Beruflich Dokumente
Kultur Dokumente
Senior Researcher
System Reliability Research Center,
Korea Automotive Technology Institute,
Estimation of Dynamic Contact
74 Yongjeong-ri, Pungse-myeon,
Dongnam-gu, Cheonan-si, Force Between a Pantograph
Chungnam 330912, South Korea
e-mail: spjung@katech.re.kr and Catenary Using the
Young Guk Kim
Chief Researcher Finite Element Method
High-Speed Rail Systems Research Center,
Korea Railway Research Institute, This paper presents a model of the catenary and pantograph and an analysis of their
176, Cheoldo Bangmulgwan-ro, dynamic interaction, using the finite element method. An analytical procedure to calcu-
Uiwang-si, Gyeonggi, 437757, South Korea late the length of droppers is introduced. The calculated dropper length is applied to the
e-mail: ygkim@krri.re.kr catenary model and the static deformation due to gravity is simulated. The presag result
of the contact wire is validated by comparison with the design specification. The wave
Jin Sung Paik propagation speed of the catenary model is acquired by applying the impact force to the
Senior Researcher contact wire. The result, regarding the wave propagation speed, agrees well with the
High-Speed Division, reference speed as defined by the UIC code. On the contrary, the pantograph model is
Korea Railway Research Institute, unified with the finite element catenary model, and the dynamic interaction of the
176, Cheoldo Bangmulgwan-ro, catenary-pantograph is simulated. An optimization technique to find the material
Uiwang-si, Gyeonggi, 437757, South Korea properties of the pantograph model is proposed. Based on the following performance test
e-mail: jspaik@krri.re.kr data, the optimum values of the material properties are found by using the response
surface analysis method. The reliability of the pantograph-catenary model is verified by
Tae Won Park1 comparing the contact force results obtained from the simulation and test. When the
Professor pantograph drives at 305 km/h, 370 km/h, and 430 km/h, the contact force variation and
Department of Mechanical Engineering, the possibility of loss of contact are discussed. [DOI: 10.1115/1.4006733]
Ajou University,
San 5, Woncheon-Dong, Yeongtong-Gu, Keywords: catenary, pantograph, finite element method, dropper length, dynamic
Suwon-City, 443749, South Korea interaction, optimization
e-mail: park@ajou.ac.kr
1 Introduction Presag is the initial deflection of the contact wire due to gravity,
and significantly affects the contact force variation. In addition,
Most railway trains employ electric traction. Electric power is
the movement of the pantograph induces flexural motion of the
transferred from an overhead catenary to the train via a panto-
catenary, and this flexural wave propagation also affects contact
graph that is mounted on its roof. Mechanically, the catenary and
force variation [7]. The displacement of the contact wire reaches
pantograph interact with each other through the contact force act-
its maximum when the driving speed of the pantograph equals the
ing on a collector strip on the head of the pantograph [1]. A panto-
wave propagation speed of the catenary. Therefore, the latter must
graph moving at high speed causes the vibration of the catenary
be carefully considered when the catenary model is simulated.
which, in turn, affects the contact force. As the trains speed
Ambrosio [8] and Rauter [9] performed a co-simulation of finite
increases, so does the variation in contact force, and contact may
element and rigid multibody dynamic codes. A linear finite ele-
even be lost [2]. The standard EN 50,119:2001 [3], which is the
ment code was used to analyze the behavior of the catenary, while
international standard in the railway industry, therefore requires a
a multibody code was used to describe the dynamics of the panto-
numerical analysis of the contact force variation when a new cate-
graph. A contact force model was developed for the interaction
nary system is designed.
between the catenary and the pantograph. This approach produces
Several studies have been devoted to estimating the contact
good simulation results, which follow the experimental data suc-
force variation between the catenary and pantograph, and most
cessfully. However, no validation was done for the presag or
recent studies are based on the finite element method (FEM). Seo
wave propagation speed of the catenary.
