Physical Modelling

ECPMG 2020
4th European Conference on

Physical Modelling
in Geotechnics
Editors: Jan Laue & Tarun Bansal
PROCEEDINGS OF THE 4th EUROPEAN CONFERENCE ON PHYSICAL
MODELLING IN GEOTECHNICS
LULEÅ, SWEDEN, 15-17 MARCH 2020
HYBRID LULEÅ/ZOOM, 7-8 SEPTEMBER 2020
4th European Conference on
Physical Modelling
in Geotechnics
Edited by:
Jan Laue & Tarun Bansal

Luleå University of Technology, Sweden
This work is licensed under the Creative Commons Attribution- NonCommercial-
NoDerivatives 4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-nd/4.0/
ISBN (print): 978-91-7790-542-4

ISBN (pdf): 978-91-7790-543-1
ECPMG 2020, Laue & Bansal (eds.), ISBN (print): 978-91-7790-542-4 & ISBN (pdf): 978-91-7790-543-1
Table of contents
Preface VII
1 Back to basics
Identification of scaling laws for the friction in centrifuge models of reinforced soil 3
Jean de Sauvage & Thierry Dubreucq
Silty sand characterization in the geotechnical centrifuge 9

Varvara Zania, Sine Maria Christensen & Irene Rocchi
The effects of sand grading on the bearing capacity of surface foundations 17

Mathias R. Jensen, Varvara Zania & Barry M. Lehane
Tests of varied sample preparation methods for centrifuge modelling 25

Eric Ritchie, Leonardo Lalicata, Sam Divall & Richard Goodey
Accelerations in dynamic centrifuge modelling of seismic liquefaction problems 31

Andreia Marques, Paulo Coelho, Stuart Haigh & Gopal Madabhushi
Boundary effects in centrifuge modelling of soil liquefaction 39

Zheng Li & Sandra Escoffier
Centrifuge model studies on dry granular soils subjected to dynamic compaction induced tamping 43
Saptarshi Kundu & B.V.S. Viswanadham
Model preparation for unsaturated soil testing in a centrifuge environment 49

Leonardo Maria Lalicata, Luc Thorel, Augusto Desideri, G.M. Rotisciani & Francesca Casini
Effect of Crushed Particles on Soil-Structure Interface Behaviour 57

Angus Pettey & Charles Heron
Carboxymethylcellulose as a Newtonian viscous fluid for centrifuge modelling 65

Jarno Terwindt, Wouter van der Star & Dirk de Lange
2 Physical modelling techniques and facilities
A new geotechnical centrifuge at Deltares, Delft, the Netherlands 75

Rob Zwaan, Jarno Terwindt, Dirk de Lange & Adam Bezuijen
A Novel Small-Scale Model Testing Device for Laterally Loaded Piles 83

Hendrik Sturm & Andrea Støren
Development of a Hammer to drive monopiles of offshore wind turbines in Centrifuge 89

Semaan Maatouk, Matthieu Blanc & Luc Thorel
Experimental testing and ring polygons for the rolling stock – track infrastructure interaction 95
and subgrade analysis
Alexander Savin, Andrey Zaytsev, Alexander Abrashitov, Andrey Sidrakov & Alexander Gorlov
III
Pneumatic load frame for long-term in-situ application of moderate tension loads on 103
foundation elements
Jorge Yannie, Claes Alén & Jelke Dijkstra
Chromatic aberration when using white LEDs for PIV: issues and remedies 109
Mir Amid Hashemi, Jingmin Xu, Charles M. Heron & Alec M. Marshall
Physical Model of Rainfall Induced Landslide in Flume Test: Preliminary Results 115
Željko Arbanas, Vedran Jagodnik, Josip Peranić, Sara Pajalić, Martina Vivoda Prodan &
Nina Čeh
3 Piles: installation and capacity
Soil plug investigation with respect to pile geometry and installation method 125
Sascha Henke
Pile group effects on installation and capacity of jacked piles in sand tested in the centrifuge 133
Dirk de Lange & Jarno Terwindt
Centrifuge modelling of shaft resistance of a rough rock-socketed pile 139

José Gregorio Gutiérrez-Ch, Geyang Song, Charles Heron, Alec Marshall & Rafael Jimenez
Physical modelling of single-helix piles in sand using calibration chamber tests 143
Diego Moreira da Silva & Cristina de Hollanda Cavalcanti Tsuha
Study on the installation effect of helical piles in very dense sand 151
José Antonio Schiavon, Cristina de Hollanda Cavalcanti Tsuha & Luc Thorel
Experimental technique for creating enhanced capacity piles in a centrifuge environment 157
Leonardo Lalicata, Andrew McNamara & Sarah Stallebrass
Model tests of grouted piles in gravel esker and fill soils 163
Wik Breure & Anders B Lundberg
4 Infrastructures
Geotechnical instrumentation of an experimental embankment dam 171

Jasmina Toromanovic, Johan Lagerlund, Peter Viklander & Jan Laue
Surface wave propagation in centrifuge testing using Electromagnetic Drop-Projectile system 177
Vipul Kumar & S.P.G. Madabhushi
Centrifuge modelling of train passage over clay subgrades 185

Scott Robinson, Andrew Brennan, Cameron Munro, Peter Woodward, David Connolly, Omar
Lagrouche & Ahmet Esen
Centrifuge modelling of tunnel-soil-pile interaction considering the presence of rigid caps 193
Jingmin Xu, Andrea Franza & Alec M. Marshall
Centrifuge Modelling of the Reverse Faulting-Tunnel Interaction in Granular Alluvial Soil 201
considering Tunnel Flexible Joints
Hamid Tohidifar, Mojtaba Moosavi & Mohammad Kazem Jafari
Soil Improvement of a Subway Station Excavated by an Earth Pressure Balance Machine 205
Mohammad Zamani Ahmad Mahmoudi, Ali Entezari Zarandi, Milad Naghibi & Ali Azad
IV
Performance of Geogrid reinforced soil walls with low permeable backfill: Centrifuge Study 213
B.V.S. Viswanadham, Midhula Jayanandan, Dipankana Bhattacherjee & Prince Kumar
Centrifuge model studies on geogrid and hybrid geosynthetic reinforced slopes subjected to 221
rainfall
Dipankana Bhattacherjee & B.V.S Viswanadham
Effect of fiber soil reinforcement on the bearing capacity of a surface strip foundation 227
Maria Diakoumi, Kevin Stone & David Roddy
5 Offshore geotechnics
Centrifuge tests of monopile foundations under one-way cyclic lateral loading 235
Kunkun Cheng, Charles M. Heron, Luke J. Prendergast & Alec M. Marshall
Use of optical fibres to measure pore water pressure development during impact pile diving: a 243
geotechnical centrifuge study
Amin Askarinejad, Tristan O. Quinten, Mario Alvarez Grima, Cornelis van ‘t Hof & Ken Gavin
The modelling of piles under multidirectional loading 247

Christophe Gaudin, Conleth O’Loughlin & Manuel Herduin
Suction caisson response under tensile and compressional vertical cyclic loading in sand over clay 255
Marc Stapelfeldt & Jürgen Grabe
Centrifuge Modelling of Suction Caissons Subjected to Cyclic Loading in Tension 263

Yuen Zhang, Kamaleshwar Sudhakaran & Amin Askarinejad
An investigation into the performance of helical piles with fins under lateral loading 269
Kevin Stone, Maria Diakoumi & Oliver Jeffcock
Pile test database on centrifuged models: tension loading for floating wind turbine anchor 275
Luc Thorel, Ismat El Haffar & Matthieu Blanc
6 Extended Abstracts
Experimental study of the interaction of shallow foundations with an emerging strike-slip fault 281
rupture
Athanasios Agalianos, Evangelia Korre, Tarek H. Abdoun & Ioannis Anastasopoulos
Investigation of the dynamic behavior of gravity-type quay wall under saturated condition 283
Yeon-Sam Kim, Kil-Wan Ko & Dong-Soo Kim
Design of a Miniature Cone Penetrometer for the Geotechnical Centrifuge of Delft University of 285
Technology
Dirk de Lange, Siavash Honardar & Amin Askarinejad
Behaviour of prefabricated concrete screw piles under axial loading 287

André Arnold & Amin Askarinejad
Validation of ESB Box and Estimation of Shear Wave Velocity in Partially Saturated Soil 289
Yoon-Ah Kim, Hae-In Lee, Satish Manandhar & Dong-Soo Kim
Retrofit, redesign, and restructuring of a large geotechnical centrifuge research center 291
April J. Bowman, Anthony F. Tessari & Christopher Price
V
Using Fibre Bragg Grating sensors to estimate the horizontal monopile behaviour in centrifuge 293
Matthieu Blanc, Zhong-Sen Li & Luc Thorel
High density soil instrumentation for a thermal pile test 295

Anders Bergström, Saqib Javed & Jelke Dijkstra
Physical Modelling of Geotechnical structure subjected to Hydraulic Scour 297

Liam Jones & Ioannis Anastasopoulos
Distribution of temperature in culverts in cold climate 299

Karina Tommik, Jan Laue, Sven Knutsson & Nina Lintzén
Geosynthetic reinforcement in lightly piled embankments: Laboratory model development 301

Per Gunnvard, Jan Laue & Qi Jia
Co-seismic response of a mat-founded structure on deposits with silty-sand layers 303

April J. Bowman, Orestis Adamidis, Shideh Dashti & Ioannis Anastasopoulos
Author Index 305
VI
Preface
Exchange of Knowledge and experience by means of written contributions, presentations and lively
discussions among researchers involved in Physical Modelling in Geotechnics is an essential part
in further developing our profession. The series of European Conferences on Physical
Modelling sponsored by the Technical committee (TC) 104 of the International Society of Soil
Mechanics and Geotechnical Engineering (ISSMGE) fills the gap between the International
Conferences on Physical Modelling in Geotechnics taking place every four years. This series has
started as Conference Series on Centrifuge Modelling (Centrifuge 1988) and the scope has later been
widened to geotechnical physical modelling in general. This conference series reaches its 10th event in
2022. Regional conferences were established in Europe (2008) and Asia (2012) alternating in the
middle of the four year period of the international conferences.
The 4th European conference on physical modelling in Geotechnics taking place in Luleå in the north
of Sweden follows its preceding ones held in London (2008), Delft (2012) and Nantes (2016). It is the
first of all international and regional conferences taking place in Scandinavia and also the first taking
place at a facility which does not have a geotechnical centrifuge yet. Contributions have been
received from Europe but also from both Americas, Asia and Australia.
The proceedings consists of 40-reviewed full paper and 12 extended abstracts. Extended abstracts
have been chosen to be a second option of contributions to allow discussion of new topics and results.
Selected contributions will be invited to be published in a special edition of the International Journal
on Physical Modelling. The extended abstracts are presented in a 3-minute pitch format to enhance
discussion. After an inspirational talk given by Prof Malcom Bolton and a motivational talk
presented by Prof Michael Davies, the other contributions are presented in 7 sessions under the
heading of “Back to the basics”, “Physical modelling techniques and facilities”, “Piles installation
and capacities”, “Infrastructure” and “Offshore Geotechnics”.
The participants were also supposed to enjoy the Vårvinter (Springwinter) in Luleå
including self-load tests on the frozen harbour. This had to be cancelled due to the COVID 19
pandemic and the conference has to be transferred to be hybrid. We hope that everyone will take the
chance to visit Luleå in winter and we promise the delegates are very welcome.
Dr.-Ing. Jan Laue

Chair of the 4th European Conference on Physical Modelling in Geotechnics
Professor in Geotechnical Engineering
Head of the Division of Mining and Geotechnical Engineering
Luleå University of Technology, Luleå, Sweden
VII
1. Back to basics
Identification of scaling laws for the friction in centrifuge models of reinforced soil
Jean de Sauvage1, Thierry Dubreucq1
Geotechnical Engineering, Environment, Natural Hazards and Earth Sciences Department, IFSTTAR, Université
1
Gustave Eiffel, France

Corresponding author: Jean de Sauvage (jean.de-sauvage@ifsttar.fr)
ABSTRACT: Centrifuge modelling requires the respect of scaling laws. For many phenomena, these scaling
laws have been studied and described but some of them are still unresolved. For example, the scaling laws
concerning the friction between the soil and an inclusion such as a nail have been only partly characterised.
As long as the size of the grains is small enough in comparison with the size of the inclusion, the peak shear
strength is the same as in the prototype but the scaling of the displacement needed to mobilize this strength
is unknown yet. This raises issues for the modelling of reinforced soil structures.
The influence of the grains size and the roughness of the inclusion on the friction law at the soil-inclusion
interface have been characterised experimentally. Mini-pull out tests have been carried out in an adapted
simple shear box. These tests have been carried out in Hostun sand and with inclusions presenting various
diameters and roughness. Discrete Element numerical simulations realised with the open source software
Yade-DEM completed this experimental study. It allowed to study the local mobilization of the shear
strength along the inclusion and to understand the micromechanical processes involved. The comparison of
experimental and numerical results allows to explain the principles governing the mobilisation of shear
strength in the centrifuge models of reinforced soil structures and to set the firsts steps towards scaling laws.
Keywords: Friction; Scaling Law; Centrifuge Modelling; Reinforced Soil.
1 INTRODUCTION tests were carried out as well as numerical

Centrifuge modelling is a cost effective way to simulations using Discrete Element Method.
study the behaviour of geotechnical structures. In
particular, it is often used for the modelling of 2 MINI PULL OUT TESTS
reinforced soil (e.g. Dubreucq 1999, de Sauvage A simple shear test apparatus has been adapted by
2018). However, it requires the respect of scaling drilling a hole in the side of the box and fixing the
laws and the ones concerning the friction between load cell to a mandrel instead of the upper half box.
the soil and an inclusion such as a nail have been This way, the pull out force and the nail’s head
only partly characterised. As long as the size of the displacement can be monitored under different
grains is small enough in comparison with the size vertical loads. The inside dimensions of the box are
of the inclusion, the peak shear strength is the same 139 mm x 145 mm and its height is 48.5 mm. The
as in the prototype but the scaling of the relative friction length of the rods is 144 mm and its
displacement needed between soil and inclusion to horizontal speed is equal to 1 mm/min.
mobilize this strength is unknown yet (Garnier et al Three different surface roughness types were
2007). The question whether the friction depends on considered: smooth rods, threaded rods and coated
the relative displacement or on the distortion is not rods. The last ones intended to model the grouting
sorted out. This raises issues for the modelling of of prototype nails and arc-welding rods were used.
reinforced soil structure. For example, the tip They are composed of a steel core and a granular
displacement of a soil-nailed wall is a consequence coating of 0.8 mm thickness. The roughness index
of the soil mass movement needed to mobilize shear Rn is defined by Rmax/d50 were Rmax is the maximal
strength along the inclusions. distance between peaks and valleys. Then, the
Milligan and Tei (1998) have carried out a study measured roughness index were:
of these scaling laws for nails in sand. They - clean smooth rods: Rn = 0.11
mentioned that the peak displacement did not - coated rods: Rn = 0.43
depend on the nail diameter or length and did not - threaded rods: 2.40 < Rn < 4.23
depend on the vertical stress. However they noticed It is difficult to study the influence of the grains
that the peak displacement for smooth nails was less size without changing the behaviour of the soil. In
than the half of the one measured for rough ones. order to investigate the size effect, it was decided to
In order to confirm these observations and to use different nail sizes. The diameters of the
observe the influence of roughness, mini-pull out threaded rods were 1.89 mm, 2.86 mm and
3
3.83 mm. The diameters of the smooth ones were
1.98 mm, 3.01 mm and 3.97 mm and the coated
ones were only found in two diameters: 3.42 mm
and 4.05 mm. The coating was intact after the pull
out tests. To simplify notations the rods will be
named “2 mm”, “3 mm”, “3.5 mm” and “4 mm”
later on.
The first experiments showed that even a very
small roughness had an influence on the pull out
force. The peak shear strength was higher with a
lightly oxidized rod than with the same rod after
cleaning through phosphorous acid even though the
roughness of the oxidised one is not sensible by
hand. Figure 2. Lateral friction against the normalized roughness
The surrounding soil is Hostun sand (HN38 type, (on a logarithmic scale).
d50 = 109 µm). A grease clog prevents the sand to
pour out of the box without disturbing the rod Figure 2 shows the lateral friction against the
displacement. Density is dynamically adjusted normalized roughness for all the tests performed. It
using a rubber hammer and the target volumic can be noticed that the most significant parameter is
weight for all tests is 1480 kg/m3, corresponding to the roughness since qs increases linearly with the
an 80% density index. The same person operated all logarithm of the normalized roughness. The slope of
the tests and the total volume of sand used allows the “qs against log (Rn)” line is proportional to the
neglecting the creation of fine particles by breakage vertical stress. For 2 mm rods this slope is equal to
during the successive tests. For every configuration, 2.3σv-32 with a regression coefficient R2=0.99.
three different vertical stresses have been imposed: The diameter of rods has a relatively small
51 kPa, 101 kPa and 153 kPa. influence on qs for the rough ones but for the smooth
ones, qs is almost doubled when the diameter drops
3 RESULTS from 4 mm to 2 mm under the maximum stress
153 kPa. Furthermore, for every vertical stress, the
For each tests, the lateral friction, computed as the shear strength is divided by 10 for smooth rods in
pull out force divided by the friction surface, was comparison with the rough ones.
plotted against the displacement of the rod (reduced As mentioned before, a specific attention was
by the rod diameter). The figure 1 gives an example dedicated to peak displacement up, defined as the
of such a plot. rod displacement needed to mobilize the maximum
shear strength at the interface between soil and
inclusion. The figure 3 plots, for three different
vertical stresses, the peak displacement reduced by
the rod diameter and corrected by the elastic
stretching. This correction is necessary to
investigate the scaling laws ruling the sole interface
friction.
Figure 1. Lateral friction against the reduced displacement

for the “4 mm” threaded rod under different vertical stresses.
The peak shear strength, qs, is defined as the

maximum of these curves. Using penetrometer tests
realised by Schiavon (2016) on HN38 sand, the
point resistance qc was estimated and, for the rough
rods, the relation between qs and qc was found to be
similar to the one mentioned in the French
Figure 3. Corrected reduced peak displacement against the
recommendations on tiebacks (CFMS 1995).
vertical stress for different rods.
4
For rough or coated rods, up increases slightly with a Radius-Expansion Friction-Decrease
with the vertical stress imposed. The rate of this approach (Chareyre and Villard 2002).
increase does not depend on the rod diameter.
However, for the smooth rods, the peak
displacement does not depend on the vertical stress.
Under a constant vertical stress, the reduced peak
displacement increases when the diameter
decreases. For example, under 101 kPa and with a
threaded rod, it goes from 0.19 (4 mm) to 0.28
(2 mm). It also shows that the absolute peak
displacement is not really constant as in the
experiments carried out by Milligan. For the smooth
rods, the influence of diameter is much more
significant.
To consider the scaling laws, one must keep in
mind that, with prototype nails (diameter ≈ 10 cm)
in sand, the order of magnitude of the peak Figure 4. View of the initial step of the DEM simulation
displacement in pull-out tests is 1 cm (Schlosser et
al 1993). The choice made to realise the initial state does
Eventually, the vertical displacement of the not model the sand pouring usually used for the
actuator imposing the vertical load has been also preparation of samples in centrifuge modelling but
monitored but the values were extremely small, i.e. is a good way to create a random pack with a desired
less than 0.05 mm. density index.
When the sample is ready, a downward velocity is
4 NUMERICAL SIMULATIONS prescribed to the wall type element until the desired
vertical stress is reached. Eventually a prescribed
To extend the numerical study and to understand the horizontal velocity is prescribed to the nail. At each
micromechanical processes underlying the friction step, the sum of forces and moments acting on each
behaviour of the soil-inclusion interface, it was grain is computed based respectively on relative
decided to use the Discrete Element Method, called displacement or overlapping with respect to the
DEM later on, and described by Cundall and Strack contact point and relative rotations and their
(1979). The open-source framework Yade acceleration is deduced. In a schematic way, on
(Šmilauer et al 2015) provided a powerful tool. This contact points, the contact normal forces are due to
method gives access to data that are difficult to the elasticity of the grains and the tangential ones
obtain in an experimental work such as the rotation are related to the normal ones by the friction law.
of particles. For more details, one can see the documentation on
Yade offers a great choice of grains shapes and the contact formulation Law2_ScGeom_
size distributions as well as a big variety of FrictPhys_CundallStrack.
mechanical behaviours. In order to limit the The nail geometry was designed on GMSH
computation time, the first model was realised with (Geuzaine and Remacle 2009) and imported as an
spheres and their behaviour was supposed elastic- STL file. It allowed assigning a geometrical
frictional. The Young modulus of the grains was roughness. There is no interaction between the nail
chosen to be 80 GPa. This value is an average of the and the box and, therefore, the only force exerted on
values used in the literature, ranging from 56 GPa the nail is due to the grains. Note that there is no
to 113 GPa. Their friction angle is 30°. friction between the grains and the box or the wall.
The setting of the initial state is a critical step in The figure 5 is an example of plotting the pull out
every DEM simulation since the arrangement and force against the displacement for a test carried out
interlocking of the granular assembly can deeply with a rough rod. The values cannot be compared to
affect the final measures. In this case, a box was the ones in the experimental tests, partly because of
formed of fix rigid facets to contain the grains, a the simple assumptions made on the shape and size
wall type element was placed on the top to impose distribution of grains. It can also be noticed that no
a vertical stress later on and the nail was realised decrease in the pull out force is observed after the
with Pfacet type elements. The size of the box was peak strength is reached. On the example presented,
chosen to be 1 cm x 1 cm x 1 cm. Then the pack of the decrease is only due to the finite length of the
grains was set up using the randomDensePacking nail: after 4 mm, the tail of the nail enters in the
function. The target void ratio was then obtained volume observed.
5
distribution cuts-off. However, they already show a
minor divergence with the results of Milligan and
Tei. More important, they clearly point the strong
influence of the roughness on the peak displacement
value. In order to find the scaling laws ruling the
friction at the interface between soil and inclusion,
a careful study of roughness has to be carried on.
The DEM simulations presented are quite simple
and cannot pretend to highlight all the phenomena
underlying the friction behaviour but they already
show their potential to explain the scaling of
interface friction.
Figure 5. Pull out force (N) against the nail displacement 6 ACKNOWLEDGEMENTS
(mm) for a rough rod. The authors want to acknowledge Christophe
Roppert and Minh-Tan Do from the Materials and
However, the tendencies observed in the structures Department in IFSTTAR, University
experimental tests are confirmed and the peak Gustave Eiffel. They did the roughness
displacement increases with the roughness. measurements.
7 REFERENCES
CFMS. 1995. Recommandations Tirants d’Ancrage TA
95 (Ground Anchor Recommandations). Eyrolles,
Paris, France.
Chareyre, B. and Villard, P. 2002. Discrete element

modeling of curved geosynthetic anchorages with
known macro-properties. In Proceedings of the First
International PFC Symposium, pp 197-203.
Cundall, P. A. and Strack, O. D. L. 1979. A discrete

numerical model for granular assemblies.
Geotechnique 29(1) : 47-65.
Dubreucq, T. 1999. Renforcement des fondations

superficielles par inclusions planes horizontales
extensibles quelques effets de taille dans le frottement
sable-inclusion. PhD thesis, École Nationale des Ponts
et Chaussées.
Garnier, J. et al 2007. Catalogue of scaling laws and

similitude questions in geotechnical centrifuge
Figure 6. Visualization of rotated particles (in red) for a modelling, International Journal of Physical Modelling
rough rod (top) and a smooth one (bottom) in Geotechnics, 7(3) : 01-23.
The figure 6 presents the two simulations realized Geuzaine, C. and Remacle, J.-F. 2009. Gmsh: a three-
with a smooth rod and a rough one. The particles are dimensional finite element mesh generator with built-
coloured according to their rotations: the red ones in pre- and post-processing facilities. International
are the most disturbed ones. This visualization Journal for Numerical Methods in Engineering
79(11) : 1309-1331
suggests that the rough rod mobilizes a bigger
volume of sand. This could explain the differences Milligan, G. W. E. and Tei K. 1998. The pull-out
observed regarding the peak displacements. resistance of model soil nails. Soils and Foundations,
38(2) : 179-190.
5 CONCLUSION
The experimental tests here presented need to be de Sauvage, J. 2018. Étude du comportement des murs de
soutènement par clouage des sols en place : application
confirmed with other soils or particle size
au dimensionnement du parement. PhD thesis, ENTPE.
6
Schiavon, J.A. 2016. Behaviour of helical anchors Šmilauer, V. et al, 2015. Yade documentation 2nd ed
subjected to cyclic loadings. PhD thesis. École centrale (The Yade Project), http:// yade-dem.org/doc/
de Nantes, Universidade de São Paulo.
Schlosser, F. et al, 1993. Recommendations Clouterre

1991, French national research project Clouterre
(English translation), Presses de l’ENPC, Paris.
7
Silty sand characterization in the geotechnical centrifuge

Varvara Zania1, Sine Maria Christensen1, Irene Rocchi1
Civil Engineering Department, Technical University of Denmark, Denmark
1
Corresponding author: Varvara Zania (vaza@byg.dtu.dk)
ABSTRACT: The geotechnical characterization of soil deposits from piezocone data can be quite
challenging when it comes to intermediate soils, like silty sand. There are well established and widely used
empirical correlations for the characterization of cohesionless and cohesive soils, however there are still
open questions when it comes to the interpretation of piezocone measurements for silty sand deposits. This
study aims at improving the understanding of the mechanical behavior of silty sand. In order to do so cone
penetration tests with pore pressure measurement at the shoulder of the penetrometer were performed on
silty sand samples in the geotechnical centrifuge. After a literature review on the current sample preparation
methods, the moist tamping method with under compaction has been employed to ensure a homogeneous
silty sand sample. Three samples have been prepared at 23%, 33% and 55% fine content (non plastic). The
stress level effect has been studied by performing CPTUs at different g levels. In addition, the CPTU results
were interpreted to evaluate strength parameters of the soil. Element triaxial tests were carried out and the
results were compared with the interpretations of the CPTU data. Sieve analysis indicate that the
reconstituted samples were homogeneous and uniform, except from the low fine content sample which was
over compacted at larger depths. The interpretation methods are capturing the classification of soil type but
underestimate the fine content for the silty sand with non plastic fines. The results indicate higher peak angle
of friction at higher fine content, which is consistent in both empirical correlations and triaxial test results.
Keywords: Silty Sand; CPTU Interpretation; Soil Classification; Fine Content; Peak Friction Angle.
1 INTRODUCTION Mayne,1990, and Robertson and Cabal, 2015). As

Cone penetration testing (CPT) is one of the most silty sand deposits can be vulnerable to earthquake
common methods for ground profiling and soil loading, their characterization is very important
characterization. In the past decades, several (Han et al, 2016).
publications have contributed to the interpretation The aim of this study is to provide further insight
of CPTs, while the main categorization with respect into the mechanical behaviour of silty sand. A
to the mechanical behaviour has been to coarse- laboratory investigation would require the
grained and fine-grained soils (Robertson, 1990). A reconstitution of silty sand samples; hence it is
unified approach has been suggested lately based on pertinent to first describe the available sample
integrated contours on the normalized soil preparation methods, their applicability in the
behaviour charts (Robertson, 2009). There are well- centrifuge and other relevant drawbacks. After the
established and widely used empirical correlations identification of an appropriate reconstitution
for the characterization of cohesionless and method, silty sand samples have been prepared at
cohesive soils (Robertson and Cabal, 2015, Lunne different fine contents and CPTUs have been
et al. 1997). Nevertheless, the interpretation of performed in the geotechnical centrifuge. The effect
piezocone measurements for silty sand deposits can of the stress levels was examined by the application
be quite challenging, as for the transitional soils it of different centrifugal acceleration. Samples were
depends on the penetration rate, the fine content and retrieved from the strongbox, and sieve analysis and
the plastic behaviour (Martinez et al, 2018). This triaxial tests at similar confinement stresses were
has led to approximate conservative suggestions for carried out in order to compare and evaluate the
the angle of friction of silty sands, like a decrease interpretation methods for silty sand
by 3o of the equivalent angle of friction of the clean characterization.
sand (EN1997-2, 2007) or even a value of 32° is
suggested for low-risk projects (Robertson and 2 SAMPLE PREPARATION METHODS
Cabal, 2015). The reconstitution of a soil sample in the laboratory
Due to the wide application of CPTs in field is expected to provide specimens with densities
testing, design methods for geotechnical problems ranging from loose to dense as found at its natural
like bearing capacity of foundations and deposit. A homogeneous sample, without particle
liquefaction assessment have been developed on the segregation must also be uniform, which can be
basis of the normalized tip resistance (Kulhawy and assessed by the void ratio throughout the sample
9
(Huang et al. 2015). Hereafter an overview of the of fines content by slicing the specimen into five
available methods for preparation of silty sand equal layers. It was concluded that the method of
samples is provided, and then the adopted procedure mist pluviation is able to produce homogeneous
in the current study is described. silty sand specimens, as described by Huang et al.
(2015).
2.1. Literature review Kuerbis & Vaid (1988) evaluated how to prepare
Reconstituted silty sand samples can be prepared in well graded or silty sand samples, such that particle
the laboratory by one of the following methods: (a) segregation is avoided. They proposed the slurry
dry or moist tamping (Amini & Qi, 2000, Kim et al. deposition method, which was further developed by
2014, Ladd, 1978), (b) dry/wet/mist pluviation (Lo Carraro & Prezzi (2008). In the first version of the
Presti et al., 1992, Huang et al., 2015), and (c) slurry slurry deposition method by Kuerbis & Vaid (1988),
deposition (Silva, 2009, Suzuki, 2015). the fines and sand are prepared as two separate
Dry or moist tamping is carried out by sequential slurries and poured into a mixing tube. In the newer
compaction of layers of given thickness with a version by Carraro & Prezzi (2008) the fines and
specified force and frequency of tamping. A sample sand are mixed dry in a closed container until the
prepared with the use of moist tamping method is mass is homogeneous, before pouring it into a tube
partially saturated, and can be loose to dense. The half filled with deaired water through a funnel,
capillary tension forces, which are present during afterwards making it a slurry by shaking and
the process allow the construction of a much looser rotating the closed tube for at least 20 minutes,
sample, than other methods, such as pluviation, removing air bubbles and topping up with deaired
(Kuerbis & Vaid 1988). A modification of the water when necessary. The mixing tube is placed in
tamping method was reported by Ladd (1978) the triaxial apparatus allowing the particles to settle
named the under compaction method, which for around 20 minutes, before the tube is moved up.
minimised the particle segregation for samples with Silva (2009) studied the rate effect of silica silt
a wide range of particle sizes and produced when conducting miniature CPT in the centrifuge.
homogeneous samples. The principle of under The silica flour samples were prepared by mixing it
compaction is to increase the compaction for each as a slurry having a water content of 120%. The
new layer applied i.e. the first layer is lightly slurry was poured into the sample container, but it
tamped, and the tamping is gradually increased with was observed that the finest silt particles were
increasing number of layers. deposited at the surface of the sample due to particle
Dry (or air) pluviation and wet (or water) segregation.
pluviation are similar techniques. The material is
poured dry into a sample container and either 2.2. Soil samples
deposited dry or in water. In the dry pluviation, the The literature review has indicated that the moist
fall height of the particles affects the relative density tamping and the mist pluviation methods can
of the sample. This is due to the energy at which the reportedly produce homogeneous silty sand
particles are deposited. Furthermore, the void ratio reconstituted samples without particle segregation.
and thereby density of the sample is also influenced In the current study, due to the limitations regarding
by the funnel size used for pluviation. A the application of the mist pluviation method in the
homogeneous specimen of well-graded soil without centrifuge, the moist tamping method with under
particle segregation is difficult to obtain with the dry compaction was employed.
pluviation method. This is attributed to both the The soil sample was prepared sequentially in 4
pluviation of sediments and the following saturation layers, with a target thickness of approximately 4cm
(Kuerbis & Vaid 1988). Lo Presti et al. (1992) each, according to the procedure described
reported also particle segregation by assessing the hereafter. Fontainebleau sand and silica flour, which
particle size distribution of subspecimens from a have reportedly 2% and 84% fine content
constructed specimen. It was observed, that the respectively (according to their particle size
finest particles were concentrated in the bottom of distribution curves), were mixed firstly in dry state
the specimen after vibration, which was applied to in pre-calculated ratios to provide different values
obtain a denser specimen. of target fine content. The homogeneous soil mass
The mist pluviation method is carried out like the was then mixed again thoroughly with water
air pluviation, but the soil material falls through a corresponding to 12% water content. The silty sand
mist zone, where the soil particles are mixed with mixture was placed in the strongbox, on top of a pre
water droplets. The mist pluviation method was compacted saturated coarse sand layer 145mm
evaluated by a series of triaxial tests and the thick, which was covered by a geotextile filter. The
determination of water content and spatial variation coarse sand layer was placed in the bottom of the
10
sample container to allow a uniform flow of water during the spinning of the centrifuge. The
through the sample during the saturation phase. The consolidation was not monitored, only the final
thickness of the coarse sand layer was measured thickness was measured at the end of the test. Table
before each sample preparation, and there was no 1 reports the thickness of the silty sand sample
further compaction observed. before the centrifuge tests and after the end of
Each layer of silty sand was compacted by placing testing. The overall settlement was between 5% and
a steel circular plate on the soil surface and using a 10% of the layers thickness depending on the fine
proctor hammer. The weight block on the hammer content.
had a mass of 2.5kg and the fall height was 30cm,
this is equal to a potential energy of 7.36J. The steel 3 CENTRIFUGE TESTING
plate was placed on top of the sample to distribute All the centrifuge tests were conducted in the 2.5m
the compaction force evenly on the entire surface of radius geotechnical beam centrifuge at DTU. A load
the soil. After the layer was compacted the surface frame was mounted on top of the strongbox with
was scarified, before applying the next layer, to four bolts. Afterwards, the centrifuge was
ensure continuity in the interface between layers. accelerated and a constant rotational speed was
The layers were tamped according to the under maintained equivalent to 50g and 70g. The mini
compaction method, which means that the tamping penetometer used for the centrifuge CPTUs has an
of the layers was gradually increased. All samples apex angle equal to 60o, a diameter (B) of 11.3mm
were saturated with deaired water. The water flow and a net cone area ratio equal to 0.55. In flight
was induced from the bottom of the container, while CPTUs were carried out at two locations for each g
the hydraulic gradient was controlled through a level, keeping a distance of at least 6B between
standpipe. The duration of the saturation increased them. The piezocone was installed at a velocity of
with the fine content from 5h to 24h. A water level 1.33mm/s to a penetration depth of about 120 to
of 2cm above the soil surface was maintained after 140mm in order to avoid any boundary effects
the sample was prepared. The water level ensured arising from the proximity to the geotextile. To
that the cone tip would be submerged in water and minimise boundary effects, the CPTUs were
therefore the pore pressure transducer would be kept undertaken at a minimum distance of 150mm from
fully saturated during the acceleration of the the strongbox boundaries. The pore pressure
centrifuge. transducer of the mini penetometer was calibrated
After the centrifuge testing, samples were taken after inflight penetration tests on clean water. The
from the undisturbed part of the container for both tip resistance and sleeve friction is measured by
sieve analysis and triaxial testing. The sieve analysis strain gauges on a full bridge connection, which
of samples from the top and the bottom of the silty were calibrated on axial loading against a reference
sand and from the triaxial samples after testing were calibrated load cell on the data logging system of the
consistent indicating a homogeneous sample. The centrifuge.
fine content was estimated based on both ISO
17892-4 (2016) and ASTM D2487 (2017) 4 RESULTS AND DISCUSSION
standards. The properties of the prepared silty sand
samples are reported in Table 1. The tip resistance, sleeve friction and pore pressure
measurements are presented hereafter. At first one g
Table 1. Properties of the silty sand samples prepared for the level is discussed and then normalized responses for
centrifuge CPTUs. both examined stress levels.
Sample MT1 MT2 MT3 The tip resistance for the three samples is shown
Bulk unit weight, γ1 (kN/m3) 19,7 19,5 19,9 in Figure 1. It can be noticed, that the tip resistance
Bulk unit weight, γ2 (kN/m3) 19,8 19,5 19,7 of the sample with the lower fine content indicates
Fine content (ISO 17892-4) 23,4 32,8 54,8 increasing strength with depth, with a clear
Fine content (ASTM D2487) 22,9 31,8 54,5 indication of change of slope at the interface of the
two intermediate compaction layers (80mm depth).
Layer thickness (mm) 162 170 175
This highlights that the all the layers beyond the
Layer thickness end (mm) 147 165 165
uppermost one were overcompacted. On the other
hand, the tip resistance of the samples with higher
The bulk unit weight has been calculated based on fine content shows that the silty layer is
both overall measurements of mass and volume in homogeneous without any variations or contrasts
the strongbox (γ1) and from the samples for triaxial around the interfaces of the compaction layers.
testing (γ2), and the two values are very close and Furthermore, any effect of the fine content on the tip
indicate a dense silty sand sample. It has to be noted resistance is hardly noticed at shallow depth;
that the thickness of the silty sand layer decreased
11
however beyond 40mm penetration the tip pressure measurement was zeroed along with all the
resistance increases with higher fine content. sensor readings before each cone penetration.
Tip resistance, qc Sleeve friction, fs
[MPa] [kPa]
0 1 2 3 4 5 6
0 0 10 20 30 40 50 60
0
23% FC
23% FC
33% FC
20 33% FC
20
55% FC
55% FC
Cone penetration depth [mm]
40

40
60 60
80 80
100 100
120 120
140 140
Figure 1. Tip resistance variation with depth at 50g for three Figure 2. Sleeve friction variation with depth at 50g for three
silty sand samples. silty sand samples.
The sleeve friction for the three samples is Pore pressure, u2

presented in Figure 2. The sleeve friction is [kPa]
gradually mobilized as the sleeve of the cone 0 50 100 150
0
penetrates into the soil layer, providing reliable
estimates beyond 55mm depth. It is observed that 23% FC
the sleeve friction for the samples with 23% and 33% FC
20
33% fine content are in agreement at shallow depths 55% FC
(up to approximately 40mm which corresponds to
the uppermost and last compacted layer), while only
40
the sample with 55% fine content attains higher
sleeve friction. Nevertheless, the sleeve friction at

deep penetration in the silty sand with 23% fine 60
content is also showing higher resistance indicating
a non homogeneous sample, denser at larger depths.
The sleeve friction for both samples at 33% and 80
55% fine content has a similar trend of increase with
depth and a change in slope at about 65mm
penetration depth. The sleeve friction attains always 100
higher values at the higher fine content.
The pore pressure measurements show initially a
very low response, which can be explained by the 120
fact that the pore pressure transducer is located at
the shoulder of the cone, hence reliable pore
pressure measurements are obtained at a penetration 140
depth greater than the width of the cone. The pore Figure 3. Pore pressure variation with depth at 50g for three
silty sand samples.
12
Higher pore pressure are recorded at the 55% fine
(a) 80
content sample, presumably as a result of the
70 23% FC
increased fine content. There is not consistent
33% FC
increase of pore pressure with fine content as the 60
55% FC
sample with the lowest fine content exhibits higher 50
qc/σ'v [-]
pore pressures than the sample with 33% fine
content. The variation with depth of the pore 40
pressure is similar for all three tests presented here, 30
except from a slight change of slope between 70mm
20
and 100mm for the higher fine content. This was not
recorded in any other CPTU carried out in the same 10
sample. 0
During the retraction of the cone pore pressures 0 2 4 6 8 10
were measured and were equal to the hydrostatic for z/B [-]
all the tests presented here. It is concluded that the
reconstituted silty sand samples were homogeneous
and uniform, apart from the sample with the lowest (b) 80
fine content, which is deemed comparable with the 70
remaining samples only up to a depth equal to 60
40mm.
50
4.1. Effect of stress level qc/σ'v [-] 40
The effect of the stress on the cone penetration has 30

been evaluated by comparing the normalised tip
20
resistance with the vertical effective stress at two
different g levels as shown in Figure 4. It is noticed 10
that the trends of the variation of the normalized tip 0
resistance, are not influenced significantly by the 0 2 4 6 8 10
stress level. On the other hand, the low fine content z/B [-]
sample appears to have a continuously increasing
resistance as a result of the overcompaction of the Figure 4. Normalized tip resistance with normalized depth at
(a) 50g and (b) 70g for three silty sand samples.
deeper layers.
The critical depth, which is defined as the
normalized depth in Figure 5 for the three silty sand
penetration where the normalised tip resistance
becomes constant (Bolton et al, 1999), appears to be samples at two g levels. Pore pressure built up
slightly influenced by the stress level. On the other develops for all the experiments after the cone was
fully embedded in the silty sand, observed as change
hand it is affected by the fine content, increasing
from z/B=2 to z/B=4 for 33% and 55% fine content in slope at about z/B= 1. The variation appears to
keep increasing up to a depth of z/B=2 for most tests
respectively. Considering that the penetration up to
40mm depth, can be reliable for the sample with and exhibit a fairly constant value thereafter. The
23% fine content, the effect of the fine content on slight inclination at greater depths can be related
with the different stress normalization exponent as
the peak response can be evaluated. It is observed
that the peak normalized cone resistance is not also in the case for the normalized tip resistance (see
Figure 4). Irrespectively of the g level the
monotonically related to the fine content. The peak
normalized tip resistance is lower at the higher g normalized excess pore pressure decreases with
decreasing fine content. Moreover, it seems that the
level regardless of the fine content. It also appears
that the stress exponent is dependent on the g level silty sand with 33% fine content has an almost
for the silty sands with 33% and 55% fine content, drained response at z/B greater than 4 with the
excess pore pressure being less than 10% of the
and more specifically it is expected to be higher at
70g compared to 50g. vertical effective stresses. On the other hand, the
silty sand sample at 55% fine content exhibits
The excess pore pressure normalized with the
vertical effective stress is plotted with the excess pore pressure larger than 50% of the vertical
effective stresses.
13
(a) type, which is in agreement with the soil sample
Δu/σ'v composition.
-1 -0,5 0 0,5 1
0
Normalized cone resistance Qtn

4
z/B [-]
10 23%FC_50g
33%FC_50g
55%FC_50g
(b) Δu/σ'v 23%FC_70g
-1 -0,5 0 0,5 1 33%FC_70g
0 55%FC_70g
Normalized friction ratio Fr

2 Figure 6. Normalized tip resistance with normalized friction
ratio for three silty sand samples. Classification scheme with
soil behaviour type zones based on Robertson 2009.
4
z/B [-]
The soil behaviour type index Ic which is

essentially the radius of concentric circles that
6
define the boundaries of the soil type was defined
23% FC
by Robertson (2009) as:
33% FC8
𝐼𝐼𝑐𝑐 = [(3,47 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑄𝑄𝑡𝑡 )2 + (𝑙𝑙𝑙𝑙𝑙𝑙𝐹𝐹𝑟𝑟 + 1,22)2 ]0,5 (1)
55% FC
10 where Qt is dimensionless cone resistance, and Fr is

Figure 5. Normalized excess pore pressure with normalized the friction ratio.
depth at (a) 50g and (b) 70g for three silty sand samples.
Table 2 presents a summary of estimated
parameters for soil classification and
4.2. CPT based soil characterization characterization. The Ic obtained for each one of the
The centrifuge CPTUs were interpreted based on tests are also reported. It is noticed that the CPTU
existing soil classification and characterization of the sample with 33% fine content exhibits a slight
schemes. In Figure 6 normalized cone resistance is decrease of the Ic with the increase of the stress
plotted with respect to the normalized friction ratio level. It has been suggested that there is a link
for all samples at 50g and 70g. The soil between the compressibility of soil and the Ic, which
classification according to the unified soil would indicate a less compressible silty sand at
behaviour type approach suggested by Robertson higher stress levels. The stress dependency is also
(2009) is also shown. It is evident that the results of represented by the stress normalization exponent,
the sample with the low fine content are unreliable which was defined by Robertson (2009) as:
especially at 70g. This is due to the high sleeve
friction recorded in this particular test. The recorded 𝜎𝜎′𝑣𝑣0
𝑛𝑛 = 0,38𝐼𝐼𝑐𝑐 + 0,05 � �− 0,15 (2)
sleeve friction in this experiment showed an erratic 𝑝𝑝𝑎𝑎
response with an increase for every subsequent
penetration of the cone. The results of the low fine where 𝜎𝜎′𝑣𝑣0is the vertical effective stress, and pα is
content at 50g show an increase of the normalised the atmospheric pressure. This expression predicts
tip resistance as the penetration progressed to the decrease of the exponent with the increase of the g
deeper more compacted layers. Overall all the level for the sample with 33% fine content and the
samples are characterized as ‘Sand mixtures: silty opposite for the sample with the higher fine content.
sand to sandy silt’ in regards with the soil behaviour Nevertheless, the centrifuge CPTUs have provided
14
a trend of increasing exponent with fine content at Triaxial tests were carried out at confinement
low stresses and higher stress exponents at higher stress equal to 60kPa corresponding to the vertical
stress level. The stress exponent provided by effective stress at the middle of the silty sand layer.
Equation 2 is in agreement with the centrifuge The shearing rate was slow enough to ensure fully
CPTUs only for the higher fine content. drained conditions. The estimated values of the
The fine content can also be estimated based on Ic angle of friction from the CPTU interpretation and
according to the following expression: those derived from the triaxial tests are reported in
Table 2. The range of friction angles provided for
𝐹𝐹𝐹𝐹 = 1,75𝐼𝐼𝑐𝑐3,25 − 3,7 (3) the sample with the low fine content is quite wide as
expected from the increase of the normalized tip
where Ic is the soil behaviour type index. Recent resistance which resulted from the over compaction
studies suggest that current correlations appear to of the deeper layers. The NTH method provides
underestimate the fines contents (Yi, 2014). This is consistently higher values compared to those by Eq.
in accordance to the findings of the centrifuge 4 and 5. It should be mentioned that the normalised
CPTUs. As seen in Table 2 the calculated fine excess pore pressure (Bq) has a very low value
content is lower than the resulting of the sieve ranging from 0,01 to 0,04 at 33% and 55% fine
analysis reported in Table 1. content respectively. These three methods do not
The drained peak friction angle was estimated appear to capture any effect of the fine content on
based on four different methods: the peak friction angle. However, an increase of the
(a) the expression by Robertson and Campanella peak friction angle with the increase of the fine
(1983): content was obtained in the triaxial tests, which was
related to change from contractive (33% fine
1 𝑞𝑞 content) to dilative (55% fine content) behaviour.
𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡′ = �𝑙𝑙𝑙𝑙𝑙𝑙 �𝜎𝜎′𝑐𝑐 � + 0,29� (4)
2,68 𝑣𝑣0 The sample with 55% fine content had a dilation
angle equal to 13o. The increase of the peak friction
where qc is the cone tip resistance and 𝜎𝜎′𝑣𝑣0is the angle with the higher fine content is also captured in
vertical effective stress. the last method (Robertson, 2009) as shown in the
(b) the expression by Kulhawy and Mayne (1990): contour plots in Figure 6.
0,5
𝑞𝑞
𝜑𝜑′ = 17,6 + 11𝑙𝑙𝑙𝑙𝑙𝑙 ��𝑝𝑝𝑐𝑐 �� 𝑝𝑝𝑣𝑣0 �
𝜎𝜎′
� (5) 5 CONCLUSION
𝑎𝑎 𝑎𝑎
An experimental study in the geotechnical
centrifuge was performed aiming to provide some
where 𝜎𝜎′𝑣𝑣0is the vertical effective stress, qc is the
insight into the mechanical behaviour of silty sands.
cone tip resistance and pα is the atmospheric
Homogeneous samples were reconstituted with the
pressure.
wet tamping method using under compaction,
(c) the NTH method as suggested by Senneset et
however only the samples with higher fine contents
al. (1989) and Mayne and Ouyang (2018)
were uniform. The classification of the samples was
and (d) the peak friction angle contours on the
in accordance to the anticipated soil behaviour.
normalized soil behaviour type chart by Robertson
However, the fine content was underestimated by
(2009).
current methods. The peak friction angle was found
Table 2. Interpretation of mechanical properties of the silty to increase with the increase of the fine content.
sand samples.
Sample
MT1 MT2 MT2 MT3 MT3 6 ACKNOWLEDGEMENTS
50g 50g 70g 50g 70g The assistance of MSc student Efthymios
Ic (eq.1 ) 1,7-2,1 2,4 2,2 2,4 2,5 Panagiotis in the centrifuge and triaxial experiments
0,58- 0,75- 0,73- 0,77- 0,82- and the support of the technical personnel at DTU is
n (eq.2 )
0,66 0,78 0,74 0,79 0,83 gratefully acknowledged.
n (Fig.4) 0,55 0,95 0,77 0,85
FC (eq.3) 6-12 18-20 15-16 19-21 22-23 7 REFERENCES
φ’ (eq.4) 37-42 33-35 35 35 34 Amini, F., & Qi, G. Z. 2000. Liquefaction testing of
φ’ (eq.5) 33-40 32 33-35 33 33 stratified silty sands. Journal of Geotechnical and
φ’ NTH >40 34-36 36-37 36-37 37 Geoenvironmental Engineering, 126(3): 208–217.
φ’ TRX 34 30 - 39 39
15
ASTM D2487-17 2017. Standard Practice for Lunne, T., Robertson, P.K., & Powell, J.J.M., 1997. Cone
Classification of Soils for Engineering Purposes penetration testing in geotechnical practice. Blackie
(Unified Soil Classification System), ASTM Academic, EF Spon/Routledge Publ., New York.
International, West Conshohocken, PA.
Martínez, M., Tonni, L., Gottardi, G. & Rocchi, I. 2018.
Bolton, M. D., Gui, M. W., Garnier, J., Corté, J. F., Analysis of CPTU data for the geotechnical
Bagge, G., Laue, J. & Renzi, R. 1999. Centrifuge cone characterization of intermediate sediments. In Cone
penetration tests in sand. Géotechnique, 49: 543–552. Penetration Testing 2018 : Proceedings of the 4th
International Symposium on Cone Penetration Testing,
Carraro, J. A. H. & Prezzi, M. 2008. A New Slurry-Based CRC Press, pp. 281–287.
Method of Preparation of Specimens of Sand
Containing Fines. Geotechnical Testing Journal, 31(1). Mayne, P. & Ouyang, Z. 2018. Calibrating NTH method
for -’ in clayey soils using centrifuge CPTu, in
DS/EN ISO 17892-4 (2016), Geotechnical investigation ‘Proceedings of the 4th International Symposium on
and testing - laboratory testing of soil - part 4: Cone Penetration Testing’.
Determination of particle size distribution.
Robertson, P.K. & Campanella, R.G. 1983. Interpretation
EN 1997, 2007, Eurocode 7: Geotechnical design - Part of cone penetration tests. Part I: sand. Canadian
2: Ground investigation and testing, Danish Standards. Geotechnical Journal, 20(4): 719–733.
Han, J.-T., Kim, T. & Cho, W. 2016. Centrifuge Robertson, P. K. 1990. Soil classification using the cone
modeling of seismic response of normally consolidated penetration test. Canadian Geotechnical Journal, 4,
deep silt deposit. Acta Geotechnica, 11(1): 71–81. 151–158.
Huang, A.-B., Chang, W.-J., Hsu, H.-H. & Huang, Y.-J. Robertson, P. K. 2009. Interpretation of cone penetration
2015. A mist pluviation method for reconstituting silty tests – a unified approach. Canadian Geotechnical
sand specimens. Engineering Geology, 188: 1–9. Journal, 46, 1337–1355.
Kim, J. H., Kim, S. R., Lee, H. Y., Choo, Y. W., Kim, D. Robertson, P. K. & Cabal, K. 2015. Guide to Cone
S. & Kim, D. J. 2014. Miniature cone tip resistance on Penetration Testing for Geotechnical Engineering, 6th
silty sand in centrifuge model tests. Physical Modelling edn, Gregg Drilling & Testing, Inc.
in Geotechnics, 1301–1306.
Senneset, K., Sandven, R. & Janbu, N. 1989. Evaluation
Kuerbis, R. & Vaid, Y. P. 1988. Sand Sample Preparation of soil parameters from piezocone tests, Transportation
- The Slurry Deposition Method. Soils and Research Record, No. 1235, 24–37.
Foundations, 28(4): 107–118.
Silva, M. 2009. Rate effects in centrifuge piezocone
Kulhawy, F. & Mayne, P. 1990. Manual on estimating penetration on dilatant silica silt and subsequent
soil properties for foundation design. dissipation. Australian Geomechanics Journal, 44, 19–
28.
Ladd, R. S. 1978. Preparing test specimens using
undercompaction. Geotechnical Testing Journal, 1(1): Suzuki, Y. 2015. Investigation and interpretation of cone
16–23. penetration rate effects, PhD thesis, The University of
Western Australia.
Lo Presti, D., Pedroni, S. & Crippa, V. 1992. Maximum
dry density of cohesionless soils by pluviation and by Yi, F. 2014. Estimating soil fines contents from CPT
ASTM d 4253-83: A comparative study. Geotechnical data. In Proc. 3rd International Symposium on Cone
Testing Journal, 15(2): 180–189. Penetration Testing, Huntington Beach, Las Vegas,
Nevada, USA, pp. 949-955.
16
The effects of sand grading on the bearing capacity of surface foundations

Mathias R. Jensen1, Varvara Zania1, Barry M. Lehane2
Civil Engineering Department, Technical University of Denmark, Denmark
1
Civil Engineering Department, University of Western Australia, Australia

2
Corresponding author: Varvara Zania (vaza@byg.dtu.dk)
ABSTRACT: The bearing capacity of surface foundations in cohesionless soil has been studied extensively
in the past decades. Experimental investigations have shown that particle size effects in the bearing capacity
problem are associated not only with strength mobilization but also with the shear band thickness. Centrifuge
testing can provide useful insights, especially when taking into consideration the scale effects, by using a
modelling of models approach. This paper describes the results of a systematic study examining the effects
of sand grading on the responses of surface footings. Four different sands, including two commercially
sourced silica sands and two mixed silica sands, were deposited by air pluviation and tested under controlled
conditions in the geotechnical centrifuge at Technical University of Denmark (DTU). To examine
correlations often employed in the field between tip resistance (qc) of Cone Penetration Tests (CPTs) and
footing bearing pressure (qf), a series of CPTs were conducted in samples prepared at similar conditions to
the samples of the footing tests. The results indicate that the bearing capacity is controlled by a general shear
failure mechanism and that further research is required to assess the range of applicability of direct
relationships between bearing capacity and qc for surface footings.
Keywords: Footings; CPT; Centrifuge Modelling; Sands; Sand Grading.
1 INTRODUCTION 𝑞𝑞 = 𝑟𝑟𝑠𝑠 · �𝑠𝑠/𝐵𝐵 (1)

The traditional bearing capacity approach first
proposed by Terzaghi in 1943 has been widely where rs is a soil parameter. In the case of sands,
investigated since its formulation and new insights many researchers have examined correlations
on the bearing resistance of footings have been between rs with the tip resistance qc, e.g. Mayne and
obtained. These include the influence of (a) the Illingworth (2010) Mayne et al. (2012), Lehane
progressive failure as strain localization can lead to (2013), Mayne and Woeller (2014) and Mayne &
hardening and softening at critical states, (b) Dasenbrock (2018). Mayne and Illingworth (2010)
anisotropy, (c) shear band thickness and (d) analysed a database of field tests and derived a mean
nonlinear peak strength envelope (Lau and Bolton, trend for footings on sand that indicates rs = 0.585qc.
2011). The effect of the particle size has been When using this value in Eq. 1 for the capacity
associated with the relative thickness of the shear criterion of s/B = 10%, the bearing pressure (q0.1) is
band (Tatsuoka et al, 1997). In current practice, a equal to 0.18qc; this value is very similar to the
direct correlation with the CPT tip resistance (qc) is value of 0.16qc determined independently by
employed, where the reference “capacity” is defined Lehane (2013).
by the well-known 0.1B Method corresponding to However, experimental and numerical studies
the bearing pressure (q) required to cause a footing have shown that footing tests on dense sand are
to settle by 10% of its width s/B = 10% (Amar et al. controlled by a general shear failure mechanism
1998). The bearing capacity (qf) calculated using the (Kimura et al. 1985; Cerato and Lutenegger 2007;
standard bearing capacity equation is usually much Mase and Hishiguchi 2009). On the other hand, the
larger than the one determined from a direct CPT mechanism reported for the footing tests conducted
method at s/B = 10% (q0.1), as it requires settlements in the field (up to s/B= 10%) is analogous to
well in excess of 10% of the foundation width to expansion of a spherical cavity (e.g. Lehane 2013).
develop, especially in the case of a local shear The direct CPT method relies on a similarity
failure (Lau and Bolton, 2011). between the failure mechanism of the footing and
Briaud and Gibbens (1999) observed that the the deep penetration CPTs.
response of footing load tests conducted in the field The aim of this study was to investigate the effect
can be unified via a “characteristic load- of sand grading on the bearing capacity and to
displacement curve” when the data are presented in evaluate the uniqueness of the relationship between
terms of bearing pressure versus normalised deep penetration qc and q0.1 for different sand types.
displacement: A centrifuge study was undertaken, where soil
17
conditions are controlled. Four silica sands with 100
Leighton Buzzard B
similar particle regularity, including two uniformly 90 Leighton Buzzard C
graded sands and two mixed sands with a wider 80 Leighton Buzzard D
Percentage passing (%)

grading variation, were therefore tested. To confirm 70
Mix 1
Mix 2
expectations of modelling of models in centrifuge 60
testing of footings (e.g. Ovesen 1975), two surface 50
footings of different diameters (40mm and 60mm) 40
were tested at two different g levels (40g and 60g) 30
for each sand. These models both represent a 20
circular footing of 2.4m diameter at the prototype 10
scale. Furthermore, two centrifuge CPTs were 0
carried out for each sand at two different g levels 0,0 0,1 1,0
Particle size (mm)
10,0
(ranging from 32.2g to 64.5g) in new samples Figure 1. Particle size distribution curves for the sands in this
prepared at a similar relative density (Dr). study.
2 SOIL PROPERTIES 2.1. Basic indices

The uniformly distributed fractions C and D of The basic classification indices of the four sands
Leighton Buzzard sand (referred to as LB-C and investigated are summarised in Table 1. All sands
LB-D in this study) were tested. In addition, two have very similar particle shapes, but a slight
sands were mixed from the LB-C and LB-D sands variance in uniformity coefficient (Cu = d60/d10) of
and a third Leighton Buzzard fraction (fraction B, 1.49 to 2.80 (where d60 and d10 are the mesh sizes
LB-B). Mix 1 comprises 1/3 of each of the three through which 60 percent and 10 percent of the sand
fractions, and Mix 2 comprises 1/5 of fractions B particles pass, respectively). LB-C and LB-D are
and C and 3/5 of fraction D. The two mixed sands commercially sourced silica sands with distinctly
were mixed with a concrete mixer. Several sieve different mean effective particle size (d50). In order
analyses were carried out for each mixed sand to to specifically investigate the effect of Cu in this
ensure that the sands were mixed homogeneously study, the two mixed sands were mixed to have a
and the particle distribution curves are presented in similar d50 to the LB-C sand. A long series of direct
Figure 1. shear tests indicated that the two mixed sands have
slightly higher critical state friction angles (φ’cs)
than LB-C and LB-D (Jensen, 2019). The void ratio
limits were determined following Danish
Geotechnical Society Guidelines (DGF 2001).
Table 1. Index properties of sands investigated

Gradation Void ratio limits Particle shape Gs φ’cs: degrees
Sand
d50: mm Cu emax emin R S ρ
Leighton Buzzard C (LB-C) 0.40 1.54 0.78 0.53 0.51 0.75 0.63 2.66 31.4
Leighton Buzzard D (LB-D) 0.19 1.49 0.87 0.59 0.46 0.70 0.58 2.66 30.7
Mix 1 0.37 2.80 0.71 0.43 0.49 0.73 0.61 2.66 32.6
Mix 2 0.38 2.31 0.74 0.47 0.50 0.74 0.62 2.66 32.7
Figure 2. Microscope images of sand particles: (a) LB-B; (b) LB-C; (c) LB-D.
18
2.2. Particle shape quantification hinge above it to allow free rotation. This piston was
Representative optical microscope images for the then used in-flight to apply load to the footings. A
sands are provided in Figure 2 (including Leighton stiff plastic beam was screwed onto the piston and
Buzzard fraction B which is used for the mixed was functioning as a reference point for the LVDT
sands), where the sands are shown to have very mounted on the side of the loading frame. The
similar particle shapes. Visual observations were vertical displacement of the footings during
used to derive the roundness (R), mean sphericity penetration was therefore monitored with better
(S) and regularity (ρ) provided in Table 1 using the precision than the one obtained from the control of
methodology described by Jensen et al. (2019). the linear actuator. The LVDT was regretfully not
mounted for Test 1 and therefore the load-
displacement curve for this test is not deemed
3 CENTRIFUGE TESTS
reliable at small displacements. All footings were
3.1. Sample preparation and test programme placed on the sand surface in the centre of the
The sand was pluviated into the strongbox (having strongbox to avoid boundary effects. The cone used
an internal diameter of 527mm and an internal for the centrifuge CPTs has a diameter (dc) of
height of 498mm) to give a final height of 317mm 11.3mm and was installed at a velocity of 1.33mm/s
when levelled off. The sand pluviation was to a penetration of about 200 to 240mm. To
performed at 1g and manoeuvred manually in a minimise boundary effects, the CPTs were
circular motion at a constant drop height above the undertaken at a minimum distance of 150mm to the
sand surface. In order to control the density of the nearest boundary.
sample, the sand flow and the drop height were
adjusted (with increasing flow and decreasing drop 3.2. Footing test results
height leading to a lower density). All tests were The variation of bearing stress (q) with the
performed in a dry condition. settlements normalised by the respective footing
The footings were 16mm thick and made of diameters (s/B) are provided in Figure 3. In
aluminium. The testing programme of the footing accordance with modelling of models in centrifuge
tests is shown in Table 2 (including Dr, g level and testing of footings (e.g. Ovesen 1975), the responses
footing diameter B). The smallest B/d50 ratio is 100, up to s/B = 10% of the 40mm diameter footings
which ensures that centrifuge scale effects are (60g) are nearly the same as those of the 60mm
negligible (Ovesen 1979). Furthermore, two diameter footings (40g). Test 1 deviates from this
centrifuge CPTs were carried out for each type of trend as a result of the low accuracy of the measured
sand at two different g levels ranging from 32.2g to displacement. The footing responses seem to differ
64.5g. The CPTs were performed in different at footing penetrations in excess of s/B = 10% for
samples to the footing tests, however, these were LB-C and Mix 2, whereas excellent agreement
prepared at nearly the same relative density; with appears for the entire footing penetrations for the
the largest deviation being the CPTs performed in LB-D and Mix 1 sands.
the Mix 2 sand at Dr = 85%. Experimental centrifuge work and numerical
simulation studies have shown that bearing capacity
Table 2. Footing testing programme
factors are size-dependent for footings on sand, with
Test Dr B B/d50
no.
Sand
(%)
g level
(mm)
the bearing capacity (qf) decreasing with footing
1 LB-C 81 40 60 150 width (or diameter) (Kimura et al. 1985; Cerato and
40 100 Lutenegger 2007; Mase and Hishiguchi 2009).
2 LB-C 82 60
60 316
These studies show that a footing placed on a dense
3 LB-D 82 40
40 211
sand is expected to fail in a general shear failure.
4 LB-D 82 60
One characteristic of this failure mode is heaving of
5 Mix 1 84 40 60 162
the soil surface adjacent to the footing. This
6 Mix 1 82 60 40 108
characteristic was observed after every footing test
7 Mix 2 80 40 60 158
in this study; see an example in Figure 4. The
8 Mix 2 80 60 40 105
bearing pressure of a footing in general shear
rapidly reaches a peak and a subsequent plateau,
All footing tests were conducted under after which the surcharge around the footing (Nq-
displacement control in the 2.5m radius term in Terzaghi’s bearing capacity formula) makes
geotechnical beam centrifuge at DTU. A load frame the bearing pressure increase to very large depths.
was secured on top of the strongbox with four bolts. The footing responses plotted in Figure 3 indicate
For the footing tests, a piston was installed with a that the footings in this study failed in general shear.
19
Figure 3. Bearing stress versus normalised settlement for footings on the four investigated sands.
Figure 4. Example of a general shear failure observed after

one of the footing tests.
When following a procedure proposed by Lau and Figure 5. Footing results of Test 3 and 4; with Terzaghi’s
Bolton (2011) with Terzaghi’s bearing capacity bearing capacity formula superimposed for different friction
formula superimposed for different friction angles, angles.
it is seen that the maximum mobilized friction
angles of the footing tests occur at or close to s/B = However, in field tests, where footings are
10%. This is illustrated in Figure 5, where Test 3 generally embedded in the sand, a sudden collapse
and 4 are plotted (note that γ’ is the effective unit (or failure) is rarely seen. Therefore, the most
weight of the sand). Therefore, it can be inferred common response shows no clear peak or plateau,
that q0.1 approximately coincides with the maximum and the bearing pressure increases to very large
mobilized friction angle when the sand fails in depths. The same type of response was observed
general shear. It is also noticed that the equivalent from embedded footing tests conducted in the
friction angle at s/B = 10% is higher than the critical centrifuge by Liu and Lehane (2019). In fact, these
state angle shown in Table 1. tests have shown that the mode of deformation at
20
s/B = 10% is similar to that of cavity expansion
without failure planes extending to ground level
(e.g. see Lehane 2013). It follows from this failure
mechanism that q0.1 vs. s/B characteristics are
largely independent of footing width (or diameter).
3.3. Effects of sand grading on q0.1

The bearing stresses measured at normalized
settlements s/B = 10% (q0.1) are generally very
similar between all four sands (1600±200 kPa), with
the highest and lowest measured q0.1 for LB-D and
LB-C, respectively (see Figure 3). Slightly higher
values of q0.1 (≈12-15% higher) were obtained for
the two mixed sands in comparison to LB-C sand
(all three sands having approximately the same d50).
Hence, some effect of the sand uniformity is
observed. Furthermore, an effect of the mean
particle size d50 appears on q0.1. This is seen with the
LB-D sand (having a smaller d50 than the other three
sands) taking ≈22% higher values of q0.1 when
compared to LB-C sand and ≈5-8% higher values
than the mixed sands. The peak friction angles Figure 6. Evidence of upward concavity in the CPT profiles.
found from direct shear tests for the different sands,
The tip resistance in the shallow depth phase is
have been higher for the mixed sands following a
associated with upward movements of the surface
1:1 association with Cu (Jensen, 2019). This
around the cone rod, which is similar to a general
indicates that besides the friction angle the bearing
shear failure of a footing. At a certain depth (deep
capacity is strongly associated with the d50 , which
penetration) a localised mechanism in the vicinity
could be attributed to the shear band thickness
of the cone tip takes over, and qc can be predicted
developed in the general shear failure.
using spherical cavity expansion theory (e.g. Yu and
Mitchell 1998; Suryasentana and Lehane 2016). In
4 RELATING qc WITH (CENTRIFUGE)
this phase the increase in normalised tip resistance
FOOTING RESPONSE slows down rapidly and enters a quasi-stationary
In the generalised direct CPT method for evaluation regime. At deep penetration in a homogeneous
of footing response, which is often used in the field, normally consolidated sand, qc increases as a
the load-displacement curves are normalised by the constant function of vertical effective stress (σ’v).
cone tip resistance (qc). The bearing capacity An ideal correction method should therefore
(evaluated at s/B = 10%) can be capped at stresses produce an equal normalised tip resistance at
corresponding to a percentage of the average different depths in the same soil at the same relative
measured qc (qc,avg), determined over a depth density (Dr).
interval which according to Mayne and Woeller The tip resistance can be normalised (or
(2014) can be taken from founding level to 1.5B corrected) to a common σ’v:
depth below. For this purpose the centrifuge CPTs
(of which some are shown in Figure 6) were ′
𝜎𝜎𝑣𝑣,𝑟𝑟𝑟𝑟𝑟𝑟
𝑐𝑐
corrected following the procedure described below. 𝑞𝑞𝑐𝑐,𝑟𝑟𝑟𝑟𝑟𝑟 = 𝑞𝑞𝑐𝑐 · � 𝜎𝜎𝑣𝑣′
� (2)
4.1. Correction for shallow depth effect on qc where σ’v,ref is a reference vertical effective stress,
For a 2.4m diameter footing, a CPT conducted in the qc,ref is qc at σ’v,ref and c is a stress normalisation
field will have penetrated a depth of more than 100 exponent.
cone diameters before the tip reaches a depth of As evident in Figure 6, the shallow penetration
1.5B. However, the centrifuge CPTs in this study phase is characterised by a parabolic increase
had penetrated 5.3dc (40g) and 8.0dc (60g) at 1.5B (upward concavity) of qc with depth; this trend was
depth. Therefore, in order to draw parallels to a field noted by Puech and Foray (2002), Senders (2010)
situation, the CPT qc values were corrected in the and Bolton et al. (1999). This upward concavity
upper soil horizon due to shallow depth effects (e.g. means that the stress exponent (c) in Eq. 2 decreases
see Gui and Bolton 1998). with depth until at a critical depth (dcr) when the tip
21
resistance enters the quasi-stationary regime and c Table 3. Estimations for all the centrifuge CPTs.
takes on a nearly constant value (steady state). In qc,avg
Sand (z/dc)cr c (steady state) Q
order to evaluate c with depth, Eq. 2 can be (MPa)
rearranged: LB-C 16 1.10 140 5.89
LB-D 10 1.05 185 6.27
𝑙𝑙𝑙𝑙�
𝑞𝑞𝑐𝑐,𝑟𝑟𝑟𝑟𝑟𝑟
� Mix 1 15 0.60 1220 9.53
𝑞𝑞𝑐𝑐
𝑐𝑐 = 𝜎𝜎′𝑣𝑣,𝑟𝑟𝑟𝑟𝑟𝑟
(3) Mix 2 14 0.55 1750 11.52
𝑙𝑙𝑙𝑙� ′ �
𝜎𝜎𝑣𝑣
It is noticeable that relatively high values of ‘c’
The variations of Eq. 3 with the normalised are reported in this study when compared with
penetration depth (z/dc) for the LB-C and Mix 2 values reported in calibration chamber tests (e.g. see
sands are plotted in Figure 7, where the transition Moss et al. 2006). However, values of ‘c’ tend to be
between shallow and deep penetration is illustrated closer to unity in centrifuge tests (Bolton et al.
with a critical ratio of penetration depth to cone 1999), which may be an artefact of the ratio of the
diameter, (z/dc)cr. The “steady state” value of c and lower cone diameter to the stress gradient. The ‘c’
(z/dc)cr for all four sands are listed in Table 3. values greater than unity seen for LB-C and LB-D
sands are not credible, since values of c>1 indicate
an upward concavity of qc even at deep penetration.
‘c’ values less than one should generally be
expected in sands with uniform relative densities
due to the reducing tendency for dilation at higher
stress levels.
Jensen et al. (2019) observed that increasing
values of uniformity coefficient (Cu) leads to sands,
which are more dilatant at small stress levels and
require smaller stress levels for the sand behaviour
to become contractive. Therefore, given that the tip
resistance is influenced by these two factors, some
variations in the steady state c could be expected. It
is possible that the uniformity of the sands (the
Figure 7. Variation of stress exponent (c) with normalized

penetration depth (z/dc)
Figure 8. Normalized tip resistance (Q) versus normalized penetration depth (z/dc) for all centrifuge CPTs.
22
mixed sands having higher Cu values than the LB-C
and LB-D sands) contributes to some extent to the
observed variations of the c exponent, although the
magnitude of differences seen in Table 3 indicate
that further research is required.
With the estimated “steady state” c values (in
Table 3), the normalized tip resistance (Q) is
determined by:
𝑄𝑄 = 𝑞𝑞𝑐𝑐 /(𝜎𝜎𝑣𝑣′ )𝑐𝑐 (4)
The variations of Eq. 4 with z/dc for all four sands

are shown in Figure 8 where Q is seen to develop a Figure 9. Variation of normalised bearing stress (q/qc,avg)
near constant value after the critical ratio of with normalised settlement (s/B) for all footing tests.
penetration to cone diameter (z/dc)cr. The best-fit
values of the near constant Q are listed in Table 3.
Note that the significantly higher values of Q for the 5 CONCLUSIONS
mixed sands compared with LB-C and LB-D sands A centrifuge surface footing and CPT investigation
are due to the lower values of c (see Eq. 4). In order conducted on four sands at the same relative density
to correct the centrifuge CPT data for shallow but with different particle size distributions has
penetration effects (which are negligible in the shown:
field), the near constant values of Q are assumed to
a) Approximately 12-15% higher values of
extend all the way to the soil surface (z/dc = 0).
q0.1 was observed with a slight increase in
uniformity coefficient (Cu).
4.2. Normalisation of (centrifuge) surface
b) An apparent increase by about 20% in q0.1
footing response with qc,avg values when the mean particle size (d50) was
The corrected qc values averaged to a depth of 1.5B lowered by 50%.
below founding level (qc,avg) are determined for each
sand and listed in Table 3. The bearing pressures (q) The trends and the normalisation using CPT qc of
measured in each footing test are then normalised centrifuge test load displacement data are the
by the corresponding qc,avg and plotted against s/B in subject of an on-going study at the Technical
Figure 9. It can be inferred that the q/qc,avg vs. s/B University of Denmark.
characteristic changes with the uniformity of the
sand. The q0.1/qc,avg ratios range from 0.25 to 0.29 6 ACKNOWLEDGMENTS
for LB-C and LB-D sands and 0.15 to 0.18 for Mix
The authors appreciate the support of technical
1 and Mix 2. This is mainly due to the large
personnel at DTU in the experimental study.
variation in qc,avg between the mixed sands and the
LB-C and LB-D sands (which is primarily due to the
7 REFERENCES
very different stress exponents; see Table 3).
Further research is required to firstly establish how Amar, S., Baguelin, F., Canepa, Y., & Frank, R. 1998.
New design rules for the bearing capacity of shallow
the stress exponent ‘c’ varies with the sand
foundations based on Menard pressuremeter tests.
properties so that the appropriate qc,avg value can be Geotechnical Site Characterization, Vols 1 and 2, 727–
obtained. It is noteworthy that Liu and Lehane 733.
(2020) showed relatively unique q/qc,avg vs s/B
characteristics for embedded footings in four Bolton, M. D., Gui, M. W., Garnier, J., Corte, J. F.,
different sand types and with a range of mean Bagge, G., Laue, J., & Renzi, R. 1999. Centrifuge cone
particle sizes, mineralogies and uniformity penetration tests in sand. Geotechnique, 49(4): 543–
coefficients. The Liu & Lehane (2020) results are 552.
comparable to those of LB-C and LB-D on Figure
9, which have ‘c’ values close to unity. Briaud, J. L., & Gibbens, R. 1999. Behavior of five large
spread footings in sand. Journal of Geotechnical and
Geoenvironmental Engineering, 125(9): 787–796.
23
Cerato, A. B., & Lutenegger, A. J. 2007. Scale effects of method based on cone penetration tests. Geotechnical
shallow foundation bearing capacity on granular Special Publication, (227), pp. 664–679.
material. Journal of Geotechnical and
Geoenvironmental Engineering, 133(10): 1192–1202. Mayne, P. W., & Woeller, D. J. 2014. Generalized direct
CPT method for evaluating footing deformation
DGF 2001. Laboratoriehåndbogen - DGF Bulletin 15. response and capacity on sands, silts, and clays.
Dansk Geoteknisk Forening, Lyngby, Denmark. Geotechnical Special Publication, (234), pp. 1983–
1997.
Gui, M. W., & Bolton, M. D. 1998. Geometry and scale
effects in CPT and pile design. Geotechnical Site Mayne, P. W., & Dasenbrock, D. 2018. Direct CPT
Characterization, Vols 1 and 2, pp. 1063–1068. Method for 130 Footings on Sands. Geotechnical
Special Publication, pp. 135–146.
Jensen, M.R., Zania, V., Lehane, B.M., Liu, Q., & Moss, R. E. S., Seed, R. B., & Olsen, R. S. 2006.
Butterworth, C. 2019. The strength of Perth sands in Normalizing the CPT for overburden stress. Journal of
direct shear. In: Proc. of the 13th ANZ Geomechanics Geotechnical and Geoenvironmental Engineering,
Conference, 1-3 April, Perth, Australia, pp. 185-190. 132(3): 378–387.
Jensen, M.R. 2019. Effects of Particle Size, Uniformity, Ovesen, N. K. 1975. Centrifugal testing applied to
and Particle Shape on the Strength and Bearing bearing capacity problems of footings on sand.
Capacity of Sands. MSc Thesis, Technical University Géotechnique, 25(2): 394–401.
of Denmark.
Ovesen, N. K. 1979. Panel discussion: The scaling law
Kimura, T., Kusakabe, O., & Saitoh, K. 1985. relationship. Proc. 7th Eur. Conf. Soil Mech. and
Geotechnical model tests of bearing capacity problems Found. Engng, Brighton, 4, pp. 319–323.
in a centrifuge. Géotechnique, 35(1): 33–45.
Puech, A., & Foray, P. 2002. Refined Model for
Lau, C.K., and Bolton, M.D. 2011. The bearing capacity Interpreting Shallow Penetration CPTs in Sands.
of footings on granular soils. II: Experimental Proceedings of the Annual Offshore Technology
evidence. Géotechnique 61(8): 639-650. Conference, pp. 2441–2449.
Lehane, B. M. 2013. Relating foundation capacity in Robertson, P. K., 2009. Interpretation of cone penetration
sands to CPT qc. Geotechnical and Geophysical Site tests – a unified approach. Can. Geotech. J. 46 (11),
Characterization 4 - Proceedings of the 4th 1337–1355.
International Conference on Site Characterization 4,
ISC-4, 1, pp. 63–81. Senders, M. 2010. Cone resistance profiles for laboratory
tests in sand. Proc. 2nd Int. Conf. on Cone Penetration
Liu & Lehane (2020). A systematic investigation of the Testing, Huntington Beach, CA, paper no. 2-08.
load-displacement response of shallow footings on
sand. Geotechnique, //doi.org/10.1680/jgeot.19.P.253 Suryasentana, S. K., & Lehane, B. M. 2016. Updated
CPT-based p–y formulation for laterally loaded piles in
Mase, T., & Hashiguchi, K. 2009. Numerical analysis of cohesionless soil under static loading. Geotechnique,
footing settlement problem by subloading surface 66(6): 445–453.
model. Soils and Foundations, 49(2), 207–220.
Tatsuoka, F., Goto, S., Tanaka, T., Tani, K., Kimura, Y.,
Mayne, P. W. & Illingworth, F. 2010. Direct CPT method 1997. Particle size effects on bearing capacity of
for footings on sand using a database approach. footing on granular material. Proc. of IS-NAGOYA,
Proceedings, 2nd Intl. Symposium on Cone pp. 133–138.
Penetration Testing, Vol. 3, pp. 315-322.
Yu, H. S., & Mitchell, J. K. 1998. Analysis of cone
Mayne, P. W., Uzielli, M., & Illingworth, F. 2012. resistance: Review of methods. Journal of
Shallow footing response on sands using a direct Geotechnical and Geoenvironmental Engineering,
124(2): 140–149.
24
Tests of varied sample preparation methods for centrifuge modelling

