Professional Documents
Culture Documents
Abstract: This paper provides a very efficient, integrated framework for seismic analyses of long-span cable-stayed bridges. The efficiency
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 08/16/16. Copyright ASCE. For personal use only; all rights reserved.
comes from the dramatic reduction in formation time and the degrees of freedom (DOF) associated with the structure, using the integrated finite
strip method (IFSM) along with the application of a very robust and efficient time history method (THM) using the Newmark scheme for dynamic
analysis of the bridge structure. The previous versions of the finite strip method are limited to modeling the bridge deck only, whereas other struc-
tural components are replaced by assumed boundary conditions. Using the IFSM, all components of the long-span cable-stayed bridge can be mod-
eled in a unified system, and consequently, the real dynamic behavior including the interactions between deck, piers, and cables can be perfectly
considered. To verify the solution, the geometric and dynamic properties of the Kap Shui Mun (KSM) Bridge, as a real example of a long-span
cable-stayed bridge, are derived by the proposed finite strip method. Then, the seismic response of KSM Bridge under uniform and nonuniform
earthquake loadings is investigated by using the THM. The results show that the IFSM can be applied successfully for seismic analysis of long-span
cable-stayed bridges, and the analysis can be performed in a minimal amount of time. DOI: 10.1061/(ASCE)BE.1943-5592.0000821. This work is
made available under the terms of the Creative Commons Attribution 4.0 International license, http://creativecommons.org/licenses/by/4.0/.
Author keywords: Seismic analysis; Time history method; Newmark method; Long-span bridge; Cable-stayed bridge; Dynamic;
Integrated finite strip method.
Introduction using FEM has been one of the most widely adopted approaches for
the study and design of long-span bridges in the last few decades
To meet the economic, social, and recreational needs of the commu- (Cook et al. 1989). With increased span lengths and successful
nity for safe and efficient transportation systems, long-span bridges application of lightweight composite materials for bridges, the
have been built throughout the world. These bridges represent some dynamic characteristics of these structures have become more com-
of the most challenging kinds of structures designed in civil engi- plex and the conventional FEM approach might not always be suffi-
neering. Today, long-span cable-stayed bridges are of great interest, cient. In contrast, the formulation of a FEM model and setting up of
mainly as an alternative and a more economical solution than sus- the boundary conditions of a three-dimensional bridge structure for
pension bridges. Because of their great flexibility, modern long-span dynamic analysis is very complicated and time consuming, espe-
cable-stayed bridges are usually very susceptible to dynamic loads. cially for long-span cable-stayed bridges, and because of the
Therefore, the wind-resistant and earthquake-resistant designs are extremely high number of degrees of freedom (DOF) involved,
key issues for successful construction of these bridges. such a procedure becomes inconvenient in practice. The conver-
The structural response of long-span bridges under dynamic and gence rate of the conventional FEM in dynamic problems also is
aerodynamic loads is very complicated, and powerful computa- usually slow, because the nonlinearities associated with the flexible
tional techniques that can accurately analyze and predict the struc- bridge structures lead to significant redistribution of the internal
tural behavior of these massive structures are required. This is a forces. One of the solutions for improving the convergence rate is to
major point in successful and optimized design of long-span cable- use very small-sized elements throughout the structure, resulting in
stayed bridges. Most of the bridge structural analyses are performed a large number of DOF. Therefore, in the case of long-span cable-
by the aid of the well-known FEM. Quasi-static numerical analysis stayed bridges, with thousands of DOF, and at the same time with
high nonlinear behavior under dynamic and aerodynamic forces,
1
the finite element analysis does not seem to be efficient enough.
Ph.D. Student, Dept. of Civil and Environmental Engineering, Univ. The finite strip method, as a semianalytical numerical analysis pi-
of Ottawa, Ottawa, ON, Canada K1N 6N5.
2 oneered by Y. K. Cheung in the 1970s (Cheung and Cheung 1971),
Professor, Dept. of Civil and Environmental Engineering, Hong Kong
Univ. of Science and Technology, Kowloon, Hong Kong; Western China has shown great potential for analyzing bridge structures. However,
Earthquake and Hazards Mitigation Research Centre, Sichuan Univ., because of some limitations, this method is not as popular as the
Chengdu 610065, China. FEM. For instance, the ordinary finite strip method can deal only
3
Dept. of Civil and Environmental Engineering, Hong Kong Univ. of with prismatic members with simply supported end conditions.
