Professional Documents
Culture Documents
Track-Bridge Interaction Analysis Using Interface Elements Adaptive To Various Loading Cases PDF
Track-Bridge Interaction Analysis Using Interface Elements Adaptive To Various Loading Cases PDF
Abstract: In this paper, a numerical model using interface elements is proposed for the sequential analysis of nonlinear track–bridge interac-
tion by taking the loading history into consideration. Longitudinal resistance tests were conducted on track fastening systems to determine
changes in track resistance that occur at the instant of application or release of a vertical load. On the basis of the test results, longitudinal track
stiffness laws for a nonballasted track were established for different loading cases including constant vertical load, sudden application of verti-
cal load, and sudden release of vertical load. In addition, an interface element adaptive to various loading cases that can be incorporated into a
conventional FEM was formulated for modeling the track resistance. Furthermore, a numerical algorithm for a practical solution of the nonlin-
ear system equation was developed. The validity and applicability of the proposed analysis method were verified by conducting extensive pa-
rameter studies on a simply supported bridge and a multispan bridge. DOI: 10.1061/(ASCE)BE.1943-5592.0000916. © 2016 American
Society of Civil Engineers.
Author keywords: Nonlinear track–bridge interaction; Continuous welded rail; Track stiffness law; Longitudinal track resistance;
Loading history.
ance tests by applying a specified vertical load before the applica- lasted track, it is necessary to conduct a longitudinal resistance test
tion of the longitudinal force and maintained it unchanged for the of the fastening system.
entire duration of the test. The track resistance changes suddenly To establish the longitudinal behaviors of a fastening system
when a vertical load is applied or released, and a restoring force can under the conditions of different types of vertical loadings including
be generated accordingly. Thus, it is difficult to make realistic sudden change in the vertical load, a test apparatus for controlling
assumptions about such a complex phenomenon. This makes it nec- the longitudinal and vertical loads according to the test schedule
essary to verify the assumptions that have been made about the was installed. Fig. 1 shows a schematic and photograph of the test
loading–unloading paths when a vertical load is suddenly applied or apparatus. The length of the tested segment of the rail was 500 mm.
released. In addition, appropriate track stiffness laws need to be When applying the longitudinal and vertical loads together, fric-
established. tional resistances may be generated at the vertical contact surface
In the present study, a numerical model using interface ele- between the longitudinal rod and the rail end and at the horizontal
ments was developed for use in the sequential analysis of nonlin- contact surface between the vertical rod and the top surface of the
ear TBIs by taking the loading history into consideration. rail. This may produce errors in the measurements of the longitudi-
Longitudinal resistance tests were conducted on the fastening sys- nal and vertical forces. To minimize the frictional resistances,
tems to determine the changes in the track resistance at the instant grease was applied to the vertical surface of the rail end and a roller-
a vertical load is applied or released. On the basis of the test zig was installed at the end of the vertical rod. Two displacement
results, constitutive laws regarding the track–bridge interface of a transducers were installed parallel to the rail at the front and rear
nonballasted track were established for different loading cases sections, respectively, and the rail displacement was determined as
including constant vertical load, sudden application of vertical the average of the measurements of the two transducers.
load, and sudden release of vertical load. An interface element for The longitudinal resistance test of fastening systems was con-
modeling the track resistance was also formulated. The interface ducted in accordance with EN 13146-1 (CEN 2012), which requires
element is adaptive to various loading cases and can be readily force control. However, after reaching the plastic state beyond the
assembled with the finite elements of the rail and bridge. A numer- elastic limit, the rail displacement increases drastically under the
ical algorithm that can be used to solve the nonlinear system equa- controlled force, and this makes it difficult to apply the next vertical
tion was also developed. The validity and applicability of the pro- load. Hence, to measure the longitudinal resistance of the fastening
posed numerical model were verified by conducting extensive system under the train load (vertical load) after the application of
parameter studies on a simply supported bridge and a multispan the temperature load (longitudinal load), it was necessary to apply
bridge, respectively, and comparing the results with those obtained the longitudinal load under controlled displacement. To simulate
by a separate analysis method. very slow application of the temperature load, the longitudinal load
Vertical Loading
(force control) Roller zig
Longitudinal Loading
(Displacement control)
Rail
Grease
Fastening system
(a) (b)
Fig. 1. Longitudinal resistance test of fastening system under a controlled vertical load: (a) schematic of test apparatus; (b) photograph of test
apparatus
tem with a baseplate has a hard rail pad, and its elasticity is mainly critical elastic displacement was not much larger than 1.0 mm.