Eric Ritchie1, Leonardo Lalicata1, Sam Divall1, Richard Goodey1
1
Department of Civil Engineering, City, University of London, London, United Kingdom
Corresponding author: Eric Ritchie (eric.ritchie.1@city.ac.uk)
ABSTRACT: Centrifuge modelling is an established technique capable of investigating the ground’s

response to complex geotechnical events. Centrifuge models are often created from reconstituted soil, with
well-defined boundary conditions and known soil parameters. Clay soil models may be prepared by mixing
clay powder with distilled water to form a slurry. This slurry is placed within a soil container and subjected
to a vertical stress (usually in a consolidation press or consolidated inflight). This creates an isotropic model
but there is a fundamental difference between this soil model and naturally occurring soil deposits. The
structure and fabric present within a naturally occurring clay is not reproduced by this preparation process.
It is well-established that structure and fabric in naturally deposited soils are as significant in their effect on
soil behaviour as, for instance, the stress history. Inherent structure and fabric within clay soils creates
anisotropy which can vary with depth, this is particularly apparent when considering the permeability.
Creating a soil model for centrifuge modelling with representative permeability anisotropy would allow for
a better representation of consolidation driven events and the ability to observe long-term behaviour of
complex geotechnical events. Currently, there are limited methods of doing so, leading to a considerable
gap in knowledge associated with the behaviour of layered ground. This paper describes the development
of the equipment and experimental procedure for quantifying the structure developed by different sample
preparation techniques for centrifuge modelling.
Keywords: Laboratory Testing; Centrifuge Modelling; Soil Structure.
1 INTRODUCTION fundamental disparity between this and a site such

Centrifuge modelling has enabled researchers and as Whitechapel station. London like many other
engineers to investigate construction and cities is a sedimentary clay which has been
geotechnical events. Most commonly centrifuge deposited in layers creating inherent anisotropy. A
modelling has been utilised to determine homogenous clay model cannot simulate the
mechanisms and patterns of movement. different dissipation rates of excess pore pressures
Predominantly these patterns are as a result of the as seen at Whitechapel station.
deformations experienced due to a construction There have been efforts to conduct centrifuge tests
event i.e. the short-term movements. However, in a in layered ground conditions such as Grant (1998),
clay soil there can be continuing movements that Hird (2006) and Hossain and Randolph (2010), in
occur years after the construction event has finished which the interaction between different stiffness
as excess pore pressure dissipate. These long-term layers were the primary objective. However, these
movements can be significant as reported in Hill and models are not representative of a sedimentary
Staerk (2016). Long-term settlement between structure. Laboratory sedimentation is not a new
construction stages and after construction has nearly procedure and has been used previously to create
doubled the settlement induced by tunnelling. Hill laboratory samples. Stallebrass et al. (2007)) split
& Staerk (2016) also highlight the difference open a sedimented sample and a reconstituted
between two observation points at Whitechapel sample and the difference in structure is evident
Crossrail station, Vallance Road and Kempton (Figure 1). Structure can be defined as the
Court buildings where no settlement mitigation difference in interpedicular bonding and the
could be used. Due to the differing invert levels at different arrangement of particles (fabric).
the two locations the ground conditions Divall et al. (2018) created layered models within
subsequently varied. The rate of consolidation at a geotechnical centrifuge by sedimenting high water
Vallance Road decayed nearly twice as fast as that content slurries in flight. Further developing this
at Kempton Court, due to the presence of sand beds preliminary method could create a model in which
near the invert at Vallance Road. permeability anisotropies could be designed. Such
Current centrifuge models are idealised and that representative dissipation of pore pressures
conducted in homogenous blocks of clay, such that could be observed and the long-term patterns of
there are well known properties and well-defined movements can be investigated through centrifuge
boundary conditions. However, there is a modelling.
25
away any part of the clay that has been in contact
with grease or contaminants and cutting the
remainder into small pieces. These are then soaked
in distilled water and mixed to form a second slurry
with a water content of 120%. The slurry is then
consolidated as described previously.
Work undertaken by Phillips (2016) found that
when natural and reconstituted clay cuttings are
agitated in water they do not break down to the
particulate size. They will disaggregate into slurries
that contain a higher percentage of silt size
agglomerates of clay particles also referred to as
‘clay peds’. This effectively changes the particle
Figure 1. Photographs of split sedimented and reconstituted size distribution of the material.
London Clay samples Stallebrass et al. (2007). The fabric of a sample consolidated from a slurry
consisting of recycled material may be noticeably
The properties of any sedimented samples within different from a sample consolidated from clay
a centrifuge need to be quantified, the permeability
powder. As soil behaviour is partly governed by the
anisotropy needs to be measured. The compression influence of structure (Leroueil & Vaughan, 1990),
behaviour needs to be understood such that a desired the difference in structure may be significant
height of centrifuge model can be achieved. Triaxial enough to change its engineering properties
tests also need to be completed such that they can compared to a sample prepared from powder and
be used for numerical analyses. These parameters
distilled water.
measured will be compared to the standard
techniques of making centrifuge modelling to assess
2.3 Mixing under a vacuum
the difference in behaviour. This highlighted a
further question, there is no single way that Slurries prepared from powder or recycled material
centrifuge models are created, and does the method can be subjected to vacuum to de-air the mixture.
of sample preparation effect the resulting model Slurries can be mixed under a vacuum and then
properties. pumped directly into a strongbox container
This paper details the preliminary testing of the reducing air being introduced to the sample.
difference between centrifuge sample preparation Alternatively, the slurry can be mixed as previously
techniques. Oedometer tests are presented on a described but before being poured into the
variety of slurries prepared in different ways strongbox it is subjected to a vacuum (Hossain &
representing the different sample preparation Randolph, 2010). The slurry is then carefully
procedures commonly used in centrifuge modelling. placed into the strongbox container and
consolidated as described previously.
2 MODEL MAKING PROCEDURES The use of a vacuum to de-air the sample may
change the compressibility of the sample as well as
2.1 Kaolin clay powder and distilled water increase the consistency of the sample preparation
Cairncross (1973) was cited in Mair (1979) as the technique. Reducing the chance of air pockets being
earliest example of a method of creating trapped in the sample, which if in contact with a
homogenous clay models consisting of powdered structure being modelled, could be significant. This
clay and distilled water to twice its liquid limit method removes variability from researcher to
(120% water content). Once fully mixed the slurry researcher, as it does not require any manual
is carefully placed into a strongbox container to agitation.
avoid entrapping air. The clay is agitated to remove
any possible air bubbles created during pouring. The 2.4 Consolidating models inflight
slurry is then consolidated in stages to achieve a
Alternatively, from consolidating a sample within a
desired stress history. This is done through either a
press, a slurry can be consolidated under self-weight
hydraulic or a pneumatic press, in a series of loading
within a centrifuge. A slurry prepared by any of the
stages arriving at a homogenous sample.
methods above is poured into a strongbox and then
placed onto a centrifuge swing. Alternatively, the
2.2 Recycling previous tests in Kaolin clay slurry can be pumped into the strongbox inflight. In
Tests previously performed in kaolin clay models both cases, typically the slurry is placed on top of a
created as described above, can be recycled to make geofabric matt or porous plastic sheet to allow for
another centrifuge model. By cutting the remainder drainage at the base of the sample. The slurry is then
26
accelerated in stages to consolidate under self- 3 EXPERIMENTAL PROCEDURE
weight at varying g-levels. This creates a stress 3.1 Sample preparation
profile that increases with depth. To create a
uniform stress distribution or an overconsolidated 3.1.1 Preparation of slurry from powder
sample a surcharge can be applied to the surface of To create the first slurry speswhite kaolin clay was
the sample. used, the same as what is used for centrifuge models
To avoid generation of high excess pore pressures at City, University of London. The clay powder was
that creates pipping and preferential drainage paths mixed to approximately twice the liquid limit or
during consolidation, the model is accelerated 120% water content, in a 9-litre Hobart paddle
slowly to the desired g-level. Subsequently, mixer, for approximately 2 hours until the clay was
consolidation of samples can be a lengthy procedure a smooth slurry.
and the total model preparation length is
comparable to that of a hydraulic or pneumatic 3.1.2 Preparation of slurry from recycled
press. material
Work completed by Mikasa & Takada (1986) To create a slurry from recycled material, a 6-inch
recognised segregation caused when consolidating California bearing ratio mould (CBR mould) of
soft soils within the centrifuge, when using a soil kaolin clay which had been prepared as described in
consistency (water content) that did not allow for section 3.1.1, was consolidated to a stress of 500
particle segregation. Using post-test investigation kPa to create a triaxial sample. The remainder of the
techniques, it was determined that segregation clay was trimmed removing any parts in contact
occurred from a marked drop off in the effective with the CBR mould. The sections were cut into
stress void-ratio curve. The authors suggested using small cubes approximately 40-50mm3. The cubes
a water content equal or less than twice the liquid were soaked in distilled water and mixed to a water
limit of the clay to avoid segregation. Townsend et content of 120% in the same Hobart mixer.
al. (1986) also recognised the issue of segregation To create a different slurry using recycled
in centrifuge modelling and combined with other material a CBR mould was filled with clay slurry
researchers have proposed to avoid segregation by mixed from powder and loaded to a maximum stress
using low initial water content samples and staged of 200 kPa. The sample was then removed from the
consolidation procedures. press and cut into small cubes of 40-50 mm3 and
Sorta et al. (2012) investigated the effect of an mixed with distilled water to a target water content
increased acceleration field to the segregation of 120%. Both slurries had been mixed for a
characteristics of a variety of clay-sand mixtures. minimum of 5 hours, it was sometimes necessary to
Samples where left to consolidate under their own stop the mixer to break the larger lumps by hand.
self-weight at 1g for a month, then the segregation
index was calculated as defined in Donahue et al.
3.1.3 Preparation of a slurry using a vacuum
(2018). The same slurries were then consolidated
inflight for the equivalent of a month at relevant g- The samples subjected to a vacuum were prepared
levels. The segregation index was calculated and the in the same was as that from Speswhite kaolin
segregating boundaries was plotted on a ternary powder in section 3.1.1. Once the slurry was mixed
diagram to categorise different properties. it was transferred to a vacuum chamber where it was
This concluded that the segregation boundaries subjected to a vacuum and agitated for 30 minutes.
are significantly shifted at elevated g-levels (i.e. a After which the sample was removed from the
sample may not segregate under normal conditions chamber further agitated by hand and subjected to
but may during consolidation inflight). The the vacuum for a second time for a further 30
boundary is also dependent on the source of the minutes. The sample was then ready for testing in
material. If the resulting structure of the sample an oedometer.
consolidated inflight is different from that of sample
prepared in a mechanical press then the sample 3.1.4 Consolidating samples within a
behaviour may also differ. From existing literature, centrifuge
it is inconclusive if consolidating a sample on a To consolidate samples in flight for testing in an
centrifuge would yield different material properties. oedometer or a standard element test, clay is poured
To investigate the influence of different sample into a sedimentation column designed to be used
preparation techniques a series of one-dimensional within the centrifuge as shown in Figures 2 and 3.
compression tests were completed to compare These sedimentation columns will be used for a
coefficients of compression and swelling, as well as future test series to create sedimented samples
the reputability of results. similar to Stallebrass et al. (2007) but using a
27
centrifuge. In which the resulting samples by attaching a top cap. 100mm of clay slurry
permeability and mechanical properties will be prepared from a powder was poured into the tube.
measured. These sedimented samples will be The column was then accelerated to 10g and
compared to the aforementioned standard allowed to consolidate under self-weight for 24
techniques of creating centrifuge models to assess hours, after every 24 hours the g-level was doubled
any differences. to a maximum g level of 100g is achieved. The
effective radius is set to two thirds of the slurry
height within the columns for all tests as per Taylor
(1995). After allowing for consolidation at
maximum acceleration the centrifuge was
decelerated and the columns removed for further
testing.
3.2 Summary of tests completed

Table 1 summarises the slurries created, that were
tested within the oedometer. All tests were prepared
using distilled water.
Table 1. Summary of tests undertaken
Subjected
Test Kaolin Recycled
to a
Number powder material
vacuum
T1 
T2 
T3  
T4  
T5 
T6 
T7 
Figure 2. Sedimentation column arrangement
3.3 Oedometer testing
Once the samples were created they were then tested
within a standard 75mm oedometer. The loading
platern and the bottom porous stone were saturated
within a vacuum chamber. The slurry was placed
carefully in the centre of the oedometer cell then
spread to create an even layer. The clay was then
agitated manually prior to testing to remove any air
introduced when pouring the clay samples.
The samples created using the centrifuge, are
extruded into a custom made oedometer with an
oedometer ring diameter the same as the internal
diameter of the sedimentation columns. This is to
minimise disturbance caused by trimming samples.
The oedometer ring initially sits inside a recess
Figure 3. Sedimentation columns on centrifuge swing during within a PMMA holder. Once the base has been
commissioning tests.
removed from the sedimentation column it is
The sedimentation columns are manufactured secured to PMMA holder and the clay sample was
from 3mm thick, 54mm internal diameter PMMA extruded into the oedometer ring. The
tubing that has a total height of 800mm. The sedimentation column is removed and resealed for
column can be split into two sections the lower future testing. The oedometer ring fits into an
section being 180mm tall allowing for the samples adapter such that it can be used in a standard 75mm
to be extracted or converted to a permeameter cell
28
cell. A custom piston has also been manufactured is mixed for a longer period of time there was more
and applies a stress of 1.5kPa. ‘peds’ present creating a less compressible sample.
The samples where then one-dimensionally Oedometer test number 3 with recycled material
compressed and swelled within the oedometer to is very similar to test number 1. These were both the
determine the compression and swelling indexes. first samples to be prepared from the slurry directly
Samples were subjected to a maximum vertical after initial mixing. From these preliminary tests it
effective stress of 400 kPa, unloaded to a range of is inconclusive to what effect stress history had on
different stress then reloaded back to a maximum of the cuttings. There could be a more complex
stress of 400 kPa. The displacement was measured interaction between cuttings, mixing duration,
continuously with a linear variable displacement mixing speed and waiting and remixing a slurry.
transducer (LVDT) resting on the arm of the loading
frame with a reading taken at one a second.
3.6
T1 - CC= 0.672
T2 - CC= 0.700
4 RESULTS 3.4
T3
T4
- CC= 0.642
- CC= 0.639
For this paper, the results for samples consolidated T5
T6
- CC= 0.567
- CC= 0.521
inflight have been omitted due to insufficient data. 3.2 T7 - CC= 0.576
The one-dimensional compression data for the other 3

sample preparation techniques are compared in
Specific volume
Figure 4. It can be seen there is a 5% difference in 2.8
compression indexes between the samples
consolidated from a powder versus those that have 2.6
been subjected to a vacuum prior to testing. The

samples subjected to the vacuum are always less 2.4
compressible, this is thought to be due to the volume 2.2

of air being expelled from the sample during the
consolidation stages. Thus, creating a higher 2
compression index. The samples prepared using a 1 1.5 2 2.5 3
vacuum had a greater consistency with the Log σ'v (σ'v in kPa)
T1: Powder and distiled water
difference in compression index being less than 1%. T2: Powder and distilled water 2
The samples prepared from powder and mixing with T3: Powder and distilled water subjected to a vacuum
T4: Powder and distilled water subjceted to a vacuum 2
distilled water are consistent within a 5% bracket, T5: Recycled material
T6: Recycled material 2
thought to be due to the varying amount of air T7: Recycled material 3
present within the sample.

The slurries consisting of recycled material as Figure 4. One-dimensional compression curves for different
opposed to powder were around 20% less sample preparation techniques
compressible than the other sample preparation
techniques. The results of the tests are also more The average swelling index for each of the sample
varied than the others and fall within a 10% bracket. preparation techniques was 0.11. There is no
The reason for this is thought to be due to the obvious trend and all results sit within a 10%
different grading curve achieved after recycling the bracket of the average value. There is the smallest
clay cuttings (after Phillips, 2016). The silt variation in swelling index with the samples
agglomerates within the slurry are not broken-down prepared from recycled material, the reason for this
during mixing and once loaded in the oedometer it trend is not clear and requires further testing.
creates a less compressible sample.
Phillips (2016) noted that when disaggregating 5 CONCLUSIONS
kaolin clay cuttings after 30 minutes of mixing there
The compression test results show there is a
was an increase in sub 63μm particles. However,
significant difference between the sample
after a further 30 minutes of mixing there was a
preparation procedures that deserves further
reduction in sub 63μm particles.
investigation. From the preliminary tests it can be
Oedometer test 1 and 2 with recycled material
determined that;
were from the same batch of slurry. The slurry was
1. Samples consisting of kaolin powder and
sealed and then remixed prior to testing. The second
distilled water become slightly less
oedometer test conducted of recycled material is
compressible after being subjected to a
less compressible than the first, this could be
vacuum
variation within the sample or as the second sample
29
2. Slurries consisting of recycled clay cuttings Hill, N., & Staerk, A. (2016). Long-term settlement
are significantly less compressible than following SCL-tunnel excavation. Crossrail project
samples prepared from powder thought to Infrastructure design and construction- 3, pp 227-248.
be due to presence of peds.
Hird, C.C., Emmett, K.B.& Davies, G. 2006. Piling in
3. The swell indexes fall within a similar
layered ground: risks to groundwater and archaeology.
range for all preparation techniques. Science Report SC020074/SR. Environment Agency:
The difference between clay properties are Bristol.
potentially significant for centrifuge modelling if
within a series of tests there are some models Hossain, M. S. & Randolph, M. F. (2010). Deep-
prepared from powder and some prepared from penetrating spudcan foundations on layered clays
recycled material. The response of any test may centrifuge tests. Géotechnique 60, No. 3, 157-170.
differ producing a variation in results. This also
influences numerical modelling, models are often Mair, R. J. (1979), Centrifugal modelling of tunnel
compared to centrifuge test data. If the soil model construction in soft clay. PhD thesis, University of
Cambridge.
properties do not reflect the test parameters this
could affect the results for such models. Mikasa, M, & Takada, N, (1984). Self-weight
There needs to be further testing on samples consolidation of very soft clay by centrifuge. In
consolidated inflight to see where this fits within the Sedimentation Consolidation Models—Predictions
other sample preparation techniques. Further testing and Validation (Yong RN and Townsend FC (eds)).
on samples consisting of recycled material that have ASCE, New York, NY, USA, pp. 121–140.
also been subjected to a vacuum should be
investigated. It is predicted that this will further Philips, N. S. (2016). Disaggregation of soil during slurry
reduce the compressibility as seen with the samples pipe jacking. PhD thesis, City University of London,
prepared from powder. Geotechnical Engineering Research Group, School of
Mathematics, Computer Science and Engineering
6 ACKNOWLEDGEMENTS Sorta, A. R., Sego, D. C. & Wilson, W. (2012). Effect of
This research was supported by the George Daniels Thixotropy and Segregation on centrifuge modelling.
scholarship at City, University of London. International journal of Physical Modelling
Additional thanks should also go to the members of Geotechnics, 2012, 12(4), 143-161.
the research centre for Multi-Scale Geotechnical
Engineering London for their support during the Stallebrass, S. E., Atkinson, J. H. & Mašín, D. (2007).
Manufacture of samples of overconsolidated clay by
project.
laboratory sedimentation. Géotechnique 57, No. 2,
249–253.
7 REFERENCES
Cairncross, A.M. (1973). Deformations around tunnels in Taylor, R. (1995). Geotechnical centrifuge technology.
stiff clay. PhD thesis, University of Cambridge. London: Blackie Academic & Professional.
Divall, S., Stallebrass, S.E. Goodey, R.J, & Ritchie. E.P. Townsend FC, McVay MC, Bloomquist DG and
(2018). Development of layered models for Mcclimans SA (1986) Reclamation of Phosphatic Clay
geotechnical centrifuge tests. Physical Modelling in Waste Ponds by Capping. Volume 1: Centrifugal
Geotechnics – McNamara et al. (Eds) 2018 Taylor & Model Evaluation of Reclamation Schemes for
Francis Group, London, ISBN 978-1-138-34419-8. Phosphatic Waste Clay Ponds. Florida Institute of
Phosphate Research, Florida, Bartow, FL, USA, FIPR
Grant, R. J. (1998). Movements around a tunnel in two- publication no. 02-030-056.
layer ground. Doctoral thesis, Department of Civil
Engineering Geotechnical Engineering Research
Centre. The City University, London.
30
Accelerations in dynamic centrifuge modelling of seismic liquefaction problems

Andreia Marques1, Paulo Coelho1, Stuart Haigh2, Gopal Madabhushi2
Dep. of Civil Engineering, University of Coimbra, Portugal
1
Dep. of Engineering, University of Cambridge, UK

2
Corresponding author: Andreia Marques (as.silvamarques@gmail.com)
ABSTRACT: Earthquake-induced soil liquefaction continues to be a major subject of continuing research

over the last decades. In fact, despite the recent advances on the comprehension of this phenomenon, it is
repeatedly observed in recent earthquakes and have devastating effects in modern societies increasingly
dependent on liquefaction-susceptible infrastructures. Earthquake-induced liquefaction effects are
particularly important in shallow foundations built in seismically active areas on saturated deposits of
granular soils, being quite complex to model, as significant loss of the soil’s strength and stiffness due to
excess pore pressure build-up, complex soil-structure interaction (SSI) mechanisms and very large
deformations are typically observed. Dynamic centrifuge modelling is accepted as a powerful research tool
to experimentally shed light on these problems, as it can effectively mimic the soil’s effective stress during
and after the seismic motion and realistically capture SSI effects. This paper discuss the ability of this
experimental research tool to accurately assess the mechanisms governing the behaviour of shallow
foundations built on liquefiable ground, focussing mostly on the mechanisms governing the propagation of
horizontal and vertical accelerations in the ground and foundations observed in centrifuge test performed
on a model built inside an ESB container and representing neighbouring shallow foundations built on ground
undergoing liquefaction due to a seismic simulation generated by the Stored Angular Momentum (SAM)
actuator. Several instruments were place through the model to understand in detail the mechanisms,
including acceleration MEMS placed on the footings and accelerometers placed in the ground and model
boundaries. It was observed that boundary effects may affect the performance of the model and that footings
exhibit non-planned vertical accelerations that may not result from individual rocking mechanisms but rather
from some rocking of the all centrifuge model. Overall, the unique value of centrifuge modelling in
liquefaction research is demonstrated, especially if the testing conditions are fully understood.
Keywords: Liquefaction; Centrifuge Modelling; Shallow Foundations; ESB Container; SAM Actuator.
1 INTRODUCTION more recent 2010-11 New Zealand earthquakes

Earthquake induced soil liquefaction is a complex (Cubrinovski et al. 2011) or the 2018 Indonesia
problem that has been continuously affecting Sulawesi earthquake (Sassa and Takagawa 2019).
modern society. Even if it is seldom responsible for Figure 1 shows some recent examples of
a significant number of human fatalities, this liquefaction induced effects on buildings resting on
catastrophic phenomenon seriously affects human shallow foundations, highlighting the need for
life as it can cause severe damage to buildings, further research into the complex behaviour of
bridges and other structures and infrastructures, shallow foundations built on liquefiable soils,
causing significant and prolonged economic losses including studying efficient techniques to ensure
and social disruption. foundation safety through ground improvement.
This phenomenon is commonly observed in loose This is only possible through advanced physical
saturated deposits of sandy soil, under structures modelling or properly-calibrated numerical
resting on shallow foundations. In fact, settlement modelling, which usually requires the use of
and tilting of shallow foundations on saturated dynamic centrifuge modelling.
sandy/silty soils has been repeatedly observed
throughout the world as a consequence of
liquefaction or softening of the foundation soil – for
instance in Mexico City during Michoacan
earthquake of 1985 (Mendoza and Avunit 1988), in
the city of Adapazari due to 1999 Kocaeli
earthquake (Karaca 2001; Bakir et al. 2002; Yılmaz
et al. 2004), during the 2010 Maule (Chile)
Earthquake (Bray and Frost 2010), in 2011 Tohoku Figure 1. Seismic-liquefaction damage on buildings resting
(Japan) Earthquake (Ashford et al. 2011), and in the on shallow foundations. (adapted from Cubrinovski, 2013)
31
The effects of seismic liquefaction are indeed a 2 CHARACTERISTICS OF THE
very complex phenomenon to model, as significant DYNAMIC CENTRIFUGE MODEL
loss of the soil’s strength and stiffness due to excess To highlight the behaviour of neighbouring shallow
pore pressure build-up, complex soil-structure foundations under liquefaction effects, a dynamic
interaction (SSI) mechanisms and very large centrifuge test was performed at Cambridge
deformations are typically observed. In the physical University’s Schofield Centre. The centrifuge
modelling of engineering problems, scaled models model represented an 18 m deep liquefiable deposit
of field prototypes are normally used to simulate with two nearby shallow foundations placed on
field conditions. The reliability of such scale models loose deposits of fully saturated sandy soil, as
in geotechnical engineering may be undermined by schematically shown in Figure 2.
the reduced stress level they might be subjected to, The test was conducted using the 10-m diameter
resulting in a noticeably different response with Turner Beam Centrifuge (Schofield, 1980). The
respect to prototype geotechnical models. This centrifuge model was prepared inside an ESB
particular limitation of 1g testing of scaled models container (Zeng and Schofield, 1992), in order to
may be overcome if the stresses and strains at minimise undesirable boundary effects. The
homologous points in the prototype and model actuator used in the centrifuge test to generate
structures could be identical. This can be achieved seismic simulations was the SAM actuator
by spinning the model built at 1:N scale on a (Madabhushi et al., 1998), which is a simple and
geotechnical centrifuge to increase the gravitational reliable mechanical actuator that uses the energy
field by a factor of N, so that the prototype stresses stored in a pair of flywheels to generate the input
are matched (Muir Wood, 2004). motion. Despite not being able to reproduce real
Centrifuge modelling is based on the principle of seismic actions, it generates nearly sinusoidal
creating scaled-down models of geotechnical horizontal acceleration motions of chosen duration
structures in order to depict the behaviour of a and amplitude, which is considered valuable for
prototype model, which corresponds to a non-scale fundamental research on earthquake effects, as it
simplified representation of the real problem. The avoids the difficulties introduced by more complex
behaviour of soils has been established to be highly dynamic loading.
non-linear and hence true prototype behaviour can Hostun sand, whose properties are summarized in
only be observed in a model under stress and strain Table 1 and described in detail by Flavigny et al.
conditions similar to the prototype. A geotechnical (1990), was selected to create the model. It is a fine-
centrifuge enables us to recreate the same stress and grained, clean and uniform silica sand typically very
strain level within the scaled model by testing a 1:N susceptible to liquefaction as it lies well within the
scale model at N times earth's gravity, created by boundaries for most liquefiable soil (Figure 3).
centrifugal force. The similarity of stress, strain and
material behaviour ensures the realism of model Table 1. Properties of Hostun sand (Stringer, 2008).
behaviour as long as the boundary conditions are
Angle of D10 D50
correctly implemented. However, despite centrifuge C =D /D e
repose (o) (mm) (mm) u 60 10 max
emin Gs
modelling being a very attractive tool to model such
complex problems, it still may carry some 33 0.286 0.424 1.59 1.067 0.555 2.65
unexpected phenomenon during its performance.
Figure 2. Model layout with the type, position and identification of the instruments used during the test (prototype scale).
32
system are described fully by Stringer and
Madabhushi (2009), subsequent improvements to
the system being discussed by Stringer and
Madabhushi (2010). With this system the saturation
procedure is automated, not requiring a constant
human presence through the entire process.
3 RESULTS AND DISCUSSION

The data presented in this work focuses mostly on
the discussion of the acceleration motions measured
on the footings through the MEMS installed in both
structures.
Figure 3. Particle size distribution curve for Hostun sand
superimposed on liquefaction susceptibility curves proposed by 3.1. Input motions
Tsuchida (1970) – adapted from Stringer (2008).
The centrifuge model was subjected to an input
Considering the aim of the test, structural models seismic motion at the base, applied parallel to the
had a simple design, consisting of solid steel blocks long side of the model and designed to replicate a
with prototype dimensions (length x width x height) relatively strong real earthquake, planned to last
of 3×3×0.75 m3, in the case of the lighter structure about 25 seconds, have a predominant frequency of
(L), and 3×3×1.225 m3, in the case of the heavier 1 Hz and impose maximum peak horizontal
structure (H). The pressure transmitted through the accelerations close to 0.3-g. The long earthquake
foundations basis equals 58 and 95 kPa, for the duration planned aimed at intensifying liquefaction
heavier and lighter structures, respectively. effects and facilitating model behaviour analysis.
A series of pore-pressure transducers (PPTs), The actual time history of the horizontal seismic
accelerometers (accs) and linear variable loading applied to the base of the model container is
displacement transducers (LVDTs) were installed in depicted in Figure 4a, showing that the planned
the model to assess the soil behaviour during the earthquake simulation was successfully imposed to
centrifuge test (Figure 2). A series of the model. The seismic simulation applied is not
microelectromechanical system accelerometers purely single-frequency, although the predominant
(MEMS) were carefully attached to the footing to frequency matches the desired value of 1Hz.
measure the vertical and horizontal accelerations
(a)
induced in the footings during the earthquake.
Several accs were attached to the bottom and lateral
walls of the container to measure the vertical and
horizontal input motions and to evaluate the
horizontal motion propagation through the ESB
walls.
Full saturation of the model ground is an essential
part of the model preparation in centrifuge-based
liquefaction research, as liquefaction is strongly
affected by the degree of saturation. The saturation
process requires time and strict control to avoid
(b)
model disturbance and thus yield superior results. A
viscous solution of hydroxypropyl methylcellulose
in water was used as the pore fluid (Madabhushi,
1994), in order to achieve the so-called viscosity
scaling (Lambe and Whitman 1982) and overcome
the conflict between time scaling in flow and
dynamic phenomena that occur simultaneously
during liquefaction. The fluid used had a viscosity
50 times that of water to match the centrifuge
acceleration used in the centrifuge test ( 50-g). This
saturation procedure was processed with a novel
automatic saturation system (Cam-Sat), available at Figure 4. Time history of the seismic motion applied to the
Cambridge. The design and application of this model: (a) horizontal accelerations, (b) vertical accelerations.
33
An accelerometer was also placed at the base of of the horizontal accelerations is considerably
the model, in the vertical direction, to assess a different in the soil and in the lateral boundary
possible unplanned vertical motion imposed to the formed by the ESB vertical walls. This is certainly
model (Figure 4b). The measurements confirm that due to the fact that the container walls, designed to
the total input motion applied to the base of the match the stiffness of the sand used in the design,
model also includes a vertical component. Still, this maintain a similar response during the earthquake.
component is significantly smaller than the In contrast, as the sand liquefies, the stiffness
horizontal input motion and should not affect degradation caused by the large effective stress
meaningfully the overall model behaviour. reduction strongly affects the propagation of the
accelerations through the deposit. In fact, it is clear
3.2. Excess-pore-pressure generation and that large attenuation of the horizontal accelerations
settlements occur as the waves travel upward within the
The measurements of the excess-pore-pressure in liquefied soil. Therefore, the different response of
the ground at different locations and the settlements the soil and the container walls, which is a logical
of the footings during and after the earthquake result of the differences in the stiffness of the
undoubtedly prove that the ground liquefied due to materials once the sand liquefies, may induce
the seismic loading and that the settlements of the boundary effects that should be avoided by placing
footings are compatible with the performance of the model footings sufficiently away from the
real structures in the field when affected by container walls. It should be emphasised, however,
liquefaction effects. More detail on the data that soil softening due to excess pore pressure
obtained is shown in Marques et al (2012). generation is believed to strongly mitigate boundary
effects by reducing the soil's ability to transmit
reflected seismic waves from the model box
3.3. Accelerations in the footings boundaries toward the footing models. Therefore,
The values of the horizontal accelerations measured no significant amplification of the acceleration
in each footing and in the soil under each footing records should be expected once the soil starts to
axis at different depths are depicted in Figure 5. The liquefy, namely at a reasonable distance from the
first major conclusion is that the transmission to the walls. In this test, as suggested by Figure 5, no
soil of the horizontal shaking applied to the base of apparent increase is visible in the soil response
the model is quite successful, as the horizontal under the footings, where the boundary effects
accelerations recorded at the bottom of the soil would be more relevant. Consequently, the footings
model are very similar to those measured at the base seem to be placed as far from the walls as necessary
of the container. Contrarily, the upward propagation to prevent unrealistic accelerations.
CT-A/H CT-A/L ESB

0.4 0.4 0.4
Footing (g)
acc hat the
0.2 0.2 0.2

0.0 0.0 0.0
-0.2 -0.2 -0.2
-0.4 -0.4 -0.4
5 15 25 35 5 15 25 35 5 15 25 35
0.4 0.4 0.4
acc h near the
0.2 0.2 0.2

surface (g)
0.0 0.0 0.0

-0.2 -0.2 -0.2
-0.4 -0.4 -0.4
5 15 25 35 5 15 25 35 5 15 25 35
0.4 0.4 0.4

Inputh
acc h near the
0.2 0.2 0.2

base (g)
0.0 0.0 0.0

-0.2 -0.2 -0.2
-0.4 -0.4 -0.4
5 15 25 35 5 15 25 35 5 15 25 35
time (s) time (s) time (s)
Figure 5. Analyses of the time histories of horizontal accelerations at the footings and at different positions between the
centre of each footing and the walls of the ESB container in model.
34
0.4
0.3
0.2
0.1
acch (g)
0.0
Heavier structure
-0.1
-0.2 Lighter structure
-0.3 1m under the

Heavier structure
-0.4
7 12 17 22 271m under the Lighter
structure
time (s)
Figure 6. Horizontal motions measured in both footings and in the soil underneath.
0.4 0.4
accv left (g)
accv right (g)

0.2 0.2
0.0 0.0
-0.2 -0.2
-0.4 -0.4
9 11 13 15 17 19 9 14 19
time (s) time (s)
Figure 7. Comparison of the vertical motions measured at each structure during a short duration of time.
0.4 0.4
accv at Heavier structure (g)
accv at Lighter structure (g)
0.2 0.2
0.0 0.0
-0.2 Heavier-left -0.2
Heavier-right
-0.4 -0.4
Vertical Input 9 11 13 15 17 19
9 11 13 15 17 19
time (s) time (s)
Figure 8. Comparison of the vertical motions measured at the left and right sides of both structures during a short period.
time (s)
31.1 31.6 32.1 32.6 33.1
-0.5
settlements (m)
-0.6
-0.7
-0.8
-0.9
Figure 9. Oscillation in the LVDT readings during a short period due to possible shaking induced by model rocking.
35
For a more detailed and perceptible analyses of global rocking of the model may result from an
the horizontal accelerations measured in the test, a unplanned rocking motion introduced in the model
zoom of the data measured in each footing and near during the simulation of the earthquake or from a
the surface are presented in Figure 6 corresponding response of the soil model caused by the low-
to a relatively short period of time. Figure 6 clearly friction interface between the soil and the container.
shows that the horizontal motions measured in each Nevertheless, the overall performance of the results
footing are not in phase with each other. measured should not be significantly affected by
Furthermore, the horizontal motions that reach the this unexpected behaviour, especially if it is
heavier structure (H) are in phase with the ground properly understood.
motion propagation, while the motions that reach
the lighter footing (L) are not in phase with the rest 4 CONCLUSIONS
of the horizontal data (soil and heavier structure). The experimental data obtained in a dynamic
This observation is certainly unexpected, centrifuge model representing two neighbouring
suggesting that the lighter structure might be shallow foundations built on liquefiable ground
experiencing some individual mechanism show the unique value of this powerful
governing its motion that diverges from the motion experimental research tool in assessing the
of the ground and heavier structure. mechanisms governing the performance of the
On the other hand, soil softening due to seismic structures during and after a seismic event. In fact,
liquefaction may facilitate rocking of the soil model the stiffness degradation of the soil induced by large
in the vertical plane under horizontal base shaking. excess-pore-pressure degradation and its effects on
Although this is not visible in the soil data, as the motion upward propagation is clearly captured
vertical accelerometers were not installed by the model, as are the excess-pore pressure
throughout the soil deposit, experimental results on generation and the footing and ground vertical
the footing’s vertical motions emphasizes the settlements.
occurrence of significant rocking motion. For a However, as the stiffness of the ESB container
clearer observation of this phenomenon, Figure 7 remains the same during the earthquake, significant
shows the comparison of the left and right vertical unrealistic boundary effects may occur in the
motions measured at both sides in each footing, ground near the physical limits of the model ground.
while Figure 8 compares the vertical motions In order to avoid these possible boundary effects
measured in both sides of the same footing and and the negative consequences on the behaviour
compares these with the measured vertical input. exhibited by the footings, it is important to place the
Analysing each of these figures, it is clear that the structures sufficiently distant from the container
vertical accelerations measured in the footings are walls, namely when using a ESB container. It
not a mere consequence of the vertical input, should be noticed that using other containers, such
suggesting also that a rocking mechanism causing as laminar containers, may reduce this problem but
the vertical accelerations in the footings does not create other challenges in centrifuge modelling.
seem to occur in each footing individually, as it is Last but not the least, the analysis of the rocking
visible in Figure 8. Instead, the rocking mechanism mechanisms observed in the models should take
seems a result of a global rocking mechanism into consideration that unexpected components of
affecting all the model (Figure 7). In fact, both sides acceleration, which are often not even measured,
of each footing show vertical motion that are in may be applied to the models. Furthermore,
phase, the footing’s motions being completely out additional vertical accelerations in the footings may
of phase with each other. This disturbs the response result from complex rocking mechanisms that may
of the footing models, being also visible when not be totally realistic, as they result from global
comparing the vertical oscillations obtained in both rocking of the model instead of a realistic response
shallow foundations during typical cycles of of the footing to the dynamic loading. This
shaking, through LVDT´s placed on top of the unforeseen response, only observed because of the
footings (Figure 9). Footing oscillation may be unusual number and position of the accelerometers
assumed to be equal to the ground surface used in the test, may result from an unplanned
oscillation at that location. At the time represented rocking motion introduced in the model during the
by the analysed cycle of base shaking (~31.1-33.1s), simulation of the earthquake or from a response of
the soil has already been fully softened by the the soil model caused by the low-friction interface
generation of epp, resulting in a fluid-like ground between the soil and the container. In any case, the
surface oscillation. These results highlight the fact consequences of this problem should not affect the
that both footings have similar levels of vertical overall value of the data obtained.
oscillations but are not in phase. This unexpected
36
Considering the validated ability of centrifuge Flavigny, E., Desrues, J. and Palayer, B. 1990. Le sable
modelling to model problems as complex as the one d’Hostun «RF».Revue Française de Géotechnique,
considered and the large limitations of potential 53, 67–69.
alternatives, progress in the understanding of the Karaca, G. 2001. An Investigation into Large Vertical
behaviour of shallow foundations built in Dis-placement Experienced by the Structures in
liquefiable ground and improvements in the Adapazari during 17 August 1999 Earthquake. MS
development of more efficient liquefaction thesis, Middle East Technical University, Ankara,
mitigation measures strongly depends on the use of Turkey.
centrifuge modelling as a research tool, especially if
some limitations are appropriately dealt with in the Lambe, P.C. and Whitman, R.V. 1982. Scaling for
earthquake shaking tests on a centrifuge. Proceedings
design of the model and analysis of the data. International Conference on Soil Dynamics and
Earthquake Engineering, Vol. 1, Southampton,
5 ACKNOWLEDGEMENTS England, July 1982: 367-378.
This research was funded by the European
Madabhushi, S.P.G. 1994a. Scaling laws in Dynamic
Community’s 7th Framework Programme Centrifuge Modelling. Proc. VIIIth Int. Conf. of the
[FP7/2007-2013] for access to the Turner Beam Association for Computer Methods and Advances in
Centrifuge, Cambridge, UK, under Grant Geomechanics, West Virginia University,
Agreement nº 227887. The first author would also Morgantown, USA.
like to thank FCT – Fundação para a Ciência e a
Tecnologia, Portugal – for sponsoring her PhD Madabhushi, S.P.G., Schofield, A.N. and Lesley, S.
1998. A new Stored Angular Momentum (SAM)
programme (grant no. SFRH/BD/73170/2010). The
based Earthquake Actuator. Proc. Centrifuge’98, Int.
kind and skilled support of the technicians and Conf. on Centrifuge Modelling, Tokyo, Japan:111-
researchers involved in the experimental work 116.
carried out in the Schofield Centre is also
acknowledged. Marques, A.S., Coelho, P.A.L.F., Cilingir, U., Haigh,
S.K. & Madabhushi, S.P.G. 2012. Centrifuge
modeling of liquefaction-induced effects on shallow
6 REFERENCES foundations with different bearing pressures. 2nd
Ashford, A, S., Boulanger, R., and Brandenberg, S. 2011. Eurofuge conference on Physical Modelling in
Recommended design practice for pile foundations in Geotechnics. Delf.
laterally spreading ground. Report PEER 2011/04,
Pacific Earthquake Engineering Research Center Mendoza, M.J. and Avunit,G. 1988. The Mexico
(PEER), Berkeley, CA. Earthquake of September 19, 1985-Behavior of
Building Foundations in Mexico City”, Earthquake
Bakir, B.S., Sucuoglu, H. and Yilmaz, T. 2002. An Spectra, 4(4): 835-853.
Overview of Local Site Effects and the Associated
Building Damage during 17 August 1999 Izmit Muir Wood, D. 2004. Geotechnical modelling, Volume
Earthquake. Bulletin of seismological Society of 1. Taylor & Francis.
America, 92(1): 509-526.
Sassa, S., Takagawa, T. 2019. Liquefied gravity flow-
Bray, J. D. and Frost, D. 2010. Geo-engineering induced tsunami: first evidence and comparison from
Reconnaissance of the 2010 Maule, Chile the 2018 Indonesia Sulawesi earthquake and tsunami
Earthquake. Geotechnical Extreme Events disasters. Landslides 16, 195–200.
Reconnaissance (GEER) Association report No. https://doi.org/10.1007/s10346-018-1114-x
GEER-022, Chapter 7.
Schofield, A.N.1980. Cambridge geotechnical centrifuge
Cubrinovski, M., Bradley, B., Wotherspoon, L., Green, operations, 20th Rankine Lecture. Geotechnique,
R.A., Bray, J.D., Wood, C., Pender, M., Allen, J., 30(3):227-268.
Bradshaw, A., Rix, G., Taylor, M., Robinson, K.,
Henderson, D., Giorgini, S., Ma, K., Winkely, A., Stringer, M.E. 2008. Pile Shaft Friction in Liquefaction.
Zupan, J., O'Rourke, T., DePascale, G., Wells, D. First Year Report, University of Cambridge, UK.
2011. Geotechnical Aspects of the 22 February 2011
Christchurch Earthquake, Bulletin of the New Stringer, M.E. and Madabhushi, S.P.G. 2009. Novel
Zealand Society for Earthquake Engineering, 44(4): Computer Controlled Saturation of Dynamic
205-226. Centrifuge Models using High Viscosity Fluids.
ASTM Geotechnical Testing Journal, 32 (6), 53-59.
Cubrinovski, M. 2013. Liquefaction-Induced Damage in
the 2010-2011 Christchurch (New Zealand) Stringer, M.E. and Madabhushi S.P.G. 2010. Improving
Earthquakes. 7th Int. Conf. on Case Histories in Model Quality in Dynamic Centrifuge Modelling
Geotechnical Engineering, Chicago, USA. through Computer Controlled Saturation. In: Proc.
International Conference on Physical Modelling in
37
Geotechnics S. M. Springman, J. Laue, and L. Zeng, X., and Schofield, A. N. 1992. Design and
Seawards, eds., Zurich, Switzerland: Balkema. performance of an Equivalent Shear Beam (ESB)
model container for earthquake centrifuge modelling.
Yılmaz, M. T., Pekcan, O., Bakır B. S. 2004. Undrained Geotechnique, 46(1), 83-102.
cyclic shear and deformation behavior of silt-clay
mixtures of Adapazarı, Turkey. Soil Dynamics and
Earthquake Engineering, 24(7), 497-507.
38
Boundary effects in centrifuge modelling of soil liquefaction

Zheng Li1, Sandra Escoffier1
1
Université Gustave Eiffel, Département Géotechnique, Environnement, Risques naturels et Sciences de la terre
(GERS), Geomaterials & Models in Geotechnics Laboratory (GMG), France
Corresponding author: Sandra Escoffier (sandra.escoffier@ifsttar.fr)
ABSTRACT: Centrifuge modelling is commonly regarded as a valuable tool for seismic analysis of
geotechnical problems. However, the correct recovery the loading condition under seismic loadings greatly
depends on the boundary of the container, whereas the soil was placed. This issue is of great importance
especially for soil liquefaction. The issue of significant soil stiffness and shear strength degradation with
increasing strain, which should ideally be reproduced by the physical boundaries of the centrifuge model,
becomes very important. Recently at new laminar container was designed at Université Gustave Eiffel
centrifuge laboratory, which advances the correctly modelling of boundary condition under
seismic/dynamic loadings. The newly designed laminar box composes metal frames and rollers trying to
have almost zero lateral stiffness, which minimizes greatly the boundary effects. This paper presents a series
of centrifuge tests on soil liquefaction problem using this newly designed laminar container. Loose and
dense Hostun HN31 sand with relative density of 50% and 80% were used. The influence of boundary
effects induced by laminar container on the liquefaction of sand with different relative densities was
highlighted and quantified.
Keywords: Liquefaction; Centrifuge Modelling; Boundary Effects; Laminar.
1 INTRODUCTION displacement profiles and phase-lag between the

As a very effective approach, centrifuge modelling laminar container and soil column.
has played very important role improving the
understanding not only phenomena’s, but also the 2 EXPERIMENTAL SET-UP
mechanism of many geotechnical problems. For the In this section, the experimental set-up of the
simulation of seismic effects, dynamic centrifuge centrifuge test will be introduced. In this study, the
modelling has been used for nearly two decades to test was carried out under the gravity level 60g. All
investigate the seismic behaviour of various soil- the units will be presented in prototype scale
structure systems. The artificial boundary of a otherwise mentioned.
model container may cause distortion in the stress
and strain field in the model compared with those in 2.1. Centrifuge models
the prototype modelled (Teymur & Madabhushi, 2.1.1 Laminar container
2003).
In order to enhance the value of the experimental Recently, a new laminar container was developed in
results resulting from centrifuge tests, proper the centrifuge team of Université Gustave Eiffel in
simulation of boundary conditions is required. This Nantes. Figure 1 shows the 3D model of the laminar
issue remains one of the most controversial aspects container. The inner dimensions of the container are
of centrifuge modelling of soil’s dynamic 719*379 mm. Each ring in the laminar container
behaviour, especially if the seismic event leads to has a height of 10mm with weight of 1.025 kg. The
liquefaction, since the dramatic reduction of soil number of rings can be adjusted according to the
stiffness and shear strength caused by liquefaction height of the soil model. In this study, 22 rings were
should ideally be reproduced by the model in the laminar container. The aim of the laminar
boundaries (Coelho, Haigh, & Madabhushi, 2003). container is to have a ‘zero’ lateral stiffness
In this paper, the boundary effects induced by a container. Rollers were installed between rings.
newly designed laminar container at Université Therefore, the rings can move freely in the
Gustave Eiffel are evaluated. A series of 1D horizontal direction. A very thin rubber film with 1
centrifuge shaking table tests on saturated sand mm was installed inside the container to minimize
model were conducted to simulate level deposits as much as possible constrains of rubber film. Two
subjected to base shaking with different magnitude buffer stops were placed at the side of the container
and frequencies. The boundary effects are to limit the extreme unwanted lateral displacement.
evaluated mainly in terms of maximum A rigid frame is also installed on the side of the
container to ensure the 1D shear deformation.
39
in dynamic test can be evaluated. Pore pressure
sensors were also used to monitor the built-up of
excessive pore pressures.
Figure 2. Sensor plan of the centrifuge model (model scale)
Figure 1. Laminar container recently developed in Université A large number (totally 50 shaking events) of
Gustave Eiffel centrifuge laboratory simple sine base input signals with different
frequencies and amplitudes were used in the test.
2.1.2 Material properties The input sequence of the shaking events is shown
Hostun HN31 sand was used in the soil model, the in Figure 3. where the input frequency and
properties of the sand are summarized in Table. 1. amplitude are also noted.
Table 1. Characteristics of Hostun HN31 sand.

D50 ρs
Sand Cu emin emax
(mm) (g/cm3)
HN31 0.35 1.57 0.656 1 1.60
2.1.3 Model preparation

Instead of the normally used air pluviation
preparation technique, moisture tamping method
was used to prepare the sand sample. In order to
achieve a homogenous sand model, under-
compaction method (Ladd, 1978) was adopted. This
method was originally used for preparation small Figure 3. Shaking events (Acceleration and frequency are in
scale triaxial laboratory tests which considers the prototype scale)
influence of the tamping of upper layers to the under
layers. The initial water content of the moist sand 3 RESULTS AND DISCUSSION
is 3.3%, loose sand model with relative density of As it was explained in section 2.1.4, large number
50% was then prepared following the under- of base shakings were input to the model. As a
compaction method. result, the initially loose sand became denser and
The method proposed by Ueno (1998) was denser. The initial relative density is around 50%
utilized to saturated the sand model at 1g. HPMC and nearly 2.0 cm settlement was observed on the
viscous fluid was prepared and the viscosity was surface of the sand. The final relative density is
around 60 cSt. The degree of saturation was estimated around 98%. The data from pore pressure
determined by the method proposed by Okamura sensor confirms this. After the 30th shaking event,
and Inoue (2012). The final degree of saturation was instead of the built-up of pore pressure, a suction
estimated around 92.0%. behaviour of water was observed in the sand, which
corresponds to the dilatancy of sand under shear
2.1.4 Sensor plan and experimental program loading. Therefore, the boundary effects of the
Figure 2. shows the sensor plan in the dynamic laminar container should be evaluated according to
centrifuge tests. Three arrays of accelerometers different state of the sand model, i.e., loose, medium
marker with red triangles were instrumented. In dense and dense states.
each sensor array, there are 5 accelerometers. The The boundary effects of the laminar container are
depths of these 5 accelerometers from bottom to top first analysed for the sand model in loose state. Take
in prototype are 13.5, 10.5, 7.5, 4.5 and 1.5m. The the 6th shaking event (0.1 g) as an example, the
base shaking was input at the bottom of the soil maximum displacement profiles of central, middle
column at the depth 15.0 m. By comparing the and side array are shown in Figure 4. The profiles
lateral responses of the central, middle and side of central and middle array are almost identical.
array, the boundary effect of the laminar container Different behaviour is observed for the side array.
40
Figure 4. Maximum displacement profiles of central, middle
and side array, 6th shaking event (0.1 g input)
By analysing the time history of displacements

near the soil surface (depth 1.5m, prototype scale,
Figure 7. Maximum displacement profiles of central, middle
CH09, CH14 and CH21), the displacement and side array, 26th shaking event (0.1 g input)
amplitudes are different. In addition, the
displacement on the container is lag behind to the At the shallow depth 1.5 m under the soil surface,
displacement in the soil. the displacement time-history is shown in Figure 8.
The displacements of central and middle in the soil
are identical. However, great phase-lag is observed
between the soil and container.
Figure 5. Displacement time-history of CH09, CH14 and

CH21, depth 1.5m, prototype scale. 6th shaking event (0.1 g
input)

However, with the increase of depth, the phase-
lag between the side wall and soil column decreases. input)
Only very small phase-lag can be found. The
differences mainly come from the displacement Even in deeper depth, phase-lag can be observed.
amplitudes, see Figure 6. For example, at depth 7.5m, the side wall is always
lagged behind the soil.

CH19, depth 7.5m, prototype scale. 6th shaking event (0.1 g Figure 9. Displacement time-history of CH06, CH12 and
input) CH19, depth 7.5m, prototype scale. 26th shaking event (0.1 g
input)
For the sand model in medium dense state
(shaking event 26), the maximum displacement Comparing with the sand model in medium dense
profiles of central, middle and side array are shown state, similar boundary effects of model with very
in Figure 7. Similar as the sand in loose sate, the dense sand are observed. The behaviour of the
profiles of central and middle array are almost central and middle array is identical; however, great
identical. Different behaviour is observed for the phase-lag was observed between the side wall and
side array. the soil. The results are shown in Figures 10, 11 and
12.
41
4 CONCLUSIONS
In this paper, the boundary effects of a new laminar
container designed and built in centrifuge team in
Université Gustave Eiffel were evaluated. In
general, the behaviour of the soil columns in the
center and in quarter of the container are almost
identical. However, great phase-lag was observed of
the side-wall of the laminar container, especially for
medium and very dense sand model. At the same
depth, the ring of the laminar container is lagged
behind the motion of the soil column.
5 ACKNOWLEDGEMENTS
The authors would like also to thank the valuable
support and help of the technical staff of the
Figure 10. Maximum displacement profiles of central, middle Université Gustave Eiffel centrifuge team.
and side array, 42th shaking event (0.1 g input)
6 REFERENCES
Coelho, P. A. L. F., Haigh, S. K., & Madabhushi, S. P.
G. 2003. Boundary effects in dynamic centrifuge
modelling of liquefaction in sand deposits. 16th
ASCE Engineering Mechanics Conference, 1–12.
Seattle.
Figure 11. Displacement time-history of CH09, CH14 and Ladd, R. S. 1978. Preparing test specimens using
CH21, depth 1.5m, prototype scale. 42th shaking event (0.1 g undercompaction. Geotechnical Testing Journal,
input) 1(1), 16–23.
Okamura, M., & Inoue, T. 2012. Preparation of fully

saturated models for liquefaction study. International
Journal of Physical Modelling in Geotechnics, 12(1),
39–46. https://doi.org/10.1680/ijpmg.2012.12.1.39
Teymur, B., & Madabhushi, S. P. G. 2003. Experimental

study of boundary effects in dynamic centrifuge
modelling. Géotechnique, 53(7), 655–663.
input)
Ueno, K. 1998. Methods of preparation of sand samples.
Proceedings of the International Conference
Centrifuge 98, 1047–1056. Tokyo.
42
Centrifuge model studies on dry granular soils subjected to dynamic compaction

induced tamping
Saptarshi Kundu1, B.V.S. Viswanadham2
Research Scholar, Department of Civil Engineering, Indian Institute of Technology Bombay, INDIA
1
Professor, Department of Civil Engineering, Indian Institute of Technology Bombay, INDIA

2
Corresponding author: B.V.S. Viswanadham (viswam@civil.iitb.ac.in)
ABSTRACT: The objective of this paper is to present centrifuge model studies on dry granular soils
subjected to dynamic compaction (DC) induced tamping. A custom developed and designed inflight DC
actuator was used. The developed actuator is equipped with a motorized lifting and dropping mechanism to
induce repeated drops on the soil surface at high gravities, with prototype energies varying from 50 ton-
meter to 400 ton-meter. The actuator can thus simulate both low-energy blows and high-energy DC
processes adopted in the field. The present paper discusses the results of two centrifuge model tests
performed on dry loose granular soil deposit using the 4.5 m radius large-beam centrifuge facility available
at IIT Bombay, India. The analysis of data captured by accelerometers, coupled with GeoPIV based Image
Analysis was employed to demonstrate the effectiveness of DC in granular soils using the developed
actuator for various tamper energies. The results are interpreted in terms of crater depth, induced ground
velocities and peak accelerations and increase in relative density of the soil post DC.
Keywords: Centrifuge; Dynamic Compaction; Scaling Laws; Ground Improvement.
1 INTRODUCTION improvement depth and degree of compaction can

In recent times, rapid industrialization and be gathered from the above due to soil heterogeneity
infrastructure developments have created acute and problems related to instrumentation and data
scarcity of suitable construction sites. This has led acquisition. Further, the results of low-energy DC
to renewed interest in finding means of improving tests by Feng et al. (2000), Arslan et al. (2007),
the geotechnical characteristics of sites with Bonab and Rezaei (2009), Bonab and Zare (2014),
deficient subsoil types (like loose soil deposits, Kundu and Viswanadham (2016) and Kundu and
collapsible soils, reclaimed fills, coal ash deposits, Viswanadham (2018) lack field validation owing to
waste dump sites, etc.). Among the available ground dissimilar stress conditions in the model and full-
improvement techniques, Dynamic compaction scale prototype. Limited centrifuge model
(DC) has gained popularity in view of its simplicity, simulations on dry soil conducted by Oshima and
low-environmental impact and cost-effectiveness. Takada (1998) and Merrifield and Davies (2000)
The method transmits kinetic energy onto the also exhibited major restrictions in modelling the
soil/material by repeatedly dropping heavy tamper time interval between successive tamper drops.
from a significant height at regularly spaced Hence, there arises the necessity of developing a
intervals (square pattern with print spacing of the robust actuator for studying DC induced ground
order of 4 m - 6 m). This eventually leads to improvement and factors influencing its efficiency.
optimization of foundations for structures, thereby The present study aims at investigating the
economizing the project. The current practice for response of dry, loose sand deposit subjected to DC.
evaluating the feasibility of DC and choosing the Centrifuge-based physical modelling technique was
operational parameters at the site are not well adopted in this regard to replicate identical stress-
established. The decisions have to be made strain response in model and prototype soil deposits.
depending mainly on field trials, past experience, Tests were conducted at high gravities on a typical
and empirical equations derived from field case dry, loose sand using a custom-developed in-flight
studies. The above factors necessitate a detailed actuator. Design details of the actuator are discussed
investigation of DC process for establishing design in subsequent sections. The influence of tamping
parameters governing its efficiency. energy during DC was also investigated. Data
Field testing and related case studies on the interpretation was done through data captured by
effectiveness of DC were documented by various accelerometers, coupled with GeoPIV based Image
researchers including Mayne et al. (1984), Zou et al. Analysis on photographs captured during centrifuge
(2005), Bo et al. (2009) and Feng et al. (2011). tests. The results have been presented in terms of
However, the results presented were extremely site- crater depths, effective depth of improvement and
specific, and limited information on the ground vibrations induced during DC.
43
2 IN-FLIGHT DC ACTUATOR during centrifuge tests by de-activating the motor.
An actuator was custom designed and developed for Thus, the present actuator has been designed to
simulating DC within a centrifuge. In case of enable variation of tamper striking frequency
centrifuge modelling, the body forces of model through control of motor rotational speed, which is
geomaterials are increased by inducing centripetal advantageous while modeling DC in saturated soils.
acceleration N times that of earth’s gravity (g) so as Detailed description of the in-flight actuator is
to replicate full-scale stresses in the centrifuge presented in Kundu and Viswanadham (2020).
model reduced by N times at homologous points.
Table 1 Scaling laws for simulating DC in centrifuge
Accordingly, appropriate scaling laws need to be
Parameters S.F.
established for linking the model behaviour to that Angle of internal friction (ϕ) (°) 1
of the prototype. The relevant scaling laws adopted Unit weight of soil (γ) (kN/m3) N
in this study related to simulation of DC in Relative density of soil (RD) (%) 1
centrifuge are presented in Table 1, based on which Mass of tamper (m) (t) 1/N3
the present actuator has been designed. Radius of tamper (r) (m) 1/N
The various components of the actuator include Drop height (H) (m) 1/N
the support system, impactor assembly, tamper- Time interval between successive blows (ti) (min) 1/N
hook arrangement and guiding rod-assembly. Tamper velocity (v) (m/s) 1
Essential components of the actuator are shown in Ground acceleration (A) (m/s2) Na
Ground velocity (V) (m/s) 1
Figure 1 along with instrumentation details. The
Crater depth (dc) (m) 1/N
support system consists of beam, columns and Depth of improvement (di) (m) 1/N
flanges, and carries the weight of the whole Energy of tamper (E) (t-m) 1/N3
actuator. The impactor assembly as shown in Figure N: Gravity level or scale factor; S.F.: Scale factor; aFor
2 consisting of drive shaft, sheave, follower, steel example, Am/A p = N; m: model; p: prototype.
hoist rope and stopper for hook arrangement
controls the lifting and dropping of tamper. Three Impactor assembly
distinctive components, namely base plate, guide

24V PMDC motor
shafts and collar are integrated to constitute the
Stopper for
tamper used for DC (Figure 3). The hook hook arrangement
arrangement (Figure 4) is used for engaging and
disengaging tamper in-flight during DC. In
addition, guiding rod-assembly (Figure 5) Beam Guiding rod assembly
Column Support system
consisting of central and peripheral rods enable Flanges Tamper-Hook
linear guidance of tamper and nullify the effects of arrangement
Coriolis force generated at high gravities

respectively.
0.16 m
A brief explanation of the mechanism involved in -xm xm
lifting and dropping of tamper is provided here. 0.12 m
ym
Permanent
markers
Activation of motor facilitates rotation of drive shaft
in clockwise direction which causes the cam 0.33N m
mounted on drive shaft to engage the follower of

sheave. This results in simultaneous and
equiangular rotation of sheave and subsequent Soil strata Strong box
lifting of tamper using hoist rope. Further rotation

of driving shaft causes hook arrangement to reach a Figure 1. Schematic representation of DC simulator assembly
position where hook comes in contact with hook Sheave with slot for hoist rope Follower attached to sheave
stopper and rotates anticlockwise about the hinge. Bearing housing
Bearing housing
for sheave
The above process results in release of the tamper, Cam attached to for drive shaft
thereby initiating free-fall on the soil surface. driving shaft

Driving shaft
Electric motor
connected to
Rotation of the drive shaft beyond this point creates Stopper for sheave
drive shaft
a situation when the cam rotates out of contact with
the follower. Subsequently, the hook arrangement Plate supporting
Hoist rope
descends under the influence of gravity and slides central guide rods
over tamper collar to reach the initial position. The Stopper for
hook arrangement Plate supporting
afore-mentioned sequences are repeated to facilitate Peripheral
peripheral guide rods
subsequent blows on the soil surface, and the guide rods Central guide rods
process can be terminated at any point of time Figure 2. Schematic representation of impactor assembly
44
Through holes preparation. The front side of the container was
provided with a transparent perspex sheet of 50 mm
Collar thickness for enabling view of front elevation of the
Sloping surface model during testing. The plane of the perspex sheet
served as a plane of symmetry, on account of which,
Guide shafts only half of the soil mass and cylindrical tamper
were modelled. With this, axial symmetry of DC on
Base plate
the soil surface was modelled. A set of twelve
permanent markers made of thin transparency
Figure 3. Details of tamper sheets were pasted along inner side of perspex sheet,
thereby constituting a series of quadrants consisting
Hoist rope of four markers along the corners, as shown in
Figure 1. These markers serve as reference points
Hook for carrying out GeoPIV analysis on images
holding block captured during centrifuge tests. Further, in order to
simulate homogeneous field conditions in the small-
Hook scale centrifuge model, the sand models were
Hinge prepared in a loose dry state by adopting the air-
Platform to
support tamper pluviation technique and placed at a relative density
Sloping
(RD) of 35%. During model preparation
51° surface accelerometers were placed on the soil surface at
Figure 4. Hook arrangement and associated components desired locations.
A large beam centrifuge of 4.5 m radius and
capacity of 2,500 g-kN at the National Geotechnical
Hook arrangement Centrifuge Facility (NGCF), Indian Institute of
Technology Bombay, INDIA was used to conduct
Tamper centrifuge tests. Details of the centrifuge equipment
Guide rods support and associated specifications are available in
Chandrasekaran (2001). After completion of
various stages involved in model preparation, the
model container was mounted on the swing basket
Central guide rods of centrifuge. The in-flight DC actuator along with
Peripheral other accessories were subsequently assembled and
guide rods
Soil strata
fixed thoroughly. In addition, the front elevation of
model was illuminated to enable clear visualization
Figure 5. Guiding-rod assembly
of the model at all stages of centrifuge testing.
After taking 1g readings, centrifuge was started
3 MODEL MATERIAL AND and the desired g-level was attained. Then motor
METHODOLOGY ADOPTED was activated remotely for raising and dropping
The model soil used in the study is Goa sand having semi-cylindrical steel tamper (mass and radius of
a specific gravity of 2.654 with emax = 0.94 and emin tamper scaled as per modelling considerations in
= 0.63. The sand is poorly graded with negligible centrifuge) 16 times on the surface of the soil
fines, and can be classified as SP [Cu = 2.065; Cc = deposit from a prefixed drop height. Selection of 16
1.117] according to the Unified Soil Classification number of blows was made considering standard
System (USCS). The model soil thus represents a field observations. After imparting the 16th blow to
typical granular material existing in loose state in the soil surface, the motor was deactivated and
the field, and exhibits characteristics of soils stopping of centrifuge equipment was initiated.
typically selected for DC projects. Peak ground accelerations and velocity induced
The present paper discusses the results of two during tamper drops were analysed with the help of
centrifuge model tests performed at 30 gravities recorded accelerometer data. A GoPro digital
using the developed in-flight DC actuator. The first camera (Hero 5 Black, USA), having a resolution of
test was performed on dry sand with 56 t-m (ton- 12 megapixels was used to record proceedings of
meter) energy, while the second test involved 208 t- the experiment in-flight condition by capturing one
m energy. A rigid container with internal image in every two seconds. Geo-PIV based image
dimensions of 720 mm in length, 450 mm in breadth analysis [White et al. (2003)] on images captured
and 410 mm in depth was used for model during flight was used to plot crater formations,
45
displacement contours, volumetric strain contours centrifuge test results, di and di,e obtained herein are
and increase in relative density post DC. compared to the standard formula [Equation (1)]
suggested by Menard and Broise (1975).
di or di,e = n(mH)0.5 (1)
The results of centrifuge tests on dry sand are
discussed briefly in this section. All values are where, m is mass of tamper and H is height of fall.
reported in prototype scale. The variation of crater Number of blows
0 2 4 6 8 10 12 14 16
depths (dc) and crater radius (rc) corresponding to 16 0.0 0.0
blows are shown in Figure 6 for both the models. E = 56 t-m
TC1 depth
E = 208 t-m
TC2 depth
0.2 0.5
Crater depth and radius increased progressively TC1 width TC2 width
with successive blows. The maximum crater depth
Crater radius, rc (m)

Crater depth, dc (m)
0.4 1.0
and radius are observed to be about 0.49 m and 1.8
0.6 1.5
m respectively in Model TC1 at 56 t-m energy level.
Upon increasing the energy level to 208 t-m, the 0.8 2.0
maximum crater depth and crater radius increased
1.0 2.5
to 1.16 m and 2.8 m respectively in Model TC2.
In addition, the displacement contours of Models 1.2 3.0
TC1 and TC2 in radial and vertical directions at the Figure 6. Effect of energy level on crater depth
end of 16 blows are presented in Figure 7 and Figure Radial distance from tamper center, xp (m)
8 respectively. In general, the contours depict Depth from initial ground surface, dp (m)
-10
00
-10 -8
-8 -6
-6 -4
-4 -2 0 22
Tamper
4 66 88 10
10
considerable displacement of soil owing to

densification, which gradually reduced with depth 22
and radial distance from the center of tamper. It can 44

be inferred that the vertical and radial boundaries of
the displaced zone substantially increased on 66
account of higher energy levels simulated in Model 88

TC2 as compared to Model TC1. Model TC1
E = 56 t-m
Model TC2
E = 208 t-m
Further, to study the depth of improvement 10
10
Radial
induced by DC, the increase in relative density (RD) displacement (m) 0 0.01 0.02 0.04 0.08 0.16 0.24
or volumetric strain (εv) contours post DC are Figure 7. Effect of energy level on radial displacement of soil
presented in Figure 9 for Models TC1 and TC2. In Radial distance from tamper center, xp (m)
the present study, depth of improvement (di) for -10 -8 -6 -4 -2 0 22 4 66 88 10
Depth from initial ground surface, dp (m)
-10 -8 -6 -4 10
00 Tamper
models prepared with sand is considered as the
thickness of soil strata measured from initial ground 22
surface to a depth below which ΔRD is less than

44
10%. (εv) corresponding to ΔRD of 10% is 1.7%.
The corresponding variation of di,e with all 16 blows 66
are shown in Figure 10. The effective depth of

88
improvement (di,e) is defined as the depth of Model TC1 Model TC2
improvement measured from the base of crater. This 10
10
E = 56 t-m E = 208 t-m
Vertical
is denoted as di,e, which is equal to difference of displacement (m) 0 0.01 0.02 0.04 0.08 0.16 0.32 0.64 1.15
depth of improvement and depth of crater (di - dc). Figure 8. Effect of energy level on vertical displacement
The value of di,e in dry sand increased from 4.1 m
for low energy DC in Model TC1 to 5.59 m for high -10 -8 -6
Radial distance from tamper center, xp (m)
-4 -2 0 22 4 66 88 10
Depth from initial ground surface, dp (m)
-10 -8 -6 -4 10
energy DC in Model TC2. Further, the average 00
dc = 0.49 m
Tamper
dc =
radial width of improvement (ri) increased from 2 m 22 di =
1.16 m
to 4 m with increase in blow energy. Thus the 4.6 m di,e = 4.1 m

di,e = di =
approximate volume of soil (πri2di,e) improved by 44 5.59 m 6.75 m
DC increased from 51 m3 to 281 m3 corresponding 66

to an increase in energy from 56 t-m to 208 t-m.
Thus, an increase in DC energy in the order of 3.71 88
Model TC1 Model TC2
times resulted in increased soil volume by about 5.5 10
10
E = 56 t-m E = 208 t-m
Final relative density ΔRD (%) 35 40 45 55 75 95
times. Further, increase in ri value by 2 times with Increase in relative density ΔRD (%) 0 5 10 20 40 60
increasing energy signifies that for higher energy Volumetric strain εv (%) 0 0.85 1.70 3.40 6.80 13.6
levels, double the tamping spacing can be adopted Figure 9. Effect of energy level on contours of relative density
in the field. In order to gain insight into the
46
Number of blows 1.6 16
Radial distance from tamper center
0 2 4 6 8 10 12 14 16 1.4 4.5 m (Radial) 4.5 m (Vertical) 14
0 9.0 m (Radial) 9.0 m (Vertical)
Effective depth of improvement, di,e (m)
1.2 13.5 m (Radial) 13.5 m (Vertical) 12
Acceleration, PGA (m/s2)

Acceleration, PGA (g)
1
E = 56 t-m E = 208 t-m 1.0 10
Tamper
2 0.8 8
TC1 TC2
0.6 6
3 Accelerometers
0.4 4
4 0.2 2
0.0 0
5
0 2 4 6 8 10 12 14 16
Number of blows
6 (a) Model TC1 [Dry sand; E = 56 t-m]
Figure 10. Effect of energy level on (di,e) with blows
1.6 16
In the present study, the value of the empirical 1.4
4.5 m (Radial)
9.0 m (Radial)
4.5 m (Vertical)
9.0 m (Vertical)
14
constant n is computed for both (di) and (di,e). For 1.2
13.5 m (Radial) 13.5 m (Vertical)
12
Acceleration, PGA (m/s2)

Acceleration, PGA (g)
Model TC1 n (di) = 0.61 and n (di,e) = 0.55 and for 1.0 10
Model TC2 n (di) = 0.47 and n (di,e) = 0.39. This 0.8 8
indicates that with increasing energy level the n 0.6 6
value decreases.
0.4 4
In addition, the PGA and PGV values along radial
0.2 2
and vertical directions induced during DC in soils
0.0 0
for energy levels of 56 t-m and 208 t-m after the 1st, 0 2 4 6 8 10 12 14 16
4th, 8th and 16th blow are presented in Figs. 11(a) - Number of blows
(b) Model TC2 [Dry sand; E = 208 t-m]
11(b) to Figs. 12(a) - 12(b). Figs. 11(a) - 11(b) Figure 11. Variation of peak ground acceleration (PGA) with
indicate that the PGA values after 16th blow are successive blows
0.44g [radial] and 0.38g [vertical] for Model TC1,
which increased to 1.5g [radial] and 1.23g [vertical] Hence, the observed PGV induced during DC are
respectively in Model TC2 upon increasing the well within the maximum magnitude of 76 mm/s at
tamper energy. Based on Figs. 12(a) - 12(b) the peak a radial distance of 13.5 m away from tamper,
radial velocity near tamper after 16th blow is thereby indicating relevance of the technique.
observed to be about 40 mm/s in Model TC1, and 225
about 211 mm/s in Model TC2. The corresponding 200
Peak ground velocity, PGV (mm/s)
peak vertical velocity is about 25 mm/s in Model 175

9.0 m (Radial)
13.5 m (Radial)
9.0 m (Vertical)
13.5 m (Vertical)
TC1 and about 112 mm/s in Model TC2. Hence, as 150
evident from the PGA and PGV plots, an increase in 125 Tamper
energy level from 56 t-m in Model TC1 to 208 t-m 100
in Model TC2 resulted in substantial increase in 75
PGA and PGV values monitored at the same blow 50
Accelerometers
number. The test results in terms of crater depth, 25
effective depth of improvement, PGA and PGV 0

0 2 4 6 8 10 12 14 16
values are summarized in Table 2. The Number of blows
specifications of FHWA [Lukas (1995)] restricts (a) Model TC1 [Dry sand; E = 56 t-m]
PGV to 76 mm/s to avoid hazards to adjacent 225
structures. In the present study, the PGV at 13.5 m 200 4.5 m (Radial) 4.5 m (Vertical)
Peak ground velocity, PGV (mm/s)

distance from tamper center are 17.32 mm/s and 175 13.5 m (Radial) 13.5 m (Vertical)
32.8 mm/s for Model TC1 and TC2 respectively. 150

