See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/337155496

3D DYNAMIC MODELING OF RAILWAY TRANSITION ZONES IN SOFT SOIL

Conference Paper · July 2017

CITATIONS READS

0 391

1 author:

Edina Koch
Széchenyi István University, Gyor
26 PUBLICATIONS 36 CITATIONS

SEE PROFILE

All content following this page was uploaded by Edina Koch on 10 November 2019.

The user has requested enhancement of the downloaded file.

 Aktuálne geotechnické riešenia a ich verifikácia, Bratislava 05.- 06. júna 2017

3D DYNAMIC MODELING OF RAILWAY TRANSITION ZONES IN

SOFT SOIL
Edina Koch1

ABSTRACT:
The substructure stiffness of the railway lines play an important role regarding the internal
forces in the rails, sleepers and crest zone of the embankment. Excessive stresses in any of
these components can result in permanent deformations of the track. The problem is
especially pronounced in locations where there are abrupt changes in support stiffness due to
structural differences in the track. The typical problem zones are points where the open lines
run onto bridges or into tunnels, as well as transition zones connecting the open track and the
track above culvert and drainage tunnel crossings. Examples prove that inadequate technical
solutions can generate damage that may require long term speed restrictions or lead to short
maintenance cycles and significantly increasing the total cost of ownership. The author
reports the first results obtained from the investigation of a 3D numerical model of a transition
zone subjected to dynamic train loads. The structure is embedded in soil environments with
different strength and stiffness properties. The mechanical behavior of the “B” size category
structure and its soil environment is presented in the study with special regards to factors like
train speed, embankment height and the settlement differences developing in the transition
zone.

1 Introduction
Railway transition zones are always present at the boundaries between regions of different
stiffness. Figure 1 shows this problem and relates it to another one; namely the differential
settlement between these two zones. According to Insa (2008), the key factor for reducing the
maintenance costs in railways consists of defining an optimum vertical stiffness below the
track and maintaining it. This is clearly a problem in transitions between embankment and
bridges or tunnels. One of the objectives of the transition is to provide a gradual stiffness
variation as shown in Fig. 2.
ERRI (1999) indicates that the factors influencing the behavior of the track in transition
zones can be: (1) external to the track (axle loads, weather conditions, speed and vibrations),
(2) geotechnical issues (sub-grade and soil conditions), (3) structural conditions (bending
stiffness, lateral movements and interaction between track and bridge) or (4) related to the
track design and layout (stiffness, location of track dilation devices or presence of CWR).
A number of different solutions for transition zones have been proposed and applied. These
transitions are built to smooth the stiffness variation between the “soft” approach section and
the “stiff” section over the structure. Transitions based on smoothing the stiffness variation on
the “soft” side (Kerr and Moroney, 1993; Li and Davis, 2005; Read and Li, 2006) include the
use of oversized sleepers, varied spacing sleepers, underlayments of hot-mix-asphalt,
geotextiles, or soil-cement, additional rails, approach slabs, and others.
1,
Edina Koch, PhD, Széchenyi István University, Egyetem tér 1, 9026 Győr, Hungary,
phone No.: +36-96-613633, E-mail: [email protected]
 Transitions based on lowering the stiffness on the “stiff” section (Kerr and Moroney, 1993;
Sasaoka and Davis, 2005; Read and Li, 2006; Li et al., 2010) include the use of soft rail pads
under sleeper pads, plastic sleepers or ballast mats. According to Li and Davis (2005),
transition zones must address the specific stiffness issues of the corresponding track
discontinuities in order to be effective.
The problems associated with the transition zones require a complex analysis. For efficient
modelling of the mechanisms resulting in gradual line deteriorations in the transition zones,
understanding the 3D and dynamic effects associated with the problem are essential. To better
understand the problem, a 3D numerical model has been developed and presented for time
domain analysis.

Fig. 1. Stiffness problems Fig. 2. Transition Solutions

2 Modeling approach

The effect of moving train loads over a culvert has been analysed using the Plaxis 3D
Dynamic package. The culvert itself consists of a square concrete box (2 m by 2 m). Figure 3
shows a section of the culvert and the soil profile. The top 5m of the subsoil in the current
model was soft silt, resting on 15 m of stiff sand. On top of the soft layer, a sand embankment
was built to support the railway line. Height of the embankment was H=0-2-4m, with a slope
of 1:1.5. The ballast layer was designed with 0.35 m thickness.