provided by the resilient baseplate pad (resilient baseplate pad Therefore, as shown in Fig. 3, the maximum longitudinal displace-
type). Conversely, the elasticity of the fastening system without a ment during the test under vertical loading was set to 3 mm so as that
baseplate is provided by the resilient rail pad (resilient rail pad the fastener was in the plastic state before and after application of the
type). Fig. 2 shows examples of the resilient baseplate pad and resil- vertical load. (2) The next step is to apply a linearly increasing verti-
ient rail pad type fastening systems, which are commonly used in cal load from 0 to 30 kN over 1 s (Dt = 1 s) without applying addi-
nonballasted tracks of high-speed and main lines. The longitudinal tional longitudinal rail displacement. (3) The last step is to increase
resistance tests were conducted to assess the longitudinal resistance displacement of the rail at the same speed as in Step 1 until the plastic
characteristics of the two types of fastening systems. state was attained. Fig. 3(a) shows the variations in the longitudinal
The loaded test was conducted by applying a specified vertical displacement and vertical load with time. Although the duration of
load (30 kN or 60 kN) before the application of a longitudinal force, the longitudinal load in the unloaded state during the loaded to
and both were maintained for the entire duration of the test. In the unloaded state test was set to 3,000 s, the results showed that the plas-
unloaded test, only the longitudinal force was applied; there was no tic state was attained within 10 min [see Fig. 5(a)]. Thus, in the from
vertical load. The vertical load of 30 kN used for the loaded test is loaded to unloaded state test, the loading duration was reduced to 30
the approximate value of the vertical load to which a resilient fas- min (1,800 s) to save time.
tening system of a nonballasted track is generally subjected. The
maximum axle load of high-speed railway lines in Korea is 17 t, Longitudinal Rail Restraint Test under Sudden Change from
and the wheel load is approximately 85 kN. The wheel load is dis- Loaded to Unloaded State
tributed by the flexural rigidity of the rail, and the load on a single A longitudinal rail restraint test was conducted to investigate the
fastener can be determined by the well-known theory of a beam on variation in the longitudinal resistance and the restoring force when
an elastic foundation (Esveld 2001): a vertical load was suddenly released. The longitudinal and vertical
loads were applied in stages as follows: (1), (2), and (3) in the test of
P ¼ W ½1 e b a=2 cos ðb a=2Þ (1) sudden application of a vertical load were implemented; (4) the ver-
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi tical load was released in the plastic state over a short time (Dt = 1 s)
where b ¼ 4 ðk=aÞ=ð4EIÞ; P is the load on a single fastener; W is without applying additional longitudinal rail displacement; and (5)
the wheel load; k is the vertical spring constant of the fastener; a is the displacement of the rail was allowed again in the unloaded state.
the distance between fasteners; E is the modulus of elasticity of the Fig. 3(b) shows the variations in the longitudinal displacement and
rail; and I is the second moment of inertia of the sectional area of the vertical load with time for this case.
the rail. For the Korean high-speed railway, it can be assumed that
EI = 6,415 kN · m2 (UIC 60 rail), a = 0.625 m, and k = 30 103 kN
Test Results
· m; hence, P 0.351W 29.9 kN. Based on this, 30 kN was
selected as the vertical load for the loaded test. A vertical load of 60 Unloaded and Loaded State Tests
kN was also used to investigate the variation in the track resistance From the test results shown in Fig. 4, it can be seen that the resistance
with an increase in the vertical load. increases elastically for small displacements and remains constant
Rail pad
ڿڼۋڈۇۄڼڭ
yT
Rail pad
(a) ڿڼۋڈۀۏڼۇۋۀێڼڝ
Baseplate pad ۀۏڼۇۋۀێڼڝ
Baseplate (b)
Fig. 2. Types of fastening systems used in nonballasted tracks: (a) resilient baseplate pad type; (b) resilient rail pad type
5
3
Time(s) Time(s)
Force (kN)
Force (kN)
Loaded Loaded
30 30
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
Unloaded Unloaded
3000 s Time(s) 1800 s 1800 s Time(s)
Δt = 1 s
(a) (b) Δt = 1 s Δt = 1 s
Fig. 3. Variations in longitudinal and vertical loads with time under sudden change in vertical train load: (a) from unloaded to loaded state; (b) from
loaded to unloaded state
30 30
25 25
Loaded (60kN)
Longitudinal resistance (kN)
20 20
Unloaded
10 10
Unloaded
5 5
0 0
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9
Fig. 4. Results of longitudinal rail restraint tests of fastening systems under unchanged vertical load: (a) resilient baseplate pad type; (b) resilient rail
pad type
35 15
1st test 1st test
2nd test Vertical load 14
2nd test
30
3rd test 13 3rd test
11
20
Load (kN)
10
15
9
7.0 kN
1.7 kN
8
10 Longitudinal resistance
2.0 kN
7
2.0 kN
5 Δu
6
Δu
Δu
0 5
0 10 20 30 40 50 60 70 4.7 4.8 4.9 5.0 5.1 5.2 5.3 5.4 5.5 5.6
Fig. 5. Results of longitudinal rail restraint tests of fastening systems under sudden change from unloaded to loaded state: (a) longitudinal and vertical
loads versus time; (b) longitudinal resistance versus displacement
35 25
1st test 1st test
2nd test Vertical load 2nd test
30 Δu
20 Δu
Longitudinal resistance (kN)
25
10.6 kN
Longitudinal resistance
10.5 kN
15
20
Load (kN)
9.0 kN
9.0 kN
15 10
1.4 kN
2.7 kN
10
5
5
Δu : longitudinal displacement of rail during vertical loading time Δt
0 0
0 10 20 30 40 50 60 70 80 5.85 5.90 5.95 6.00 6.05 6.10 6.15
Fig. 6. Results of longitudinal rail restraint tests of fastening systems under sudden change from loaded to unloaded state: (a) longitudinal and vertical
loads versus time; (b) longitudinal resistance versus displacement
However, Fig. 6(b) shows that when the longitudinal rail displace- tances in the track stiffness law of the previous loading case; and Fi
ment begins to increase again after the vertical load is released the is the track resistance induced by the previous loading case.