Science and Technology, Kowloon, Hong Kong (corresponding author). Moreover, the method cannot handle shear forces, internal supports,
E-mail: zhenyuanshen@gmail.com and concentrated loads. In spite of the large number of references on
4
Assistant Professor, Dept. of Civil and Environmental Engineering, the use of the finite strip method in structural analysis, the application
Univ. of Ottawa, Ottawa, ON, Canada KIN 6N5.
of the existing finite strip methods in seismic analysis is restricted to
Note. This manuscript was submitted on August 20, 2014; approved
on June 17, 2015; published online on December 30, 2015. Discussion pe- bridge superstructures or bridges with assumed pier conditions only.
riod open until May 30, 2016; separate discussions must be submitted for In the conventional finite strip methods, the piers and the cables of
individual papers. This paper is part of the Journal of Bridge the bridge are normally replaced by certain boundary conditions, and
Engineering, © ASCE, ISSN 1084-0702. this assumption is reasonable only when the stress distribution along
coincide with the knots on the nodal lines, thus leading to more
considered. In the following subsection, the development of the accurate results. Also, the introduction of unequally spaced interior
dynamic properties of a long-span cable-stayed bridge using IFSM knots enables one to describe the accurate response in the region of
is briefly presented. high-stress gradients, or at the locations of abrupt geometric
changes, by spacing the knots more closely. The unequally spaced
B3-spline function centered at ym can be expressed as
Displacement Functions 8
>
> 0 y < ym2
>
>
The displacement of a strip in the IFSM is obtained by applying B3- >
> Am ðy ym2 Þ3 ym2 y < ym1
spline functions (Prenter 1975) in the longitudinal direction. For the >
< A ðy y Þ3 þ C ðy y Þ3
m m2 m m1 ym1 y < ym
transverse direction, the concept was adapted from the FEM, in Um ðyÞ ¼
>
> B ðy yÞ 3
þ D ðy yÞ 3
ym y < ymþ1
which the cubic polynomials are used to represent the transverse >
>
m mþ2 m mþ1
>
> B ðy yÞ 3
ymþ1 y < ymþ2
shape function of the strip. The same approach is used by the spline >
:
m mþ2
0 y y
finite strip method. mþ2
Assuming that the bridge deck is a shell structure, both in-plane (2)
and out-of-plane DOF are included in the analysis. A flat shell
spline strip is shown in Fig. 1, in which each knot of a nodal line has in which
Cm ¼ ðymþ2 ym2 Þ½ðymþ2 ym1 Þðymþ1 ym1 Þðym ym1 Þðym1 ym2 Þ1
Dm ¼ ðymþ2 ym2 Þ½ðymþ1 ym2 Þðymþ1 ym1 Þðymþ1 ym Þðymþ2 ymþ1 Þ1 (3)
The membrane displacement functions u and v and the flex- equations are cubic Hermite polynomial functions used for vertical
ural displacement function w are expressed as the product of displacement variation and linear interpolation for in-plane dis-
transverse polynomials and longitudinal B3-splines, as in the placements as
following:
N1 ¼ 1 X; N2 ¼ X; N3 ¼ 1 3X 2 þ 2X 3 ;
X
rþ1
u¼ ðN1 U1m ðyÞuim þ N2 U5m ðyÞujm Þ (4)
m¼1 N4 ¼ xð1 2X þ X 2 Þ; N5 ¼ ð3X 2 2X 3 Þ (7)
X
rþ1 where X ¼ x=b. ½U1m to ½U8m are row matrices, and each matrix
v¼ ðN1 U2m ðyÞvim þ N2 U6m ðyÞvjm Þ (5) has (m þ 3) local B3-splines. ½U1m , ½U2m , ½U5m , and ½U6m are
m¼1 related to displacements u and v of nodal lines i and j, respec-
tively, whereas ½U3m , ½U4m , ½U7m , and ½U8m are related to dis-
X
rþ1 placement w.
