Professional Documents
Culture Documents
1 s2.0 S0308016104001607 Main PDF
1 s2.0 S0308016104001607 Main PDF
1 s2.0 S0308016104001607 Main PDF
www.elsevier.com/locate/ijpvp
Abstract
The primary objective of this paper is to develop a simulation model for the fluid–structure interactions (FSI) that occur in pipeline systems
mainly due to transient events such as rapid valve closing. The mathematical formulation is based on waterhammer equations, traditionally
used in the literature, coupled with a standard beam formulation for the structure. A new finite element formulation, based on flow velocity,
has been developed to deal with the valve closure transient excitation problems. It is shown that depending on the relative time-scales
associated with the structure, fluid and excitation forces, there are situations where the structural vibration response increases with FSIs. This
is in contrast to what is normally accepted in the literature, i.e. FSI reduces the structural displacements.
q 2004 Elsevier Ltd. All rights reserved.
1. Introduction used to reduce the size of the problem and care has been
exercised to include axial mode shapes since the interaction
Even though many researchers have used hybrid models occurs through the axial equations of the beam.
for waterhammer problems, with the method of charac- Tijsseling [1] presented a very detailed review of
teristics (MOC) modeling the waterhammer equations and transient phenomena in liquid-filled pipe systems. He
the finite element method (FEM) modeling the structure, dealt with waterhammer, cavitation, structural dynamics
few have used the wave equation resulting from the and fluid–structure interaction (FSI). The main focus was on
elimination of one of the variables from the waterhammer the history of FSI research in the time-domain. One-
equation in FEM. The wave equation can be formed with dimensional FSI models were classified based on the
flow velocity as the fluid variable, which is appropriate for equations used. The two-equation (one-mode) model refers
the valve closure excitation. This equation is elliptical in to classical waterhammer theory, where the liquid pressure
nature and hence can be readily modeled using FEM. In this and velocity are the only unknowns, the four-equation (two-
investigation, this feature is exploited to develop a coupled mode) model allows for the axial motion of straight pipes;
FEM formulation of both the structure and the fluid. Effects axial stress and axial pipe-wall velocity are additional
such as junction coupling and Poisson coupling are included variables. The six-equation model is necessary if radial
while friction coupling has been neglected due to the short inertia forces are to be taken into account; hoop stress and
time-scales associated with the excitation. Model reduction, radial pipe-wall velocity are the additional unknowns. The
based on the structural and fluid vibration modes, has been state-of-the-art fourteen equation model describes axial
motion (liquid and pipes), in and out-of-plane flexure, and
torsional motion of three-dimensional pipe systems.
* Corresponding author. Tel.: C91-44-2257-8192; fax: C91-44- Wiggert et al. [2] used the MOC to study transients in
2257-0509
E-mail address: mouli@iitm.ac.in (C. Padmanabhan).
pipeline systems. They identified seven wave components,
1
Currently with Government College of Engineering, Thiruvanantha- coupled axial compression of liquid and pipe material,
puram, Kerala, India. coupled transverse shear and bending of the pipe elements
0308-0161/$ - see front matter q 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2004.06.009
2 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
in two principal directions and torsion of the pipe wall. The rw V_ C p 0 Z 0 (5)
fourteen characteristic hyperbolic partial differential
equations were converted to ordinary differential equations 2 2
p_ C rw a V 0 K 2rw a nu_ 0 Z 0 (6)
by the MOC transformation. The formulation was applied to
two systems of three mutually perpendicular pipes. where
Heinsbroek [3] reported an application of FSI in the 2 ðKf =rw Þ E
nuclear industry. His analysis was based on a combination a Z ; E Z ;
of MOC and FEM. His conclusion was that while the MOC 1 C Kf D=E t 1 K n2
technique was superior for axial dynamics, FEM was more Kf is the fluid bulk modulus, mp, E, G, n, Ip, Ap, D, rw, t, u, v, w,
robust for transverse/lateral dynamics. The investigation p and V, are the mass per unit length of pipe, Young’s
also highlighted the fact that FSIs do take place and a model modulus of elasticity, Poisson’s ratio, the second moment of
based only on the fluid gives erroneous results. This is area, the cross-sectional area, the inner diameter, density of
corroborated by data from experiments. Lee and Kim [4] fluid, the thickness of the pipeline, displacement of pipe in
used a finite element formulation for the fully coupled x-direction, displacement of pipe in y-direction, displace-
dynamic equations of motion and applied it to several ment in z-direction, pressure and velocity of flow, respec-
pipeline systems. Wang and Tan [5] combined MOC and tively. If the derivative of Eq. (5) with respect to the axial
FEM to study the vibration and pressure fluctuation in a direction and Eq. (6) with respect to time respectively is
flexible hydraulic power system on an aircraft. Casadei et al. taken, one of the variables can be eliminated. Two wave
[6] presented a method for the numerical simulation of FSI equations can then be obtained, either in terms of pressure or
in fast transient dynamic applications. They had used both in terms of velocity. The wave equations obtained are
finite element and finite volume discretization of the fluid elliptical in nature and suitable for solution by the FEM.