longitudinal resistance suddenly increases by 1.4–2.7 kN, returning Furthermore, the minimum and maximum elastic displacements
to the plastic longitudinal resistance in the unloaded state. This is are given by the following:
caused by the recovered friction between the top of the rail foot and
min ¼ umin ^
uiþ1 i
the fastening clip. ui (4)
The results of this test reveal that when the vertical load is
released while exhibiting the plastic behavior in the loaded state the
max ¼ umax ^
uiþ1 i
ui (5)
stiffness of the longitudinal resistance momentarily changes to that
of the unloaded state, and a restoring force is generated because of
where uimin and uimax are the minimum and maximum elastic dis-
the difference between the longitudinal resistances in the loaded
placements in the track stiffness law of the previous loading case;
and unloaded states. Thus, as also suggested by Ruge and Birk
and ^
u i is defined as follows:
(2007), in a sequential analysis of the TBI it is necessary to consider
the restoring force that is generated when the vertical load is 8
< uo ui uo
released in the plastic zone in the loaded state. ^
ui ¼ ui uo ui < uo (6)
:
uo ui < uo
Establishment of Longitudinal Track Stiffness Law for
Nonballasted Track where ui is the relative displacement of the rail and bridge generated
by the previous loading case. The four parameters calculated by
Eqs. (2)–(5) completely determine the track stiffness law for the
Unloaded and Loaded States
next loading case.
The UIC 774-3-R (UIC 2001) notes that a bilinear function can be
used to represent the track resistance with a level of precision suffi-
cient for calculation purposes. Therefore, in the present study, the Sudden Change from Vertically Unloaded to Vertically
longitudinal track stiffness law of the track resistance was expressed Loaded State
as a bilinear function. To establish the track stiffness law for the case of sudden applica-
The track stiffness law established from the test results in Fig. 4
tion of a vertical load, i.e., sudden change from the unloaded to the
for the case of no preloading (initial loading case) is shown in
loaded state, the following assumptions were made based on the test
Fig. 7. If the vertical loading condition of the next loading case is
results in Fig. 5. (1) For conservative evaluation and simplified cal-
the same as that of the previous loading case, the track stiffness law
culation, the slight reduction in the longitudinal resistance at the
can be established using three different patterns, depending on
instant of application of the vertical load is ignored. (2) The track re-
whether the track resistance after the previous loading case lies
sistance after application of the vertical load follows the loaded re-
sistance-displacement curve, and the plastic resistance increases
from Fu to Fl .
2uo Based on these assumptions, the track stiffness law was estab-
positive lished for three different patterns (Fig. 9). In Fig. 9, the origin of the
F imax ྛ plastic zone coordinate is shifted to the point (ui ; Fi ) corresponding to the track
resistance and displacement induced by the previous loading case.
Therefore, the equations of the minimum and maximum plastic
iþ1 iþ1
2F u or 2F l
uimin
resistances (Fmin and Fmax , respectively) in the track stiffness law
uimax
ې are as follows, regardless of whether they are in the elastic or plastic
ྚ elastic zone zone:
F imin
iþ1
Fmin ¼ ðFl Fu Þ þ Fmin
i
Fi (7)
ᐭ
negative iþ1
Fmax ¼ Fl Fu þ Fmax
i
Fi (8)
plastic zone
Furthermore, the minimum and maximum elastic displacements
Fig. 7. Track stiffness law for the case of no preloading
are given by the following:
Fi
min max
ې
2F u or 2F l
Fi
2F u or 2F l
2F u or 2F l
ui ui
ui
Previous
Previous loading case
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
Fi
loading case
ې F i+1
min F i+1
min
F i+1
min ui+1
min
ui+1
max
(a) (b) (c)
Fig. 8. Track stiffness law for the case of unchanged vertical load from the previous loading case: (a) negative plastic zone; (b) elastic zone; (c) posi-
tive plastic zone
F i+1
max
F i+1
max F i+1
max
F l-F u
F l-F u
F l-F u
Fi
ui+1
min
max
ې
Fi
ui ui+1
max ui
2F u
2F u
ui
2F u
ui+1
min Previous loading case
Previous loading case
Fi
ui+1
F l-F u
2uo max
F i+1
min
F i+1
min
F i+1
min
(a) (b) (c)
Fig. 9. Track stiffness law for the case of sudden change from the unloaded to the loaded state: (a) negative plastic zone; (b) elastic zone; (c) positive
plastic zone
F l-F u
Fl-F u
F l-F u
Fi
ui+1 Fi+1
max Restoring force
Fi+1
max ui+1 Restoring force min
Fi
min
Fi+1
max ې ui+1
ui ui ui max
ui+1
max
2F u
2F u
2F u
O O O
Previous loading case
ui+1
min
ui+1
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
max
ې Fi+1
min Fi+1
min
Restoring force Fi+1
F l-F u
F l-F u
min
(Fl-Fu)
Fi
Fig. 10. Track stiffness law for sudden change from loaded to unloaded state based on the assumption of Scenario I: (a) negative elastic zone; (b) elas-
tic zone; (c) positive plastic zone
iþ1
and Fmax , respectively) in the track stiffness law are given by the ^
u i ¼ uimin for ui uimin (15)
following, regardless of whether they are in the elastic or plastic
zone:
^
u i ¼ uimax for ui uimax (16)
iþ1
Fmin ¼ ðFmin
i
Fi Þ=a (11)
The track stiffness law for unloaded and loaded states, Eqs. (2)–
iþ1
Fmax ¼ ðFmax
i
Fi Þ=a (12) (6), can be applied to Scenario II, because the loaded state of the
track resistance is unchanged.