125
Table 2 Summary of centrifuge test results 100
Test E adi,e (dc)max a
PGA a
PGV 75
legend (t-m) (m) (m) (g) (mm/s) 50

25
Radial Vertical Radial Vertical
0
TC1 56 4.10 0.49 0.202 0.128 17.32 10.9 0 2 4 6 8 10 12 14 16
TC2 208 5.59 1.16 0.191 0.284 32.8 18.2 Number of blows
All values are reported in prototype scale; aReported after (b) Model TC2 [Dry sand; E = 208 t-m]
Figure 12. Variation of peak ground velocity (PGV) with
16th blow at 13.5 m from tamper center; E: Tamper
successive blows
energy in each blow; (dc)max: Maximum crater depth;
47
5 CONCLUSIONS Hajialilue-Bonab, M. and Zare, F.S. 2014. Investigation
on tamping spacing in dynamic compaction using
The present study was intended to investigate the model tests. Ground Improvement, 167(3): 219-231.
response of dry, loose granular soil deposits
subjected to DC within a geotechnical centrifuge. A Kundu, S. and Viswanadham, B.V.S. 2016. Studies to
custom-designed actuator was developed in this evaluate the impact of tamper on the depth of
regard, design details of which are presented. The improvement in dynamic compaction. Japanese
actuator is found to be capable of replicating low- Geotechnical Society Special Publication, 2(59):
energy as well as high-energy DC processes in- 2033-2037.
flight within a geotechnical centrifuge. The results
obtained in the course of the present study indicated Kundu, S. and Viswanadham, B.V.S. 2018. Influence of
tamper shape on dynamic compaction of granular
the effectiveness of both low-energy and high-
soil. Proceedings of the 9th International Conference
energy DC processes in improving the on Physical Modelling in Geotechnics, London, UK,
characteristics of loose granular soils. The crater 17-20 July 2018, Andrew McNamara, Sam Divall,
depths, effective depth of improvement and ground Richard Goodey, Neil Taylor, Sarah Stallebrass,
vibrations induced during DC were found to Jignasha Panchal (Eds), Taylor & Francis Group
increase with higher tamping energy. Further, the (Pubs.), Vol. 2, pp. 1205-1209.
close correlation of depth of improvement obtained
during centrifuge tests using the developed actuator Kundu, S. and Viswanadham, B.V.S. 2020. Design and
with corresponding data reported in the field development of an in-flight actuator for modelling
indicated the robustness and sound performance dynamic compaction in a geotechnical centrifuge.
characteristics of the actuator. Geotechnical Testing Journal, ASTM, Manuscript
number: GTJ-2019-0169R2
6 ACKNOWLEDGEMENTS Lukas, R.G. 1995. Geotechnical Engineering Circular
The authors would like to thank the staff of the No. 1 - Dynamic compaction. Federal Highway
National Geotechnical Centrifuge Facility and of Administration Report No. - FHWA-SA-95-037,
Geotechnical Engineering Laboratory at the Indian Washington D.C.
Institute of Technology Bombay, Mumbai, INDIA,
for their support throughout the present study. Mayne, P.W., Jones, J.S. and E’Dinas, D.C. 1984.
Ground response to dynamic compaction. Journal of
Geotechnical Engineering, ASCE, 110(6): 757-774.
7 REFERENCES
Arslan, H., Baykal, G. and Ertas, O. 2007. Influence of Menard, L. and Broise, Y. 1975. Theoretical and
tamper weight shape on dynamic compaction. practical aspect of dynamic consolidation.
Ground Improvement, 11(2): 61-66. Géotechnique, 25(1): 3-18.
Bo, M.W., Na, Y.M., Arulrajah, A. and Chang, M.F. Merrifield, C.M. and Davies, M.C.R. 2000. A study of
2009. Densification of granular soil by dynamic low-energy dynamic compaction: field trials and
compaction. Ground Improvement, 162(3): 121-132. centrifuge modelling. Géotechnique, 50(6): 675-681.
Chandrasekaran, V. S. 2001. Numerical and centrifuge Oshima, A. and Takada, N. 1998. Evaluation of
modelling in soil structure interaction. Indian compacted area of heavy tamping by cone point
Geotechnical Journal, 31(1): 30-59. resistance. Proceedings of the Centrifuge’98, Tokyo,
T. Kimura, O. Kusakabe and J. Takemura (eds),
Feng, S. J, Shui, W. H, Tan, K., Gao, L. Y. and He L.J. Rotterdam: Balkema (pubs.), Vol. 1, pp. 813-818.
2011. Field evaluation of dynamic compaction on
granular deposits. Journal of Performance of White, D.J., Take, W.A. and Bolton, M.D. 2003. Soil
Constructed Facilities, ASCE, 25(3): 241-249. deformation measurement using particle image
velocimetry (PIV) and photogrammetry.
Feng, T.W., Chen, K.H., Su, Y.T. and Shi, Y.C. 2000. Géotechnique, 53(7): 619-631.
Laboratory investigation of efficiency of conical-
based pounders for dynamic compaction. Zou, W. L., Wang, Z., Yao and Z. F. 2005. Effect of
Géotechnique, 50(6): 667-674. dynamic compaction on placement of high-road
embankment. Journal of Performance of Constructed
Hajialilue-Bonab, M. and Rezaei, A.H. 2009. Physical Facilities, ASCE, 19(4): 316-323.
modelling of low-energy dynamic compaction.
International Journal of Physical Modelling in
Geotechnics, 9(3): 21-32.
48
Model preparation for unsaturated soil testing in a centrifuge environment

Leonardo Maria Lalicata1, Luc Thorel2, Augusto Desideri3, G.M. Rotisciani3, Francesca Casini4
1
School of Engineering and Mathematical Sciences, City, University of London, London, UK
2
GERS Department, Geomaterials and Modelling in Geotechnics Laboratory, Univ. Gustave Eiffel, Nantes Campus,
Bouguenais, France
3
Department of Structural & Geotechnical Engineering, University of Rome La Sapienza, Rome, Italy
4
Dipartimento di Ingegneria Civile e Ingegneria Informatica, Università` degli Studi di Roma “Tor Vergata”, Rome,
Italy
Corresponding author: Leonardo Maria Lalicata (leonardo.lalicata@city.ac.uk)
ABSTRACT: This paper presents selected aspects of the model preparation and the experimental procedures
used for testing the unsaturated soil behaviour in centrifuge environment. The study is based on the results
of centrifuge tests carried out to investigate the influence of partial saturation on the behaviour of laterally
loaded piles. The soil used in the experimentation is a high permeability silty clay soil, named B-grade
kaolin. The models were statically compacted at two different densities with the same water content. In
order to reduce the after compaction suction, the samples were subjected to an imbibition process at 1×g
connecting the bottom of the model with the water reservoir. The influence of the compaction process and
the imbibition stage on the subsequent inflight soil conditions are discussed and analysed. The centrifuge
tests were realised at the centrifuge centre of the University Gustave Eiffel, in Nantes, France, in the
framework of the GEOTRANSALP project.
Keywords: Unsaturated Soils; Static Compaction; Water Retention Curve; Centrifuge Modelling; Hydro-Mechanical
Stress Paths.
1 INTRODUCTION be measured although the horizontal stresses are in

Centrifuge modelling represents a valid general unknown.
methodology to investigate the performance of This paper presents the details of the model
different types of geotechnical works such us preparation of the experimental investigation
shallow and deep foundations, tunnels, slopes and conducted by Lalicata et al. (2019). The study
embankments. Traditionally, studies have been aimed at assessing the influence of partial saturation
conducted on fully saturated or completely dry soil on the response of laterally loaded piles installed in
models. However, in some cases the variation of the fine-grained soils. The centrifuge tests (N = 100×g)
water content in the soil due to the water table were carried out using the centrifuge facilities of the
oscillation, rainfall events, interaction with the University Gustave Eiffel, located close to Nantes
atmosphere and exceptional events (i.e. pipe rupture (France) in the framework of the GEOTRANSALP-
etc…) may be crucial for the geotechnical works PILE-UNSAT agreement.
behaviour. These and other problems have been
recently investigated by centrifuge modelling 2 MATERIALS
(Casini et al., 2009; Soranzo et al., 2015; Thorel et 2.1. The soil used
al., 2011). Scaling laws for unsaturated soils have The material used is a commercial soil named B-
been experimentally investigated, among others, by
Grade kaolin (90% fine silt and 10% clay). This soil
Depountis et al. (2001) and Caicedo et al. (2006) has a low plasticity index (IP=14%) and a low liquid
under different soil condition and g-levels. The limits (wL=42%), the plastic limit is wP=28%. A
authors found that capillary rise and diffusion time complete hydro-mechanical characterisation was
in the centrifuge could be scaled by 1/N and 1/N2, conducted in the geotechnical laboratory using
respectively. Details of the analytical formulations standard apparatus for unsaturated and saturated
can be found in Caicedo and Thorel (2014). soils. Results are reported in Lalicata (2018) and
Compaction is the most common technique to Lalicata et al. (2020). For the sake of brevity, only
prepare unsaturated soil samples (Caicedo and the flooding oedometric tests and the suction
Thorel, 2014). Dynamic compaction allows to controlled oedometer tests are presented here. The
control the density but not the compaction stress. On
hydro-mechanical parameters of the soil are listed
the other hand, with the uniaxial static compaction in Table 1.
procedure both density and compaction stress can
49
Table 1. Hydro-mechanical properties of B-Grade kaolin. ratios are selected (0.93 and 0.75) prepared with the
ρs CC CS N0* ϕ' cv same water content w=15%. The two densities are
(g/cm3) (°) (mm2/s) considered representative, respectively, of a
2.66 0.26 0.078 1.36 22 1.0 collapsible soil and a swelling soil under wetting
*N0 is the voids ratio at σ’v=1 kPa under oedometric conditions. processes. The higher void ratios of 1.12 was
discarded as showed deformation too elevated,
Several imbibition tests in the oedometer Figure 1b.
apparatus were undertaken to define the after
compaction conditions (w, e) for centrifuge models.
The samples (18 mm thick and 70 mm in diameter)
were prepared with different compaction features
and subjected to different total vertical stresses σv
applied before the wetting stage. In the main series
of the tests, the vertical stress was 150 kPa
corresponding to the vertical stress active at half of
the pile embedded length (z=8m). Samples were left
equalise under the external load before filling the
oedometer cell with water.
The soil powder was dried at 105 °C for 24 hours
before adding the demineralized water to reach the
target water content. The moisture was manually
mixed and then it was kept in sealed bags for 24-48
hours to homogenise in a temperature-controlled
environment. The specimens were statically
compacted (v=1.5 mm/min) directly in the
oedometric ring.
Three voids ratio are taken into account (e0 = 0.77,
0.92 and 1.12), for water content ranging from 10%
to 26%, with steps of 4%. For each grid point, at Figure 1. Wetting induced deformation versus: a) initial void
least two samples were tested. ratios, b) total vertical stress.
Figure 1a shows the wetting induced deformation
under σv = 150 kPa in function of the compaction The soil water retention curve (SWRC) was
void ratios. Water content measurements at the end obtained by means of the suction controlled
of the test have shown that all the samples reached oedometer (Romero et al. 1995). The data
degrees of saturation closed to one (Srmin=0.95). corresponding to the main wetting curve a two
Results are grouped for the initial water content. For different void ratios, 0.93 and 0.75, are presented in
the material, the latter does not have a significant Figure 2 in terms of suction s (in log scale) versus
influence on the wetting deformation while the the degree of saturation Sr. In the suction range
results strongly depends on the initial void ratios. experimentally studied, the data exhibit a marked
The deformations reduce approximately linearly as influence of porosity in the retention behaviour. In
the voids ratio decrease. Little swelling was fact, as porosity decreases, air entry value increases
observed for the more compacted samples and the slope of the transition zone reduces in the
(e0=0.77). Sr–s plane. The Van Genuchten (1980) equation was
The influence of the applied external load during used to fit the data:
samples inundation is depicted in Figure 1b, for
m
stresses ranging from 0 to 600 kPa. Samples were  1 
prepared at e0= 1.12 and w= 14%. Soil swelling S r =S r ,res + ( S r , sat − S r ,res ) ⋅   (1)
 1 + (α s ) n 
occurs only in absence of external load applied,  
while since from σv= 50 kPa the soil experience
positive deformations that increase up to σv= 300 Where Sr,res and Sr,sat are respectively the residual
kPa and then they slightly reduce in magnitude. and saturated degree of saturation;  is the inverse
Thus, in the centrifuge, this soil would have of the air entry value, n governs the slope of the
experienced modest swelling in the shallower layer transition zone and m controls the inflection point at
and volume reductions increasing with depth. the residual values of suction. The values of the
In order to explore the influence of the compaction parameters are listed in Table 2. Neglecting the
conditions of soil in the centrifuge tests, two void hysteresis of the SWRC, eq (1) can be used to
50
estimate the as-compacted suction and thus the The model was extensively instrumented in order
initial effective stress in the model according to the to follow both the equalisation phases and pile
Bishop effective stresses definition: σ ' =σ + S r ⋅ s loading (Figure 5).
(Bishop and Blight, 1963). As the total stress is zero
after the compaction the effective stresses in the soil
reduces to the product S r ⋅ s , and are 387 kPa for
e0= 0.93 (Sr0= 0.43) and 1768 kPa for e0=0.75
(Sr0=0.52) respectively.
10000
suction, s = ua - uw (kPa)
1000
100 Figure 3. Model pile and calibration set-up.

e =0.75
0
e =0.93
0
10
VG e =0.75
0
VG e =0.93
0
1
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
degree of saturation, Sr (-)
Figure 2. Soil water retention curves, SWRC, for the two
densities used in the centrifuge tests.
Table 2. Van Genuchten parameters for different void ratios.

e0 Srsat Srres α (1/kPa) n m Figure 4. Model assembled in the strongbox
0.93 1.0 0.16 0.02 1.4 0.25
0.75 1.0 0.16 0.006 1.3 0.23
2.2. The model pile

The model pile (scale 1/100) is a close-ended
aluminium tube (external diameter D=12mm,
thickness t= 1mm) embedded into the soil for 150
mm. The pile was instrumented with 10 pairs of
strain gauges arranged every 15 mm, providing the
bending moment profile along the pile length. The
strain gauges were calibrated in the laboratory
adopting the cantilever scheme with point load. The
experimental arrangement for the calibration is
presented in Figure 3.
2.3. The experimental apparatus

The axial symmetric models 180 mm deep were
prepared in 300 mm diameter rigid tub. As shown
in Figure 4, two twin containers were spin up at the
same time: one was used for CPT measurement
before, during and after the test. The other was used
for the load test on the pile that was bored in the
centre of the model, at 1×g, by means of a manual
screwing system. The screw was mounted into a
Figure 5. Model instrumentation.
rigid frame to maintain position and verticality.
51
The bottom of the model was connected to a water The moisture was prepared following the same
reservoir, the level of which was governed by an procedure used for the flooding tests. After the
electro-pneumatic valve system remotely controlled moisture homogenisation, the soil was then passed
by the operator from the centrifuge control room. in a mechanical mixer to size down the aggregates
A laser sensor measured the water height in the tank. and stored again in sealed plastic bags.
Five LVDT sensors measured soil settlements, three
tensiometers, installed on the side of the container, 3.2. Compaction
measured the pore water pressure (negative and The soil models were statically compacted, in a
positive) in the soil. The lateral load was applied at rigid tub of 300 mm in diameter. An extension is
35 mm from ground level by a hydraulic actuator. required to compact the soil to the target density. A
The loading phase was displacement-controlled 10 mm thick sand layer, surrounded by geotextile,
(v=0.003 mm/s at model scale), and a load cell allowed the drainage at the bottom of the model. A
measured the lateral load. One LVDT built in on a 2 mm thick plastic sheet coated the inner side to
rotational joint gave the pile’s vertical displacement reduce skin friction during model preparation and
and rotation. during the test.
The soil is compacted in six layers of final height
3 CENTRIFUGE TEST PROCEDURE 30 mm with a displacement rate of 1.5 mm/min. The
In total four centrifuge tests were undertaken, compaction is a displacement-controlled process:
varying both the elevation of the water table, zw, and the press stops when the target displacement was
the initial void ratio of the soil, e0 (Table 3). In the reached. In Figure 6 typical results of the
two tests where the water table was 70 mm from the compaction process are presented for one loose soil
ground level the pile was loaded until a lateral sample (e0=0.93).
displacement, y, of 30-40% of the diameter, D, was After the compaction of each layer, its surface was
reached. In the following step, the water table level scarified to achieve a good connection with the
was raised to ground level keeping the load sequent one.
constant. For the sake of comparison, two additional The soil surface was covered with a plastic film to
tests were conducted up to soil failure in fully prevent water evaporation in the upper part of the
saturated condition (zw=0) on soil samples prepared model.
at the same voids ratio of the main tests.
Measurements of pile displacements, lateral load compaction stress, σ (kPa)
vc
and bending moments were recorded. 0 200 400 600 800
In order to reduce the as compacted suction, the 210
samples were subjected to an imbibition process at
180 0
1g, connecting the bottom of the model with the
model height (mm)
water reservoir. 150 30

Before pile loading, equalisation of pore pressures depth, z (mm)
in the soil was allowed. The load was applied with 120 60
a very slow displacement rate to avoid the creation
90 90
of excess pore pressures.
At the end of each test, undisturbed specimen were 60 120
sampled to obtain the water content and voids ratio
distribution with depth. 30 150
0 180
Table 3. Initial conditions of the centrifuge tests.
Figure 6. Typical results of the compaction process for one of
Test e0 zw w0 σvc Sr0
the low density models.
ID (mm) (kPa)
T 05 0.93 0 15.03 580 42.02
The reasonable homogeneity of the distribution of
T 06 0.93 70 14.67 559 42.01
vertical compaction stresses, Figure 7, and the good
T 08 0.75 70 14.72 1395 51.03 repeatability of the results in the different tests
T 09 0.75 0 14.72 1395 51.03 confirm the validity of the procedure used. It has to
be noticed, however, that the dispersion increases
3.1. Model preparation with the soil density. The compaction stress
increases with dry density from a mean value of 500
Samples were prepared by static compaction at two
kPa for the loose soil (e0 = 0.93) to 1400 kPa for the
different voids ratios (e0= 0.93 and 0.75) and the
dense soil (e0 = 0.75).
same water content (w=15%).
52
compaction stress, σ (kPa)
vc
0 500 1000 1500 2000 2500

0
30
depth, z (mm)
60
90
120
150
180
e = 0.93 e = 0.75
0 0
σ = 500 kPa σ = 1400 kPa
vc vc
Figure 7. Compaction stress profiles in the tests.
4 ANALYSIS OF THE RESULTS

In this section, selected results are presented to
illustrate first the influence of the centrifuge test Figure 8. 1g imbibition: a) adsorbed volume, b) vertical
procedure on the soil behaviour. The complete displacement of soil.
results are presented in Lalicata et al. (2019) and
Lalicata (2018). 4.2. In flight phases
Figure 9 shows the temporal evolution of the
4.1. 1g imbibition vertical displacements at the ground surface during
The absorbed water volume in this stage is the in-flight phase for T06 (e0= 0.93) and T08
presented in Figure 8a. Figure 8b shows the (e0= 0.75). The increase in the gravity loading,
temporal evolution of the vertical displacements at together with the hydraulic boundary condition
the soil surface for the two densities considered. The imposed at the base, leads to an instantaneous
wetting process induces the development of vertical settlement of about 3 mm and 7 mm for the dense
displacements directed upwards attaining a final and the loose soil respectively. The vertical
values of approximately 3 mm for the loose soil and displacements increase in time up to 3.5 mm and 10
~6 mm for the dense soil. The water infiltration into mm in the two cases.
the soil layer entails an increase in water volume by After the pile loading, the water table was raised
around 2 lt in both cases. up of 70 mm saturating the whole model. The loose
The vertical displacements increase as the wetting sample globally settles as the volumetric collapse of
front advances. Despite the absorbed water volume the unsaturated part is greater than the volumetric
is approximately coincident in the two cases, the swelling of the deeper soil layer. On the other hand,
wetting-induced swelling in the dense soil is the LVDTs measured very little increase in soil
roughly the double of that found in the loose soil. volume in the dense sample.
This outcome reflects the relevant difference in the
time (min)
as-compacted suction and the degree of saturation 0 120 240 360 480 600 720
(see Table 8). Because of the different shape of the
soil displacements (mm)
0
SWRC, (see Figure 2), the same approximate
2 e =0.75
variation in Sr leads to a suction decrease and, thus 0
4
to a stress unloading, greater in the dense than in the in flight equ.
loose soils. 6 w.t. rising
The overall imbibition process induced by 8
capillary rise at 1g with zero pore pressure at base 10
had not still reached stationary conditions, because 12
e =0.93
0
the main aim of this phase was to increase the

suction to speed-up the equilibrium in the following Figure 9. Soil vertical displacements versus time in test T06
(e0=0.93) and T08 (e0=0.75).
in-flight phase.
53
The instantaneous soil volume reduction observed considerations about the soil state at the end of the
as g increases is possible because the soil is not fully tests. The reference test on saturated soil conditions
saturated. Indeed, the increment of total stresses is for the dense soil, T09 (e0=0.75), shows a profile of
taken all by the effective stresses, as the presence of voids ratio approximately constant with depth to a
air in pores allows for a soil skeleton compression. value of ~0.75. For the loose soil (T05) the voids
According to the previous imbibition stage, the soil ratio decreases with depth from a value of ~0.9 at
degree of saturation is not uniform in the soil but the ground surface to a value of 0.76 for a depth of
reduces moving towards the ground surface (far 170 mm. The latter distribution well fit the
from the bottom boundary condition). Thus, the oedometric normal consolidation line, NCL, below
initial settlement is likely to be related to the 40 mm of depth, while the denser ones intercept it
compression of the upper part of the model. at z= 120 mm. The models were subjected to the
Conversely, the consolidation settlement has to be same experimental procedure (1g imbibition,
attributed to the excess pore pressures dissipation in increase of total stress and in-flight equalisation).
the lower part of the model. Thus, these differences are related both to the
The pore pressure evolution for the test T06 different initial void ratio and the different shape of
shown in Figure 10 qualitatively confirms these the SWRC that controls the variation of mean
findings. In fact, the in lower part of the model effective stress and the preconsolidation pressure
(z= 140mm), the pore pressure rises as g increases during hydro-mechanical stress paths.
and then reduces with time. In the upper part
(z= 40mm), it reduces slightly at the beginnings.
After four hours of consolidation the trends seems
to be stabilised but the absolute values are different
from those expected. The tensiometers correctly
capture the pore pressure variation during the rising
of the water table; they measured approximately
+70 kPa at the end the test. Since the instruments
worked well at 1g environment both before and after
each test, further investigations are needed to
understand the validity of this tools in high
gravitational fields.
T06 pressure at the base
200 e =0.93
pore pressure (kPa)
0
w.t. rising
100 z=140mm
z=90mm
0 z=40mm
Figure 11. Pile loading and rising of the water table stages: a)
lateral displacement versus time, b) latera load versus time.
-100
0 120 240 360 480 600 720
Void ratios, e
time (min)
0.6 0.7 0.8 0.9 1
Figure 10. Pore pressures versus time in T06 (e0=0.93).
0
The pile head displacement and force evolution

30
versus time are presented in Figure 11 a-b for the
two soil densities. During the saturation at constant 1D NCL
depth, z (mm)
load, the piles head displacements increased more 60

for the sample with e0= 0.93 than the one with
e0= 0.75. In both cases, the inundation event 90
provoked the collapse of the pile. In fact, the
displacements keeps growing even if the pore 120
T05 (e =0.93)
pressures tend to the equilibrium (Figure 10). 0
T06 (e =0.93)
150 0
4.3. End of test measurements T09 (e =0.75)
0
The voids ratio profiles with depth, at the end of the 180
tests, in Figure 12 allows drawing same general Figure 12. Void ratios profiles with depth at the end of the tests.
54
In test T06, after the water table rising, below Caicedo, B. and Thorel, L. 2014. Centrifuge Modelling
z= 90 mm, the voids ratio is the same of T05 of Unsaturated soils. Advances in the Mechanics of
(saturated reference test). From 40 to 90 mm of Unsaturated soils. Journal of Geoengineering
depth the effects of the volumetric collapse upon Sciences, 2(1-2): 83-103. doi:10.3233/JGS-130013.
wetting are very significant while this has a small Caicedo, B., Medina, C., and Cacique, A. 2006.
influence in the shallower part of the model (z<40 Validation of time scale factor of expansive soils in
mm). centrifuge modeling. In Proceedings of the
International Conference on the Physical Modelling
5 CONCLUDING REMARKS in Geotechnics, Hong Kong, 4-6 August 2006.
Balkema, Rotterdam, pp. 273–277.
The paper presents the experimental details of the
centrifuge testing on an unsaturated fine graded soil Casini, F., Munoz, J., Lorenco, S., Thorel, L., Vaunat, J.,
focusing on the soil behaviour from the model Delage, P. and Gallipoli, D. 2009. Centrifuge
preparation to the end of the test. modelling of an unsaturated collapsible soil. In
The study demonstrate the feasibility of centrifuge Between theory and practice in unsaturated soil
modelling in unsaturated soils highlighting the mechanics. Rotterdam: Millpress.
influence of the initial density on the soil response
during the flight: Depountis, N., Davies, M.C.R., Harris, C., Burkhart, S.,
Thorel, L., Rezzoug, A., Konig, D., Merrifield, C. and
• Low compacted soils experience significant Craig, W.H 2001. Centrifuge modelling of capillary
positive deformations during both flight and the rise. Engineering Geology, 60(1-4): 95-106.
rising of the water table.
• Dense soils show a moderate compression during Lalicata L.M., Rotisciani G.M., Desideri A., Casini F.,
the g-level increase and the application of the Thorel L. 2020. Numerical Study of Laterally Loaded
hydraulic boundary condition at the bottom. Pile in Unsaturated Soils. In Geotechnical Research
Moderate swelling was observed during the for Land Protection and Development. Springer,
Cham. doi.:10.1007/978-3-030-21359-6_76
water table rising.
Useful insights on the soil response in flight can Lalicata, L. 2018. Effect of saturation degree on the
be collected from simple flooding oedometer tests. mechanical behaviour of a single pile subjected to
On the other hand, the knowledge of the SWRC, and lateral forces (in Italian). Ph.D. thesis, Department of
its dependence from void ratios, is crucial to analyse Structural and Geotechnical Engineering, Sapienza
the hydraulic behaviour of the soil during the Università di Roma, Rome.
centrifuge tests.
Lalicata, L., Desideri, A., Casini, F. and Thorel, L. 2019.
The 1g imbibition phase prior testing efficiently
Experimental observation on a laterally loaded pile in
reduces the soil suction and speed up the in-flight unsaturated silty soil. Canadian Geotechnical Journal
consolidation time. Some attention is required to not 56(11): 1545-1556. doi.:10.1139/cgj-2018-0322.
over saturate the material before the centrifuge test
beginning. Romero, E., Lloret, A. and Gens, A. 1995. Development
The tensiometers correctly reproduced the pore of a new suction temperature controlled oedometer
pressures variations in-flight but more work must be cell. In Proceedings of the first International
done to investigate the shift observed from the Conference on Unsaturated Soils, Paris, France, 6-8
expected values. September 1995. Balkema, Rotterdam, pp 533-539.
To date, the procedure for the model preparation Soranzo, E., Tamagnini, R. and Wu, W. 2015. Face
requires a significant effort from the researcher that stability of shallow tunnels in partially saturated soil:
limits the applicability to large experimental centrifuge testing and numerical analysis.
campaigns. Further reflections on possible Géotechnique, 65(6): 454–467.
improvements of the experimental procedures are
needed. Thorel, L., Ferber, V., Caicedo, B. and Khokhar, I. 2011.
However, the findings of this study encourage the Physical modelling of wetting-induced collapse in
use of centrifuge modelling to investigate the soil embankment base. Géotechnique, 61(5), 409–420.
structure interaction in unsaturated soils. Van Genuchten, M. T. 1980. A closed-form equation for
predicting the hydraulic conductivity of unsaturated
6 REFERENCES soils. Soil Science Society of America Journal, 44(1):
Bishop, A. and Blight, G. 1963. Some aspects of effective 892–898.
stress in saturated and partly saturated soils.
Géotechnique, 13(3): 177–197.
55
Effect of Crushed Particles on Soil-Structure Interface Behaviour

Angus Pettey1, Charles Heron 1
Nottingham Centre for Geomechanics, University of Nottingham, UK
1
Corresponding author: Angus Pettey (angus.pettey@nottingham.ac.uk)
ABSTRACT: When shear is induced at a soil-structure interface fines will be generated either by particle
crushing or the breakage of asperities. In a standard monotonic interface test these fines will have a
negligible impact upon the interface behaviour. However, during cyclic loading the volume of fines
generated can become significant and hence the long-term effect on the interface behaviour needs to be
investigated. Traditionally interface tests are designed such that the soil sits above the interface (an ‘upright’
test) and hence gravity and segregation would lead to the fines migrating towards the interface thus
amplifying their impact on the observed behaviour. However not all prototype interfaces are orientated in
this upright direction; for example a pipeline will have an ‘upright’ interface at the crown of the pipe, an
‘inverted’ orientated interface at the invert of the pipe (where fines will migrate away from the interface)
and an interface orientated at ninety degrees at the spring-line (side) of the pipe (where fines will migrate
along the interface). As such, it is important to consider every potential prototype case when developing,
for example, constitutive models. The development of such models to predict interface behaviour is
becoming more commonplace, though consideration of the issues from the perspective of real engineering
problems will enhance the utility of any models developed.
A bespoke cyclic interface shearing device has been adapted to facilitate testing of interfaces in a variety
of orientations. This equipment is also capable of providing visual observation of the movement and
crushing of the soil particles during shearing using transparent model containers and high-resolution
imaging. A series of interface tests conducted using this apparatus will be presented in this paper and the
impact of fines, and fines migration, on the interface behaviour will be discussed.
Keywords: Soil-Structure Interface; Granular Mechanics; Particle Crushing.
1 INTRODUCTION driving, interfaces may be sheared hundreds or

The physical changes in soil-structure interfaces thousands of times. The mechanism underlying this
occur in the interface layer. This is a distinct band shearing is yet to be fully understood at cumulative
of soil that is in contact with the structural interface displacements in the order of tens of millimetres, let
and far-field soil. The layer has been defined by alone thousands. It has been shown by
various studies as five to ten times the mean particle experimentation on soil-soil interfaces, that cyclical
diameter (DeJong et al. 2006; DeJong and Westgate shearing gives rise to crushing of granular matter.
2009). For this study, Leighton Buzzard Sand (LBS)
The confinement condition of a soil-structure Fraction B is used as the testing medium due to the
interface has been widely studied. Three distinct wealth of published research into its particle
conditions are found to affect the response of the characteristics. Cavarretta et al. (2010, 2016);
interface: Constant Normal Load (CNL), K = 0; Senetakis et al. (2013), detail the particle crushing
Constant Normal Stiffness (CNS), K = constant; characteristics of LBS Fraction B in particle-particle
Constant Volume (CV), K = ∞ (DeJong and contact. Such research provides a good reference
Westgate 2009; Evgin and Fakharian 1996). The point for developing the more complex model of
condition that occurs at the interface can vary for soil deformation at the interface.
different applications. For example, a piled Another shortcoming of interface research thus
foundation, will experience an approximately far, is that there has only been consideration of the
constant stiffness boundary condition in the radial upright condition of the interface, that is, where fine
direction. material migrates towards the interface layer due to
One major shortcoming in the published literature gravity. An alternate case to model would be the
regarding interface behaviour, is that cyclic inverted interface, where fine material developed
behaviour of the interface has only been considered falls away from the interface. This case is of
up to 45 cycles by DeJong et al. (2006). Other importance when considering the boundary
studies have considered cyclical behaviour, albeit at conditions of the interface, a buried pipeline for
lower cycle number. In ‘live’ engineering problems example, will have both an upright and inverted
such as the thermal cycling of pipelines or pile interface at the top and bottom of its circumference.
57
A third case also exists, where an interface is
orientated at ninety degrees at the spring-line of the
pipe, where fines propagate parallel to the interface.
A diagram of the three principal orientations is
shown in Figure 1.
For the purposes of this study, only the upright
and inverted cases will be considered.
Figure 2. Schematic of updated Toton Shear apparatus
2.2. Test plan

A basic test plan was used to simulate the interface
condition at large cumulative shear displacements.
The test plan is shown in Table 1. Each test consists
of 4000 shear reversals of 10 mm displacements,
peak to peak (2000 full cycles). Although numerous
previously published studies have used the cycle
number to denote the length of a test, this can
Figure 1. Diagram of the three principal interface orientation disguise the true amount of shearing a sample has
conditions.
experienced. For example, a shallow pipeline
experiencing thermal cycling will not undergo
2 METHODOLOGY repeated cycling at a constant displacement for its
2.1. Test apparatus entire lifespan. It would be more accurate to
To identify the impacts of large cumulative consider the cumulative displacement and number
displacements on interfaces, testing is carried out of shearing reversals.
using a bespoke interface shear-testing device, The normal stress, 𝜎𝜎𝑙𝑙 , an d sh ear di splacement
developed within the Nottingham Centre for magnitude, |𝑢𝑢𝑥𝑥 |, are constants throughout testing, to
Geomechanics (Hashemi and Heron 2017). isolate the effect of a change in interface orientation
Since the first uses of this apparatus, several angle, 𝜃𝜃. T he s hear r ate, 𝑣𝑣, is al so co nstant
improvements have been made. Firstly, imaging of throughout testing at 1 mm/s. Each test is repeated,
the soil sample during testing is now possible, using for validation of results, and the identification of
a transparent sample container. Imaging of the soil trends. Note that subsequent use of ‘upright’ and
sample allows for mechanical changes, identified in ‘inverted’ will refer to an interface orientation angle
the interface layer, to be matched to recorded of 0° and 180° respectively, relative to the
changes in sample stress and volumetric behaviour. horizontal.
Such an approach gives more context to trends
identified in results data. Table 1. Test Plan
Secondly, the entire assembly can be rotated Test Code 𝜎𝜎𝑛𝑛 (kPa) |𝑢𝑢𝑥𝑥 |(mm) 𝜃𝜃(°)
through 180° for the testing of the inverted 100-10-000-1 100 10 0
condition. Thirdly, the vertical load motor has been 100-10-000-2 100 10 0
replaced by a pneumatic actuator and pressure 100-10-180-1 100 10 180
controller. The controller has a much faster response 100-10-180-2 100 10 180
time than the previously used motor and encoder
control loop. Proportional–Integral–Derivative
2.3. Testing medium
control is used to set a target pressure in the actuator,
to apply a constant normal load on the sample. This 2.3.1 Soil
improvement allows for a greater stiffness in the far- Prior to testing, the Fraction B LBS was washed and
field soil, giving a closer approximation of the CNL dried, with a particle size distribution analysis
condition than previously published. A schematic of carried out in accordance with BS 1377-1:2016
the Toton Shear apparatus, in its upright (British Standards Institution 2016). It is found that
configuration, with the described improvements is 90% of material passes between 2 mm and 1.18 mm.
displayed in Figure 2. Samples are poured from the same height, through
a sieve of size 3 mm, to limit the rate of each pour
and ensure samples have equivalent relative density
prior to testing. Each test sample is of mass 300 g,
with a cross-sectional area of 80 × 80 mm. At the
58
end of testing, the sample is collected by spatula. frequency of 0.4 Hz. Whilst during initial loading of
The same particle size distribution analysis is then the sample, images are recorded with a 1 Hz
carried out, to assess the change in sample frequency. A high refresh rate is used as it is hoped
constitution. in future studies to utilise the recorded image data
Hutchings and Shipway (2017), show that silica in the use of advanced Particle Image Velocimetry.
sand, such as the LBS used in this study, has a Each frame of image data can be matched to
Vickers hardness value of 700-1200 HV. This is of controller data by the indexing of each frame in the
key importance when considering the tribological recorded data. As such, the micromechanisms that
implications of the shearing of the interface. take place at the interface, can be identified with
added context of the volumetric and stress response
2.3.2 Structural interface plate of the sample. One drawback of this analysis is that
A 3 mm stainless steel plate is used to simulate the the image data only represents a 2D plane of the
structural interface for the tests presented in this sample. An assumption must be made that the
study. Vickers hardness tests have been conducted mechanism witnessed in this plane, must also exist
upon the material used, and an average hardness throughout the depth of the sample.
value of 188 HV has been found across three During post-processing, the raw Portable
separate plates. Network Graphics (.png) files are cropped to the
Each plate of dimension 300 x 108 mm has two size of the sample in the frame, and the frame
tests conducted upon it, in virgin areas. number is added to the image. An example of the
recorded image data is shown in Figure 3. The
2.4. Data collection and processing individual frames are stitched together as a video
file that can be used to examine the evolution of the
2.4.1 Rolling resistance micromechanisms present at the interface.
The Toton Shear apparatus utilises two parallel
1 kN capacity load cells, to transfer rotational force
from the ball screw and nut, applied by a stepper
motor, to linear force upon the shear platen.
The shear platen is mounted upon rails that incur
a rolling resistance. Prior to testing the samples, this
rolling resistance is quantified by a series of 20
calibration shear displacements of 12 mm. A
0.001 inch feeler gauge is passed below the soil Figure 3. Sample image after post-processing
container before calibration, to ensure that the soil
container does not impart any resistance upon the 2.5. Data smoothing
shear platen and has a consistent offset from the The data gathered by the DAQ unit has an
platen, outside of the cross section of the sample. acquisition loop length of 0.1 s. As such, load data,
When processing the raw data, acquired by a particularly the horizontal data, is required to be
National Instruments Data Acquisition (DAQ) unit, smoothed by an appropriate method to remove the
the horizontal loads for forward – load cells in noise in the recorded data.
compression – and backward – load cells in tension Firstly, the average load, across the two parallel
– movement are treated separately, with calibration load cells, for each shear displacement (half cycle)
loads calculated for each. Across all four tests, the is calculated. Hence, for the tests conducted in this
average forward and backwards rolling resistances study, 4000 average loads are presented. The same
were found to be 42.74 N, and -42.58 N method has been carried out for the vertical load
respectively. The corresponding resistance is data, to allow the calculation of stress ratio, µ, for
subtracted, as appropriate, from all load data in each cycle, shown in Equation 1,
subsequent calculations and plots.
𝐻𝐻𝐿𝐿
𝜇𝜇 = (1)
2.4.2 Image data 𝑍𝑍𝐿𝐿
Image data is captured by a Teledyne DALSA

where 𝐻𝐻𝐿𝐿 is the average horizontal load of a single
Genie Nano camera unit and 16 mm lens, mounted
shear displacement, and 𝑍𝑍𝐿𝐿 is the average vertical
adjacent to the sample container. Due to the long
load of a single shear displacement. The
focal length, an exposure time of 0.08 s is used to
unprocessed data is retained to allow for closer
capture each frame. Images of resolution
analysis of response in each displacement.
4112 x 3008 pixels are recorded every 2.5 mm of
travel during shearing, giving an equivalent
59
Figure 4. Plot of the stress ratio, µ, vs cumulative shear displacement
Secondly, a local regression smoothing method the stress ratio of each specimen, which has
(loess) (MATLAB 2019) is used to remove noise in previously been unrecorded in published work. For
the data. Smoothing is carried out upon load data in the test pressure of 100 kPa and cyclic displacement
addition to vertical displacement data, recorded by of 10 mm, an upper limit exists at a stress ratio of
the vertical Linear Variable Differential 0.7. This value seems independent of the orientation
Transformer (LVDT), as to increase the clarity of of the interface, with inverted tests reaching no
general trends across test data. higher value than upright tests. However, the
inverted tests seem to show a more pronounced peak
3 RESULTS AND DISCUSSION in stress ratio before steadily falling.
3.1. Stress response It is evident that there is a similar trend in the rate
of change in stress ratio during the rapid increase.
Figure 4 displays a plot of the evolution of the stress The similarity in the rate of change in stress can be
ratio throughout each of the four tests, up to a better shown by plotting the derivative of stress ratio
maximum cumulative shear displacement of against itself- shown in Figure 5. It is shown that the
40,000 mm. It should be noted that due to a safety
limit within the controller software, tests 100-10-
000-2 and 100-10-180-2 did not complete the full
4000 shear displacements. The safety limit exists as
to prevent damage to the apparatus. Tests that fail in
this manner are unable to be restarted as the sample
will have travelled outside of the tested area of the
interface plate. Hence, unexpected rotations of
particles at the interface may occur. Furthermore,
when the test is stopped for a prolonged period,
shear stress degradation occurs as the linear driving
force on the shear platen is released by the stepper
motor. As such, data for these tests has been cut at
their relevant end points of 37,410 mm and
23,150 mm of displacement respectively.
Fortunately, the data from these tests shows good
similarity to the repeated tests.
An immediate observation that can be made is that
at large displacements, there is a steep increase in Figure 5. Plot of first derivative of stress ratio, d(µ), vs
stress ratio, µ
60
earlier the onset of an increase in stress ratio, the rapidly at the same point, and crushing is also seen
greater the peak rate of increase. Interestingly, once to begin to occur.
the peak stress ratio has been reached, regardless of The crushing of the sample can also be confirmed
the stress path taken, the residual stress ratio is very by grain size distribution analysis. Figure 7 shows a
similar for a test of the same type. This could plot of grain size analysis on each of the four tested
suggest the existence of a critical state style samples and a baseline of the clean, untested
condition. Note also, the inverted tests display a Fraction B. With consideration of the vertical
clear negative gradient after the peak stress ratio has displacement data in Figure 6, samples with a
been reached. greater total contraction, have a greater proportion
of fine particles at the end of the test.
3.2. Volumetric response
The eventual increase in stress ratio may be
explained by the volumetric response of the sample
over large shear displacements. The samples all
contract over the duration of the test in a similar way
to that presented in DeJong et al. (2003). On each
reversal of shear displacement, the sample
experiences a small dilation before steadily
contracting for the rest of the displacement, as the
soil skeleton rearranges to its lowest energy state.
The volumetric data recorded by the vertical LVDT
is shown in Figure 6.
Figure 7. Grain size distribution analysis for each test and a

clean, untested sample
Test 100-10-180-1 experienced the greatest total

contraction. The gradient of its vertical
displacement seems to fall close to zero as a
cumulative displacement of 40,000 mm is reached.
Figure 8 shows a frame of image data taken at the
end of the 100-10-180-1 test, with a cumulative
displacement of 40,000 mm. When compared with
the sample image in Figure 3, it can be seen that the
interface layer, shown at the bottom of the images,
is saturated with fine material. It could be
Figure 6. Plot of vertical displacement against cumulative
shear displacement
hypothesised that when the interface layer is fully
saturated with fine material, an ultimate contraction
Similar to the stress response, there seems to exist is reached, where a sliding mechanism occurs again,
a steady state where the rate of contraction of the and unbroken particles in the far-field soil do not
sample remains constant after a point. During contact the interface. The stress state at this point
testing, the samples all experienced a plastic may reflect a standard direct shear test, comprised
contraction due to crushing of the grains in the
interface layer, confirmed by an audible crushing
sound.
The crushing experienced can be witnessed in the
image data. For example, in test 100-10-000-1
where the gradient becomes almost constant, at a
cumulative shear displacement of 15000 mm and
1500 reversals, the stress ratio begins to increase Figure 8. Frame taken from test 100-10-180-1 at 40,000 mm
displacement
61
of the fine material. Further investigation into the experimental response is not caused by boundary
stress state in the interface layer when saturated with effects. As the sample cycles through a stroke of
fines, will need to take place. 10 mm, the leftmost and rightmost 10 mm of
It was observed during the upright tests that a interface plate boundaries of the interface plate are
period of sliding occurred, prior to the crushing. It only worn for half of the amount of travel, as soil
is believed that any sample will experience crushing particles travel over this area. Hence, it would be
after an amount of interface shearing. Although the expected for the boundaries to experience a smaller
increase in stress ratio and plastic contraction amount of wear than central areas.
occurred at smaller cumulative displacements for
the inverted tests, it should not be assumed that all 3.4. Micromechanical response
inverted tests will act in a similar manner. Future Using the image data, the micromechanisms that
testing will seek to clarify if this immediate occur during interface shearing at high cumulative
response is a function of orientation angle. displacements, can be identified. For the first time,
a distinct micromechanism has been identified for
3.3. Damage to interface interface shearing. After many shear reversals, a
After each test is conducted, surface fine material is random translation of a single particle within the
cleaned from the plates. An image showing the interface layer occurs, causing a cascading
plates of the one upright and one inverted test area rearrangement of particles in the remainder of the
is shown in Figure 9. The tested area is clearly interface layer, and in the far field soil. This
identified, with micro abrasions and ploughing wear mechanism has not been previously described, and
on the interface plate. Plastic flow occurs via third- is likened to popcorn cooking, as the translation of
body abrasion when a normal load is applied to an one particle causes the translation of many more.
abrasive particle with hardness 1.2× greater than the This popcorning can be easily identified within the
first (damaged) body (Hutchings and Shipway image data when viewed as a video, though is more
2017). As demonstrated in Section 2.3.1, the difficult to identify in discrete frames, and therefore
hardness of silica sand far exceeds the value given is omitted from this paper.
for the structural interface plate. When the silica The effect of inversion upon the fine material
particles are forced into the relatively soft steel generated at the interface, can more easily be shown
interface repeatedly, ploughing of troughs and by comparison of the upright and inverted tests after
abrasion occurs. Further study should be given to crushing has begun. Figure 10 displays a
the relative hardness of the interface, and what comparison of the samples from the 100-000-10-
effect that has on shear behaviour. 4000 and 100-180-10-4000 tests, after the crushing
mechanisms have begun, and near to the end of the
tests. The fine material in the upright test, Figure 10
(a), has formed a clear band of packed fine material,
with a peak quantity of fines in the centre of the
sample. This is to be expected as fines tend towards
the interface under the effect of gravity. In contrast,
in the inverted sample (b), fines have fallen
throughout the sample, coating particles in the far-
Figure 9. Image of damaged interface plate after tests field soil. However, a concentration does exist in the
100-10-000-1 (left), 100-180-10-2 (right) interface layer.
The absence of fine material in the far-field soil of
One could hypothesise that the increase in stress the upright sample, and the concentration of fines in
ratio occurs due to fine particles falling into the the interface layer of the inverted test, suggests that
abrasions on the interface plate. As the fine material, crushing only occurs in the interface layer. If
some of which has a diameter <63 µm, is worked crushing was to occur in the far-field soil, fines
into the abrasions, the fine soil skeleton transfers would be seen to coat the particles throughout the
more normal stress into shear stress. upright sample, with a greater concentration in the
The damage upon the interface plate is worse in interface layer; this does not occur. Further work
the middle of the sample area, note the darker band will be conducted to determine the origins of the
of wear in the middle of the sample area, caused by fine/crushed material.
deeper grooves that reflect less light. This
observation is key in validating that the
62
Interface
(a)– upright test
Interface
(b)– inverted test
Figure 10. Comparison of crushing mechanisms for inverted and upright testing
4 CONCLUSION Cavarretta, I., Coop, M., O’Sullivan, C., 2010. The

influence of particle characteristics on the behaviour
The experimental work presented in this study of coarse grained soils. Géotechnique 60, 413–423.
represents the first insight into interface shear
behaviour at large cumulative displacements in the Cavarretta, I., O’Sullivan, C., Coop, M.R., 2016. The
order of thousands of millimetres of travel and shear relevance of roundness to the crushing strength of
reversals. The existence of non-linear shear granular materials. Géotechnique 67, 301–312.
behaviour at large displacements is described and
analysed with respect to the volumetric and stress DeJong, J., White, D., Randolph, M., 2006. Microscale
responses of the test samples. Future work observation and modeling of soil-structure interface
conducted will seek to specifically identify the behavior using particle image velocimetry. Soils and
cause of the steep rise in stress ratio, as well as the Foundations 46, 15–28.
crushing of grains in the interface layer. DeJong, J.T., Randolph, M.F., White, D.J., 2003.
Interface load transfer degradation during cyclic
5 ACKNOWLEDGEMENTS loading: a microscale investigation. Soils and
This research is funded with an Engineering and foundations 43, 81–93.
Physical Sciences Research Council grant in
addition to funding from the Nottingham Centre for DeJong, J.T., Westgate, Z.J., 2009. Role of Initial State,
Material Properties, and Confinement Condition on
Geomechanics, University of Nottingham.
Local and Global Soil-Structure Interface Behavior.
With thanks to Mir Hashemi Afrapoli, Luke J. Geotech. Geoenviron. Eng. 135, 1646–1660.
Bedford, Steve Lawton, for their efforts to fabricate
and assemble the testing apparatus and control Evgin, E., Fakharian, K., 1996. Effect of stress paths on
systems. the behavior of sand–steel interfaces. Canadian
Geotechnical Journal 33, 191–199.
6 REFERENCES
Hashemi, M.A., Heron, C.M., 2017. Development of a
British Standards Institution, 2016. Methods of test for
soil-interface device for cyclic and large deformation
soils for civil engineering purposes. Part 1.
shearing, in: Proc. 19th ICSMGE.
63
Hutchings, I.M., Shipway, P., 2017. Tribology: friction Senetakis, K., Coop, M.R., Todisco, M.C., 2013. The
and wear of engineering materials. Elsevier Science & inter-particle coefficient of friction at the contacts of
Technology. London, UK Leighton Buzzard sand quartz minerals. Soils and
Foundations 53, 746–755.
MATLAB, 2019. MATLAB Documentation
9.7.0.1190202 (r2019a). The MathWorks Inc.,
Natick, Massachusetts.
64
Carboxymethylcellulose as a Newtonian viscous fluid for centrifuge modelling

Jarno Terwindt1, Wouter van der Star2, Dirk de Lange1
Geo-engineering, Deltares, the Netherlands
1
Subsurface and groundwater quality, Deltares, the Netherlands

2
Corresponding author: Jarno Terwindt (jarno.terwindt@deltares.nl)
ABSTRACT: In geotechnical centrifuge modelling, the unification of time-scaling factors is particularly

important for studies regarding coupling consolidation and dynamic phenomena, for example pile
installation problems and earthquake loading. A common method to achieve this unification is the use of a
substitute pore fluid, which has a viscosity which is larger than that of the prototype pore fluid (generally
water). Aqueous methyl celluloses have proven very valuable for this purpose, mainly because the viscosity
can be increased while the density remains relatively unaffected and the solution can be characterised as
near-Newtonian. In this paper we describe the use of a low molecular weight sodium
carboxymethylcellulose (designated lCMC) in 1% to 4% solutions as an alternative to methyl celluloses
through physical chemical, (density, temperature dependence) and geotechnical (permeability in sand and
friction angle) analyses and describe the experience with this pore fluid at the Delft Geotechnics centrifuge
over the past 20 years.
The lCMC was characterized as a near-Newtonian fluid up to the limit of the rheometer of 2000 s-1, for a
1% solution. The fluid was density varied from 1.004 to 1.007 kg/L at concentration from 1.5% to 3.5%.
Temperature dependence of viscosity was relevant (a change in dynamic viscosity of 25 mPa∙s to 20 mPa∙s
at a temperature change form 20°C to 25°C, for a 1.5% solution) and should thus be taken into account
during interpretation. Colouring with biodegradable dyes from the food industry (red E124, blue E132) at
concentrations of 10 g/L enabled tracking of flow patterns through regular film and photography in various
sand types with a negligible effect on density and viscosity. Prepared solutions were stable for over 180
days when stored at 10 °C. These characteristics are compared to other methyl cellulose solutions used in
geotechnical testing such as metholose. The characteristics of these pore fluids turn out to be similar to the
extent that they have been reported in literature. The study is concluded with a testing program for new pore
fluids and proposals for additional recording of fluid characteristics during experiments.
Keywords: Viscous Pore Fluid; Newtonian Behaviour; Centrifuge Modelling.
1 INTRODUCTION 2 VISCOUS PORE FLUID

The geotechnical centrifuge is a valuable modelling REQUIRMENTS
tool, as it allows cost-effective examination of real The requirements of an ideal substitute pore fluid
scale problems (Leung et al., 1994, Timkura et al., are, amongst others, discussed by Allard &
1998). However, it is well documented that not all Schenkeveld (1994). The primary requirement is
relevant quantities scale in the same manner. that the overall constitutive behaviour (e.g. stress
Garnier et al. (2007) provides an overview with strain behaviour, pore pressure generation and shear
respect to scaling factors. In coupled consolidation strength) at n-g testing must be the same as the
and dynamic phenomena, the unification of time- prototype soil and pore fluid at 1-g testing. When
scaling factors is particularly important. A common the prototype pore fluid is water, the substitute pore
method to achieve this is the use of a substitute pore fluid should have:
fluid, which has a viscosity of n-times the viscosity • a density close to water. This is important for
of the prototype pore fluid, where n refers to the correct reproduction of hydrostatic pore
geometrical scaling factor (i.e. applied g-level). pressure, effective stress and inertia and
Several substitute pore fluids have been used by seepage forces;
different studies, such as silicone oil (Ko, 1994), • the same rheological behaviour as water. The
proprietary fluids (Allard & Schenkeveld, 1994), dynamic viscosity should be constant for all
aqueous solutions of glycerol (Whitman et al., shear stress levels (i.e. Newtonian behaviour).
1981) or methylcellulose (Dewoolkar et al., 1999, This ensures that the soil-fluid interaction is
Adamis & Gopal, 2015). This paper presents the the same in the model as in the prototype;
chemical and physical properties and the application • the same compressibility as water. This
of a variation on the latter substitute pore fluid. ensures comparable behaviour between the
65
model and prototype with regard to the soil place to minimize health risk to personnel hampered
skeleton during (dynamic) loading and efficient operation.
unloading; Following this, an evaluation of possible solutions
• a dynamic viscosity proportionally n times started and in 1999. Deltares developed a pore fluid
greater than water, where n is the applied g- based on sodium carboxymethylcellulose (lCMC 1).
level. This allows the model soil permeability The selected product (AF0305, Akzo Nobel) has a
to be comparable to the prototype soil degree of substitution of 0.7-0.95 three available
permeability, following Darcy’s law. cellulose-groups in each cellulose subunit has been
The pore fluid must also be chemically polar in replaced by the functional group carboxymethyl
order to enable its use with silts and clays, in (Figure 1).
addition to sands. To the authors’ knowledge no
viscous pore fluids have been applied in silts of
clays in geotechnical centrifuge testing.
Apart from these primary requirements, there are
also practical considerations, as discussed by Allard
& Schenkeveld (1994), Kuttert et al., (1998) and
Dewoolkar et al. (1999). These are as follows:
• the substitute pore fluid should be available in
a range of viscosities to suit geotechnical
centrifuge testing. Typically, g-levels vary
between 10 g and 150 g. Therefore, dynamic
viscosities of 10 mPa·s to 150 mPa·s are of
interest;
• preparation of the model with the substitute
pore fluid should not take excessively more
time when compared to preparing a soil model
with water, although it is recognized that
preparation times will always increase to a Figure 1. Functional unit of cellulose with 3 locations (-R),
certain extent using a substitute pore fluid; where the -H end group is replaced by carboxymethyl.
• production times should be acceptable and
fluid properties between batches should be Other types of lCMC, with a higher chain length,
consistent; were also investigated but proved to exhibit non-
• fluid properties should not change Newtonian behaviour at low shear stress and were
significantly over time, particularly in- causing clogging of pore volumes. lCMC comes in
between fluid preparation and test execution; an odourless, off white, fee flowing white powder
and is widely used in the food industry as a
• the substitute pore fluid should be non-toxic,
thickening agent.
bio-degradable and inert towards equipment
Although this was -back in 1999- the first use of a
and sensors for safe handling and disposal.
carboxymethylcellulose, the use of modified
celluloses as a pore fluid had been coined already in
3 FLUID PROPERTIES 1998 by Kutter et al. through using methylcellulose
3.1. Selection of CMC (which is chemically similar). In the past years
The original Delft Geotechnics (Deltares’ another modified (methyl)cellulose has been
predecessor) pore fluid (Allard & Schenkeveld, heavily used: hydroxypropyl methylcellulose
1994) fulfilled the geotechnical requirements (HPMC).
described above and was successfully used in
several modelling studies (e.g. Allard et al., 1994). 3.2. Physical Properties
However, it was composed of various volatile and A number of different variations of lCMC solutions
flammable chemicals, and soon the need for a fluid were tested on basic physical properties. Different
which would less hazardous and bio-degradable concentrations were prepared by mixing lCMC in
became apparent. High costs were associated with hot water and agitating the mixture while permitting
the disposal of soil models and the procedures in the solution to cool. In this paper, the employed
1
We designate the used CMC in the remainder of this
paper as lCMC, as it is on the lower end of the viscosity-
application range due to a lower molecular weight.
66
definition of concentration for lCMC solution is dry In order to validate if the lCMC solutions behave
mass of lCMC over mass water. An elevated like a Newtonian fluid, four different concentrations
temperature prohibits the lCMC particles to were prepared and tested using a HAAKE MARS
dissolve in water, allowing optimal wetting of rheometer (Thermo Fisher, Karlsruhe, DE). The
particles. Upon cooling, the suspended fully wetted shear stress was measured with shear rates varying
particles dissolve into the aqueous solution. This is from approximately 1 s-1 to 2000 s-1, where
a common method for preparing methylcellulose equipment limitations so permitted. The results are
type solutions and is, amongst others, described by presented in Figure 4, and show that, although fluid
Dewoolkar et al. (1999) and Adamidis & Gopal behaviour is not fully Newtonian over the tested
(2015). shear rate range, it can be assumed to be Newtonian
Although temperature variations are typically in the range of interest for (dynamic) centrifuge
limited, say 20°C ± 2°C, during geotechnical testing. The range of dynamic centrifuge testing has
centrifuge testing, it is still desirable that adopted from Adamidis & Gopal (2015).
temperature related variations in model pore fluid Additionally, Figure 4 shows that a substitute pore
viscosity follow the same trend as the prototype fluid with a viscosity up to 300 mPa·s can easily be
pore fluid. In order to validate temperature stability prepared, thus allowing centrifuge testing up to
both the specific gravity and the viscosity were 300 g.
measured at different temperatures and
concentrations. Temperatures ranged from 5°C to
30°C and concentrations were varied from 0 %
(water) to 3.5 %. The specific gravity was measured
using a hydrometer and the viscosity was measured
using an Ubbelohde viscometer.
Figure 4. Dynamic viscosity at varying shear rate for various

concentrations. Dashed lines indicate measurement limits of
the rheometer. Exponential fits are applied to the data in order
to guide the eye.
The long-term stability of the solution was also

evaluated. During a period of 11 months, the
Figure 2. Specific gravity versus temperature for water (0.0% viscosity of a single 2.0 % solution stored at 10 °C
lCMC) and various concentrations of lCMC.
was periodically measured using an Ubbelohde
Figure 2 and Figure 3 show that the specific viscometer at 20 °C. Results are shown in Figure 5.
gravity and the viscosity decrease with temperature. The viscosity over almost 1 year was stable,
To allow for correct interpretation of results it is consistent with observations of Dewoolkar et al.
important to monitor the model temperature during (1999) for 1-3% solutions of hydroxypropyl methyl
centrifuge testing. cellulose (HPMC: Metolose 90SH-100).
Figure 3. Dynamic viscosity temperature for water (and Figure 5. Viscosity versus time for a 2.0% solution. The
various concentrations of lCMC. Exponential fits are applied viscosity was monitored for a period of about 11 months and
to the data in order to guide the eye. shows no significant change over time.
67
Figure 6 shows the effect of the lCMC- Table 1. Sand Colorants in Water, measured at 20°C.
concentration on the solution viscosity at different Parameter Unit RED BLUE
temperatures. The results vary for the same product Colourant - E124 E132
per batch differ in the order of about 10% (data not - Ponceau Indigo
Name
shown) and thus batch-by-batch adjustment of the 4R carmine
used lCMC-concentration may be necessary Concentration (g/L) 10 10
through addition or evaporation of water. Density at 20°C (kg/L) 1.005 1.003
Viscosity (mPa∙s) 1.053 1.048
3.3. Geotechnical Properties

Constant head permeability tests were conducted on
Baskarp B15 Sand (Baskarp sand properties are
summarized in Table 2) using a standard constant
head set up (ISO 17892-11:2004) and sand samples
saturated with water and samples saturated with
three different lCMC solutions.
Following Darcy’s law, the permeability should
increase linearly in a 1:1 relation with increasing
Figure 6. Viscosity versus concentration, at different pore fluid viscosity. Figure 8 shows the ratio of the
temperatures. Depending on the batch found viscosities may permeability of water saturated samples to the
vary slightly, in the order of about 10 %. Viscosity has been
determined using and Ubbelohde viscometer. Exponential fits
permeability of lCMC solution saturated samples.
are applied to the data in order to guide the eye. Results follow a 1:1 slope, thus satisfying the
scaling requirement.
Pigments have been added to lCMC fluid on
several occasions in order to show flow patterns. Table 2. Soil Properties of Baskarp B15 Sand.
Adding colour to the pore fluid facilitates Parameter Unit B15
visualization of ground water flows, as shown in Particle density, γs kN/m3 2.65
Figure 7. Such pigments should be well visible on Minimum void ratio, emin - 0.52
the sands concerned and change the physical Maximum void ratio, emax - 0.89
characteristics of the fluid as little as possible. The
Mean mass diameter, d50 mm 0.13
colours red and blue, as shown in Table 1, have been
added for visualization with the lCMC and showed Coefficient of uniformity, d60/d10 - 1.4
good visibility at long term (2 months) stability.
Figure 8. Ratio between vertical hydraulic conductivity of

water and lCMC viscous substitute pore fluid versus dynamic
viscosity. Three different concentrations were prepared, and
the hydraulic conductivity was determined at 1-g using a
standard constant head permeability test on Baskarp B15 sand.
The straight line indicates a 1:1 relation.
Figure 7. Example of visualization of ground water flow using
E132 (in water). Red dyed pore fluid is introduced at point A, Additionally, consolidated anisotropic drained
while an under-pressure is introduced at point B (Image F.M.
Schenkeveld, 2000). triaxial tests were conducted on Baskarp B15 sand.
The aim was to investigate any possible change in
friction angle due to the different pore fluid.
Samples were prepared in both loose and dense state
and saturated with water and lCMC solution.
68
Results are presented in Table 3 and show no water will start to flow out of the model and will be
significant difference in friction angle between the replaced by the viscous fluid. Viscosity
water saturated and lCMC solution saturated measurements of the fluid exiting the model can
samples. provide indications whether saturation with viscous
fluid is complete. Great care should be taken to
Table 3. Summarized Consolidated Drained Triaxial Test apply a small enough under pressure in order to
Results on Baskarp B15 Sand. Vertical and horizontal stress
was 300 kPa and 162.5 kPa, respectively. allow for homogeneous saturation and avoid
preferential flow path in the sand.
e μ ϕ
Pore Fluid
(-) (mPa∙s) (°) 5 APPLICATION IN GEOTECHNICAL
Water 0.57 1.0 43.7
CENTRIFUGE TESTING
Water (coloured) 0.57 1.0 42.8
The lCMC-based substitute pore fluid has been
CMC 3.5% 0.57 299.5 44.0
developed by Deltares in 1999 under a research
CMC 3.5% (coloured) 0.57 297.4 42.9 project and has been mainly applied in commercial
Water 0.71 1.0 40.0 projects of which the results cannot be disclosed. As
Water (coloured) 0.70 1.0 40.9 such no data was published. However, publicly
CMC 3.5% 0.71 299.5 40.0 available applications also exist.
CMC 3.5% (coloured) 0.70 297.4 40.1 Hölscher et al. (2012) and Nguyen (2019) applied
this pore fluid in several series of rapid pile load
4 SOIL MODEL PREPERATION tests and static pile load tests in the geotechnical
centrifuge at Deltares at 40 g. The rapid load tests
This section describes best practices regarding were carried out with several loading rates. Pore
lCMC. A batch of substitute pore fluid is prepared pressures were measured at different distances from
at the desired viscosity using guidelines on the pile tip. The tests confirm that a rapid load test
concentration presented in Figure 6. As described can overestimate static capacity due to pore water
earlier, the lCMC is added to de-aired water (heated pressure when testing piles in medium to fine sands.
to about 60°C) while being agitated. When all The results were compared with field measurement
lCMC particles are fully suspended, the solution is and it has been concluded that the centrifuge tests
allowed to cool to room temperature. This may be represent realistic behaviour.
achieved by starting at a high concentration and More recently, Askarinejad et al. (2017)
adding cold (5-10°C) water later as is shown for compared the lCMC based pore fluid with a
HPMC (Adamidis & Gopal, 2015). Additional glycerine solution and normal water in pore
tweaking is sometimes used in order to fine tune the pressure development of laterally loaded offshore
viscosity. When using food grade lCMC, insoluble piles. It was concluded, that although the problem
impurities are still present in the aqueous solution was assessed to be a drained situation, the density of
and prohibit infiltration of the lCMC solution into the pore fluid played a major role in the
the sand model due to clogging. Therefore, the development of pore pressure. It was observed that
lCMC solution requires to be filtered before in can it directly affects the state of the soil and the
be applied as a substitute pore fluid. Such filtering stiffness and lateral capacity of the pile.
may also reduce the encountered reduction in
permeability reported in methyl celluloses (Tobita, 6 COMPARISON OF PORE FLUIDS
2020). The authors’ experience is that glass-fibre
Roughly four different pore fluids have been used
filters with a pore opening of 0.5 µm work well.
for centrifuge modelling to properly scale the result:
After filtration, the model can be saturated with
silicon oil-water mixtures (Zeng et al., 1998), water-
the substitute pore fluid. The starting point for
silicon oil mixtures, xanthan as well as various
obtaining a sand model saturated with viscous fluid
substituted celluloses. Especially the latter is
is a fully water saturated sand model. A fully
showing behaviour corresponding with the
saturated sand model can be prepared using
requirements stated in the introduction. The effect
techniques described by Poel & Schenkeveld
of a pore fluid is however not only determined by
(1998). The phreatic level of the water saturated
the substituent of the cellulose (carboxymethyl in
sand model should be equal to the top of the sand
case of lCMC, and methyl in case of methyl
model. A volume of viscous fluid in excess of the
cellulose), but is also determined by the % of
amount required to replace the water in the pore
available spots which is substituted. In general, a
volume should now be placed on the surface. The
higher degree of substitution with oxygen
excess volume is required to avoid air entrapment.
containing substituent (e.g. carboxymethyl in lCMC
By applying an under pressure to the sand model the
69
Table 4. Overview of modified methyl celluloses as viscous fluids.
Abbre- Name Type Substituent(s) Viscosity of Degree of Source
viation (brand) 2% solution sub-
(mPa∙s) stitution 2
HPMC hydroxypropyl Metolose 1) hydroxyproxy 80-120 1) 47% Dewoolkar et al.
methyl cellulose 90SH- 2) methoxy 2) 6% 1999, Shin-Etsu
100 2018.
HPMC hydroxypropyl Methocel 1) hydroxyproxyl 40-60 1) 27-30% Adamidis et al.
methyl cellulose F50 2) methoxy 2) 4-7.5% 2015, Steward,
1998. Dupont
2019.
carboxymethyl Akucell carboxymethoxy 40-60 21-30% This study, Akzo
lCMC
cellulose (sodium) AF0305 2010
MC methyl cellulose Metolose methoxy 80-120 26-33% Tobita et al. 2008,
SM100 Shin-Etsu 2018.
and hydroxytmethyl in HPMC) will increase the The lCMC solution shows Newtonian behaviour in
viscosity because there will be more interaction the shear rate region of interest for dynamic
between individual lCMC chains through hydrogen centrifuge testing. Homogeneous fully saturated
bridges, which leads to a cross-linking soil samples containing the viscous pore fluid were
characteristic. prepared and recommendations for soil model
Furthermore, cellulose polymers are available in preparation are given in this article. Also, the
several chain lengths, which has a very strong effect influence of the viscous pore fluid on the shear
on viscosity. An overview of characteristics of pore strength of the soil model compared with water is
fluids used in centrifuge modelling (as far as can be not significant. Three cases of successful
inferred from the paper in combination with application of the pore fluid in dynamic
publicly available characteristics of the used geotechnical centrifuge testing are cited. The
products) is shown in Table 4. Deltares geotechnical centrifuge that is under
Table 4 shows that a full comparison of construction (Zwaan et al., 2020) will therefore
characteristics is complicated, as they are generally likely also use this pore fluid in many of its
not consistently described and supplier information experiments.
is needed to obtain all available information. We
propose therefore to systematically analyse and 8 ACKNOWLEDGEMENTS
describe pore fluids in order to aid the The authors would like to thank F.M. Schenkeveld
reproducibility of experiments in the future. In for his efforts on the development of the lCMC pore
addition to the characteristics for pore fluids fluid.
described above it is therefore recommended to also
describe: 9 REFERENCES
• chemical characteristics; Adamidis, O. and Madabhushi, G. 2015. On preparation
• degree of substitution; of viscous pore fluids for dynamic centrifuge
• average molecular mass; modelling. International Journal of Physical
• density and viscosity of a 2% solution at the Modelling in Geotechnics, 15(3): 141-149.
temperature employed.
Akzo Nobel. 2010. Product Specification Akucell
AF0305.
7 CONCLUSION
Deltares developed a viscous pore fluid for Allard, M.A. and Schenkeveld, F.M. 1994. The Delft
geotechnical centrifuge modelling based on sodium Geotechnics model pore fluid for centrifuge tests.
Centrifuge 94, 133-136.
carboxymethylcellulose. The effects of
concentration, temperature, ageing and shear rate on Allard, M.A., Andersen, K.H., and Hermstad, J. 1994,
these aqueous lCMC solutions were examined. The Centrifuge Model Tests of a Gravity Platform on
rheological behaviour is also investigated. Very Dense Sand; 1: Testing Technique and Results.
Viscosities suitable for centrifuge testing up to 300 In Proceedings of the 7th International Conference on
g proved to be achievable. The solutions are stable the Behaviour of Offshore Structures, Cambridge
over time and the change in characteristics due to Massachusetts, USA, July 1994, pp. 231-254.
temperature changes is limited, and similar to water.
2
Calculated on on a mass-basis
70
Askarinejad, A., Philia Boru Sitanggang, A., and Poel, J.T., & Schenkeveld, F.M. 1998. A preparation
Schenkeveld, F. 2017. Effect of pore fluid on the technique for very homogeneous sand models and
cyclic behavior of laterally loaded offshore piles CPT research. In Proceedings of the International
modelled in centrifuge. In Proceedings of the 19th Conference Centrifuge 98, Tokyo, Japan, 23-25
International Conference on Soil Mechanics and September 1998. Taylor & Francis Group, London,
Geotechnical Engineering, Seoul, South Korea, 17-21 pp. 149–154. Shin-Etsu. 2010. Metolose®. Metolose®
September 2017, pp. 897-900. DuPont, 2019. SR. Brochure.
MethocelTM Brochure.
Tobita, T. 2018. Permeability of sand with a
Dewoolkar, M., Ko, H., Stadler, A., and Astaneh, S. methylcellulose solution. In Proceedings of the
1999. A Substitute Pore Fluid for Seismic Centrifuge International Conference on Physical Modelling in
Modeling. Geotechnical Testing Journal, 22(3): 196- Geotechnics, London, UK, 17–20 July 2018. Taylor
210. & Francis Group, London, pp. 271–276.
Garnier, J., Gaudin, C., Springman, S.M., Culligan, P.J., Tobita, T. 2020. Variation of Permeability of Viscous
Goodings, D., Konig, D., Kutter, B., Phillips, R., Fluid During Liquefaction Model Testing. In Model
Randolph M.F. and Thorel. L. 2007. International Tests and Numerical Simulations of Liquefaction and
Journal of Physical Modelling in Geotechnics, 7(3): Lateral Spreading. Kutter , B. Manzari, M.T., Zeghal,
pp 01-23. M. pp. 649-652.
Hölscher, P., van Tol., A.F., Huy, N.Q. 2012. Rapid pile Whitman, R. V., Lambe, P. C., and Kutter, B. L. 1981.
load tests in the geotechnical centrifuge. Soils and Initial Results from a Stacked Ring Apparatus for
Foundations, 52(6): 1102-1117. Simulation of a Soil Profile. In Proceedings of the
International Conference on Recent Advances in
ISO/TS 17892-11:2004(en), Geotechnical investigation Geotechnical Earthquake Engineering and Soil
and testing — Laboratory testing of soil — Part 11: Dynamics, Missouri-Rolla, US, 26 April – 3 May
Determination of permeability by constant and falling 1981. pp. 1105–1110.
head.
Zeng, X., Wu J., and Young B. 1998. Influence of viscous
Ko, H. Y. 1994. Modeling Seismic Problems in fluids on properties of sand. Geotechnical Testing
Centrifuges. Centrifuge 94, pp. 03–12. Journal 21(1): 45-51.
Kutter, B.L., Stewart, D.P. and Chen, Y.R. 1998. Zwaan, R., Terwindt, J., de Lange., D.A., Bezuijen, A.
Experience with the use of methylcellulose as a 2020. A New Geotechnical Centrifuge at Deltares,
viscous pore fluid in centrifuge models. Geotechnical Delft, the Netherlands. In Proceedings of the 4th
Testing Journal, 21(4): 65-369. European Conference on Physical Modelling in
Geotechnics, Luleå, Sweden, 15–17 March 2020.
Nguyen, T.C. 2017. Modelling of pile load tests in
granular soils: Loading rate effects. Doctoral thesis.
Delft University of Technology.
71
2. Physical modelling techniques and facilities
A new geotechnical centrifuge at Deltares, Delft, the Netherlands

Rob Zwaan, Jarno Terwindt, Dirk de Lange, Adam Bezuijen
Geo-engineering, Infra and Structures, Deltares, The Netherlands
Corresponding author: Rob Zwaan (rob.zwaan@deltares.nl)
ABSTRACT: A new geotechnical centrifuge has been commissioned by Deltares, Delft, The Netherlands,
as a replacement of the previous facility, which has expended its serviceable life. The new centrifuge is an
Actidyn C72-31 beam type centrifuge, and has a 260 g-tonne capacity and a platform radius of 5.0 m. The
platform can house test set-ups with dimensions up to 1.2 m × 1.2 m × 1.2 m (length × width × height). The
data-acquisition system is supplied by HBM and has 40 channels with a sampling rate of up to 100 kHz.
Two high speed cameras and four monitoring cameras are available for imaging. A 4-axis robotic system
allows in-flight excavation and installation operations. The centrifuge is fully prepared for later installation
of an earthquake simulator. Supporting facilities consist of a clay mixer and a consolidation frame. This
paper details the choices made leading to this centrifuge, the specifications of the centrifuge and supporting
facilities.
Keywords: GeoCentrifuge; Deltares; Geotechnical; Centrifuge; C72.
1 INTRODUCTION tunnel interaction when the first TBM tunnels were