[4,5] created a catenary model using the finite element absolute
Cho [10] created a catenary-pantograph model using the FEM,
nodal coordinate formulation (ANCF) [6]. The ANCF beam ele-
and suggested a new formulation for a geometrically nonlinear
ments were used to model the catenary, and a rigid multibody
dropper. Through a few experiments, Cho validated the reliabil-
pantograph model was created. A sliding joint element was devel-
ity of the dropper model and showed that the wave propagation
oped to connect the pantograph to the catenary. Seo validated the
speed of the contact wire follows the experimental result well.
reliability of the model by comparing the simulation result with
Cho proved that the amplitude of the contact force variation
test data. However, the stagger, which is the enforced displace-
changes according to the trains speed, and an increase in the
ment of the contact wire to reduce the wear of the collector strip,
uplift force can reduce the amplitude of the contact force varia-
was not described in the model. Also, the presag and wave propa-
tion. The presag of his catenary model, however, was not
gation speed of the catenary were not verified.
validated.
Recently, custom analysis programs [11,12] for the catenary-
1
Corresponding author. pantograph have been developed. Simulation results have been
Contributed by the Design Engineering Division of ASME for publication in the
JOURNAL OF COMPUTATIONAL AND NONLINEAR DYNAMICS. Manuscript received October
shown to match well with test data. The limitation of those pro-
28, 2011; final manuscript received April 2, 2012; published online June 13, 2012. grams is that they cannot calculate the proper length of droppers,
Assoc. Editor: Jose L. Escalona. and is treated merely as an input condition, as in previous other
In Eq. (6), the maximum presag is unique for a given span, weight x1
Tm c1 RA x1 wm x1 0 (15)
per unit length, and tension. The maximum presag, however, is vari- 2
able since the vertical forces F1,F2,, and Fn are changed according
to the length of droppers. During construction of the catenary, the
presag of the contact wire is controlled by adjusting the length of
droppers, while the maximum presag of the contact wire is given as
a design specification. Using Eq. (6), Eq. (5) can be re-written as Eq.
(7), which is a function of the maximum presag dc. Equation (7) is
generally used when designing the catenary [21]
4dc
di xi x1 S1 xi x1 (7)
S1 2
Additionally, the presag of the messenger wire at (c1) is where E is the encumbrance defined in Fig. 3. Thus, by substituting
Eqs. (7) and (19) into Eq. (20), the length of a dropper can be
1 wm x1 2 calculated as
c1 RA x1 (16)
Tm 2 !
1 X
i1
w m xi 2
li E R A xi Fk xi xk
In Fig. 5(b), the moment equilibrium at the location of the second Tm k1
2
dropper () is (21)
4dc
2 xi x1 S1 xi x1 0
x2 S1
Tm c2 RA x2 F1 x2 x1 wm x2 0 (17)
2
3 Verification of the Catenary Model
Additionally, the presag of the messenger wire at (c2) is
Figure 6 shows the geometrical information of a catenary to
calculate the length of droppers. The span length is 50 m, eight
1 wm x2 2
c2 RA x2 F1 x2 x1 (18) droppers are attached between the messenger and contact wires,
Tm 2 and the distance between droppers is shown. Table 1 shows the
Component element Sectional area (mm2) Elastic modulus (GPa) Poissons ratio Mass density (kg/m) Damping
properties of the catenary. The tension and density of the contact 3.3 Validation of the Wave Propagation Speed of a
and messenger wires are shown and the weight of a dropper is Catenary. The wave propagation and reflection of a catenary
assumed to be 1.0595 N. Here, dc is the maximum presag set to have significant effects on the dynamic interaction between the
the design specification. catenary and pantograph. Since resonance of the contact force
occurs when the driving speed of the train is equal to the wave