The minimum and maximum elastic displacements are given by If the track resistance is larger than the unloaded plastic re-
sistance before the release of the vertical load, the track resist-
min ¼ umin ^
uiþ1 i
ui (13) ance would momentarily decrease to the unloaded plastic resist-
ance during the release of the vertical load. The restoring force
is generated by the difference between the track resistance for
max ¼ umax ^
uiþ1 i
ui (14) the previous loading case and the plastic resistance for the next
loading case, as shown in Fig. 10; the restoring force DC is thus
where given by
8
< Fl Fu ðFmax
i
Fi Þ for ui uo = a uo þ uimax
DC ¼ 0 for uo = a þ uo þ uimin ui < uo = a uo þ uimax (17)
:
ðFl Fu Þ ðFmin Fi Þ for ui < uo = a þ uo þ uimin
i
Numerical Formulation the rail and the top surface of the bridge deck; ur and utb are,
respectively, the longitudinal displacements of the rail and the
top surface of the bridge deck; ucb and f cb are, respectively,
Formulation of Stiffness Matrix of Interface Element
the longitudinal displacement and rotational angle of the
It can be assumed that the TBI force acts on both the neutral axis of neutral axis of the bridge deck; and h is the height from the
the rail and the top surface of the bridge deck. Thus, when the track neutral axis of the bridge deck to the top surface of the bridge
resistance is modeled by continuously distributed longitudinal deck.
springs that connect both the rail and the bridge, the interaction When modeling the rail and bridge deck using rod and beam ele-
force in the elastic state can be expressed as ments of the same length, respectively, as shown in Fig. 11, the rela-
tive displacement of the rail and bridge deck can be expressed in
F ¼ ct ur=b ¼ ct ður utb Þ ¼ ct fur ðucb þ h f cb Þg (18) terms of the shape functions and nodal displacements of the two ele-
ments as follows:
where ct is the spring coefficient of the longitudinal track
resistance per unit length; ur=b is the relative displacement of ur=b ¼ f Ue (19)
where f is the shape function vector, which can be expressed as fol- < 1; 1; 1; 0; h; 1; 0; h>T (26)
lows when the longitudinal displacements of the elements are
assumed to change linearly: where Fplastic is the longitudinal plastic resistance, which has a value
of Fmin or Fmax .
f ¼ < 1 j ; j ;ð1 j Þ; 0; ð1 j Þh; j ; 0; j h > (20) Whether the interface element is in the elastic zone or plastic
zone is determined by the relative displacement of the rail and the
Ue is the nodal displacement vector, which can be expressed as bridge deck at the center of the interface element
When maintaining a short element length within a certain Plastic zone: jucenter
r=b j uo (28)
range, the longitudinal spring resistance of the elements can be
assumed to be in either the elastic zone or the plastic zone. Such
an assumption hardly holds true when the stress changes suddenly The relative displacement at the center of the interface element
within a length of the element. However, the point at which the ucenter
r=b is obtained from Eq. (19) as
stress of the rail or the bridge deck suddenly changes on-site
ucenter
r=b ¼ fð j ¼ 1=2Þ Ue (29)
occurs at an expansion joint of the rail or the bridge or at the end
point of the distributed vertical load. Hence, because the nodes of
an element can be selected to prevent the point from occurring
inside the element, the preceding assumption can be used in ana-
lyzing the TBI. Numerical Algorithms
When the longitudinal resistance between the rail and the bridge The equilibrium equation for the first loading case of the track–
is in the elastic zone, it can be modeled by the interface element bridge system can be expressed as
sharing nodes of the rail and bridge elements. In such a case, the
interface element becomes a linear longitudinal spring, and the Krb U1 þ F1 ðU1 Þ ¼ P1 þ T1 (30)
potential energy Ep of the interface element can be expressed as fol-
lows based on Eqs. (18) and (19): where Krb is the stiffness matrix of the track and bridge; and
F1 ðU1 Þ is the force vector of the longitudinal resistance correspond-
ðL
1 1 ing to U1 in the track stiffness law of the first loading case, as calcu-
Ep ¼ ct ðut=b Þ2 dx ¼ ðUe ÞT Kef Ue (22) lated using Eq. (24) or Eq. (26). P1 and T1 are, respectively, the exter-
2 0 2
nal force vector and temperature load vector for the first loading case.
