Академический Документы
Профессиональный Документы
Культура Документы
Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct
a r t i c l e i n f o a b s t r a c t
Article history: This paper studies the dynamic properties of aqueduct–water coupling system in bent-type aqueduct
Received 17 April 2012 structures using the Arbitrary Lagrangian–Eulerian (ALE) method. A three-dimensional fluid–structure
Revised 18 July 2012 interaction model was established, with plate rubber supports. The speed-time sequence of fluctuating
Accepted 29 July 2012
wind acting on the aqueduct was simulated by the Auto-regressive Moving Average (ARMA) model.
The natural vibration characteristics, seismic responses, and wind responses of the aqueduct structure
were calculated and comparatively analyzed in different conditions of water depth. The simulation
Keywords:
results show that the application of isolation technology can reduce aqueduct stiffness and change the
Isolated aqueduct bridge
ARMA model
vibration characteristics of an aqueduct structure. The application of isolated technique is able to elevate
Fluid–structure interaction the earthquake resistance performance of aqueduct structure. However, the isolation remarkably
Wind response increases the wind stress response and reduces wind resistance performance of the aqueduct bridge.
Seismic response Ó 2012 Elsevier Ltd. All rights reserved.
0141-0296/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.engstruct.2012.07.019
H. Zhang et al. / Engineering Structures 46 (2013) 28–37 29
Fig. 1. Dongshen aqueduct in China. The motion of an incompressible viscous fluid, described by do-
main X, is governed by a set of partial differential equations that
are derived from the laws of the conservation of mass (continuity)
low calculating accuracy of the connection of spring oscillator and
and the conservation of momentum. The Navier–Stokes equations
the aqueduct body. Boundary element method was used to analyze
of the fluid in the ALE description can be written as follows
water sloshing height and the distribution of velocity potential in
[15,18,19].
[16], however, effects of the aqueduct on the water is not analyzed.
In the ALE method, the grid can be moved to follow boundary mo-
Continuity equation:
tion, resolve complex flow features, and prevent the grid from tan-
gling. The study results in [11] show that the ALE method can give @ u_ b
¼0 ð2Þ
a full consideration of the coupling effects of both water and struc- @xb
ture in solving the nonlinear fluid–solid coupling problem.
Momentum equation:
In order to fully take into account the impact of the isolated
support on the wind-induced dynamic performance of the @ u_ a @ u_ b @ sab
water–aqueduct coupling system, a three-dimensional finite ele- þ ðu_ b u_ 1b Þ ¼ þ fB ð3Þ
@t @xb @xb
ment model of an isolated aqueduct structure is introduced in this
paper. The ARMA model is used to simulate the fluctuating wind where the subscripts a and b represent the three spatial coordinate
speed time series. The dynamic characteristics of the aqueduct– components and represented indices imply summation; u_ a is the
water coupling system are studied by using the ALE method. fluid velocity component; fB is the body force. The term of
Dynamic responses of the water–aqueduct coupling system under pseudo-stress is defined as:
lateral earthquake and wind loads, respectively, are calculated in
order to study the different influences of the isolated support on @ u_ a @ u_ b
seismic and wind resistance of the aqueduct structure. This work sab ¼ Pdab þ ðl þ lt Þ þ lt ð4Þ
@xb @xa
can provide scientific basis for the wind resistance design of iso-
lated aqueducts. where P = p/q is the kinematic pressure; p the static pressure; q the
This paper is organized as follows. Section 2 proposes the fluid density; dab the Kronecker delta; l and lt the dynamic viscos-
fundamental equations of fluid–solid coupling system. Section 3 ity and dynamic eddy viscosity, respectively.
presents the stochastic wind field simulation based on the AR mod- To apply the weighted residuals to the Eqs. (2) and (3), multiply
el and the ARMA model. Section 4 gives the numerical simulation each equation by a test (weight) function, where use function w1
analysis. The last part, Section 5, offers discussions and for the mass Eq. (2) and w2 for the momentum Eq. (3). Finally,
conclusions. the weak form of the mass and momentum equations are as
follows:
Z
2. Fundamental equations of fluid–solid coupling system @ u_ b
w1 dX ¼ 0 ð5Þ
X @xb
2.1. ALE description
Z Z Z
@ u_ a @ u_ a @w2
w2 dX þ w2 ðu_ b u_ 1b Þ dX P dX
The ALE method is a hybrid method comprising Lagrangian and X @t X @x b X @xa
Eulerian methods. In the implementation of ALE, the mesh is gen- Z I Z
@ u_ a @w2
erated in a reference coordinate system which is independent of þ ð l þ lt Þ dX w2 nb sab dC w2 f B dX ¼ 0 ð6Þ
X @xb @xb @X X
the structure and fluid movement. The mesh node denotes the ref-
erence node, which can move arbitrarily in space. The physical where nb is the b component of the surface outward unit normal
H
quantities in the fluid and structure domain can be described and vector and () indicates an integral over the surface boundary.