domain and the peculiarities of each with respect to the Since the boundary condition for the valve closure event is in
interaction process were highlighted. terms of flow velocity it is easier to use the wave equation in
An earlier study carried out by Kellner et al. [7] showed terms of velocity and is given by:
that FSI reduced displacements and the corresponding loads
on the snubber below the elbow by a factor of almost four. v2 V 1 v2 V v3 u
K K 2n Z0 (7)
In this investigation junction coupling was considered vx2 2
a vt
2 vx2 vt
whereas Poisson coupling was neglected. Lavooij and The 3D beam element with six degrees-of-freedom per node
Tijsseling [8] suggested a provisional guideline to judge is used to model the pipe Eqs. (1)–(4) resulting in the
when the FSI is important. This guideline is based on the equation below
characteristic time-scales of the system under consideration.
One of the objectives of this study is to re-examine those ½Mf€u g C ½Kfug K ½S2 fpg Z ff ðtÞg (8)
proposed guidelines using the new finite element formu- where [M] and [K] are the mass matrix and stiffness matrix of
lation based on flow velocity. the pipe and the interaction of pressure
Ð with the structure due
to the Poisson coupling ½S2 Z 2n 0l ½N s T ½Np0 dxHere ½N s
2. Finite element formulation represents the shape function matrix for the axial displace-
ment of the structure and [Np] the shape function matrix for
fluid pressure. The matrix [N 0 p] represents the gradient of the
2.1. Waterhammer problem
shape function in the x-direction. The junction coupling is
modeled as a force term {f(t)} at the nodes on the junctions
For studying the FSIs in pipelines, the model proposed by
given by the area of cross-section multiplied by the pressure
Wiggert et al. [2] has been used. The first four equations are
at the respective node. The finite element form of the wave
related to the structure while Eqs. (5) and (6) are the
Eq. (7) is formulated using the Galerkin technique and is
waterhammer equations. This model accounts for the
Poisson coupling, which appears in the axial structural € g C ½HfVg K ½Sfug
½GfV _ Z f0g (9)
equation (Eq. (1)) and the influence of the structural response
on the pressure (Eq. (6)). The set of pipe dynamic equations where
suggested by Wiggert et al. (1987) is shown below: ðl ðl
1
½G Z 2
NvT Nv dv; ½H Z Nv0T Nv0 dv;
00 0
EAp u K m€u C 2nAp Z 0 (1) a 0 0
ðl
½S Z 2n Nv0T N s0 dv
EIp w 0000 C mw
€ Z0 (2) 0
½Afpg
_ C ½BfVg K ½S1 fug
_ Z f0g (10) 2.2. Pressure transient problems
8 9 2 0 ½I 0
38 9 8
0 0 9
> u_ > > u> > 0 >
>
> >
> 6 >
7> >
> >
> >
>
> u€ >
> > 6 7>
> >
> >
> >
>
>
> > 6 K½MK1 ½K 0 0 0 ½M ½S2 7>
K1
_
u > > ½M ff ðtÞg >
< > > > > >
K1
= 6 7< = < =
V_ Z 66 0 0 0 ½I 0 7 V C
7> > > 0 (11)
>
> >
> 6 7> > > >
>
>
> > 7> > > >
>
> V€ >> 6
> 4 0 ½GK1 ½S K½GK1 ½H 0 0 >
>
5>
_
V >
>
>
>
>
> 0 >
>
>
: >
> ; > >
: ; : > >
;
p_ 0 ½A ½S1 K½A ½B 0
K1 K1
0 p 0
½Afpg
_ C fBg½4f fVg K ½S1 ½4S fxg
_ (16)
The respective reduced matrices are used in the fourth order 3. Validation studies
Runge-Kutta integration scheme. The values of the
structural variables as well as that of the fluid flow velocity 3.1. Benchmark 1
available in the modal coordinates are transformed to the
nodal coordinates by multiplying with [4s] and [4f], Heinsbroek [3] used the water hammer theory for the
respectively. The pressure values can be used directly as fluid coupled with beam theory for the pipe to model FSI
no transformation is carried out. problems in non-rigid pipelines systems. He compared two
4 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
Fig. 1. The Heinsbroek [3] pipeline system and pressure comparison at valve.