From 1987 until 2017 Deltares has been operating a constructed in The Netherlands. It was used to
geotechnical centrifuge for research on soil check the influence of ship impact on bridge
mechanics, the interaction between soil and foundations in Hongkong and Incheon. It has been
structures and other related research topics. In 2017 used to study pile capacity, rapid pile load testing
this centrifuge with a beam of 6 m and a capacity of and pile group effects. Furthermore, numerous
1050 g-tonne was closed after relocation of the offshore projects were performed (dragging of
complete soil mechanical department of Deltares to anchors, jack-up-seabed interaction etc.).
a new location in Delft, the Netherlands. This After the relocation of the soil mechanical
relocation was part of an infrastructural department to the new location the centrifuge
reorganization in which all the departments of became a “remote” location. This fact combined
Deltares are located together as much as possible on with increasing difficulties to find spare parts for the
one location in order to combine forces on the use almost thirty-year-old machine, an increasing
of facilities, experts and knowledge. amount of (ir)regular maintenance work and the
Before the forming of Deltares in 2008 the partially outdated test control, data- and video
centrifuge was part of “Delft Geotechnics” and from acquisition systems has led to the decision to close
1999 “GeoDelft”. The four walled shape of the the centrifuge in April 2017. The centrifuge was
centrifuge basket, the large size of the basket (2 m dismantled, the large underground centrifuge
× 1 m × 1 m, l × w × h) and payload capacity of the chamber became a rain water storage and on top a
centrifuge made it possible to create very large soil new seventeen floors high building was constructed
samples inside the basket itself without using a to accommodate students from the Delft University
strongbox. It was also possible to create relatively of Technology.
big test set-ups with large strongboxes and/or an In 2010 the first plans to acquire a new centrifuge
extensive amount of utilities for use during were discussed and up until 2017 these plans were
centrifuge flights. elaborated to a detailed list of demands for the new
The centrifuge contributed to a large amount of centrifuge. This list was based on the thirty years of
geotechnical research projects. During these experience with the old centrifuge combined with
projects the centrifuge was one of the steps between the possibilities needed for the expanding scope of
small scale tests up to field scale tests and had an work since GeoDelft became a part of Deltares.
important role in optimizing and the reducing of With the detailed list of demands a public tender
risks in designs where the interaction between soil was started after which it was decided to choose the
and structures was of importance. It contributed for C72-3 beam centrifuge with a capacity of 260 g-
example to validate improved calculation methods tonne manufactured by Actidyn Systems.
for dike stability and the fact finding when a canal The funding for the new centrifuge is split
dike failed near the village of Wilnis. It helped to between the Ministry of Economic Affairs of the
understand the tunneling processes and the pile- Netherlands and Deltares. This funding also
75
included the construction of the new centrifuge dimensions of 1.2 m × 1.2 m × 1.2 m designed for a
building, new data- and video acquisition system, payload of 2600 kg at 100 g or 1000 kg at 150 g.
new test control system, controlled tool system, soil
consolidation frame, hydraulic power pack and a
new clay mixer with a 500 l capacity.
2 GENERAL CONSIDERATIONS
The previous in-house designed geotechnical
centrifuge was one of the biggest geotechnical
centrifuges worldwide. However, in daily use that
appeared not only advantageous. The centrifuge
could only operate under vacuum, which made
starting and stopping a relatively long-time
procedure since the vacuum had to be created before
spinning-up and had to be released before stopping.
For this vacuum and for safety, the centrifuge was
built in a cellar which decreased the accessibility.
Analysing the projects that were run, it appeared
that the full capacity in payload nor g-level was
hardly ever used. Therefore, it was decided that the Figure 1. The new C72-3 centrifuge at Deltares.
new centrifuge could be smaller than the old one but
should allow easier access and a shorter turn-around To maintain the balance of the centrifuge an
time for one test. Furthermore, maintenance should automatic unbalance detection system is integrated
be simpler and less costly. This is realized by using consisting of four detection strain gauges in the
‘standard equipment’ instead of an in-house design, floor connection of the pedestal, a counterweight
a centrifuge on the same floor level as the Soil with a mass of 14,896 kg and two hydraulically
Mechanical Laboratory and a centrifuge that can run actuated small weights of 367 kg each inside the two
at atmospheric pressure. centrifuge beams for fine adjusting purposes. The
Since it can be expected that soft soil (peat and hydraulic system and the controlling PLC
clay) will be used quite frequently, a stand-alone (Programmable Logic Controller), both needed for
mode during re-consolidation of that soil was the automatic balance system, are completely
essential. integrated on board of the centrifuge.
When the centrifuge starts, up to a level of 10 g,
3 THE C72-3 CENTRIFUGE the position of the large counterweight will be
The chosen C72-3 is a modified version of the adjusted until the desired position has been reached
previously manufactured C72 centrifuges. The C72- for coarse balancing. Then when the centrifuge g-
3 is for a large part comparable with the centrifuges level increases to the level needed for the centrifuge
delivered to the Centre for Offshore Foundation experiments, the small counterweights are deployed
systems at the University of Western Australia for fine balancing purposes. The small
(Gaudin et al., 2018) and the Korea Advanced counterweights also maintain the balance of the
Institute of Science and Technology (Kim et al., centrifuge during flight, when during an experiment
2006). The main difference is that the C72-3 is the weight or the balance of the payload changes.
developed to make future installation of the Actidyn On top of the centrifuge beams, a user platform is
Earthquake simulator possible. For this purpose, the constructed that provides easy access to the two 19”
C72-3 is equipped with an earthquake simulator racks that provide housing for on-board equipment.
prepared platform, extra high-pressure hydraulic Besides the centrifuge related equipment such as the
sliprings to power the simulator and the whole unbalance control PLC, (there is enough space for
centrifuge construction is re-designed to a higher additional equipment such as the test control
payload which increased the capacity to 260 g-tonne system, NAS (network attached storage) units,
instead of 240 g-tonne. patch panels and power supplies. The 19” racks are
The C72-3 has opted for a gearbox to connect the situated above the heart of the rotation axis to
electric motors with the rotor drive shaft instead of minimize the g-level on the equipment within the
a drive-belt. This choice is made because of the racks.
higher payload capacity and larger g-level range. The centrifuge is driven by two frequency
The platform radius is 5 m with platform controlled electric motors that drive the gearbox
which is coupled to the central axis of the
76
centrifuge. The total installed main power supply
for driving the centrifuge is 270 kW. In addition to
this, 70 kW is reserved for on-board systems and
experiment power supply and 170 kW for the
hydraulic power unit to charge the earthquake
simulator and drive hydraulic devices during
experiments. To deliver the power needed for the
complete centrifuge and surrounding utilities, a new
separate high voltage line from the public energy
grid is laid out to a designated transformer in the
centrifuge building.
4 THE CENTRIFUGE BUILDING

4.1. General Situation
For the new centrifuge a new building had to be
created on the Deltares location at the
Figure 2. Layout of the centrifuge building foundation.
Boussinesqweg in Delft.
To get a good connection with the already existing
The centrifuge chamber wall and ceiling are
Soil Mechanical Laboratory and the 1g geotechnical
constructed on top of this floor, the steel
model facilities, it was decided to attach the new
reinforcement of the floor concrete is thoroughly
centrifuge building to this laboratory. In this way
connected to the four-layered steel reinforcement of
facilities, staff and materials can easily be shared
the 40 cm thick wall and ceiling. Throughout the
since the centrifuge is a logical addition to the
whole foundation, floor, wall and ceiling, the steel
facilities of this lab. It also decreased the amount of
reinforcement was designed to the maximum
space needed for the new building since test set-up
possible density to ensure a firm centrifuge base and
construction and model preparation for the
chamber that can resist a fatal unbalance. As a
centrifuge can be combined with the already
further precaution a highly reinforced wall of 40 cm
existing activities in the lab. In this way only a
thick is placed outside the steel centrifuge chamber
centrifuge chamber with a second floor for the air
door. If the door fails to resist an impact due to fatal
conditioning unit and hydraulic power pack, a
unbalance, this wall is the final measure needed in
control chamber and a connecting hallway had to be
this safety chain.
built. Although this seems not too comprehensive, it
The rest of the centrifuge building will be a
still was a challenge to construct this on the soft
normal steel structure with construction-, glass- and
soils of the Netherlands. For the foundation of this
isolation panes.
additional building of 255 m2 it was necessary to
In Figure 3 a rectangular hatch is visible on top of
drive 77 piles (see Figure 2) into the ground to a
the centrifuge building. This hatch is also present in
depth of 19 metres. The huge amount of piles was
the centrifuge chamber ceiling and enables the
needed on this location to keep the building
hoisting of the centrifuge in- and out of the chamber.
stabilized, even in case of a fatal unbalance of the
centrifuge (loss of payload).
The ground floor of the centrifuge building is
attached to the floor of the Soil Mechanical
Laboratory and is wide enough to drive a forklift in
to the centrifuge chamber when necessary. The only
interruptions in the floor surface from the Soil
Mechanical Laboratory to the centrifuge room are
two narrow expansion and/or vibration joints of 2
cm each. These joints are filled with flexible
synthetic spacers to maintain a flat surface. This flat
floor surface from the preparation area to the
centrifuge chamber is for instance essential for the Figure 3. Artist-impression of the new centrifuge building.
shockless transport of vulnerable soil samples with
low density sand beds.
77
4.2. Control chamber conditioning unit. This circulating airflow also
The centrifuge control chamber is situated in the provides cooling for the hydraulic power unit and
rounded end of the hallway opposite of the the centrifuge power cabinet.
centrifuge chamber. This ensures a safe distance The utility room is connected to the second floor
from the centrifuge chamber while having a clear of the Soil Mechanical Laboratory with a small
view on the hallway leading to the centrifuge hallway enabling easy access when necessary.
chamber. The control chamber houses the
centrifuge operating desk, the test control desk 5 CENTRIFUGE CONNECTIONS
(including video and data acquisition), 19” racks for To connect the centrifuge with the surrounding
NAS, test control modules and other relevant utilities and mains power the centrifuge is equipped
equipment and additional desks for project leaders with rotary joints. On top of the two 19” racks a
and/or clients that want to attend during centrifuge Fiber Optic Rotary Joint is installed to connect the
tests. on-board network with the network in the
The centrifuge operating desk with a stand-alone centrifuge. This provides two 10 Gbps fiber optic
computer and designated software enables the connections from the two on-board ethernet
operator to control and monitor the centrifuge. switches to the two ethernet switches in the control
Parameters such as desired g-level and rate of room. From these Juniper EX2300-C switches cat6
acceleration to reach this g-level can be controlled. ethernet cables are used to create the necessary
The real-time monitoring and registration of network connections to further equipment on both
centrifuge parameters such as temperature and sides. This network is used for data- and video
unbalance are also integrated in this system. This acquisition, centrifuge control purposes, the test
system also incorporates safety systems such as an control system, the 4-D robot and other equipment
automatic centrifuge stop when unbalance reaches that requires network connections. From the control
threshold limits or when doors providing access to room switch the centrifuge network is also
the centrifuge chamber or utility room are opened. connected to the Deltares network through fiber
LCD screens display video images of the four safety optic cables to provide remote control and
cameras in the centrifuge building provide real-time monitoring possibilities and access to the Deltares
sight on what is happening in the several areas. data storage building.
The test control desk houses everything that is On top of the 19” racks also signal slip rings are
necessary to perform a centrifuge experiment such available for copper connection of 3 BNC fitted
as real-time controlling of valves, hydraulic video lines and 32 signal lines each capable of 150
actuators, lighting, etc. during flight. The Vdc/1 A. The signal lines are used to copper-
controlling and monitoring of the data- and video connect I/O ports of the on-board test control
acquisition providing real-time test information are systems with the control room.
also integrated in the test control desk. The test The power sliprings and fluid rotary joints are
control desk and the centrifuge operating desk are mounted below the central axis of the centrifuge.
situated next to each other. The power sliprings provide mains power to the on-
board equipment and experiments in the centrifuge
4.3. The utility room basket up to 220 V/ 100 A. The fluid rotary joints
The centrifuge building is partially equipped with a provide connection of two hydraulic lines up to 200
second floor above the centrifuge chamber. This bar, one hydraulic line up to 300 bar, pressurized air
houses the utility room with the centrifuge power up to 10 bar, water and de-aired water up to 20 bar
cabinets, the climate installation for the centrifuge and vacuum down to -0.8 bar. The fact that 3
chamber and the hydraulic pressure unit. This utility hydraulic lines are installed is caused by the
room is part of the air conditioning loop that keeps decision to make the centrifuge ready for possible
the temperature in the centrifuge chamber below future installation of an earthquake simulator
30 °C. The air conditioning loop consists of the
utility room, the centrifuge chamber and large 6 CENTRIFUGE EQUIPMENT
airducts in the floor and wall of the centrifuge The centrifuge is equipped with several systems
chamber. This creates a closed loop system in which necessary to perform experiments. These systems
the air is conditioned and recirculated to ensure are partially designed and manufactured by Deltares
proper temperature conditions during centrifuge and partially bought from third parties. The systems
flights. The air flows from the conditioning unit into are designed and purchased based on 30 years of
the utility room, then down through a square hole of experience with the previous centrifuge and adapted
2 m × 2 m in the ground into the centrifuge chamber to modern standards.
and then through the airducts back to the
78
6.1. Test control system eight MX410B 4-channel high-dynamic universal
The test control system is designed to perform any amplifiers, one MX879B multi-I/O module and 1
remote action that could be needed during CX27C ethernet gateway. For flexible purposes
centrifuge experiments. The system is completely during centrifuge testing two additional MX840B
designed and built by Deltares and uses and one MX879B modules are available. The
CompactRIO hardware and software from National MX410B can handle data rates up to 100 kS/s
Instruments as heart of the system. The (maximum 38 kHz signal bandwidth) and the
CompactRIO system consists of a controller with a MX840B can handle 40kS/s (maximum 7.2 kHz
processor and user-programmable FPGA (field- signal bandwidth). These sample rates are expected
programmable gate array) that is equipped with to be more than enough, even for possible future
conditioned I/O modules. These modules provide a earthquake simulator experiments. The MX879B
wide range of functions and come in a rugged module provides 8 real-time math possibilities,
industrial form suitable for use in harsh output of calculated input signals, signal generator
environments. features and 32 digital I/O ports. The CX27C
For the centrifuge two CompactRIO systems are module provides 1 Gb/s transmission and PTP
designed and coupled over an ethernet connection (Precision Time Protocol) functionality for the
with each other. One system is built in the 19” racks modules in the backplane.
of the centrifuge and the other one in the centrifuge The systems are connected to each other and the
test basket. Both systems are fitted with a wide control room through ethernet cables and the fiber
range of I/O ports ensuring flexible use of this optic rotary joint. The modular system ensures
system to enable centrifuge testing. Partially the I/O flexible addition of extra modules for data-
ports from the on-board systems are copper wired acquisition purposes and easy maintenance or repair
over the signal slip rings to terminations in the 19” work. The CatMan AP software ensures easy set-up
rack in the centrifuge control room. These wires can of data-acquisition sessions, flexible real-time
be used to exchange signals between the centrifuge monitoring during tests and extensive data-
and the control room. This could occur when for processing possibilities.
instance read-out equipment that is not able to
withstand the g-level in the centrifuge needs to be 6.3. Video acquisition system
hard-wired with a test running in the centrifuge. Since computing capacity and processing speed
The software user interface enables full control of have developed strongly in the past years more
the whole system from the control room. This possibilities for video-based data gathering became
system will be used to switch valves, control available. This now has become a strong addition on
lighting, steer plungers, provide and switch power, the conventional data-acquisition and increased the
transfer digital and analogue I/O signals between need for images with better resolution, high-
the basket, the 19” rack and the control room. definition video and high-speed video. For these
The versatility of the system enables easy purposes four 4K color cameras and two high-speed
modifications if other functions that are yet color cameras have been added to the centrifuge
unforeseen are needed in the future. equipment. The 4K cameras are manufactured by
Flir Systems Inc. and are of the type ORYX 31Mpix
6.2. Data acquisition system with a resolution of 31 MP (megapixel) at 26 fps
For the centrifuge data acquisition system HBM frame rate. These cameras can be monitored real-
QuantumX modules are chosen in combination with time on the test control desk in the control room and
CatMan AP software. After this decision, the Soil the images are also stored on a NAS in the on-board
Mechanical Laboratory also switched to this system 19” racks for later analysis purposes. When
since the centrifuge is part of the facilities in this lab performing dynamic centrifuge experiments, it
and therefore it is logical to choose one system to occurs that events in centrifuge experiments happen
align the data-acquisition process, output files and extremely fast. To capture these events high-speed
sensor connectors. The data-acquisition hardware is cameras are essential. Two cameras manufactured
divided into a system in the control room and a by AOS Technologies AG of the type L-VIT 2500
system in the centrifuge basket. are available to capture images with a resolution of
The control room system consists of a BPX002 2 MP at 2500 fps frame rate for a period of 4 s.
19” backplane that houses two MX840B 8-channel These images cause a data stream that is way too
universal amplifier units, one MX879B multi-I/O large to transmit real-time to the control room, so
module and one CX27C ethernet gateway. the storage of these images is done in NAS units in
The system on board of the centrifuge basket the on-board 19” racks of the centrifuge. After the
consists of a BPX002 19” backplane that houses
79
experiment is finished the images are downloaded The tool head of the robot can pick up tools that
to the control room NAS for further analysis. are fitted with a universal baseplate that enables
Besides the cameras for video acquisition, the throughput between the ACTS and the tool of two
centrifuge has two additional fixed on-board hydraulic ports, two pneumatic or water ports and
cameras with a resolution of 5 MP and a frame rate power and data lines. This makes it possible to
of 26 fps to monitor the centrifuge itself. These develop, build and use specific tools that can
cameras are mounted on the centrifuge beam near perform a wide array of actions using these provided
the hinges. One camera overlooks the whole utilities.
centrifuge basket and the other camera is pointed at The ACTS is controlled by a Rockwell
the 19” racks and the platform on-board. The Automation RSLogix 5000 PLC that is situated in
images can be monitored in the control room on a the on-board 19” racks of the centrifuge and a
separate screen. dedicated PC on the test control desk in the control
room.
6.4. The 4-D Controlled Tool System The software of the ACTS gives full control on
The new Deltares centrifuge has the possibility of the functions. It is possible to upload the three-
using hydraulic or electric actuators as in the dimensional drawing of the centrifuge test set-up
previous centrifuge. In the old centrifuge, the and let the software calculate the most efficient
actuators had to be fixed on the locations where routing to the different points in the model where
actions were required: actions at different locations actions are required during flight while avoiding
required several actuators. For the new centrifuge, collisions with the model using no-go zones.
the 4-D ACTS (Actidyn Controlled Tool System) is The robot was initially designed, built and used at
acquired. This system provides actuator and tool IFSTTAR (former LCPC, Derkx et al., 1988), and
flexibility during flight up to a g-level of 100 g. is now identical to that used at the Centre for
Offshore Foundation systems at the University of
Western Australia (Gaudin et al., 2018) and the
Korea Advanced Institute of Science and
Technology (Kim et al., 2006).
7 UTILITIES
For the new centrifuge, several utilities are available
to facilitate soil model preparation and testing.
7.1. Strongboxes and consolidation cells

The strongboxes are structurally reinforced to
Figure 4. The 4-D controlled tool system.
minimize mechanical deformation during flight and
can also be used as consolidation cell using the
The ACTS creates the possibility to use 4 different consolidation frame or during centrifuge flight.
tools on any location covering almost the whole The platform of the C72 centrifuge is 1.2 m ×
surface of the centrifuge basket during flight. The 1.2 m, which is large enough to use the rectangular
system is placed on brackets on the centrifuge and cylindrical strongboxes that were already
container side walls at a height of 1.2 m from the available from the previous centrifuge. The
floor of the basket. cylindrical cells are available in 30 cm, 60 cm and
The system is fitted with two perpendicular 90 cm internal diameter and can be stacked up to the
electric displacement axes in a plane parallel to the desired height. The rectangular strongbox can hold
bottom of the centrifuge basket and a dual electric a model of 80 cm × 20 cm × 40 cm (l × w × h) and
axis of displacement and/or rotation perpendicular has one transparent wall. The strongboxes are fitted
to the previous mentioned plane. The axis can be with a drainage system in the bottom and have feed-
used to maneuver tools and are able to produce a through options in the bottom and walls for water
force of up to 5 kN, a torque of up to 5 N∙m and a and/or air hoses, these can also be used as entry for
speed of up to 50 mm/s with a positioning accuracy sensor cables.
within 1 mm on each axis and 1° on the rotation axis.
The tool holder can move along all four axes
7.2. Clay slurry mixer
simultaneously and can pick up or place back the
different available tools during flight enabling On the previous centrifuge location, a Nauta clay
flexible use of different functionalities. slurry mixer with a capacity of 200 l was used to
create clay slurry. Since this mixer slowly started to
80
show more and more signs of wear, the costs for 7.3. Consolidation frame
moving the mixer to the new location should be To consolidate the clay slurries from the clay mixer
considerable and the fact that the capacity was lower an Actidyn soil consolidation frame was acquired.
than desired, it was decided to acquire a new clay This frame consists of a highly reinforced and stiff
mixer. frame, a hydraulic actuator with a stroke of 1.5 m, a
hydraulic power supply up to 15 MPa and a control
system. The frame can be used for uniform
consolidation or a depth variable consolidation
based on the hydraulic gradient method.
During consolidation, parameters such as
pressure, flow, temperature, are independently
controlled and monitored. The control system also
incorporates safety circuitry that are essential to the
safe operation of the system. After the user has
programmed the desired consolidation path, the
system operates fully automatic until consolidation
is finished.
Besides the option of using this consolidation
frame, it is also possible to consolidate samples
during flight using the increased g-force of the
centrifuge.
Figure 5. The Heilig 500 l clay mixer.
The new clay mixer is manufactured by Heilig BV

from Heerhugowaard (NL) and is delivered
completely with the steel construction and platform
giving access to the top of the mixer. The mixer has
a work volume of 500 l, can handle grains up to
3 mm and can be vacuumed down to 50 mbar(a) in
order to create a slurry that is free of air bubbles.
The conical mixer has a sanitary open mixing
screw to avoid lump formation while mixing and is
provided with a CIP (clean-in-place) system that
ensures easy and thorough cleaning after use. The
outlet of the mixer on the bottom is provided with a
ball valve and a flange of 125 cm diameter. Below
this flange, a table with a hydraulic jack can be Figure 6. The Actidyn consolidation frame under test.
placed. A clay consolidation cell will be placed on
top of this table. The hydraulic jack function pushes 7.4. Additional equipment
the consolidation cell firmly to the flange ensuring For the centrifuge experiments a wide array of
an airtight connection. After this, the consolidation additional equipment is available. Several hydraulic
cell will be vacuumed to the same pressure level as 50 kN hydraulic actuators with strokes of 10 cm up
inside the clay mixer. In this way the air free slurry to 70 cm, hydraulic and electric pumps, pneumatic
from the clay mixer will flow air free into the and electric valves and penetrometers in several
consolidation cell when the ball valve is opened. diameters, lengths and fitted with or without T-bar
Then the consolidation cell can slowly be brought and/or pore pressure transducers. Besides this an
to atmospheric pressure again and installed into the extensive range of sensors is available such as soil
consolidation frame. pressure transducers, displacement sensors (LVDT
and laser), temperature sensors and force
transducers.
81
8 ACKNOWLEDGEMENTS Gaudin, C., O’Loughlin, C.D., Breen, J. 2018. A new 240
g-tonne geotechnical centrifuge at the University of
The realization of the new geotechnical centrifuge Western Australia. In Proceedings of the International
is made possible with funding from the Dutch Conference on Physical Modelling in Geotechnics,
Ministry of Economic Affairs and Deltares. The London, UK, 17–20 July 2018. Taylor & Francis
effort of the management and the Department of Group, London, pp. 501-506.
Real Estate & Construction Matters of Deltares, the
members of the Centrifuge Construction Team and Kim, D.S., Cho, G.C., Kim, N.R. 2006. Development of
the perseverance and enthusiastic commitment of all KOCED geotechnical centrifuge facility at KAIST. In
the people involved in this, is gratefully Proceedings of the 9th International Conference on
acknowledged. Physical Modelling in Geotechnics, Hong-Kong, pp.
147–150.
9 REFERENCES
Derkx, F., Merliot, E., Cottineau, L.M., Garnier J. 1998.
On-board remote-controlled centrifuge robot.
Centrifuge 98, pp. 97–102.
82
A Novel Small-Scale Model Testing Device for Laterally Loaded Piles

Hendrik Sturm1, Andrea Støren2
Norwegian Geotechnical Institute (NGI), Oslo, Norway
1
Norconsult, Sandvika, Norway

2
Corresponding author: Hendrik Sturm (hendrik.sturm@ngi.no)
ABSTRACT: In 2014, NGI and BP America Inc. developed and built a novel model testing device for
measuring the response of cyclic lateral loaded piles. Objective of this device is to assess site-specific cyclic
p-y backbone curves on intact soil specimens under in-situ stress and density conditions. Primary application
of these curves is in structural fatigue analysis. For that purpose, the model pile is subjected to sequences
with symmetric cyclic displacement amplitudes, where the load level is well below the pile capacity limit.
The performance of the p-y apparatus, as the device is called, has been demonstrated by means of an
extensive testing program on different soil types. The results were compared with centrifuge test data. It
was found that the measurements made in the p-y apparatus agreed very well with both the centrifuge test
results and the numerical framework, which was proposed some years prior to the development of the
device. Despite the simplicity of the cyclic load history used in these tests, the actuation system is capable
to apply more complex cyclic load histories allowing both asymmetric loading, force-controlled testing, and
even random load histories. This enables to study the pile response under general monotonic and cyclic
loading. An important design aspect, in particular for renewable offshore energy structures, is the
Serviceability Limit State (SLS). The general suitability of the p-y apparatus for that application case was
investigated recently and the plausibility of the measurement was checked. This article introduces this novel
model testing device, shows the performance for its original purpose, and presents preliminary results of the
recently conducted more complex cyclic load histories.
Keywords: 1G Pile Model Testing; Cyclic P-Y Curves; Fatigue; Serviceability.
1 INTRODUCTION respond with a cyclic resistance amplitude Δp,

Soil structure interaction has become in the recent which gradually decreases with number of applied
years increasingly important in offshore load cycles N. The existence of the cyclic attractor
geotechnical engineering, particularly with the rise postulates that Δp will approach a constant value;
of offshore wind energy turbines. But also, the denoted in the BP design framework steady-state.
development of very deep and remote oil and gas At steady state can be calculated a secant stiffness
reservoirs demands more advanced design viz.
approaches to reduce the risk of failures.
Foundation capacity requires information of only Δ𝑝𝑝
the soil strength, while strains are only relevant for 𝐾𝐾𝑆𝑆𝑆𝑆𝑆𝑆.,𝑆𝑆𝑆𝑆 = (1)
Δy
compatibility reasons at failure of the soil body.
However, soil-structure interaction design requires A cyclic p-y backbone curve can be constructed
information of the complete stress-strain response from the secant stiffness values for different
of the soil. Even more complicated is the description displacement amplitudes.
of the soil response under cyclic loading, as it This framework has been developed based on a
depends on the load amplitude and mean value, load series of centrifuge tests. The results of these tests
frequency, duration and drainage conditions. were normalized with the respective static shear
In order to cope with this complexity in the strength of the intact specimen. This allows to apply
design, dedicated engineering models are required. the framework to all type of soils, given that the
One such model has been developed by BP for the static shear strength is known. However, since only
structural well conductor fatigue design (Zakeri et a limited number of soil types have been tested so
al., 2019). The model proposes a framework for far, an application in real projects requires
cyclic p-y backbone curves. The basic assumption laboratory testing to increase confidence in the
is the existence of a cyclic attractor of the secant framework. For that purpose, NGI developed the
stiffness for a given symmetric and constant so-called p-y apparatus shown in Figure 1 (Zakeri
displacement amplitude. A pile segment subjected et al., 2017).
to a constant cyclic displacement amplitude Δy will
83
2 THE P-Y APPARATUS AND TYPICAL stiffness for low displacement amplitudes.
TEST RESULTS However, a closer inspection of the trend lines in
The underlying idea of this apparatus is to assess the this area reveal an unrealistic shape of the p-y curve
steady state secant stiffness on intact specimens where the pressure does not start from zero at zero
taken with standard piston samplers. That means displacement, but rather approaches a constant
that the soil specimen has an outside diameter of value. The measurements, however, predict a
68mm and a length of 100mm or 200mm, consistent and realistic p-y curve.
respectively. The specimen is fully confined by
means of a rigid cylinder and two special designed
endcaps. Through these endcaps is installed a model
pile with D=10mm in diameter, which goes through
the whole specimen.
Figure 2. Normalized secant stiffness and trendlines

Figure 1. p-y apparatus including 2 different sized specimen developed based on a comprehensive series of centrifuge tests
cylinders and model piles. (caption shown in Figure 3)
After installation of the model pile, the specimen

is consolidated to in-situ stress level. At end of the
consolidation, the model pile will be attached to a
loading frame and a hydraulic actuator applies
packages with constant cyclic displacement
amplitudes. The number of cycles per package
depends on the soil type and behaviour, but
typically ranges between 500 and 2500. The cyclic
displacement amplitude y/D varies between 0.125%
and 8%, where the amplitude is doubled from
package to package. Typical loading frequencies are
between 0.1Hz and 1Hz.
Although only the response in the last cycle is
required in the BP design framework, the complete
development of the secant stiffness with number of
cycles is reported. In addition, the corresponding
hysteretic damping is evaluated and reported, which
is of importance for structural fatigue damage Figure 3. Equivalent damping ratio at steady state and
analysis. Example results of the normalized steady- trendline based on the shown data points
state secant stiffness and corresponding
damping ratio are presented on Figure 2 and The damping values show a plausible trend,
Figure 3. but with a large scatter. The dashed line in
Figure 2 shows in addition two trend lines based Figure 3 represents a best fit with a large standard
on the BP design framework, which was deviation for increasing cyclic displacement
established based on best fit of centrifuge data. amplitudes.
The measured secant stiffness values agree well
with the trend line, but seem to underestimate the
84
3 SERVICEABILITY TESTING The series comprises tests with different cyclic
The p-y apparatus uses an MTS hydraulic actuation load histories, all performed at the same material
system for applying monotonic and cyclic loading. with the same initial stress and density state. The
This opens the possibility to test the pile response to tests were conduced such that the specimen height
generalized cyclic loading. Of particular interest for was kept constant during cyclic loading. Due to
designers is the foundation serviceability, that constructional reasons, the specimen is not water
means the accumulated displacement due to tight and could drain through the endcaps. A back
repeated cyclic loading. pressure cannot be applied.
Table 1. Index properties of Kaolin 4 TEST RESULTS

Parameter Unit Value Figure 5 shows a test with a constant average load
of 40kPa and 5 packages with cyclic load
Specific density ρs [g/cm3] 2.59 amplitudes of 40kPa, 60kPa, 80kPa, 60kPa and
Clay content [%] 55.7 40kPa. 500 cycles per package were applied except
for the largest amplitude where 1000 cycles were
Grain size D60 [mm] 0.003
applied. The loading frequency was 0.25Hz.
Plastic limit wP [%] 28 It can be seen, that the accumulated average
Liquid limit wL [%] 56.1 displacement increases continuously, and the
accumulation rate becomes larger for larger load
Plasticity index IP [%] 28.1
amplitudes. When decreasing the load amplitude
again, the accumulation rate also decreases and is
In 2018/19, a series of tests has been conducted to eventually smaller than in the corresponding
study the general suitability of the p-y apparatus for package with same load amplitude applied prior to
the application to serviceability problems. For that the peak phase.
purpose, the first-time load-controlled tests were The cyclic displacement amplitude is small(er)
performed. To ensure repeatability and consistency and almost constant for the two first load packages
of the tests, two bins with re-constituted Kaolin clay but increases continuously in the third package with
were prepared. The properties are summarized the 80kPa load cycles. The specimen shows a
in Table 1. After consolidation to 80kPa, the softening behaviour. When decreasing the load
blocks of Kaolin were divided into 9 square amplitude again, the corresponding displacement
samples (see Figure 4). Only one specimen can amplitude becomes continuously smaller. The
be obtained from each sample. specimen shows a hardening or setup behaviour and
seems to regain resistance. But the cyclic amplitude
remains larger than measured in the corresponding
package prior to the peak phase.
A second test is shown on Figure 6 where a
constant cyclic amplitude of 40kPa is applied but
the average load is varied in 8 packages from 40kPa,
80kPa, 40kPa, 0kPa, -40kPa, -80kPa, -40kPa and
0kPa. Again, 500 cycles per package with a
frequency of 0.25Hz was applied.
It can be seen, that the accumulated average
displacement follows the average load and seems to
be completely reversable when changing
the loading direction. As for Test 4, shown on
Figure 5, the accumulation rate is a function of
Figure 4. Reconstituted blocks of Kaolin for p-y testing the load intensity.
The cyclic displacement amplitude, however,
The sample was than trimmed to fit into the seems to be almost constant and independent on the
specimen cylinder and built in into the p-y average load. An exception are the packages
apparatus. After installation of the model pile, the without an average load (i.e. pav= 0kPa), where
specimen was consolidated in three steps to a final significant larger displacement amplitudes are
consolidation pressure of 120kPa. The DSS shear observed.
strength of this material is approximately 25.5kPa.
85
Figure 5. Test 4 with constant average and variable cyclic loading
Figure 6. Test 5 with variable average and constant cyclic loading
86
Figure 7. Test 2 with constant cyclic to average load ratio but alternating direction of average load and increasing cyclic loading
The last test presented is Test 2 shown on Figure resistance, since the amplitude slightly decreases
7. Like test 5 shown on Figure 6, the direction of within each package at constant cyclic load
the average load was changed. In addition, though, amplitudes.
also the value of both the average and the cyclic
amplitude changed such that the ratio between 5 DISCUSSION
cyclic and average load was equal 1 or -1, The three examples presented in the previous
respectively. The number of load cycles varied section show, that plausible test results can be
between 500 and 1000, all at a frequency of 0.25Hz. obtained using the p-y apparatus. Comparing the
Like for the other two examples, an increase of the results qualitatively with laboratory tests
average accumulated displacement can be seen for (Wichtmann, 2016) and numerical models
increasing load intensity. Furthermore, the average (Niemunis et al., 2005), reveal a quite consistent
displacement seems to follow the average load level behaviour. The cyclic flow rule proposed by
and even changes the direction accordingly. Niemunis et al. (2005), postulates an increased
However, after each reversal, the specimen seems to accumulation rate (i.e. the average strain per load
be "softer", following the previous tendencies cycle) depending on the cyclic amplitude and the
independent of the reversal. On Figure 7 is shown a ratio of cyclic to average load. This agrees with the
trend line over the fourth and second last package. observations in Test 4 and test 5 shown on Figure
The reversal in the third last package seems to not 5 or Figure 6, respectively.
affect the accumulation rate nor the absolute value. Very limited literature is found on laboratory or
In fact, it seems that the direction did not matter, and model tests with reversals, which could explain the
it may be expected that the accumulation rate and decrease in accumulated displacements as seen in
absolute value of the average displacement would Test 5 (Figure 5). Also, the setup effects, that
be the same in the second last package if no load means the increase in cyclic resistance during
reversal would have been made in the third last loading, as observed in Test 2 (Figure 6), has not
package. been explicitly reported in the literature. However,
Also, the cyclic displacement amplitude gets setup effects in laboratory tests are known and have
larger for increasing load amplitude. However, there been reported by Andersen (1988) at the example of
seems to be a tendency for an increased cyclic undrained cyclic DSS tests with consolidation
87
periods between the cyclic load packages. The specimen to consolidate continuously. Such tests
cyclic resistance increased in each package due to are currently conducted, but with focus on the
re-consolidation and compaction in the resting secant stiffness. For that purpose, symmetric cyclic
periods. Although, the specimen height was kept load amplitudes are applied. The first tests showed
constant in the p-y apparatus, local consolidation promising results and agreed qualitatively well with
effects are still plausible to happen; the largest pore numerical predictions. It is thus expected, that the
pressure will be generated around the pile, but due same configuration can be used for serviceability
to the high consolidation coefficient of Kaolin clay, testing as well.
it is likely that the pore pressure dissipated radially
(and axially through the endplates). Specimens of 7 ACKNOWLEDGEMENTS
natural clay teasted in the p-y apparatus showed The authors gratefully acknowledge the support
after dismounting a zone around the pile which from the Wave Loads and Soil Support for Extra-
seemed to have a different water content than the Large Monopiles (WAS-XL) project (NFR grant
rest of the specimen (Zakeri et al., 2017). Kaolin 268182) and from the reducing cost of offshore
clay, however, is white and such a zone can hardly wind by integrated structural and geotechnical
be identified. design 2 (REDWIN 2) project (NFR grant 296511).
The soft response in Test 5 at zero average load
may indicate a gap around the pile. However, since 8 REFERENCES
the displacement amplitude was very small, the
Andersen, K.H., 1988. Properties of soft clay under static
observation may be also caused by a highly
and cyclic loading. Presented at the International
disturbed zone in the vicinity of the pile. The fact Conference on Engineering Problems of Regional
that the amplitude in the last package is almost twice Soils, Beijing, China, pp. 11–26.
the amplitude of that in the fourth package supports
this assumption. Only the soil at the side of the pile Niemunis, A., Wichtmann, T., Triantafyllidis, Th., 2005.
to which an average load was applied in the A high-cycle accumulation model for sand.
previous packages is softer, whereas the soil at the Computers and Geotechnics 32, 245–263.
other side is not significantly affected. When https://doi.org/10.1016/j.compgeo.2005.03.002
applying an average load also in the opposite
direction, the soil at both sides become equally soft, Wichtmann, T., 2016. Soil behaviour under cyclic
loading - experimental observations, constitutive
hence the cyclic amplitude in the 8th and last
description and applications (PhD Thesis).
package is almost twice as large.
Zakeri, A., Sturm, H., Dyvik, R., Jeanjean, P., 2017.
6 CONCLUSION Development of novel apparatus to obtain soil
The results show the general suitability of the p-y resistance–displacement relationship for well
apparatus for serviceability testing. Using conductor fatigue analysis. Can. Geotech. J. 54,
reconstituted Kaolin allowed to compare the results. 1435–1446. https://doi.org/10.1139/cgj-2016-0528
But the drainage coefficient of Kaolin is high, and it
Zakeri, A., Sturm, H., Jeanjean, P., 2019. Validation and
is believed that water drained out through the Extension of Soil Response Framework for Fatigue
endcaps. Analysis of Offshore Wells and Piles, in: Offshore
It is planned to perform further tests varying more Technology Conference. Presented at the Offshore
systematically the cyclic load histories. Technology Conference, Offshore Technology
Furthermore, it is considered to perform the tests Conference, Houston, Texas.
under constant axial loading allowing the soil https://doi.org/10.4043/29236-MS
88
Development of a Hammer to drive monopiles of offshore wind turbines in

Centrifuge
Semaan Maatouk1, Matthieu Blanc1, Luc Thorel1
GERS-GMG, Gustave Eiffel University, Nantes, France
1
Corresponding author: Semaan Maatouk (Semaan.maatouk@ifsttar.fr)
ABSTRACT: A new in-flight hammer has been especially designed at the Gustave Eiffel University to drive
in centrifuge at 100×g a model monopile for offshore wind turbine. The model pile driving device uses an
electro-mechanical lift system to drive open ended aluminium piles into saturated dense sand. A recap of
centrifuge pile drivers are presented. Then the paper describes the development of a powerful pile driving
hammer. At the end, the driving curve of a 5 m diameter prototype pile is presented.
Keywords: Centrifuge; Hammer; Open-Ended Monopile; Offshore Wind Turbine; Sand.
1 INTRODUCTION systems that studies the impact driving method

Offshore wind farms are gradually beginning to installation for large diameter pile (Heins and al.,
flourish since 1991 to produce clean and 2018, Fan and al., 2019). Therefore, the present
renewable energy. In shallow coastal waters, paper describes the development of a new in-flight
where the seabed is beneath 35 m, monopile electro-mechanical hammer with blow energy of
supports are to this day the preferred foundation 3.2 J to drive in centrifuge at 100×g a model
solution for offshore wind turbines. These piles are monopile of 50 mm large into a saturated dense
single large-diameter steel tubes, often installed in sand. Hammer blow count is studied against the
dense sands, typically 8 to 10 m in diameter, 30 to penetration depth of the monopile.
50 m long and weighing up to 1000 tonnes (Byrne
and al., 2019). 2 MATERIALS AND METHODS
Monopile foundations are commonly installed 2.1. Centrifuge modelling
using dynamic method so called “Impact pile Centrifuge modelling is used here on small scale
driving”. Offshore pile driving can last up to models installed in a macro gravity allowing for
several hours depending upon pile dimensions, soil the replication of stress state occurring with
conditions and input energy of the hammer. This prototype soil. The Gustave Eiffel University’s
process usually involves thousands of hammer geotechnical centrifuge (Figure 1) has a radius of
blows at the pile head. Each of the three majoring 5.5 m and its swinging basket permits the
characteristics of hammering, i.e. frequency, ram installation of a container with embanked mass of
mass and drop height can be adjusted to 2000 kg at 100×g.
accommodate different prototype hammers. A The test presented is carried out on 1/100 scale
recap of the different in flight hammers developed model piles at 100 times the earth’s gravity
in centrifuge for driving piles are presented in table (100×g). In this case, the main scaling laws
1. (Garnier and al. 2007) used to link the full scale to
In fact, the increase in the turbines capacity the small scale model are listed in Table 2: for
reduces the cost of the renewable energy. Thus, a instance x*=xmodel/xprototype.
powerful hammer is necessary to drive large Concerning the flow of water it follows the
diameter monopile. However, there is still limited Darcy law. Since diffusion processes take place in
Table 1. Recap of centrifuge pile drivers.
Frequency
Type of pile driver g-level Ram mass (g) Drop height (mm) References
(Hz)
62
Pneumatic 100 95 2.5 Nunez, and al., 1988
Double acting hammer
0 – 20 De Nicola, and al.,
Pneumatic 100 Max. value : 70 0 – 20
Free fall 1994
0 – 32 Levacher, D. and
Electro-magnetic 31.2 163.8 0.6
Free fall Cottineau, L.M.1996
40 – 55 Van Zeban and al.,
Electro-mechanical 50 Max. value : 240 35
Free fall 2018
89
R = 5.5 m
Swinging basket
L = 1.4 m
w = 1.15 m
Max. acceleration: 100×g H = 1.5 m
Figure 1. Gustave Eiffel University’s geotechnical centrifuge.
Table 2. Scaling law for different parameters (with N = placed over the sand sample reconstitution to be
100). saturated by a vertical rising flow through four
Parameter Notation Unit Scaling factor
draining channels located at the bottom part of the
Distance x* L 1/N
strongbox. The attained water table is about 40 mm
Stress σ* M/LT2 1
Density ρ* M/L3 1 above the sand surface.
Gravity g* L/T2 N
Deformation ε* - 1 2.3. Model pile
Velocity v* L/T 1 The model pile is a 525 mm (length) by 50 mm
Acceleration a* L/T2 N (outside diameter) by 2.5 mm (thickness) open-
Frequency f* 1/T N
ended circular aluminium tube. This corresponds
Force F* ML/T2 1/N2
Mass m* M 1/N3
to a prototype monopile with a length of 52.5 m
Energy E* ML2/T2 1/N3 and an outside diameter of 5 m and a wall thickness
Blow count n* - 1 of 0.25 m when tested at 100×g.
centrifuge tests N2 times faster than at the 2.4. Hammer description

prototype test, and the hammer is operated at N The experimental set-up presented in Figure 2(a)
times the prototype rate, the rate of fluid diffusion permits in-flight installation of the monopile by
should be reduced by a factor of N in the model. impact driving. The deliver energy chosen to drive
This may be achieved either by reducing the the model pile up to an embedment depth of 210
particle size of the soil or by using a viscous pore mm is 3.2 J (3200 kJ in prototype scale). This set-
fluid (Kutter B.L. 1994). In contrast, in this study, up consists of a hollow hydraulic jack fixed over
the soil sample is saturated with water instead of a the raiser. An electric motor is placed over the
viscous fluid. Thus, excess of pore water pressure vertical actuator to rotate a rod inside the jack. The
cannot be generated during the hammering process end of this rod is attached to a fork which allows
which can induce drained fluid response and the vertical translation of the rotating ram (Figure
therefore changing in the driving curve. 2(b)). On this ram, there exist two diametrically
opposed cylinder bearings which are rested on a
2.2. Model soil fixed wave-shaped support. It permits the ram free
The model soil consists of a poorly graded NE34 fall until 25 mm to impact driving the pile followed
Fontainebleau sand (Table 3) with a relative by a ride up of this ram thanks to an electric motor.
density of 80% obtained by air pluviation into a During the pile installation, the vertical jack
rectangular strongbox of internal dimensions 1200 follows the penetrating pile maintaining a
mm × 800 mm × 360 mm (length × width × specified impact drop height thanks to a laser that
height). detects a target object fixed on the monopile
Once the sand level reaches the top of the (Figure 2(a)).
strongbox, the sand pluviation is stopped. Then, a The input driving energy into the system can be
raiser with a height of 120 mm (Figure 2(a)) is varied depending on the drop height and on the
Table 3. Characteristics of Fontainebleau NE34 sand (Li and al., 2020).
Sand d60/d10 d50 (µm) 𝜌𝜌𝑑𝑑−𝑚𝑚𝑚𝑚𝑚𝑚 (kg/m3) 𝜌𝜌𝑑𝑑−𝑚𝑚𝑚𝑚𝑚𝑚 (kg/m3)

Fontainebleau NE34 1.53 210 1434 1746
90
ram mass of the hammer. The ram used weighs 160 3 RESULT
grams, which cannot be changeable during the pile On field (at full scale), the most common dynamic
installation. Therefore, a drop height of 20 mm is method of offshore pile installation is the impact
chosen to drive the pile with a model blow energy pile driving. Using this installation method, a
of 3.2 J. model monopile is hammered in flight at 100×g.
The rotational speed of the motor chosen is 150 Figure 3 shows the cumulative hammer blows
round per minute which corresponds to a blow rate recorded during the installation. The ordinate at
of 5 blows per second. By modifying this speed, the origin axis represents the pile penetration under
the number of blows per minute can be adjusted its self-weight at 100×g. Thus, the pile penetrated
until 33 Hz which corresponds to 20 blows per to a depth of around 2.8 m (56% of the pile
minute in the prototype. diameter) in prototype scale before being driven.
4932 blows are necessary to drive the pile to the
(a)
Electric motor
Load sensor
Displacement
Laser
sensor
Target
Hydraulic jack
Driving guide
Positioning beam Monopile
Metallic support
Raise supplement part
Sand strongbox
12 cm
36 cm
(b)
Fork
Ball bearing Cylinder bearing
Ram
Anvil Wave-shaped
ram support
Figure 2. (a) Experimental set-up of pile driving hammer (inner dimensions), (b) Details of the hammering process.
91
embedment depth of 21 m. As expected, the Fontainebleau sand is 3.2 J. Pile driven takes 4932
number of blows required per unit meter blows to penetrate the pile to the embedment
penetration increases exponentially with the length of 21 m.
penetration depth (Figure 4).
5 PERSPECTIVES
4 CONCLUSION In order to well understand the hammering
A new in-flight hammer has been designed to drive process, an instrumented monopile with optical
in centrifuge a 1/100 scale model open-ended Fiber Bragg Grating sensors (Li and al., 2020) and
monopile of 50 mm diameter. This hammer an accelerometer under the anvil would be
enables the free fall on 20 mm of a 160 grams ram. necessary. Moreover, sand sample should be
Thus, the deliver energy necessary to drive the saturated by viscous fluid in order to achieve
model monopile into a dense saturated with water “scaling law”. Once the pile is installed, without
Accumulated
Start driving
Impact driving at 100×g
Figure 3. Prototype data driving curve
1400
1200
1000
blows number
800
600
400
200
0
2.8-3.8
3.8-4.8
4.8-5.8
5.8-6.8
6.8-7.8
7.8-8.8
8.8-9.8
0-2.8
9.8-10.8
10.8-11.8
11.8-12.8
12.8-13.8
13.8-14.8
14.8-15.8
15.8-16.8
16.8-17.8
17.8-18.8
18.8-19.8
19.8-21
Pile displacement (m)
Figure 4. Hammer blow count plotted against the penetration depth
92
stopping the centrifuge, an electromechanical jack laws and similitude questions in geotechnical
positioned horizontally could be launched to load centrifuge modelling. International Journal of
the pile horizontally either monotonically or physical modelling in Geotechnics, 17(3), 1-24.
cyclically.
Heins, E., Bienen, B., Randolph, M.F., Grabe, J. 2018.
Effect of installation method on static and dynamic
6 ACKNOWLEDGEMENTS load test response for piles in sand. International
The first author wish to thank the Gustave Eiffel Journal of physical modelling in Geotechnics.
University and the Région pays de Loire for his https://doi.org/10.1680/jphmg.18.00028.
PhD grant.
This work was supported by France Energies Kutter, B.L. 1994. “Recent advances in centrifuge
Marines and State finance managed by the modelling of seismic shaking,” state-of -the-Art
National Research Agency under the Investments Paper. Proceedings 3rd International Conference
for the Future program bearing the reference on recent advances in Geotechnical Earthquake
SOLCYP+ANR-10-IEED-0006-18. Engineering and Soil dynamics, St. Louis, 2, pp.
Assistances and discussions from the technical 927-942.
staff of the centrifuge group are highly
acknowledged. Levacher, D., Cottineau, L.M. 1996. Pile driving
hammers in centrifuge modelling, 6th
7 REFERENCES International Conference and Exhibition on
Piling and deep foundations, Bombay, India, 15
Byrne, B.W., Burd, H.J., Zdravkovic, L., McAdam,
R.A., Taborda, D.M.G., Houlsby, G.T., Jardine, R.J.,
– 17 January, 8p.
Martin, C.M., Potts, D.M., Gavin, K.G. 2019. PISA:
new design methods for offshore wind turbine Li Z. S., Blanc, M., Thorel, L. 2020. Using FBGS to
monopiles. Revue Française de Geotechnique, estimate the horizontal response of a monopile in a
158(3). https://doi.org/10.1051/geotech/2019009. geotechnical centrifuge. International Journal of
physical modelling in Geotechnics.
De Nicola, A. and Randolph, M.F. 1996. Development https://doi.org/10.1680/jphmg.19.00022.
of a miniature pile driving actuator. Proceedings
Centrifuge 94, Editor Leung, Lee and Tan, Nunez, I.L., Hoadley, P.J., Randolph, M.F., Hullet, J.M.
Rotterdam: A.A. Balkema, 473-478. 1998. Driving and tension loading of piles in sand on
a centrifuge. Proceeding Centrifuge 88, Editor
Fan, S., Bienen B. and Randolph, M.F. 2019. Centrifuge Kimura, Rotterdam: A.A. Balkema, pp. 353-362.
study on effect of installation method on lateral
response of monopiles in sand. International Journal Van Zeban J.C.B, Azua-Gonzalez C., Alvarez Grima
of physical modelling in Geotechnics. M., Van’t Hof C., Askarinejad A., 2018. Design and
https://doi.org/10.1680/jphmg.19.00013. performance of an electro-mechanical pile driving
hammer for geo-centrifuge. Proceedings of the
Garnier, J., Gaudin, C., Springman, S.M., Culligan, P.J., International Conference on Physical modelling in
Goodings, D., Konig, D., Kutter, B., Phillips, R., Geotechnics, London, UK, 1, pp. 469-473
Randolph, M.F., Thorel, L. 2007. Catalogue of scaling
93
Experimental testing and ring polygons for the rolling stock – track infrastructure
interaction and subgrade analysis
Alexander Savin1, Andrey Zaytsev2, Alexander Abrashitov2, Andrey Sidrakov3, Alexander

Gorlov4
1
Administration Russian University of Transport, vice-rector for science, Russia
2
Track and track facilities Department Russian University of Transport, Russia
3
Institute of Track, Construction and Structures of Russian University of Transport, Deputy of Director, Russia
4
High-speed train moving Department JSC MOSGIPROTRANS, leading engineer, Russia
Corresponding author: Andrey Zaytsev (andrei.zaitsev2010@yandex.ru)
ABSTRACT: This paper presents the calculation and the results of experimental testing of events that
compensate for the local subsurface sleepers of the ballast of the railway track and influence of the junction
of the rail on lower substructure. The measures were developed with the aim of improving the work to
reduce the rate of local subsidence of the ballast prism leading to the appearance of local splashes, which in
turn leads to an increase in the probability of failure of the railway structure as a whole. Modelling was
carried out on the experimental ring railway track of the Experimental Testing Ring polygon (Sherbinka
station). The development of local under-sleeper subsidence of the ballast bed (Kosmin, A and Kosmin, V.
2006) of the railway track from crushed granite, with the formation of splashes and modelling of measures
that reduce the rate of development of these processes on the experimental ring railway track of the JSC
VNIIZhT (Railway Research Institute) training ground (Sherbinka station). The new tests on the
experimental ring are part of the general complex of tests of the ballastless track and its elements (laboratory
tests, testing grounds on the ring, operational tests on the operating line). The results of tests on the
Experimental ring should be the basis for the continuation of tests on the high-speed test site at speeds up to
400 km / h. This test site will be the first stage of the high-speed highway from Moscow to Kazan and heavy
haul operation (freight trains with high axle load). Three tasks solving at this Polygon are presented in this
paper.
Keywords: Field Testing; Experimental Polygon Ring; Stone Blowing Technique; Anchors; Ballastless Track.
1 INTRODUCTION concentrated in educational institutions, mainly in

During the railway exploitation, engineers and the Saint-Petersburg and Moscow institutes of
scientists provided a wide range of researches on the railway engineers.
exploitation sections of the railways, especially in The first branch research center in Russia, the
the part of rail – wheel interaction, rolling stock – Experimental Institute of Railways, was established
railway track interaction. Approximately, in the by a resolution of the Board of the People's
same period, the 80-90s years of the XIX century Commissariat of Railways on April 18, 1918. At
important experiments were performed by that time, the country's Railways were acutely
Zimmerman on the Eltzas – Lotqaringia railways, concerned about the reconstruction of the traction of
by Eng. Stetsevich on the Tambov – Saratov and rolling stock. To study this issue, the Institute
Baltic railways and Eng. Vasyutinsky on the needed to organize a special polygon in the form of
Warsaw – Wien railway (MIIT Report. 1958, a regular circle on an absolutely flat site, for the
Zaytsev, A. 2015.). construction of which a significant territory was
Thus during 1897–1898 an Engineer required.
A.Vasyutinsky did some researches for the A proposal to build a "Special Railway Circle
permanent way on an experimental site nearby Poligon" for experiments with railway rolling stock
Warsaw: a railway gauge 1435 mm, the rails with and infrastructure was put forward by Alexey
weight 31.45 and 38 kg per meters, wooden sleepers Dolinzhev, an employee of the Institute for traction
with length 2.7 meters and two variants of the reconstruction, this idea was put into practice in
ballast prism: from coarse sand in mixture with 1932 when creating an Experimental Ring Polygon
gravel and rock stones. that was built on the Sherbinka (Butovo) station of
Before the Russian Revolution in 1917, the the Moscow Railway, the design and construction
scientific developments and research in the field of of which was supervised by Nikolai Belokon’
railway transport and railway infrastructure were (Stepov, V. and Savin, A. 2017).
95
(a) PK 25+50 Track axis
Track axis
43 2 1
R = 956m – first 5
track, length 6000 m 6
4 7
8
2
3 Track axis Track a
oncrete
CS
Figure 2. Cross-sections of the railway track a) section of

Figure 1. Part of interactive view of the Sherbinka Ring zero place subgrade type (Morozova A. 2013); b) the section
Poligon (1-station track Park 2-ballastless track polygon, 3- with ballastless construction in the tests of A. Savin (1-a
contact network laboratory 4-Experimental Ring centre, mixture of soils treated with polyphilizers; 2 – a crushed
Interactive Presentation. 2016, Savin, A. and Matveeva, stone-gravel-sand mixture)
E. 2017, Stepov, V. and Savin, A. 2017).
high-speed traffic (the solution of the last task is not
The Experimental Ring (Figure 1) has: the 1st ring yet completed and is planned to be carried out
track with a length of 6000 m with a constant curve during large-scale experimental work in 2020).
radius of 956 m without slopes; the 2nd and 3rd ring
tracks with a length of 5700 m with curves with a 2 THE PROBLEMS SOLVED AT THE
radius of 390 to 1200 m and a longitudinal slope of
TEST RING POLYGON IN
track - 12‰ (GOST R 57076-2016).
According to the analysis of engineering,
SHCHERBINKA
geological elements of typical soils and 2.1. Elimination of splashes by stone-blowing
construction materials: on the Experimental ring technique
laid (Figure 2a): 1 - gravel ballast; 2 - embankment The traditional bearing of ballast in the plan and
soil: dark fine sand, medium density with a high profile tamping violates the established under the
content of crushed stone; 3 – embankment soil: dark sleeper bottom (compacted ballast under the action
fine sand, medium density, with a high content of of dynamic forces the surface of the ballast),
crushed stone; 4 – embankment soil: sand gravel, resulting in re-packing the ballast under the sleeper
moistered, medium density, with grass and gravel; granules and the formation of broken and worn
5 - clay; 6 – loam; 7,8 - clay (Morozova 2013). particles of the ballast, reducing the drainage
This paper presents the results of solving several properties of under-sleeper ballast foundation and
problems using the capabilities of the test site, the settlements of the track (Esveld, C. 1989, Peych,
depending on their importance for Russian Y. 2003.).
Railways: The settlements of the track lead to an increase in
(1) is the development of modern technical the number of track linings, which, in turn, leads to
solutions to reduce the dynamic impact in the rail a deterioration in the quality of the ballast under the
joint zone sleeper. In addition, with small lifts, there is a so-
(2) is testing the structures of anchors for the called edge or false padding (non-falling of the
subgrade infrastructure at the testing site: including ballast stone along the edge of the raised 10-15 mm
as braces for contact network supports and as sleeper).
structures for strengthening the subgrade slopes, The method of adding ballast under a sleeper was
(3) is the testing of existing and promising first used before 1910 and became common on the
ballastless upper structures for the permanent way London and North-West Railway (UK), and then it
of the track for the movement of heavy trains and
96
was quickly extended to other British Railways. did not fit tightly with the ballast material, two on
Since 1934, the Great southern railway in Ireland each side of the joint.
has introduced a method of filling the butt end of the The ballast excavation showed that the bottom
sleeper with long-handled shovels. bed of the sleepers was strong (damped). Between
In the USSR, the souflage was first used in 1936 the sleepers, there were dried traces of a splash of
on crushed stone ballast, in 1937, already on two dirty - white pulp that covered the space of eight
Railroads souflage was applied, with the pouring of sleepers: three on the outgoing side of the rail joint
material from butt end of the sleeper. and five on the receiving side (Figure 3).
In the 70s XX century, in the USSR, souflage was
prohibited completely for two reasons: the need to
the digging up the ends of sleepers, which could
lead to a loss of cross-stability of the rail-sleeper
grid, and with a rapid increase in the laying of the
unjointed track.
Mechanization of the souflage was proposed by
John Waters (Waters, J.1981, McMichael, P and
McNaughton, A. 2003.), the essence of the new
proposal was the introduction of a new rubble "air
blowing of stone" (the fascination of the rubble with
the flow of air in the working tube), thus the
technology of stone-blowing was developed and
implemented.
At first stage, the comparative analysis was Figure 3. The uncut wet splash, Zhilevo, Moscow Railways
performed for stone blowing and tamping
techniques by 1G physical modelling of recovered These tests revealed that when using the
ballast. The focus on contamination levels and technology of pneumatic stone blowing on a wet
lateral resistance values and preliminary tests were splash, the time between adjacent alignments
performed to develop correct physical models increased by almost six times. In general, these field
(Zaytsev, A., Abrashitov, A., Sydrakov, A. 2018.). tests have shown a possible advantage of
The results of the laboratory experiments clearly straightening with a souflage (stone blowing) over
demonstrated that the ballast adjusted by stone straightening with mechanical and pneumatic
blowing possesses has better mechanical tipping of sleepers in the area of small wet splashes
characteristics in comparison to ballast adjusted by in the presence of a strong sleeper bottom bed.
tamping. To prove the advantages of the stone The settlements growth up 20 millimeters for 6
blowing technique, experiments are being carried million tons of rolling stock tonnage, and up to 80
out in the real-world environment. percent of the settlements after straightening with
In parallel, the field tests were carried out on the pneumatic stone blowing during the first 25
track alignment technology using stone blowing thousand loading cycles (Figure 4).
technic on the 4th section of the Zhilevskaya
distance of the Moscow railway (not cycle
Polygon). The load capacity of the section is 113
million tons, the passed tonnage is more than 500
the ground points
million tons, and the set speed of freight trains is 80

km/h. Capital repair has been done five years before
the experiments. The construction of the track is an
unjointed track on reinforced concrete sleepers,
crushed stone ballast, anchoring the rails to the
sleepers with an anchored ARS-type of rail
fastenings.
As an experiment object, an uncut wet splash was the number of trains
selected in an insulating joint nearby the road
crossing. At the rail-joint zone, there was a non- Settlements along internal rails
attachment of the sleeper to the ballast at a length of
4 sleepers with subsidence of the second degree. Settlements along external rails
There were sleepers on either side of the joint that Figure 4. The average rail settlements under train loading
after straightening with pneumatic stone blowing at the site
97
For further research of pneumatic stone blowing
technology in autumn of 2018, a special site was
prepared on the Experimental Polygon Ring (see Number of sleepers
Figure 1).
Settlements, mm
On the curve adjacent to the experimental section,
there were numerous wet splashes of up to 8 before alignment
sleepers at the joints in the plan of the railway track
on the section directly, a dry splash was recorded, 16 million tons
which manifested itself in the absence of a clear fit
of the sleepers to the ballast.
On the curve adjacent to the experimental section,
there were numerous wet splashes of up to 8 25 million tons
sleepers in the rail joints in the plan of the railway
track, on the experimental section, a dry splash was Figure 6. Power profile of the left joint rail line 291-292
founded, which manifested itself in the absence of a
clear fit of the sleepers to the ballast. The all-gross
weight of trains which passed through section was 2
billion 200 million tons. (a) (b)
Pieces of a flat square geogrid measuring 75 by 30
cm with 32 mm mesh were placed under the two
sleepers close to the joint on both sides, covered
with a plastic mesh with 5 mm mesh (masonry
mesh). These geogrids were laid on both sides of the
sleepers under the supporting parts of the sleepers Figure 7. The state of the browning layer after passing 25
(four in total). In the joint numbered 293-294, a million tons of traffic: a) crushed stone with the addition of
mixture of crushed stone and rubber was used as a fine particles б) crushed stone without the addition of rubber
souflage (stone blowing) material, in a ratio of 4 to crumb
1, respectively. The rubble was painted blue color
(Figure 5). the subsequent use of cut rubber of the 5-8 mm
fraction helped prevent the ballast material from
Rail Joint
being worn under the sleeper for 25 million tons of
missed traffic. These Polygon tests have confirmed
the high efficiency of the stone blowing technology
for straightening the track even in particularly
difficult operating conditions with low labor and
material costs for carrying out the straightening
Stone blowing material plastic geogrid
operation.
The possible application of pneumatic stone
blowing technology in some cases include:
blowing geogrid alignment of a long-used track on ballast, including
Figure 5. The design scheme of test sites on dry and wet splashes, preventing the formation
of splashes; alignment of the track on overpasses
During the tests, compact blower equipment was with driving on ballast; alignment of the track in
used, and the sleeper boxes were partially unearthed problem areas (crossbars of switches, joints of
and staggered. equalizing spans; alignment of a ballastless tracks.
According to the results of profile measurements
in combination with leveling data and values of rail 2.2. Checking the effectiveness of using
deflection measurement by flexometers (LVDTs), ground self-opening anchors on Ring Polygon
power longitude profiles of the rail head under the
On the 3rd kilometer of the 3rd track of the
rolling stock were obtained. The most characteristic
Experimental Ring of JSC "VNIIZHT" at the
power profile was obtained on the left thread of the
Sherbinka station, tests were conducted of
rail joint 293-294 (Figures 6, 7).
prototypes of "self-opening anchors of the AS-
The blowing crushed stone of the 5-10 mm
railway series" and sets of equipment for their
fraction was completely embedded in the pores of
installation.
the ballast particles in the absence of a plastic mesh,
98
A line of self-opening ground anchors, the most The placement of the supports and the drawback
popular models are shown in figure 8. These of the contact network support is shown in Figure 9.
anchors have a load-bearing capacity of up to 12
tons, they use the same anchor rods, as well as nuts
with rings or end fittings.
On figure 8b the installation of the anchors is
shown. This document describes methods for
installing the most common anchor models: both
manually and using power equipment. The process
of installing anchors to a greater depth differs from
the one described only in that longer steel pushers
and anchor rods are required. embankment
Lanhor
AS-ZD 4
Figure 9. The placement of the anchor to hold the contact

network support
Defective support with a bearing capacity of 98

AS-ZD 2 AS-ZD 1 kNm, on an embankment with a size of 3 m. Ground
conditions: single-layer soils, loam soft-plastic,
(a)
watered, with one experimental drawback with an
anchor AS-ZHD-2, the size of 360x100 mm, the
area of the bearing surface S=330 mm2 (Figure 9).
(b)
Screw rod with a diameter of 20 mm, length4 m.
The required forces in the cables of the braces (in
the traction of the ground anchor) are 11.77 kN for
temporary fastening of defective concrete supports.
As a result, the AS-ZHD2 anchor was installed
during the installation time of 45 minutes, the depth
of the anchor is 3.5 meters.
The anchor was loaded to a maximum load of
15kN, and the working load on the anchor was 13
kN (after holding and stabilizing the indicators). A
temporary drawback is fixed on the anchor, the
Figure 8. Three types of ground self-opening anchors force in which is 2 kN.
Another example will be shown with the
The anchors are designed for installation using a placement of an anchor and a plate to protect the
hydraulic jackhammer of a weight class of 41 kg. slope of the embankment on the polygon.
After the anchor is submerged to a depth Embankment with a steep slope with a height of 3
corresponding to the length of the anchor rod, the meters. Ground conditions: medium-density gravel
installation tool ("steel pusher") is removed. Then with sand, used AS-ZHD-4 anchor, size 205x65
the anchor turns from the longitudinal position to mm, bearing surface area S= 110 mm2. Screw rod
the perpendicular, so-called "anchor lock" position. with a diameter of 16 mm and a length of 2 m.
The direct indication measuring device involving As a result of the work, the AS-ZHD4 anchor was
the hydraulic jack, which is equipped with a locking installed, with an installation time of 20 minutes,
mechanism, allows easy and quick check of the and an anchor depth of 1.5 meters. Testing was
anchor load. If the soil resistance is not hard enough, performed with an anchor loading up to a maximum
the operator will immediately determine that load of 13 kN.
additional measures are needed: installing a second Load on the anchor working 11 kN (after exposure
anchor, increasing the depth of the dive, or using a and stabilization of indicators). The plate with
larger anchor model. dimensions 1x1 m is fixed on the anchor (Figure
10).
99
A new structure is currently being prepared for
installation on the ring. When constructing a section
of the interface of the roadbed, two protective layers
with layer-by-layer compaction are poured with
vibro-rollers (RSRS, 2019).
The thickness of the second layer with layer-by-
layer compaction with vibratory thickness shall be
not less than 1.8 m. The Thickness of the first
protective first protective layer should be 0.4 m with
ballastless construction of the top structure
assuming the presence of a hydraulically bound
base layer under the condition of the presence of a
hydraulically bound base layer at least 30 cm and
0.7 m at ballast construction of the permanent way.
When dumping of the upper protective layer of the
subgrade to form a recess in the form of "wedge»
and the ballastless structures form a recess in the
form of a "wedge". The length of the interface
section is assumed to be 20m.
Figure 10. Anchor and plate for the strengthening fill slope Then the resulting recess is filled with the material
of the protective layer with geogrid reinforcement.
2.3. Using a polygon to check the structures of
a ballastless upper structure of a track 4 CONCLUSION
Before laying the experimental structures of the The results of solving several problems using the
ballastless upper structure of the track, special capabilities of the test site, depending on their
attention was paid to the construction of the base importance: task number (1) is the development of
(subgrade), which had the same type of device for modern technical solutions to reduce the dynamic
all commercially available structures on Russian impact in the rail joint zone.
railway industry: LVT, MaxBögl, Alstom, Tines The possible application of pneumatic stone
and was built by a single contractor (Savin, A. 2015, blowing technology in some cases include:
Savin, A. 2017, Ulanov, I. and Gorlov, A. 2019.). alignment of a long-used track on ballast, including
The device of the reinforced layer is standardized on dry and wet splashes, preventing the formation
on Russian Railways for strengthening, liquid and of splashes; alignment of the track on overpasses
powder additives are added to the soil mixture with driving on ballast; alignment of the track in
(Technical conditions N 2640r of JSC RZD. 2012). problem areas (crossbars of switches, joints of
Concentrate number 1 is mixed with water. equalizing spans; alignment of a ballastless tracks.
Concentrate number 2-a mixture of surfactants, Task number (2) is testing the structures of
cement, lime, slags and other components, added to anchors for the subgrade infrastructure at the testing
the soil in the form of a finished powder. site: including as braces for contact network
At the base of the four designs of ballastless track supports and as structures for strengthening the
(LVT, MaxBögl, Alstom and Tines), arranged on subgrade slopes. The anchors are designed for
the 2nd path of the Pilot ring, laid politicization installation using a hydraulic jackhammer of a
reinforced layer and a layer of gravel-sand mixture weight class of 41 kg. After the anchor is submerged
(SGPS), see Figure 2b. to a depth corresponding to the length of the anchor
The thickness of these layers is 0.5 and 0.7 m, rod, the installation tool ("steel pusher") is removed.
respectively. their width is 7 m, the slope of the Two examples for current net pole and for
surface in the cross section is 0.04. The thickness of reinforcing of subgrade slope are presented.
these layers is 0.5 and 0.7 m, respectively. their Task number (3) is the testing of existing and
width is 7 m, the slope of the surface in the cross promising ballastless upper structures for the
section is 0.04. permanent way of the track for the movement of
At the end of strengthening the soil with heavy trains and high-speed traffic (the solution of
polyphilizers, the strength for uniaxial compression the last task is not yet completed and is planned to
was determined in the laboratory dry samples, be carried out during large-scale experimental work
which amounted to 2.08 MPa, water saturation at in 2020).
the same time 1.47 %, from the mass of dry soil. The On the Experimental Ring, six types of UAVS
strength of water-saturated samples was 1.32 MPa. from different manufacturers (Rheda, LVT,
100
MaxBögl, Alstom, Tines) were tested to determine Peych, Y. 2003. Interaction of path elements in the
more than 30 parameters (track geometry, track conditions of clogging, ballast dampening, and
draft, etc.). subluxation of the ground base, Doctor of Science
The values of deflections and stresses Degree Thesis, Russia, Moscow: 312.
experimentally obtained during heavy traffic with a
RSRS, 2019. Temporary railway track of ballast-free
load of 235 kN on the axis for all types are construction. track structures on the section of the 3rd
approximately two times lower than the critical track Experimental ring of the Shcherbinka station of
values, 294 which allowed us to conclude that one the Moscow railway Branch of a LLC "RSRS GmbH
of the areas of application is cargo traffic (Savin, A. Railway Infrastructure projects» "The Railway
2017). Tracks» documentation, Russia, Moscow: 28.
Analysis of rainfall intensity paths on the gravel
bottom and the intensity of the sediment showed Savin, A. 2015. Combined Method of Ballastless Track
that the sediment of ballastless track occurs 4.8 Examination. VNIIZhT Bulletin. Railway Research
times slower than the ballast, with the exception of Institute Bulletin. – 2015. – No. 2: 33-36.
the transitional areas. However, the precipitation is
Savin, A. 2017. The conditions for using the ballastless
uneven. On a section of 75 m after passing 600 Track, Doctor of Science Degree Thesis, VNIIZT,
million tons of gross precipitation varies from 2 to Moscow, Russia: 444 (in Russian).
11 mm.
The reason for such roughness slopes is the Savin, A. and Matveeva, E. 2017. For the First time in
uneven deformation modulus of the chemically Russia. Experimental ring – 85 years old. Railway
fortified soil and the layer of crushed stone-sand- engineering (Technika zeleznih dorog). No. 3: 79-86.
gravel mixture along the length of the test site. (in Russian).
Stepov, V. and Savin, A. 2017. Experimental ring of JSC

5 REFERENCES
"VNIIZHT": history, advantages, prospects. Bulletin
Esveld, C. 1989.Track Geometry and Vehicle Reactions.
of the JSC RZD, Moscow. (in Russian).
Rail Engineering International Edition. 4: 13.
Technical conditions N 2640r of JSC RZD. 2012.
GOST R 57076-2016 Field test centres for the railway
Technical conditions of for crushed stone-gravel-sand
rolling stock and the infrastructure elements. Technical
mixture, used for the laying as a sub-ballast protective
requirements, Russia, Moscow: Standartinform, 2016
layers at the railway track infrastructure. Edition
(in Russian).http://docs.cntd.ru/document/1200139389
20.12.2012.
Interactive Presentation. 2016. Interactive presentation of
Ulanov, I. and Gorlov, A. 2019. Estimation of structural
the Experimental Ring of VNIIZHT JSC 11.11.2016.
deformations of the transition section rigidity of the
http://vniizht.ru/3d/new/
subgrade and artificial structures of the high-speed
highway. Eurasian Union of Scientists (ESU). No. 3
Kosmin, A and Kosmin, V. 2006. English-Russian
(60): 46-55.
railway dictionary, Marshrut, Moscow, Russia: 960
ISBN 5-89035-084-6
Waters, J.1981. Pneumatic Ballast Injection a means of
adjusting track level. 4th International Rail and
McMichael, P and McNaughton, A. 2003. The
Sleeper, conference Adelaide, Australia.
Stoneblower-Delivering the Promise: Development,
Testing and Operation of a New Track Maintenance
Zaytsev, A. 2015. Heavy Train Traffic Conditions and
System. TRB Annual Meeting CD-ROM.
Peculiarities. Technology and Art, 6, Vilnius,
Litwania, ISSN 2029-400X: 14-19.
Morozova (Konon), A. 2013 The bearing capacity of the
railway track foundation, Candidate of Science Degree
Zaytsev, A., Abrashitov, A., Sydrakov, A. 2018. 1g
Thesis, PGUPS, Saint-Petersburg, Russia: 184.
physical modelling of the stoneblowing technique for
the improvement of railway track maintenance
MIIT Report. 1958. The report and materials of the
Physical Modelling in Geotechnics – McNamara et al.
experiments on the Oshmyany-Gudogay Belarusian
(Eds), Taylor & Francis Group, UK, London, ISBN
railways. Typescript, Moscow Institute of Transport
978-1-138-34419-8: 203-206.
Engineers (MIIT), Moscow, (in Russian).
101
Pneumatic load frame for long-term in-situ application of moderate tension loads
on foundation elements
Jorge Yannie1, Claes Alén2, Jelke Dijkstra2
Infrastructure, NCC, Sweden
1
Architecture and Civil Engineering, Chalmers University of Technology, Sweden