30 H. Zhang et al. / Engineering Structures 46 (2013) 28–37
Z
2.3. The Galerkin form for fluid equation
mij ¼ wiu_ wju_ dX ð16Þ
X
The approximation of velocity and pressure can be defined
Diffusion:
using the Finite Element Method (FEM) as a linear combination 2 3
of the basis functions of wu_ and wP in the discretized domain Xh. kij ðxxÞ kij ðxyÞ kij ðxzÞ
X
N 6 7
K ¼ 4 kij ðyxÞ kij ðyyÞ kij ðyzÞ 5 ð17Þ
u_ a ðx; y; zÞ ¼ u_ ja ðtÞwju_ ðx; y; zÞ ð7Þ
j¼1
kij ðzxÞ kij ðzyÞ kij ðzzÞ
X
M
Z !
Pðx; y; zÞ ¼ Pj ðtÞwjP ðx; y; zÞ ð8Þ @wju_ @wiu_ @wju_ @wju_
j¼1
kij ðabÞ ¼ dab ðl þ lt Þ þ lt dX ð18Þ
X @xb @xb @xa @xb
where N and M are the total number of velocity nodes and pressure
nodes. The substitution of Eqs. (7) and (8) into (5) and (6) yields: Advection:
Z !
@wju_ K ¼ kðu_ u_ 1 Þ ¼ diagðnij Þ; nij
w1 dXh u_ jb ¼ 0 ð9Þ Z
Xh @xb @wju_
¼ ðu_ u_ 1 Þ wju_ wku_ dX ð19Þ
" # X @xb
Z Z
@ u_ a j
@wju_
w2 wju_ dXh þ ðu_ b u_ 1 Þ w2 wku_ dXh u_ ja Gradient:
X h @t Xh @xb
Z
" # T @wiu_
j @w2 j @wju_ @w2 j C ¼ ½ cij ð1Þ cij ð2Þ cij ð3Þ ; cij ðaÞ ¼ wjP dX ð20Þ
wP P þ ðl þ lt Þ u_ X @xa
@xa @xb @xb a
Z ! I Force vector:
@wju_ @w2 h j I Z
þ l t dX u_ b w2 nb sab dC F ¼ ½ fi ð1Þ fi ð2Þ fi ð3Þ ; f i ðaÞ ¼
T
wiu_ nb sab dC wiu_ f B dX ð21Þ
Xh @xa @xb @ Xh @X X
Z
w2 f B dXh
Xh As discussed in the previous section, the approximate solution
¼0 ð10Þ is expressed as a function of a set of basis functions, i.e., shape
The above system contains 3N velocity unknowns and M pres- functions. An arbitrary function u as a linear combination of the
sure unknowns. We can obtain the essential number of equations shape functions can be expressed as:
by introducing one weighting function per unknown. According Ne
X
to the Galerkin method, the weight functions are chosen to be u N i ui ð22Þ
the same as the basic functions which are used to define the i¼1
approximate solution. So we define w1 ¼ wiP and w2 ¼ wiu_ and apply where Ni is a globally defined shape function; Ne the number of ele-
this method to Eqs. (9) and (10) yields: ment; and the coefficients ui the unknown nodal values. Hence, for
Z !