H = 0,0 m 0,35 m
H sand embankment
2,0 m
4,0 m culvert 2m
5,0 m
soft silt
2m

15 m

Figure 3. Longitudinal view of the track passing over the culvert (not to scale)
 The length of the model was 96 m and the width was 45 meters. Standard fixities and
absorbent boundaries were applied to model free-field conditions and reduce wave reflections
at the boundaries. The rail was modeled by a beam element along 96 m of profile in the Y-
direction. The properties of the beam section were selected in a way that it had the same
flexural- and tensile stiffness as a UIC60 rail. The standard sleeper, B70, was modeled as a
beam element by duplicating the moment of inertia and area. Input properties for rail and
sleepers are shown in Table 1. A total of 121 sleepers were placed in the model with a center-
to-center distance of 60 cm. Figure 4 shows the PLAXIS 3D model. The moving train was
modeled with the LM71 Eurocode load model consisting of eight dynamic point loads of 125
kN vertical force.
PLAXIS 3D defines dynamic loads using a time-force signal. In the model, every single
dynamic point load has its own multiplier. These load multipliers turn on and off when the
rolling vehicle passé over it. The dynamic time step changed to simulate different travel
velocities (80 km/h and 250 km/h), while the distance between dynamic point loads were held
constant. For example, a train with 80 km/h speed passes 80 cm in 0.036 sec, hence, the time
interval must be chosen 0.036 sec for the fixed dynamic point loads. The total elapsed time
between the first load and the last load 4.32 sec. An additional time of 2.68 sec was
considered to allow complete dissipation of the waves induced by the passing train.

Rail

Sleepers
Train Load
Ballast
Soft Silt
Embankment
Backfill
Culvert
Stiff Sand

Figure 4. Plaxis 3D model

Table 1. Input properties in PLAXIS 3D for rail and sleeper

Sleeper Rail
Parameter
B70 UIC60
Cross section area A (m2) 0,0513 0,0077
Unit weight  (kN/m3) 25 78
Young’smodulus E (MPa) 36000 200000
4
Moment of inertia third axis I3 (m ) 0,0253 0,00003
4
Moment of inertia second axis I2 (m ) 0,00024 0,00000513

The following construction phases were defined:

1–initial phase, 2–excavation, 3–placing the backfill material underneath the culvert, 4–
construction of the culvert, 5–placing the backfill material at both sides of the culvert, 6–
construction of the embankment, 7–placing the ballast, 8–placing the sleepers, 9–placing the
rails, 10–moving train (80 km/h and 250 km/h).
A plastic drained calculation condition was chosen in phase 1-9. In phase 10, the dynamic
calculation option was selected to model the stress and strain waves and vibrations in the soil.
In this phase, all dynamic point loads on the rails were active.

3 Material properties

It has been discovered from dynamic response analysis (Seed & Idriss, 1970), that most soils
exhibit nonlinear stress-strain relationships. The shear modulus G (see Figure 5) is usually
expressed as the secant modulus found at the extreme points of the hysteresis loop. The
damping factor is proportional to the area found inside the hysteresis loop. The applied
terminology of damping usually means the dissipation of strain energy during cyclic loading.
From the definition of both physical properties, it is clear that each of them depends on the
magnitude of the strain for which the hysteresis loop is determined. Thereby, both the shear
moduli and damping factors must be determined as functions of the strain level experienced
by the soil. Several studies have shown that the shear moduli of most soils decay
monotonically with strain. Cavallaro et al. (1999), Mayne & Schneider (2001), and Benz et al.
(2009) suggest that the maxima are at very small strain levels, i.e. less than 10-6 to 10-5, which
is associated to recoverable strains, the material behavior is almost purely elastic (see Figure
6).

a) b)

Fig. 5. a) Definition of the secant shear stiffness Gsec of the hysteresis loop, b) decrease of
Gsec from its maximum value Gmax with increasing shear strain amplitude ampl