where the (i, j)th element of the stiffness matrix Kef is given by The equilibrium equation for the ith loading case is as follows:
ð1 X
i X
i X
i
where fi ð j Þ is the ith element of the shape function vector in The incremental equilibrium equation derived from Eq. (31) is
Eq. (20). as follows:
When applying the Lagrange equation to the potential energy of
Krb DUi þ Fi ðDUi Þ ¼ Pi þ Ti (32)
the interface element, the equilibrium equation of the interface ele-
ment is obtained as follows:
where
Kef Ue ¼ Feelastic (24) DUi ¼ Ui Ui1 (33)
where Feelastic is the nodal force vector of the interface element, and In Eq. (32), Pi and Ti are, respectively, the external force vector
the stiffness matrix Kef is as follows: and temperature load vector for the ith loading case.
X
n X
n
elastic and plastic zones in a single interface element. However,
U¼ DUi ; Se ¼ DSei (35)
such a scenario may not be encountered in practice for an element
i i
length of less than 1 m, and reduction in the length of the element is
thus recommended in the event of failure of convergence.
A nonlinear analysis is required to solve Eq. (32) for every load-
Fig. 12 shows the flowchart of the TBI analysis for cases of se-
ing case. For a track–bridge system in which the elements except
quential loading. The authors developed a FORTRAN program
the interface elements behave linearly and the interface elements
based on the procedure illustrated in the flowchart.
behave nonlinearly, and that complies with the bilinear functions,
the solution for each loading case can be feasibly obtained as
described next. Numerical Examples
All the interface elements are initially assumed to be in the elas-
tic zone. The displacement increments are obtained by solving the
Example 1: Simply Supported Bridge
equilibrium equation using the track stiffness law established by
analysis of the previous loading case. Whether an interface element Verification of Proposed Method
is in the elastic or plastic zone is determined based on the calculated The validity of the proposed method was verified by conducting
displacement increments. If an element moves from the elastic zone nonlinear analyses of a track–bridge (single-span bridge) system to
(i=1,2,..,N)
Setting the i-th load ( Pi + Ti )
Yes
Calculating the cumulative displacements
and stresses of the rail and bridge
End
at the embankment outside the bridge was maintained in the ferences between the present results and those for the second con-
unloaded state. The length of the interface element was set to 0.1 m dition of Ruge et al. (2009). However, Fig. 14 shows that the
as in the example of Ruge and Birk (2007). maximum stresses of the rail—which are important from the per-
Because the axial and bending stiffness of the bridge are much spective of engineering design—determined by the present
larger than those of the track, the structural function of the track
method are mostly equal to those obtained by Ruge et al. (2009)
could be neglected and all the deformations of the bridge could be
for the two analysis conditions.
assumed to be borne by only the axial and bending stiffness of the
bridge for a given load. The bridge deformation was first deter- Effects of Size of Interface Element
mined with respect to the load and was used as the prescribed dis-
To evaluate the effect of the length of the interface element on
placement in the TBI analysis. The TBI analysis of Ruge and Birk
the accuracy of the analysis, a parameter study was conducted
(2007) was also based on these assumptions. The same conditions
by varying the length of the interface element as 0.1, 0.5, 1.0,
that they used were applied in the first example of the present study
2.0, 5.0, and 10.0 m. Two different approaches were applied to
to enable comparison of the results. The conditions were artificially
the analysis. In the first approach separate independent analyses
realized by multiplying the elastic modulus of the bridge by a very
large factor (1,000) and increasing the vertical load by the same were conducted for the different loads, and the results were
factor. combined linearly. The other approach involved sequential
Fig. 14 compares the analytical results obtained by the present
method with those obtained by Ruge and Birk (2007), and there is Table 2. Geometrical and Material Properties and the Load Applied to the
very good agreement. Simply Supported Bridge
The second numerical example is the track resistance under sud- Description Property
den change from the unloaded state to the loaded stage by the appli-
cation of a vertical load. In this case, the sequential loading was Longitudinal axial stiffness of rail EA = 3.23 106 kN
implemented by temperature increment followed by the application Distributed stiffness of unloaded track (ct)u = 6.0 104 kN · m2
of a vertical load. The track resistance complied with the track stiff- Distributed stiffness of loaded track (ct)l = 12.0 104 kN · m2
ness law in Fig. 9 for the application of a vertical load after tempera- Critical elastic relative displacement uo = 5.0 10−4 m
ture loading and complied with the law in Fig. 8 for temperature Length of bridge L = 60 m
loading. Stiffness of elastic support KA = 6.0 105 kN · m
Fig. 15 compares the analytical results obtained by the present Distance from neutral axis to upper surface ho = 1.21 m
method with those obtained by Ruge et al. (2009). The dashed line Distance from neutral axis to lower surface hu = 4.79 m
represents the results of Ruge et al. (2009) for the following first Cross-sectional area of bridge AB = 1.7535 m2
condition, and the dotted-dashed line represents the results for the Moment of inertia of bridge IB = 16.5208 m4
second condition: Dynamic magnification factor of bridge UB = 1.05
1. Temperature increment followed by the application of a verti- Elastic modulus of bridge EB = 3.4 107 kN · m2
cal load (DT – vertical load); and Temperature change in bridge DT = þ30°C
2. Temperature increment followed by the application of a verti- Thermal expansion coefficient of bridge aB = 1.0 10−5
cal load together with the action of the restoring force [DT – Braking force p = 20 kN · m
(DC þ vertical load)]. Vertical load V = 80 kN · m
ho
hu
KA Bridge
L = 60 m
Fig. 13. Schematic of the simply supported bridge with single track
-20
Vertical load (q = 80 kN/m)
-40
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
-60
-80
Temperature→Vertical→Braking
-100
Bridge section
-120
-20 0 20 40 60 80 100 120
Distance (m)
Fig. 14. Longitudinal rail stresses determined by sequential analysis for the application of three loads (temperature increment – vertical load – brak-
ing force)
100
The presented method (ΔT → Vertical)
80 Ruge et al. (2009) (ΔT → Vertical)
Ruge et al. (2009) (ΔT → (ΔC + Vertical))
60
Longitudinal rail stress (MPa)
40
20
-20
-40
-60
-80
Bridge section
-100
-20 0 20 40 60 80 100
Distance (m)
Fig. 15. Longitudinal rail stresses obtained by sequential analysis for the application of two loads (temperature increment – vertical load)
analysis in which the correct sequence of the loads was taken the analysis of TBIs involving interface elements that exhibit the
into consideration. nonlinear behavior of a bilinear function.