2
Corresponding author: Jelke Dijkstra (jelke.dijkstra@chalmers.se)
ABSTRACT: A cost-effective load frame is developed in order to study pile setup or creep of foundations
elements under sustained loading in natural soil deposits in a controlled manner. The load frame is
specifically designed for long-term in-situ loading of foundation elements in tension up to loads of 500 kN.
The novelty is the use of pneumatic springs, actually commercially available pneumatic lifting bags that are
demonstrated to function in two modes of operation. In the first constant gas mass application, a control
precision of ±10% is easily attainable with few manual adjustments over the course of weeks. In the second
regulated application, the addition of an electrical air compressor with accumulator tank increases the
precision to <1% without manual intervention.
Keywords: Load Frame; Pile Load Test; Tension Loads; Long-Term Loading.
1 INTRODUCTION multiple testing sites or for large 1-g test pits, giving
Field tests are not frequently performed due to the repeatability under well controlled procedures. It is
considerable amount of time, work and resources based on utilising the potential energy of a (gas)
required. In some cases, custom-made load frames spring system with a very low spring constant so
need to be developed in order to test the pile that any additional displacement will only result in
response, and sometimes the test equipment is lost a small load variation. This potentially will open up
with each test. Furthermore, the level of control and more well controlled pile testing to be performed by
consistency is limited, as opposed to physical model research groups with modest means, such as for
tests at 1-g using a calibration chamber (e.g. Brown example universities.
et al. 2004, Jardine et al. 2009) or at elevated stress
levels in the centrifuge (e.g. Nicola et al. 1999, 2 PNEUMATIC LIFTING BAGS
Zhang et al. 2010). Another issue is that in-situ pile The main criterion for long-term testing is to have a
load tests cannot be easily generalised, as the results constant load within the target accuracy over the test
depend on the test site, pile type, testing procedures duration. In case the potential energy of a spring
and experimenter. Nevertheless, field tests are the system is considered, this translates to having a
final step in validation especially when the local sufficiently small spring constant so that any change
soils are difficult to replicate in the laboratory, such in the displacement range of interest does not
as the sensitive clay in the current case. significantly change the applied load. Many
Common loading methods used in the field are solutions exist for conventional spring elements
kentledge (dead weight) or complex hydraulic with several geometries and materials. For example,
loading systems, (e.g. Fellenius 1972, Karlsrud et al typical spring types are compression/tension/torsion
2014). Dead weight loading requires strict safety coils, flat-rolled, flat-stacked and disc springs.
measures and large support structures while These traditional mechanical springs made out of
regulated hydraulic systems are costly and prone to steel are not easily designed to the specific constant
control failure. Furthermore, for reliable studies on target stiffness over a large displacement range.
pile-setup or pile-creep displacement, the load Hence, the somewhat unconventional solution
needs to be applied for long periods of time considered here: the gas (air) spring. The use of gas
(Karlsrud et al 2014). springs is not new to pile load application, as the
In the current research, a new loading method controlled combustion of a gas is the standard
based on pneumatic lifting bags was developed to method used for pile driving, dynamic testing or
overcome some of the aforementioned limitations. even quasi-static testing (Rausch et al. 1985, Janes
This new approach is designed to apply long-term et al. 1991). In the current application, however, the
moderate loads (<500 kN) in a robust and simple load duration are orders of magnitude longer than in
manner, while being cost-effective and safe. It is a previous studies. Gas springs are available in
flexible solution that can be easily adopted on multiple geometries and materials, however, the
103
pneumatic lifting bag type provided the most
economical solution for the air pressures and
contact surface areas considered.
2.1. Configuration
The lifting bags used in this research were sourced
from Trelleborg AB (Figure 1). These bags are
made of high-quality rubber reinforced with layers
Figure 2. Illustration of possible bag configurations. Four
of aramid fibre, making them robust and safe to bags gives increased stability against eccentricity as
work with in the field. The bags are square shaped compared to two bags.
with dimensions 658 X 658 X 25 mm3 under
deflated state and provide a maximum 380 mm 2.2. Analytical model
lifting height. These are designed for max 800 kPa
The reaction load from the bags is a function of the
(8 bars) of gas pressure and provide different load
lifting height H that governs the contact area A, the
capacities depending on their contact area. The bags
applied gas pressure and the gas temperature. Three
are found to be the optimal solution for the long-
possible loading scenarios are considered:
term load application based on their simplicity,
versatility and low cost. Two bags with model • Constant pressure: using a constant pressure
number TLB-32 were used in the current source, e.g. air compressor with a mechanical
application, with a maximum capacity of 320 regulator valve.
kN/bag for the largest contact area. • Constant gas mass: the bags are pre-loaded
A frame with a minimum of two bags is required until the target load and subsequently sealed
for applying a tension load on the pile head or to retain the air mass.
anchor. The loading capacity of the frame is • Constant load: using an automatically
proportional to the number of bags and the regulated pressure source that is adjusted as
configuration of the bag (i.e. pressure and contact function of the pile head load, e.g. solenoid
area). The actual configuration used and other valves and electrical air compressor.
options to scale the load are shown in Figure 2. Two In the first case, temperature effects will not play
bags balance in their common axis (y axis in Figure any role since the pressure is held constant.
2), keeping the frame aligned. However, if the Therefore, the bags can be modelled using simple
tension loading rod is not centred, the frame can tilt. equilibrium equations that are based on volume
The latter is not a problem in the case of three or change only (Equation 1). In the second case,
four bags, as these balance in all directions to keep however, the load capacity becomes more complex
the same internal gas pressure, since all bags are since changes in volume and temperature can take
interconnected to the same supply line. place simultaneously. For this case, Boyle's law is
Finally, it is important that the bags are able to used to account for volume and temperature
redistribute the load on the ground surface without changes as shown in Equation 2. The third case will
inducing excessive settlements. The latter need is keep the load constant by automatically regulating
site specific, however, shallow foundation pressures the pressure in the bags as function of a sensor
up to 50 kPa did not prove any issues in the trial reading. Therefore, changes in temperature or
tests. volume will be compensated for by the control
equipment.
The generated load Q depends on the contact
length L of the bag and the gas pressure p:
𝜋𝜋𝐿𝐿2
𝑄𝑄 = 4
𝑝𝑝 (1)
Changes in gas pressure caused by volume and/or

temperature variations are estimated using Boyle's
law:
Figure 1. Pneumatic lifting bag model TLB32 from Trelleborg 𝑛𝑛𝑛𝑛𝑛𝑛
AB. 𝑝𝑝 = (2)
𝑉𝑉
104
Figure 4. Geometrical model for the lifting bag.
Figure 3. TBL32 load versus displacement curves for different 𝜋𝜋2 𝐻𝐻 2 𝜋𝜋𝜋𝜋
pressure and loading conditions. 𝐿𝐿 = � 4
− 2𝐻𝐻 2 + 8𝑅𝑅2 − (6)
2
where p is the gas pressure, n is the gas mass, R

where 𝐴𝐴𝑠𝑠𝑠𝑠 is the surface area of the sphere
the gas constant, T the gas temperature and V the
volume occupied. For a fixed mass of gas with representing the free inflated bag, 𝐴𝐴∗𝑠𝑠𝑠𝑠 is the surface
constant temperature T, the volume V occupied by area of the compressed sphere, R is the radius of the
the gas is inversely proportional to the pressure. The sphere, L is the contact surface diameter, H is the
gas mass n and the gas constant R are combined in maximum lifting height of the bag and α is the angle
a single constant k: of the half ring radius to the vertical axis.
The correction factors for the sphere radius R=𝜒𝜒R
𝑘𝑘𝑘𝑘 and the diameter of the contact surface L= 𝜂𝜂L are
𝑝𝑝 = (3)
𝑉𝑉 calibrated using the load curves provided by the
manufacturer. The values for the TLB-32 bag are
These three different scenarios are elaborated in approximately 1.15 and 1.25, for 𝜒𝜒 and 𝜂𝜂
Figure 3 for an initial load of 50 kN using 200 kPa respectively, see Figure 5. Having the contact area
of gas pressure. In all cases, the equilibrium calculated using L in Equation 6, the volume of the
equations are required to calculate the resulting load compressed sphere is determined using Equation 7.
from the applied gas pressure. These equations
require the contact area between the bags and the 𝐿𝐿+𝐻𝐻
𝐻𝐻 2 𝐿𝐿 2 𝜋𝜋𝐻𝐻𝐻𝐻2
reaction points, which is approximated by 𝑉𝑉 = ∫ 𝐿𝐿
2
4𝜋𝜋𝜋𝜋 �� − �𝑟𝑟 − � 𝑑𝑑𝑑𝑑 + (7)
2 2 4
modelling the bag geometry under two conditions, 2
as illustrated in Figure 4. In the first condition, the

bags have a complex shape when fully inflated with where V is the volume of the lifting bag when
free boundaries and low gas pressure; with a compressed and r is the radius until the side edge.
maximum lifting height based on the bag model. The effect of changes in temperature and height on
This complex form is approximated using a sphere the bag's behaviour is estimated by combining
geometry with a modified radius in order to account Boyle's law and Equation 7. First, the constant k is
for the difference in volume from the true bag shape. obtained for a given gas pressure, bag volume and
In the second condition, the bags are placed between gas temperature. Then any changes in temperature
two plywood sheets and compressed, resulting in a and/or volume must be compensated by the gas
shape similar to a merged cylinder and a half-circle pressure to keep k constant.
ring. Again, this shape is an approximation and the
contact surface diameter (cylinder base) is modified
to account for the difference in contact from the true
bag shape. Finally, the additional assumption of a
rigid membrane is required to impose that both
geometrical models maintain the same surface area.
After making these simplifications the contact
surface diameter L is then derived from Equations 4
and 5.
𝐴𝐴𝑠𝑠𝑠𝑠 = 4𝜋𝜋𝑅𝑅 2 (4)
𝜋𝜋 𝐿𝐿 𝐻𝐻 𝐻𝐻 2𝜋𝜋𝐿𝐿2
𝐴𝐴∗𝑠𝑠𝑠𝑠 = ∫0 2𝜋𝜋 � + sin𝛼𝛼� 𝑑𝑑𝑑𝑑+ (5)
2 2 2 4
Figure 5. Fit of analytical model against manufacturer data
for bag TLB-32.
105
The basic characteristics of the model pile (or
anchor) and relevant soil conditions need to be
known in advance in order to design the most
appropriate loading system. For the field test
reported here, two bags were used per pile in order
to have more flexibility in the maximum load and
allowable displacement.
3 IMPLEMENTATION
The pneumatic load frame was tested in the field by
applying a tension load on 4.20 m long and 0.235 x Figure 7. Change in load due to lifting height variations for
constant pressure case.
0.235 m2 square pile elements embedded 13 m
below the ground surface. At this depth, the soil
comprises of sensitive clays with an undrained shear 3.1. Constant gas mass
strength in the range of 30 - 40 kPa leading to a After load application the bags were sealed and left
maximum expected short-term failure load ofabout unattended whilst monitoring the load with a load
138 kN. A tension loading rod fixed at the pile toe cell (Anyload 363YH; 30 tons) located at the
guided the load back to surface level. Hence, the connection joint of the tension loading rod and the
shallow load from the reaction frame did not affect HEB beams. This configuration was already
the tested pile sections and the pile element was satisfactory for quick maintained load testing
loaded in compression. The latter prevents tension (where a constant air mass was applied at the
cracks in the concrete. beginning of each load step for < 30 min) and for
The load frame consisted of two parallel HEB260 long-term testing (where for an allowed regulation
steel beams resting on top of two lifting bags and precision of ±10% change in applied Q, manual
connected to the tension-loading rod as shown in adjustments were only needed every couple of
Figure 6. The bags were located between two weeks).
double-layered waterproof plywood sheets and
rested on top of a 0.30 m thick timber mat. The 3.2. Constant load
pressure in the bags pushed the steel beams upward,
A simple control system using solenoid valves, an
pulling the pile section and creating a reaction load
electrical air compressor with an accumulator tank
against the timber mat at the ground surface. Care
of 22 litres and a control computer was developed
was taken in centering the tension loading rod and
to obtain more accurate load regulation. The air
the frame in order to avoid load eccentricities. In the
compressor and accumulator tank were
current application a working pressure of 100 - 200
mechanically limited to a constant target pressure.
kPa and an initial lifting height of 8 - 10 cm in each
Based on the load cell measurements the control
bag conservatively leads to up to 150 kN load
loop implemented in LabView sent a digital signal
capacity with a spring stiffness of 3 kN/cm for
to a relay that controlled the solenoid valves in
constant pressure loading (see Figures 3 and 7).
between the compressor and the lifting bags. The
system was designed to adjust any arbitrary load
variation resulting from the pile head displacement
or changes in ambient temperature. Furthermore, it
is protected against power loss since the solenoid
valves are of the normally-closed type, with zero
differential pressure and very low internal leakage.
Consequently, during a potential power
interruption, the system temporarily converts to a
constant mass system that maintains the load. The
total system cost of compressor, two lifting bags,
valves, load cell, data acquisition (National
instruments NI-cDAQ9178 equipped with two
NI9219 and two NI9237 modules) and control
Figure 6. Loading frame components at ground level. computer is below €3000.- at 2017 price level. All
parts can be re-used afterwards.
106
Figure 8. Constant gas mass loading. Stage 1 for test Pile 2. Figure 9. Constant load scenario. Stage 1 for test Pile 5.
4 RESULTS 6 ACKNOWLEDGEMENTS
Two long-term tests were performed to evaluate the The kind provision of the test site by PEAB AB and
lifting bags performance for pile testing. One test the financial support from Trafikverket in the
used the constant mass (CM) loading scheme while framework Branch samverkan i Grund (BIG) is
the other utilised the constant load (CL) control. The greatly acknowledged.
total duration of the CM and CL test was
approximately 3000 hours and 1200 hours 7 REFERENCES
respectively. Complete details of these tests can be Brown, M., Anderson, W., & Hyde, A. 2004. Statnamic
found in Yannie (2016). Here the most relevant load testing of model piles in a clay calibration chamber. Int.
stages are compared to demonstrate the Journal of Phys. Modelling in Geot., 4(1), 11-24.
effectiveness of the loading rig.
Fellenius, B.H. 1972. Down-drag on piles in clay due to
For the CM test, the pile was initially loaded to negative skin friction. Can. Geot. Jnl., 9(4), 323-337.
52.5 kN. The load variation was approximately ∆Q
= 5 kN (9.5% of the initial loading) after 100 hours Janes, M.C., Horvath, R.C., & Bermingham, P.D. 1991.
as shown in Figure 8. Figure 9 shows the CL pile An innovative dynamic test method for piles. In Intl.
test, which was initially loaded to 54 kN. This load Conf. on recent advances in geotechnical earthquake
was applied for approximately 45 hours before the engineering an soil dynamis, paper 2.
next load stage. The initial tolerance of the control
system allowed a load variation of ±1 kN, which Jardine, R.J., Zhu, B., Foray, P., & Dalton, C. P. 2009.
Experimental arrangements for investigation of soil
was further adjusted to ±0.5 kN by decreasing the
stresses developed around a displacement pile. Soils
allowed tolerances in the control system. and foundations, 49(5), 661-673.
5 CONCLUSIONS Karlsrud, K., Jensen, T.G., Lied, E. K.W., Nowacki, F.,

A new non-standard mechanical loading framework & Simonsen, A. S. 2014. Significant ageing effects for
has been developed for in-situ testing of pile axially loaded piles in sand and clay verified by new
field load tests. In Offshore Technology Conference,
elements in tension up to 500 kN (4 lifting bags).
Paper number OTC-25197-MS.
This setup proves to be robust, cost-effective and
easy to adapt to different loading regimes and Rausche, F., Goble, G.G., & Likins Jr, G.E. 1985.
anticipated pile head displacements. The variation Dynamic determination of pile capacity. Journal of
of the applied load in time for the constant mass Geotechnical Engineering, 111(3), 367-383.
solution (±10% Q) has been further improved (<
±1%Q) with the addition of a simple control system Nicola, A.D., & Randolph, M. F. (1999). Centrifuge
composed of solenoid vales, a portable electric air modelling of pipe piles in sand under axial loads.
compressor and an electronic control system. Géotechnique, 49(3), 295-318.
107
Yannie, J. On the long-term behaviour of tension loaded Zhang, C., White, D., & Randolph, M. (2010). Centrifuge
piles in natural soft soils. PhD Thesis Chalmers modeling of the cyclic lateral response of a rigid pile in
University of Technology, Gothenburg, Sweden soft clay. Journal of Geotechnical and
Geoenvironmental Engineering, 137(7), 717-729.
108
Chromatic aberration when using white LEDs for PIV: issues and remedies
Mir Amid Hashemi1, Jingmin Xu1, Charles M. Heron1, Alec M. Marshall1
Department of Civil Engineering, University of Nottingham, UK
1
Corresponding author: Mir Amid Hashemi (ezzmh2@nottingham.ac.uk)
ABSTRACT: Particle Image Velocimetry (PIV) is a powerful tool used in geotechnical centrifuge
modelling to measure displacement and deformation fields in a 2D plane of a model container. Usually, this
plane is a lateral wall of the model held by a relatively thick block of transparent acrylic. Consequently,
light passing through the acrylic wall undergoes refraction at an angle that is dependent on the wavelength
of the light. The varying levels of refraction, which are of greatest impact near the corners of images where
incident angles are largest, can result in images suffering from something similar to a radial blur, which
affects image quality and PIV analysis performance. Common practice in centrifuge testing is the use of
white light emitting diodes (LEDs), which are a combination of red, green and blue LEDs, hence containing
light of a variety of wavelengths. This paper examines the issues related to the use of white LEDs and
provides an assessment of PIV performance when using white and monochromatic LEDs. Two experiments
were carried out using dry sand; the first with a rigid body movement simulating a no shear strain
displacement field, and a second test with sub-surface ground displacements from a greenfield tunnel
volume loss test. PIV performance is quantified using the variation of local shear strain in both experiments
and a comparison of displacements in the second experiment. Results show a higher level of noise error on
the images taken with white LEDs compared with those taken with the monochromatic amber LEDs. It is
recommended that LEDs with a single-wavelength colour that closely matches the colour of the medium
(i.e. soil) should be used as this will be the wavelength that will be reflected the most by the medium.
Keywords: Particle Image Velocimetry; Chromatic Aberration.
1 INTRODUCTION where 𝑛𝑛 is the refraction index between air and

Particle Image Velocimetry (PIV) is a widely used PMMA acrylic and 𝜆𝜆 is the wavelength expressed in
tool to measure the displacement field between two µm.
images of a deforming object. The geotechnical
centrifuge research community has benefitted from
the GeoPIV tools described by White et al. (2003),
which was later improved by Stanier et al. (2016)
under the name GeoPIV-RG.
The methodology used is to allow light to pass
through one side of the centrifuge model by means
of a transparent material usually made of acrylic. To
maintain a rigid boundary condition and keep
structural integrity, the thickness of the acrylic wall
is usually relatively thick (usually 50 to 100 mm).
A side effect of such thickness is the impact on
quality of the image resulting from refraction of
light for different wavelengths. It is well known that
light refracts when it enters into different media. Figure 1. Dispersion curve for PMMA acrylic.
However, the index of refraction depends on the
wavelength of the light. This dependency is known This dispersion affects the quality of the images
as the dispersion curve. For acrylic (PMMA) the through the effect of chromatic aberration. An
relationship between the refraction index and the illustration is shown in Figure 2 where an object
wavelength is described by Eq. 1 (Sultanova et al. behind a thick acrylic wall is seen by the camera as
2009), with the relationship also presented in if it is in different locations for different colours.
Figure 1. The resulting image suffers from an effect similar to
radial blur and the sharpness and quality of the soil
1.1819𝜆𝜆2
𝑛𝑛2 − 1 = 𝜆𝜆2 −0.011313 (1) texture is reduced. Moreover, control markers
(especially in the corners) show the same effect and
109
the accuracy of the detection of their centroid is 2 METHODOLOGY
diminished. An example of such an effect is shown Two series of tests were carried out. The first test
in Figure 3 where the same corner region of an involved a rigid body movement of a textured plate,
image is compared between lighting with white whereas the second test included a typical
LEDs (Figure 3a) and monochromatic LEDs greenfield tunnelling volume loss scenario, both of
(Figure 3b); the image in Figure 3a is visually more which were carried out at 1g (i.e. not in the
blurry than that in Figure 3b. centrifuge).
A 100mm thick acrylic wall was used for both
Acrylic wall tests. The camera was placed perpendicular to the
Camera wall at a distance of 200mm. The camera was a
Teledyne DALSA Nano M4020 with a resolution of
4112 x 3008 and a Tamron 1.1" 8mm F1.8 lens. The
acrylic wall coordinate system was fixed using
Real object control markers (black dots over white squares; see
Figure 4), with control marker locations determined
using a sheet of photogrammetric targets. The
Apparent object images were saved in the lossless TIFF format.
For both experiments, PIV analyses were carried
Figure 2. Chromatic aberration effect (exaggerated) of a out on specific areas using an in-house MATLAB
white object and its appearance on the camera. code based on Ncorr (Blaber et al. 2015) that uses
the inverse compositional Gauss-Newton method
To improve the image quality, it is therefore for subpixel correlation similar to GeoPIV-RG
beneficial to use lighting that outputs a sharp (Stanier et al. 2016). The analyses used patches
spectral response, such as monochromatic LEDs, to measuring 4141 pixels and spaced 10 pixels apart
decrease chromatic aberration. This paper provides both horizontally and vertically.
an assessment of the impact of using white LEDs
versus monochromatic LEDs on the PIV 3 RIGID BODY MOVEMENT
performance for a relatively standard geotechnical
centrifuge test setup. The first experiment was made using fine sand
(Leighton Buzzard Fraction E sand with an average
grains size of 𝑑𝑑50 = 0.122mm) glued on an
aluminium plate. The plate was placed against the
acrylic wall and moved using a sub-micron
precision actuator. The plate was moved in a
rectangular pattern, as shown in Figure 4, starting
from the bottom right, moving anti-clockwise, with
20 steps (images) taken at the locations indicated by
an ‘*’ in Figure 4 and detailed in Table 1, with the
coordinate system indicated in Figure 4.
For this study, the main region of interest is where
the effects of chromatic aberration are prevalent
(near the outer edges of the image), hence the area
denoted by the dashed white line in Figure 4 was
considered.
Since the plate moves as a rigid body, it does not
undergo deformation. Any deviation from zero
deformation comes from measurement errors and is
a good indicator for quantifying the quality of the
PIV analysis.
Figure 3. Example of the same picture taken with (a) white

LEDs, and (b) monochromatic amber LEDs.
110
Figure 4. Rigid body movement of a plate following the path
drawn in black; the region of interest for PIV tracking is shown
with the dashed white line. Images are taken in locations
indicated by a ‘*’.
Table 1. Displacement path of the rectangular plate.

Step x-displacement y-displacement
number (mm) (mm)
0 0 0
1 0 10 Figure 5. Shear strain comparison between white LEDs (a)
2 0 20 and amber LEDs (b).
3 0 30
To quantify the level of variability in shear strain,
4 0 40
a variance filter with a 3x3 square was applied on
5 -10 40
the shear strain field. This calculates how much
6 -20 40 every patch varies from its neighbouring patches. A
7 -30 40 histogram of local variances is shown in Figure 6,
8 -40 40 where it is shown that variability in the shear strain
9 -60 40 is higher in the case of white LEDs compared to the
10 -85 40 amber LEDs.
11 -110 40
12 -110 20
13 -110 0
14 -85 0
15 -60 0
16 -40 0
17 -30 0
18 -20 0
19 -10 0
20 0 0
Figure 5 shows the shear strain calculated from

the displacement field between step 0 (initial image)
and step 4. Results show that, on average for all Figure 6. Histogram of local variance of strain for white
displacement fields, a higher variability of shear LEDs (dashed line) and amber LEDs (solid line).
strain for the white LEDs than for the
monochromatic amber LEDs.
111
The average of this variance filtered field is then
taken to quantify the variability level into a scalar
value for each image. This scalar is shown in Figure
7 for each image and for both white LEDs and
amber LEDs. Results demonstrate that images lit
with monochromatic amber LEDs performed better
than those lit with white LEDs, with 30% less
variance on average.
Figure 8. Tunnel volume loss experiment with two regions

chosen for comparison. A-A’ is the section used for
displacement measurement shown in Figure 10.
The same process of shear strain variability

quantification as with the rigid body movement test
was carried out. Results of local variance are shown
in Figure 9a for the Top region and Figure 9b for the
side region. Figure 9 shows that there is a consistent
Figure 7. Average of local variance of shear strain for all
reduction of shear strain variability using amber
images compared between white LEDs and amber LEDs. LEDs as opposed of white LEDs.
4 TUNNEL VOLUME LOSS

A typical greenfield tunnel volume loss experiment
conducted at 1g was also carried out to determine
the effect of white versus monochromatic LEDs.
The greenfield tunnelling experiment consisted of a
plane-strain strongbox with a 100mm thick acrylic
front wall and a cylindrical model tunnel running
from the front wall through the soil to the back wall
of the container. The model tunnel consisted of a
water-filled latex membrane sealed onto a rigid
aluminium shaft with enlarged ends that were fixed
within the front and back walls of the strongbox (see
Zhou et al. 2014 for full details). Tunnel volume
loss was imposed by extracting a known volume of
water from the model tunnel, and is defined as the
volume of water (per unit length) extracted from the
tunnel divided by the initial cross-sectional area of
the tunnel. The same fine sand from the rigid body
movement test was used and was prepared to a loose
state (relative density of approximately 30%). Two
regions were compared, as shown in Figure 8. The
top region is located over the tunnel where
maximum displacement occurs. The side region is
located where a minimal amount of displacement
occurs. A series of tunnel volume losses of 0.1%,
0.2%, 0.3%, 0.4%, 0.5%, 0.6%, 0.8% and 1% were
imposed. For each volume loss, two images were Figure 9. Average local variance of shear strain for tunnel
taken, one with white LEDs and another with amber volume loss experiment for (a) top region, and (b) side region.
LEDs.
112
In order to assess the difference in PIV light source can provide better images. Two tests
performance in terms of displacements, a were carried out with images taken through a
comparison of vertical displacement is shown in 100mm thick acrylic wall and using either white
Figure 10 for the soil along a horizontal line 100mm light LEDs or monochromatic amber LEDs. The
above the tunnel centre (see section A-A’ in Figure first test included a sand-coated plate that was
8). The x axis is the horizontal distance from the moved as a rigid body along a rectangular pattern
centre of the tunnel. It can be seen that the two near the edge of the image. The second test was a
curves closely match each other when the typical greenfield tunnel volume loss experiment.
measurements are made around the centre of the Both experiments showed a lower level of
camera. However, there is a small difference variability on the shear strain measurement when
between -150mm and -100mm where the data is using monochromatic LEDs rather than white
close to the edge of the camera (see Figure 8). This LEDs. The difference in measured displacements
difference, which is approximately 0.01mm, is very was found to be small, however it may be important
small, however it still may be important when in tests where displacements or strains are needed to
attempting to determine displacements or strains be measured with high levels of accuracy.
with great accuracy.
6 REFERENCES
Blaber, J., Adair, B. and Antoniou, A. 2015. Ncorr:
Open-Source 2D Digital Image Correlation Matlab
Software. [online] Experimental Mechanics, 55(6):
1105–1122. Available from
http://doi:10.1007/s11340-015-0009-1.
Stanier, S.A., Blaber, J., Take, W.A. and White, D.J.

2015. Improved image-based deformation
measurement for geotechnical applications. [online]
Canadian Geotechnical Journal, 53: 727–739.
Available from http://doi:10.1139/cgj-2015-0253.
Sultanova N.G., Kasarova S.N. and Nikolov, I.D. 2013.

Characterization of optical properties of optical
polymers. [online] Optical and Quantum Electronics,
Figure 10. Vertical displacement along horizontal line (section 45: 221-232. Available from http://doi:
A-A’ in Figure 8) 100mm above tunnel axis at 0.3% tunnel 10.1007/s11082-012-9616-6.
volume loss.
White, D.J., Take, W.A. and Bolton, M.D. 2003. Soil
5 CONCLUSIONS deformation measurement using particle image
This paper demonstrated the effect of chromatic velocimetry (PIV) and photogrammetry.
Geotechnique, 53(7): 619–631.
aberration on the quality of images used for particle
image velocimetry analysis under relatively typical Zhou, B., Marshall, A. M. and Yu, H. S. 2014. The effect
conditions for geotechnical centrifuge testing. of relative density on greenfield settlements above
Results showed that using a light source that has a tunnels in sands. In Geoshanghai 2014 - International
wide colour spectrum range (e.g. white light) affects conference on geotechnical engineering, Shanghai:
the sharpness of images and that a monochromatic ASCE. 96–105.
113
Physical Model of Rainfall Induced Landslide in Flume Test: Preliminary Results

Željko Arbanas1, Vedran Jagodnik1, Josip Peranić1, Sara Pajalić1, Martina Vivoda Prodan1,
Nina Čeh1
1
Faculty of Civil Engineering, University of Rijeka, Croatia
Corresponding author: Željko Arbanas (zeljko.arbanas@gradri.uniri.hr)
ABSTRACT: Physical modelling of landslides using scaled landslide models behavior began in 1970s in
Japan at a scaled natural slope physical model. The laboratory experiments of landslide behavior in a scaled
physical model (also called flume or flume test) started in 1980s and 1990s in Canada, Japan and Australia
under 1g conditions. The main purpose of the landslide physical modelling in the last 25 years was research
of initiation, motion and accumulation of fast flow like landslides caused by infiltration of water in a slope.
In October 2018, at the Faculty of Civil Engineering University of Rijeka, started a four-year research
Project Physical modelling of landslide remediation constructions behavior under static and seismic actions,
funded by the Croatian Science Foundation. The main Project aim is the modelling of landslide remedial
constructions’ behavior in physical models of scaled landslides in static (rainfall triggered landslides) and
seismic conditions (earthquake triggered landslides) and their combination under 1g conditions. In this
manuscript we will present the preliminary results obtained in landslide initiation test of a sandy slope
(constructed of 0-1.0 mm the Drava River sand) exposed to an artificial rain typical for local conditions in
Croatia by rainfall simulator. The results of landslide development were monitored by observation of
volumetric water content and pore water pressure as well as by of surface displacement by structure from
motion (SfM) surface observation displacement monitoring inside the model displaced mass. In this paper
the preliminary results of provided test will be presented related to initiation and development of the
observed instability of the sandy slope model.
Keywords: Landslide; Flume Test; Sandy Slope; Artificial Rainfall; Monitoring.
1 INTRODUCTION 1982; Yagi et al. 1985; Yamaguchi et al. 1989) on

Landslides are one of the major geohazards that natural slopes exposed to artificial rainfall. The
threaten vulnerable areas in mountains, cities, along laboratory experiments of landslide behaviour in a
riverbanks, coasts and islands causing severe scaled physical model (also called flume or flume
consequences expressed in lost human lives and test) started in 1980s and 1990s in Canada (Hunger
economic losses all around the world. Due to and Morgenstern 1984), Japan (Yagi et al. 1985)
climate change we are witnesses of an increase in and Australia (Eckersley 1990) under 1g conditions.
the frequency and/or intensity of heavy rainfall, as The research Project Physical modelling of
well as shift of locations and periodicity of rainfall, landslide remediation constructions’ behaviour
which significantly increase landslide risk in under static and seismic actions, funded by the
landslide prone areas. Urban expansion due to Croatian Science Foundation, started in October
population growth, development of coastal and 2018 at the Faculty of Civil Engineering, University
mountain areas, construction of roads and railways of Rijeka, Croatia. The main Project aim is
increase exposure to the landslide hazard. Strong modelling behaviour of landslide remedial
earthquakes are the second triggering factor for constructions in physical models of scaled
numerus rapid and long runout landslides in landslides in static and seismic conditions (Arbanas
seismically active areas. et al. 2019). In this manuscript we will present the
The answers of scientists, engineers, and all others initial results obtained in landslide initiation test of
included in landslide risk reduction is in significant a sandy slope exposed to an artificial rain typical for
developing of landslide science. Some of the local conditions in Croatia by rainfall simulator. The
answers are in different types of modelling of results of landslide development were monitored by
landslide related researches especially in landslide observation of surface displacement by structure
triggering and landslide process evolution. from motion (SfM), terrestrial laser scanner (TLS)
Landslide modelling for a long time was based on and a pair of high speed cameras, as well as by pore
only numerical modelling techniques. Physical water pressure and displacement monitoring inside
modelling of landslide analysing small scale the model displaced mass. While analyzes of the
landslide models behaviour was started in 1970s observed small scale landslide movements are still
and 1980s in Japan (Oka 1972; Kutara and Ishizuka ongoing, in this paper we will describe the surface
115
observations of landslide movements and compactor at the water content of w = 2%. The
conclusions that followed from movement analyses. specific gravity of sand is Gs = 2.7.
The strength properties of the Drava River Sand
2 MATERIALS AND METHODS was determined in direct shear apparatus at the same
Physical model of a scaled landslide under static relative density as it is in the flume of Dr = 0.5 and
actions caused by artificial rainfall was designed to low normal stresses appropriate to the stresses
enable initiation of a landslide caused by controlled present in the flume slope. The average friction
artificial rainfall and equipped with adequate angle is φ = 34.9o.
photogrammetric equipment and complex sensor The hydraulic conductivity of the Drava River
network with ability to measure displacements, soil Sand at the same relative density as it is in the flume
moisture, pore pressures and soil pressures within a of Dr = 0.5 determined in oedometer at low normal
landslide body (Figure 1a). The flume is constructed stresses appropriate to the stresses present in the
as a steel base construction with transparent flume slope is k = 5x10-5 m/s. Measured matric
Plexiglas sides in layout dimensions of 2.30 x 1.00 suction at the water content of w = 2% varies from
m that lies on the working table (Figure 1b). It was 6 to 8 kPa.
provided that the maximal height of the soil material
in the slope model is to 30 cm. An inclination of the 2.2. Monitoring and observations
flume slope can be adjusted for a type of used soil Monitoring system established at the flume test
material from 10o to 45o. To prevent a sliding of soil followed the principles used in observation of real
mass at the contact of the flume base, the geogrid is landslides; it was consisted of geotechnical
fixed to the flume base. monitoring systems and geodetic monitoring
systems (Mihalić Arbanas and Arbanas 2015).
2.1. Slope construction Geodetic monitoring is based on innovative
For the first test in the constructed flume, the fine photogrammetric equipment for multi-temporal
sandy material was chosen to be built-in in the landslide analysis (Zanuta et al. 2006) from stereo
flume slope at an inclination of 30o. Fine graded 0- image sequences obtained by a pair of high speed
1.0 mm (Figure 2) the Drava River Sand (D60 = 0.37 cameras (Feng et al. 2016). The high speed cameras
mm, D10 = 0.19 mm, Cu = 1.947) was built-in in the have characteristics of max 2000 fps, resolution of
flume in 6 cm height 5 layers, 30 cm in total, at 1280 x 1024 pixel and ISO: 6000 ISO Mono 2000
relative density Dr = 0.5 and reached void ratio of e ISO Color.
= 0.78. Each layer was compacted by a manual
a b
Figure 1. a Schematic view of physical model: 1-tensiometers, 2-pore pressure transducers, 3-strain gauges, 4-accelerometers, 5-
rainfall simulator (sprinkler system), 6-high speed cameras, 7-terrestrial laser scanner (TLS), 8-infrared camera (Arbanas et al.
2019); b Photo of physical model.
116
Figure 2. Grain size distribution curve of the Drava Sand built-
in in the slope.
Terrestrial laser scanner (TLS) as well as

structure from motion (SfM) photogrammetry
surveys enabled determination of reached 3D
landslide surface or high resolution digital model of
a landslide model before and after sliding very Figure 3. Photo of one branch with four axial-flow full-cone
precisely (Bitelli et al. 2004). The TLS Faro Focus spraying nozzles.
HD 130 with the range of 0.6 – 130 m and resolution
of 70 Mpixel as employed for the model slope pressure valve. The minimal intensity that can be
surface scanning. achieved with only one nozzle at each branch is 10
Geotechnical monitoring is consisted of a mm/hour while ta maximal intensity can reach more
complex network of miniature sensors equivalent to than 600 mm/hour using 4 nozzles at each branch.
those used in geotechnical monitoring in the field Before the use, the rainfall simulator system was
(Mihalić Arbanas and Arbanas 2015). Built-in tested and calibrated in different regimes to enable
sensors have ability to measure displacements adequate simulation of real precipitations applied to
MEMS (Ooi and Wang 2014), moisture and pore the model slope.
water pressures (pore pressure and moisture
transducers), soil pressures and contact forces (load 3 RESULTS AND DISCUSSION
cells, strain gauges) and soil suction (tensiometers) After the model slope construction and installation
(De Dios et al. 2010; Ooi et al. 2014; Scaioni et al. of the monitoring equipment in the slope, the test
2013; Lu et al. 2015). All sensors used in the test are was conducted by exposing the slope to the artificial
connected to data loggers for continuous data rain from three nozzles, one in the upper part of the
collection during the test. slope (H), one in the middle of the slope (M), and
one in the foot of the slope (L) reaching an uniform
2.3. Rainfall simulator artificial rainfall intensity of 58.23 mm/hour. The
One of the most important issues in physical models same intensity was maintained during the test 121
to simulate landslide initiation and motion caused min, when the intensity was increased on 121.52
by rainfall infiltration is full controlled intensities of mm/hour and after 7 minutes (necessary for
artificial rainfall. Although it is possible to use some intensity adjustment using water pressure valve)
of the already developed rainfall simulators (Iserloh reduced on 81.01 mm/hour that as maintained to the
et al. 2012, 2013; Lora et al. 2016), there was end of the test, Figure 4.
necessity for developing of a rainfall simulator that The first movements in the model slope were
can be adjusted to reach the rainfall intensities registered after 141 minute (126 minutes after the
characteristic for precipitation events in Croatia. rainfall started) Figure 5a. The landslide initiation
Construction of rainfall simulator includes ordinary was started 11 minutes after complete saturation of
elements usually used for watering plants and has an the slope foot (115 min after the rainfall started),
ability to control intensity, flow, and uniformity of and landslide occurrence was presented in complete
wetting as well as the time of rainfall simulation. loss of soil strength in the slope toe and
Rainfall simulator is consisted of a system of tree development of main scarp in the slope material,
sprinkler branches with four different axial-flow Figure 5a. It was noted that after the movement, the
full-cone nozzles (Figure 3). The supply of the ground water started to run out through the sliding
system is controlled by flow meters and water mass.
117
After the landslide initiation, the further
instability distribution was retrogressive with
developing of new scarps 135, 139 (Figure 5b), and
148 (Figure 5c) minutes after rainfall started.
Figure 4. Diagram of rainfall intensity during the test. The

rainfall started after 15 minutes of sensors adjustment.
a b
c
b
Figure 6. Photos of landslide final retrogressive development.

a Landslide retrogressive development (149 minutes after the
c rainfall started); b Initiation of the debris flow from the top of
the slope (149 minutes after the rainfall started); c Final
landslide development to the top of the slope (150 minutes after
the rainfall started).
After developing of the fifth crack, the rapid

spread of the landslide to the top of the slope as
occurred in next two minutes including an initiation
of a debris flow originated from the crest of the
slope (Figure 6 a,b,c). In this moment (150 minutes
after the rainfall started), the rainfall was suspended
that affected stopping of landslide activity. While
Figure 5. Photos of landslide initiation and retrogressive this test was conducted with the task to identify
landside development. a Landslide initiation (126 minutes after conditions of a landslide initiation and mechanism
the rainfall started; b Development of the third and fourth main and type of landslide occurrence through the
cracks (139 minutes after the rainfall started); c Development reached level of landslide development, final
of the fifth main crack (48 minutes after the rainfall started).
collapse of the slope was not enabled.
118
Analysis of measured data of the volumetric water a
content during the test in the upper (H), middle (M),
and lover (L) part of the slope enables the
reconstruction of the landslide development, Figure
7 a, b and c.
In the first stage of the test the water infiltration in
the slope started and the rise of volumetric water
content was faster in shallower layers of the slope in
all three parts of the slope (Figure 7 a, b and c). In
the lowest and middle part of the slop it is clearly
noticeable that the upper layers reached constant
values of volumetric water content around 0.2
m3/m3 from 5 to 30 min after rainfall starting in the
foot of the slope and from 10 to 40 min in the middle
part of the slope. In the top of the slope, the same
values of volumetric water content around 0.2
m3/m3 form 5 to 20 min and stayed almost constant
to the end of the test. In the same time, the initial
suction in the top of the slope took the constant b
value of 11.5 kPa (Figure 7c). The difference in
these volumetric water contents is caused by
establishing of a seepage trough the slope, and by
increasing of hydraulic conductivity caused by
suction decreasing and saturation of the slope from
bottom layers to the surface.
An increased inflow and rising of the ground
water level in the foot of the slope begun 30 min
after rainfall started that is evident at the deepest
sensor (L at -24 cm), and this process was continued
till the ground water level flooded the surface of the
slope foot (117 min after the rainfall started), Figure
8a, maintaining through the test. The rising of the
ground water level in the middle of the slope begun
30 minutes later, Figure 7b. A ground water level
was not identified at the top of the slope; infiltrated
water was drained to the lower slope parts.
The failure of the slope was occurred immediately c
after the ground water level flooded the surface of
the slope foot when that caused decreasing of soil
strength and sliding (Figure 5a) and further
retrogressive landslide development to the top of the
slope. Finally, in the moment when the slope was
completely saturated, the rain intensity was too high
for further water infiltration and forming of the
debris flow from the top of the slope, Figure 8.
Although the analyses of spatial and temporal
movements data is still ongoing in the moment of
the manuscript preparation, and the results of these
analyses are not a subject of this paper, it should be
noted that applied methods for model surface
observations using photogrammetric equipment for
multi-temporal landslide analysis from stereo image
sequences obtained by pair of high speed cameras,
Figure 7. Diagrams of measured values of volumetric water
terrestrial laser scanner (TLS) as well as structure content, suction and temperature during the test. (a) In the foot
from motion (SfM) photogrammetry surveys have of the slope; (b) In the middle of the slope; (c) In the top of the
slope. Numbers in parentheses designate depth of a sensor.
119
result of landslide model surface motions. Stereo
image sequences obtained by pair of high speed
cameras enable a base for very precise temporal
analysis of surface displacements but the method is
depended on surface grid mark point density.
Terrestrial laser scanner (TLS) survey can ensure
the most precise 3D point cloud of model slope
surface, but restrictions are related to time period
necessary to complete a survey. An extended time
necessary for scanning of about 30 min per scan,
points on possibility to use TLS in extremely slow
sliding processes or at the start and end of a test.
Structure from motion (SfM) photogrammetry
surveys is less precise than TLS survey, but using
Figure 8. Debris flow occurred at the end of the test. the high resolution cameras on small slope models,
the obtained point clouds are not of significantly
an important role in understanding and explanation less quality. In a relatively not so fast sliding
of landslide initiation in a slope model. processes, a series of from a handmade high
The sliding process in the sandy slope model in resolution photos would provide a high quality
previously presented conditions was slower than temporal 3D point cloud. As an example of the TLS
expected, excepting the debris flow process at the and SfM 3D point clouds at the end of described
end of the test. The preliminary analyses have test, without of detailed analyses of their
shown that an integrated approach using all three differences, are presented in Figure 9.
surface observation methods would give an optimal
a b
Figure 9. A view at the 3D point cloud provided form a Terrestrial laser scanner (TLS) survey; b Structure from motion (SfM)
photogrammetry at the end of the sliding process.
120
4 CONCLUSION laboratory equipment used for laboratory testing
In this manuscript we presented the preliminary was provided in the frame of Project Research
results obtained in landslide initiation test of a sandy Infrastructure for Campus based Laboratories at the
slope (constructed of 0-1.0 mm the Drava River University of Rijeka, co-funded in a part by the
sand) exposed to an artificial rain typical for local Ministry of Science, Education and Sports of the
conditions in Croatia by rainfall simulator. The Republic of Croatia and the European Fund for
results of landslide development were monitored by Regional Development (ERDF). The authors also
observation of volumetric water content, soil want to thank to laboratory technicians Juraj Stella
suction and displacement monitoring within the and Duje Kalajžić and their help in test conducting,
displaced mass model. and TLS and SfM surveys.
Under artificial rainfall the model slope built-in in
sandy slope started to slide after 126 minutes, after 6 REFERENCES
the ground water level flooded the surface of the Arbanas, Ž., Pajalić, S., Jagodnik, V., Peranić, J., Vivoda
slope foot 117 min after the rainfall started, and the Prodan, M., Đomlija, P., Dugonjić Jovančević, S., Čeh,
landslide as developed to the top of the slope. N. 2019. Development of physical model of landslide
Finally, in the moment when the whole slope was remedial constructions’ behaviour. In: Proceedings of
the Proceedings of 4th Regional Symposium on
completely saturated, the rain intensity was too high
Landslides in Adriatic Balkan Region, Sarajevo,
for further water infiltration and forming of the Bosnia and Herzegovina, 23-25 October 2019.
debris flow from the top of the slope. Geotechnical Society of Bosnia and Herzegovina,
The landslide initiation in the foot of the model Sarajevo, pp. 103-108.
slope, as well as further developing of landslide to
the top of the model slope was explained based on Bitelli, G., Dubbini, M., Zanutta, A. 2004. Terrestrial
results of the volumetric water content and soil laser scanning and digital photogrammetry techniques
suction observation during the test. to monitor landslide bodies. In: Proceedin of XXth
Although the analyses of spatial and temporal ISPRS Congress: Geo-Imagery Bridging Continents,
movements data is still ongoing in the moment of Istanbul, Turkey, 12–23 July 2004. International
Archives of the Photogrammetry, Remote Sensing and
the manuscript preparation, it should be noted that
Spatial Information Sciences, Vol. XXXV, Part 5, pp.
applied methods for model surface observations 246–251.
using photogrammetric equipment for multi-
temporal landslide analysis from stereo image De Dios, R.J., Enriquez, J., Victorino, F.G., Mendoza,
sequences obtained by pair of high speed cameras, E.A., Talampas, M.C., Marciano, J.J.S.Jr. 2010. A tilt,
terrestrial laser scanner (TLS) as well as structure soil moisture, and pore water pressure sensor system
from motion (SfM) photogrammetry surveys have for slope monitoring applications. Science Diliman,
an important role in understanding and explanation 21: 15-27.
of landslide initiation in a slope model. The
preliminary analyses have shown that an integrated Eckersley, D. 1990. Instrumented laboratory flowslides.
Géotechnique, 40:489–502.
approach using all three surface observation
methods would give an optimal result of landslide Feng, T., Mi, H, Scaioni, M., Qiao, G., Lu, P., Wang, W.,
model surface motions. Tong, X., Li, R. 2016. Measurement of Surface
Further analyses of data provided by conducted Changes in a Scaled-Down Landslide Model Using
test as well as results of further test conducted in High-Speed Stereo Image Sequences. Photogrammetry
different slope conditions under different rainfall Engineering and Remote Sensing, 82(7): 547-557.
regimes will give better insights in landslides
initiation of sandy slopes exposed to an artificial Hunger, O., Morgenstern, N.R. 1984. Experiments on the
rainfall. flow behaviour of granular materials at high velocity in
an open channel. Géotechnique, 34:405–413.
5 ACKNOWLEDGEMENTS Iserloh, T., Fister, W., Seeger, M., Willger, H., Ries, J.B.
The research presented in this paper was supported 2012. A small portable rainfall simulator for
by Croatian Science Foundation under the Project reproducible experiments on soil erosion. Soil &
IP-2018-01-1503 Physical modelling of landslide Tillage Research, 124: 131–137.
remediation constructions behaviour under static
and seismic actions (ModLandRemSS), and Iserloh, T., Ries, J.B., Arnaez, J., Boix-Fayos, C., Butzen,
University of Rijeka under the Project Uniri- V., Cerda, A., Echeverria, M.T., Fernandez- Galvez, J.,
Fister, W., Geissler, C., Gomez, J.A., Gomez-
Tehnic-18-113 Laboratory research of static and
Macpherson, H., Kuhn, N.J., Lazaro, R., Leon, F.J.,
cyclic behaviour at landslide activation. These Martinez-Mena, M., Martinez-Murillo, J.F., Marzen,
supports are gratefully acknowledged. The part of
121
M., Mingorance, M.D., Ortigosa, L., Peters, P., Ooi, G.L., Wang, Y.H. 2014. Applying MEMS
Reguees, D., Ruiz-Sinoga, J.D., Scholten, T., Seeger, Accelerometers to Measure Ground Vibration and
M., Sole-Benet, A., Wengel, R., Wirtz, S. 2013. Characterize Landslide Initiation Features in
European small portable rainfall simulators: a Laboratory Flume Test. In: Proceedings of Geo-
comparison of rainfall characteristics. Catena, 110: Congress 2014, Atlanta, Georgia, February 2014,
100–112. ASCE, pp. 2019-2028.
Kutara, K., Ishizuka, H. 1982. Seepage flow in the Ooi, G. L., Wang, Y-H., Tan, P. S., So, C. F., Leung, M.
embankment and stability of slope during rain (in L., Li, X., Lok, K. H. 2014. An Instrumented Flume to
Japanese). Tsuchi--Kiso Paper No 1330, Japan. Characterize the Initiation Features of Flow
Landslides. Geotechnical Testing Journal, 37(5): 1-21.
Lora, M., Camporese, M., Salandin, P. 2016. Design and
performance of a nozzle-type rainfall simulator for Scaioni, M., Lu, P., Feng, T., Chen, W., Qiao, G., Wu,
landslide triggering experiments. Catena, 140: 77-89. H., Tong, X., Wang, W., Li, R. 2013. Analysis of
spatial sensor network observations during landslide
Lu, P., Wu, H., Qiao, G., Li, W., Scaioni, M., Feng, T., simulation experiments. European Journal of
Liu, S., Chen, W., Li, N., Liu, C., Tong, X., Hong, Y., Environmental and Civil Engineering, 17(9): 802-825.
Li, R. 2015. Model test study on monitoring dynamic
process of slope failure through spatial sensor network. Yagi, N., Yatabe, R., Enoki, M. 1985. Laboratory and
Environmental Earth Sciences, 74(4):3315–3332. Field Experiments on Prediction Method of Occurring
Time of Slope Failure due to Rainfall. Journal of Japan
Mihalić Arbanas, S., Arbanas, Ž. 2015. Landslides: A Landslide Society, 22:1-7_1.
guide to researching landslide phenomena and
processes. In: Handbook of Research on Yamaguchi, I., Nishio, K., Kawabe, H., Sibano, H., Lida,
Advancements in Environmental Engineering. C. 1989. Initiation and fluidization of an artificial
Gaurina-Međimurac, N. (ed.). IGI Global, Hershey, pp. landslide: Field experiment in Yui Shizuoka
474-510. Prefecture, Japan (in Japanese). Shinrin Kosoku Areal
Survey, 158:3–9.
Oka, H. 1972. Impacts by the “artificial landslide”: re-
examine the rage of nature (in Japanese). Kagaku Zanutta, A., Baldi P., Bitelli, G., Cardinali, M., Carrara,
Asahi, 32:152–153. A. 2006. Qualitative and quantitative photogrammetric
techniques for multi-temporal landslide analysis.
Annual Geophysics, 49:1067-1080.
122
3. Piles: installation and capacity
Soil plug investigation with respect to pile geometry and installation method
Sascha Henke
Department of geotechnics, Helmut Schmidt University, Hamburg, Germany
ABSTRACT: Soil plug formation in open-ended piles depends on several factors like installation method,
installation depth and pile geometry as well as soil density. To investigate the influence of these factors on
soil plug formation an extensive centrifuge testing programme has been carried out. During these tests, piles
with different geometry (tubular piles, U-piles and sheet-piles) were installed by quasi-static jacking, cyclic
jacking and impact driving. All tests were carried out in silica sand with different soil density. During
installation the reaction force as well as stress development inside and outside the piles were measured. In
the present paper, the centrifuge testing programme is discussed. Furthermore, exemplary results are
presented in detail. The focus lies on stress development inside the piles as an indicator for soil plugging
tendency. The results are used to develop a simplified approach to predict stress development inside open-
ended piles with respect to the parameters investigated throughout the centrifuge testing programme.
Keywords: Soil Plugging; Centrifuge Testing; Beam Centrifuge; Silica Sand; Pile Installation.
1 INTRODUCTION
Soil plugging in open-ended piles like tubular piles
as well as sheet-piles and H-piles is of special
geotechnical interest as it on the one hand might
increase bearing capacity but on the other might also
increase driving resistance. Therefore, research was
carried out over the last decades to predict the soil
plugging tendency with respect to different
influencing parameters like pile diameter,
installation method, soil conditions etc., see Kishida
(1997), Paikowsky and Whitman (1990), Brucy et
al. (1991), Gavin and Lehane (2003), Jardine et al.
(2005) and Lüking (2010) for example. Figure 1. Schematic setup of the beam centrifuge
It has to be stated that the above mentioned papers
are only exemplary and many other researchers also All tests were executed at an acceleration level of
focused on different aspects regarding soil plugging 50g. Thus, the conversion from model to prototype
and identified many influencing factors like pile scale is done using a factor of N = 50.
diameter, penetration depth, pile geometry, soil The acceleration filed was adjusted that the
conditions etc. acceleration of 50g acted at 1/3 of the pile’s final
To gain deeper insight into all the above penetration depth.
mentioned factors related to soil plug formation an
extensive centrifuge testing programme was carried 2.2. Silica Sand
out which is presented in this paper focusing on the All tests were carried out in very fine dry silica sand,
influence of pile geometry, installation method as a pure sand with a mean grain diameter of
well as soil density. In special, the soil plugging d50 = 0.19 mm. The maximum and minimum void
process is measured directly by means of total ratios are determined as emax = 0.79 and emin = 0.49.
pressure measurement inside and outside the piles.
2.3 Piles
2 TESTING PROGRAMME
Piles with different cross-section were investigated:
In the following section the centrifuge testing • Four tubular piles with different diameter,
programme is described. • three U-profiles (U) with different height
and
2.1. General information • three sheet-piles (SW) with different
All tests were carried out in the geotechnical beam opening angles.
centrifuge at the University of Western Australia,
see Figure 1. The different piles are shown in Figure 2.
125
installation was modelled force-controlled. The
mass of the driving weight is 50 g. Maximum
penetration depth was set to 150 mm.
2.5 Instrumentation
Main focus laid of the development of total stresses
inside and outside the piles. Stress sensors were
located inside and outside the piles 5 mm and
40 mm above the pile toe. The maximum capacity
of the sensors was approx. 1 MPa. Only the small
tubular pile was not instrumented due to space
reasons.
Furthermore, the penetration resistance was
continuously measured at the pile head during
installation (jacking, cyclic jacking). As impact
driving was modelled force-controlled the
penetration resistance was only measured during the
Figure 2. Piles tested in the geotechnical centrifuge
static pile load tests. The general sensor layout for a
tubular pile is depicted in Figure 3.
2.4 Installation methods
All piles are installed by quasi-static jacking with
constant velocity. Furthermore, the tubular pile with
0.83 m diameter in prototype scale as well as the U-
pile with a 0.45 m height in prototype scale are
installed by cyclic jacking as well as impact driving.
2.4.1 Pile jacking

The piles were installed displacement controlled
with constant velocity of 0.2 mm/s (1 cm/s in
prototype scale). The final depth was defined as
150 mm. At 75 mm, 100 mm and 150 mm depth the
installation process was stopped to carry out a static
load test.
2.4.2 Cyclic jacking

Idealising vibratory pile driving the piles were
installed with 0.2 mm/s velocity and an additional
sinusoidal movement of the actuator until 150 mm Figure 3. Measurement layout.
depth was reached. The frequency of the sinusoidal
movement was set to f = 0.25 Hz with an amplitude 2.6 Test preparation
of 1 mm. In one test box in general six or seven piles were
In these tests, the installation process was also installed. The distance between two piles was set to
stopped at 75 mm, 100 mm as well as 150 mm depth at least six times the pile diameter to avoid influence
for static load testing. between different piles.
The tests were carried out in loose as well as dense
2.4.3 Impact driving silica sand. The resulting void ratio was calculated
Impact driving was carried out using the pile driving out of the soil volume and the weight. To verify
hammer developed in de Nicola and Randolph homogenous soil conditions cone penetration tests
(1994). For a detailed description of the pile driving were carried out in advance of the tests.
hammer refer to the above-mentioned paper. The relative soil density ID for loose packings
As installation parameters a driving frequency of ranged between 0.25 and 0.35. The dense soil
10 Hz with a constant falling height of 10 mm (in showed relative densities ID between 0.92 and 0.95.
loose soil) or 20 mm (in dense soil). Thus, the
126
3 RESULTS it is evident that during jacking the stress increase
3.1 Influence of installation method inside the pile is much higher compared to impact
driving. This is due to dynamic effects during
Exemplary, the penetration resistance of a U-profile impact driving which leads to soil compaction as
with a prototype height of 0.45 m during jacking as well as inertia effects such that total stresses inside
well as cyclic jacking is compared in Figure 4. the pile decrease.
For better comparison, the total stresses are
normalized with the initial vertical stress state in the
relevant depth such that a dimensionless factor KPf
is calculated, see Figure 6.
KPf can be interpreted as effective earth pressure
coefficient which indicates soil plugging inside the
profile. If KPf inside a pile is significantly higher
compared to the external value arching inside the
profile leads to soil plugging.
In the present case, it can be seen that after approx.
2 m penetration, a constant KPf between 9 to 10 is
measured inside the jacked U-profile. For the
impact driven pile the development of KPf is similar
during the first 2 m but afterwards significantly
decreases. This is due to dynamic effects as the plug
which starts to form continuously is loosened due to
Figure 4. Penetration resistance of a U-profile during jacking the dynamic forces.
and cyclic jacking (loose soil ID ≈ 0.3).
It is evident that the penetration resistance is

higher during jacking. Furthermore, it can be seen
that due to the cyclic motion of the U-profile the
penetration resistance continuously changes during
cyclic jacking.
As the pile resistance does not directly indicates
the tendency of soil plugging, in Figure 5 the total
pressures at the pile toe inside and outside the U-
pile are depicted with respect to installation depth.
Figure 6. Normalized total pressure KPf at the toe of a U-

profile during jacking, cyclic jacking and impact driving
(loose soil ID ≈ 0.3)
Figure 5. Total pressure development at the toe of a U-profile
during jacking, cyclic jacking and impact driving (loose soil 3.2 Influence of pile geometry
ID ≈ 0.3)
3.2.1 Tubular piles
First, it can be seen that in all cases the total The main influencing factor regarding the geometry
pressures inside the pile are significantly higher of tubular piles is the pile diameter. Thus, the
compared to the outer side of the U-profile. This normalized resistance (sgr = 0.05D) during the static
indicates that in all cases arching occurs which load tests after 50 mm, 100 mm and 150 mm (3.5 m,
might lead to soil plug development. Furthermore, 5 m, 7.5 m in prototype scale) is compared for
127
jacked piles with different pile diameter, see Figure To justify this assumption, the total stresses at the
7. pile toe for the piles with 0.83 m as well as 1.44 m
diameter in prototype scale are compared in Figure
8. In Figure 8, a sudden increase of total internal
stresses becomes evident for the 1.44 m diameter
pile at a penetration depth of about 5 m. This
indicates soil plug formation at this depth.
3.2.2 Sheet-piles
Furthermore, the influence of the opening angle α of
sheet piles on soil plugging is investigated. For the
definition of the relevant geometric values of a sheet
pile, refer to Figure 9.
Figure 7. Normalized resistance of jacked tubular piles with

different diameter at different penetration depths (loose soil
ID ≈ 0.3)
Figure 9. Relevant geometric values of a sheet pile
It can be seen in Figure 7 that the normalized
resistance is the highest for the pile with diameter As already investigated by Radomski (1968) α is
Di,p = 0.83 m whereas the bigger pile (Di,p = 1.44 m) an important influencing factor regarding soil
has significantly lower normalized resistance at plugging tendency inside sheet piles. In the present
3.5 m and 5 m depth. At 7.5 m depth the normalized study, sheet piles with the geometric values depicted
resistance of the biggest pile increases over- in Table 1 are investigated.
proportionally. This indicates that soil plugging
starts to develop at this depth for the biggest pile Table 1. Geometric values for the different sheet piles
whereas the other piles developed a soil plug at investigated in the present study
shallower penetration depths. name width b1 width b2 opening
angle α
SW 0.50/1.10 m 1.10 m 0.59 m 38.29°
SW 0.54/0.98 m 0.98 m 0.54 m 30.07°
U 0.75/0.38 m 0.75 m 0.75 m 0°
Summarizing, all piles described in Table 1 have

a similar outlined area Au. The penetration
resistance for all sheet piles during jacking into
loose silica sand is plotted in Figure 10.
It can be seen that the penetration resistance for
all sheet piles is similar during the first 3 m of
penetration. Afterwards, differences in penetration
resistance are obvious. The U-pile with α = 0°
develops higher resistance compared to the other
investigated piles. It is obvious that the penetration
resistance increases with decreasing opening angle
from this depth on.
This can be explained as arching is supported by
Figure 8. Total stress at the pile toe over penetration depth
for different tubular piles during jacking (loose soil ID ≈ 0.3) lower opening angles such that the tendency of soil
plug formation decreases with increasing value of α,
see Figure 11.
128
also evident that the depth at which the internal
stresses increase over-proportionally is much
shallower in dense soil (approx. 2 m compared to
5 m in loose sand).
Figure 10. Penetration resistance for different sheet piles

during jacking (loose soil ID = 0.35)
Figure 11. Bearing capacity with respect to the sheet pile’s

opening angle after 5 m of jacking (loose soil ID = 0.35)
3.3 Influence of soil density

As expected, soil density mainly influences
penetration resistance as well as internal stress
development inside open-ended piles. The denser
the sand the higher is the penetration resistance as
well as the total stress development.
Figure 12 shows the penetration resistance Figure 12. Penetration resistance of a 0.83 m diameter
development for the tubular pile with prototype tubular pile during jacking and cyclic jacking into a) loose
diameter of 0.83 m during jacking as well as cyclic silica sand and b) dense silica sand
jacking into loose sand and dense sand respectively.
Furthermore, the total stress development at the Regarding all test data, the quotient of the depths
pile toe inside and outside the 0.83 m as well as the at which the over-proportional stress increase for
1.44 m diameter tubular pile is shown in Figure 13. different soil densities was identified is calculated
It is evident that in both cases total stresses are as
significantly higher if the piles are installed into
𝑧𝑧𝑣𝑣,𝐼𝐼𝐷𝐷 =0,32
dense sand. Regarding the 1.44 m diameter pile it is 𝜔𝜔 = 𝑧𝑧 (1)
𝑣𝑣,𝐼𝐼𝐷𝐷 =0,92
129
validated such that additional data could be
produced using numerical simulations. Due to lack
of space these numerical simulations are not
discussed in the present paper. For further
information regarding the numerical simulations,
refer to Henke (2013).
Resulting out of numerical and experimental data
a simplified approach to calculate KPf with respect
to pile diameter and soil density could be
established.
Regarding jacking of tubular piles the following
mechanical model is considered, see Figure 14.
Figure 14. Mechanical model to calculate soil plug resistance

inside tubular piles
Mainly, two parameters are unknown. On the one

hand KPf and on the other hand the plug height hPf.
First, the work of Randolph et al. (1991) is
considered to estimate the plug height as 10 % to
20 % of the penetration depth such that only KPf has
to be determined.
For estimation of KPf with respect to pile diameter
and soil density the following correlations are
evaluated using numerical and experimental data:
Figure 13. Total pressure development for the a) 0.83 m and
b) 1.44 m diameter tubular pile during jacking into loose and
dense silica sand 𝐾𝐾𝑃𝑃𝑃𝑃 (𝐷𝐷𝑖𝑖 ) = 506.1𝑒𝑒 −3.564𝐷𝐷𝑖𝑖 − 11.63 (2)
With respect to the present tests, this quotient ω and

lies between 2.6 and 3.7 which more or less
corresponds to the quotient of investigated relative 𝐾𝐾𝑝𝑝𝑝𝑝 (𝐼𝐼𝐷𝐷 , 𝐷𝐷𝑖𝑖 ) = 𝐾𝐾𝑃𝑃𝑃𝑃 (𝐷𝐷𝑖𝑖 ) + 0.6894𝑒𝑒 6.42𝐼𝐼𝐷𝐷 − 8.99 (3)
soil densities which is about 2.9. Thus, it can be
assumed that a rather simple mathematical These formulae are valid for 𝐼𝐼𝐷𝐷 > 0.40 and 𝐷𝐷𝑖𝑖 ∈
correlation exists between these two values which [0.40; 0.90].
has to be investigated in further research as the The above mentioned results are graphically
present data basis is too small. visualized in Figure 15 and 16.
It has to be stated that these correlations are
4 SIMPLIFIED CALCULATION MODEL evaluated out of a rather small amount of data
Using the centrifuge test results numerical available such that they have to be detailed in
modelling of pile installation processes was further research work.
130
• Pile geometry significantly influences soil
plugging:
o In tubular piles a decreasing pile
diameter increases soil plugging
tendency
o For U-piles the profile’s height
mainly influences soil plugging.
This means, that the smaller the
height, the higher is the arching
effect inside the pile.
o The opening angle α of a sheet pile
influences soil plugging. Higher
opening angle reduces the internal
stress development.
• The installation method significantly
influences the soil plug formation process.
Figure 15. Correlation between KPf and Di compared with During quasi-static jacking the internal
exemplary test data (ID ≈ 0.30)
stresses are much higher compared to
dynamic installation methods. This leads to
the conclusion that dynamic effects like soil
densification and inertia effects act against
soil plugging.
• As expected, the denser the soil packing the
higher is the internal stress build-up which
acts as an indicator for soil plug formation.
Summarizing, the centrifuge test results provide

valuable data to better understand the mechanism of
soil plugging.
Furthermore, the data was used to validate
numerical simulations of pile installation processes.
Finally, a simplified approach to predict the soil
plug’s capacity is presented which has to be detailed
in further research.
Figure 16. Correlation between KPf and ID compared with 6 ACKNOWLEDGEMENTS

exemplary test data (Di = 0.83 m)
The author thanks the German Research Foundation
(DFG) for funding this project under the number HE
5 CONCLUSIONS 5445/2-1. Furthermore, special thanks belongs to
In the present paper an extensive centrifuge testing the Centre for Offshore Foundation Systems at
programme was presented. In the geotechnical UWA in Perth for their valuable support.
beam centrifuge piles with different geometry were
installed on-flight by means of pile jacking, cyclic 7 REFERENCES
jacking as well as impact driving.
Brucy F., Meinier J., Nauroy J.-F. 1991. Behaviour of
As all tests were carried out in dry silica sand with Pile Plug in Sandy Soils During and After Driving.
different soil density such that it was possible to 23rd Annual Offshore Technology Conference in
investigate the influence of soil density on soil Houston, Texas, pp. 145-154.
plugging tendency, too.
During the pile installation process different de Nicola A., Randolph M.F. 1997. The plugging
values like penetration resistance and total stress behaviour of driven and jacked piles in sand.
development inside and outside the piles were Géotechnique 47(4):841-856.
continuously recorded.
Out of all these tests, the following general Gavin K.G., Lehane B. 2003. The shaft capacity of pipe
piles in sand. Canadian Geotechnical Journal, 40:36-
findings regarding soil plug formation can be
45.
summarized.
131
Henke S. 2013. Untersuchungen zur Pfropfenbildung Schriftenreihe Geotechnik, Universität Kassel, Heft
infolge der Installation offener Profile in granularen 23.
Böden. Veröffentlichungen des Instituts für
Geotechnik und Baubetrieb der TU Hamburg-Harburg, Paikowski S.G., Whitman V.R. 1990. The effects of
Heft 29. plugging on pile performance and design. Canadian
Geotechnical Journal, 27:429-440.
Jardine R., Chow F., Overy R., Standing J. (2005). ICP
design methods for driven piles in sand and clays. Radomski H. 1968. Untersuchungen über den Einfluss
Thomas Telford, London. der Querschnittsform wellenförmiger Spundwände auf
die statischen und rammtechnischen Eigenschaften
Kishida H. 1967. The ultimate bearing capacity of pipe (German). Mitteilungen des Instituts für
piles in sand. Proceedings of the 3rd Asian regional Wasserwirtschaft, Grundbau und Wasserbau der
Conference, SMFE, 1:196199. Universität Stuttgart, Heft 10.
Lüking J. 2010. Tragverhalten von offenen Randolph M.F., Leong E.C., Houlsby G.T. 1991. One
Verdrängungspfählen unter Berücksichtigung der dimensional analysis of soil plugs in pipe piles.
Pfropfenbildung in nichtbindigen Böden (German). Géotechnique, 41(4):587-598.
132
Pile group effects on installation and capacity of jacked piles in sand tested in the
centrifuge
Dirk de Lange1, Jarno Terwindt1
Geo-engineering, Deltares, the Netherlands
1
Corresponding author: Dirk de Lange (Dirk.deLange@deltares.nl)
ABSTRACT: During installation of a foundation pile the surrounding soil state (stress level and void ratio)
in the vicinity of the pile will be affected. In which way and to what extend will depend on the installation
method, pile dimensions and soil properties. When piles are installed relatively close to each other, the
behaviour of a pile within such a group might therefore differ from a single pile. Two centrifuge tests were
performed in order to investigate the effects of pile installation on an adjacent pile. Three piles were installed
in-line, with a pile to pile spacing of four diameters. The central pile was instrumented such that the
resistance and pore water pressure at the tip could be measured. In addition, also the force acting on a friction
sleeve and the head of the pile could be measured. During installation of the piles all these parameters were
recorded. The capacity of the central pile was measured before and after the installation of the neighbouring
piles. The results of these tests are presented and discussed in this paper.
Keywords: Pile-Group; Centrifuge Testing; Pile Installation Order.
1 INTRODUCTION final load test. The results regarding pile set-up were
A good understanding of the behaviour of a pile published in De Lange et al. (2014). This paper
group and of individual piles in a pile group is discusses the effects of pile installation on an
important, depending on the stiffness of the adjacent pile similar to the publication of Le Kouby
superstructure, since many constructions are built et al. (2016). A different installation sequence was
upon it. However, existing pile group models and applied for the pile group during the two tests.
measurements are still subject of discussion (Le
Kouby et al., 2016). In general, it can be stated that 2 MATERIALS AND METHODS
the behaviour of a pile group differs from the 2.1. Test set-up
behaviour of a single pile multiplied by the number The test set-up existed of a steel cylindrical sample
of piles. This is often expressed by an efficiency container, hydraulic cylinders mounted on a stiff
coefficient and will generally depend on the pile steel frame, for pile installation and testing, and the
configuration (spacing and number), the pile model piles, see Figure 1. The frame was placed on
properties (dimensions), the installation order and top of the container.
method, and the soil conditions. Le Kouby et al.
(2016) give an overview of performed research in 2.2. Dimensions and layout
this field and present results of calibration chamber
tests on the installation order effects on the response Closed-ended piles with a diameter, D, of 16 mm
of a pile group in silica sand. They also conclude were used. The pile group existed of one
that pile spacing and installation order are important instrumented pile in the centre and two non-
parameters. The local pile responses differ from instrumented piles at a (centre-to-centre) distance of
those of the other piles in the group. Further, they 64 mm. These three piles were installed in a line, see
concluded that installing the central pile lastly will Figure 2. The inner diameter of the container was
lead to the “best” group effect and a stiffer response 0.9 m and the height of the sand sample was 0.6 m.
(both for tip resistance and shaft friction). They also The distance from a pile to the wall was 300 mm.
observed a negative group effect for the tip The (centre-to-centre) distance from the single pile
resistance and a positive group effect for the shaft to the pile group was also 300 mm. The piles were
friction. installed to a penetration level of approximately
In 2013, two tests were run in the Deltares beam 0.32 m.
centrifuge (arm length of 5.5 m). The tests were run
at 40g. During these tests, both a single pile and a 2.3. Soil material and preparation method
group of three piles were installed (in flight) in a The tests were performed in fully water saturated
single sand sample and have been load tested Baskarp B15 sand. The main characteristics are
regarding pile set-up. The centrifuge continued given in Table 1. The sand was prepared by dynamic
spinning from the start of the installation until the compaction of a fully saturated sample (Rietdijk et
133
al., 2010). Relative densities of respectively 66.3% Table 1. Baskarp B15 sand characteristics.
and 66.8% were achieved for test 1 and 2. Parameter Unit B15
Particle density, ρs kg/m3 2.65
Mean mass diameter, d50 µm 118
Coefficient of uniformity, d60/d10 - 1.4
Minimum porosity, nmin % 35.2
Maximum porosity, nmax % 47.6
2.3. Model piles and instrumentation

The forces acting on the pile head, the pile base and
the friction element could be measured separately
for the instrumented piles. Also, the pore water
pressure at the base could be measured, see
Figure 3. For all piles the displacement of the
hydraulic cylinders, which were connected to the
piles, was measured.
A rough pile surface was created to pronounce the
interaction between the pile shaft and the sand.
According to Garnier and König (1988), the piles
had totally rough interfaces, since the normalized
roughness, Rn, was equal to 0.27 (Rmax = 32 µm), see
300 mm 300 mm 300 mm Figure 4. The suitable roughness was obtained by
machining a continuous profile of notches over the
entire shaft surface, see Figure 3.
600 mm
Figure 1. Drawing of test set-up: soil model container

together with the hydraulic cylinders and reaction frame.
Figure 2. Top view of the pile layout: 16 mm diameter in a 0.9

m diameter container. Pile 1 and 2 were instrumented piles;
pile 3 and 4 were dummy piles. Figure 3. Geometry, surface and sensor locations of the
instrumented piles.
134
Table 3. Test scheme of pile in group, test 2
Step Activity
1 Installation pile 3
2 Installation pile 2
3 Waiting for 1 min
4 1st capacity test (20%D)
5 Waiting for 100 min
Figure 4. Definition of normalized roughness (Garnier and
König 1998). 6 2nd capacity test (20%D)
7 Waiting for 1000 min
2.4. Test program and procedure 8 3rd capacity test (20%D)
To study possible ageing effects, the instrumented 9 Installation of pile 4
piles were jacked into the soil over a certain distance 10 Waiting for 1 min
to mobilize the axial capacity (compression), 11 4th capacity test (20%D)
multiple times after installation, see Table 2 and 3. 12 Waiting for 100 min
During test 1, pile 2 was installed and load tested
13 5th capacity test (20%D)
several times, before pile 3 and 4 were installed.
After installation of pile 3 and 4, which were
The installation of and the tests on the piles were
installed simultaneously, pile 2 was again load
done displacement controlled. The installations of
tested several times. During test 2, pile 3 was
the piles were performed at a displacement rate of 1
installed first. After installation of pile 2, pile 2 was
mm/s. The pile load tests were performed at a
load tested several times before pile 4 was installed.
displacement rate of 0.002 mm/s up to a pile head
After installation of pile 4, pile 2 was load tested
displacement of 10% (test 1) or 20% (test 2) of the
again several times.
pile diameter, Dp. After each load test, the piles were
Some other differences between test 1 and test 2
unloaded by pulling the piles upwards.
should be noticed as well. First, for test 2, the load
test at 10 minutes after a prior load test was
cancelled. Second, due to operational constraints 3 RESULTS AND DISCUSSION
with the centrifuge during test 2, pile 2 could not be 3.1. Correction and normalization of data
tested at 1000 min after installation of pile 4. And All measured forces are corrected by subtracting the
third, the piles were not unloaded by pulling in initial values, measured just before installation. In
between installation and load testing during test 1. this way, the own weight of the piles has been
eliminated. The influence of the increasing
Table 2. Test scheme of pile in group, test 1 acceleration field with depth on the weight of the
Step Activity piles is negligible (also due to opposite buoyant
1 Installation of pile 2 forces). The total friction is defined as the difference
2 Waiting for 1 min between the forces at the pile head and the pile tip.
3 1st capacity test (10%D) Further, all presented results are normalized. The
4 Waiting for 10 min displacements are normalized by the pile diameter,
5 2nd capacity test (10%D) Dp. The measured stresses at the pile tip are
6 Waiting for 100 min normalized by the calculated vertical effective stress
level at the pile tip. The mean shear stresses at the
7 3rd capacity test (10%D)
shaft are normalized by the calculated vertical
8 Installation of pile 3 and 4
effective stress at half the height of that part of the
9 Waiting for 1 min shaft, what carries the load.
10 4th capacity test (10%D)
11 Waiting for 10 min 3.2. Pile installation
12 5th capacity test (10%D) During test 1, pile 2 was installed before pile 3 and
13 Waiting for 100 min 4, while during test 2, pile 2 was installed after pile
14 6th capacity test (10%D) 3. Figure 5 shows the normalized tip resistance
15 Waiting for 1000 min during installation of pile 2 in test 1 and 2. Initially,
16 7th capacity test (10%D) a stiffer tip response was observed for pile 2 during
Cyclic displacements of 0.1mm at 50% test 1 (“first pile”). However, a higher tip resistance
17 of capacity measured during prior step was observed at larger penetration levels during
18 8th capacity test (10%Dp) test 2 (“second pile”).
135
fatigue (Lehane, 1992). The measured resistance at
the friction sleeve was not reliable during test 2.
Therefore, this data is not presented in this paper.
3.3. Installation of adjacent piles

During the installation of adjacent piles, all gauges
of the instrumented central pile were recorded, and
the displacement of the central pile was controlled
to be zero, see Figure 7. During test 1, pile 3 and 4
were installed simultaneously, while during test 2,
only pile 4 was installed after installation of the
central pile (pile 2).
Figure 8 shows the normalized total stress at the
tip of pile 2 during installation of the adjacent pile(s)
as function of the penetration of the adjacent pile(s).
During test 1, the total pressure acting against the
tip of pile 2 started to decrease immediately when
the adjacent two piles started penetrating the sand.
However, after a penetration of about 14D the
pressure acting on the tip of pile 2 increased to a
Figure 5. Normalized tip resistance during installation of pile similar level as before the installation of the
2. adjacent piles. During test 2, a decrease was
observed after a penetration of the adjacent pile of
about 10D. The total pressures became more or less
constant after a penetration of about 14D. Figure 9
shows the normalized total pore water pressure at
the tip of pile 2 during installation of pile 4 in test 2
as function of the penetration of pile 4. Initially, the
pore water pressure increased, after 9D penetration
of pile 4, it decreased, after 14D penetration it
increased again and after 17D it became stable. It is
thought that the soil around the tip of pile 2 initially
is being pushed away from the pile (downwards) by
the installation of the adjacent pile(s), however, for
deeper penetration of the adjacent pile(s), soil
displacement around the tip of pile 2 becomes
mainly sideward, see Figure 10.
d/D (-) d/D (-)
19.885 19.888 19.890 19.892 19.895 19.110 19.113 19.115 19.117 19.120
0.0 0.0
2.5 2.5
d/D (-)
5.0 5.0
Figure 6. Normalized shaft friction, total and at friction 7.5 7.5
element, during installation of pile 2.