The small strain stiffness implementation in PLAXIS is based on the small strain overlay
model (Benz et al., 2009). Parameters G0 and 0.7 are required input parameters. In the
absence of experimental data for the determination of these two required parameters,
approximations through correlations can be appropriate.
To model the soil behavior, the HS-small constitutive model was applied. The ballast layer
is modeled by MC, for the concrete culvert Linear Elastic model was applied. Soil basic and
advanced properties in models are listed in Table 2. The applied Poisson’s ratio for all layers
in the HS-small model is the default value given by PLAXIS (ur = 0.2).
 Fig. 6. Characteristic stiffness-strain behavior in logarithmic scale

Table 2. Material properties of soil layers

backfill /
subsoil ballast culvert
parameter embankment
soft clay dense sand gravel concrete
model HS-small HS-small MC LE
E (kPa) 100 000 3 E+7
ref
E50 (kPa) 4 000 36 000
Eoedref (kPa) 4 000 36 000
Eur ref (kPa) 12 000 108 000
G0 ref (kPa) 6 000 100 800
m (-) 0,70 0,51
0,7 (-) 0,00279 0,00014
c'ref (kPa) 10 1,0 10,0
'ref (deg) 25 35,5 40,0
ψ (deg) 0 5,5

4 Results

The aim of modeling was to determine the settlement in the culvert transition zone while
varying embankment height and train speed. In order to evaluate the settlement due to a
moving train, several cross-sections were selected (3 on open track, 5 on backfill and 1 on the
culvert) and total displacements were determined on top of the ballast when the moving train
was exactly above the cross-section.
Figure 7 shows the deformed mesh of a model. One can see the effect of the moving train
as it pushes the embankment into the soft subsoil.
 Fig. 7. Deformed mesh of a model

The influence of train speed and embankment height on on track settlement is summarized
in Figure 8. It can be established that a higher speed causes greater settlement and the higher
the embankment, the lower the settlement due to the moving train. The difference is not
much; especially in case of low embankment height due to the softness of the subsoil.

height of embankment (m)

0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0
0
displacement due to first moving train (mm)

-20

-40

-60

-80
soft silt - 80 km/h
soft silt - 250 km/h
-100
Fig. 8. Relationship between embankment height and settlement

Figure 9 shows the vertical velocity for five checkpoints in different layers on open track
(A: +3,35 m-top of the ballast, B: +2 m in the embankment, C: 0,0 m-top of the subsoil, D: -
1,5 m in the subsoil, E: -5 m in the subsoil). Results relate to the case of 3.0 m high
embankment and 80 km/h train speed. Velocity amplitudes are smaller as you go deeper,
which is matched to the engineering expectation. The highest velocity belongs to checkpoint
A that is located on top of the ballast. The checkpoints B, C, D and E show smaller velocities
as the wave goes deeper in Z-direction.
Figure 10 presents the result of total displacements due to repetitive moving load. The same
load model run seven time on the 96 m long model with the speed of 80 km/h. One can state
that the first moving load causes high immediate settlement, the residual one is much lower.
The incremental settlement between two load steps slightly decreases as the running step
increases.
dynamic time s (sec)
0 1 2 3 4 5 6 7
0,06

0,04
vertical velocity vs (m/s)

0,02

0,00

-0,02

-0,04

-0,06

-0,08 A B C D E

-0,10

-0,12

-0,14

-0,16

Fig. 9. Vertical velocity vs dynamic time at 80 km/h

Figure 10 shows the total displacement in a longitudinal section of the tranzition zone when
the train approaches to the backfill zone. Red colour means higher settlement. The effect of
moving train is clear. The displacement decreases with depth. The residual settlement on the
open track is less than the immediate.