@wju_ any node j:
dX u_ jb ¼ 0
h
ð11Þ
Xh @xb Ne
X
uj ¼ uðxj Þ Ni ðxj Þui ð23Þ
Z " Z # i¼1
@ u_ ja @wju_
wiu_ wju_ dXh þ ðu_ b u_ 1 Þ wiu_ wku_ dXh u_ ja The element of eight nodes was chosen in this paper [20]. The
Xh @t Xh @xb
" # " # velocity shape functions in terms of the eight nodes element’s
i j i
@w @w @wu_ j parametric coordinates (n, g, f) are given as:
wjP u_ Pj þ ðl þ lt Þ u_ u_
@xa @xb @xb a 1 1
"Z # N1 ðn; g;fÞ ¼ ð1 nÞð1 gÞð1 fÞ;N 2 ðn; g;fÞ ¼ ð1 þ nÞð1 gÞð1 fÞ
I Z 8 8
@wju_ @wiu_
þ l t dX u_ jb
h
wiu_ nb sab dC wiu_ f B dXh 1 1
N3 ðn; g;fÞ ¼ ð1 þ nÞð1 þ gÞð1 fÞ;N 4 ðn; g;fÞ ¼ ð1 nÞð1 þ gÞð1 fÞ
Xh @xa @xb @ Xh Xh 8 8 ð24Þ
1 1
¼0 ð12Þ N5 ðn; g;fÞ ¼ ð1 nÞð1 gÞð1 þ fÞ;N 6 ðn; g;fÞ ¼ ð1 þ nÞð1 gÞð1 þ fÞ
8 8
which has 3N + M ordinary differential equations. 1 1
N7 ðn; g;fÞ ¼ ð1 þ nÞð1 þ gÞð1 þ fÞ;N 8 ðn; g;fÞ ¼ ð1 nÞð1 þ gÞð1 þ fÞ
8 8
2.4. Matrix form for fluid equation And the pressure shape functions are:
The Eqs. (11) and (12) can be written as a matrix form as NP ðn; g; fÞ ¼ 1 ð25Þ
follows: The derivatives of the shape functions in terms of the global
€ þ ðK þ K Þu_ þ CP ¼ F
Mu ð13Þ coordinate system can be expressed in matrix form as:
C T u_ ¼ 0 ð14Þ T T
@Ni @Ni @Ni @N i @N i @N i
@n @g @f ¼J @x @y @z
ð26Þ
where the matrices are computed from integrals of the shape func-
tions. The definitions of these matrices are given below. where J is the Jacobian matrix. Therefore, the matrix of Eq. (13) can
be evaluated by the transformation:
Mass: Z Z
2 3 udX ¼ uðx; y; zÞdx dy dz
mij 0 0 Xe Xe
6 7 Z Z Z
M¼4 0 mij 0 5 ð15Þ 1 1 1
0 0 mij
¼ uðn; g; fÞjJjdndgdf ð27Þ
1 1 1
H. Zhang et al. / Engineering Structures 46 (2013) 28–37 31
€ ðtÞ þ C uðtÞ
Mu _ þ KuðtÞ ¼ PðtÞ þ Pf ðtÞ ð28Þ where es and ex are tolerances for stress and displacement conver-
gence, respectively. The calculation procedure of the iterative
_
where uðtÞ; uðtÞ € ðtÞ are the displacement, velocity and acceler-
and u fluid–structure coupling method is shown in Fig. 2.
ation vectors at time t, respectively. P(t) is the transient external
load vector, which is the earthquake or wind load. Pf(t) is the fluid
force on the structure. M, K and C are the mass stiffness and damp- 3. Stochastic wind field simulation
ing matrices, respectively. The Rayleigh damping matrix used in this
paper is defined as: Turbulent wind can be described by Gaussian steady process
C ¼ ar M þ br K ð29Þ according to the statistical analysis of wind samples. The stochastic
wind field on membrane structures is synthesized in this study
where ar and br are the proportional damping coefficients of mass based on the power spectrum density (PSD) of wind velocity and
and stiffness. an Auto-regressive Moving-Average (ARMA) model. The merit of
the ARMA model is that it uses few parameters to approximate a
2.6. Coupling system equation and solving method target PSD [23,24].
The target PSD of the turbulent wind velocity can be expressed
Fluid–structure interaction occurs when fluid flow cause defor- by the Davenport power spectrum as follows [25]:
mation of the structure. This deformation, in return, changes the
4x2 u2
boundary conditions of the fluid flow. Under earthquake or wind SV ðnÞ ¼ ð34Þ
loads, movement of the aqueduct on the coupling interface leads nð1 þ x2 Þ4=3
to the sloshing of water, and the dynamic water pressure acts on 1200 nz
x¼ ; f ¼ ð35Þ
the aqueduct wall on the interface which changes the distribution zf VðzÞ
of deformation and stress of the aqueduct. The fluid and structure 0:4VðzÞ
material only interact their interface. Hence, the coupling occurs u ¼ ð36Þ
lnðz=z0 Þ
through the boundary conditions. The fluid and structure on the
coupling interface satisfy the conditions of displacements continu-
ity and equilibrium of forces.