Fig. 16 shows satisfactory agreement between the results for ele-
ment lengths within 1.0 m. The numbers in parentheses in Table 3 Effect of Considering Axial Stiffness of Rail in Evaluation of
indicate the numbers of iterations required for the analyses of the Bridge Deformation
temperature load, vertical load, and braking force, respectively. The In the two examples considered in the verification of the proposed
table shows that the shorter the element, the better is the agreement model, the bridge deformation was calculated by structural analysis
between the analytical results, but a greater number of iterations are of the bridge alone for a given load and treated as prescribed dis-
required. However, even if the interface element was shorter than placements for the TBI analysis. However, the TBI may cause
0.1 m, the maximum number of iterations required for the analysis the axial stiffness of the rail to affect the behavior of the bridge to
of a particular load would not exceed six. These results confirm that some extent. Hence, to investigate the effect on the rail stress when
the numerical solution procedure proposed in this paper is valid for the axial stiffness of the rail is considered in the evaluation of the
-20
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
-62
-64
-40 -66
-68
-60 -70
-72
-80 56 57 58 59 60 61 62 63 64
Bridge section
-100
-20 0 20 40 60 80 100
Distance (m)
Fig. 16. Longitudinal rail stresses with respect to the length of the interface element
Number of iterations
Analysis method Li = 0.1 m Li = 0.5 m Li = 1.0 m Li = 2.0 m Li = 5.0 m Li = 10.0 m
Separate analyses 14 (=6 þ 4 þ 4) 13 (=6 þ 4 þ 3) 13 (=6 þ 4 þ 3) 10 (=5 þ 3 þ 2) 10 (=5 þ 3 þ 2) 9 (=5 þ 3 þ 1)
Sequential analysis 16 (=6 þ 5 þ 5) 16 (=6 þ 5 þ 5) 15 (=6 þ 5 þ 4) 13 (=5 þ 4 þ 4) 13 (=5 þ 4 þ 4) 11 (=5 þ 3 þ 3)
Note: Li = the length of interface element. The numbers in parentheses are the numbers of iterations required for the analyses of the temperature load, vertical
load, and braking force, respectively.
bridge deformation, a numerical analysis was conducted using the investigating the effect of the increase in the rail stress. In Fig. 4, the
system data in Table 2 without an artificial increase in the elastic highest increase was 100% and a was thus assumed to vary between
modulus of the bridge and the vertical load. The track resistance 1.0 and 2.0.
was assumed to increase between the unloaded state and the loaded From Fig. 18, which shows the maximum tensile and compres-
state when the vertical load was applied. The length of the interface sive stresses of the rail with respect to a, the maximum compres-
element was set to 0.1 m. sive stress determined by separate analyses is constant regardless
Fig. 17 shows the longitudinal rail stresses with respect to of the value of a. Using sequential analysis the maximum com-
whether or not the effect of the axial stiffness of the rail on the pressive stress becomes constant after increasing by about 5.9% at
bridge deformation is considered, and the results obtained by con- a= 1.25. Conversely, the maximum tensile stress increases with
sidering the axial stiffness of the rail are somewhat lower. This increasing a for both analytical approaches. Regarding the com-
can be attributed to the structural function of the rail, namely, to pressive and tensile stresses, only the higher absolute maximum
constrain the displacement of the bridge during TBI. Table 4 gives between the two is generally important in the design of a nonbal-
the maximum values of the tensile and compressive stresses of the lasted track. Therefore, it should be noted that the effect of the ver-
rail obtained from Fig. 17. Table 4 shows that the maximum rail tical load is more pronounced in a sequential analysis because
stress when the axial stiffness of the rail was considered was both the tensile and compressive peak stresses increase with
underestimated by 1.9% using sequential analysis and by 5.1% increasing longitudinal resistance to a certain extent. On the con-
using separate analyses. Thus, it is necessary to consider the axial trary, the effect is less important in a separate analysis because the
stiffness in TBI analysis, especially when using the separate analy- compressive stress, which is higher, is unaffected by the increase
ses approach. of the longitudinal resistance.