3.1 Finite Element Model of a Catenary. Figure 7 shows a propagation speed of the catenary, design specifications of the cat-
finite element model of a single span of a catenary. Three dimen- enary such as the tension, presag, and wave propagation speed
sional beam elements are used to model the messenger and con- decide the maximum driving speed of a train. Thus, an exact
tact wires, and the length of each beam element is 0.05 m. Thus, a estimation of the wave propagation speed of the catenary is very
single span of the catenary is composed of 2000 beam elements. important. However, calculation of the wave propagation speed of
Table 2 shows the sectional area and material properties of the the catenary is difficult since the catenary consists of many com-
beam element. Each dropper is modeled by using a tension-only ponents. In this study, a simplified equation, defined by the UIC
spring whose stiffness is 50 kN/m. The tension is activated when standard, which is the international standard maintained by the
the length of the spring element reaches the calculated dropper international union of railway, is used as the reference wave prop-
length. Ground spring elements, which are vertical springs agation speed of the catenary. The reference wave propagation
between the ground and object, are used to constrain the messen- speed is compared with the wave propagation speed of the simula-
ger and contact wires, and their stiffness is obtained as previously tion model, and then the reliability of the simulation model is vali-
described in Ref. [15]. The ground spring element which con- dated. In UIC 799 OR, the wave propagation speed is defined as
strains the contact wire represents the registration arm of the s
actual catenary system. Initial tensions of 14 kN and 20 kN are Tm Tc
applied to both ends of the messenger and contact wires, as shown C (22)
qm qc
in Fig. 7. Static deformation of the contact wire is obtained when
a gravitational acceleration of 9.81 m/s2 is applied to the catenary.
The length of a dropper (li) was calculated using Eq. (21) and where T and q are the tension and mass density, and the subscripts
applied to the catenary model. Table 3 shows the calculated length m and c represent the messenger and contact wires, respec-
of each dropper, and one can see that all values are symmetric tively. In the case of the catenary shown in Fig. 6, the reference
with respect to the 4th and 5th droppers.
Table 3 Calculated results for the dropper length
3.2 Validation of the Presag of a Catenary. Figure 8 shows
the static analysis results of the catenary model. In Fig. 8, the Dropper 1st 2nd 3rd 4th 5th 6th 7th 8th
maximum presag of the contact wire is 0.0247 m, and has an
error of 1.2% compared with the design specification of 0.025 x (m) 4.5 11.25 18.0 22.75 27.25 32.0 38.75 45.5
(Span/2000 shown in Table 1). This verifies the presag of the li (m) 1.2501 1.1249 1.049 1.0251 1.0251 1.049 1.1249 1.2501
finite element model of a catenary.
P2 P1 325 275
V1 3:6 486:5 km=h
t2 t1 1:3643 0:9763
Fig. 10 Displacement variation of P1, P2, and P3 due to the
impact force
P3 P2 375 325
V2 3:6 486:5 km=h
t3 t2 1:7163 1:3643
(23)
SFY
Rx (24)
SFF
Since the static life force Fs is constant and does not affect the
vibration characteristic of the pantograph, Fs can be neglected.
Thus, the receptance of the three-DOF pantograph model can be
calculated as
Y1 s CE D2
R1 s
Fc s ACE AD2 EB2
where A m1 s2 c1 s k1
Fig. 14 Three DOF model for the following performance B c1 s k1
simulation (27)
C m2 s2 c1 c2 s k1 k2
stiffness of a spring, c is damping, Fc is the contact force, Fs is the D c2 s k2
static lift force, and y is the generalized coordinate. E m3 s2 c2 c3 s k2 k3
According to Newtons 2nd law, the equations of motions of
the three rigid bodies are derived as
By substituting jx instead of s in Eq. (27), the receptance can
m1 y1 c1 y_1 y_2 k1 y1 y2 Fc 0 be obtained as a function of the frequency.