Displacements continuity:
u_ f ¼ u_ s ð30Þ
Equilibrium of forces:
rf nf rs ns ¼ 0 ð31Þ
where u_ f and u_ s are the fluid velocity and the structure velocity on
the interface, respectively; rf and rs are the fluid stress and the
structure stress on the interface, respectively; nf and ns are the unit
outward normal vectors of the fluid and of the structure on the
interface, respectively.
The fluid equation and structure equation are unified to get the
finite element equation of the coupled fluid–structure system. The
iterative fluid–structure coupling method is used to solve the cou-
pling system in this paper. In this method, the fluid and structure
equations are solved individually, in succession, always using the
latest information provided by the other part of the coupling sys-
tem, which might be more applicable to solve very large problems
than the direct fluid–structure coupling method.
The coupling system is solved at each instant of time using the
incremental Newton–Raphson solution method. The solution vari-
ables of the fluid and structure are coupling, so the fluid equation
and the structure equation at each time step can be solved by iter-
ation in turn until the solutions of the coupling system are conver-
gent. For each time step, the displacements of the fluid structure
interface node may update the entire fluid mesh and change the
ALE velocity at each node. Criteria that are based on stress and dis-
placement are used to check for convergence of the iterations. Fig. 2. The procedure of the iterative fluid–structure coupling method.
32 H. Zhang et al. / Engineering Structures 46 (2013) 28–37
where f is Monin coordinate; VðzÞ is the average wind velocity and 4. Numerical analysis
z0 = 0.03 m is the ground roughness. Then the correlation function is
thus obtained: 4.1. A 3-D calculation model of the coupling system
Z xc
C ij ðkÞ ¼ 2 S0ij ðxÞ cosðkxsÞdx ði; j ¼ 1; 2; . . . nÞ ð37Þ The 3-D model of a large bent-frame rectangular aqueduct with
0
plate rubber bearings is introduced as shown in Fig. 3. The dynamic
where s is the sample time and x = 2pn is the cut off frequency. characteristics of the aqueduct–water coupling system are studied
using the ALE method. The span L of the aqueduct is 10 m; the
3.1. M-dimensional finite auto-regressive (AR) model height H1 of the H-shaped supporting frame is 24 m; the width B
is 7.0 m; the depth H2 is 5.6 m; the thickness is 0.5 m; the width
An m-dimensional discrete time sequence Yr (r = 1, 2, . . .) is writ- and height of the tie bar are 0.3 m and 0.4 m, respectively. The
ten as: aqueduct body and supporting frame are built using C50 and C30
concrete, respectively. For the C50 concrete, Young’s modulus E
X
P
e i Y ri , B0 X r
A ð38Þ is 3.45 104 MPa and Poisson’s ratio l is 0.167; and for the C30
i¼0 concrete, Young’s modulus E is 3.00 104 MPa and Poisson’s ratio
e i and B0 are coefficient matrices obtained by the predefined l is 0.167.
where A
The water and the structure are modeled using 3-D fluid ele-
auto-correlation function CYY(k); Xr is a normalized Gaussian white
ments and solid elements, respectively. The plate rubber bearing
noise sequence, with the auto-correlation function satisfying:
is installed between the aqueduct body and supported frame.
Fig. 3. Diagram of finite element model of aqueduct: (a) aqueduct structure and (b) water body.
Table 1
The natural frequencies of vibration of the aqueduct (unit: Hz).
to the local geographic and geologic conditions. In this analysis the the water depth 3.748 m. According to the displacement curves,
duration T is 16 s, and the time step Dt is 0.02 s. the application of isolation technology can increase the maximum
The local basic wind pressure is 0.35 kN/m2. The fluctuating transversal displacement of the aqueduct.
wind speed time series at different heights of the aqueduct struc- Figs. 11 and 12 respectively show the maximum transversal dis-
ture is simulated by using the Auto-regressive Moving Average placement curves at the top of the aqueduct in different water
(ARMA) model. The Davenport power spectrum is employed as depth conditions under earthquake action and wind load. It can
the target spectrum. In this analysis, the duration T is 120 s, and be observed from Figs. 11 and 12, with the depth of water in aque-
time step Dt is 0.1 s. Seven fluctuating wind speed time series at duct increasing, the maximum transversal displacement of the
different heights of aqueduct structure are obtained by MATLAB structure increases. Furthermore, for the aqueduct with water
simulation. Fig. 6 shows the time history of fluctuating wind veloc- depth of 3.748 m, the displacements of the isolated aqueduct un-
ity at the top of the aqueduct (30.8 m above ground). der the earthquake and the wind effect are larger than those of
Fig. 7 compares the power spectrum between the simulation re- the aqueduct with rigid bearing by 2.6 times and 2.5 times, respec-
sult and the target. It can be seen that fit agreement is good be- tively. Hence, the application of isolation reduces the stiffness of
tween the two curves in Fig. 7, which indicates the fluctuating the aqueduct and increases the seismic and wind displacement
wind is effectively simulated by ARMA model. The time history response.