Effect of Increase in Track Resistance Caused by Vertical Sudden Change from Vertically Loaded to Vertically Unloaded
Loading State
The results of the longitudinal rail restraint test of the fastening sys- As indicated in Fig. 6, the longitudinal resistance is reduced and a
tem shown in Fig. 4 reveal an increase in the longitudinal track re- restoring force is generated at the instant a vertical load is released.
sistance with increasing vertical load. Thus, in the TBI analysis of To investigate the variation in the rail stress when the vertical load
this study, the ratio of the longitudinal resistances in the loaded and is suddenly released, a TBI analysis was conducted with two differ-
unloaded states (a ¼ Fl =Fu ) was used as the parameter for ent assumptions about how the longitudinal track resistance
100
Considering
the axial stiffness of rail
80 Not considering
the axial stiffness of rail
40
20
Temperature
0
Fig. 17. Longitudinal rail stresses according to whether or not the axial stiffness of the rail is considered in the bridge deformation
Table 4. Maximum Rail Stresses with Respect to Whether or Not the in the longitudinal direction on the double track and the train load
Axial Stiffness of the Rail Is Considered in Determining the Bridge (comprising the vertical load and acceleration/braking forces) were
Deformation set up. Two load models were considered (Fig. 21) based on the
Sequential analysis Separate analyses movement direction of the train on each track. The length of the
interface elements of the analytical models was set to 1 m. The anal-
Maximum Maximum ysis was conducted using two different approaches, separate analy-
rail stress rail stress
ses and sequential analysis.
Maximum rail (MPa) (not Maximum rail (MPa) (not
stress (MPa) considering stress (MPa) considering Fig. 22(a) shows the longitudinal stress of the rail for each load-
Sign of rail (considering axial (considering axial ing scenario comprising a temperature increment, a vertical load,
stress axial stiffness) stiffness) axial stiffness) stiffness) and acceleration/braking forces when the train load was applied at
300 m (i.e., 300 m from the left A1-abutment. Fig. 22(b) shows the
Tension 72.3 69.4 86.3 81.0
rail stresses determined by sequential analysis and separate analy-
Compression −72.6 −71.2 −97.9 −92.9
ses. The numbers in parentheses in the legend of the figure indicate
the number of iterations required for convergence of the analysis of
changes after the sudden release of the vertical load. These two the respective loads. The number of iterations did not increase de-
assumptions correspond to Scenarios I and II, which were used to spite the great length of the bridge. The maximum rail stresses
establish the track stiffness law for sudden change from vertically determined by sequential analysis and separate analyses are 68.8
loaded to unloaded states. and 105.0 MPa, respectively. The separate analysis produced higher
Fig. 19 shows the results of the TBI for temperature increment rail stresses at some points.
followed by the application and release of the vertical load. The Fig. 23 shows the maximum longitudinal rail stresses deter-
dotted line in the figure corresponds to Scenario I, and the dotted- mined by sequential analysis and separate analyses when the
dashed line corresponds to Scenario II. The notations in parenthe- train load (Fig. 21) moved in steps of 1 m from 0 to 970 m while
ses in the legend indicate the longitudinal track resistance state the bridge temperature increased or decreased. With sequential
with respect to being unloaded (Fu ) or loaded (Fl ). From Fig. 19, it analysis the rail stress during increase in the bridge temperature
can be seen that, after the release of the vertical load, the rail stress attained a maximum value of 86.6 MPa when Load Model II
does not fully recover to the value before the application of the was at 630 m. With separate analyses the rail stress during
vertical load; a residual stress remains. This is caused by the non- decrease in the bridge temperature attained a maximum value of
linear characteristics of the longitudinal track resistance. However, 118.8 MPa when Load Model II was at 630 m. Fig. 23 shows
the difference between the rail stresses for Scenarios I and II is that the rail stresses determined by sequential analysis satisfy
insignificant. the permissible additional rail stress of 92 MPa prescribed in EN
1991-2 (CEN 2003), whereas those determined by separate anal-
yses far exceed the prescribed limit. A more economical track–
Example 2: Multispan Bridge
bridge system design is achieved by using sequential analysis to
The practical applicability of the proposed method to the design of a estimate the rail stress. However, Fig. 23 shows that both se-
track–bridge system was verified by conducting a TBI analysis of a quential analysis and separate analyses indicate that the variation
16-span bridge to be built on a conventional line. Fig. 20 shows the in the rail stress significantly depends on the position of the train
longitudinal section of the bridge, and Table 5 gives the system data load. This means that it would be difficult to intuitively predict
and properties. Two trains assumed to be located at the same point the position of the train load that maximizes the rail stress. Thus,
95 -95
Compression (MPa)
85 -85
Tension Tension
Tension (MPa)
80 -80
75 -75
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
70 -70
65 -65
60 -60
1.00 1.25 1.50 1.75 2.00
α ( = F l /Fu )
Fig. 18. Maximum longitudinal rail stresses with respect to a (ratio of the loaded track resistance Fl to the unloaded track resistance Fu )
100
ΔT(Fu)
80 ΔT(Fu)→Vertical(Fl)
ΔT(Fu)→Vertical(Fl)→ΔC(Fu)→-Vertical(Fu)
60 ΔT(Fu)→Vertical(Fl)→{ΔC(Fl) + -Vertical(Fl)}
Longitudinal rail stress (MPa)
40
20
-20
-40
-60
Bridge section
-80
-20 0 20 40 60 80 100
Distance (m)
Fig. 19. Longitudinal rail stresses for sudden change from vertically loaded to unloaded state
for worst-case design, it is necessary to conduct the TBI analysis characteristics of the track resistance with respect to the vertical
by varying the position of the train load. loading condition, especially cases of sudden change in the verti-
cal load. The findings were used to establish longitudinal track
stiffness laws for a nonballasted track. In addition, an interface
Conclusions element for modeling the nonlinear track resistance was devel-
oped, and a numerical algorithm for solving the nonlinear system
This paper proposed a numerical method using interface ele- equation was formulated.