Normalized Displacement Installed Pile
10.0 10.0
Figure 6 shows the normalized shaft friction

during installation of pile 2 in test 1 and 2. The 12.5 12.5
displacement halfway the pile shaft of interest has

been taken for the normalized displacement in this 15.0 15.0
figure. The total friction was higher during test 2 17.5 17.5
compared to test 1. These observed tip and shaft

responses are in line with the finding of Le Kouby 20.0
A
20.0
B
et al. (2016). The normalized friction at the friction Figure 7. Vertical displacement of pile 2 during installation of
sleeve is substantially larger than the normalized the adjacent pile(s) during test 1 (left) and 2 (right). The vertical
total friction, which can be explained by friction axis is the displacement of the pile being installed.
136
increased, and after a penetration of the adjacent
pile of about 14Dp, the total friction decreased.
Figure 10. Different mechanisms during installation (adopted

from Le Kouby et al. 2016).
Figure 8. Normalized stress at the tip of pile 2 during

installation of the adjacent pile (s). The vertical axis is the
displacement of the pile being installed.
Figure 11. Normalized stress acting on (a part of) the shaft of

pile 2 during installation of the adjacent pile (s). The vertical
axis is the displacement of the pile being installed.
Positive shaft friction is acting as an upward force

and negative shaft friction as a downwards force.
An increase in shaft friction can only be explained
by upward soil movement relative to the pile, while
Figure 9. Normalized pore water pressure at the tip of pile 2 a decrease can be explained by downward soil
during installation of the adjacent pile (s). The vertical axis is movement. Therefore, it can be concluded that
the displacement of the pile being installed. between 9D and 14D penetration of the adjacent
pile(s), the soil movement along the friction sleeve
A similar trend has been observed for the friction of pile 2 (bottom 5D) was mainly downwards, while
at the friction sleeve during test, see Figure 11. the soil movement along the upper part of the pile
However, the total friction showed the opposite shaft (15D) was mainly upwards. Movements in the
trend. Initially, the total shaft friction decreased (test opposite direction can be argued for penetration
1) or was stable (test 2), after a penetration of the levels of the adjacent pile(s) between 14D and 20D.
adjacent pile(s) of about 9Dp, the total friction All observed trends were more pronounced during
test 1, since two piles were installed simultaneously.
137
3.4. Influence on single pile capacity
Before and after installation of the adjacent pile(s),
the central pile has been load tested several times
regarding pile set-up. The results are given in Figure
12 (tip resistance) and Figure 13 (shaft friction) as
function of normalized tip displacement. Before
installation of the adjacent pile(s), during test 1 and
2, the central pile has been load tested three times.
Higher tip resistance and total friction were
mobilized during the first load tests during test 2,
which could be explained by the installation
sequence. The results from the first load test during
test 1 are not comparable to the results from the
other load tests, since no unloading had been
applied before the first load test of test 1. After
installation of the adjacent pile(s), a weaker tip
response and an increase in shaft friction, during the
fourth load tests, can be clearly observed.
Figure 13. Measured parameters for the instrumented piles
during capacity testing.
5 ACKNOWLEDGEMENTS
The authors would like to acknowledge Frans Kop,
Rob Zwaan, Ferry Schenkeveld, Thijs van Dijk,
Jack van der Vegt and Ruud van der Berg for their
contribution to the preparation and performing of
the presented tests. Further, Ruud Stoevelaar, Jelke
Dijkstra, Frits van Tol and Bert Everts are
acknowledged for their guidance. The tests were
funded by CURNET and Deltares.
6 REFERENCES
Garnier, J., and König, D. 1998. Scale effects in piles and
nails loading tests in sand. In Proceedings of the
International Conference Centrifuge 98, Tokyo, Japan,
Vol. 1, pp. 205–210.
Le Kouby, A., Dupla, J.C., Canou, J., Francis., R. 2016.

The effects of installation order on the response of a
pile group in silica sand. Soils and Foundations, 56(2):
Figure 12. Measured parameters for the instrumented piles 174-188. doi: 10.1016/j.sandf.2016.02.002
during capacity testing.
De Lange D.A., Van Tol, A.F., Dijkstra, J., Bezuijen A.
4 CONCLUSIONS & Stoevelaar, R. 2010. Set-up effects of piles in sand
tested in the centrifuge. In Proceedings 8th Int. Conf.
Results from centrifuge tests on the effects of pile Phys. Model. Geo., Perth, Australia, 14 – 17 January
installation (jacking) on an adjacent (instrumented) 2014, Taylor & Francis Group, London, pp. 721-727.
pile are presented in this paper. The results indicate
directions of soil movement along the adjacent pile Lehane, B.M. 1992. Experimental investigations of pile
during pile installation. The results also indicate the behaviour using instrumented field piles. PhD thesis,
effect of the installation sequence and clearly show University of London (Imperial College).
a weaker tip response and an increase in shaft
Rietdijk J., Schenkeveld F.M., Schaminee P.E.L. and
friction of a single pile after installation of (an)
Bezuijen A. 2010. The drizzle method for sand sample
adjacent pile(s). These findings are in line with the preparation. In Proceedings 8th Int. Conf. Phys. Model.
findings of Le Kouby et al. (2016) and these are Geo., Zürich, Switzerland, 28 June – 1 July 2010,
especially important for non-stiff superstructures, Taylor & Francis group, pp. 267-272.
which are founded on pile groups.
138
Centrifuge modelling of shaft resistance of a rough rock-socketed pile

José Gregorio Gutiérrez-Ch1, Geyang Song2, Charles Heron2, Alec Marshall2, Rafael Jimenez1
ETSI Caminos, Canales y Puertos, Universidad Politécnica de Madrid, Spain
1
Nottingham Centre for Geomechanics, University of Nottingham, United Kingdom

2
Corresponding author: José Gregorio Gutiérrez-Ch (jg.gutierrez@upm.es)
ABSTRACT: Previous research and field experience have demonstrated that socket roughness is a crucial
factor that affects the shaft resistance of rock-socketed piles. This research aims to present a methodology
to evaluate the shaft resistance behaviour of a rough axially loaded pile socketed in soft rock.
Geotechnical centrifuge tests were conducted at 50-g using small scale aluminium model piles fabricated
with different roughness profiles. The model piles were socketed into an artificial soft rock, which consists
of sand, cement, bentonite and water. Fibre Bragg grating sensors were used to measure the axial
strains/loads along the piles. Results presented in this paper relate to a single tests and are used to
demonstrate the viability of the approach to investigate the shaft behaviour of rock-socketed piles
considering socket roughness.
Keywords: Rock-Socketed Piles, Shaft Resistance, Socket Roughness, Centrifuge Tests, Fibre Bragg Grating.
1 INTRODUCTION instrumentation, as well as the model set-up, are

Load tests and numerical modelling (Pells et al. described next.
1980, Horvath et al. 1983, Gutiérrez-Ch et al.
2019a, 2019b) on rock-socketed piles suggested 2.1. Pseudo rock
that socket roughness is an important aspect that The effect of socket roughness on shaft resistance
affects the shaft resistance behaviour. Seidel and of rock-socketed piles is particularly significant in
Collingwood (2001) indicated that socket piles socketed in soft rocks having an intact UCS
roughness resulting in real projects depends mainly of 𝜎𝑐 = 1 − 12 MPa (Seidel and Collingwood
on the drilling tools employed, as well as the type 2001). However, the preacticalities of preparing
and strength of the rock. For instance, O’Neill et centrifuge models on such intact rocks is complex
al. (1996) reported that sockets drilled with an (Leung and Ko 1993). For this reason, an artificial
auger tool in soft rock tend to have a sinusoidal soft rock was prepared by mixing sand, cement,
pattern, and Nam and Vipulanandan (2008) bentonite and water with a particular proportion
indicated that sockets drilled by core barrel are (see Table 1). After 44-days of normal curing, an
30% rougher than those drilled with an auger tool. average value of 𝜎𝑐 = 1.14 MPa was obtained,
However, socket roughness is not usually which was based on several cube tests.
considered in current design methods, and the
estimation of shaft resistance of rock-socketed Table 1. Pseudo rock mix proportions.
piles is commonly based on empirical correlations Mix proportions by percent mass (%)
as a function of the uniaxial compressive strength Sand Cement Bentonite Water
(UCS) of the intact weaker material (rock or pile) 52.3 12.2 6.5 29.0
(Rezazadeh and Eslani 2017).
This research aims to present a methodology 2.2. Model pile
used to study the shaft resistance response of an
2.2.1 Manufacturing and instrumentation
axially loaded rock-socketed pile considering
socket roughness using a geotechnical centrifuge To achieve an axial stiffness (𝐸𝐴; where 𝐴 is cross
test. sectional area and 𝐸 is Young’s modulus) of a
0.8 m diameter concrete solid pile (𝐸 = 30 GPa) in
2 MATERIALS AND METHODS prototype scale, a 15.875 mm diameter (𝐷) model
pile (model scale) with a 2.032 mm thickness (𝐸 =
In this study, a test was conducted at 50-g using
70 GPa) was used (see Figure 1). The nominal
the geotechnical centrifuge facility at the
University of Nottingham Centre for Geomechanic length of the pile is 80 mm at model scale
(NCG). The centrifuge has a 2 m beam radius and (equivalent to 4 m at prototype scale).
is capable of reaching accelerations up to 150-g. Also, to assess the influence of socket roughness
The pseudo-rock employed, the model pile and its on rock-socketed pile behavior, the model pile was
machined with a sinusoidal surface (see Figure 1),
139
with an asperity amplitude and a wavelength of 0.8 the wavelength shift for the 𝐹𝐵𝐺𝑖 sensor, and 𝜆𝑂𝑆𝑖
mm and 10 mm, respectively (model scale). is its wavelength offset. Though there may be
To measure the axial load along the model pile, slight changes between the results among the
two optical fibres with corresponding fibre Bragg calibration tests, the 𝐾𝐹𝐵𝐺𝑖 is considered to be
grating (FBG) sensors were used. Each optical unique for each FBG sensor (𝐹𝐵𝐺𝑖 ), so that its
fibre line has three FBG sensors (see Figure 1) final value, employed to compute the axial load at
bonded (via super glue and UV sensitive adhesive) the location of each 𝐹𝐵𝐺𝑖 sensor during the
to the internal surface of the aluminium model pile; centrifuge tests is taken as the average from the
these are labelled according to their distance from three calibration tests conducted for each FBG
a 0 mm line (𝐻) normalized by the model pile sensor.
radius (i.e., 𝐻/𝑅).
4000
𝐹 = −2.0∆𝜆𝐹𝐵𝐺 + 32.8 𝑅 = 0.
3000
Load, F (N)
2000
1000
FBG5
𝐹𝐵𝐺
Linear fitted line
0
-2000 -1500 -1000 -500 0
Wavelength shift, ∆𝜆𝐹𝐵𝐺(pm)
Figure 2. Example of calibration of FBG sensors conducted
on the model pile for sensor FBG5: load-wavelength shift
Figure 1. Aluminium model pile: schematic illustration of the relationship.
geometric and instrumentation details.
2.3. Centrifuge model preparation
2.2.2 Calibration of FBG sensors
The centrifuge model (see Figure 3) was prepared
The strain measured from FBG sensors is not by the following steps. A piece of soft polystyrene
necessarily equal to the strain of the aluminium (length equal to the model pile diameter) was glued
pile (physical strain from the model pile is to the bottom of the model pile to eliminate base
translated to the FBG via the super glue and UV resistance. The pre-mixed pseudo-rock mixture
sensitive adhesive), hence the FBG sensors need to was placed into a steel cylinder container, which is
be calibrated. The calibration procedures described 20 cm in diameter and height. Then, the model pile
by Song (2019) were adopted, where three uniaxial was pushed into the mixture (just after pouring)
compression loading tests were carried out for each and set to the designed position using a temporary
pile, so that the relationship between axial load support frame mounted to the top of the container.
along the pile (𝐹𝑖 ) and shift in FBG wavelength Finally, the container was stored and cured under
(∆𝜆𝐹𝐵𝐺𝑖 ) can be obtained. high humidity condition for 44 days.
Figure 2 shows one of the calibration results.
Each load increment was held for 15 s; the 2.4. Centrifuge test
maximum axial load applied to the model piles was
The centrifuge test was performed after 44-days of
4000 N, which is above the maximum axial load
curing. The pseudo-rock container was placed on
applied during the centrifuge tests (approximatly
the centrifuge platform and steel plates (30 mm
3600 N). The wavelength shift, as a function of the
total thickness) were placed on the top of the
applied axial load, is shown in Figure 2; a linear
pseudo-rock to induce (at 50-g) an external
relationship was obtained, so that the load-
surcharge of 120 kPa to represent a 6 m soil
wavelength shift relationship can be described as:
stratum (with an average unit weight of 20 kN/m3),
see Figure 4. Next, the loading system –i.e., the
𝐹𝑖 = 𝐾𝐹𝐵𝐺𝑖 ∆𝜆𝐹𝐵𝐺𝑖 + 𝜆𝑂𝑆𝑖 (1) loading frame, two actuators, a load cell, and a
connector– was installed and connected to the
where 𝑖 is the FBG sensor number, 𝐹𝑖 is the axial model pile. A single linear variable differential
load recorded by the 𝐹𝐵𝐺𝑖 sensor, 𝐾𝐹𝐵𝐺𝑖 is the transformer (LVDT) was positioned on an
slope of the linear fitted line of Figure 2, ∆𝜆𝐹𝐵𝐺𝑖 is aluminium plate located above the pile cap (see
140
Figure 4) to measured the model pile settlement. 40% gradual reduction of the pile load capacity at
The axial force along the pile was obtained using 𝛿 ⁄𝐷 = 25% is noted.
the wavelength shift recorded at each FBG sensor.
The acceleration of the centrifuge was gradually
increased to 50-g and then the model pile was
loaded axially with a displacement controlled
procedure by the loading system (actuator rate of
0.1 mm/s).
Figure 5. Load-settlement curve from the centrifuge test of a

rough rock-socketed pile with a polystyrene base.
3.2. Shaft resistance-settlement response

Figure 6 shows the evolution of the mobilised
Figure 3. Drawing of the centrifuge model. average shaft resistance (𝑓𝑎𝑣𝑒 ) with settlement at
different portions of the pile (referred to by the
normalized depth 𝐻/𝑅). The 𝑓𝑎𝑣𝑒 was computed
as:
𝐹𝑖 −𝐹𝑖+1
𝑓𝑎𝑣𝑒 = 𝜋𝐷∆𝐿
(2)
where 𝐹𝑖 and 𝐹𝑖+1 are the mobilised axial loads at

two consecutive locations (e.g. between the FBG
sensors located at 𝐻/𝑅 = 2.5 and 𝐻/𝑅 = 5.0, see
Figures 1 and 6 for reference), ∆𝐿 is the nominal
length between the sensor locations, and 𝐷 is the
pile diameter.
Results show that the shaft resistance is first fully
mobilised in the upper portion of the pile for a 𝛿/𝐷
of 3%; as the applied load increases further, the
Figure 4. Centrifuge test set-up. 𝑓𝑎𝑣𝑒 decreases in the upper portion of the pile, and
the full mobilization of 𝑓𝑎𝑣𝑒 in the lower portions
3 RESULTS AND DISCUSSION of the pile occurs for a 𝛿/𝐷 = 𝛿𝑝𝑒𝑎𝑘 /𝐷 of 6.6%.
3.1. Load-settlement response After such a threshold, a reduction of 𝑓𝑎𝑣𝑒 is noted.
The load-settlement curve from the centrifuge test In addition, for lower pile head settlements (𝛿/𝐷 ≤
conducted on a rough rock-socketed pile with the 3%) much of the loads are concentrated around the
polystyrene base (i.e., considering shaft resistance front of the asperities located in the upper part of
only) is shown in Figure 5. The test was conducted the pile; then, with increasing pile head settlement
until a limit pile head settlement (𝛿) of 25% of the (𝛿/𝐷 > 3%), a damage at the rock-pile interface
pile diameter (𝛿 ⁄𝐷 = 25%) was reached. The might occur, and the maximum values of 𝑓𝑎𝑣𝑒 shift
results are presented at prototype scale. downwards (towards the pile base), where, with
As can be seen in Figure 5, the pile load capacity further loading, a similar behaviour is observed.
increased rapidly and almost linearly until the peak This experimental result is qualitatively congruous
head load 𝐹𝑝𝑒𝑎𝑘 = 8.85 MN (and its with the results of DEM3D piles tests conducted on
corresponding pile head settlement, 𝛿𝑝𝑒𝑎𝑘 /𝐷 = rough piles socketed in sanstone (𝜎𝑐 = 11.5 MPa)
6.6%) was reached. After this load threshold, a by Gutiérrez-Ch et al. 2019b, in which the
degradation of asperities and significant bond
141
breakages becomes more relevant in the front of boundary conditions. Engineering Geology, 240: 123–
the asperities for 𝛿/𝐷 > 3%. 139. https://doi.org/10.1016/j.enggeo.2018.04.017.
Mobilised Average Shaft Resistance,𝑓𝑎𝑣𝑒 Gutiérrez-Ch, J.G., Melentijevic, S., Senent, S. and
Jimenez, R. 2019a. DEM models to predict side shear
(MPa)
-0.3 0.3 0.9 1.5 2.1 resistance of rock-socketed piles considering socket
-5 roughness. In Proceedings of the 53rd U.S. Rock
Fully mobilised 𝑓𝑎𝑣𝑒 Fully mobilised 𝑓𝑎𝑣𝑒 Mechanics/Geomechanics Symposium, New York
in the lower region in the upper region
0 City, New York, 23–26 June. ARMA 19–1757.
5 Gutiérrez-Ch, J.G., Melentijevic, S., Senent, S. and

Jimenez, R. 2019b. DEM models to predict side shear
𝛿⁄𝐷 (%)
10 resistance of rock-socketed piles considering socket

roughness. In Proceedings of the 1st Mediterranean
15 Young Geotechnical Engineers Conference, Mugla,
Turkey, 23–24 September, pp. 47–54.
𝐻 ⁄𝑅 = 0
𝐻 𝐻 ⁄ 𝑅 = 2.5
20
𝐻 ⁄ 𝑅 = 5.0
Horvath, R.G., Kenney, T.C. and Kozicki, P. 1983.
25 𝐻 ⁄ 𝑅 = 8.8
Methods for improving the performance of drilled
0.0 − 2.5
2.5 − 5.0
5.0 − 8.8
8.8 − 10.1
𝐻⁄𝑅 𝐻 ⁄ 𝑅 = 10.1 piers in weak rock. Canadian Geotechnical Journal,
30 2𝑅
20(4): 758–772. https://doi.org/10.1139/t83-081.
Figure 6. Mobilised average shaft resistance -settlement
curves at different depths from the centrifuge test of a rough Leung, C.F. and Ko, H-Y. 2001. Centrifuge model study
rock-socketed pile with a polystyrene base. of piles socketed in soft rock. Japanese Society of Soil
Mechanics and Foundation Engineering, 33(33): 88–
4 CONCLUSION 91. https://doi.org/10.3208/sandf1972.33.3_80.
This paper presented a centrifuge test methodology
Nam, M.S. and Vipulanandan, C. 2008. Roughness and
to study the shaft resistance behaviour of an axially
unit side resistances of drilled shafts socketed in clay
loaded rock-socketed pile considering socket shale and limestone. Journal of Geotechnical and
roughness. To do that, an artificial soft rock made Geoenvironmental Engineering, 134(9): 1272–1279.
of a mixture of sand, cement, bentonite and water https://doi.org/10.1061/(ASCE)1090-
having 𝜎𝑐 = 1.14 MPa was employed. The model 0241(2008)134:9(1272).
pile was machined from an aluminium tube with a
sinusoidal surface, and instrumented with fibre O’Neill, M.W., Townsend, F.C., Hassan, K.M., Buller,
Bragg grating (FBG) sensors to record the axial A. and Chan, P. S. 1996. Load transfer for drilled
strains/loads along the pile. Results show that the shafts in intermediate geomaterials. U.S. Department
methodology developed in this work can be of Transportation, FHWA-RD-95–172, Final Report.
employed to analyse the load- and shaft resistance- Pells, P.J.N., Rowe, R.K. and Turner, R.M. 1980. An
settlement response of a rock-socketed pile experimental investigation into shear for socketed
considering socket roughness, also providing piles in sandstone. In Proceedings of the International
interesting insight into the load transfer Conference on Structural Foundations on Rock,
mechanisms. Sydney, Australia, 7-9 May 1980, pp. 291–302.
5 ACKNOWLEDGEMENTS Rezazadeh, S. and Eslami, A. 2017. Empirical methods

for determining shaft bearing capacity of semi-deep
This research was funded, in part, by the Spanish foundations socketed in rocks. Journal of Rock
Ministry of Economy, Industry and Mechanics and Geotechnical Engineering, 9(6):
Competitiveness under Project BIA 2015-69152-R. 1140–1151.
The first author was the recipient during 2019 of https://doi.org/10.1016/j.jrmge.2017.06.003.
one Fellowship for PhD research provided by the
José Entrecanales Ibarra Foundation. The support Song, G. 2019. The use of protective structures to
of both institutions is gratefully acknowledged. reduce tunnelling induced damage to buildings. Ph.D.
thesis, Faculty of Engineering, University of
Nottingham, UK.
6 REFERENCES
Gutiérrez-Ch, J.G., Senent, S., Melentijevic, S., and Williams, A.F. 1980. The design and performance of
Jimenez, R. 2018. Distinct element method piles socketed into weak rock. Ph.D. thesis, Monash
simulations of rock-concrete interfaces under different University, Melbourne, Australia
142
Physical modelling of single-helix piles in sand using calibration chamber tests

Diego Moreira da Silva1, Cristina de Hollanda Cavalcanti Tsuha1
Department of Geotechnical Engineering, University of São Paulo, Brazil
1
Corresponding author: Cristina de Hollanda Cavalcanti Tsuha (chctsuha@sc.usp.br)
ABSTRACT: Large calibration chamber tests have been used in different studies to investigate pile
performance under soil controlled conditions. For full scale field tests, pile and soil measurements are
challenging and expensive; in addition, the field soil variability and the lack of information on soil
characteristics are obstacles to make comparisons and evaluate geometrical effects on the pile response.
Therefore, calibration chamber tests on model piles are particularly attractive to explore completely the
some aspects of pile-soil interaction. The current paper presents a large calibration chamber constructed
recently at the University of São Paulo (ESSC/USP) in Brazil to investigate the behaviour of helical piles in
sand. The main aspects that will be studied using this equipment are: installation forces; effect of installation
on the pile behaviour; pile uplift and compressive capacity; cyclic performance; effect of wing ratio and
helix pitch on the pile performance and installation forces; stresses in the soil during pile installation,
monotonic and cyclic loading; and effects of rotation speed and advance rate during installation. For this
purpose, instrumented model piles of different configuration were fabricated. Some preliminary results of
pile installation and loading tests are presented in this paper.
Keywords: Calibration Chamber; Physical Modelling; Helical Piles; Sand.
1 INTRODUCTION Due to some limitations of 1g-model tests (it is not

Helical piles are widely used to resist tensile and adequate to study problems dependent on the
compressive forces in transmission line towers, confining stresses), this type of pile has also been
wind towers, solar panels, and other types of investigated by different authors using centrifuge
structures. More recently, they have been suggested model tests (Levesque 2002, Tsuha 2007, Bian et al.
for offshore applications (Byrne and Houlsby 2015, 2008, Wang et al. 2010, Schiavon et al. 2017, Al-
Spagnoli and Gavin 2015, Al-Baghdadi et al. 2015, Baghdadi 2018).
Cerfontaine et al. 2019, etc.) due to the important
uplift capacity provided by the anchor effect of the
helix between other advantages. This type of pile is
composed of one or more helical plate welded to a
steel shaft. During installation, they are "screwed"
into the ground by the application of torque on the
shaft with a hydraulic motor (Figure 1), with a
penetration rate of one pitch per revolution. The
installation torque along the depth is registered
during pile installation as a tool to control the pile
capacity.
The current study aims to investigate the
behaviour of helical piles in sand using calibration
chamber modelling. In field tests, the soil conditions
are normally not controlled, thus preventing the
evaluation of the effects of some parameters on the
problem investigated. Additionally, the soil
instrumentation is not feasible in field studies, and Figure 1. Helical pile installation at the EESC-USP Campus.
only physical models in laboratory are able to
explore completely some aspects of pile-soil Calibration chamber tests is another type of
response. experimental technique used for tests on reduced
The performance of helical foundations have been scale models, in which the field stress condition can
studied using 1-g models tests by different authors be reproduced in laboratory. Foray (1991)
(Clemence & Pepe 1984; Mooney et al. 1985; commented that this physical modelling technique
Narasimha Rao et al. 1989; Ghaly et al. 1991; etc.). can be used to study geotechnical axisymmetric
143
problems involving a high stress level, such as long research, nine different instrumented model piles
offshore piles, anchors, nailing etc. were fabricated to be installed and tested in dry
For calibration chamber tests, homogeneous soil dense sand prepared by air pluviation technique. For
samples can be prepared under different controlled the tests, soil stress sensors are placed in the sand
boundary conditions. This physical modelling mass to measure the variation of the stresses in the
technique allows the use of sand samples under soil during pile installation and under loading. Some
known and controlled stresses. A pressurized soil preliminary results of this study are presented in this
sample in a calibration chamber reproduces what paper.
occurs in a “portion” of soil at a given depth, (can
simulate field overburden pressure). Therefore, 2 CALIBRATION CHAMBER TESTS
tensile and compressive loading tests on a single- The EESC/USP calibration chamber is composed
helix pile model in a calibration chamber can of: i) a steel tank with a top cover; ii) a rubber
reproduce the helix bearing resistance at a particular pressurised membrane; iii) air pluviation system; iv)
depth in the field and the shaft resistance just above installation and reaction structures.
the helix, as illustrated in Figure 2.
2.1. EESC/USP calibration chamber
2.1.1 Steel tank
The dimensions of the calibration chamber for the
study on helical foundations were calculated based
on the results of numerical simulations (Silva 2020)
and on the findings on boundary effects in
calibration chamber tests available in the literature.
The calibration chamber constructed consists of a
steel tank with 25.4mm thickness, 1.50 m high and
1.20 m diameter. The bottom of the tank was
Figure 2. Simulation of the bottom section of a helical pile
installed 2 meters below the ground surface at the
(Silva, 2020)
Experimental Foundation Site of the University of
There are few studies on helical piles using Sao Paulo at Sao Carlos, Brazil. Figure 3 shows the
calibration chamber tests. The first case is the work tank and the top cover of the chamber.
of Nagata & Hirata (2005). These authors
performed installation and uplift tests on a helical
pile model in a tank of dry sand under controlled
confining pressure. These authors evaluated the
shape of the failure zone and the magnitude of
confining stresses action on this zone. Malik et al.
(2017, 2019) investigated the end bearing capacity
of a single-helix pile and the helix bending
deflection in sand under controlled confining
pressure. However, in these studies the pile was not
installed by rotational driving reproducing the field
installation procedure. Nagai et al. (2018)
performed installation and uplift loading tests on a Figure 3. Steel tank and top cover.
helical pile model in a calibration chamber. They
tested different pile installation methods, and 2.1.2 Pressurised membrane
evaluated the helix bearing capacity under tensile For the calibration chamber tests, a pressurised
loading considering the change in earth pressure at rubber membrane (Figure 4) with a central hole of
the pile tip. 50 mm diameter (to allow the penetration of the
The large calibration chamber of the ESSC/USP, model pile in the pressurised soil inside the
presented in this paper, was constructed to chamber) is installed between the sand sample
investigate the effect of pile geometry on the surface and the top cover to apply a controlled
installation forces, and the performance of helical pressure on the soil sample in the tank. A latex
piles under monotonic and cyclic tensile and rubber sheet was also placed along the rigid walls of
compressive loading. The aim of the chamber is to the chamber lubricated with grease.
simulate the field stresses condition of the The upper membrane is filled with compressed air
surrounding soil of a helical pile in sand. For this to simulate different vertical stress conditions on the
144
soil sample. For example, an applied pressure of 100 2.2. Experimental programme
kPa can reproduce the helix bearing resistance of a 2.2.1 Soil sample preparation
helical pile installed at a depth of approximately 6m
in dry dense sand. The dry sand specimens for this study are prepared
in the chamber tank by air pluviation technique as
2.1.3 Pile installation system illustrated in Figure 6. The characteristics of the
silica sand of Analândia used in this study are
The installation system are connected to the tank presented in Table 1. The sample density shown in
cover during installation. After the soil sample this table is the average value obtained from ten
preparation, the model pile is installed using a calibrated boxes placed in the tank at different
particular device to ensure a pile penetration rate positions during the preparation of the soil.
into the soil sample of one pitch per revolution. As
shown in Figure 5, a threaded bar (and nut) with the Table 1. Sand properties
same pitch of the pile helix is connected to the pile Specific gravity of the sand Gs 2.67
shaft to control the penetration rate during Maximum dry density (kg/m3) ρd(max) 1794
installation. Minimum dry density (kg/m3) ρd(min) 1463
Sand sample average dry density ρd 1784
Average grain-size (mm) d50 0.18
Coefficient of uniformity Cu 1.73
Figure 4. Pressurised rubber membrane.
Figure 6. Sand pluviation system.
After the chamber tank is filled, the bottom part

of the model pile is placed on the soil sample before
the installation of the pressurised membrane. The
diameter of central hole of the membrane is smaller
than the helix diameter of the model piles; therefore,
the pile tip (helix) is placed under the membrane
before pile installation. After that, the chamber is
carefully closed.
2.2.2 Soil stress sensors

Pile-soil interactions are governed by the stress field
induced by the pile installation process. Therefore,
for a better understanding of the installation forces
and failure mechanisms of helical piles, six soil
Figure 5. Model pile installation. stress sensors are placed (vertically or horizontally)
145
in the soil samples at different depths. These sensors 2.2.4 Helical pile installation
are also installed to check the initial vertical stresses For the model pile installation, a reaction frame
in the soil when the vertical pressure is applied by system (Figure 10) is connected to the top cover of
the rubber membrane. The stress sensors used for the calibration chamber. An electric motor coupled
this study were fabricated by the Geoinstrumentos, laterally to a reducer apply the torque to the top of
with diameter of 20 mm and 7 mm thickness (Figure the helical pile as shown in Figure 5. The motor
7). rotates the thread bar, which is coupled to the top of
the pile with a "connection glove". When the helix
reaches 1.0 m depth below the soil surface, the
installation is stopped. After that, the system
illustrated in Figure 10 is removed, and another
reaction frame system is connected to the top cover
for the pile loading tests.
Figure 7. Soil stress sensors.
2.2.3 Helical pile models

For this research, nine different instrumented model
piles (Table 2) with 1.5 m length were fabricated to
be installed and tested in dry dense sand. The steel
pipe (shaft) diameter of all model piles is 42.2 mm
with a 9 mm-thick wall, and the thickness of the
helices is 6.4 mm (¼ ”).
Figure 8 illustrates three model piles with helix
pitch of 40mm (P4, P5, and P6). The preliminary
test presented in this paper was carried out on the
pile model P4 (helix diameter of 84.4 mm and pitch
of 40 mm).
Table 2. Dimensions of the model piles (Silva, 2020).

Helix diameter Shaft D/d Pitch
Type
D (mm) diameter (d) ratio (mm)
P1 84.4 42.2 2 30
P2 126.6 42.2 3 30 Figure 8. Instrumented helical pile models with helix pitch of
40 mm (Silva, 2020)
P3 168.8 42.2 4 30
P4 84.4 42.2 2 40
P5 126.6 42.2 3 40
P6 168.8 42.2 4 40
P7 84.4 42.2 2 50
P8 126.6 42.2 3 50
P9 168.8 42.2 4 50
Strain gages were installed at the pile surface of all

model piles to measure the axial forces and
installation torque acting on the shaft at 20 mm
above the helix (Figure 9). For the mechanical
protection of the strain gages during installation, a
cold solder paste (Plasteel) was used (Figure 9b).
Figure 9. (a) Strain gages and (b) Mechanical protection of
the strain gages.
146
2.2.5 Pile loading tests pressure of 100 kPa was used. Figure 12 shows the
For the model tests, the load is applied by a results of measured torque obtained by the
hydraulic jack attached to a test frame (for tensile instrumentation installed just above the pile helix.
and compressive loading) as illustrated in Figure 11. As the bottom part of the pile is placed in the soil
During the loading tests, the load at the pile head sample before the pile installation, the values of
was recorded using a digital load cell, and the head torque were registered after a tip depth penetration
displacement were measured using two supported of approximately 300 mm.
dial gauges (Figure 11).
Figure 12. Installation torque measured just above the helix

during the model pile installation.
Figure 12 illustrates a peak value obtained in the

beginning of installation that decreases along the
depth. During the model pile installation, the soil
sample becomes disturbed, and the torque value is
reduced during the penetration.
For this first test, a soil stress sensor registered the
variation of the vertical stress in the soil at the depth
of 1.25 m (the sensor was placed at 250 mm below
the final embedment depth of the pile, and 300 mm
distant from the pile center). The sensor results are
presented in Figure 13.
Figure 10. Reaction frame for pile installation.
Figure 13. Variation of vertical stress measured by the sensor

installed at the soil depth of 1.25m during pile installation.
Figure 11. Tensile and compressive loading tests.
The results shown in Figure 13 illustrates that
3 RESULTS AND DISCUSSION before pile installation the vertical pressure at the
Some preliminary results of the first test performed 1.25 m was approximately 94 kPa. When the pile tip
using the EESC/USP calibration chamber are reached a depth of 0.75 m, the installation was
presented in this section. For the first test, an applied stopped for few minutes. After the pause, the
147
installation starts again, and for a tip penetration 6 REFERENCES
between 75 cm and 100 cm (final depth) the vertical Al-Baghdadi, T.A., Brown, M.J., Knappett, J.A.,
pressure at 1.25 m was reduced from around 95 kPa Ishikura, R. 2015. Modelling of laterally loaded screw
to 83 kPa. This observation indicates that the piles with large helical plates in sand. In Meyer, V.
vertical stresses on the soil below the helix (Ed.), Frontiers in Offshore Geotechnics III, Taylor &
decreases when the helix approaches. These results Francis Group, London, pp. 503-508.
obtained by the soil stress sensor can explain the
reduction of installation torque with the depth Al-Baghdadi, T. 2018. Screw piles as offshore
observed in Figure 12. However, more tests are foundations: Numerical and physical modelling
(Doctoral dissertation, University of Dundee).
necessary to confirm this finding.
After installation, a tensile pile load test was Byrne, B. W., Houlsby, G.T. 2015. Helical piles: An
carried out on the model pile. Figures 14 presents innovative foundation design option for offshore wind
the vertical load x displacement curve of the uplift turbines. Phil. Trans. R. Soc. A 373(2035).
test. From the results of the instrumentation of the doi:10.1098/rsta.2014.0080.
model pile just above the helix it could be separated
the fraction of uplift capacity related to the helix Cerfontaine, B., Knappett, J. A., Brown, M. J., Davidson,
bearing from the shaft resistance. Although the helix C., Sharif, Y. 2019. Screw pile design optimisation
diameter of the pile tested has only 84.4 mm (two under tension in sand. Proceedings of the XVII
times the shaft diameter) the helix bearing is 88% of ECSMGE-2019 - Geotechnical Engineering
foundation of the future, doi: 10.32075/17ECSMGE-
the total uplift capacity (Figure 14). This finding
2019-069
illustrates the advantage of the anchor effect of
helical piles. Clemence, S. P. & Pepe, F. D., Jr. 1984. Measurement of
Lat-eral Stress Around Multihelix Anchors in Sand,
Geotech-nical Testing Journal 7(3): 145-152.
Foray, P., 1991 “Scale and Boundary Effects on

Calibration Chamber Pile Test”, in Calibration
Chamber Testing, pp. 147-160.
Ghaly, A., Hanna, A. & Hanna, M. 1991. Uplift behavior

of screw anchors in sand. I: dry sand. Journal of
Geotechnical Engineering 117(5): 773-793.
Levesque, C. L. 2002. Centrifuge modelling of helical

anchors in sand. Thesis M.Sc.E., University of New
Brunswick, Canada.
Figure 14. Uplift loading x displacement curve. Malik, A. A., Kuwano, J., Tachibana, S., Maejima, T.
2017. End bearing capacity comparison of screw pile
4 CONCLUSION with straight pipe pile under similar ground
conditions. Acta Geotechnica, 12(2), 415-428.
The current paper describes the EESC/USP
calibration chamber, constructed recently at the Malik, A. A., Kuwano, J., Tachibana, S., & Maejima, T.
University of São Paulo to investigate the 2019. Effect of helix bending deflection on load
performance of helical piles, and presents some settlement behaviour of screw pile. Acta
preliminary results. The outcomes of this study Geotechnica, 14(5), 1527-1543.
should clarify some fundamental aspects for the
Mooney, J.S., Adamczak, S., Clemence, S.P. 1985. Uplift
understanding of the helical pile behaviour. capacity of helical anchors in clay and silt. In Uplift
behavior of anchor foundations in soil: 48-72. ASCE.
5 ACKNOWLEDGEMENTS
The authors would like to thank Vértice Engenharia, Nagai, H., Tsuchiya, T., & Shimada, M. (2018).
Influence of installation method on performance of
and the National Council for Scientific and
screwed pile and evaluation of pulling resistance.
Technological Development - CNPQ (Process Soils and Foundations, 58(2), 355-369.
425770/2016-8 and 310881/2018-8), and for
granting the doctoral scholarship of the first author. Nagata, M. & Hirata, H. 2005. Study on uplift resistance
of screwed pile. Nippon Steel Technical Report, (92),
73-78.
148
Narasimha Rao, S.N., Prasad, Y.V., Shetty, M.D. & Schiavon J.A., Tsuha, C.H.C., Thorel, L. 2017. Cyclic
Joshi, V.V. 1989. Uplift capacity of screw pile and post-cyclic monotonic response of a single-helix
anchors. Geotechnical Engineering 20(2): 139-159. anchor in sand. Géotechnique Letters 7(1): 11-17.
Silva, D. M. 2020. Behaviour of helical anchors Tsuha, C.H.C. 2007. Theoretical model to control on site
subjected to cyclic loadings. Doctoral Thesis (In the uplift capacity of helical screw piles embedded in
preparation), São Carlos school of Engineering, sandy soil. Ph.D. thesis, Department of Geotechnics,
University of São Paulo, São Carlos. São Carlos School of Engineering, University of São
Paulo, São Carlos, Brazil. [In Portuguese.]
Spagnoli, G., Gavin, K. 2015. Helical Piles as a Novel
Foundation System for Offshore Piled Facilities. In Wang, D., Merifield, R.S., Gaudin, C., Hu, Y. 2010.
Abu Dhabi International Petroleum Exhibition and Centri-fuge model tests of helical anchors in clay. In
Conference, 9-12 November, Abu Dhabi, UAE, Proceedings of the 7th International Conference on
https://doi.org/10.2118/177604-MS Physical Modelling in Geotechnics, Zurich,
Switzerland, pp. 1069–1074.
149
Study on the installation effect of helical piles in very dense sand

José Antonio Schiavon1, Cristina de Hollanda Cavalcanti Tsuha2, Luc Thorel3
1
Civil Engineering Division, Aeronautics Institute of Technology, Brazil
2
Department of Geotechnical Engineering, University of São Paulo, Brazil
3
GMG Laboratory, GERS Department, Gustave Eiffel University, France
Corresponding author: José Antonio Schiavon (schiavon@ita.br)
ABSTRACT: The increasing interest on the use of helical piles to support offshore facilities has motivated
the development of researches on this type of foundation. Literature presents some studies on helical piles
in dense sand that disregard the effects that pile installation could cause on soil as a “wished-in-place”
condition is considered. Normally this condition is adopted because of the difficulty in reproducing the pile
installation according to the in-situ procedure. However, some previous studies have shown that the failure
mechanism in uplift is controlled by the disturbed-non disturbed condition of the soil around the pile, mainly
for dense sand condition. On the other hand, the literature also brings some studies arguing that the pile
installation little influences both failure mechanism and uplift performance. Therefore, the current study
uses results of centrifuge modelling to compare the monotonic uplift performance of single-helix piles in
very dense sand installed via two different procedures: i) by placing the model pile during sand bed
pluviation, which intends to avoid any soil disturbance (“wished-in-place” condition) and; ii) by screwing
the model pile in flight with torque and downward crowd force applied simultaneously. The results show
that the model pile installed in flight showed uplift capacity at least 60% greater than the “wished-in-place”
model pile, also accompanied by a greater axial stiffness. The unexpected poorer performance of the
“wished-in-place” model piles suggests that the model may have acted as an obstacle for the sand grains
during pluviation, which resulted in density variation of the sand around the model and, consequently,
provided lower uplift capacity compared to the in-flight installation case.
Keywords: Helical Piles; Sand; Soil Disturbance; Installation Procedure, Centrifuge Modelling.
1 INTRODUCTION capacity is less important in dense sand than in loose

Because of the increasing interest on the use of sand. This finding suggests that, in addition to the
helical piles in both onshore and offshore structures, interaction between helices (Hao et al. 2019), the
many questions and issues have been raised. In this difference in compactness is considerable if the
scenario, centrifuge modelling has been an dense sand is traversed by a helix one, two or three
important tool for the study of helical piles since it times. However, for loose sand, the penetration of
allows for parametric studies reproducing the same the lower helix shall loosen totally the sand inside
field of forces and mass in a full-scale structure with the cylinder circumscribed the helix. As a result, the
closely controlled test conditions (Garnier, 2001). sand above all the helices present equivalent sand
However, reliable simulations of the pile behaviour disturbance and, consequently, similar
depend on replicating the in-situ soil conditions compactness. The above observations highlight the
(e.g. deposition process, stress history) and the pile importance of reproducing the prototype installation
installation procedure, in addition to avoiding scale to replicate the disturbance caused on dense sands.
effects. Gavin et al. (2014) reported full-scale load tests
Helical piles are installed by screwing it into the on single-helix piles in dense sand and analysed the
ground usually using a hydraulic motor attached to results using Finite Element (FE) Modelling. Under
a backhoe loader or a rotary rig equipment. compressive loading, the numerical simulation
Consequently, the soil penetrated by the pile showed very good agreement with the experimental
experiences displacement and shearing, which can result in terms of load-displacement response.
cause changes in the soil around the pile. Based on Under uplift loading, in contrast, the numerical
a simulation via Discrete Element Modeling, Sharif simulation overpredicted both axial stiffness and
et al. (2019) noted that pile installation modifies the uplift resistance. According to the cited authors, the
sand density close to the pile, with the relative parameters of sand used in the FE model derived
density (Dr) changing from 80% to 55%. from triaxial compression tests and, therefore,
In centrifuge tests with single and multi-helix provided a very good prediction in compressive
model piles in sand, Tsuha et al. (2012) observed loading. In this case, the soil below the helix, which
that the contribution of the upper helices to the uplift controls the pile response, is relatively unaffected
151
HP1
HP2
HP3
Figure 1. Helical pile models
by the installation. For the uplift case, however, the helical piles used for onshore structures.
sand above the helix experience disturbance during The helical pile models consist of a single helical
installation. Consequently, the parameters from the plate welded to the tip of a round rod, both steel
triaxial tests with undisturbed samples are not fabricated (Fig. 1). Table 1 presents the dimensions
suitable to be assigned for the disturbed sand above of the three model piles. Both pitch (p) and
the helix. thickness (th) of the helical plates are not scaled
The hypothesis by Gavin et al. (2014) was according to the dimensions of the same prototype.
investigated in the study of Pérez et al. (2017) that The model HP1 was used in previous research
showed that assigning modified values for the conducted in the same laboratory (Tsuha, 2007),
parameters of disturbed soil can provide good and the other two models were fabricated later for
agreement with experimental results. In contrast, the research of which the present study is part
when the installation effect is neglected and (Schiavon, 2016).
undisturbed sand parameters are considered, the
uplift capacity is overpredicted. 2.2. Sand bed
The testing campaign reported here was designed The sand used in the experiments is the HN38
to evaluate the installation effect on the sand fraction of Hostun sand, which is a very fine and
condition and, consequently, on the uplift load- uniform silica sand extracted from deposits located
displacement response of the single-helix model in the commune of Hostun, France. Table 2 presents
piles. To isolate the installation effect, the model some characteristics of the Hostun sand. The model
piles were installed via two different procedures and piles were tested in dry sand beds reconstituted by
then tested in uplift loading. At first, the model piles means of the raining deposition technique, in which
under “wished-in-place” condition (no installation dry sand is placed in an automatic hopper and pours
effect) were expected to exhibit larger ultimate load through a slot to produce a uniform sand rain over a
compared to the model piles installed simulating the container.
in-situ procedure. However, the results have shown
the opposite. Therefore, the current paper aims to Table 1. Dimensions of helical pile models.
discuss the experimental procedure and results to D d p th
Model pile
contribute to future experiments. (mm) (mm) (mm) (mm)
HP1 20.0 6.0 1.6 0.5
2 MATERIALS AND METHODS HP2 26.6 8.0 1.4 0.8
The centrifuge testing conducted for the current HP3 33.0 10.0 1.5 1.8
study was undertaken using the IFSTTAR geo-
centrifuge. Different g-levels for three single-helix Table 2. Characteristics of HN38 Hostun sand.
pile models of different sizes were used to simulate Parameter Unit Value
the same prototype. In addition to the comparison of Specific gravity of particles, Gs - 2.64
two different installation procedures, the use of Maximum dry density, ρd(max) kg/m³ 1554
different model sizes aimed to ensure the non- Minimum dry density, ρd(min) kg/m³ 1186
occurrence of scale effects. Average grain size, d50 mm 0.12
Coefficient of uniformity, CU - 1.97
2.1. Model piles deg. 47
Angle of friction*, φ
The current tests simulated a single-helix pile *
from triaxial tests with Dr between 90% and 95% and
prototype with a helix diameter D = 330 mm, a shaft confining stresses of 50 kPa, 100 kPa and 200 kPa.
diameter d = 100 mm, and installed with a helix
depth H = 8D. Helix and shaft diameters were The hopper moves into roundtrips while the sand
established to provide a helix-to-shaft diameter ratio deposes on the container bottom. Hooper horizontal
(D/d) equals to 3.3, which is a common value for speed and slot width are the parameters set to
152
control the drop height and the flow rate of sand. container was expected to be subjected. The vertical
The pluviation process is known for providing displacement rate for pull-out tests was 0.3 mm/s at
homogenous specimens and good repeatability model scale.
(Garnier, 2001; Küs, 1992; Miura and Toki, 1982).
The final dimensions of the sand bed are 1200 mm 3 RESULTS AND DISCUSSION
in length, 800 mm in width and 360 mm in depth. Figures 3 to 5 present the load-displacement
To assess the achieved density, at least two density response, at prototype scale, for the monotonic
pots were placed on the container bottom. The uplift tests with models under condition “wished-in-
average relative density (Dr) of sand beds resulting place” (WIP) and screwed into the sand bed in flight
from sand raining ranged from 95% to 99%. (SCR).
Since the use of WIP condition intended to avoid
2.3. Installation and testing of model piles any soil disturbance caused by the model pile
Two different procedures were followed to install installation, both axial stiffness and ultimate load
the model piles at an embedment ratio H/D = 8 were expected to be larger than those obtained with
(where H is the helix embedment depth), which is the SCR condition. However, regardless the
expected to provide an intermediate behaviour anomalous behaviour observed for the 8FH SCR
between shallow and deep anchor (Pérez et al. test with uplift displacement between 5 and 20 mm
2017). The first procedure consisted of pausing the (Fig. 4), the model piles exhibited lower stiffness for
pluviation, positioning the model piles on the sand the WIP condition.
bed surface, and then recommencing the pluviation Moreover, the WIP condition resulted in
(Fig. 2). The use of this procedure aimed to evaluate significantly lower ultimate uplift load (Qult)
the performance of the model piles in a sand bed compared to the SCR condition. The values of Qult
unaffected by the model pile installation, which is for the WIP condition are respectively 63%, 65%
recognised here as the “wished-in-place” (WIP) and 40% lower than those of SCR condition.
condition. Two nylon threads were used passing Table 3 summarises the results of the monotonic
laterally the top of each model pile to ensure uplift tests. The larger values of Qult for HP3 WIP
verticality. Figure 2 shows the positioning of and HP1 SCR are likely due to sand bed
models during the pluviation. The whole procedure heterogeneity, and not necessarily scale effect since
is carried out at 1×g outside the centrifuge. no trend with increasing size can be identified.
The second procedure is similar to that reported in Despite the dispersion, the results indicate a
current authors’ previous studies (e.g. Tsuha and tendency of WIP condition to provide lower
Aoki, 2010; Schiavon et al., 2016). In this ultimate load and stiffness in monotonic uplift tests
procedure, the model anchors are screwed into the compared to the SCR condition.
sand bed similarly to field practice (SCR condition). The unexpected lower axial stiffness and ultimate
The vertical feed rate corresponds to 1 helix pitch load for the WIP tests may be due to possible local
per revolution, with a rotating rate of 5.3 rpm (0.56 changes in density around the pile resulting from
rad/s). The whole procedure is carried out under “umbrella” effects (Hao et al., 2019). Similar effect
macro-gravity (in flight). on density is observed in the vicinity of the
In both procedures, a minimum centre-to-centre container walls, where the sand bed is less dense
spacing of 10D was kept between model piles. compared with the central zone of the container.
Ternet (1999) verified inhomogenous zones in the
periphery of sand beds reconstituted via sand
Table 3. Results of uplift model tests (prototype scale).

Cont. Dr avg H/D Qult U(peak)
Test
No. (%) (kN) (mm)
HP1 WIP 1 95 8 38.5 45.9
HP1 SCR 4 99 8 104.5 49.7
HP2 WIP 1 95 8 31.4 51.2
HP2 SCR 2 96 8 89.3 99.4
Figure 2. Sand bed preparation for WIP condition. HP3 WIP 1 95 8 55.8 52.6
HP3 SCR 3 96 8 88.0 63.9
Before the first test in the container, three cycles raining in containers similar to those of the current
of centrifugation taking three minutes each study. In addition to the action of air vortices on the
were conducted up to the maximum g-level the falling grains, the sand deposition with reduced
153
kinetic energy in those zones results from the are deposed in a less dense assembly than the rest of
collision of grains with the container walls. the sand bed. In addition, a small amount of sand
In the current work, however, no wind vortex is grains can remain deposed in a loose state on top of
assumed to occur near the model piles because the the model pile, which can further reduce the kinetic
tests were carried out in a zone free from boundary energy of falling grains.
effects on the sand deposition. Therefore, the model Figure 6 shows photos from a quick experiment
piles may be acted as an obstacle for the falling conducted with fine silica sand and a threaded rod
grains that collide with the pile top and side. with 9.3 mm in diameter. Sand rain was simulated
similarly to the procedure described in Miura and
120 Toki (1982), but with no volume control of rain
flow. The sand rain was recorded in slow motion
100
video aiming to observe the movement of grains
after colliding with the rod. The figure shows that
Tensile force (kN)
80
significant change of direction occurs for the grains
60 colliding with the rod. In addition to the loss of
kinetic energy with the collision with the rod,
40 deflected grains can strike other free falling grains
HP1 SCR (with no previous collision) and, thus, cause a chain
20 10%D
of collision events, which will result in a broader
HP1 WIP
0
effect of deposition with reduced energy.
0 20 40 60 80 100 For investigations considering the “wished-in-
Vertical displacement (mm)
place” condition using sand pluviation for the sand
Figure 3. Load-displacement responses for HP1 model. bed preparation, “umbrella” effect could be
minimised by dividing the pile shaft into extension
120 segments and attaching them to the pile composition
each time the sand bed height reaches the targeted
100 location for the next segment. This procedure was
conducted in the study of Hao et al. (2019) with
Tensile force (kN)
80
single and multi-helix pile models in sand. The
60
above-mentioned procedure can minimise but not
eliminate variations on sand density, since short
40 shaft segment extensions can also cause “umbrella”
HP2 SCR
effects. Alternatively, when the sand deposition via
10%D
20 pluviation is not a requirement for the WIP
HP2 WIP
condition, the sand bed could be prepared with
0
0 20 40 60 80 100
compaction, vibration or a combination of both.
Figure 4. Load-displacement responses for HP2 model.
120
100
Tensile force (kN)
80
60
40
10%D HP3 SCR

20
HP3 WIP
0
0 20 40 60 80 100
Figure 5. Load-displacement responses for HP3 model.
After collision, the sand grains lose kinetic energy

and change direction. Consequently, the sand grains Figure 6. Pluviation of fine sand over a rod.
154
3.1. Comparison with numerical simulations hand, good agreement with experimental results
A Finite Element modelling in axisymmetric was obtained when the sand parameters
condition was conducted to simulate the helical pile corresponded to the constant-volume shearing
as a “wished-in-place” element. Two different set of condition (FE model 2), which may be compared to
characteristics for the sand mass were considered in the loose state of sands.
order to compare numerical and experimental
150
results. The characteristics assigned for pile and
sand are listed in Table 4.
Tensile force (kN)

Table 4. Sand parameters for numerical simulations. 100
Sand parameter FE model 1 FE model 2
γ (kN/m³) 15.3 15.3
𝑟𝑟𝑟𝑟𝑟𝑟
𝐸𝐸50 (kPa) 5.0×104 2.2×104
𝐸𝐸𝑜𝑜𝑜𝑜𝑜𝑜 (kPa) 2.6×104 2.0×104 50
𝐸𝐸𝑢𝑢𝑢𝑢 (kPa) 1.0×105 6.0×104
m 0.5 0.5
HP2
νur 0.2 0.2 0
HP1
pref (kPa) 100 100 0 10 20 30 40

𝐾𝐾0𝑁𝑁𝑁𝑁 0.32 0.5 Vertical displacement (mm)
Rf 0.9 0.9 Figure 7. Load-displacement responses from numerical and
experimental tests with WIP condition.
cref (kPa) 1.0 1.0
φ (deg.) 47 30
4 CONCLUSION
ψ (deg.) 17 1
The current study intended to investigate the
where, 𝐸𝐸50 𝑟𝑟𝑟𝑟𝑟𝑟
= secant stiffness in standard drained triaxial test,
installation effects of helical pile models in very
𝐸𝐸𝑜𝑜𝑜𝑜𝑜𝑜 , tangent stiffness for primary oedometer loading, m =
power for stress-level dependency of stiffness, 𝐸𝐸𝑢𝑢𝑢𝑢 𝑟𝑟𝑟𝑟𝑟𝑟

=
dense sand. Model piles were subjected to uplift
unloading/reloading stiffness, νur = Poisson's ratio for loading in centrifuge after being installed during the
unloading-reloading, pref = reference stress for stiffnesses, 𝐾𝐾0𝑁𝑁𝑁𝑁 sand bed preparation (“wished-in-place” condition
= coefficient of earth pressure at rest for normal consolidation, – avoiding installation effects) or installed in flight
Rf = failure ratio. by screwing it into the sand bed.
The pile models at “wished-in-place” condition
The constitutive model used for the sand was the exhibited lower ultimate uplift load, which at first
Hardening Soil Model, HS-soil (Schanz et al., could lead to the conclusion that the screwing
1999). The FE model 1 uses the parameters for the process causes strength increase of the soil around
undisturbed HN38 Hostun sand, as listed in Table 1. the pile. This is not the case. The model piles
The sand parameters for the FE model 2 were positioned on the surface of the sand bed during
determined through a parametric study searching for pluviation (WIP condition) caused umbrella effects
a pile load-displacement response with reasonable on the sand deposition around the models. As a
agreement with those from the WIP tests. The result, local soil density variation occurred in the
simulation of these two conditions aimed to sand.
demonstrate the difference between the values of The findings emphasize the importance of the
parameters if an undisturbed condition (FE sample preparation to reproduce the intended test
model 1) were in fact achieved. condition and, thus, avoid misinterpretation of
The pile element was modelled as linear-elastic results.
material having Young’s modulus E = 2.0×108 kPa
and Poisson’s ratio ν = 0.3. Interface elements were 5 ACKNOWLEDGEMENTS
assigned to the pile-sand interface with interface
The authors would like to thank WEAMEC
angle of friction δ = 20 deg. in both FE models.
(Projects REDENV-EOL and JE-CORECT), the
Both sand and pile were modelled with 15-node
Pays de la Loire Region, Nantes Metropole,
solid elements. Meshing update technique was used
CARENE, Coordenação de Aperfeiçoamento de
to avoid excessive mesh distortion.
Pessoal de Nível Superior - Brazil (Finance Code
Figure 7 compares the experimental results in
001) and USP-COFECUB Project (Grant No. UcMa
WIP condition with the numerical ones. The FE
132/12) for their financial support, which enabled
model 1 resulted an uplift resistance significantly
the study to be carried out. Special thanks go to the
larger than the 6 experimental tests. On the other
Gustave Eiffel University centrifuge team for their
155
technical support and assistance during these dissertation, University of Caen, Caen, France (in
experimental campaigns. French).
Tsuha, C.H.C. 2007. Theoretical model to control on site

6 REFERENCES
the uplift capacity of helical screw piles embedded in
Garnier, J. 2001. Physical models in geotechnics: state of sandy soil. PhD dissertation, University of São Paulo,
the art and recent advances (1st Coulomb Lecture). In São Carlos, Brazil (in Portuguese).
Proceedings of the Caquot Conference, Paris, France.
Presses de l'Ecole Nationale des Ponts et Chaussées, Tsuha, C.H.C. and Aoki, N. 2010. Relationship between
Paris, pp. 1-59. installation torque and uplift capacity of deep helical
piles in sand. Canadian Geotechnical Journal, 47(6):
Gavin, K., Doherty, P. and Tolooiyan, A. 2014. Field 635-647.
investigation of the axial resistance of helical piles in
dense sand. Canadian Geotechnical Journal, 51(11): Tsuha, C.H.C., Aoki, N., Rault, G., Thorel, L. abd
1343-1354. Garnier, J. 2012. Evaluation of the efficiencies of
helical anchor plates in sand by centrifuge model tests.
Hao, D., Wang, D., O’Loughlin, C.D. and Gaudin, C. Canadian Geotechnical Journal, 49(9): 1102-1114.
2019. Tensile monotonic capacity of helical anchors in
sand: interaction between helices. Canadian Schanz, T., Vermeer, P.A., and Bonnier, P.G. 1999. The
Geotechnical Journal, 56(10): 1534-1543. hardening soil model: formulation and verification. In
Beyond 2000 in computational geotechnics. Balkema,
Küs, B. 1992. Pluviation des sables. DEA report, Rotterdam, pp. 281-296.
Université de Nantes, Nantes, France (in French).
Schiavon, J.A. 2016. Behaviour of helical anchors
Miura, S. and Toki, S. 1982. A sample preparation subjected to cyclic loadings. PhD dissertation,
method and its effect on static and cyclic deformation- University of São Paulo and LUNAM University, São
strength properties of sand. Soils and foundations, Carlos, Brazil.
22(1): 61-77.
Schiavon, J.A., Tsuha, C.H.C. and Thorel, L. 2016. Scale
Pérez, Z.A., Schiavon, J.A., Tsuha, C.H.C., Dias, D. and effect in centrifuge tests of helical anchors in sand.
Thorel, L. 2018. Numerical and experimental study on International Journal of Physical Modelling in
influence of installation effects on behaviour of helical Geotechnics, 16(4): 185-196.
anchors in very dense sand. Canadian Geotechnical
Journal, 55(8): 1067-1080. Sharif, Y.U., Brown, M.J., Ciantia, M.O. et al. 2019.
Numerically modelling the installation and loading of
Ternet, O. 1999. Reconstitution et caractérisation des screw piles using DEM. In 1st International
massifs de sable: application aux essais en Symposium on Screw Piles for Energy Applications
centrifugeuse et en chambre de calibration. PhD (ISSPEA 2019), Dundee, Scotland. University of
Dundee, Dundee, pp. 101-108.
156
Experimental technique for creating enhanced capacity piles in a centrifuge

environment
Leonardo Lalicata1, Andrew McNamara1 and Sarah Stallebrass1
Civil Engineering, City, University of London, UK
1
Corresponding author: Leonardo Lalicata leonardo.lalicata@city.ac.uk
ABSTRACT: The paper describes experimental techniques developed in the geotechnical centrifuge facility
at City, University of London to test bored pile foundations with profiled bores designed to increase pile
capacity. Improvements were made to existing equipment to ensure accurate measurement of the load
displacement response at working loads and potential errors quantified.
Keywords: Pile Foundations; Load-Displacement; Stiff Clays.
1 INTRODUCTION The following sections outline how the test, model

Researchers at City, University of London are preparation equipment, displacement and load
collaborating with Keltbray Piling, on an Innovate monitoring systems were developed to fulfil the
UK funded project to develop an enhanced capacity above criteria.
bored cast in situ pile. This research builds on work
by Gorasia and McNamara (2016) which explored 3 THE CENTRIFUGE TEST
the enhanced capacity of bored cast in situ piles with Bored cast in situ piles are commonly used
discrete ribs or a rib shaped like a spiral along the foundations in London, constructed mainly in the
length of the pile. London Clay, a stiff overconsolidated marine clay,
The research at City, University of London with a casing provided to support overlying sands
consists primarily of a substantial series of and gravels. They are most efficient when they do
centrifuge model tests, 13 tests to date, exploring the not extend into underlying unstable strata and the
influence on the load-displacement response of the depth of London Clay available is often sufficient to
piles to changes in the geometry of a novel form of prevent this. In the field tests piles were constructed
profiling applied to the pile bore. These centrifuge in the London clay at Southall. Consequently, the
model tests are supplemented by field trials and the centrifuge tests needed to simulate the construction
results of the tests are being used to develop the of pile foundations in overconsolidated clay. The
profiling equipment used in the field by Keltbray overlying strata would not be included in the model
Piling. to simplify the analysis of the piles and ensure it was
straightforward to distinguish the effect of the
2 METHODOLOGY profiling of the bore which would not be possible in
There were a number of criteria that needed to be the cased sand and gravel layer. The soil used in the
fulfilled to ensure that the programme of centrifuge model was created from Speswhite Kaolin clay as
testing complemented the developments and field discussed below.
tests undertaken by Keltbray Piling: The Geotechnical Engineering Research Group at
 Centrifuge tests should relate sufficiently City University makes use of an Acutronic 661
closely to the field tests in terms of stress state, beam centrifuge, described in detail by Schofield
pile geometry and soil conditions and yet be and Taylor (1988). The model was tested at 50g so
well defined and straightforward to analyse. that piles representing typical diameters used in the
field can be constructed at a sensible size for the
 Tests should result in consistent load
purposes of model preparation. To maximise the
displacement data once allowance has been
number of piles tested a “plane strain” strongbox
made for changes in soil properties
was used of nominal internal dimensions 550 x 200
 It should be possible to measure displacements
x 375mm. The piles were bored, profiled and cast,
and loads accurately at working loads as well as
see section 4 at 1g and the loading apparatus was
when the pile reaches ultimate capacity.
then assembled on the plane strain strongbox.
 The method of profiling the pile bores should be Constructing the piles at 1g means that the contact
reproducible and the geometry of the between the piles and the surrounding soil is
indentations representative of those used in the optimised leading to values of the adhesion factor α
field. typically in the region of 0.75 – 1.0 (Gorasia and
McNamara, 2016). This eliminates uncertainties
157
that occur in the field due to the softening of the soil measurements are largely consistent in the tests. As
near the pile bore and will ensure that increases in might be expected, the undrained strength increases
capacity caused by profiling will be underestimated slightly with depth as water content reduces.
by the centrifuge tests.
Once pore pressure equilibrium was reached in
flight, see below, the piles were loaded until failure
at a displacement rate of 1 mm/min. Due to the high
displacement rate, pile loading takes place under
undrained soil conditions.
3.1. Preparation and characterisation of the

model soil
The Speswhite Kaolin clay used in the tests is
prepared from slurry with an initial water content of
120% which is twice the liquid limit. The slurry was
created by mixing dry powder and distilled water in
an industrial ribbon blade mixer.
The slurry was then carefully poured into the
model container and manually stirred to expel the
main air bubbles. The model container was Figure 1. Undrained shear strength profiles with depth.
previously coated with pump water grease to
minimise skin friction. Beneath the slurry there was 3.2. The experimental apparatus
a filter paper and a 3 mm porous plastic sheet, with For each test the soil sample provided up to four
an aluminium drainage plate at the base. On top of testing sites within the rectangular strong box,
the slurry a second filter paper and porous plastic Figure 2. The piles are positioned at the centreline
sheet were placed and drainage was allowed of the model, 100 mm from the sides, which is far
through holes in the loading plate in the press. The enough to minimise boundary effects (Craig, 1995).
initial slurry height (~575 mm) was much greater
than the height of the strongbox (375 mm) requiring
the use of an extension in the consolidation stage.
Consolidation was achieved by means of a
hydraulic press over a period of 9 days including 1
day of swelling. The models were compressed to a
vertical stress of 500 kPa that was then reduced to
250 kPa, inducing an in-flight over consolidation
ratio (OCR) that decreases with depth. In flight, the
water table was maintained at 10mm below the
surface of the clay by means of a standpipe supplied
continuously with water from outside the centrifuge
and connected to the bottom of the model. The top
surface was sealed with a synthetic rubber coating
to prevent clay drying during the test. (Panchal et
al., 2019; Gorasia and McNamara, 2016). This
allows the clay to achieve pore pressure equilibrium
with drainage to the base of the model after a period
of around 44 hours in flight before the piles are
loaded.
At the end of the test, the undrained shear strength Figure 2. Model geometry.
was estimated from hand vane tests and water
content samples. The water contents were converted The piles used in tests were 16 mm in diameter
to strengths using critical state theory (Wood, and 180 mm long, giving a prototype pile 800mm in
2004). Soil strength measurements for the tests are diameter 9 m long. The pile length to diameter ratio
given in Figure 1. The symbols in the Su plot are the is small if compared to typical piles used in the field
values derived from water content measurements, but a reasonable compromise given the need to
while the lines correspond to the hand vane results. provide sufficient soil under the base of the piles
In general, the latter give higher strengths. The and the limited depth of clay available. Smaller
158
diameter piles could have been used with a higher g movement of the beam if only one pile was loaded,
level, but this would have made the profiling particularly if the pile was at the end of the beam.
operation impractical. In addition, the relatively As it was not feasible to ensure that the beam
small length of the piles ensures a faster in flight encountered all piles simultaneously the
consolidation prior to pile loading. displacement of the pile heads was monitored
The piles, are spaced 110 mm apart and loaded independently.
simultaneously by means of a rigid beam. The In order to avoid any undesired eccentricity
apparatus is devised such that it is possible to obtain during the loading of the pile, a ball bearing laying
independent load and settlement data for each pile. on top of a concave plate was used to apply the load,
Several pieces of apparatus were designed and Figure 4(a). Three load cells, sandwiched between
developed to improve the experimental technique. two aluminium plates, measured the load on each
These included the loading equipment, the pile ensuring that errors due to bending in the load
measurements of pile displacement and the guides, cell are eliminated Figure 4(b).
jigs and impression tools needed to create the model
piles, as discussed below.
4 LOAD-DISPLACEMENT
MEASUREMENT
The axial load is applied from the actuator to the
piles by means of a rigid reverse T-shaped beam
(Taylor et al., 2013). The initial concept was to
measure the load whilst assuming that the loading
arrangement was sufficiently stiff to ensure that the
displacement applied to all piles was the same.
Thus, it should have only been necessary to measure
displacement on the beam. To reduce the interaction
between the beam and the loaded piles the system
was stiffened with two silver steel guides (16 mm in
diameter) positioned at the extremes of the beam.
Each guide travels through a linear bearing 100 mm
long and is connected to the other guide above the
plate carrying the actuator by a 40 mm thick silver
steel plate, Figure 3.
Figure 4. Load measurement system (a) vertical cross section,

(b) detail of the load cells plate.
Several attempts were made to obtain reliable

readings of the head displacement of the pile, the
results and the consequences for the response of the
pile will be discussed later in the paper. The current
measuring system is sketched in Figure 5. A 2 mm
Figure 3. Loading apparatus. thick aluminium plate, 175 mm long and 16 mm
wide, is a tight fit to the pile collar and two LVDTs
Although this resulted in a stiff loading system measure the plate displacement at its extremes. The
that would not deflect when loading all piles head displacement of the pile is given by the mean
simultaneously, it was not possible to prevent some value of the LVDT measurements. A 1 mm gap
159
between the plate and the concave dish guarantees resin was found to have a Young’s Modulus equal
that load and displacement are independent to 1.1 GPa and a yield stress of 35 MPa. These
readings. values confirm that the pile behaves as a linear
elastic material in the range of the applied loads.
Figure 5. System for the measurement of the head

displacement of the pile.
5 MODEL PILES
As noted in Section 3, the piles were cored, profiled
and cast at 1g. For results to be repeatable it was key
that the piles were vertical, a constant length into the
clay and the profiling was precisely located and well
defined.
The piles were cut using a hypodermic thin walled
tube with an external diameter of 16 mm. The tube
was mounted onto a brass handle to allow for easy
cutting (Gorasia & McNamara, 2016). The
verticality and position of the piles was achieved Figure 6. (a) guide for pile boring, (b) impression.
using the system of guides shown in Figure 6 (a). To
minimise soil disturbance during boring several
precautions were undertaken: the internal part of the
tube was sprayed with a lubricant oil and the edge
of the tube was sharpened. In addition, the boring
was always undertaken in three steps of equal
length.
The enhanced capacity piles were created from
the plain pile bore. A custom designed tool was
inserted into the guide and used to profile the pile
bore Figure 6 (b). After profiling, the piles were cast
using a ‘fast cast’ polyurethane resin, Sika Biresin
G27 (McNamara, 2001; Gorasia and McNamara,
2016). The pot life of the resin is approximately 3 Figure 7. Exhumed model piles some with profiled bores
minutes. Aluminium powder was used as filler in an
equal mass ratio with the two components of the These values also allowed the pile-soil stiffness
resin to ensure that the pile is not buoyant. The ratio to be estimated from standard relationships
mixture was designed to have a good fluidity to fill between soil stiffness and strength. For a low
the profiles. Figure 7 shows some of the exhumed plasticity, heavily overconsolidated clay such as
piles and demonstrates the success of the that used here, the stiffness/strength ratio is
methodology adopted. Several uniaxial typically given by Eu/Su =300 leading to relative
compression tests were undertaken to measure the stiffness ratio, between pile and soil of
mechanical properties of the resin when set. The approximately 90. Therefore, the pile is expected
160
behave as a compressible pile with most of the that. This is likely to be related to some small
interaction concentrated in the upper part of the (~1/400 of rotation) rocking of the beam due to the
shaft as might be expected of a full length pile used temporary asymmetry of the loading and the
in the field tests. relatively high stiffness of the piles. The other piles
(2 and 3, in the central position) loaded last, displace
6. ANALYSIS OF RESULTS with approximately the same loading rate as that
6.1 Repeatability imposed. That is because the external piles have
“balanced” the beam by applying roughly the same
Figure 8 shows data from all tests on straight-sided reaction and because piles 2 and 3 are closer to the
model piles interpreted in a consistent manner to centre of the beam, where the effect of rocking is
give the ultimate capacity of the pile. These ultimate lower. This behaviour is not affected by the
capacities have been plotted against undrained shear configuration of the straight-sided and profiled
strength calculated using both water content piles.
measurements and vane test data. In general, the
ultimate loads correlate well with undrained
strengths calculated from water content data and
show a slight increase in ultimate load with
undrained strength as would be expected.
Figure 8. Ultimate load of all straight sided model piles plotted

against the undrained shear strength at z=L/2=90 mm, where
L is the length of the pile.
6.2 Pile-beam- pile response

The typical evolution with time of displacement and
load for a set of four straight-sided and profiled piles
are presented in Figure 9(a) and Figure 9(b)
respectively. The dashed line represents the
imposed displacement. For each pile, the Figure 9. Evolution with time of: (a) displacements of the piles,
displacement is given by the average of the LVDTs (b) loads of the piles
and the load is the sum of the LCs measurements.
The piles are displaced over a time interval of 2.5 The consequence of this effect is depicted in
minutes that approximately corresponds to the Figure 10(a) and Figure 10 (b) that show the load-
accuracy of the collar installation into the bore (± 1 settlement response for an external (1) and an
mm), given that the displacement rate of the beam internal pile (3) respectively. In each figure, the
should be 1 mm/min. Inspection of the measured load-settlement curve is compared with
displacements, Figure 9 (a), shows that in the early the one that would be obtained assuming that the
stages of loading, the pile head displacements are pile displacement was equal to that imposed. The
different from those imposed. underestimation of the initial pile stiffness in the
This is particularly true for the first two piles (4 latter case is apparent in both cases: 63% for pile 1
and 1) that are at either end of the loading beam. For and 43% for pile 3. The difference reduces in the
pile 4 the displacement curve is concave at the second case as the pile is loaded after piles 4 and 1
beginning and then, as the beam touches pile 1, the and it is closer to the centre of the beam. It is thus
concavity changes and the trend become parallel to clear that in order to obtain reliable measurements
161
at low deformation levels (i.e. working load) it is displacements separately. The consequence of
critical to measure the head displacement of the pile measuring the pile head movements directly
directly. However, the value of the failure load is compared to measuring the displacement of the
unaffected by errors in the displacement loading beam has been quantified.
measurements. Compromises in the modelling of the overall pile
geometry have been discussed and it has been
established that the soil/pile stiffness used ensures
that the pile will respond as a compressible pile even
though it is comparatively short. It has been shown
that consistent measurements of pile capacity can be
obtained if changes in undrained strength are
accounted for, demonstrating that the methodology
for the preparation of the soil bed and the
construction of the piles provides repeatable results
that can be used to generate data for a parametric
study of the influence of the profiling on pile
behaviour.
8 ACKNOWLEDGEMENTS
The authors would like to acknowledge the support
of Innovate UK and Keltbray Piling.
9 REFERENCES
Craig, W.H. 1995. Geotechnical centrifuges: past,
present and future. In Geotechnical Centrifuge
Technology (Taylor RN (ed.)). Blackie Academic and
Professional, Glasgow, UK, ch. 1, pp. 1–19.
Gorasia, R.J. and McNamara A.M. 2016. High-capacity

ribbed pile foundations. Proceedings of the Institution
of Civil Engineers: Geotechnical Engineering, 169(3):
264–275.
Figure 10. Direct measures of pile displacement compared with
that expected from the beam movement: (a) external pile, (b)
central pile.
McNamara, A.M. 2001. Influence of Heave Reducing
Piles on Ground Movements Around Excavations. PhD
thesis, City University, London, UK.
7 CONCLUSIONS
A series of centrifuge tests have been designed to Panchal, J.P., McNamara, A.M., Halai, H. and Divall, S.
characterise the increase in pile capacity obtained 2019. Centrifuge modelling to determine the influence
by profiling the bore of a bored cast in situ pile. In of pile stiffness on pile capacity. Proceedings of the
addition to pile capacity the tests also needed to XVII ECSMGE-2019 Geotechnical Engineering
measure changes in load displacement response at foundation of the future. doi: 10.32075/17ECSMGE-
2019-0036.
working load, which required accurate
measurement of both load and in particular Schofield, A.N. and Taylor R.N. 1988. Development of
displacement at displacements of less than 0.2mm. standard geotechnical centrifuge operations. In
The aim was to produce a substantial series of tests Centrifuge 88 (Corte JF (ed.)). Balkema, Rotterdam,
that would form a parametric study to be compared the Netherlands, pp. 29–32.
to a limited number of field tests.
Apparatus has been developed that loads four Taylor, R.N., Rose, A.V. and Gorasia, R.J. 2013. Pile and
piles always including a straight unprofiled pile for Pile Group Capacity: Some Findings from Centrifuge
comparison. The loading system, despite being Tests. International Journal of Geo-Engineering, 5(2),
modified to improve its rigidity, did not provide a 5-15.
sufficiently accurate or reliable means of measuring
Wood, D.M. 2004. Geotechnical modelling. New York:
the displacement of the pile head and consequently Spon Press.
a system was developed to measure these
162
Model tests of grouted piles in gravel esker and fill soils

Wik Breure1, Anders B Lundberg1
Geotechnical Department, ELU Konsult AB, Sweden
1
Corresponding author: Wik Breure (wik.breure@elu.se)
ABSTRACT: Grouted piles consist of an internal steel member used for drilling and grouting, and a grout
body created during installation. The foundation method is highly efficient in permeable soils, such as the
esker gravels frequently found in Sweden. During the preliminary design phase for the Slussen project in
the centre of Stockholm, a series of in-situ model tests were performed and evaluated at the start of the
project when grouted piles were considered as a foundation method. The objective of the model tests was
to evaluate the installation procedure in the specific soil conditions, assess the load distribution and load
bearing capacity during the static load tests, and the corrosion protection of the central steel member of the
grouted piles. The tests were carried out at two different locations within the specific project. The challenge
was to estimate and generalize the efficiency of the foundation method in the entire area. This paper
discusses the in-situ model tests, especially the load test results, and elaborates on the suitability of the
grouted piles for the specific project.
Keywords: Grouted Piles; Model Testing; Esker Geology.
1 INTRODUCTION is similar to grouted anchors: a grout mix is injected

The suitability of ground engineering methods is at low pressure into the ground during rotation or
highly depending on the local soil conditions. penetration. Laboratory tests of grouted piles in
Figure 1 shows an idealised cross-section of the calcareous soil have been conducted by Lee and
Swedish soil strata, which is dominated by soft soils Poulos (1991), and on grouted anchors by Ostmayer
deposited during the Holocene period as a result of and Scheele (1978). Miltenberger (2001)
the Weichselian glaciation. The soft clays and recommends a characteristic shaft resistance for
shallow bedrock in the Eastern part of the country grout anchors. However, the standard design
permit end-bearing piles (Lundberg and Li 2015), method outlined by Miltenberger requires that the
while the shaft-bearing piles are more suitable for ground soil conditions are suitable for grouted
the deeper clay deposits in Western Sweden anchors/piles, especially regarding the development
(Jendeby 1991). of the grout body. In the current paper a case study
Eskers are a particular feature of the Swedish of in-situ model tests in esker/fill soil layers is
geology, sand and gravel ridges located along elaborated to assess the suitability of the method in
subglacial rivers, in which the flow of water resulted the project specific soil strata.
in a soil mass with a homogenous grain size. These
are highly permeable and are frequently used as 2 CASE STUDY: SLUSSEN AREA IN
groundwater reservoirs. STOCKHOLM
The traditional piling methodology using end- The geological conditions in the Slussen area result
baring piles in which the bearing capacity is verified both from the natural deposition of sediments and
by dynamic pile load tests as outlined in Andersson the development of the city of Stockholm since the
et al. (2016) is frequently not suitable for esker soils medieval period. The city of Stockholm was
since the bedrock is typically located at a deeper originally built on an esker ridge, extending from
level along the esker ridges. Application of the the South to the North of the city.
European continental piling methods, i.e. bored Deposition of fine sediments from the Mälaren
piles or continuous flight auger (CFA) piles is often drainage basin occurred along the original coastline
not possible, due to the boulders located in the esker before the current quays were completed: as a result,
soil mass. the slopes around the original coastline consisted of
An alternative solution is the application of massive layers of gyttja and organic soil.
grouted piles. These are constructed through a The city of Stockholm was founded in 1252
casing, which is drilled through layers containing (Hall 2008). The urban development soon extended
boulders. Various types of grouted piles are to the South and North of the city. The organic and
discussed in the scientific literature, e.g. large inorganic waste was deposited along the coastline,
diameter grouted piles, e.g. Abbs et al. (1985) and consisting everything from food waste to massive
Nauroy et al. (1985). The principle of grouted piles boulders, as the city gradually expanded. The layers
163
of soft soil were covered by this mixed fill, and piles, such as CFA piles, tenuous, and the
quays were built along the new coastline. significant depth to the bedrock (often exceeding 50
The Slussen area shown in Figure 2 is now m) makes end-bearing piles time-consuming and
undergoing an extensive reconstruction. During the uneconomical.
project the quays are widened, and superstructures The use of grouted piles was therefore
for a subway and bus terminal, and a new bridge are investigated to assess the suitability of the method,
being built, all requiring piles for ground support. and a series of model tests were performed, in which
The soil strata presents specific complications the whole installation and testing procedure was
for standard foundations methods: the large boulder executed in-situ.
content in the fill makes the construction of bored
Figure 1. Idealisation of the soil strata in Sweden.
Figure 2. The Slussen area in Stockholm.
164
3 MODEL TESTS
3.1. Aims of the tests
The aims of the in-situ model tests were to assess
the feasibility of the method with regards to:
- The construction of the piles

- Bearing capacity of the piles
- Load distribution along the piles.
The construction of the piles consisted of drilling

of a casing through the fill layer, after which the pile
was grouted through the casing. The high
permeability of the esker material was considered a
risk, since a large amount of grout would
presumably be needed for the piles, resulting in
environmental and economic costs. Assessment of
the resulting bearing capacity from the model tests
was essential, since they provided the representative
shaft resistance used during pile design. The
distribution of the load along the grouted piles was
also if interest: the aim was to separate the casing
and the grout body with a plastic tube to reduce the
influence of negative skin friction if settlements
would occur in the fill layer.
3.2. Model test set-up

For the model tests drilling and grouting equipment
was applied for the construction and
instrumentation of the nine test piles, after which
compression and tension pile load tests were
performed.
3.2.1 Installation
The principle of the model test piles is shown in
Figure 3. Drilling was conducted from the ground
surface through the fill layer. A 140 x 8 mm steel
casing with a length of 21-27 m was drilled through
the fill layer and terminated at the top of the esker
layer. The stratification is interpreted from site
investigation. At the top of the esker layer, a hollow
73 mm diameter steel pile with 53 mm internal
diameter is drilled through the soil with a drill bit,
which is simultaneously pumping cement grout
through the soil. Before the drilling and grouting of
the pile starts, a plastic pile is put on top of the
hollow steel pile to separate the steel casing and the
grouted pile. The grouted length varies between 4-6
m. A water-cement ratio of 0,5-0,7 is used for the Figure 3. Schematic representation of the principle of the
grouting. Spoil was collected during drilling. model pile tests, length in mm.
Compression tests were carried out on the spoil
from the grouting process with water-cement ratio 3.2.2 Instrumentation
of 0,5. The cube compression strength after 7 days The grouted pile was instrumented with strain
was 39,0 MPa, after 17 days 41,5 MPa and after 29 gauges at c/c 1 m through the hollow steel pile,
days 48,9 MPa. shown in Figure 3, as well as through the casing. A
temperature sensor was installed in the middle of the
165
grout body to detect any anomalies during grouting. measurements by the strain gauges. The strain
Vibrations and settlements were measured during gauges mounted on the steel casing display a
the installation of the piles, and the impact on the relatively uniform elastic elongation of the steel
ground water level was monitored during and after casing, indicating a relatively uniform mobilization
the tests. of the shaft friction along the steel casing.
The Strain measurements inside the grout body
3.2.3 Compression and tension pile load tests display a more gradual mobilization with less strain
Compression and tension static pile load tests were at the lower end of the pile.
performed on the piles. A beam frame with a
hydraulic apparatus was used for the test and the 4.4. Dynamic pile load tests
surrounding piles were used to mobilize the reaction Dynamic pile load tests were performed on some of
for the tests. Dynamic pile load tests were the grouted piles. Figure 6 shows the results of a
subsequently executed. dynamic pile load test on one pile, displaying the
force and velocity multiplied by the pile impedance
4 RESULTS output. These results show a different response
4.1. Grout flow during installation compared to the static pile load tests, since the
reflection of the steel casing is less visible. The
The grout flow showed very large variation return signal from the grouted piles is more
between the piles, possibly resulting from the significant. The reason behind this could be the
variation in permeability of the esker layer. The viscous behaviour of the plastic tube.
flow was at a maximum level for the piles at 21
m grouting depth, surpassing 2000 L/m. At 27 4.5. Excavation of the piles
m depth the grout flow was lower, but with a The construction of grouted piles was also tested at
maximum level of around 1400 L/m and a another location in the area. These piles were
minimum level of 200 L/m. installed by means of the described method. At a
4.2. Static pile load tests later stage during the construction of the Slussen
area, these test piles were excavated, and the result
Compressive and tension pile load tests were carried of the grouting process could be examined.
out on the piles. Figure 4 shows a representative Figure 7 show the grout body including the visible
time versus load response of a tested pile in tension. the intersection between the man-made fill, which is
Both the total load and the load retained by the steel mixed with organic soil and contains boulders, and
casing are shown in Figure 4. The steel casing ratio the esker gravel. This indicates that for this specific
of the load was calculated through the measured pile the steel casing ended above the intersection
load on the top strain gauge on the hollow steel tube between the fill and esker material. The esker has a
compared to the total pile load in tension. lower boulder content, but still shows a significant
Theoretically the plastic tube would permit all the irregularity at the lower end of the grouted pile.
load being transferred to the grout body, but a
significant part of the load was retained by the steel
4.6. Other measurements
casing through shaft friction.
Figure 4 displays a very stiff response during the The vibration level and impact on the ground water
initial loading stages, which is gradually level were also monitored during the installation of
transformed to a more viscous response at around the piles and the subsequent testing activity, but
800-900 kN loads. The load ratio of the steel casing these did not display any large effect compared to
appears to reduce gradually when the load increases. the background noise, which was significant, since
the test was conducted in the centre of Stockholm.
4.3. Load distribution Settlement measurements at the testing location
showed settlements of around 6-9 mm between the
Figure 5 shows a representative obtained strain piles.
distribution along a tested pile through
166
Figure 4. Representative response static pile load test.
Figure 5. Representative load distribution response during static pile load tests. The legend represents the distance of the strain
gauges with respect to the surface in meters.
167
5 CONCLUSIONS AND DISCUSSION
The in-situ tests showed that:
- The method resulted in the construction of
grouted piles with an expected bearing
capacity.
- The installation process resulted in
excessive grout flow.
- It was not possible to separate the load
between the steel casing and the grouted
pile in the static pile load tests, as the
results showed that a significant part of the
Figure 6. Response dynamic pile load test. load was retained by the steel casing
through shaft friction.
Due to these reasons, the grouted pile method was
not used for a large volume of piles within the
Slussen project, but could still be used as an
alternative for some areas, due to the minor risk of
settlements and the negligible impact on vibration
and ground water levels.
6 REFERENCES
Abbs, A. F. and Needham, A. D. 1985. Grouted piles in
weak carbonate rocks. In Offshore Technology
Conference. Offshore Technology Conference.
Andersson, A. Axelsson. G. and Lundberg, A.B. 2016.