different beam theories and two different solution methods the result of Heinsbroek [3]. Once again higher frequencies
in the time domain. First he used a hybrid method, i.e. the are present in the results of the present formulation, which is
fluid equations are solved by the MOC and the pipe due to the possible smaller time steps taken during
equations are solved by the FEM in combination with a simulation. Fig. 4 shows the displacement histories of the
direct time integration scheme. In the second method, he z-direction at the bend, of which the z displacement becomes
used only the MOC for the pipe as well as for the fluid unstable without FSI but with FSI it is much smaller
equations. The system analyzed consists of two pipes with and stable. This is similar to the example shown in Kellner
lengths of 310 and 20 m. The diameter of the pipe is et al. [7]. The natural frequencies of the structure and that of
0.2064 m and its wall thickness is 6.35 mm. The material the fluid are found out separately and are given in Table 1.
properties are rsZ7900 kg/m3, EZ210 GPa, nZ0.3, These were obtained using LAPACK [9] eigenvalue solver
k2Z0.53, rfZ880 kg/m3, KZ1.55 GPa. routines. The added mass effect of the fluid is included while
The structural boundary conditions for the pipeline evaluating the structural frequencies. The fluid frequencies
system are no displacements at the ball valve as well as at are evaluated from the finite element form of Eq. (9) without
the upstream reservoir end. Further the vertical motion at
every 10 m along the pipe is arrested by supports such that
only horizontal motion is allowed. Hydraulic transients are
generated by closing the valve in 0.5 s. It is assumed that the
flow velocity decreases linearly.
Fig. 1 shows the pressure history at the valve due to valve
closure; a comparison of the results from the present
formulation with those of Heinsbroek [3] shows good
agreement. Some higher frequency ripples are seen in the
results of the present formulation. It can be noted that while
the magnitudes agree very well, there seems to be a phase
difference of 1808 in the pressure response predicted, as can
be seen from Fig. 1. Fig. 2 shows the pressure time histories
at the valve with and without FSI. The effect of the vibration
of the structure on the fluid is to increase the peak values
of pressure, when interaction is included in the model.
The displacement history of the pipe at the bend is shown
in Fig. 3 and the maximum magnitude matches well with Fig. 2. Pressure at the valve with and without FSI.
J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14 5
including the FSI term, i.e. the last term. From the table it Table 1
can be seen that the structural frequencies are much below Structural and fluid frequencies, in Hz, for Heinsbroek [3] geometry
the fluid frequencies. Hence, very good FSI can be expected, Serial no. Structural frequency Fluid frequency
as seen from the pressure and displacement plots in Figs. 2 1 178.1 55.3
and 4, respectively. At the same time, in the y-direction, 2 178.1 166.1
there is not much interaction as seen in Fig. 5. 3 191.5 277.3
4 202.7 389.2
5 308.5 502.3
3.2. Benchmarks 2 and 3 6 308.5 616.7
7 377.0 732.8
8 409.2 850.9
Wiggert et al. [2] analyzed the liquid and structural
9 436.7 971.3
transients in piping by the MOC. The pipe and fluid 10 437.1 1094.3
dynamic equations presented in Ref. [2] are made use of in 11 553.4 1220.3
the present study also. The formulation was demonstrated 12 619.8 1349.5
for two cases of a system with three pipes directed 13 620.9 1482.3
14 716.9 1618.7
orthogonally and connected in series as shown in Fig. 6. 15 736.0 1759.1
For the first case (benchmark 2), the piping is made of
Fig. 4. Displacement in the z-direction, at the bend, with and without FSI. Fig. 5. Displacement in the y-direction, at the bend, with and without FSI.