Fig. 15 Comparisons of the following performance test and simulation results: (a) response at
0.5 Hz: 40 mm, (b) response at 6.5 Hz: 2 mm, (c) response at 9.5 Hz: 0.5 mm, and (d) response at
10 Hz: 0.9 mm
Y abs f 1 simulation f 1 test abs f 2 simulation f 2 test (28) minimize the objective function. This is assumed by the results of
the sensitivity analysis, however, the increase or decrease of
where abs is the absolute value, f 1 and f 2 are the first and second design variables may be changed during optimization.
frequencies, and the subscripts represent the simulation and test. The sequential quadratic program algorithm [25] was used to
Table 4 shows the design variables. There are nine design varia- find the optimum values of the design variables that minimize the
bles in the three DOF model shown in Fig. 12: (m1, m2, m3), (k1,
k2, k3) and (c1, c2, c3). Since damping has little effect on the fre-
quency, the damping elements c1, c2, and c3 are excluded from the
design variables for optimization. The values of the damping ele-
ments are set as c1 30 Ns/m, c2 0 Ns/m, and c3 64.9 Ns/m;
those values are defined in the design sheet offered by the panto-
graph manufacturer. In addition, k3, which is the stiffness of the
air compressor, is set to 0 N/m since the air compressor produces
a static life force and is fixed when the pantograph is driving.
Here, k1 represents a nonlinear spring that is installed inside the
pan head; as a result, it is not easy to find a representative value of
k1. In this study, k1 is set to 12,340 N/m, which is the largest of
the values offered by the pantograph manufacturer, since the high
frequency vibration is more important. Four design variables, as
shown in Table 4, were selected for optimization.
Twenty-five simulations were carried out, according to the cen-
tral composite design table [24]. The design table and simulation
results are shown in the Appendix. Using these results, the
response function that shows the relationship between the objec-
tive function and design variables is derived as
objective function. Table 5 shows the optimization results, and the Each mass was connected by using a translational joint. The fre-
objective function is the minimum at m1 ( 1), m2 (1), k2 (0), and quency (x) and amplitude (A) of the excitation were varied and
m3( 1). After optimization, the sum of the differences regarding the contact force variation was measured.
the two peak frequencies of the following performance simulation The following performance simulation results are shown in
and test decreases by 95.93% (1.314 Hz ! 0.0535 Hz). Table 6 Fig. 15. Figure 15(a) shows the contact force variation under the
compares the 1st and 2nd resonance frequencies of the test, origi- excitation of 0.5 Hz40 mm. The value of 0.5 Hz means the pan-
nal, and optimum models. The first frequency has been changed tograph stays in a nearly static environment, and the simulation
more than the second frequency after optimization, and this can result follows the test data well in the aspect of magnitude and fre-
definitely be seen in Fig. 13. quency. However, the high frequency components are seen in the
test result, while these vibration modes are not seen in the simula-
4.4 Validation of the Optimization Results. After optimiza- tion result. This is because the simplified three-DOF pantograph
tion, the receptance of the simulation model successfully follows model was used. In Figs. 15(b)15(d), the simulation result cannot
the test data, as shown in Fig. 13. To validate the dynamic charac- represent a high frequency effect. As pointed out in Sec. 4.1, a
teristics of the optimized three DOF pantograph model, the fol- high frequency of the pantograph is not important since the theo-
lowing performance simulation model was created, as shown in retical frequency of the vertical force that is applied to the panto-
Fig. 14. The material properties of the optimum model shown in graph is lower than 20 Hz. Therefore, the high frequency effect on
Table 5 were used, and sinusoidal excitation was applied to the
model. The stiffness of the contact spring was set to 50 kN/m, as
Table 8 Statistical analysis results regarding the simulation
noted in the EN 50,318 standard [26]. To maintain the average and test data at V 5 305 km/h
contact force, a vertical static force of 175 N was applied to m3.