(a) Reliability analysis of pile groups based on dynamic
pile testing. In Nordic geotechnical conference 2016
pp. 503-512.
Hall, T. 2008. Stockholm: the making of a metropolis.

Routledge, Abingdon.
Jendeby, L. 1991. Load transfer in friction piled

foundation in soft clay. In Proc. 10th Eur. Conf. on
SMFE (Vol. 1, pp. 449-452).
Lee, C. Y. and Poulos, H. G. 1991. Tests on model

(b)
instrumented grouted piles in offshore calcareous soil.
Journal of geotechnical engineering, 117(11), 1738-
1753.
Lundberg, A. B. and Li, Y. 2015. Probabilistic

characterization of a soft Scandinavian clay supporting
a light quay structure. Geotechnical safety and risk V.
IOS Press, Amsterdam, 170-175.
Miltenberger, M. A. 2001. Capacity design of grouted

anchors. Transactions, SMiRT 16
Nauroy, J. F. and LeTirant, P. 1985. Driven piles and

drilled and grouted piles in calcareous sands. In
Offshore Technology Conference. Offshore
Technology Conference.
(c)
Ostermayer, H. and Scheele, F. 1978. Research on
Figure 7. Excavated grout body.
ground anchors in non-cohesive soils. Revue Française
de Géotechnique, (3), 92-97.
168
4. Infrastructures
Geotechnical instrumentation of an experimental embankment dam

Jasmina Toromanovic1, Johan Lagerlund1,2, Peter Viklander1,2,3, Jan Laue1
1
Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, Sweden
2
Vattenfall Research & Development, Sweden
3
Vattenfall Vattenkraft AB, Sweden
Corresponding author: Jasmina Toromanovic, jasmina.toromanovic@ltu.se
ABSTRACT: An experimental embankment dam was constructed in Älvkarleby (Sweden) in 2019, in order
to allow for geotechnical experiments under realistic and, at the same time, in a controlled environment. The
embankment dam is a zoned earth fill embankment dam with a central core of fine-grained till. The
embankment dam is 4 m high, 20 m long with a base of 15 m, i.e. a medium scale model. The embankment
dam is built in a “concrete box” and a tent has been raised over the site, which will protect from freezing
temperatures and precipitation, thereby creating controlled conditions. In the paper, the construction of the
embankment dam and the geotechnical instrumentation is presented. No results of the geotechnical
measurements are presented, as this is an early stage of the embankment dam experiment and as data is yet
to be acquired.
The geotechnical instrumentation installed are deformation measurements, pore pressure- and strain
monitoring devices. All geotechnical instrumentation is placed in one section of the dam. The deformations
are monitored by shape accelerometer arrays, which allows for 3D monitoring of deformations. Vibrating
wire piezometers are placed in the core and downstream filter, mainly with focus on the saturation phase of
the core soil. Strain measurements with fibre optics are performed in the bottom of the dam and done
perpendicularly to the dam axis. The reservoir of the embankment dam will be filled during winter/spring
2020. The measurements are planned to be ongoing until summer 2021. In the end, different loading
scenarios as well as grouting will be tested. Data from the geotechnical monitoring will be used as input for
inverse modelling in order to back-analyse the values of material parameters. Re-calibration of the numerical
models will be done in order to detect possible changes in the values of material parameters.
Keywords: Embankment Dam; Deformation; Pore Water Pressure; Vibrating wire Piezometer; Strain; Fibre Optics.
1 INTRODUCTION Today, numerical modelling is often used as

In Sweden, most of the existing hydropower dams support when predicting how an embankment dam
were built before the 1960s. Thus, there are a is going to behave over time. In order to do these
significant number of ageing structures to maintain predictions, data describing the mechanical
today. As an embankment dam ages, deterioration properties of the dam are required. Sufficient data to
of the structure may occur. Erosive processes both describe the material in the different dam zones are
inside and outside the embankment dam may affect seldom available as documentation from the
its performance as well as the structural stability of construction time are lacking. As internal erosion
the dam negatively. processes takes place inside the dam, mechanical
Erosion processes taking place on the surface of properties may also change over time. Even if data
the embankment dam are easier to detect than may be available, it is not necessarily that the
internal erosion processes taking place inside the current situation is described properly or matches
embankment dam. Typical damages on the surface today’s state. Since the extraction of physical
of the embankment dam may be settlements, samples from embankment dam’s core material may
sinkholes and erosion of blocks from the slopes. have negative consequences, it would be beneficial
Internal erosion processes inside the embankment to develop non-destructive methods for obtaining
dam are more difficult to detect and can go on for mechanical properties. Such a non-destructive
decades without being detected, ICOLD (2013) method is inverse analysis, which has been applied
high lights four types of internal erosion. If erosion to embankment dams (Toromanovic et al. 2020).
processes are left unattended, they may pose the In 2019, an embankment dam was constructed, as
embankment dam at serious risk of being damaged. a collaboration between Vattenfall AB, Luleå
Damages that could lead to collapse and in a worst- University of Technology, Uppsala University,
case scenario - casualties. There is a need of Lund University of Technology and HydroResearch
detecting these processes before they become AB. The scope of the project is two folded. One part
critical. is to study the mechanical behaviour of an
171
embankment dam during its first filling, allowing to
follow the behaviour over time and thereafter to use
the data in inverse analysis. Another part was to
localise a number of imbedded damages within the
embankment dam by geophysical methods (not
reported in this paper).
Measurements will be done before, during
reservoir filling as well as operation of the
embankment dam. Surveillance of the mechanical
behaviour of the embankment dam is performed by
Luleå University of Technology.
Typical damages that could be caused by erosion
processes were built in at known locations within
the dam core. After the construction of the Figure 1. A view of the supportive concrete structure with the
embankment dam, the built-in damages are to be raised embankment dam inside.
found and located by a team of researchers using
temperature cables, resistivity measurements and 3 THE EMBANKMENT DAM
seismic methods. The locations of the damages are
3.1. Design
not known by this team, as it is a sort of blind test.
Responsible for localising the damages by An embankment dam is a type of earthfill dam built
geophysics are Uppsala University, Lund with soil materials, i.e. gravel, sand and till and
University of Technology and HydroResearch AB. rockfill. The embankment dam is constructed with
different zones. In the centre, the core is built. The
2 THE SITE AND SUPPORTIVE core in Swedish embankment dams are most
commonly built with till, which is a broadly graded
STRUCTURES
type of soil commonly found in Sweden. The till has
The site for the test dam is in Älvkarleby, Sweden, a low hydraulic conductivity when compacted well
located approximately 150 km north of Stockholm, and is hydraulic barrier within the dam keeping the
where Vattenfall AB has its R&D laboratory. In water in the reservoir. Outside of the core, one or
order to build an embankment dam on dry land and several filter zones (e.g. fine and coarse filter) are
have the ability to operate it as a proper built. These filter zones are built both upstream and
embankment dam, a support structure was needed. downstream of the core to keep the core soil in place
The support structure had to be water tight and and prevent internal erosion i.e. fine grained
strong enough to support an embankment dam. particles from being washed out with the seepage.
Thus, it was constructed with reinforced concrete. Filter zones are typically built with sand and/or
Its dimensions were 20 m x 16 m and founded 4 m gravel.
below the natural ground surface. The bottom Outside the filters on both sides, the shoulder fill
concrete plate was made with glass fibre is to be found. The shoulder fill may consist of soil
reinforcement bars, in order not to interfere with any material ranging in size from sand to rock. The
geophysical methods, i.e. to simulate natural rock shoulder fill provides the embankment dam with
surface. The base of the support structure is leaned structural stability.
1° towards the downstream side of the embankment When designing an embankment dam it is vital to
dam where seepage water is collected. The two side choose the materials constituting each type in such
walls leaned outward with a ratio of 1H:8V in order way that internal erosion is prevented. If the fine-
to improve the compaction of the soil materials grained material is eroded from the core, its water
against the abutments. An overview of the built sealant ability will be diminished. This process may
embankment dam is shown in Figure 1. damage the core soil and make the embankment
In order to provide a controlled laboratory dam less safe in terms of structural stability.
environment, a tent was raised over the structure, Wherever seepage crosses a boundary from a finer
protecting the dam from precipitation and freezing to coarser grained soil, the potential exists for
temperatures. Water supply to the dam is provided particles to escape from the finer soil into the
from Dalälven (Dala river), which is located nearby coarser. By combining the particle size of each
the test site. adjoining zone correctly, internal erosion may be
172
Figure 2. Section of the embankment dam.
Core Fine filter Coarse filter Shoulder

100
80
Ammount passing [%]
60
40
20
0
0,01 0,1 1 10 100 1000
Particle diameter [mm]
Figure 3. Particle size distributions for the materials used in the embankment dam.
Table 1. Values of material parameters in the embankment dam.

Parameter Unit Core Fine filter Coarse filter Shoulder
Unit weight, γ kN/m3 22.5 19 17 18
Friction angle, φ ° 42 36 40 41
Hydraulic conductivity, k m/s 10-7 10-5 10-2 10-1
prevented. The proper combination of particle size project. A total of 1500 ton of material was used in
distribution curves is described by Terzaghi (1996). the construction of the embankment dam. An
In Sweden, the design and construction of a dam is overview of some material characteristics in the
governed by dam safety guidelines given in RIDAS dam is presented in Table 1.
(Energiföretagen, 2019). The embankment dam
herein was designed to have four material zones; 3.2. Construction
core, fine-, coarse filters and shoulder, Figure 2. It took four weeks to complete the construction of
The core was a fine-grained till 0-32 mm with the embankment dam. The embankment dam was
19% content of fines and the other zones were made built in layers, by following a specific pallet plan.
from crushed rock in various fractions. All particle Each layer was 20 cm thick and was compacted
size distributions can be found in Figure 3. These before next layer was placed.
distributions are from the design stage of the
173
Figure 4. Geotechnical instrumentation in the dam
The sequence of construction was always to start devices. All instrumentation was placed in one
each new layer with the core soil followed by the section of the embankment dam. The location of the
fine filter, coarse filter and lastly the shoulder fill instrumentation is shown in Figure 4.
and thereafter compaction was done with a
vibratory rammer. The compaction was done in the 4.1. Deformation
same order as the materials were placed. After each The deformations in the dam were monitored by
layer, the compaction of the core soil was measured shape accelerometer arrays (SAA), which allow for
with balloon test. The degree of compaction was 90- deformation monitoring in x, y and z directions. The
95% of modified Proctor density. technology is based on MEMS technology (micro-
At six predetermined locations, predefined electro-mechanical system). More information is
damages were built into the embankment dam and found in Danisch et al. (2007). The system is able to
their position was measured with a total station. The simultaneously measure deformations (3D) as well
built-in damages are resampling defects that could as accelerations, based on the tilt from the
be found in reality. The upstream shoulder was not gravitational direction.
compacted and the downstream shoulder was just The installed sensors are of type Measurand
compacted by an excavator. SAAV250. The rigid segments used were 0.25 m, in
Throughout the construction, measurement order to fit the scale of the project. Three
equipment was placed into the embankment dam. inclinometers have been installed, with position and
The positioning of the geophysical equipment length as follows (given in Figure 4):
(cables for resistivity, temperature and seismic)  Upstream, 3.5 m
were placed in the fine filter. The positioning of the
 Middle, 4.0 m
geotechnical measurement equipment can be seen in
Figure 4. All cables from the equipment, except  Downstream, 3.0 m
fibre optics, were led into a logging station. X-direction is from upstream to downstream, with
Other measurements include the collection of just a slight deviation. This was documented and
seepage downstream the embankment dam. A taken into consideration when performing the
20 cm deep drainage ditch, running parallel to the interpretation of movements in the dam body. The
toe of the dam, collected the leakage water and led y-direction is along with the dam axis and z is down-
it to a well. The embankment dam was divided into /upwards. During the construction of the dam,
seven sections. At the drainage ditch, the seepage PipeLife tubes were installed. When installing the
from each section could be measured separately by sensors, they were guided down in the tubes.
Thomson weirs. Continuous measurements of the deformations are
on-going. The data is available online, providing the
4 GEOTECHNICAL possibility to follow the development in real time.
INSTRUMENTATION Temperatures are also measured by the devices.
Measurements from finished construction until
The geotechnical instrumentation was consisting of February 2020 have shown minor movements. This
deformation-, pore pressure- and strain monitoring was expected and was mainly attributed to the
174
activities around the test embankment facility as cables have to cross all material zones. Thus, the
levelling of the ground and setup of the tent. material is varying from the fine-grained till in the
core to a more boulder-sized material in the
4.2. Pore pressure shoulder of the dam. In the area of the shoulder and
Vibrating wire piezometers were placed in the core filter zones, geotextile was first placed and
and filters of the dam. The pore pressure transducers backfilled with sand. Bentonite was used to seal the
are of the type Geosense VWP-3400, with a cables in the core. The cables were pre-tensioned
measuring range of -70 kPa to 345 kPa. The when installed.
vibrating wire within the sensor is in a closed The fibre optic cables were installed in a loop,
system, which will react to the pressure changes consisting of four sections through the dam, see
outside the transducer. However, this also mean that Figure 6. When construction of the embankment
the measurements are affected by the atmospheric dam was finished, the clamps providing the pre-
pressure and needs to be corrected due to that. tension were loosened, as the cables were already
12 vibrating wire piezometers were installed fixed in the material and thereby “locked” in place.
during construction of the dam, where they were
placed directly on a compacted layer. The
placement of the piezometers was done according to
Figure 4. A pore pressure transducer is shown in
Figure 5.
Figure 5. Pore pressure transducer placed in the

fine filter.
Continuous measurements of the pore pressure in

the dam will be performed. Measurements so far
have only shown that the equipment is working
Figure 6. Installation of fibre optics cables.
since water has not yet been filled in the reservoir.
With time, saturation of the core soil and filters will
Measurements will be performed at certain times,
be monitored.
especially when there are changes in the loading
scenarios as when the reservoir of the dam is to be
4.3. Strain filled. The zero-measurement readout was
Strain measurements with fibre optics are successful, showing that the cables were working
performed in the bottom of the embankment dam. without losses. This was conducted in end of
The measurements are done perpendicularly to the November 2019.
dam axis in four sections. The application of strain
measurements used here was developed in the work 5 CONTINUED MONITORING AND USE
by Iten (2011). OF THE DATA
Installations have been done with two fibre optic
cables; one for measuring strain and one for The data from the geotechnical instrumentation is
temperature compensation. The cables have been mainly planned to be used in inverse modelling.
installed in a shielded bed of sand. Since the cables Inverse modelling of the embankment dam is
are placed perpendicularly to the dam axis, the aiming to identify the values of the material
parameters of the soil in the different zones.
175
With time, if changes are occurring in the The authors would furthermore like to thank
embankment dam, this could be reflected in the Vattenfall AB for their contribution in building the
results of the inverse analysis. The analysis could test facility and to care for its proper operation.
then be resulting in changes in values of material Energiforsk’s Dam safety program is also
parameters with time. The amount of data to be acknowledged for support. Cautus Geo AS and
acquired will allow for re-calibration and an Marmorta Engineering AG are acknowledged for
upgrade of the values of the material parameters. the support installing the measurement system.
In reality, this amount of geotechnical
instrumentation is very seldom found in existing 7 REFERENCES
dams. Having so many installations as in this Danisch, L.A., Lowery-Simpson, M.S., and Abdoun, T.
project, gives an opportunity to determine how 2007. Shape-acceleration measurement device and
much instrumentation would be sufficient to method, US Patent US7296363 B2.
perform a reliable inverse analysis. This could lead
to further recommendations on how to conduct ICOLD. 2013. Bulletin 164: Internal erosion of existing
inverse analysis for other embankment dams. dams, levees, and dikes, and their foundations –
Volume 1: Internal erosion processes and engineering
assessment. ICOLD.
6 ACKNOWLEDGEMENTS
The research presented in this paper was carried out Iten, M. 2011. Novel Applications of Distributed Fiber-
as a part of ‘Swedish Hydropower Centre – SVC’. optic Sensing in Geotechnical Engineering. ETH
SVC was established by the Swedish Energy Zurich, Switzerland.
Agency, Energiforsk and Svenska kraftnät together
with Luleå University of Technology, KTH Royal Energiföretagen. 2019. RIDAS - Kraftföretagens
Institute of Technology, Chalmers University of riktlinjer för dammsäkerhet. Energiföretagen,
Stockholm, Sweden. (In Swedish)
Technology and Uppsala University. Participating
companies and industry associations are: Andritz Terzaghi, K., Peck, R., Gholamreza, M. 1996. Soil
Hydro, Boliden, Fortum Generation, Holmen Mechanics in Engineering Practice, Third Edition.
Energi, Jämtkraft, Karlstads Energi, LKAB, John Wiley & Sons Inc., New York, USA.
Mälarenergi, Norconsult, Rainpower, Skellefteå
Kraft, Sollefteåforsens, Statkraft Sverige, Sweco Toromanovic, J., Mattsson, H., Knutsson, S. and Laue, J.
Energuide, Sweco Infrastructure, Tekniska verken i 2020. Parameter identification for an embankment dam
Linköping, Uniper, Vattenfall R&D, Vattenfall using noisy field data. Proceedings of the Institution of
Vattenkraft, Voith Hydro, WSP Sverige, Civil Engineers - Geotechnical Engineering, Available
Zinkgruvan and AFRY. from: https://doi.org/10.1680/jgeen.19.00163.
176
Surface wave propagation in centrifuge testing using Electromagnetic Drop-

Projectile system
Vipul Kumar1,2, S.P.G. Madabhushi1
Schofield Centre, Department of Engineering, University of Cambridge, UK
1
Geotechnical Engineer, Mott MacDonald, UK

2
Corresponding author: Vipul Kumar (vk337@cam.ac.uk)
ABSTRACT: Surface wave propagation can be experimentally investigated with the help of centrifuge
modelling. In this paper, we discuss the development of an electromagnetic drop-projectile apparatus which
will be used to generate vibrations in the soil upon impact onto a shallow foundation. This apparatus was
developed as an improvement over the drop-ball apparatus developed by previous researchers. The drop
projectile assembly allows to vary drop height and perform multiple drop tests during a single flight of the
centrifuge package. This system can be used to generate vertical vibrations in the soil to replicate the loading
pattern generated by high speed trains such as upcoming HS2 in the UK. Attenuation of accelerations in soil
suggest that friction between soil particles and confining stress have a significant effect on the transmission
of surface waves.
Keywords: Rayleigh Waves; Centrifuge Modelling; Drop-Projectile System; High-Speed Trains.
1 INTRODUCTION railway track(s) (Fryba 1972; Kenney 1954; Krylov

High speed trains have emerged out as an attractive 1995; Payton 1967).
alternative for inter-city travel and commerce as
compared to air transport. High speed rail networks
not only tend to promote growth between their
major hub cities, but they can also promote growth,
act as a catalyst of socio-economic development for
population centres along their routes.
European Union directive defines a high speed
rail as lines which are purpose-built for train speeds
equal or greater than 250 km/hr or lines which are
upgraded to allow train speeds greater than
200km/hr. UK is now developing its new high speed Figure 1. Generation of ground-borne vibrations due to
rail network (HS2), with projected operational wheel-axle load on rails (Krylov 1996)
speed of HS2 trains is about 360 km/hr, connecting
the densely populated and commercial centre of Generally, the Rayleigh surface velocity of
London with the other major northern cities. subgrade is in the range of 100-200 m/s and it was
Ground-borne vibrations generated by high speed not possible for conventional trains to reach this
trains can not only cause discomfort to the onboard critical velocity. However, it was only after 1970s
passengers but the vibrations propagating within the that with the increasing operational train speeds,
ground can also affect railway catenary structures, multiple instances of high dynamic train-track
adjacent buildings and any sensitive mechanical amplification were observed in the field where the
equipment which might fall within the zone of track alignment passed over ground with low
disturbance. Rayleigh surface wave velocity (in the range of 40-
The main mechanism of generation of ground- 60 m/s) consisting of soft soils such as soft silty
borne vibrations is the wheel-axle pressure acting clays, peats and organics at shallow depths
on the rails (Figure 1). Dynamic amplification of (Adolfsson et al. 1999; Madshus and Kaynia 2000;
train-track vibrations has been a subject of Woldringh and New 1999). In these instances, high
theoretical research in the last century. For a non- speed rail operations could only be fully resumed
oscillating train load moving with a constant after undertaking ground remediation measures to
velocity on the surface of an elastic half-space, stiffen the subgrade and consequently increasing the
train-track vibrations tend to amplify as the train Rayleigh surface wave velocity.
speed approaches a certain critical velocity Analytical solutions are generally valid only for
(Rayleigh wave velocity) of the medium supporting simpler geometry and boundary conditions. A
177
number of researchers have used numerical methods load. The speed of moving load achieved in their
for propagation of surface waves and vibration centrifuge tests conducted at 50g was about 89 m/s,
mitigation problems including Buonsanti et al. which was approximately half of the surface wave
(2010); Correia et al. (2007); Holm et al. (2002); velocity reported by the authors. In this speed
Banimahd et al. (2007) among others. The regime, the train-track (foundation) dynamic
advantage of numerical models lie in their ability to amplification effects in the experiments could be
model complex geometries such as embankments, quite limited based on the data collected from case
cuttings, piled embankments etc. and wave histories (Woldringh and New 1999).
propagation in multi-layered ground. However, the
soil behaviour in most of these analyses have been
limited to linear elastic and nonlinear elastic
behaviour, which is not the case for soils, especially
soft soils. The accuracy of numerical solutions for
prediction of vibrations depend on multiple factors
including constitutive model for soils, element
discretisation, consideration of soil damping and
simplification for vehicle-body dynamics.
Early experimental research in the form of field
tests into propagation of vibrations from a stationary
mechanical source and the effectiveness of vibration
isolation measures were carried out by Barkan Figure 2. Schematic sketch of moving load simulator (Itoh et
al. 2002)
(1962) and Woods (1968) among others.
Full-scale field studies can be quite expensive and In another set of experiments, Itoh et al. (2005)
can also be limited in their flexibility to undertake developed a vibration generator to apply harmonic
parametric studies. Centrifuge modelling is an loads to circular bearing plates. In contrast to the
attractive alternative to develop small-scale multiple ball dropping system (Itoh et al. (2002), the
physical models which are tested under the vibration generator applied both horizontal and
enhanced field of centrifugal acceleration. vertical vibrations. Centrifuge tests performed at
Centrifuge models can replicate the correct stress- 50g were used to investigate the effect of epoxy
strain behaviour of soil at prototype scale in the
poly-styrol (EPS) barrier and crumb rubber-
model (Madabhushi 2017). However, geotechnical modified asphalt (CRMA) underneath the
centrifuge research into surface wave propagation foundation as vibration mitigation measures. The
has still remained a relatively less explored field as authors suggested that the performance of drop-ball
compared to earthquake engineering. system (Itoh et al. 2002) was better than the
Luong (1996) carried out comparative centrifuge-
vibration generator system in terms of impact
based experiments to study the effectiveness of a energy and mobilisation of strains within soil.
stress-wave mitigation barrier subjected to impulse Yang et al. (2013) undertook centrifuge and
loading generated by explosives and a novel drop- numerical modelling to investigate ground-borne
ball apparatus. Luong (1996) concluded that the vibrations from surface sources. The vertical
drop-ball apparatus is more convenient to set-up and
excitation was generated by an electromagnetic
the characteristics of the impulse such as frequency shaker which was driven by sinusoidal input
and amplitude, can be more easily customised than waveform of varying frequency content. It was
the impulse generated by blast loading. Semblat and concluded that to calibrate the response from
Luong (1998) carried out further experiments using numerical models at surface and at depth, it is
the drop-ball experiments developed for advisable to use a nonlinear stress-dependent soil
investigating surface wave propagation by Luong stiffness rather than assuming a constant average
(1996). It was noted by the authors that the value.
nonlinear effects near the source might affect wave In this paper we focus on the development of
propagation and limit the applicability of empirical modified version of drop-ball system and its
equations of wave attenuation. application for generating surface waves in
Itoh et al. (2002) further expanded on this idea by homogeneous and layered soil profiles. This paper
developing a multiple ball dropping system also discusses the measurement of shear wave
(MBDS) to simulate a moving high speed train load velocity (Vs) to characterise in-situ ground stiffness,
in centrifuge experiments. Figure 2 shows the
which is not often discussed in the context of
schematic sketch of drop-ball system with two
geotechnical centrifuge testing for surface wave
model foundations for modelling of moving point
propagation. A brief discussion of effect of soil
178
stiffness contrast from the preliminary analysis of
surface wave propagation in soil is also presented.
2 CENTRIFUGE MODELLING
With the help of scaling laws, centrifuge
experiments can be used to model correct stress-
strain behaviour at a certain depth in 1g prototypes.
The centrifuge tests discussed in this paper were all
performed at 40g. For centrifugal acceleration of
‘Ng’, the scaling laws reproduced from
Madabhushi (2017) are presented in Table 1. Two
models will be discussed in this paper – Model 1 Figure 3. Miniature Air Hammer (Ghosh 2003)
(homogeneous dry sand) and Model 3 (dense sand
layer over soft clay). 2.2. Instruments and Data Acquisition
Centrifuge tests reported in Section 1 were limited
to homogeneous, dry sand medium. As reported in Piezoelectric-accelerometers of type A/23 were
the case studies (Section 1), high dynamic train- placed at various depths in the soil during model
track amplification scenarios generally occur when preparation to measure the stress-pulse generated by
soft, compressible soil is present underneath the Air Hammer. These accelerometers were
track. One of the current practices for reducing waterproofed with wax before placing them in clay.
dynamic amplification of train-track vibrations is to MEMS-type accelerometers of type ADXL78 with
improve the subgrade stiffness in order to increase ±70g were used to measure vertical acceleration at
the critical surface velocity. To investigate surface the source of vibration and within the soil. Previous
wave propagation for situations where a stiff layer centrifuge experiments performed by Stringer et al.
overlies soft ground, Model 3 was prepared with a (2010) showed that MEMS-type accelerometers can
47 mm thick (prototype thickness of about 1.9 m) have a better accelerometer-soil coupling than the
layer of dense sand overlying soft clay. piezoelectric-type accelerometers.
Table 1. Centrifuge scaling laws

Parameter Dimensions Model/Prototype
Length L 1/N
Velocity LT-1 1
Acceleration LT-2 N
Strain - 1
Stress ML-1T-2 1
Mass M 1/N3
Time (dynamic) T 1/N Figure 4. ADXL78 type MEMS accelerometer glued to HDPE
plastic
Frequency T-1 N
These MEMS accelerometers have typical
2.1. Equipment dimensions of 5 mm by 5mm by 2mm. These
The main equipment used was the 10 m beam accelerometers were oriented in vertical direction
centrifuge with a radius of 4.125 m from the centre by gluing them to a hand-cut piece of HDPE plastic
of rotation to the top of swinging platform. The as shown in Figure 4. Accelerometer-soil coupling
model container used in the centrifuge testing is a was improved by gluing dry sand to the
circular tub with an internal diameter of 850 mm accelerometers and HDPE plastic. Electrical signals
and a height of 400 mm. In-situ value of small-strain from the instruments were transmitted via junction
stiffness of soil was evaluated by measuring the boxes and through on-board computers on the beam
time-lag between arrival of peak shear wave computer and digitised by DASYLab data
amplitude between known locations of any two acquisition software.
piezoelectric-accelerometers. The shear waves were
generated with the help of Air Hammer (Figure 3), 2.3 Model Preparation
originally developed by Ghosh and Madabhushi In the preparation of sand bed for Model 1 and
(2002), placed near the bottom of the model Model 3, Hostun HN31 silica sand was used. The
container. minimum (emin) and maximum (emax) void ratio of
179
Hostun sand are 0.555 and 1.01 (Mitrani 2006). The yielded a poor signal-to-noise ratio and therefore,
sand was poured using an automatic sand pourer the possibility of using APA400MML piezoelectric
designed at the Schofield Centre (Madabhushi et al. actuator was abandoned and an alternative was
2006) to achieve relative density of about 50% and developed as discussed in Section 2.4.2.
a total thickness of 232 mm for Model 1. For Model
3, the lower clay layer was prepared by 2.4.2 Drop-Projectile System
consolidating slurry made out of speswhite Kaolin The drop-ball system of previous researchers
powder mixed with water to 120% of its liquid limit. discussed in Section 1 were limited to a single drop
The slurry was consolidated to a pressure of 200 kPa height. To investigate the relationship between
under a stress-controlled consolidation rig at 1g in impact load (or, drop height) and surface wave
an 850 mm diameter circular tub. After reaching the propagation, the first iteration (Figure 6) of drop-
required stress state, the tub was unloaded from the ball system was developed at the Schofield Centre.
rig and the clay surface was hand-trimmed to It could perform multiple drop-tests during a single
achieve a thickness of 157 mm. Hostun HN31 sand flight of centrifuge package. An electromagnet of
was poured over this clay layer using automatic type 58-50 with 24 V DC rating supplied by SG
sand pourer to a total thickness of 47 mm and Transmission was used to hold and pick-up a metal
relative density of about 67%. spherical ball of mass 34 g. The drop-height could
be adjusted with the movement of a 1D linear
2.4 Vibration Generation Source actuator manufactured by SKF.
2.4.1 Piezoelectric Actuator
At first, the feasibility of a piezoelectric actuator
was explored as source of generating vertical
vibrations. An APA400MML amplified
piezoelectric actuator (APA) was connected to an
aluminium disk 30 mm in diameter and 5 mm in
thickness. The actuator was controlled and driven
by LA75B and LC75B supplied by Cedrat
Technologies. The actuator was excited with
voltage-controlled sine waves generated by a
TG330 function generator with excitation
frequencies in the range of 20-150 Hz.
For 1g trials, Hostun HN31 sand was poured in an
850 mm tub; piezoelectric accelerometers were
Figure 6. Schematic representation of first iteration of drop-
placed about 15 mm below the sand surface. The
ball system (for Model 1)
piezoelectric actuator was bolted onto a gantry
spanning the width of the tub to provide a rigid A ‘ball drop’ was performed by simply turning
support. The test assembly in shown in Figure 5. OFF the current to electromagnet. The impact
occurred on an aluminium disk 60 mm in diameter
and with an average thickness of approximately 10
mm. The disk was placed on the surface of model
ground but with a lateral offset accounting for
Coriolis effect on the falling spherical ball under the
field of centrifugal acceleration ‘Ng’. The values of
lateral offset (dlat) due to Coriolis force were
calculated for drop-heights (Hcor) of 75 mm, 100
mm and 125 mm using Equation 1 (Madabhushi
2017).
2 1.5
2√2 𝐻𝐻𝑐𝑐𝑐𝑐𝑐𝑐
𝑑𝑑𝑙𝑙𝑙𝑙𝑙𝑙 = (1)
3 4.1250.5
Figure 5. Test package with APA400MML actuator During the centrifuge flight of Model 1, it was
found out that due to the Coriolis effect, the
A test up similar to Model 3 soil profile was run in spherical ball would fall outside the effective
the centrifuge with piezoelectric actuator which diameter of the electromagnet for it to be picked up.
180
Therefore, for the testing of Model 3, a second
iteration (Figure 7) of drop-ball system was
developed to overcome these issues. To allow the
movement of electromagnet within the Perspex
tube, a smaller electromagnet (type 58-30) was
used. A small gap was left between the end of
Perspex tube and upper face of aluminium disk to
vent out the air pressure (drag) which would
develop in front of the projectile during its fall.
Figure 9. Assembled centrifuge package (Model 3) loaded

onto the swing platform of 10 m beam centrifuge
impact tests were done using drop-ball (Model 1)/

projectile (Model 3) system. The three different
drop heights were kept same between Model 1 and
Model 3 for consistency. The resulting vertical
accelerations were measured at the source (disk)
and in soil. All signal processing operations were
performed in MATLAB R2019a.
Figure 10 and Figure 11 show the sketch of
instrumentation layout for Model 1 and Model 3.
Figure 7. Schematic representation of second iteration of

drop-ball system (for Model 3)
In this iteration, a Perspex guide tube was

provided to channel a custom-made projectile
(Figure 8). This projectile has a mass of about 34 g
and a maximum diameter of 30 mm. Centrifuge
experiments showed that this second iteration
performed was significantly more reliable than the
first iteration.
Figure 10. Sketch of instrumentation for Model 1
Figure 8. Custom-made projectile for second iteration of

drop-ball system (Drop-Projectile System) for Model 3 Figure 11. Sketch of instrumentation for Model 3
2.5 Experimental Program 3 CENTRIFUGE RESULTS

After loading the centrifuge package onto the beam 3.1. Soil Stiffness (Vs)
centrifuge, it was spun upto 96 rpm to get to 40g. Time-traces of vertically upwards propagating shear
Figure 9 shows the assembled package of Model 3 waves were obtained from horizontally oriented
with second iteration of drop-projectile system in piezoelectric-accelerometers buried at different
the middle. depths in model soil layer(s). The time delays in the
Air Hammer test was done until three clean pulses arrival of peak pulse amplitude were used to
were recorded for piezoelectric-accelerometers calculate in-situ shear wave velocity profile (Vs).
buried throughout the model depth. Afterwards
181
Figure 12. Acceleration time histories at prototype scale
(Model 1)
Figure 14. Vs profile at prototype scale – Model 3
Figure 12 shows the acceleration time histories at
prototype scale as measured by the piezoelectric- 3.1. Wave propagation in ground
accelerometers for a representative case (Model 1). Vertical accelerations generated due to the impact
Although the amplitude decays as the distance from of falling ball (Model 1)/ projectile (Model 3) were
the Air Hammer increases, but the waveform even measured at the source and at different locations in
near the surface was clean enough to obtain a soil. Figure 15 shows a representative example of
reasonably accurate measurement of time-lag in the vertical acceleration time histories (prototype
arrival of first peak. Empirical expressions (Hardin scale) at lateral distance away from the source
and Drnevich 1972; Oztoprak and Bolton 2012) of (shallow depth) for Model 3 generated due to the
small-strain shear modulus (Go) and known values projectile falling through a drop height of 75 mm.
of mass density (for sand) were used to derive Vs
profile as a function of depth; there is a lack of
similar relationships for clays. The comparisons for
Model 1 and Model 3 are presented in Figure 13 and
Figure 14, respectively. It was observed that the
experimental Vs profile for sands was slightly softer
than the one obtained from empirical relationships.
A significant Vs (soil stiffness) contrast was
observed between the upper dense sand layer and
lower soft clay layer for Model 3.
Figure 15. Vertical accelerations measured at shallow depth

in soil (prototype scale) for Model 3
Attenuation of surface waves could be inferred

from the decay of peak amplitude of first arrival of
vertical accelerations in soil. These peak amplitudes
were averaged for vertical accelerations obtained
from all the three drop heights. Figure 16 shows the
comparison of these averaged accelerations for
Figure 13. Vs profile at prototype scale -Model 1
182
MEMS buried at shallow depth for Model 1 and 4 CONCLUSIONS
Model 3 against normalised distance (d/R), where d This paper describes an experimental research into
is the magnitude of distance vector of MEMS from the generation and attenuation of surface waves in
source of vibration and R is the inner radius of homogeneous and layered soil profiles. An
model container. The surface wave attenuation electromagnetic drop-projectile was developed as
results tend to suggest that for the case of stiff soil an enhancement over the original drop-ball system
overlying soft soil (Model 3), the stiffer layer could developed by Luong (1996) to increase reliability
experience higher accelerations near the source and and repeatability of experiments. Following
at shallow depths as compared to model without a conclusions were drawn from the centrifuge tests:
significant soil stiffness contrast (Model 1). 1. Air Hammer test results provided in-situ
measurement of shear wave velocity (Vs) to allow
for more accurate soil characterisation for wave
propagation than using empirical relationships
based on index properties.
2. The results suggest that regardless of the soil
stiffness contrast, the impulse generated at the
source has sufficiently attenuated due to geometric
attenuation and frictional work between sand
particles (material damping) before reaching the
model container boundary.
3. Soil stiffness contrast (stiff sand over soft clay)
tend to create a waveguide effect within the upper
layer which tend to amplify accelerations near the
source at shallow depth and also for a longer lateral
Figure 16. Comparison of averaged vertical accelerations
(prototype scale) at shallow depth
distance away from the source.
For each depth location, normalised accelerations 5 ACKNOWLEDGEMENTS

were derived wherein for a given depth, the The authors would like to acknowledge peers,
averaged maximum acceleration at any location was technical staff at the Schofield Centre for their
normalised with respect to the first MEMS at that incredible support, meaningful discussions and
depth. Figure 17 shows comparison of normalised sharing their experiences in overcoming the
average accelerations for MEMS buried at shallow challenges. The first author would also like to thank
and deep depth plotted as a function of normalised Dr. Barnali Ghosh, Tony O’Brien, Imran Farooq for
distance d/R for Model 3. The results suggest that their guidance and generous grant from Mott
for approx. similar value of normalised distances, MacDonald’s Research Fund in supporting the
the MEMS at deeper depth have higher normalised research at Cambridge University.
amplification compared to MEMS at shallow depth,
which might suggest a waveguide effect occurring 6 REFERENCES
within stiff sand layer due to soil stiffness contrast. Adolfsson, K., Andreasson, B., Bengtsson, P. E., &
Zackrisson, P. 1999. High speed train X2000 on soft
organic clay - measurements in Sweden. Geotechnical
Engineering for Transportation Infrastructure.
Proceedings of the 12th European Conference on Soil
Mechanics and Geotechincal Engineering,
Amsterdam, June 1999. Vol. 3.
Banimahd, M., & Woodward, P. K. 2007. Numerical

Study Of Train Speed Effect On Railway Track
Response. Railway Infrastructure: Railtex -
International Railway Engineering Conference, 4(1).
Retrieved from http://www.xitrack.com/paper6.pdf
Barkan, D. D. 1962. Dynamics of Bases and

Foundations. McGraw-Hill Book Co., Inc., New York.
Figure 17. Comparison of normalised vertical accelerations
at different depths for Model 3 Buonsanti, M., Leonardi, G., & Scopelliti, F. 2010.
Analyses of ground-borne vibrations induced by high
183
speed trains. 10th International Workshop on Railway Krylov, V. V. 1996. Vibrational impact of high‐speed
Noise. Nagahama. trains. I. Effect of track dynamics. The Journal of the
Acoustical Society of America.
Fryba, L. 1972. Vibration of Solids and Structures under https://doi.org/10.1121/1.417123
Moving Loads. In Vibration of Solids and Structures
under Moving Loads (pp. 83–93). Madabhushi, G. 2017. Centrifuge modelling for civil
https://doi.org/10.1680/vosasuml.35393.0007 engineers. Centrifuge Modelling for Civil Engineers.
https://doi.org/10.1201/9781315272863
Gomes Correia, A., Cunha, J., Marcelino, J., Caldeira, L.,
Varandas, J., Dimitrovová, Z., Gonçalves Da Silva, M. Madabhushi, S. P. G., Houghton, N. E., & Haigh, S. K.
2007. Dynamic analysis of rail track for high speed 2006. A new automatic sand pourer for model
trains. 2D approach. Applications of Computational preparation at University of Cambridge. Physical
Mechanics in Geotechnical Engineering - Proceedings Modelling in Geotechnics, 6th ICPMG’06 -
of the 5th International Workshop on Applications of Proceedings of the 6th International Conference on
Computational Mechanics in Geotechnical Physical Modelling in Geotechnics, 217–222. CRC
Engineering. Press.
Ghosh, B., & Madabhushi, S. P. G. 2002. An efficient Madshus, C., & Kaynia, A. M. 2000. High-speed railway
tool for measuring shear wave velocity in the lines on soft ground: dynamic behaviour at critical train
centrifuge. International Conference on Physical speed. Journal of Sound and Vibration.
Modelling in Geotechnics, 119–124.Luong, P. M. https://doi.org/10.1006/jsvi.1999.2647
(1996). Centrifuge modelling of a seismic barrier using
soil energy dissipation. Eleventh World Conference on Mitrani, H. 2006. Liquefaction remediation techniques
Earthquake Engineering. for existing buildings. PhD thesis, University of
https://doi.org/10.13140/2.1.2613.1202 Cambridge.
Ghosh, B. 2003. Behaviour of rigid foundation in layered Oztoprak, S., & Bolton, M. D. 2012. Stiffness of sands
soils during seismic liquefaction. PhD thesis, through a laboratory test database. Géotechnique.
University of Cambridge. https://doi.org/10.1680/geot.10.p.078
Hardin B.O., & Drnevich V.P. 1972. Shear modulus and Payton, R. G. 1967. Transient motion of an elastic half-
damping in soils: design equations and curves. Soil space due to a moving surface line load. International
Mechanics and Foundations Division, 98(7), 667–692. Journal of Engineering Science, 5(1), 49–79.
Holm, G., Andreasson, B., Bengtsson, P., Bodare, A., & Semblat, J. F., & Luong, M. P. 1998. Wave propagation
Eriksson, H. 2002. Mitigation of Track and Ground through soils in centrifuge testing. Journal of
Vibrations by High Speed Trains at Ledsgard, Sweden. Earthquake Engineering, 2(1), 147–171.
Report 10. 1–60. https://doi.org/10.1080/13632469809350317
Itoh, K., Koda, M., Lee, K. I., Murata, O., & Kusakabe, Stringer, M., Heron, C., & Madabhushi, S. 2012.
O. 2002. Centrifugal simulation of wave propagation Experience using MEMS-based accelerometers in
using a multiple ball dropping system. International dynamic testing. Physical Modelling in Geotechnics,
Journal of Physical Modelling in Geotechnics, 2(2), 7th ICPMG’ 10 - Proceedings of the 7th International
33–51. https://doi.org/10.1680/ijpmg.2002.020203 Conference on Physical Modelling in Geotechnics (pp.
389–394). https://doi.org/10.1201/b10554-61
Itoh, K., Zeng, X., Koda, M., Murata, O., & Kusakabe,
O. 2005. Centrifuge simulation of wave propagation Woldringh, R. F., & New, B. M. 1999. Embankment
due to vertical vibration on shallow foundations and design for high speed trains on soft soils. Geotechnical
vibration attenuation countermeasures. JVC/Journal of Engineering for Transportation Infrastructure.
Vibration and Control. Proceedings of the 12th European Conference on Soil
https://doi.org/10.1177/1077546305054150 Mechanics and Geotechincal Engineering,
Amsterdam, June 1999. Vol. 3.
Kenney, J. T. 1954. Steady-state vibrations of beam on
elastic foundation for moving load. Journal of Applied Woods, R. D. 1968. Screening of surface waves in soils.
Mechanics-Transactions of the ASME, 359-364. Journal of the Soil Mechanics and Foundations
Division, 94(4), 951–980.
Krylov, V. V. 1995. Generation of ground vibrations by
superfast trains. Applied Acoustics, 44(2), 149–164. Yang, W., Hussein, M. F. M., Marshall, A. M., & Cox,
https://doi.org/10.1016/0003-682X(95)91370-I C. 2013. Centrifuge and numerical modelling of
ground-borne vibration from surface sources. Soil
Dynamics and Earthquake Engineering.
https://doi.org/10.1016/j.soildyn.2012.09.003
184
Centrifuge modelling of train passage over clay subgrades

Scott Robinson1, Andrew Brennan1, Cameron Munro1, Peter Woodward2, David Connolly2,
Omar Lagrouche3, Ahmet Esen3
1
Division of Civil Engineering, University of Dundee, Scotland, UK
2
School of Civil Engineering, University of Leeds, England, UK
3
School of the Built Environment, Heriot-Watt University, Scotland, UK
Corresponding author: Scott Robinson (s.z.robinson@dundee.ac.uk)
ABSTRACT: The modelling of large geotechnical structures on the centrifuge is now well established, but
there remain technical challenges with each new application. This work arose from a larger project
investigating alternative improved embankment designs. The project required full scale modelling of new
embankment designs using the GRAFT II simulator at Heriot Watt University together with consideration
of the large scale settlement problem over a wider area. To model foundation soil beneath the embankment
required the design of a rail simulator at centrifuge scale, together with appropriate measurement techniques
and appropriate modelling of useful foundation soil types. This paper therefore describes the design
decisions taken to produce a centrifuge mounted rail simulator including details of modelling a translating
cycling load. To measure total stress at the base of the embankment during a train passage was a key
challenge faced, overcome by use of thin Singletact stress transducers. These transducers are low deflection
capacitance-based force sensors that will be further discussed in the paper. In addition to the translating load
and the instrumentation, the third aspect presented will be the formation and characterisation of a soft clay
subgrade, required due to the nature of soils that are most of concern to UK rail industry. To accelerate
consolidation without compromising the ability to achieve a clay-type mechanical response, a hybrid clay-
sand-silt mixture known as KSS 541 is installed in the model container as slurry and consolidated in-flight.
Experience with this material is discussed.
Keywords: Railways; Embankments; Centrifuge Modelling; Clays; Stress Measurements.
1 INTRODUCTION embankments will be discussed in forthcoming

This research was carried out as part of the UK papers, but this involved precisely scaling model
EPSRC funded LOCORPS project (Lowering the geogrid, facing walls, fill materials, model railway
Cost Of Railways using Preformed Systems), which ballast and model track stiffnesses to ensure the
aims to investigate and assess more efficient miniature 1/50th scale embankments correctly
alternative embankment solutions for high speed replicate the behaviour of the full-size problems.
rail that can be more rapidly constructed compared
to conventional designs. The project consists of
multiple approaches, including testing of full scale
embankment sections using the GRAFT II rail test
facility at Heriot-Watt University in Edinburgh,
focusing on the behaviour of the embankment itself
under train loading, as well as centrifuge modelling
of the proposed embankment designs at the
University of Dundee investigating the
embankment performance on different subsoils and
the interaction between them. Modelling the subsoil
is difficult to achieve using full-scale testing due to
the space and volume required, and is an area where
centrifuge modelling is advantageous. The research
investigated GRS-RW (Geosynthetic Reinforced
Soil Retaining Wall) embankments (2.5 m high at
prototype, see Figure 1) of the type widely used in
Japan, conventional ballasted embankments (1.2 m
Figure 1. A section of 1/50th scale (50mm high) geosynthetic
high) and a new preformed geogrid embankment reinforced soil retaining wall (GRS-RW) embankment under
system proposed by the project (1.2 m high). Details construction in the centrifuge rail simulator, on a soft clay
of the scaling and miniaturisation of these subsoil.
185
The centrifuge modelling aimed to investigate the In centrifuge modelling of high-speed rail, it is
embankment and subsoil settlements on both sand necessary to ensure that the conditions in the model
and soft clay, as well as how the subsoil movements are in keeping with the full size problem. Two of
affect the overlying embankments, and measuring these considerations which arise in respect of high-
how the train loads transfer through the track and speed rail are the dynamic effects from train
embankment creating the resulting stress changes vibrations, and also the scaling of the train velocity
on the top of the subsoil. This is an area that the UK to match the real-world drainage conditions.
Rail Safety and Standards Board (RSSB) states is an First dealing with the dynamic effects, the
area of poor understanding, on which there is little vibrations generated by a train whilst in motion
to no ongoing research (RSSB 2011). The focus of could potentially increase the rail settlements
this paper will be the development of methods for beyond those caused only by the stress change in the
simulating model train loading in-flight in the embankment and subsoil due to the weight of the
centrifuge, instrumenting the model embankments train. This would occur if the shear strains induced
(including total stress measurements at the subsoil by the vibrations were greater than the elastic shear
interface) and the preparation of the subsoil models strain threshold of the subsoil, meaning that plastic
used in the testing. deformation from the vibrations could accumulate.
Connolly et al. (2014) conducted field monitoring
2 TRAIN MODELS of a high-speed rail line to measure these vibration-
Before setting out the centrifuge modelling, it is induced shear strains in the surrounding soil. It was
important to define the train characteristics being found that the shear strains varied depending on the
studied. The research in this paper is based on a distance from the track, but that the maximum strain
TGV-R train (Connolly et al. 2014), which consists recorded for the case where an embankment is
of power cars weighing 68 tonnes supported by two present was γ = 5.9 x 10-6. For clay soils, the elastic
bogies each with two axles, resulting in an axle load shear strain threshold is of the order of γ = 1 x 10-4
of 17 tonnes (34 tonnes per bogey). The passenger (for Kaolin, the main component of KSS 541),
cars are significantly lighter, but instead share nearly two orders of magnitude higher than the
bogies at the carriage connections, resulting in a maximum measured shear strain from vibrations.
similar axle load of 16 tonnes (31 tonnes per bogey). This suggests that deformations generated by the
Bogies are spaced at 14 m centre to centre on the vibrations could be considered to be elastic (and
power cars and 18.7 m on the passenger cars. The recoverable) and would be unlikely to significantly
train model used in the centrifuge was based on the affect the model embankment settlements measured
TGV-R power car, but due to the similarity in the during the cyclic train loadings investigated in this
axle loads between the power and passenger cars, research, and that modelling of these dynamic
the axle passes modelled in the centrifuge could be effects is unnecessary. Hence the train was
considered to be representative of either. This modelled as a smooth passing load.
required a 1/50th scale train model (Figure 2) Secondly, the train velocity in the field determines
weighing 550 g, with model bogies (Igus the drainage conditions that occur. An indication of
WJ200QM-01 bearings) spaced at 280 mm centre to these drainage conditions can be obtained from the
centre. train’s normalised velocity (Eq. 1) where v is the
train velocity, d is the representative dimension
(assumed to be the track width of 2.5 m in this case)
and cv is the coefficient of consolidation (cv =17.2
m2/year for KSS 541). For a full-scale train moving
at 300 km/h (Connolly et al. 2014) this would
correspond to V = 3.81 x 108. A value of V greater
than 10 to 30 indicates fully undrained conditions
(Randolph and Gourvenec 2011) with no pore
pressure dissipation during the passage of the train,
hence, the normalised velocity of the model train
must be greater than 30 to correctly simulate the
undrained nature of the problem. The maximum
actuation velocity possible with the system used
was 108 m/h and considering the centrifuge scaling
Figure 2. 1/50th scale model train used to simulate a 68 tonne of the problem this means that the normalised
TGV-R power car (550 grams, 280 mm bogey spacing). velocity of the model train was V = 2750,
186
significantly greater than that required to correctly embankment ahead and behind each bogey. The
simulate undrained conditions. comparatively high GRS-RW embankments would
also have a similar effect. A 2D plane strain
vd approach such as the type used in the preliminary
V= (1) tests would fail to capture this behaviour and mean
cv
that the stresses which reach the subsoil beneath the
embankment (which are of interest in this research)
3 MODELLING APPROACH would be unrealistic. To overcome these two issues,
Consideration was given to the loading approach it was decided to adopt a novel approach and model
adopted. Two options were possible; to model the the train loading and embankments using a 3D
embankment in the centrifuge as a 2D problem with approach with a 1/50th scale model train moving
a plane strain approach, or to simulate the train along a 1 m length of model track whilst in flight at
using a 3D approach with a moving load. 50 g in the University of Dundee’s 3 m radius beam
Preliminary 1 g testing was undertaken to centrifuge. To the knowledge of the authors, this is
investigate the performance of a 400 mm length of the first time that moving train loads on
50 mm high GRS-RW embankment placed in a embankments have been tested in a centrifuge, with
model container of dimensions 400 mm square and all previous studies using 2D plane strain methods
filled with 500 mm of loose HST95 sand. Uniform (Zhang et al. 2018; Vorster et al. 2017;
cyclic loading was applied to the embankment via a Viswanadham et al. 2012).
50 mm wide, 400 mm long force applicator
representing the width of track. The force was
applied using an Instron 5985, equivalent to a
uniform stress of 1 kPa across the track area. The
stress was estimated based on the 34 kN bogey load
being evenly distributed over the area underneath
the bogey (2.5 x 3 m) giving a stress of 47 kPa,
which was scaled down by N = 50 to account for the
1 g stress conditions in the subsoil. The
embankment was subject to more than 800 cycles of
loading, resulting in a final settlement of 0.22 mm
(11 mm at full scale) (Figure 3).
Figure 4. Model train and track on a model conventional
0.25 ballasted embankment in the centrifuge actuation system.
Maximum Settlement at Model
0.2 This introduced further scaling requirements,

such as the need to correctly scale the longitudinal
Scale (mm)
0.15 bending stiffness of the track to ensure the

distribution of the train load in the direction of the
0.1 track is correctly modelled. Hence, the bending
stiffness, EI, was scaled down by N4, including
0.05
accounting for the differing material moduli of the
steel and concrete used at full scale against the
0
0 200 400 600 800 1000 aluminium used at model scale. To achieve this, two
Number of Cycles 1.08 m long model track lengths were created, each
Figure 3. Maximum settlement in each cycle for a 1 g test of incorporating integral low friction Igus WSQ-06-30
1/50th scale GRS-RW embankment on loose HST95 sand. double bearing rails for the model train to move
along (Figure 4). One length of track was scaled to
However, the preliminary testing raised a number match the bending stiffness of UIC60 rails typically
of issues. Firstly, the stresses (and hence also used in combination with conventional G44
stiffnesses) in the subsoil were N times too low due sleepers on ballasted embankments (with model
to the 1 g nature of the tests, and would not be sleepers also attached to match the interface with the
consistent with the stresses in the full scale problem. ballast), whilst a stiffer model track was created
Secondly, high speed rail lines increasingly adopt with a bending stiffness designed to match the
high stiffness concrete slab tracks on top of the scaled stiffness of the concrete slab track (made up
embankments, and these slab tracks act to distribute of hydraulically bonded layer (HBL), slab track and
the load from the train not just laterally across the rails) used in the full scale testing (Figure 5). The
embankment, but also longitudinally along the
187
longitudinal bending stiffness of the GRS-RW track an instrumented section in the middle of the track
facing walls was also scaled for the same reason. length by 150 mm and ensure the stress from the
train at this point was fully relieved before the
actuator reversed direction and the train passed in
the opposite direction. Each cycle of the actuator
resulted in the train passing this point twice,
delivering four bogey passes at a speed of 30 mm/s.
The cyclic testing continued for 6 hours, allowing
up to 2000 bogey passes to be simulated in each
centrifuge test.
Figure 5. Comparison between the conventional UIC60 rail

and sleeper model (l) and the rail and slab track model (r).
4 METHODOLOGY
4.1. Actuation system
Figure 6. Rail simulator and 1.5 m long strong box installed on
To accommodate the track lengths required, the the University of Dundee’s 3 m radius beam centrifuge.
testing was carried out in a centrifuge strongbox
with internal dimension 1500 mm long, 400 mm The actuation system itself consisted of a moving
wide and 650 mm deep (Figure 6). A viewing platform above the model running on Igus drylin
chamber with a Perspex face was placed at one end WSQ-16 linear rails at each side of the box driven
of the box to allow a cross-section of the track and via a belt system by a Parvalux SD12-LWS high
embankment to be viewed in-flight, which reduced torque 220V DC motor (Figure 7). Displacement of
the length of track and embankment to 1080 mm. the platform was measured by a Multicomp SP1-50
The 1/50th scale train model weighing 550 g ran on draw wire transducer. The system is described in
the track, connecting via two 3D joints to a moving detail in Robinson et al. (2019), but for this testing
platform on an actuator attached to the top of the it was upgraded to allow programmable movement
box such that the train was free to move vertically. such that the cyclic testing could be automated via
Using this system, it was possible for the model Labview. This was achieved by interfacing the
train to have a travel distance of up to 600 mm motor’s reversing DC controller with an
which was sufficient for the train to pass and clear inexpensive set point isolator, allowing the
Figure 7. Schematic cross section of centrifuge rail simulator showing actuator, strong box, model and instrumentation.
188
motor’s direction and speed to be easily controlled thin film pressure sensors which using printed
by a 0 to 5 V signal and an existing relay unit within conductive ink, which require an amount of
the centrifuge’s National Instruments CompactRIO compression (‘switch-on distance’) before pressure
9047 chassis. Control signals for the motor were changes begin to be detected, which is unacceptable
generated by an NI 9264 voltage output unit within if correct in-situ stresses are to be measured.
the chassis, which already provides individual Capacitive sensors do not suffer from these issues.
software programmable supply voltages for the 32 If a comparatively thick sensor is stiffer (or less
instrumentation channels on the centrifuge which stiff) than the surrounding soil, then stress arching
pass through a custom built low-noise power around the sensor can occur. The Singletact sensors
conditioning unit. This allows the low current avoid these issues due to their thin profile which
voltage signals from the NI 9264 to be used to minimises the volume of the inclusion used and also
independently power instruments with a higher due to the fact that they have a stiffness which is
current demand with a highly regulated voltage similar to the stiffness of the surrounding materials
supply. By using the same CompactRIO system for used in the embankments and subsoil. The sensors
both the instrument logging and the motor control have a Young’s modulus of 7 MPa (based on
signals, it was possible to implement a feedback manufacturer’s data), which is comparable in
control loop with the draw wire transducer in magnitude to the stiffness of the model embankment
Labview to carry out the cyclic testing using ballast (Eo′ = 4 to 8 MPa) and the KSS 541 subsoil
software control. (Eo′ = 2 to 3 MPa) as well as HST95 sand (Eo′ = 5 to
12 MPa, used in tests not described in this paper).
4.2. Instrumentation An array of five of these sensors (See Figure 8)
A range of instrumentation was used to during the were placed in the middle of the length of track
testing. Three miniature Honeywell displacement ranging from the centre of the embankment to the
sensors were used to monitor the track, embankment edge, to allow a 3D model of the stress field around
and subsoil settlements at the middle of the track a bogey to be measured at the subsoil as the bogey
length, capable of measuring settlements less than 1 approaches and passes the instrumentation point.
micron. A GoPro Hero 4 in the viewing chamber The sensor electronics were coated with epoxy resin
was also used for PIV analysis of a cross section of whilst the exposed edges of the sensing area were
the subsoil underneath the embankment at one end sealed with a fine layer of flexible adhesive.
of the track through a Perspex window which was
fitted with markers, allowing the displacement field
under the embankments to be investigated.
However, the most challenging measurement was
the total stress at the interface between the
embankment and the subsoil. This is a key
measurement in the tests as it could indicate how
effective the various embankment designs were in
distributing the load of the train and how the stress
is transferred down through the embankment to the
subsoil. Total stress measurements can be
problematic to achieve, as total stress sensors can in
themselves change the stress distribution in their
vicinity as well as being difficult to miniaturise for
use in such small models.
The total stress measurements were achieved
using Singletact S8-10N capacitive force sensors, 8
mm in diameter and 350 microns thick with a Figure 8. Array of Singletact S8-10N pressure sensors used for
sensing range of up to 200 kPa and a sensing total stress measurement.
resolution of 0.4 kPa. These operate using a stiff
elastomer between two foil layers covered with a Figure 9 shows typical results (from the 5th and 6th
thin protective polyamide outer layer, with small train passes) from a total stress sensor located in the
changes in the thickness of the elastomer due to the centre of the embankment at the ballast-subsoil
applied stress causing measurable changes in the interface. The change in stress generated by the
capacitance. The sensors have Singletact SE units passage of the individual bogies is clearly visible as
which convert capacitance into a straightforward four peaks in the stress measurements. The
analogue voltage output. These differ from other maximum change in vertical stress due to the bogey
189
load ranged from 13 to 22 kPa, with an average of silica silt) for the clay material was that it still
18 kPa. These maximum stresses were relatively provides a cohesive soil, but with less reduction in
constant, and showed no indication of changing height during consolidation, as well as shorter
with increasing cycle numbers. consolidation times than pure Kaolin. The
Expected real-world vertical stress changes from properties of KSS 541 are given in Table 1. If
a passenger train on the subsoil immediately Kaolin had been used, it would not have been
beneath ballasted embankment range from 20 to 30 possible to have a sufficiently thick layer of clay to
kPa depending on the ballast thickness (RSSB avoid boundary effects. The KSS 541 slurry was
2011). These figures are based on ballast mixed at a moisture content of 50% and placed (400
thicknesses (including sub-ballast) ranging from mm deep) into the strong box on top of a 30 mm
350 to 1000 mm. The ballast thickness used in this thick gravel drainage layer covered with a filter
case was slightly thicker than this range at 1250 mm membrane and filter paper. The moisture content of
at prototype scale (to match other embankment the slurry was critical as it was necessary to mix it
weights and stresses studied in the project), which at the lowest possible moisture content which
would distribute the train load over a greater area of avoided air entrapment occurring, in order to
the subsoil. Hence, the stress change for the minimise the height loss during consolidation.
embankment studied would be expected to be just Higher moisture contents would also have risked
below the lower end of the recommended range (i.e. segregation of the sand fraction in the KSS 541
just below 20 kPa), meaning that the average during spin up. The clay was consolidated at 50 g
measured value of 18 kPa would be in keeping with for 3 days, with monitoring of the mass of water lost
this. indicating this was sufficient to achieve near full
The measurements also clearly indicate that the consolidation. The drainage outlet location was set
subsoil experiences the stress change from the at 10 mm below the expected final surface level,
model train bogies approximately 220 mm (11 m at such that the water table was held at this location,
prototype) ahead of the arrival of the bogey itself. resulting in a final clay thickness of 300 mm.
This suggests that the effect of the embankment
distributing the load of the train along the length of Table 1. Properties of KSS 541.
the track is an important effect, validating the Parameter Unit Value
decision to adopt a 3D approach, modelling the train Gradient of CSL, M - 0.73
as a moving load. Critical friction angle, φ′crit ° 19
Clay fraction % 41
Plastic limit % 18
Liquid limit % 38
Plasticity index % 20
Coefficient of consolidation, cv m2/year 17.2
The GRS-RW embankment sections were

constructed at 1g in the lab and placed onto the
model. However, prior to this it was necessary to
use an aluminium bar of the same area and mass as
the embankment to pre-consolidate the
embankment area to prevent excessive deformation
during consolidation of the embankment in-situ.
Figure 9. Typical output from a Singletact pressure sensor The aluminium preload was consolidated at 50 g for
underneath the track (conventional ballast, 25 mm thick) for 24 hours before the actual embankment sections
one cycle of train loading on KSS541 were installed and consolidated for a further 24
hours. Oedometer testing of KSS 541 indicated that
4.3. Subsoil preparation 90% consolidation would require 12 hours, hence
The soft clay subsoil required to be consolidated in- the 24 hour periods were selected to ensure
flight from slurry. This was necessary to achieve a consolidation settlements would be negligible
realistic shear strength variation with depth, as compared to the expected embankment settlements.
opposed to the uniform profile that would occur if a During testing, the exposed surfaces of the KSS 541
press was used to prepare the clay at 1 g. One of the were covered to prevent moisture loss. The shear
primary reasons for using KSS 541 (a mixture of strength and moisture content profiles with depth
50% Kaolin, 40% HST95 silica sand and 10% A50 for KSS 541 are shown in Figure 10, which had a
final saturated unit weight of 19.1 kN/m3. Shear
190
strengths were measured using a Controls Testing track itself reaches a significantly larger settlement
pocket shear vane. of 13.52 mm (676 mm at prototype). These
settlements are clearly large compared to the
Shear Strength, s u Moisture Content,
settlement limits on rail lines, highlighting a need
(kPa) w for better understanding of the mechanisms behind
0
0 10 20 0 0.2 0.4 0.6 rail embankment behaviour on soft clay soils.
0
The reasons for the track displacement being
50 50 greater than the displacement of the embankment
Depth at Model Scale (mm)
Depth at Model Scale (mm)

shoulder are clear in Figure 12, which shows the
100 100 surface displacements of the subsoil for a cross-
section of the model visible at the Perspex viewing
150 150
chamber. The subsoil displacements are primarily
200 200
focussed around the centreline of the track, reducing
significantly with distance from this point and are
250 250 far lower at the distance from the centreline (40 mm
from centreline) at which the embankment shoulder
300 300 displacements are measured. This highlights the
Figure 10. Shear strength and moisture content with depth need to study the behaviour of rail embankments
variation (free field) for KSS 541 clay prepared in flight at 50g. considering the embankment and subsoil in
combination together, rather considering the
embankment alone; a task which is well suited to the
use of centrifuge modelling.
.
5 CONCLUSIONS
A range of challenges faced during the modelling of
a rail simulator for centrifuge application have been
discussed. Vibrations from a moving train were
considered but identified as secondary to the effects
of the change in stress caused by the train load, as
existing research suggests that shear strains induced
by the vibrations are significantly below the elastic
shear strain threshold for clays.
A 3D approach with a moving train load was
Figure 11. Model track and embankment shoulder settlements chosen to ensure the true behaviour of the
with increasing cycles of train loading for a conventional embankment and track combination was captured,
ballasted embankment (25 mm high) on KSS 541, measured at
the middle of the length of track (at 50 g in the centrifuge).
instead of a 2D plane strain method. This required
careful scaling of the track bending stiffness and a
Figure 11 shows the measured settlements of both sufficiently long model container to permit a
the track and the embankment shoulder with the complete cycle of load for each train pass.
application of 1750 bogey passes for a model Total stress measurement was achieved through
conventional ballasted embankment on KSS 541. the use of Singletact force sensors. These produced
The embankment shoulder settlement increases results comparable with the stresses expected in this
with the number of bogey passes reaching a final application. Total stresses measured during testing
value of 5.39 mm (269 mm at prototype), whilst the indicated that the model container was sufficiently
10
Displacement (mm)
0
-150 -100 -50 0 50 100 150
Surface
-10
-20
-30
Distance from Centre of Track (mm)
Figure 12. Variation of the subsoil surface displacement (relative to original surface levels) across the width of the model
(in cross-section) after 438 bogey passes at the Perspex viewing window.
191
long to model the full approach and pass of a train Randolph, M. and Gourvenec, S. 2011. Offshore
over the instrumentation point, and that in these geotechnical engineering, Taylor and Francis, London,
tests the subsoil experienced stress from the train UK.
bogey at a distance of 11 m (at prototype scale).
Robinson, S., Brown, M.J., Matsui, H., Brennan, A.,
Preparation of a suitable soft clay subsoil was
Augarde, C., Coombs, W. and Cortis, M. 2019.
accelerated by use of a KSS 541 clay-sand-silt Centrifuge testing to verify scaling of offshore pipeline
mixture. This soil shares cohesive mechanical ploughs. International Journal of Physical Modelling in
behaviour with clay but has a higher coefficient of Geotechnics, 19(6): 305-317.
consolidation, accelerating the consolidation phase.
Even so, the models required 72 hours of initial RSSB. 2011. Report RSSB 1386 (revised) - the effects of
consolidation, followed by 24 hours consolidation railway traffic on embankment stability. Rail Safety
between the construction of the embankment and and Standards Board, London, UK. Available from
the application of the rail loads. A soil with https://cdn.networkrail.co.uk/wp-
undrained shear strengths varying with depth from content/uploads/2017/02/RSSB-report-The-effects-of-
railway-traffic-on-embankment-stability.pdf [accessed
2 kPa near surface to 15 kPa at 300 mm deep was
30 January 2020].
achieved as a result.
Viswanadham, B.V.S., Das, A. and Mathur, V.K. 2012.
6 ACKNOWLEDGEMENTS Centrifuge model tests on rail embankments
This work was funded by the EPSRC under grants constructed with coal ash as a structural fill material. In
EP/N009207/1 and EP/N009215/1: LOCORPS: Proceedings of GeoCongress 2012: State of the Art and
Lowering the cost of railways using preformed Practice in Geotechnical Engineering, Oakland,
California, USA, 25-29 March 2012. ASCE, pp. 3786-
embankment systems, which is acknowledged with 3795.
thanks.
Vorster, P., Gräbe, P.J. and Jacobsz, S.W. 2017.
7 REFERENCES Centrifuge modelling of railway embankments under
Connolly, D.P., Kouroussis, G., Woodward, P.K., Alves static and cyclic loading. Journal of the South African
Costa, P., Verlinden, O. and Forde, M.C. 2014. Field Institution of Civil Engineering, 59(3): 2-10.
testing and analysis of high speed rail vibrations. Soil
Dynamics and Earthquake Engineering, 67: 102-118. Zhang, C., Jiang, G., Liu, X. and Su, L. 2018. Centrifuge
modelling and analysis of ground reaction of high-
speed railway embankments over medium
compressibility ground. KSCE Journal of Civil
Engineering, 22(12): 4826-4840.
192
Centrifuge modelling of tunnel-soil-pile interaction considering the presence of

rigid caps
Jingmin Xu1, Andrea Franza2, Alec M. Marshall1
1
Canales y Puertos, Universidad Politécnica de Madrid, Spain