Fig. 10. Longitudinal section of transition

Figure 11 shows the total displacements (construction and moving train) for seven
checkpoints in different layers (open track: A: top of the ballast +2,35 m, B: top of the subsoil,
C: -3,0 m in the subsoil; backfill zone: D: top of the ballast +2,35 m, E: top of the backfill, F:
-3,0 m in the backfill zone, G: top of the ballast above the culvert).
Based on figure 12, the following observations can be stated:
− the greatest displacement on open track occurs on top of the ballast (A),
− in the backfill zone, the displacement reduces significantly (D),
− the “bump” is conspicuous when the load is directly above the culvert (G),
− reduction of the settlement with the depth is obvious (B-C and E-F),
− at -3 m in the subsoil, the effect of passing train is definitely decreased (C).

dynamic time s (sec)

0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0
0

20
total displacement u (mm)

40

60

80

100

120

140
A B C D E F G
160

Fig. 11. Total displacement in different checkpoints

Figure 12 shows the peak displacements for the first passing of the train along the
longitudinal profile for four different cases, varying the embankment height and train speed.
The following could be stated:
− the highest settlement on open track occurs in the case of no embankment at 250 km/h train
speed,
− the lowest settlement on open track occurs in the case of 2 m embankment at 80 km/h train
speed,
− differential settlement between the open track and culvert is the highest in the case of no
embankment and 250 km/h train speed,
− case of no embankment and 250 km/h needs the longest transition,
− shapes of settlement curves change in the backfill zone due to the stiffest backfill material.

5 Ground improvement for reducing differencial settlement

The second part of our research was to investigate how reduce the differential settlement
between the two different stiffness structure (embankment and culvert) and how provide a
gradual stiffness variation. For this purpose the 2,0 m height embankment and 80 km/h train
speed model was analyzed. The basic results of the model were
− settlement on open track was 95 mm after placing the rails,
− moving train caused almost 50 mm extra settlement,
− the differential settlement between the two different stiffness structure was about 40 mm,
− necessary length of the trainsition was 2,5 times longer than the length of the backfill
zone.

distance (m)
culvert
0 12 24 36 48 60 72 84 96
vertical displacement due to first moving train load (mm)

0
backfill backfill

-20

-40

-60

-80

-100 soft silt - 0 m embankment - 80 km/h soft silt - 2 m embankment - 80 km/h

soft silt - 0 m embankment - 250 km/h soft silt - 2 m embankment - 250 km/h
-120

Fig. 12. Settlements of different models

For achieving the design criteria namely such as settlement is less than 20 mm on open
track, transition is smooth between the open track and the culvert and differential settlement is
less than 10 mm ground improvement is recommended before starting construction. One of
the solutions is deep-mixing stabilization of the subsoil.
The goal of deep-mixing is to improve the soil characteristics, e.g. increase the shear
strength and/or reduce the compressibility, by mixing the soil with some type of chemical
additives that react with the soil. The improvement occurs due to ion exchange at the clay
surface, bonding of soil particles and/or filling of voids by chemical reaction products.
One type of treatment namely partial mass stabilization was evaluated; 1.8-m diameter
equivalent columns were placed in 3.0×3.0-m square grids. Young’s modulus for the columns
was Eref=20 000 kPa, the unconfined compressive strength was qu=300 kPa and the value of
Poisson’s ratio was = 0.2.
Figure 13 shows the PLAXIS 3D model. Figure 14 shows the peak displacements for the
first passing of the train along the longitudinal profile for both cases, without treatment and
with treatment of subsoil. The following could be stated:
− the settlement on open track is reduced drastically,
− settlement reduction is significant on the backfill zone,
− the smooth transition is visible,
− differential settlement between the open track and culvert was less than 10 mm.
 Fig. 13. PLAXIS 3D model with the partial mass stabilisation

2,0×2,0 m culvert distance (m)

0 12 24 36 48 60 72 84 96
0,0
vertical displacement due to first miving train (mm)

backfill backfill

-10,0

-20,0

-30,0

-40,0

-50,0
soft silt - 2 m embankment - 80 km/h

soft silt - 2 m embankment - 80 km/h - column type deep mixing ground improvement
-60,0