6 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
Fig. 7. Pressure history comparison between (a) Wiggert et al. [2] and (b) present formulation.
J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14 7
Table 2
Frequencies of structure and fluid, in Hz, for benchmark 2 geometry [2]
Fig. 8. Pressure at the valve, for Benchmark 2 case, with and without FSI.
pressure with FSI are higher than that without FSI resulting
from the flow of energy from the structure to the fluid.
Consequently, the structural displacement reduces. In the
present finite element formulation, it is observed that higher
Fig. 10. FFT of the structural response without the effect of structure on
fluid.
Fig. 9. Velocity of the pipe at the bend C in x and z-directions, (a) Ref. [2]
and (b) present formulation. Fig. 11. FFT of fluid response without effect of structure on fluid.
8 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
Fig. 13. Pressure history comparison for benchmark 3 between (a) Wiggert et al. [2] and (b) present formulation.
J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14 9
Fig. 15. Structural velocity at C in x-direction, (a) from Ref. [2] and (b) present formulation.
10 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
Table 3 Table 4
Structural and fluid frequencies, in Hz, for Wiggert et al. [2] benchmark 3 Structural and fluid frequencies, in Hz, for modified Heinsbroek [3]
case geometry (with each segment 165 m long)
Serial no. Structural frequency Fluid frequency Serial no. Structural frequency Fluid frequency
1 0.03 1.0 1 0.01 0.9
2 0.04 3.0 2 0.03 2.6
3 0.08 5.0 3 0.06 4.4
4 0.10 7.0 4 0.09 6.1
5 0.17 9.1 5 0.14 7.9
6 0.20 11.1 6 0.20 9.6
7 0.30 13.1 7 0.26 11.4
8 0.33 15.1 8 0.33 13.1
9 0.45 17.1 9 0.34 14.9
10 0.50 19.1 10 0.42 16.6
11 0.64 21.1 11 0.51 18.4
12 0.69 23.2 12 0.62 20.1
13 0.86 25.2 13 0.73 21.9
14 0.92 27.2 14 0.85 23.7
15 1.12 29.3 15 0.92 25.4
Fig. 18. Velocity of the structure without FSI at the bend in x and z-direction
Fig. 16. Z-direction velocities when the valve is closed in 0.15 s. for modified Heinsbroek [3] geometry with each section being 165 m.
J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14 11
Fig. 19. Velocity of the structure with and without FSI at the bend in the (a) z-direction and (b) x-direction.
3.3. Parameter study Table 4 where the fluid frequencies are same as in the
original Heinsbroek [3] case (see Table 1). The fundamental
In order to understand the role of structural and fluid time structural frequency is increased in this case as the
scales as well as the excitation time scales, in the presence maximum length of a section is reduced.
or absence of FSI, a parametric study has been carried out by
varying section lengths while keeping the total length Table 5
constant. This is done so that the fluid time-scales are Structural and fluid frequencies, in Hz, corresponding to Fig. 20
constant while the structural time-scales are varied due to Serial no. Structural frequency Fluid frequency
the change in the geometric configuration. The Heinsbroek
1 178.1 23.7
[3] piping system is considered where the total length is 2 178.1 71.1
330 m, with two sections of 310 and 20 m, respectively (see 3 191.5 118.6
Fig. 1). Now, this is divided into two sections of equal 4 202.7 166.1
length keeping all other properties the same. The first fifteen 5 308.5 213.6
6 308.5 261.3
frequencies of the structure as well as the fluid are given in 7 376.9 309.2
8 409.3 357.2
9 436.7 405.3
10 437.4 453.7
11 553.0 502.3
12 619.9 551.1
13 622.7 600.3
14 637.8 649.7
Fig. 20. Addition of another length of piping with a bend to the original 15 716.9 699.6
Heinsbroek [3] geometry.