w¼ ðN3 U3m ðyÞwim þ N4 U4m ðyÞu im þ N5 U7m ðyÞwjm The cantilever behavior of the piers and the towers is modeled
m¼1 by using the so-called column strip, which is similar to the flat shell
spline finite strip. However, this is modeled as a vertical strip fixed
at one end, for providing the support boundary conditions, and free
þ N6 U8m ðyÞu jm Þ (6) at the other end, as shown in Fig. 2. In fact, the global z-direction of
the column strip is similar to the local y-direction in the flat shell
where r = total number of longitudinal sections on a nodal line. spline finite strip, which is controlled by the in-plane stiffness, in
Transverse shape functions adopted in the aforementioned the corresponding direction, whereas the global v-direction of the
X
rþ1
v¼ vm Um ðzÞ (10) Stiffness, Mass, and Damping Matrices
m¼1
After defining the displacement functions for all the structural mem-
For the displacement–strain relationships, only the bending in the ver- bers of the long-span cable-stayed bridge, in the environment of
tical and transverse directions and the axial stress are considered, whereas IFSM, the stiffness [k] and mass [m] matrices of different compo-
the shear stress and torsional moment are assumed to be negligible, nents can be calculated by using the standard finite element proce-
because the amount of these forces are very low in a 1D column strip. dure, which will be applied in the following form:
To model the cables in the finite strip environment, the cable ð
strip is developed, which is a simplified version of the 1D column ½k ¼ ½BTi ½D½Bj dA (11)
strip. In general, a cable can only resist against the axial tension
stresses. Therefore, only the axial stress defines the strain–displace-
ment relationship for a cable strip. ð
It should be noted that all the strips in the finite strip method ½m ¼ r t½NTi ½Nj dA (12)
must come with the preset boundary conditions. There are some
techniques, such as the penalty method for considering the end
in which r = density of the strip, and t = thickness of the strip;
boundary conditions of the strips, that can be simply applied to fi-
nite strip modeling (Cheung et al. 1996). [D] and [B] are the elastic matrix and the strain matrix, respec-
tively; and [N] is the shape function matrix. Similar to the FEM,
the IFSM strip properties are converted to nodal properties during
Modeling the Transition Section the model formulation process, although the number of nodes
required are significantly reduced because of the semianalytical na-
With the concept of elements and nodes in the FEM, there is no par- ture of the IFSM method. Thus, the derived matrices can be
ticular difficulty in adding other elements to the bridge deck and con- assembled using the standard assembling techniques, making the
necting them together. However, the concept element in the longitu- global stiffness matrix [K] and the global mass matrix [M] of the
dinal direction is not defined in the ordinary finite strip method or in bridge.
the spline finite strip method. To solve this problem, a special transi- There are different methods for evaluating the damping matrix
tion section was developed within the IFSM and is applied to connect [C] (Humar 2012). In this research, the classic Rayleigh damping is
Ms Msb d€s ðtÞ þ
Cs Csb d_ s ðtÞ þ
Ks Ksb d s ðtÞ
¼
0
(17)
T
Msb Mb d€b ðtÞ T
Csb Cb d_ b ðtÞ T
Ksb Kb d b ðtÞ Pb ðtÞ
in which the subscripts b and s represent the master (support dis- and Pb ðtÞ is the reaction force vector. Ms , Cs , and Ks are the mass,
placements) and slave DOF of the long-span structure, respectively, damping, and stiffness matrices associated with the unrestrained
Eq. (20), which leads to the following equation (De Silva 2005):
or
where
Fig. 6. Towers of the KSM Bridge (data from Naderian et al. 2015)
Newmark Scheme for Numeral Integration Using the Newmark method, similar to other types of time-step-
ping methods, the bridge responses at t ¼ tiþ1 can be predicted
Assuming that at t ¼ ti and t ¼ tiþ1 , the equations of motion can be under the condition that the responses at t ¼ ti and the loading at
written as both times are known.
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 08/16/16. Copyright ASCE. For personal use only; all rights reserved.
Fig. 8. (Color) Longitudinal acceleration responses under uniform sine wave excitation at Point F (see Fig. 7) at the top of the east tower
Fig. 9. (Color) Longitudinal displacement responses under uniform sine wave excitation at Point F (see Fig. 7) at the top of the east tower
in which the response and the force increments are given by in which
Dd€i ¼ d€iþ1 d€i ; Dd_ i ¼ d_ iþ1 d_ i ; Dd i ¼ d iþ1 d i
Downloaded from ascelibrary.org by University of Nebraska-Lincoln on 08/16/16. Copyright ASCE. For personal use only; all rights reserved.
(28)
^ ¼Kþ g Cþ 1 M
K (32)
λDt λDt2
Using the defined increments, Eq. (26) can be rewritten as
Dd_ i ¼ Dtd€i þ ð g DtÞDd€i (29) M gC _ M g
^ i ¼ DPi þ
DP þ di þ þ Dt 1 d€i (33)
or λDt λ 2λ 2λ
Fig. 11. (Color) Transverse acceleration responses under uniform Chichi earthquake excitation at Point B (see Fig. 7) on the deck
Fig. 12. (Color) Transverse displacement responses under uniform Chichi earthquake excitation at Point B (see Fig. 7) on the deck
General
The main span of the KSM Bridge consists of a composite steel/
As a case study, the KSM Bridge in Hong Kong was selected (Fig. concrete structure. The side spans are made of concrete box girders.