Fig. 14. Reduction of differential settlement

6 Conclusions

The presented results show that the real life problem of a train travelling in a transition zone
can be modelled in Plaxis 3D. The current paper focuses mainly on the modelling itself, and
the results presented here are only a subset of a larger parametric study. It is clearly illustrated
that the influences of the different parameters (such as the speed of the vehicle and the height
of the embankment) can be modelled efficiently. The presented experience in dynamic 3D
modelling of the problem allowed us to make design recommendations for the required length
of the transition zone in different soil conditions and different types of structures. The effect
of the train velocity on the settlements has been demonstrated. This increment however is not
associated with the increased dynamic factor but probably the higher rate of accumulated
strains due to a more rapid impact and a same rate of energy dissipation. The extra peak
settlement in the approach zone at high speed passing is probably also due to the
accumulation of dynamic strains partially caused by the reflections from the culvert.
 Based on the cyclic response analysis that has also been presented here one may conclude
that the model is capable to produce predictions on the long term behavior of the track. The
experienced incremental settlement is due to plastic deformations of the soil mass which
supposed to cease after a number of cycles. To test this hypothesis a higher pass numbers will
be tested shortly. For the analysis of shakedown like behavior different soil model with cyclic
degradation capability should be implemented.

6 REFERENCES
Insa, R.,”Diseño de vías de alta velocidad: construción y mantenimiento“, Ciclo de Formação
Avançada na ferrovia. Porto: CSF, (2008).
European Rail Research Institute. Utrech. ERRI D 230.1/RP 3. Bridge ends. Embankment
Structure Transition. State of the Art Report, (Nov. 1999).
Kerr, A. D., & Moroney, B. E. (1993). Track transition problems and remedies. Paper
presented at the the American Railway Engineering Association, Washington, USA.
Li, D., & Davis, D. (2005). Transition of Railroad Bridge Approaches. Journal of
Geotechnical and Geoenvironmental Engineering, 131(11), 1392–1398.
Read, D., & Li, D. (2006). Research results digest 79 Transit cooperative research program
D-7/Task, 15 (pp. 38). Pueblo, Colorado: Transportation technology center, Inc. (TTCI).
Sasaoka, C. D., & Davis, D. (2005). Implementing track transition solutions for heavy axle
load service. Paper presented at the the AREMA 2005 Annual Conference, AREMA.
Li, D., Otter, D., & Carr, G. (2010). Railway bridge approaches under heavy axle load traffic:
problems, causes, and remedies. Paper presented at the the Institution of Mechanical
Engineers, Part F: Journal of Rail and Rapid Transit.
Benz, T. (2006). Small Strain Stiffness of Soils and its Numerical Consequences. Ph.D.
Dissertation. Institut für Geotechnik der Universität Stuttgart. 209 p.
Atkinson J.H., Sallfors G. (1991). Experimental determination of stress strain time
characteristics in laboratory and in situ tests, Proceedings of 10th European conference on
soil Mechanics and Foundations Engineering, 3, p. 915-956.
Seed H.B., Idriss I.M. (1970). Soil moduli and damping factors for dynamic response
analyses, Technical Report EERRC-70-10, University of California, Berkeley
Wichtmann, T., Triantafyllidis, Th. (2005): Dynamische Steifigkeit und D¨ampfung von Sand
bei kleinen Dehnungen (Dynamic stiffness and damping of sand at small strains), in
German. Bautechnik, Vol. 82, No. 4, pp. 236-246.
Cavallaro, A., Maugeri, M., Lo Presti, D.C.F. and Pallata, O., (1999). "Characterising Shear
Modulus and Damping from in Situ and Laboratory Tests for the Seismic Area of Catania".
Proceeding of the 2nd International Symposium on Pre-failure Deformation Characteristics
of Geomaterials, Torino, 28 - 30 September 1999, pp. 51 - 58.
Mayne PW, Schneider JA (2001) Evaluating drilled shaft response by seismic conw.
Foundation and ground improvement, GSP No. 113, ASCE, Reston, VA, pp 655-669
Brinkgreve R.B.J., Vermeer P.A. (2010): PLAXIS-Finite element code for soil and rock
analyses, Plaxis 3D. Manuals, Delft University of Technology  Plaxis bv, The Netherlands.
Benz, T., Vermeer, P.A., Schwab, R. (2009). A small-strain overlay model. International
Journal for Numerical and Analytical Methods in Geomechanics, Vol. 33, pp 25–44.

View publication stats