12 J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14
The pressure response with and without FSI and the fluid frequency is reduced to 23.7 Hz from 55.3 Hz as seen
velocity of the structure in the x and z-directions at the bend from Table 6. The pressure variations in Fig. 23 as well as
is given in Figs. 17–19, respectively. From the figures it can structural displacements in Fig. 24 show little change due to
be seen that the pressure peak values are altered by fluid frequency reduction.
the structural vibration. In this case, one can observe that
the effect of FSI is to increase the structural response in Table 6
Structural and fluid frequencies, in Hz, for modified benchmark 2 geometry
addition to the pressure response. This is most likely due to
(see Fig. 22)
the matching of the fluid frequency with the axial vibration
of the structure and the excitation time-scale being smaller Serial no. Structural frequency Fluid frequency
than the structural time-scale. In Fig. 20, an additional 1 0.009 1.0
section of 50 m is added to the Heinsbroek [3] configuration 2 0.03 3.0
3 0.05 5.0
with a bend. The structural and fluid frequencies are shown
4 0.09 7.0
in Table 5 of which the lowest structural frequency is 5 0.13 9.1
0.001 Hz, which implies that the structure is very flexible. 6 0.19 11.1
There is a transfer of energy from the fluid to the structure 7 0.25 13.1
and the structural response increases in this case while the 8 0.33 15.1
9 0.41 17.1
pressure response comes down. These structural response 10 0.50 19.1
results are shown in Fig. 21. 11 0.60 21.2
As a last case, the fluid frequency of Wiggert et al. [2], 12 0.71 23.2
benchmark 2, is altered by extending the last pipe section to 13 0.84 25.2
14 0.97 27.5
10 m and constraining all degrees-of-freedom of the new
15 1.11 29.1
portion of the pipe. This is shown in Fig. 22. The lowest
J. Kochupillai et al. / International Journal of Pressure Vessels and Piping 82 (2005) 1–14 13
Fig. 24. Displacement at the bend C with and without FSI, in the z-direction.
investigation of this aspect to establish guidelines, which are [5] Wang ZM, Tan S-K. Vibration and pressure fluctuation in a
better than those that exist today. flexible hydraulic power system on an aircraft. Comput Fluids 1998;
27:1–9.
[6] Casadei F, Halleux JP, Sala A, Chille F. Transient fluid–structure
References interaction algorithms for large industrial applications. Comput
Methods Appl Mech Eng 2001;190:3081–110.
[1] Tijsseling AS. Fluid–structure interaction in liquid-filled pipe systems: [7] Kellner AP, Grenenboom HL, de Jong JJ. Mathematical models for
a review. J Fluids Struct 1996;10:109–46. steam generator accident simulation Proceedings of the IAEA,
[2] Wiggert DC, Hatfield FJ, Struckenbruck S. Analysis of liquid and IWGFR/50, Specialist Meeting, The Hague, Netherlands 1983 pp.
structural transients in piping by the method of characteristics. ASME 115–121.
J Fluids Eng 1987;109:161–5. [8] Lavooij CSW, Tijsseling AS. Fluid–structure interaction in liquid-filled
[3] Heinsbroek AGTJ. Fluid–structure interactions in non-rigid pipeline piping systems. J Fluids Struct 1991;5:573–95.
systems. Nucl Eng Des 1997;172:123–35. [9] Anderson E, Bai Z, Bischof C, Blackford S, Demmel J, Dongarra J, Du
[4] Lee U, Kim J. Dynamics of branched pipeline systems conveying Croz J, Greenbaum A, Hammarling S, Mckenney A, Sorensen D.
internal unsteady flow. ASME J Vibration Acoustics 1999;121:114–21. LAPACK user’s guide. Philadelphia: SIAM; 1999.