4). It is a massive long-span cable-stayed bridge structure with a Here, the deck is modeled as a flat shell in the integrated finite strip
Fig. 13. (Color) Longitudinal acceleration at Point E (see Fig. 7) at the west tower for velocity = 500 m/s
Fig. 14. (Color) Longitudinal acceleration at Point E (see Fig. 7) at the west tower for velocity = 1,000 m/s
knot has only three translational DOF, and the rotational DOF are not as field measurement tests for vibration of the deck (Lau et al. 1999).
taken into account. The four piers of the bridge are also modeled by 1D The finite element analysis was performed by SAP2000 software. In
column strips. The boundary conditions for the towers and the piers are the SAP model, the deck is modeled as thin shell elements, the piers
all fixed at the connection with the ground. The 176 cables are assumed and the towers are modeled as beam elements, and for the cables,
to be linear elastic elements and are modeled by a cable strip in the finite truss elements are assigned. As can be seen clearly in Table 2, there
Fig. 15. (Color) Longitudinal acceleration at Point E (see Fig. 7) at the west tower for velocity = infinite
Fig. 16. (Color) Longitudinal displacement at Point E (see Fig. 7) at the west tower for velocity = 500 m/s
Fig. 17. (Color) Longitudinal displacement at Point E (see Fig. 7) at the west tower for velocity = 1,000 m/s
Fig. 18. (Color) Longitudinal displacement at Point E (see Fig. 7) at the west tower for velocity = infinite
incoherence effect, and the local site effect, may play an important vibrate under the same ground motion. An artificial sine wave and a
role in the seismic response of the long-span bridges. The wave-pas- real seismic wave are selected and applied to the bridge structure,
sage effect is a dynamic phenomenon in which seismic waves arrive and their effects are investigated using the IFSM.
Fig. 19. (Color) Vertical components of seismic waves: (a) Input 1; (b) Input 2
Fig. 20. (Color) Longitudinal components of seismic waves: (a) Input 1; (b) Input 2
Fig. 21. (Color) Vertical displacement responses at Point A (see Fig. 7) on the deck: Input 1 þ Input 1 (see Fig. 19)
Fig. 22. (Color) Vertical displacement responses at Point A (see Fig. 7) on the deck: Input 2 þ Input 2 (see Fig. 19)
Fig. 23. (Color) Vertical displacement responses at Point A (see Fig. 7) on the deck: Input 1 þ Input 2 (see Fig. 19)
Fig. 24. (Color) Longitudinal displacement responses at Point G (see Fig. 7) on the east tower: Input 1 þ Input 1 (see Fig. 20)
Fig. 25. (Color) Longitudinal displacement responses at Point G (see Fig. 7) at the east tower: Input 2 þ Input 2 (see Fig. 20)
Fig. 26. (Color) Longitudinal displacement responses at Point G (see Fig. 7) at the east tower: Input 1 þ Input 2 (see Fig. 20)
Subsequently, the interactions between the bridge deck (superstruc- De Silva, C. W. (2005). Vibration and shock handbook, CRC Press, Boca
ture) and the other components could not be considered when using Raton, FL.
the conventional finite strip methods. By using the finite strip solu- Dumanoglu, A. A., and Severn, R. T. (1990). “Stochastic response of sus-
pension bridges to earthquake forces.” Earthquake Eng. Struct. Dyn.,
tion proposed in the current research, the entire bridge can be mod-
19(1), 133–152.
eled as an integrated system, in which the structural interactions Heredia-Zavoni, E., and Vanmarcke, E. H. (1994). “Seismic random-
between the structural elements can be considered also. vibration analysis of multisupport-structural systems.” J. Eng. Mech.,
In this study, the development of the IFSM was briefly pre- 10.1061/(ASCE)0733-9399(1994)120:5(1107), 1107–1128.
sented. Using the stiffness and mass matrices derived from IFSM, Humar, J. (2012). Dynamics of structures, 3rd Ed., CRC Press, Boca
the time history analysis of cable-stayed bridges was performed by Raton, FL.
the Newmark method. The proposed dynamic finite strip analysis Kiureghian, A. D., and Neuenhofer, A. (1992). “Response spectrum method
was formulated and programmed by comprehensive computer lan- for multi-support seismic excitations.” Earthquake Eng. Struct. Dyn.,
guages and programs, including Cþþ and MATLAB. The accuracy 21(8), 713–740.