of wind pressure at different heights of the aqueduct structure
can be obtained by adding the simulated fluctuating wind speed
4.3.2. Dynamic stress analysis
and average wind speed according to the relationship between
The dynamic stresses under the earthquake wave (El-Centro)
wind speed and wind pressure P = 0.5qV2 (q is the density of the
and wind load are respectively calculated and analyzed under
atmosphere). Fig. 8 shows the time history of fluctuating wind
the following three working conditions: water depth of 2.748 m,
pressure at the top of the aqueduct.
3.748 m and no water in the aqueduct. Figs. 13 and 14 respectively
show the major principal stress of middle cross-section of the
4.3.1. Dynamic displacement analysis aqueduct with rigid and isolated bearing, with water depth
The displacements under the earthquake wave (El-Centro) and 3.748 m.
wind load are respectively calculated and analyzed under the fol- From Figs. 13 and 14, we find that the distribution pattern of the
lowing six working conditions: water depth of 1.939 m, 2.748 m, main stresses of the aqueduct with rigid bearing and isolated bear-
3.248 m, 3.748 m, 4.248 m, and no water in the aqueduct. Figs. 9 ing are similar. The maximum stress occurs at the outer bottom of
and 10 respectively show the time history of transversal displace- the aqueduct tank caused by the sloshing water. Meanwhile, the
ments at the top of the aqueduct with rigid and isolated bearing of stresses on both sidewalls are also large because of the coupling
34 H. Zhang et al. / Engineering Structures 46 (2013) 28–37
Fig. 4. The mode of vibration of the aqueduct: (a) rigid bearing and (b) isolated bearing.
Fig. 7. Comparison of the power spectrum between the simulation result and the
Fig. 5. Diagram of the El-Centro seismic wave. target.
Fig. 6. Time history of fluctuating wind velocity at the top of the aqueduct. Fig. 8. Time history of fluctuating wind load at the top of the aqueduct.
H. Zhang et al. / Engineering Structures 46 (2013) 28–37 35
Fig. 9. Curves of transversal displacement at the top of the aqueduct with water depth of 3.748 m under earthquake action: (a) rigid bearing and (b) isolated bearing.
Fig. 10. Curves of transversal displacement at the top of the aqueduct with water depth of 3.748 m under wind load: (a) rigid bearing and (b) isolated bearing.
5. Conclusions
Fig. 13. Major principal stress of middle cross-section of the aqueduct with water depth of 3.748 m under earthquake action (unit: MPa): (a) rigid bearing and (b) isolated
bearing.
Fig. 14. Major principal stress of middle cross-section of the aqueduct with water depth of 3.748 m under wind load (unit: MPa): (a) rigid bearing and (b) isolated bearing.
Table 2
Major principal stress on the aqueduct under different water level conditions (unit: MPa).
earthquake and wind load events. The application of isola- earthquake resistance performance of an aqueduct struc-
tion technology can reduce aqueduct structural stiffness ture. However, the application of isolation technology
and lengthen the vibration period. remarkably increases the wind stress response and reduces
(2) With the depth of water in the aqueduct increasing, the wind resistance. In the application of seismic isolation tech-
maximum transversal displacements of structure under nology, appropriate measures should be taken, to permit the
earthquake excitation and wind load also increase. The dis- isolated bearings function, while ensuring the normal capac-
placements of the isolated aqueduct under the earthquake ity of the isolated aqueduct under strong winds.
and the wind effects are larger than those of the aqueduct
with rigid bearing. This means damping equipment or bum-
per blocks should be used to reduce the displacement Acknowledgement
response in a seismic or wind event.
(3) Because of the sloshing water, the stresses on both sidewalls The work was supported by the grants from the National
under outer excitement are large, so the coupling effect of Natural Science Foundation of China (Project No. 50709006) and
the aqueduct and water should be taken into account. Fur- the Fundamental Research Funds for the Central Universities of
thermore, the application of isolated technique elevates China (Project No. 2010B04014) which is greatly acknowledged.
H. Zhang et al. / Engineering Structures 46 (2013) 28–37 37