Category Simulation Test Error (%)
Table 7 Standard of EN 50119 to secure stable current collec-
tion performance ( V > 200 km/h) Average contact force (N) 144.5 145.8 0.9
Standard deviation (N) 38.5 39.5 2.6
Average contact Statistical max. of the c Statistical min. of the Statistical max. of the contact force (N) 259.9 264.3 1.7
force (Favg) ontact force contact force Statistical min. of the contact force (N) 59.1 27.3 53.8
(Fsmax Favg 3r) (Fsmin Favg 3r) Actual max. of the contact force (N) 248.1 262.3 5.7
Actual min. of the contact force (N) 87.0 31.5 63.8
<200 N <350 N Positive Max. uplift at the supports (mm) 63.1 / 62.5 / 59.9
the pantograph can be neglected. Figure 15(b) shows the follow- (iv) t 20 s: The pantograph starts to drive; the speed
ing performance results at the first resonance frequency of 6.5 Hz, reaches its maximum after 0.5 s
and Figs. 15(c) and 15(d) show the following performance results
around the second resonance frequency of 10 Hz. The simulation When the pantograph drives with speeds of 305 km/h, 370 km/
results are well matched with the test data at both resonance fre- h, and 430 km/h, the contact force variation is acquired.
quencies. Therefore, it can be concluded that the optimization to
find proper values of the material properties of the three DOF pan- 5.2 Simulation Results. Figure 17 shows the configuration
tograph model was successfully performed and verified the
of the simulation model, according to the steps in the analysis.
dynamic characteristics of the simulation model.
STEP 1 : The presag of the catenary due to gravity is described.
5 Dynamic Interaction Between the Catenary and The amount of presag has already been verified in
Sec. 3.2.
Pantograph STEP 2 : The stagger is applied to the contact wire.
In this section, the catenary model shown in Fig. 7 and the pan- STEP 3 : The pantograph contacts with the contact wire. Vibra-
tograph model shown in Fig. 14 are combined and the dynamic tion of the contact wire due to the contact force must
interaction is simulated. When the pantograph drives with a speed converge before STEP 4 commences.
of 305 km/h, the contact force variation of the simulation model is STEP 4 : The pantograph is moved. The contact status between
validated by comparison with the test data. Then, the contact force the pantograph and the contact wire is of principal
variation and the current collection performance are discussed interest.
when the driving speed is 370 km/h and 430 km/h, respectively
(370 km/h and 430 km/h are the operating and maximum speeds Figure 18 shows the simulation and test results when the
of HEMU-400X, which is the next generation high-speed train driving speed (V) is 305 km/h. The contact force result shown in
under development in South Korea). Fig. 18(a) is periodically repeated. The contact force reaches its
maximum around the border between spans where the registration
arm is located. Otherwise, the contact force of the test data seems
5.1 Catenary-Pantograph Model. Figure 16 shows the sim-
not to be periodically repeated. This is because of the time differ-
ulation model of the catenary-pantograph. The pantograph is the
ence between the location signal of the train and the measured
same as the three DOF model shown in Fig. 14. Only the vertical
contact force data. The location of the train is obtained by the
static force was changed from 175 N to 145 N. A thin contact
GPS sensor and the contact force variation is acquired by the load
plate whose mass was 0.2 kg was created to apply the node to
cell (2), as shown in Fig. 11, which is specially designed by the
face contact condition [27] between the pantograph and the con-
Korea Railway Research Institute. The location and force data are
tact wire. Ten spans of the catenary, shown in Fig. 7, were cre-
collected in different acquisition systems and then synchronized.
ated. A stagger of 6200 mm was applied to the registration arm,
In this process, the time difference may have occurred. Also, the
which is modeled by using the ground spring, whose stiffness is
aerodynamic force may affect the interaction of the catenary and
213 N/m, of the contact wire. Simulation was performed as in the
pantograph in the actual test. However, if the average and stand-
following steps.
ard deviation of the contact force results from the simulation and
(i) t 010 s: Only gravity of 9.81 m/s2 is applied to the test are the same, it can be concluded that the simulation results
catenary model; presag is described follow the test data with sufficient accuracy because the standard
(ii) t 1015 s: Stagger is applied to the registration arm of judgment to determine whether the loss of contact occurs or not
of the contact wire; also, the pantograph is lifted up is the sign of the statistical minimum contact force [3]. The stand-
(iii) t 1520 s: No external forces are applied to the ard, EN 50,119 defines that the statistical minimum contact force
model to secure dynamic equilibrium must be positive to avoid the loss of contact. Table 7 shows the