Lau, C. K., Mak, W. P., Wong, K. Y., Man, K. L., Chan, W. Y., and Wong,
of the proposed technique was evaluated by comparing the natural
K. F. (1999). “Structural performance measurement and design parame-
frequencies of the KSM Bridge obtained by the finite strip method ter validation for Kap Shui Mun cable-stayed bridge.” Proc., 2nd
with those obtained by FEM, and those reported from the field-test Annual Int. Conf. on Advances in Steel Structures, Vol. 1, Elsevier,
measurements, and were found to be in very good agreement. The Amsterdam, the Netherlands, 505–512.
difference in values was lower than 4.8 and 10%, respectively. The Lau, D. T., Cheung, M. S., and Cheng, S. H. (2000). “3D flutter analysis of
seismic analysis of the KSM Bridge, under uninform and nonuni- bridges by spline finite strip method.” J. Struct. Eng., 10.1061
form excitations, used within the IFSM, showed that the displace- /(ASCE)0733-9445(2000)126:10(1246), 1246–1254.
ment and acceleration responses at different locations along the Lee, M., and Penzien, J. (1983). “Stochastic analysis of structures and pip-
bridge were higher for uniform earthquake excitations, which is ing systems subjected to stationary multiple support excitations.”
Earthquake Eng. Struct. Dyn., 11(1), 91–110.
also consistent with the results obtained from the FEM analysis of
Li, W. Y. (1988). “Spline finite strip analysis of arbitrarily shaped plates
the same bridge. and shells.” Ph.D. thesis, Dept. of Civil Engineering, Univ. of Hong
The computational time is a very important factor in the seismic Kong, Hong Kong.
analysis of long-span bridges. Because of the considerable reduction Lin, Y. K., Zhang, R., and Yong, Y. (1990). “Multiply supported pipeline
in the DOF, the developed code for the IFSM is efficient enough in under seismic wave excitations.” J. Eng. Mech., 10.1061/(ASCE)0733
complicated analyses of massive structures, such as dynamic analy- -9399(1990)116:5(1094), 1094–1108.
sis of the KSM Bridge. Complete modeling of a long-span cable- MATLAB [Computer software]. Mathworks, Natick, MA.
stayed bridge using IFSM and application of the time domain method Naderian, H., Cheung, M. M. S., Shen, Z., and Dragomirescu, E. (2015).
suggest that the seismic analysis of a complicated structure can be “Integrated finite strip analysis for long-span cable-stayed bridges.”
Comput. Struct., 158, 82–97.
performed in a minimal amount of time. The numerical results show
Prenter, P. M. (1975). Splines and variational methods, Wiley, New York.
that the convergence and efficiency of the IFSM is very high in com- SAP2000 [Computer software]. Computers and Structures, Inc., Walnut
parison with those of other numerical methods. Creek, CA.
Saxena V., Deodatis G., and Shinozuk M. (2000). “Effect of spatial variation of
earthquake ground motion on the nonlinear dynamic response of highway
References bridges.” Proc., 12th World Conf. on Earthquake Engineering, New
Zealand Society for Earthquake Engineering, Upper Hutt, New Zealand.
Berrah, M., and Kausel, E. (1992). “Response spectrum analysis of struc- Shen, Z., Cheung, M. M. S., Naderian, H., Dragomirescu, E. (2013). “An inte-
tures subjected to spatially varying motions.” Earthquake Eng. Struct. grated finite strip solution for dynamic analysis of continuous multi-span
Dyn., 21(6), 461–470. bridges.” Proc., 3rd Specialty Conf. on Mechanics and Materials, Canadian
Chang, C. C. (2004). Earthquake engineering, lecture notes, Dept. of Civil Society for Civil Engineering, Montréal.
and Environmental Engineering, Hong Kong Univ. of Science and Wang, S., and Zhang, Y. (2005). “Vibration analysis of rectangular compos-
Technology, Hong Kong. ite laminated plates using layerwise B-spline finite strip method.”
Cheng, S. H., Lau, D. T., and Cheung, M. S. (2003). “Comparison of numer- Compos. Struct., 68(3), 349–358.
ical techniques for 3D flutter analysis of cable-stayed bridges.” Comput. Zhang, Q. W., Chang, T. Y. P., and Chang, C. C. (2001). “Finite-element
Struct., 81(32), 2811–2822. model updating for the Kap Shui Mun cable-stayed bridge.” J. Bridge
Cheung, M. M. S., Shen, Z., and Chan, B. Y. B. (2009). “Integrated finite Eng., 10.1061/(ASCE)1084-0702(2001)6:4(285), 285–293.
strip solution for box girder bridges and slab-on-girder bridges.” J. Zhong, W. X., and Williams, F. W. (1995). “A precise time step integration
Comput. Model Eng. Sci., 45(2), 155–177. method.” Proc. Inst. Mech. Eng., 208C(6), 427–430.