ments for the sequential analysis of nonlinear TBI by taking the The proposed method was applied to some numerical exam-
loading history into consideration. Longitudinal resistance tests ples and good agreement was observed between the results and
were conducted on different fastening systems that are used in those obtained by Ruge and Birk (2007) and Ruge et al. (2009)
nonballasted tracks. The test results were used to investigate the based on an exact formulation of the stiffness of the track
19.0
21.5
26.5
28.5
28.5
25.5
29.5
29.5
29.5
28.5
30.5
31.5
31.5
35.0
35.5
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
Table 5. Geometrical and Material Properties and Load Data of the Multispan Bridge
resistance. The following were observed from the results obtained change from a vertically loaded state to a vertically unloaded state
for the considered numerical examples. (1) Sufficient accuracy as well as from a vertically unloaded state to a vertically loaded
from the perspective of structural design is obtained when the state. (5) The proposed method requires only a few seconds of
length of the interface element is within 1 m. In addition, in such computing time for the TBI analysis of a very long multispan
cases, the proposed numerical algorithm for solving the nonlinear bridge with double tracks subjected to three different types of
system equation requires no more than six iterations to analyze a loads.
particular load even when the DOF of the system is high. (2) For The results of the present study confirm the feasibility of using
practical evaluation of the longitudinal rail stress, it is unneces- the proposed method for the static analysis of nonlinear TBIs for
sary to consider the contribution of the axial stiffness of the rail to different types of bridges under different types of loads. The pro-
the TBI in the sequential analysis, but it is worth considering it in posed interface element may also be applicable to the dynamic anal-
the separate analysis approach. (3) The effect of an increase in the ysis of nonlinear TBIs.
track resistance caused by the application of a vertical load is When a train passes over a bridge, the vertical load varies with
more pronounced when evaluated by sequential analysis com- time. The longitudinal resistance of the fastening systems also
pared with the use of separate analyses. (4) The proposed model varies in accordance with the change in the vertical load and the rel-
can be feasibly applied to the sequential analysis of TBIs includ- ative displacement between the rail and bridge. However, in this
ing those caused by special longitudinal loading, such as sudden study the static analysis was performed with the assumption that the
Fig. 21. Train load models considering the train running direction on each track: (a) Load Model I; (b) Load Model II
80
Acceleration and braking forces, track2 subjected to braking force line (4)
40 ( ) : Number of iterations
20
-20
140
120
Sequential loading: ΔT Vertical load Braking force
100 Track1 subjected to acceleration force (14)
Track2 subjected to braking force (14)
Longitudinal rail stress (MPa)
-20
-40
Acceleration force
-60
Vertical load, Braking force
-80
Fig. 22. Longitudinal rail stresses when the train load is at 300 m: (a) for single load cases; (b) for combined loads
vertical load simultaneously acts on the entire bridge, and the actual accurate results, further investigation of the dynamic train-track-
variation of the axial stresses of the rail with time was not accurately bridge interaction is required to consider the variation of the longi-
calculated because it was generally overestimated. To obtain more tudinal resistance with time.
100 -100
80 -80
60 -60
Downloaded from ascelibrary.org by University of California, San Diego on 04/05/16. Copyright ASCE. For personal use only; all rights reserved.
40 -40
20 -20
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
(a) Position of train (m) Position of train loading (m)
160 -160
Sequential loading (-ΔT→Vertical→Bra.+Acc.) Sequential loading (-ΔT→Vertical→Bra.+Acc.)
Seperate loading (-ΔT + Vertical + (Bra.+Acc.)) Separate loading (-ΔT + Vertical + (Bra.+Acc.))
120 Load model I , Load model II -120 Load model I , Load model II
(a)
100 -100
80 -80
60 -60
40 -40
20 -20
0 100 200 300 400 500 600 700 800 900 1000 0 100 200 300 400 500 600 700 800 900 1000
(b) Position of train loading (m) Position of train loading (m)
Fig. 23. Maximum rail stresses with respect to the position of the train load: (a) temperature increment; (b) temperature decrement