positive, two roots of 6 roots are real and the others are complex. If ' 0, then the squared of all roots
are real, which at least two of them are equal. If '  0, then the squared of all roots are unequal and real.
The discriminant may be rewritten as:

' (1 / 8)[(1 / 27)K 5  (1 / 2)[(1 / 27)N  J 4 ]K 4  (1 / 3)J 2NK 3  (1 / 27)N 2K 2  (1 / 54)N 3 ] (10)

It is worth to mention it that there are critical values for velocity which the system should be unstable
under those velocities. In this case the negative values for the frequencies of coupled system are gained.
The second equation of Eqs. (7), results in s 2  O2 0 where O 2 u (1  E )(D  G u )  2 EG   G 2 . In the
case of ' ! 0, substituting G1 and G 2 from Eqs. (9), in view of this fact that G 2 G1 , into the relation
for O2 gives the parameters O12 and O22 . Therefore the solution of system of partial differential equations (6)
is readily written and y ( x,W ) is obtained as follows

y ( x) [(C1 cos O1 x  C 2 sin O1 x)e iG1 x  (C 3 cos Ȝ 2 x  C 4 sin Ȝ2 x)e iG 2 x ]e iDW , if O12 ! 0 and O 22 ! 0 (11)
2990 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993

In the case of that O12 or O22 to be negative they will replace by O12  P12 or O 22  P 22 and then the sine
and cosine functions of arguments O1 x or O 2 x will replace with hyperbolic sine and hyperbolic cosine
function with arguments P1 x or P 2 x. It is worth to mention it that this system of equations is of fourth
order and we expect for a solution with 4 coefficients regardless of this fact that the
equation s 2  O2 0 is the characteristic equation of a second order differential equation. Application of
the boundary conditions (6) to the solution (11) leads to a system of algebraic equations
as M 4u4 ^C1 C 2 C 3 C 4 `T ^0 0 0 0`T where the non vanishing entries of
matrix M 4u4 for O12 ! 0 and O22 ! 0, are:

m11 1, m13 1, m21 iG1 , m22 O1 , m23 iG 2 , m24 Ȝ2 ,

iG1" iG1" iG 2 "
m31 e cos O1", m32 e sin O1", m33 e cos Ȝ2 ", m34 eiG 2" sin O2 ",
(12)
m41 (iG1 cos O1"  O1 sin O1")eiG1" , m42 (iG1 sin O1"  O1 cos O1")eiG1" ,
m43 (iG 2 cos Ȝ2 "  Ȝ2 sin Ȝ2 ")eiG 2" , m44 (iG 2 sin Ȝ2 "  Ȝ2 cos Ȝ2 ")eiG 2" ,
2 2
If O1 or O2 to be negative then O1 or O 2 will be replaced by iP1 or iP 2 respectively. Finally the frequency
equation is det(real (M 4u4 )) 0 or det(imag (M 4u4 )) 0. To help the memory that multiplying one of the
rows of the matrix M 4u4 by complex number i gives the same results for two aforementioned equations.

3 NUMERICAL EXAMPLES AND RESULTS

The analysis performed in the preceding section allows the consideration of a finite length pipe which
is perfectly bonded on elastic support. Pipe and inner fluid specifications which are used for the following
evaluations can be found in table (1). From now on, all parameters of system in all tables are assumed to
be the same as table 1, except for the parameters that are changing in every table or we refer to new
values of them.

Table 1: Specifications used in the examples

( Di , Do ) (0.05 , 0.054) [m],

U 3 [m / s],
E 207[GPa],
( U p , U f ) (7747,997)[kg / m3 ],
" 2 [m],
k 100 [ N / m 2 ]

In the first example we ignore the stiffness of foundation ( k 0 ), in this case, some of dimensionless
natural frequencies of coupled system D n aZ n are tabled in table 2. As it mentioned before, in the earlier
work (Biswas and Ahmed 2001) the second term of Eq. (4) was vanished and an analytical 1 relation for the
natural frequencies of system as D n E ( P n 4  u 2 d nn ) was given where d nn [I nc ( x)] 2 dx and P n is
³
the roots of the equation cosh P n cos P n  1 0. Also I n (x) is 0 a mode shape
as In ( x) cosh P n x  cos P n x   cosh Pn  cos P n  sinh Pn x  sin P n x   sinh Pn  sin Pn  , n 1, 2,... .
Comparison of this study and (Biswas and Ahmed 2001) shows the eliminating the second term of Eq. (4)
 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993 2991

can be incorrect because of missing some of the natural frequencies and also there is a difference between
the others.

Table 2. Comparison of dimensionless natural frequencies of this study and (Biswas and Ahmed 2001) for k 0.
Dn aZ n D1 D2 D3 D4 D5 D6 D7
This study 4.1992 11.577 22.6964 37.5189 56.0472 78.2811 104.2208
Saroj et al. ------- 16.7995 --------- 46.3104 --------- 90.7879 -----------

Another valuable point about the (Biswas and Ahmed 2001) is the determination of the stability limit.
According to the relation D n E ( P n 4  u 2 d nn ) the stability condition is P n 4  u 2 d nn t 0 or
2
u d u crtical P n / d nn , which for n 1 we obtain P1 4.375 and d 11 12.3025 where leads to
u crtical 6.3786. For this study we set D n 0 and then we find G 1 , G 2 0, and O12 K  K 2  b ,
O 22 K  K 2  b , consequently we search for the fluid velocity ( u crtical ) such that the frequency
equation to be satisfied at this fluid velocity. In this case for the specification of table 1 we
find u crtical 0.8268. Therefore the velocity limit for inner fluid is U d U crtical 39.7914 [m / s ]. The
variation of u crtical versus 1  E U f A /( m p  U f A) as a system parameter depicted at Fig. 1. If the mass
of inner fluid per unit length in comparison to the total mass per unit length of pipe to be small the critical
velocity for the stability of system is high and reversely it would be low. The same trend can be seen in
the (Biswas and Ahmed 2001). The effects of inner fluid velocity on the dimensionless natural
frequencies of coupled system (D n aZ n ) for velocities less than U crtical are investigated using the two
first natural frequencies of table 3. We may observe that the increasing the inner fluid velocity results in
the decreasing of these dimensionless natural frequencies.

6
k=100 [N/m2]
k=0.001 [N/m2]
5

4
ucritical

0
0 0.2 0.4 0.6 0.8 1
íβ

Fig. 1: Variation of critical inner velocity for the stability of system versus 1  E
2992 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993

Table 3. Variation of first two frequencies with U

U D1 D2 U D1 D2
5 4.2027 11.5774 25 4.1499 11.5242
15 4.1851 11.5597 35 4.0967 11.4709

The variation of dimensionless frequencies with the stiffness of elastic foundation and the density of
inner fluid can be seen in table 4. As it can be seen the non dimensional natural frequencies are increased
slowly when the stiffness is increased at least one order of magnitude. Another result can be found by
changing the inner fluid density. As it can be observed that the dimensionless natural frequencies is
reduced by growing the fluid density.

Table 4. Variation of first two dimensionless natural frequencies with k and U f

k D1 D2 U f [kg / m 3 ] D1 D2
0 4.1992 11.577 510 (Propane) 4.7392 13.0525

100 4.204 11.5787 997(Water) 4.1992 11.577

1000 4.246 11.5939 13850 (Mercury) 1.6281 4.4957

On the other hand table 5 shows that the growing of the rigidity and density of pipe enhances the
dimensionless natural frequencies. The inner pipe diameter reduction (with constant thickness) and the
beam length enlargement cause the weaker beam and consequently reduction in the first two
dimensionless natural frequencies. This point may be realized from table 6.

Table 5. Variation of first two frequencies with E and U p

Material E, U D1 D2
Aluminum E 74 [GPa], U 2700 [kg / m 3 ] 3.1251 8.5953

Steel E 207 [GPa ], U 7747 [kg / m 3 ] 4.204 11.5787

Cu E 130.6 [GPa ], U 8960 [kg / m 3 ] 4.336 11.9368

Table 6. Variation of first two frequencies with Di and "

Di ( Inner diameter ), t (Thickness ) D1 D2 "[m] D1 D2

Di 0.05 [m], t 0.002 [m] 4.204 11.5787 2 4.204 11.5787

Di 0.06 [m], t 0.002 [m] 4.0239 11.0869 3 1.915 5.1613

Di 0.07 [m], t 0.002 [m] 3.8656 10.6528 4 1.3137 2.9984

 R.T. FAAL and D. DERAKHSHAN / Procedia Engineering 14 (2011) 2986–2993 2993

4 CONCLUSIONS

The main conclusions of this paper may be listed as follows:

1) Dimensionless natural frequencies of coupled system are increased by growing stiffness of
elastic support and rigidity of pipe and also by reducing inner fluid density and velocity, inner
pipe diameter (with constant thickness) and beam length.
2) In the stability analysis of the pipe vibration the parameter E plays an important rule.

References

[1] Ashley H and Haviland G, (1950). Bending vibrations of a pipeline containing flowing fluid. Journal of Applied Mechanics.
72, 229–232.
[2] Biswas SK and Ahmed NU (2001). Optimal control of flow-induced vibration of pipeline. Dynamics and Control. 11, pp.
187-201.
[3] Gorman DG, Reese JM and Zhang YL (2000). Vibration of a flexible pipe conveying viscous pulsating fluid flow. Journal of
Sound and Vibration. 230, pp. 379–392.
[4] Lee U and Kim J (1999). Dynamics of branched pipeline systems conveying internal unsteady flow. Journal of Vibration and
Acoustics. 121, pp.114–122.
[5] Lee U, Pak CH, and Hong SC (1995). The dynamics of a piping system with internal unsteady flow. Journal of Sound and
Vibration. 180, pp 297–311.
[6] Lee US, Jang IN Go HA (2009). Stability and dynamic analysis of oil pipelines by using spectral element method. Journal of
Loss Prevention in the Process Industries, 22, pp.873–878.
[7] Lee US and Park J (2006). Spectral element modelling and analysis of a pipeline conveying internal unsteady. Journal of
Fluids and Structures 22. pp. 273–292.
[8] Zwillinger (2003). Standard Mathematical Tables and Formulae. CRC Press LLC.