2
Corresponding author: Jingmin Xu (jingmin.xu@nottingham.ac.uk)
ABSTRACT: Previous research on tunnelling beneath piles has assumed that caps (or other elements such
as grade beams) above the pile head are elevated from the ground surface. This neglects the possible
mobilisation of contact stresses beneath the foundation elements at the surface, which may occur as a result
of pile settlements larger than ground surface movements caused by tunnelling. This paper presents a
centrifuge study that compares the response of capped piles (where the cap is relatively stiff or rigid and is
in contact with the soil) and simple piles (no cap) to tunnelling in dry silica sand. The paper includes results
from 5 centrifuge tests, including both pile loading and tunnelling beneath piles. To consider a critical
scenario, both simple and capped piles, subjected to a constant external load during tunnelling, are located
directly above the tunnel. For strain measurement and estimation of internal forces, model piles are
instrumented with fibre Bragg gratings (FBGs) along their shafts, while the external vertical and pile head
settlement are also measured. The responses of simple and capped piles are compared by considering
tunnelling-induced pile settlements and axial forces. Results show that simple pile settlements exceeded
surface movements up to a point of pile failure (due to the loss of pile base capacity), whereas the cap
restrained the pile settlement and satisfied equilibrium by mobilising the soil surface contact stresses
(transferring the load from the pile to the cap). The paper highlights the role and potential benefits of
considering the presence of pile caps in tunnel-pile interaction analyses.
Keywords: Tunnelling; Soil-Structure Interaction; Simple/Capped Pile; Centrifuge Modelling.
1 INTRODUCTION centrifuge models of tunnelling beneath deep

Tunnelling induces stress disturbance and ground foundations of two types: piles with an elevated cap
movements that can threaten existing man-made or having no cap are labelled as “simple pile” (SP),
assets. In urban environments, tunnels are often whereas piles with a cap in contact with the soil are
excavated beneath structures with deep foundations. “capped pile” (CP). The paper reports the results of
To prevent detrimental effects, it is important to pile loading tests (to assess the load-settlement
understand the tunnel-pile interaction mechanisms. curve of the SP and CP cases to vertical loading) and
When a pile with a relatively stiff (referred to as tunnel-pile interaction tests (to compare the
rigid in this paper) cap is subjected to tunnelling- difference between SP and CP piles when subjected
induced ground movements, it is reasonable to to tunnelling induced ground movements). The
expect the mobilization of contact stresses beneath practical implications of neglecting the presence of
the cap, contrary to the case of elevated caps or the pile cap in the tunnel-pile interaction are
where no cap is present. The load carrying capacity discussed by analysing tunnelling-induced pile
of the cap can alter the tunnel-soil-pile interaction. settlements and axial forces.
Therefore, neglecting the presence or the action of
pile caps results in uncertainties for risk assessment 2 EXPERIMENTAL DETAILS
purposes. While previous experimental studies have 2.1. Experimental set up
considered the pile response to underground 2.1.1 Model details
excavations, the presence of rigid pile caps is
generally neglected (Jacobsz et al., 2004; Marshall The plane strain experimental package designed by
and Mair, 2011; Ng et al., 2013; Williamson et al., Zhou (2014) for modelling a tunnel excavation
2017; Song et al., 2019), or elevated caps not in process was used. It includes a strongbox, a
contact with the ground are assumed (Loganathan et transparent acrylic front wall, an aluminium back
al., 2000, 2001; Boonsiri and Takemura, 2015; wall, and a flexible membrane model tunnel filled
Hong et al., 2015; Franza and Marshall, 2018, 2019; with water. The model and results presented here
Chen, 2019). are described in model scale unless otherwise stated.
In this paper, the problem is experimentally Tunnel volume loss was simulated by extracting
studied using elevated-gravity and reduced-scale water from a model tunnel with a diameter Dt =
193
90 mm. For the dry soil sample, a fine-grained silica piles. For further details on to the use of fibre Bragg
sand known as Leighton Buzzard Fraction E was gratings in pile testing, refer to Song et al. (2019).
used. The loading system developed by Song et al.
(2019) was used to apply the external load on the
Dummy pile Simple/capped pile SP/CP piles; it consists of stepper motors,
gearboxes, die springs, loading bars, and associated
#1 iLci supporting frames and connectors. A load cell (LC)
#1
#2
Cap and a linear variable differential transformer
dp (LVDT) were used to measure pile head loads and
#3 FBG sensor vertical displacements of the SP/CP piles, as shown
#2 dp in Figure 2. The loading system can achieve both
iLpi
iLpi
#4 force control (setting a target load) and

cCc
ds
displacement control (at constant speed), using the

#3 #5 db load cell and LVDT measurements as inputs to the
LabVIEW control program.
Temperature sensor
LVDT
Loading bar
Figure 1. Experimental geometry and layout.
Load cell
Figure 1 displays the test configuration for the Dummy pile for
tunnel-pile interaction tests, while pictures of the temperature sensors
setup are shown in Figure 2. Two instrumented piles Fibre Bragg gratings
are present: a simple or capped pile located above Pile cap
the tunnel and a “dummy” pile for FBG temperature Model pile
compensation placed 200 mm from the tunnel. In all
tests, model piles were placed within in the middle (a) Pile installation (b) Experimental set up
of the strongbox width. Figure 2. Pile installation and setup of a centrifuge model.
Both model piles were machined from an
aluminium alloy round hollow bar (inner diameter 2.1.2 Testing plan
of 10 mm, external diameter 12 mm, 200 mm long). Five tests including three pile loading tests and two
The pile embedment length is Lp= 150 mm with a tunnel-pile interaction tests were performed on the
60° conical tip. For the capped pile, a square 10 mm University of Nottingham Centre for Geomechanics
thick aluminium plate was bolted to the simple pile; (NCG) 4-m diameter geotechnical centrifuge. Two
the cap had a side length Lc = 48 mm. To simulate a displacement controlled loading tests were
rough surface, a thin layer of sand was bonded to the performed for the simple pile SP (to check the
under-side of the cap and along the pile shaft, giving repeatability between tests), while one loading test
a final pile diameter dp = 13 mm. As shown in was performed for the capped pile CP. A single
Figure 1, to measure the axial forces along the tunnel-pile interaction (TPI) test was performed for
model piles, two fibres with 5 fibre Bragg grating each of the SP and CP piles, as shown in Table 1,
(FBG) sensors (spacing of 45 mm) were glued on imposing a constant external force of 400 N above
the opposite inner sides of the hollow bar. the foundation. In the loading tests, the model tunnel
Importantly, the FBG positioning allowed was not included in the centrifuge container.
measurement of cap, total shaft, and base loads. To
compensate for the temperature change effects on Table 1. Testing plan.
the FBG measurements of the SP/CP pile, the No. Test type Pile type Load (N)
dummy pile was instrumented with three T1 Loading Simple pile -
temperature sensors (spacing of 75 mm). Axial T2 Loading Simple pile -
strain was taken as the mean value between two T3 Loading Capped pile -
opposing FBG sensors on the piles which was
T4 TPI Simple pile 400
corrected for temperature based on separate tests
T5 TPI Capped pile 400
conducted to assess the temperature-strain
relationship for each FBG. Finally, axial load was
obtained based on axial load calibration tests on the Tests were performed at an acceleration of 68
times gravity (68 g) to simulate, at prototype scale,
194
the load-settlement response of a 10.2 m long and 3 RESULTS AND DISCUSSION
0.88 m diameter pile and its response to the 3.1. Pile vertical loading
excavation of a 6.1 m diameter tunnel with a cover
depth (C) of 12.2 m (C/D = 2.0), while considering Figure 3 shows the load-settlement curves of both
the impact of a stiff square pile cap with a side simple (SP in subplot a) and capped (CP in subplot
length of 3.25 m. For these non-displacement b). Shaft, base and cap load are summarised in
simple piles, Franza and Marshall (2019) showed Figure 4 for a direct comparison. The vertical force
that there is potential for large pile settlements to be at the pile head inferred from the load cell (LC) in
caused by tunnelling. Tests 1, 2, and 3 (see Table 1) are shown. On the
2.2. Testing procedure other hand, FBG data of the axial force at locations
#1-5 are plotted only for Tests 2 and 3; Test 1 results
2.2.1 Test preparation are similar to data from Test 2. Cap, shaft and base
The model preparation is summarized as follows: capacity can be estimated from the FBG data for
(1) With the experimental package on the both SP and CP piles. The load withstood by the cap
centrifuge platform, the sand was manually (when present) was estimated by subtracting
poured into the container to achieve a loose soil FBG#2 from FBG#1, the total shaft friction is given
sample with a relative density, Id of by the difference between FBG#2 and FBG#5,
approximately 30%. The repeatability of the while FBG#5 approximated the base capacity. In
sample preparation was demonstrated by this paper, the ultimate load capacity of the SP pile
Franza and Marshall (2018), and by repeating a is defined for the displacement that mobilises the
loading test (presented later). full/maximum shaft resistance (8% dp in these
(2) The instrumented SP/CP pile was jacked into tests).
the soil sample at 1 g (representing a non-
displacement pile) with a vertical guide and rod,
as shown in Figure 2(a); the LVDT and loading
system were then installed.
(3) The dummy pile instrumented with temperature
sensors was pushed into the target position of
the soil sample at a 1g.
(4) The load control system was activated to add a
constant external load of 5 N (based on the LC).
Subsequently, the model was gradually spun to
68 g, followed by two stabilization cycles
(going from 68 g to 15 g and back to 68 g),
which were performed to help achieve
consistency between tests by reducing localized
high-stress zones within the soil.
2.2.2 Pile loading tests

After reaching the target gravity level, the load
control system was switched from force- to
displacement-control. Next, the pile was jacked at a
rate of approximately 0.01 mm/s until a
displacement of one pile diameter dp was reached. Figure 3. The response of (a) simple and (b) capped piles to
vertical loading.
2.2.3 Tunnel-pile interaction tests
First, simple piles SP are considered with results
At the target gravity level, the external load at the in Figure 3a. Head load measured by the load cells
pile head was increased to 400 N. Subsequently, the in test 1 and 2 show good repeatability. At a
tunnel volume loss Vl,t was simulated by extracting displacement of 8% dp, mobilising the full shaft
water from the model tunnel in increments of Vl,t of capacity, the SP piles have an average ultimate load
0.1%, up to a maximum Vl,t of 3%, while keeping a capacity of 705 N (with an average shaft capacity of
constant head load. At each Vl,t increment, digital 423 N and average base capacity of 161 N),
images were taken for measurement of soil although pile tip load marginally increases for
movements. settlements larger than 8% dp. However, it should be
noted that the readings from the top FBG sensor
195
(FBG#1) are smaller than the load measured by the settlements larger than the ultimate value of 8% dp,
load cell (LC_T2), despite being located above despite a decreased shaft friction at relatively high
ground. For instance, in test 2, FBG#1 reading is pile-soil settlements (see Figure 3).
611 N while the load cell LC gives 693 N at 8% dp.
This discrepancy (in all tests) could be due to 3.2. Pile response to tunnelling
bending action transmitted to the load cells as well Results from Tests 4 and 5 for the case tunnelling
as incorrect temperature of the top FBG sensor of beneath the SP and CP piles are summarised in
the loaded piles (FBG#1, which was above the Figures 5, 6, and 7. Considering an average ultimate
surface, was corrected using temperature sensor #1 load of 705 N from the loading tests of the simple
from the Dummy pile, which was just beneath the piles, a safety factor of approximately 1.75
surface; both of these potential issues will be (=705/400) is obtained for SP piles. On the other
resolved in future tests. hand, a safety factor of 2.3 (= 915/400) can be
associated with the CP case. Furthermore, the
presence of the tunnel (which was not included in
the pile loading tests) may affect the ultimate loads,
leading to lower values of safety factor (Franza and
Marshall, 2019). Consequently, the considered
cases deal with relatively low pile safety factors.
Figure 5 presents the pile head settlements Sp and
head loads (obtained from the load cells) during the
simulated tunnel volume loss Vl,t, along with the
centreline surface settlement of the soil Uz at the
acrylic wall (130 mm from the pile; not exactly the
greenfield condition but a reasonable approximation
of a free field displacement), obtained using Geo-
PIV (White, 2003).
Figure 4. Pile settlements versus cap, total shaft, and base
loads for vertical loading tests. .
The effect of the cap in the loading test is now

considered. Figure 3(b) shows that the CP pile also
mobilised its shaft resistance at a displacement of
approximately 8% of pile dimeter. However, for a
vertical displacement of Sp = 8~12% dp, the cap
load shows a steep rate of increase, the base load
(FBG#5) was marginally affected, while the total
shaft friction decreased due to the action of the cap
on the underlying soil and its interaction with the Figure 5. Pile settlements and loads against tunnel
pile (see Figure 4 shaft data). For large settlements volume loss (model scale).
Sp >12% dp the total load capacity of the CP pile
significantly increases with settlement level due to Figure 5(a) shows that the simple pile SP settled
greater cap load, dominating the overall pile at a higher rate than both the CP pile and the ground
response. These centrifuge results agree with surface; the SP settled 5 mm (38% dp) at a
Kumar and Choudhury (2018), who reported that Vl,t = 0.8%, where the load control system could not
the cap presence significantly affects the load- maintain the specified load (400 N) due to the
settlement response of CP piles both positively and tunnelling-induced loss in the load capacity of the
negatively; the positive aspect refers to the pile. On the other hand, the rate of settlement of the
increased load capacity and the increase of soil capped pile CP is nearly constant with tunnel
confining stress (caused by the compression of the volume loss, and slightly greater than the surface
cap on the soil beneath), whereas the negative movements. At Vl,t = 1.0%, the capped pile had an
aspect refers to the decreased shearing/relative absolute settlement of 0.72 mm (5.5% dp), which is
displacements between the pile and surrounding soil 0.22 mm (1.7% dp) greater than the surface soil
(due to the rather uniform settlements of the soil settlement. To consider a serviceability condition,
mass beneath the cap). for example, at Vl,t = 0.3%, settlements of the SP
Overall, with respect to the SP pile, the CP pile pile, CP pile and surface soil are, respectively, 1.15
has the ability to resist higher loads, especially at (8.8% dp), 0.22 (1.7% dp) and 0.14 mm .
196
Importantly, up to large ground settlements and between FBG#2 and #3, and the pile total shaft load
volume losses of Vl,t = 3.0%, the CP had an absolute in Figure 7). This means that for the SP case, due to
settlement of 1.97 mm (15.2% dp) and a settlement tunnelling, both negative friction in the bottom part
relative to the surface of 0.52 mm (4.0% dp). This of the pile and base unloading developed while
indicates the beneficial role of the pile cap to piles positive friction was mobilised in the top part. At
directly above tunnel excavations for both low and Vl,t = 0.7~0.8 %, pile base capacity was significantly
high tunnelling-induced settlements. reduced; pile failure is caused by the positive shaft
To investigate the change in the pile shaft axial friction in the top part of the pile not being able to
force during tunnelling, Figure 6 presents the pile withstand the external load due to the loss of base
shaft axial force obtained from the FBG sensors load.
(FBG#2 to #5), while in Figure 7 these data are used
to compare SP and CP loads at the cap (when
present), shaft and base. Negative FBG data
indicates that the pile underwent tensile axial forces.
Comparing FBG#2 at Vl,t = 0% in Figure 6a and b
shows that the axial forces at the top of the SP and
CP piles are identical, therefore the experimental
results from the two pile conditions can be directly
compared.
Figure 7. Pile shaft and base resistance against tunnel volume

loss.
For the CP pile, first the load distribution prior to

tunnelling (Vl,t = 0) in Figure 7 is described. The cap
load is minimal, while the external load is supported
mostly by the pile shaft (compressive force at
FBG#2-4 decreasing with depth), despite a small
base load being mobilised (FBG#5). Next,
tunnelling effects are considered. Figure 6(b) shows
that the capped pile CP could sustain the vertical
load throughout the investigated tunnel volume loss
(see Figure 5(b)). For the capped pile, the
compressive axial force at the base (FBG#5) and
along the entire shaft (FBG#2-4) decrease with Vl,t,
up to the initiation of negative (tensile) axial forces
near the pile base (FBG#5 is negative), which are
necessarily associated with the development of
negative friction. For equilibrium, the action against
the external load (and negative friction, if present)
was gradually transferred from the shaft and base to
Figure 6. Pile shaft axial load against tunnel volume loss for the pile cap (see Figure 7). Interestingly, in contrast
(a) simple pile SP and (b) capped pile CP. to the SP case, the total (positive) shaft friction of
the CP pile was nearly constant (slightly increased
In the case of the simple piles SP, the service load before Vl,t =1% and decreased after Vl,t =2%), likely
of 400 N was taken in part by the base capacity, as due the cap contact stresses increasing the soil
shown by FBG#5 in Figure 6(a) and 7 for zero confining pressure and balancing for the tunnelling-
volume loss. Furthermore, Figure 6(a) for the SP induced stress relief (Figure 7).
pile shows that, during tunnelling, the base load The experimental results indicate that the action
(FBG#5) and compressive axial load at the bottom of the pile cap (when in contact with the soil) can be
part of the pile (between FBG#4 and #5) decrease remarkable for tunnelling beneath piles with
while the positive shaft friction of the upper section relatively low safety factors. Interestingly, this
of the pile slightly increases (see the difference interaction mechanism has been shown to be
197
effective from low up to extremely large tunnel 5 ACKNOWLEDGEMENTS
volume losses, indicating the practical importance This project has received funding from the
of this work. European Union's Horizon 2020 research and
innovation programme under the Marie
4 CONCLUSIONS Sklodowska-Curie grant agreement No 793715.
This paper investigated tunnel-pile interaction The first author also recognizes the financial
through geotechnical centrifuge testing, considering support provided by the China Scholarship Council
the influence of a stiff pile cap on both mobilised (CSC) and the University of Nottingham, UK.
capacity and tunnelling-induced settlements. The
following conclusions can be drawn. 6 REFERENCES
(1) In the considered cases, loading test results Boonsiri, I. and Takemura, J. 2015. Observation of
indicate that both simple (no cap) and capped Ground Movement with Existing Pile Groups Due to
piles mobilized their shaft resistance at Tunneling in Sand Using Centrifuge Modelling.
threshold settlements of 8% of the pile diameter Geotechnical and Geological Engineering, 33: 621-
dp, while base load slightly increased for larger 640
settlements. The cap played a significant role
for pile settlements greater than the threshold Chen, S.M., Lu, H., Jiang, H.J., et al. 2019. Three-
displacement (mobilising the full shaft friction): dimensional centrifuge modelling of the effects of
tunnelling on vertical behaviour of pile group. IOP
it steadily increased the pile load capacity and it
Conference Series: Materials Science and
contributed to the decrease of shaft friction. Engineering, doi:10.1088/1757899X/592/1/012125
(2) When the piles, which had relatively low safety
factors, were subjected to tunnelling-induced Franza, A. and Marshall, A.M. 2018. Centrifuge
ground movements, the pile with no cap modelling study of the response of piled structure to
experienced settlements significantly larger tunnelling. Journal of Geotechnical and
than the surface (from small volume losses up Geoenvironmental Engineering, 144(2): 04017109.
to pile failure), due to the loss of pile base load
and positive friction near the pile base. Franza, A. and Marshall, A.M. 2019. Centrifuge and real-
Contrarily, for the capped pile, both losses of time hybrid testing of tunneling beneath piles and
piled buildings. Journal of Geotechnical and
base and shaft load were balanced by the
Geoenvironmental Engineering, 145(3): 04018110.
development of contact stresses beneath the cap
(transferring the pile load to the cap); the cap Hong, Y., Soomro, M. A., Ng, C. W. W. 2015. Settlement
was also effective at preventing pile settlements and load transfer mechanism of pile group due to
much larger than the soil surface. side-by-side twin tunnelling. Computers and
(3) Although the cap role may be negligible for the Geotechnics, 64: 105-119.
pile response to vertical loadings (before the
pile shaft resistance is mobilized), when Jacobsz, S.W., Standing J.R., Mair, R.J., et al. 2004.
tunnelling beneath capped piles, neglecting the Centrifuge modelling of tunnelling near driven piles.
presence of the pile cap (if in contact with the Soils and Foundations, 44(1): 49-56.
soil) can lead to an overestimate of the pile
Kumar, A., Choudhury, D. 2018. Development of new
settlement, leading to overconservative design prediction model for capacity of combined pile-raft
and unnecessary mitigation measures being foundations. Computers and Geotechnics, 97: 62-68.
prescribed. However, the post-tunnelling
condition of capped piles relies on the cap Lee, C.J. and Chiang, K.H. 2007. Responses of single
bearing capacity (e.g. surface excavation could piles to tunneling-induced soil movements in sandy
be detrimental). ground. Canadian Geotechnical Journal, 44: 1224-
It should be noted that, in this work, the load cell 1241
and FBG sensor measurements did not match at the
pile head. Consequently, although results describe Loganathan, N., Poulos, H.G. and Stewart, D.P. 2000.
Centrifuge model testing of tunnelling-induced
qualitatively the soil-pile interaction and the impact
ground and pile deformations. Geotechnique, 50(3):
of the cap, the results presented here should be 283-294.
considered as preliminary; further tests are ongoing
to address these issues. Loganathan, N., Poulos, H.G., Xu, K.J. 2001. Ground
Future work will consider the presence of rigid and pile-group responses due to tunnelling. Soils and
caps and flexible slabs in contact with the soil (for Foundations, 41(1): 57-67.
example, piled raft foundations) when tunnelling.
198
Marshall, A.M. and Mair, R.J. 2011. Tunneling beneath Williamson, M. G., Elshafie, M. Z. E. B., Mair, R. J., &
driven or jacked end-bearing piles in sand. Canadian Devriendt, M. D. 2017. Open-face tunnelling effects
Geotechnical Journal, 48: 1757-1771. on non-displacement piles in clay - part 1: centrifuge
modelling techniques. Géotechnique, 67(11): 983-
Ng. C.W.W., Lu, H., and Peng, S.Y. 2013. Three- 1000.
dimensional centrifuge modelling of the effects of
twin tunnelling on as existing pile. Tunnelling and White, D., Take, W., and Bolton, M. 2003. Soil
Underground Space Technology, 35: 189-199. deformation measurement using particle image
velocimetry (PIV) and photogrammetry.
Song, G., Marshall, A.M., and Heron, C.M. 2019. Load Géotechnique, 53(7): 619-631
redistribution of piles affected by tunnelling: hybrid
centrifuge tests using fibre Bragg grating. Zhou, B. 2014. Tunnelling-induced ground
Proceedings of the XVII ECSMGE-2019, doi: displacements in sand. Ph.D. Thesis, University of
10.32075/17ECSMGE-2019-0236. Nottingham.
199
Centrifuge Modelling of the Reverse Faulting-Tunnel Interaction in Granular

Alluvial Soil considering Tunnel Flexible Joints
Hamid Tohidifar1, Mojtaba Moosavi1, Mohammad Kazem Jafari1
1
Geotechnical Engineering Research Center, International Institute of Earthquake Engineering and Seismology
(IIEES), Iran
Corresponding author: Hamid Tohidifar (hamid.tohidifar@gmail.com)
ABSTRACT: Traffic, metro and water transmission tunnels in urban areas play a crucial role in supporting
the lives of the people, especially after a strong earthquake. They should maintain their stability,
serviceability, and strength in such extreme events. Previous studies have shown that the Permanent Ground
Deformations (PGD) due to earthquake faulting within the soil could severely damage the underground
linear-structures. In this paper, the behaviour of a continuous concrete tunnel in alluvial deposit under the
impact of reverse faulting has been studied. The small-scale physical modelling using the geotechnical
centrifuge has been employed for this purpose. In order to reduce the damages induced by the differential
fault displacements, flexible joints have been introduced in the tunnel lining. The prevalent construction-
joints have been replaced by these flexible joints in this experimental work. The tunnel has been modelled
using an Aluminum Tube which under 25 g centrifugal acceleration has simulated a concrete tunnel with
3m diameter. In this study, the tunnel axis and fault line were perpendicular to each other. The displacement
of a 60° bedrock reverse fault and propagation of the shear bands within the upper soil layer has impacted
the embedded tunnel. Results have shown that the presence of the transverse flexible joints could reduce the
resistance of the tunnel against surface faulting displacements. Furthermore, they make the tunnel’s
behaviour more consistent with the soil deformations. In this respect, these joints decrease the damages that
may occur in the tunnel lining. Three mechanisms, including rigid rotation, rigid movement, and locking,
have been formed in the tunnel segments after the fault offset.
Keywords: Geotechnical Centrifuge; Tunnel; Fault Rupture; Alluvium; Flexible Joint.
1 INTRODUCTION during past earthquakes (Brown et al., 1981).

Natural hazards like earthquakes, landslides, or Although different researchers have studied the
volcanoes are dangerous phenomena in the world effect of the fault movement on the tunnel behaviour
that threaten the lives of humans around the world. (Brown et al., 1981; Burridge et al., 1989; Cai et al.,
Amongst these geohazards, earthquakes are more 2019; Russo et al., 2002; Wang et al., 2012), but
important because they occur several hundred times most of the studies were in the rock media with
a day in different locations worldwide. Therefore, strike-slip faults. There is limited evidence in the
different human-made structures are influenced by literature regarding the effects of faults on alluvial
this type of extreme events and might be damaged tunnels (Burridge et al., 1989; Cai et al., 2019; Kiani
due to its related mechanisms. et al., 2016; Lin et al., 2007).
Underground structures such as pipelines or Moreover, when it is not possible to avoid the
tunnels, which serve as significant lifelines in construction of tunnels through the active fault
modern societies, are more prone to the earthquake regions, a fault crossing strategy is required to
damages owing to their long-linear configuration. reduce the tunnel damages (Hashash et al., 2001).
Hashash et al. (2001) have indicated that two There are very few published results about the
different mechanisms affect the tunnels during the feasible strategies to mitigate the influence of the
earthquakes. These are (1) ground shaking, and (2) faults on the tunnels (Romero and Caulfield, 2012;
ground failure. The former mechanism refers to the Russo et al., 2002; Shahidi and Vafaeian, 2005;
ground deformations due to the seismic waves Wang et al., 2012). Table 1 shows the strategies that
propagation within the earth’s crust. The second have been employed in past practical works to
mechanism points out to the Permanent Ground reduce fault rupture effects on tunnels. The majority
Deformations (PGD) because of the earthquake of these fault crossing techniques have focused,
displacements. These displacements could come again, on strike-slip faulting within the rock media.
from seismic-shaking or earthquake-faulting. Therefore, more research is needed to understand
The earthquake-faulting was one of the main the interaction of a tunnel constructed in soft
causes of the tunnels deformations and collapses grounds (or alluvial deposits) and subjected to the
surface dip-slip faulting.
201
The purpose of this work is to present the results different longitudinal segments of a tunnel are
of an experimental centrifuge study on a model shown in this figure. Reverse faulting with a dip-
tunnel subjected to the displacement of a 60° reverse angle equal to 60° and 3.5 cm vertical offset (0.875
fault. The surrounding medium was sandy soil, and m in prototype scale) has been applied in the
articulated crossing method has been used against hanging-wall side. The relative angle between the
fault offset. fault line and tunnel axis was 90°.
The alluvium material in centrifuge test was
Table 1. Different available strategies to mitigate tunnel Firoozkouh Sand No.161 which its properties are
damages across active fault zones (Power et al., 2006; Wang presented in Table 2 (Ahmadi et al., 2018). The total
et al., 2012).
Strategy Description
depth of the soil layer was 28cm in model scale. It
Over-excavation Enlarging the tunnel cross- is equivalent to 7m of alluvium deposit in prototype
section. scale. The relative density of the soil was Dr= 65%,
Articulated Design Ductile joints increase the which shows medium dense sand. This density
longitudinal flexibility of the obtained using the wet tamping technique and
tunnel. under-compaction method.
Grouting technique Increase the ductility of fault
zone by grouting.
Isolation technique An enlarged primary liner is
surrounding smaller inner liner
with injected frangible materials
in between them.
2 MATERIALS AND METHODS

The centrifuge modelling of the fault-tunnel
interaction has been conducted at the Geotechnical
Centrifuge Facility of the Advanced Soil Mechanics Figure 2: Problem schematic definition (all dimensions in cm).
Laboratory of the International Institute of
Earthquake Engineering and Seismology (IIEES). Table 2: Physical and mechanical characteristics of the
The IIEES’s centrifuge machine is an Actidyn C67- Firoozkouh sand No.161.
2 model with 3.5m beam radius (Figure 1). It can wc φ
USCS Gs D50(mm) emin emax
accelerate a payload equal to 1.5 Tons up to 100g or (%) (°)
0.5 Tons up to 130g. Its swinging basket dimensions SP 2.65 3 0.27 0.548 0.943 37
are 100cm × 80cm × 80cm (length × width ×
height). 2.1 Tunnel properties
A circular concrete tunnel in prototype scale was
simulated with Aluminium material in the
centrifuge. The tunnel geometrical and mechanical
properties in both prototype and model scales are
presented in Table 3. The tunnel’s embedment depth
to diameter ratio is 1.625 which demonstrate a
shallow excavated tunnel in soft soil stratum. This
tunnel could be an urban facility underground
lifeline. It has been shown in Figure 2 that three
segments of a tunnel were modelled in this study.
These segments simulated cast-in-place concrete
lining in continuous tunnels. However, instead of
the transverse expansion or construction joints that
Figure 1: Geotechnical centrifuge of the International Institute
are common in practice in 6 to 12m longitudinal
of Earthquake Engineering and Seismology (IIEES).
spacing (Halvorsen et al., 1995), flexible joints with
The centrifugal acceleration in the centrifuge test low shear and bending rigidities have been used
has been set to be 25 g. Therefore, according to the here. To this purpose, two elastomeric-rubber-foam
principles of similitude in centrifuge and Ng space tubes with 10 cm length have been placed over the
(Wood, 2004), the ratio between linear dimensions two transverse gaps in Figure 2. The extremely
in prototype and model scales was N=25. lower young’s modulus (E≈3MPa) of these
The definition of the problem studied in this paper elastomers compared to the Aluminium one
is schematically illustrated in Figure 2. Three (E=70GPa) have simulated two flexible joints along
202
the tunnel axis. These sleeve-like elastomers have shown in Figure 4. This figure shows explicitly that
also prevented the entrance of the sand into the due to the surface faulting, the deformational
tunnel. mechanisms of segment No.2 were rigid
Because of the dimensional limitations of the displacement and rotation. The rotation angle was
model container, fixed boundary conditions at 3.5 degrees counter-clockwise with 2 cm horizontal
vertical plates of foot-wall and hanging-wall sides displacement to the left.
have been implemented (Figure 2). In this study, it The displacement of the reverse fault has caused
is more desired to survey the performance of the the flexible-joints gaps between the consecutive
middle segment deformations against the surface segments, to be closed (Compare Figure 3 with
faulting. Therefore, the left and right tunnel Figure 4). Therefore, after the simultaneous rotation
segments behaviour were not important. and displacement in segment No.2, a locking
mechanism will happen. In this third mechanism,
Table 3: Tunnel properties in prototype and model scales. the tunnel segments will go partially inside each
Prototype Model other (Figure 4). The amount of the tunnel segments
Material Concrete Aluminium displacement, rotation, and the location of the
E (GPa) 35 70 locking mechanisms, are dependent on some
Diameter (m) 3 0.12
parameters. These are the extent of fault offset, fault
Lining 30 0.6
thickness (cm) angle, and the tunnel segments’ diameters and
EI (MN.m2) 72700 0.19 lengths.
Tunnel Length 17.6 0.705
(m)
Embedment 4.875 0.195
Depth (m)
2.2 Fault simulator box

A split box for simulation of the surface faulting
within the alluvium strata has been designed and
manufactured by authors for Geotechnical
Centrifuge of IIEES. This fault simulator has the
capability to model both normal and reverse faults
with 60 degrees dipping-angle and maximum
vertical offset of 50mm in the hanging-wall side. Figure 3: Reverse fault rupture propagation with 60° dipping-
The simulator also has the feasibility to simulate angle and developed shear bands.
different dipping-angles by substituting the
backward guider wedges with other desired fault
angles. A hydraulic jack, connected to the vertical
plate of the hanging-wall side, has provided the
required force for the fault displacement. The
internal dimensions of the box’s soil container are
800mm×700mm×580mm (length×width×height).
Moreover, particular components and details were
also attached to the box’s walls to model the Figure 4: Tunnel's segments deformation after fault’s 3.5 cm
boundary conditions with desired flexibility. These vertical and 2.0 cm horizontal displacements.
components could model the fixed, free, or semi-
flexible boundaries for the buried linear-structures Due to the usage of aluminium tube for modelling
like tunnels or pipes. of a concrete tunnel, segments in the centrifuge test
have not shown any failure or collapse.
3 RESULTS AND DISCUSSION Nevertheless, the segments qualitatively displayed
the dominant deformational mechanisms in the
The propagation of shear bands within the soil layer
tunnel after the fault rupture. According to Power et
after the 3.5 cm vertical and 2 cm horizontal reverse
al. (2006), because the distribution of the fault
fault displacements has been illustrated in Figure 3.
displacements in the soft soils occur in relatively
It is evident in this figure that the ground
extensive zones, using the articulation method with
displacements and shear bands influenced the
flexible joints is more desired. If concrete sections
segments No.1 and 2. Segment No.2 could freely
have enough strength and stability against the
rotate and move between segments No.1 and 3. A
imposed soil stresses, and flexible joints do not fail
longitudinal section of the tunnel deformation is
203
during the large fault displacements, then the failure Brown, I., Brekke, T., Korbin, G., 1981. Behavior of the
mode of the tunnel would be due to the locking Bay Area Rapid Transit Tunnels through the
mechanism. Hayward Fault. Final Report to the US Department of
Figure 4 also shows that a gap has been formed in Transportation. Report No. UMTA-CA-06-0120-81-
1.
the right side of the picture, just in the location that
shear bands (Figure 3) have reached the bottom of Burridge, P.B., Scott, R.F., Hall, J.F., 1989. Centrifuge
segment No.1. This gap has shaped due to the study of faulting effects on tunnel. Journal of
hanging-wall movements and the stationary shape geotechnical engineering 115, 949-967.
of the foot-wall. The tunnel in this portion acted as Cai, Q., Peng, J., Ng, C.W., Shi, J., Chen, X., 2019.
a shield and has not allowed the upper soil grains to Centrifuge and numerical modelling of tunnel
fill the underlying gap. Therefore, zero soil pressure intersected by normal fault rupture in sand.
was available in this section. Computers and Geotechnics 111, 137-146.
Halvorsen, G.T., Poston, R.W., Barlow, P., Fowler,
4 CONCLUSION D.W., Palmbaum, H.M., Barth, F.G., Gergely, P.,
This paper has argued that how is the deformational Pashina, K.A., Bishara, A.G., Hansen, W., 1995.
mechanisms of a continuous concrete tunnel with Joints in concrete construction.
flexible joints during the reverse surface faulting. Hashash, Y.M., Hook, J.J., Schmidt, B., John, I., Yao, C.,
For accommodation of the tunnel behaviour with 2001. Seismic design and analysis of underground
fault displacements, the construction joints of the structures. Tunnelling and Underground Space
tunnel have been replaced with flexible joints in this Technology 16, 247-293.
study. The flexible joints have been simulated with Kiani, M., Akhlaghi, T., Ghalandarzadeh, A., 2016.
elastomeric-rubber-foam in geotechnical centrifuge Experimental modeling of segmental shallow tunnels
test. Using these ductile joints, the tunnel could in alluvial affected by normal faults. Tunnelling and
distort into an S-shape in the fault zone with lower Underground Space Technology 51, 108-119.
probable damages. Lin, M.-L., Chung, C.-F., Jeng, F.-S., Yao, T.-C., 2007.
Three mechanisms have been obtained for tunnel The deformation of overburden soil induced by thrust
after the centrifuge test. They were rigid rotation, faulting and its impact on underground tunnels.
rigid movement, and locking mechanisms. The first Engineering geology 92, 110-132.
two mechanisms occurred simultaneously during Power, M., Fishman, K.L., Makdisi, F., Musser, S.,
the fault movement. The latter mechanism could be Richards Jr, R., Youd, T.L., 2006. Seismic retrofitting
happened due to the large displacements of the manual for highway structures: Part 2-retaining
reverse fault, along with closing the flexible joints structures, slopes, tunnels, culverts and roadways.
gap. If concrete tunnel cross-sections have enough Technical Report MCEER-06-SP11, MCEER,
strength against the induced earth stresses, the University at Buffalo, The State University of New
failure of the tunnel would be due to the locking York.
mechanism of adjacent segments. This, in turn, Romero, V., Caulfield, R., 2012. Improving the Seismic
shows that in the designing process of flexible Resilience of Lifeline Tunnels, NZSEE Annual
joints, the suitable gap-space should be obtained Technical Conference, Christchurch, New Zealand.
based on three parameters. These parameters are Russo, M., Germani, G., Amberg, W., 2002. Design and
probable maximum fault displacement, tunnel construction of large tunnel through active faults: a
segment length, and tunnel diameter. recent application, Proceedings International
Conference of Tunneling and Underground Space
5 ACKNOWLEDGEMENTS Use, Istanbul.
The authors wish to express their thanks for the Shahidi, A., Vafaeian, M., 2005. Analysis of longitudinal
financial support of the International Institute of profile of the tunnels in the active faulted zone and
Earthquake Engineering and Seismology (IIEES), designing the flexible lining (for Koohrang-III
under project number 6137. The authors also tunnel). Tunnelling and underground space
acknowledge Mr Kamran Shirazian for the support technology 20, 213-221.
in carrying out the geotechnical centrifuge test. Wang, Z., Zhang, Z., Gao, B., 2012. Seismic behavior of
the tunnel across active fault, The 15th world
6 REFERENCES conference on earthquake engineering, Lisbon,
Portugal, Sept, pp. 24-28.
Ahmadi, M., Moosavi, M., Jafari, M.K., 2018.
Experimental investigation of reverse fault rupture Wood, D.M., 2004. Geotechnical Modelling. Spoon
propagation through wet granular soil. Engineering Press.
Geology 239, 229-240.
204
Soil Improvement of a Subway Station Excavated by an Earth Pressure Balance

Machine
Mohammad Zamani Ahmad Mahmoudi.1, Ali Entezari Zarandi2, Milad Naghibi3, Ali Azad4
1
Kavoshgaran Consultant Engineers, Senior Consultant, Iran
2
Institut national de la Recherche Scientifique, Postdoctoral Research Fellow, Canada
3
The Wellington Institute of Technology, Academic Staff Member, New Zealand
4
Shahrood University of Technology, PhD candidate in Mining Engineering Faculty, Iran
Corresponding author: Mohammad Zamani Ahmad Mahmoudi (miszamani@gmail.com)
ABSTRACT: In this research, the soil stability around U7 station of Tehran Subway was physically and numerically
investigated using a miniature tunnel-boring machine (TBM) and FLAC3D finite difference program, respectively. The
results demonstrated that instability occurs around launch portal and thus, an appropriate soil improvement technique
should be adopted. After comparing the possible techniques, usage of fore pole umbrella above the tunnel and
installation of fiberglass nails at tunnel face were chosen and utilized. Finally, the number, length, diameter and
strength features of fore poles and fiberglass nails were determined through sensibility analysis and numerical
modelling.
Keywords: Soil improvement; EPB; Modelling; Plastic Concrete; Fiberglass nails.
1 INTRODUCTION In order to avoid instability in the reception and

In urban areas, the construction of subway stations launch portals, the necessary measurements must be
is conducted in two ways: in the first method, the taken; Hence at the first phase, an appropriate
tunnel is excavated and then stations will be system must be selected to stabilize the soil behind
constructed. In the second method, stations are Tunnel Eye. The Tunnel Eye is defined as a part of
already constructed and then the tunnel is excavated diaphragm wall that is located directly in front of
all over the designed route by Tunnel Boring cutter-head. A soft Eye is a wall type that can be
Machine (TBM). An exemplary of later method is a excavated and penetrated by the TBM; for example:
special supportive station constructed before TBM frozen ground, glass fibre bar reinforced concrete,
arrival. This type of stations are commonly soilcrete and non-reinforced concrete. A hard Eye
constructed in the middle of subway routes and its cannot be excavated and penetrated by the TBM and
main purpose is that, in the rest of subway route, other equipment is need to remove it. Hard Eyes
transportation of precast concrete segments and include: steel bar reinforced concrete, steel sheet
emission of excavated soil to out carries out from piling and concrete with unconfined compressive
this site rather than the first station in where TBM strength exceeding cutter-head capability.
was already set up. The supportive stations have two In second phase, geometric details and resistance
portals: the reception portal that TBM pierces it and features of this system is designed so that it must be
enters to the station; the launch portal that TBM economical and meet the possible instability of two
penetrates to soil through it and continues the portals. The Engineering design of the Tunnel Eye
excavation of remainder part of tunnel route. features has a crucial role in preventing of instability
Support members of stations are usually diaphragm occurrence in both portals.
walls, nails, shotcrete and steel struts that are
installed inside the station space. 2 LITERATURE REVIEW
When the TBM arrives near the reception portal Two papers that specifically focused on launch and
and enters through it, the support elements of wall reception portals challenges with tunnel boring
such as nails, shotcrete and steel struts are machines (TBMs) have been published by Richards
influenced by cutter-head pressure of TBM and may et al. (1996) and Dean et al. (2006). Richards et al.
fail that can bring about to instability of wall. provided a detailed discussion of launch and
Moreover, when excavation process is restarted at reception challenges for pressurized face tunnelling
the launch portal and TBM penetrated into the soil, with precast concrete segment linings. The paper
the risk of soil instability and damage to the support summarized experience on Cairo Metro Line 2 and
elements are highly expected. Financial cost and other global experience on eleven major projects.
time delays due to rehabilitation may be up to Dean et al. also provided a detailed discussion of
several times more than primary costs of stabilizing launch and reception challenges for pressurized face
the reception and launch portals ground. tunnelling with precast concrete segment linings
205
and relates experience on several recent North northern-southern and eastern-westerner parts. Both
American projects. Schürch and Jost (2006) two parts have a particular supportive station in
discussed GFRP soft Eyes within slurry diaphragm middle: U7 station in the middle of northern-
walls for TBM launch and reception with references southern part and G7 station in the middle of
to projects in Australia, Singapore, and Switzerland. northern-southern part. In this research the study
In 2008, Steven et al. gathered prior studies about area is U7 station that is located in south of Goftogo
reception and launch portals challenges at different Park and on the edge of Chamran Highway. The
subway projects excavated by Earth Pressure construction of northern-southern part of line 7
Balance (EPB) machines and presented feasible started by EPB breaking in at Navab shaft (March
solution methods to stabilizing the portals. 2012). Simultaneously, the construction of U7
According to prior studies, Portal challenges and station was begun.
the cost of mitigating depend on combinations of the
many factors such as ground type, strength, Table 1. Properties of EPB and precast segments
permeability, TBM type, initial lining method, Property Unit Quantity
station ground support method, Tunnel Eye type and Outer Diameter m 9.165
specially groundwater pressure. Risks are higher Shield diameter m 9.126
with increased groundwater heads; Excessive Shield length m 10
groundwater inflows might flood the station and Excavation step length m 1.5
TBM, damage to contractor’s equipment and Precast segment length m 1.5
adjacent property, collapse of the station ground
support system and cause drawdown settlement. The U7 station was 800 m long with the width of
Solving this problem, a mechanical seal to resist 100 m. The ground overburden above the tunnel
against water pressure should be designed and varies from 15 to 20 meters. The construction of the
assembled at Tunnel Eye. In addition, the stabilizing station was begun from surface to final floor
systems should then be designed and constructed to approximately 36 meters below ground level. First,
avoid lost ground problems, delays and costs. holes bored on the boundaries of the station box and
then reinforced concrete piles with diameter of 1 m
3 PROJECT DESCRIPTION steel pipe were taken as retaining structures. The
3.1. U7 Station and Tehran Subway Details whole number of piles at perimeter of station was
Tehran Subway is a rapid transit system serving the 88 and the interval between those centres was 2.6
public transportation in Tehran, the capital city of meter. The excavation operation of station box was
Iran. Four lines (lines 1, 2, 4 and 5) of the subway divided into six levels and 5 struts rows. After
are operational while two lines (lines 3 and 7) are removing the soil in level 1, the support elements
still under construction. The Line 7 includes 27 such as shotcrete, nails and direct and diagonal steel
kilometres of tunnel with internal diameter of 9.165 struts were installed. Then these operations were
m excavated by an EPB machine. repeated for next levels until the whole box of
EPB technology is fundamentally based on the use station was constructed and supported. The inner
of excavated soil in an excavation chamber to view of U7 station after construction is shown in
support the ground. During the excavation, the soil Figure 1.
is extruded through the openings of the cutter head
and into a screw conveyor situated within the EPB.
By balancing the thrust speed of the EPB and the
rotation of the screw conveyor, it is possible to
utilize the excavated material within the screw
conveyor to establish a controlled volume and
controlled face support pressure. EPB machine
excavates as it drives forward then erects a precast
concrete segmental ring in its rear shield that seals
against the previous ring. The TBM then drives
forward again off the newly installed segment ring.
The detailed properties of the EPB Machine and
Figure 1. U7 station inner view after construction
precast concrete segments used in line 7 Tehran
Subway project are presented in Table 1.
The line 7 of subway begins in the north from
3.2. Geotechnical and geological properties
Yousef Abad and end at Takhti Stadium in the According to engineering geological survey report,
south-east of Tehran city. The line is included the soil profile throughout U7 station comprises two
206
layers; the first layer is upper layer with average 4.2. Problems at Launch portal
thickness of 4 m, containing rubbles, stone, etc. The Figure 3 illustrates an EPB machine that has
second layer is a mix of very dense gravel soils that penetrated from the station into the improved
continue to depth of 50 m. No exploration holes met ground behind the launch portal. After cutting the
underground water in U7 station ground until depth shotcrete, the primary challenge for EPB launch is
of 55 meter. If underground water existed, the cost to achieve face stability until at least the cutter-head
of ground improvement to achieve the desired can be advanced past a suitable internal mechanical
permeability reduction and suitable strength would seal and the excavation chamber may be filled and
exceed. The moisture of in-situ soil was reported brought to the necessary pressure for ground
10%. The Geotechnical properties of these layers control. After the EPB and at least one full segment
collected from the samples of boreholes at U7 ring have past the mechanical seal, the tail gap can
Station are summarized in Table 2. be thoroughly grouted. When EPB passes the
improvement zone and enters the soil, the necessary
Table 2. Soil properties at U7 station site
face pressure is prepared from muck in chamber.
Upper gravel
Parameter Unit
layer mix
Depth m 0-4 4-50
Young’s modulus, 𝐸𝐸 MPa 0.025 0.090
Poisson’s ratio, ν - 0.2 0.3
Cohesion, C MPa 0.015 0.010
Density, p Kg/m3 1700 2200
Friction angle, φ Degrees 30 34
4 PROBLEMS
4.1. Problems at Reception portal
Figure 2 illustrates the EPB excavation near Figure 3. EPB machine penetrating into the launch portal
reception portal. As EPB approaches to the wall
station, it’s necessary to reduce the face pressure to The main problems in the launch portals are
approximately zero value in front of shotcrete. After explained at below.
cutting the shotcrete by equipment inside the
station, the EPB can enter the station safely. If 4.2.1 Soil instability after cutting shotcrete
analytical assessments or numerical modelling The first step is cutting shotcrete that acts as the soil
results show any instability at the reception portal, support member and exists directly in the front of
then the necessary measurements include ground cutter head at the launch portal. At this moment, if
improvement methods must be adopted and used. the soil behind of the removed shotcrete was
Crossing of EPB at reception portal needs less instable and flowed into the station, there would not
measurements than launch portal. In case of launch be balancing force to prevent of soil instability. In
portal, at first the shotcrete must be cut and then case of U7 Station, by physical and numerical
EPB slowly penetrates to soil; this act may cause modelling, it was found that after cutting the
instability of the soil behind the shotcrete. In U7 shotcrete at Tunnel Eye, the soil behind the launch
station, the problems including instability of face portal is instable. The main methods to meet this
and deviation of cutter head did not exist at problem are:
reception portal but a mechanical seal was needed
to isolate tail-gap grout. 1) Use of grouting methods as jet grouting,
penetration grouting and chemical grouting; these
techniques improve the soil properties like cohesion
and inner friction angle. This method is used when
underground water exists because it may flow to
station and cause soil instability. In case of U7
station not only the length of grouting holes would
exceed than 35 meters and so it had huge
operational costs but the interference of grouting
equipment and steel struts caused many problems.
2) Soil replacement with stronger materials like
Figure 2. EPB machine near the reception portal plastic concrete; in this technique, many holes are
207
bored from ground surface to remove the soil behind 2) use of “start ring” in front of Tunnel Eye; this
the shotcrete and simultaneously are filled with method often is a preferred option against prior
plastic concrete. The cement and bentonite increase method because in addition to control of excavation
the strength and deformation plasticity of plastic route and preventing of cutter head deviation, it
concrete. Use of plastic concrete needs many makes the grouting operation behind the first
laboratory tests to achieve the suitable mix in regard segment ring much easier.
to uniaxial compress strength and allowable
settlement of plastic concrete. The uniaxial 4.2.4 Grouting behind the first segment
compress strength of plastic concrete must be less This problem is composed two challenges:
than cutting tools of the EPB. Preventing the instability of the soil above the first segment and
deviation from main route, the plastic concrete grout scape. In case of U7 station. To ensure the
settlement must be in allowable ranges. Since a stability of the soil above the first segment, it was
certain method to calculate uniaxial compress decided to excavate first 3 meters (two segments
strength and allowable settlement according to that each of them had 1.5 m length) of the tunnel
amount of cement, bentonite and water portion manually.
don’t exist, determination of mix layout is To solve the second challenge, a mechanical seal
expensive and time-consuming. was provided to isolate the tail-gap grout for
3) Supporting the soil by installation of fiberglass completion of the first segments.
bolts at Tunnel Eye; in this method, fiberglass bolts
are installed into pre-drilled holes and then cement 5 PHYSICAL AND NUMERICAL
is grouted into them. Axial behaviour of FG bolts is
MODELLING
the main property that increases the soil strength.
This behaviour is due to remarkable tensile strength 5.1. Modelling of U7 station
and large axial stiffness of FG bolts. This method For constructing the U7 station, the reinforced
was decided to support of the soil at U7 station. concrete piles as retaining wall was installed before
the excavation start. First, a vibro-hammer drove a
4.2.2 Soil instability after tunnel excavation casing into the pre-determined piles position and
When EPB starts to excavate the soil, the excavation then the soil within the casing removed by auger.
chamber is vacant and the face pressure to support Then a crane put the steel cages in the boreholes and
the soil does not exist yet. Thus, it is quite vital that concrete was poured into the boreholes to form the
the soil behind the launch portal must be stable in piles. Then the vibro-hammer extracted the casing
the several meters behind the launch portal until the from the ground. This work repeated for 88 piles
excavation chamber is filled with muck. By physical around the station.
modelling of tunnel excavation by EPB at launch Then soil was excavated to the first level and steel
portal of U7 Station, it was found out that FG bolts struts were installed. Then, mesh-reinforced
cannot ensure the soil stability when the chamber is shotcrete and steel nails were used to support the
empty. So the roof umbrella method (pipe roof station walls before the excavation would proceed
method) was decided as an auxiliary method to the lower level. The soil excavated to the second
minimize the soil displacements near the launch level and prior steps were repeated. This process
portal. This pipes act like beams and bear the weight was progressed until the excavation reached to final
of overburden ground. Geometric details and depth. Finally, reinforced concrete slab or base slab
resistive features of FG bolts and roof pipes were was constructed.
designed and obtained by sensitivity and parametric Five rows of struts were installed. Due to
analysis in FLAC3D program. increasing of soil lateral pressure with depth, the
lower steel struts had larger section than upper ones.
4.2.3 Cutter-head deviation from main route In this research, the U7 station construction and
then tunnel excavation by EPB machine, were
Main factor in preventing the cutter-head deviation physically and numerically modelled using a
is operator’s skill to use properly the trust force and miniature tunnel boring machine (TBM) and
excavation parameters based on ground properties. FLAC3D finite difference program, respectively.
Two methods is usually adopted to preventing of the The miniature TBM was constructed to simulate the
cutter head deviation: process of tunnelling. The main box was a
1) Excavation manually of a little length of tunnel; 300×800×800 mm stainless steel box with soil and
in this method, first several meter of soil is the displacement measurement system (Figure 4).
excavated by labours and then EPB machine The shield was made up of three tubes, a stainless
pursues excavation in this pre-excavated route. steel tube with a cutting-head for soil excavation,
208
the middle tube as the tunnel lining and finally, a Table 4. Properties of nails
steel pipe for simulating the tail void formation. The Parameter Unit Quantity
driving part consisted of two engines; the first for Section area mm2 800
advancement of the shield and the second, for Tensile strength MPa 0.22
driving the cutter. The machine was used in a Cohesion, C MPa 4.4
centrifuge under appropriate range of acceleration. Friction angle, φ Degrees 25
Next, the each level inside the station was

removed and simultaneously support members were
installed.
5.2. Modelling of tunnel exaction at reception

portal
Physical modelling at reception portal indicated
stability of the wall. No instability was occurred and
the miniature machine could enter into the station
without any collapse. Numerically, to excavate the
Figure 4. Layout of physical model of TBM and soil Box
tunnel, the EPB features and concrete segment
characteristics were. To achieve the real EPB’s face
FLAC3D program is widely used for deformation
pressure, the tunnel excavated with several face
and stress analysis around underground structures.
pressures. Eventually the value of 50Kpa was
In this present study, the main aim is to probe the
decided for EPB’s face pressure. Also, the grout
displacements and stresses around the reception and
pressure behind segments, that is effective in
launch portals when tunnel is excavated by EPB.
decreasing soil convergences, was decided equal to
The explicit, Lagrangian, calculation scheme and
face pressure plus 150 Kpa.
the mixed-discretization zoning technique used in
First, the soil displacements were assigned to zero
FLAC3D ensure that plastic collapse and flow are
value to investigate the effect of tunnel excavation
modelled accurately. Because no matrices are
by EPB. Then tunnel was bored. Figure 5 shows that
formed, large three-dimensional calculations can be
when cutter-head (white mesh) reaches the 12 m
made without huge memory requirements (FLAC3D
behind the reception portal, the prior plastic zone
manual).
(that was made in station construction phase) is
To emit the boundary conditions effect and
combined with plastic zone that caused by tunnel
demonstrate the real behaviour of portals when are
excavation. The maximum horizontal displacement
affected by tunnel excavation, several grid of U7
of reception portal was 5 mm in this time.
station was built according to real conditions.
Eventually the most suitable mesh was decided. The
size of grid dimensions were 40 m length, 30 m
width and 50 m height. The centre of coordinates of
grid was on tunnel axial at exactly in centre of
tunnel Eye. The U7 Station and Tunnel geometries,
soil properties, boundary conditions and in-situ
stresses were applied to the model. The Mohr-
Coulomb criterion was considered for geo-
mechanical behaviour of soil. The steel struts, nails
and shotcrete were modelled as beam, cable and
shell elements in FLAC3D program. Features of
shotcrete and nails have been presented in tables 3
and 4. For steel struts the steel properties were used.
Table 3. Properties of shotcrete

Parameter Unit Quantity Figure 5. Combination of plastic zone around reception portal
Thickness m 0.25
After excavation of whole tunnel and installation
Young’s modulus, 𝐸𝐸 GPa 21 the final segment, the shotcrete was cut and no signs
Poisson’s ratio, ν - 0.25 of instability were found. Therefore, reception
Density, p Kg/m3 2400 portal was stable and EPB machine successfully
crossed it. Same with numerical modelling, the EPB
209
crossed at the reception portal without any excavation is carried out without EPB’s pressure. In
instability occurrence of soil on 12 March 2018 addition, to insure the stability of work face after
(Figure 6). cutting the shotcrete, fiberglass nails are installed at
the tunnel section. The fiberglass pipes have
become popular due to their cost-effectiveness, and
some technical advantages such as high axial
strength while relatively brittle in the transverse
direction hence easily breakable during excavation.
This technique is often combined with fore pole
umbrella to eliminate the entirely unsupported
spans in weak grounds.
5.3.2 Geometry and resistive features of

stabilizing system
After cutting the shotcrete, three main zone with
highest displacements towards inner space of
Figure 6. EPB break out at reception portal without instability
station appeared. The shape of these zones was
similar to ellipse and their extent increase from
5.3. Modelling of tunnel excavation at launch
tunnel centre to out (Figure 7).
portal
After crossing the miniature machine at reception
portal, the necessary measurements were taken to
continue the tunnel boring from the launch portal.
Therefore the shotcrete was cut and the machine
was placed at front of wall. Through physical
modelling, the soil behind the launch portal was
unstable after cutting the shotcrete. The numerical
results confirmed the findings obtained from
physical modelling.
5.3.1 Choosing the stabilizing system

In order to improve the resistance and permeability Figure 7. Three zones with highest displacements
of the soil behind the launch portals, methods as
dewatering, ground freezing, grouting techniques, The principal criterion in determining the
plastic concrete, fore pole umbrella and fiberglass geometry and resistive features of fiberglass nails is
nails are commonly used. According to geological preventing the expanding of these unstable zones. If
and geotechnical conditions, financial analysis and the nails’ position is properly designed, it will need
technical restrictions, the execution of fore pole minimum number of nails.
umbrella at tunnel crown and installation of
fiberglass nails at the tunnel face had been chosen. 5.3.3 Safety factor of launch portal’s stability
This technique is quite appropriate for soils that the To design the fore poles and fiberglass nails,
peril of underground water flow does not exist when different arrays of both systems were modelled
the excavation of tunnel is started. above tunnel and in three critical ellipses,
Fore pole umbrella includes several poles carried respectively. In each case, the maximum of soil
out above the tunnel crown to sustain the displacements and the maximum of tensile stress in
overburden pressure and control the displacements fiberglass nails were recorded. The safety factor of
caused by tunnel excavation. When the excavation launch portal’s stability is equal to portion of
of station reaches tunnel’s crown level, the fore pole allowable tensile strength to maximum of tensile
s are bored. Then a steel bar is placed in each pole stress that was existed in fiberglass nails.
and the spout is closed with a bonnet. Finally the Due to difference between tensile stresses in
grout is pumped into the overall length of the pole. distinct nails, it is necessary to consider the
Cohesion and frictional angle of grout cause to maximum and minimum of tensile stresses in the
increase the shear strength of soil and decrease the nails. One of the best guides on determination of
extent of plastic zone as soon as excavation is walls safety factor in civil projects has been
started. Most important role of fore pole umbrella is published by Shaw-shong in 2005. According to
maintaining the stability of work face when
210
Showshong, the minimum of safety factor for urban stabilizing system consist of length and diameter of
underground subway walls supported by nails is fore poles and fiberglass nails.
equal to 2.
Since the excavation of first 3 m of tunnel is 5.3.5 Effect of length and diameter of fore
manually done, three safety factors were recorded in poles
models: the safety factor of portal stability in cutting The primary length of fore poles in the Figure 8 was
of shotcrete phase, in excavation phase of first 1.5 15 meter; According to Figure 9, increasing of
m of tunnel length (equal to length of first precast length for values more than 15 meter has not a
concrete segment) and finally in excavation phase remarkable influence on safety factor of launch
of second 1.5 m of tunnel length (equal to length of portal. Increasing of poles length causes to
second precast concrete segment). increasing of safety factor of both shotcrete cut and
According to the geometry and numbers of fore excavation of first 1.5 meter phases but it has not
poles and fiberglass nails in each displacement any effect on safety factor of excavation of second
ellipse, many different arrays of them were 1.5 meter of tunnel.
modelled and their safety factors were recorded. In Similar to length, increasing of diameter causes to
each case, upon to maximum of tensile stresses in increasing of safety factor of both shotcrete cut and
the fiberglass nails, the safety factor of launch portal excavation of first 1.5 meter phases but it has not
has been calculated. The stabilizing system must any effect on safety factor of excavation of second
have at least safety factor of 2 in cutting of shotcrete 1.5 meter (Figure 10).
phase. In all models the fiberglass nails had 10 m
length with diameter of 10 mm and an allowable
tensile strength equal to 688 N/mm2. The length of
10 meter has been decided upon the plastic zone
extent in the soil behind the Tunnel Eye. 2,4
2,2
Eventually, With regards to this point that the
Safety Factor
2
portal stabilizing system must be safe and 1,8
1,6
economic, a layout that is showed in Figure 8, was 1,4
decided as the best design. This layout with the 1,2
1
minimum of fore poles at tunnel crown and 7,5 10 12,5 15 17,5 20 22,5
fiberglass nails in tunnel face, ensures the safety Length [m]
factor of 2. After deciding this layout as the proposal Figure 9. Effect of fore pole length on three safety factors
design, the numerical modelling of tunnel
excavation was executed that was a stable process
without any collapse and instability.
Layout Description 2,4

2,2
Safety Factor
-6 fore poles at 0.5 m 2

1,8
above the tunnel 1,6
- 6 fiberglass nails in 1,4
1,2
central ellipse 1
- 10 fiberglass nails in 5 7,5 10 12,5 15 17,5 20
middle ellipse Diameter [cm]
- 6 fiberglass nails in Figure 10. Effect of fore pole diameter on three safety factors
external ellipse
- Maximum of 5.3.4 Effect of length and diameter of
displacement in tunnel fiberglass nails
face: 2.5 cm The primary length of fiberglass nails in the Figure
-Safety factor: 2.4 8 was 10 meter; According to Figure 11, effect of
changing in fiberglass nails length has two part.
Figure 8. The layout of treatment system at tunnel eye The first part is related to lengths less than 8
meter. In this case, the length of fiberglass nails is
5.3.4 Sensibility analysis less than plastic zone length (10 meter) behind the
Sensibility analysis is necessary for assessment of tunnel Eye. If fiberglass nails length be less than 5.7
geometric features effect of fore poles and meter, the collapse of tunnel face will definitely
fiberglass nails on the launch portal stability. So it occur in the excavation of second 1.5 meter phase.
was done by changing the geometric features of In addition, for fiberglass nails between 5.7 m and 8
211
m, the safety factor of cutting of shotcrete and length is extended around the both reception and
excavation of first 1.5 meter phases decrease while launch portals. Although the reception portal is
the safety factor of excavation of second 1.5 meter stable and safe, that dangerous zone will subject the
phase rapidly increases. soil of launch portal to instability in primary phases
The second part occurred when the length is more of excavation the tunnel. The tunnel face instability
than 10 meter. In this case, the length of fiberglass in cutting of shotcrete phase is reduced by
nails is more than plastic zone length behind the installation of fiberglass nails. Additionally, the
tunnel Eye. According to Figure 11, increasing in face instability after start the excavation of tunnel,
length of fiberglass nails causes to increase the executed without EPB pressure, is omitted by
safety factor of cutting of shotcrete and excavation installation of fore poles.
of first 1.5 meter phases but it has not any effect on
safety factor of excavation of second 1.5 meter. 7 CONCLUSION
The length of fiberglass nails has a remarkable
influence on the face stability in excavation of
second 1.5 m phase. If the length is less than 5.7 m,
2,4 the safety factor of this phase will be less than 1 and
2,2 soil instability definitely occur. The length of
Safety Factor
2
1,8 fiberglass nails must be at least 10 m.
1,6 Increasing of fore pole length and diameter causes
1,4
1,2
to slowly increasing of safety factors of both
1 shotcrete cutting and excavation of first 1.5 meter
5 6 7 8 9 10 11 12 13 phases but it has not any effect on safety factor of
Length [m] excavation of second 1.5 meter. Increasing grout
Figure 11. Effect of fiberglass length on three safety factors cohesion and tensile strength causes to increasing of
all safety factors remarkably.
The primary diameter of fiberglass nails in the
Figure 8 was 10 mm; According to factory
brochures, the change in fiberglass nails diameter 8 ACKNOWLEDGEMENTS
causes to change in their tensile strength and axial Experimental studies, were supported by Tehran
stiffness. Increasing in diameter causes to decrease Subway Organization grant.
the tensile strength and increase the axial stiffness.
The axial stiffness of nails is equal to multiply its 9 REFERENCES
section area by its axial deformability module. So in Dean, A., and Young, D. J. 2006. A Framework for
addition to change in nails diameter, their tensile Design of Tunnel Eyes. Proceedings 19th Canadian
strength and axial stiffness change too. In according Tunnelling Conference, 17-20 September 2006,
to Figure 12, increasing in diameter of fiberglass Vancouver, Canada.
nails causes to increase the safety factor of shotcrete
cutting phase remarkably but has less effect in Richards, D. P., Burchell, A. J., Campo, D. W. and
Raymond, P. 1996. Review of break-in and breakout
excavation of first and second 1.5 m safety factors.
concepts for tunnel boring shields in saturated soft
ground. Geotechnical engineering international
resources centre, 6(2): 79-98.
Schürch, M. and Jost, P. 2006. GFRP Soft-Eye for TBM

6 Breakthrough: Possibilities with a Modern
5 Construction Material. International Symposium on
Safety Factor
4 Underground Excavation and Tunnelling, 2-4 February

3 2006, Bangkok, Thailand.
2
1 Shaw-Shong, L. 2005. Soil nailing for slope
0
5 6 7 8 9 10111213141516171819202122
strengthening. Geotechnical Engineering, Gue &
Diameter [mm] Partners Sdn Bhd, Kuala Lumpur, Malaysia, 30-31
May, p. 1-9.
Figure 12. Effect of fiberglass diameter on three safety factors
Steven, W. H., & Andrew, J. F. 2008. Mitigating Tunnel
6 RESULTS AND DISCUSSION Launch and Reception Challenges. The 11th
The results of modelling indicated that after the international conference of IACMAG, 19-24 June,
construction of U7 station, a plastic zone with 10 m Torino, Italy.
212
Performance of Geogrid reinforced soil walls with low permeable backfill: Centrifuge
Study
B.V.S. Viswanadham1, Midhula Jayanandan2, Dipankana Bhattacherjee3, Prince Kumar4
1
Dept. of Civil Engineering, Professor, Indian Institute of Technology Bombay, India
2
Dept. of Civil Engineering, Research Scholar, Indian Institute of Technology Bombay, India
3
Dept. of Civil Engineering, Formerly Research Scholar, Indian Institute of Technology Bombay, India (Currently
Assistant Professor, IIEST, Shibpur)
4
Dept. of Civil Engineering, Former Postgraduate Student, Indian Institute of Technology Bombay, India
Corresponding author: B.V.S. Viswanadham (viswam@civil.iitb.ac.in)
ABSTRACT: The objective of the present study is to assess the performance geogrid reinforced soil walls with
low permeable backfill subjected to rainfall in a large beam centrifuge facility available at IIT Bombay. The
model low-permeable soil containing 20% fines was formulated by blending locally available fine sand and
commercially available kaolin in the ratio of 4:1 by dry weight. A model hybrid geosynthetic was prepared by
combining reinforcing geogrids with drainage non-woven geotextile in an integral manner. Subsequently, a
robust rainfall simulating system was used for inducing rainfall at high gravities by applying relevant scaling
laws related to rainfall modelling in a geotechnical centrifuge at 40 gravities. Results of two centrifuge tests
soil walls of 10 m height, one with the conventional geogrid reinforcement layers and other with hybrid
geosynthetic reinforcement layers, were reported and tested. The surface settlements, movements along the
wall face and pore water pressure profiles developed during rainfall were monitored with the help of data
recorded by pore pressure transducers during centrifuge tests, and by performing image analysis. Furthermore,
displacement contours experienced by geosynthetic layers with the progress of rainfall. The results showcased
the effectiveness of hybrid geosynthetic layers in reducing crest settlements and wall face movements during
rainfall by providing necessary reinforcement action coupled with in-plane drainage. The adoption of this
technique helps in using locally available low-permeable backfills in reinforced soil wall construction.
Keywords: Hybrid Geosynthetics; Low Permeable Soils; Reinforced Soil Walls; Rainfall; Centrifuge Modelling.
1 INTRODUCTION permeable native soils with high fines content are

There is a steady increase in the frequency of heavily used. Such soils are typically characterized
rainfall-triggered retaining wall and slope failures by their low peak shear resistance, loss of shear
around the world, coincident with the effects of strength on wetting, poor drainage characteristics
climate change. Many cases of such failures have and volume changes and creep potential. The
been reported in retaining walls, excavated slopes problems associated with the low-permeable
and road subgrades due to rainwater infiltration. The backfill soils have been studied by various
instability arises due to the inefficiency of the soil researchers like Christopher et al. (1998),
to dissipate the excess pore water pressure Christopher and Stuglis (2005), Koerner and
generated within the reinforced retaining walls and Koerner (2011) and Koerner and Koerner (2018).
slopes during rainfall due to its low permeability. Marginal soils with low hydraulic conductivity
While Koerner et al. (1998) recommends a backfill build up excess pore water pressure during rainfall
material which is completely devoid of fines, due to its inability to drain water quickly and cause
AASHTO (2009) and FHWA (Berg et al. 2009) failure of geostructures. Hence, application of
mandate the use of backfill materials with less than structural mitigation measures is important. In the
15% of fines passing 0.075 mm sieve and with present study, a nonwoven geotextile bonded to a
plasticity index less than or equal to six. A recent geogrid will be used, wherein the geotextile
report by NCMA recommends soils having less than component provides in-plane drainage while the
35% fines and MORTH (Ministry of Road geogrid provides tensile reinforcement.
Transport and Highway, Govt. of India) restricts the A few full scale tests/ field tests were carried out
fines to 15%. Similarly, the boundary of poor to address the effect of climatic events such as
drainage material is defined by Casagrande with a rainfall on geostructures (Stuglis 2010, Portelinha
value of 10−6 m/s for the saturated coefficient of and Bueno 2012, Portelinha et al. 2013). Similarly,
permeability (Holtz and Kovacs 1981). However, Garcia et al. (2007), Olivares et al. (2009), Lee et al.
due to the unavailability and high cost implications (2011), Portelinha and Bueno 2013 and Yoo and
of such good quality soils, poorly draining/low Jang (2013) have simulated rainfall through small-
213
scale physical model tests. While full scale studies fines percentage equal to 20%, close to the
are time consuming and cost intensive, small scale mandated value of 15% suggested by AASHTO and
physical model tests fail to replicate actual stress FHWA in construction of reinforced soil structures.
conditions and are usually considered to have In addition, the coefficient of permeability of the
limited scope while simulating rainfall events. blended soil was observed to be 1.54 x 10-6 m/sec by
Centrifuge based physical model studies, which conducting falling head test on moist-compacted
ensure the most precise replication of field soil samples placed at γdmax and OMC. The
conditions, have been employed in a very limited properties of model soil are summarized in Table 1.
number of works to study the effect of rainfall on
retaining walls. Craig et al. (1991), Ling et al. Table 1 Properties of model soil
(2009), Tamate et al. (2010), Tristancho et al Soil properties Unit Values
(2012), Ling and Ling (2012), Portelinha et al. Specific gravity, Gs 2.62
(2013) and Yoo and Jang (2013) have simulated Particle size distribution
Sand [>0.075 mm - < 4.75 mm] % 80
rainfall conditions in centrifuge at enhanced gravity Silt [>0.002 mm - < 0.075 mm] % 10
conditions to assess the response of reinforced Clay [< 0.002mm] % 10
slopes. However, the existing rainfall simulators Atterberg limits
produced only uniform intensity of rainfall. Scaling Liquid limit, LL % 11.86
Plastic limit, PL % 9.78
down of rain droplets into mist size to reduce impact Plasticity index, PI % 2.08
pressure was explored by few researchers like Ling Soil classification (USCS) SM
et al. (2010), Tamate et al. (2010), Tristancho et al Compaction characteristics a
(2012). The rainfall simulator used in the present Maximum dry unit weight, γdmax kN/m3 18.75
study designed based on the Modified Mariotte’s Optimum moisture content, OMC % 9
Permeability b
principle overcomes all the above discussed Coefficient of permeability, k m/s 1.54 x 10-6
drawbacks. Details of the custom-designed in-flight Shear strength parameters c
rainfall simulator are explained in (Bhattacharjee Cohesion, c' kPa 7

and Viswanadham 2017, 2019). Researchers have Angle of internal friction, ϕ' ° 33
a: Standard Proctor compaction; b: Falling head tests
explored the behaviour of geosynthetic reinforced performed on moist-compacted soil sample at γdmax and
soil slopes (Raisinghani and Viswanadham (2011) OMC; c: CU triaxial tests conducted on moist-compacted soil
and walls with rigid facing under seepage samples at γdmax and OMC.
conditions (Viswanadham et al. (2017), Razeghi et
al. (2019), Mamaghanian et al. (2018, 2019) with 2.1.2 Model Reinforcement
varying drainage means including geocomposite A number of geosynthetic samples were tested to
layers, sand chimney drains, geotextile chimney determine the tensile load-strain characteristics,
drains etc. However, hybrid geosynthetics transmissivity properties and soil-geosynthetic
(combination of non-woven geotextile and geogrid) interaction properties. Finally, a model geogrid
which have the dual functionality of in-plane sample (G1) and a model geotextile sample (N1)
drainage and reinforcing capability, especially in were chosen as parent materials to develop the
low permeable marginal soils, were incorporated model hybrid geosynthetic, referred to as G1N1, by
only by a few researchers like Raisinghani and satisfying the scaling considerations. G1 exhibited
Viswanadham (2011) and Bhattacherjee and an ultimate tensile load (Tgu) of 0.93 kN/m and
Viswanadham (2017, 2018, 2019), Mamaghanian et ultimate tensile strain (εgu) of 18.9% along the
al. (2019) and Razeghi et al. (2019). In this study, machine direction for the wide-width tensile tests
two centrifuge tests are compared to understand the (as per ASTM D4595-2010). N1 has a much lower
improvement in behaviour of a rigid facing peak tensile load load of 0.25 kN/m in the machine
reinforced wall with low permeable backfill soils direction at an ultimate strain value of 9.89%.G1
subjected to rainfall when geogrid reinforcements was scaled based on tensile load-strain and interface
are replaced with dual functional hybrid characteristics criteria to represent bandwidth of
geosynthetics. commercially available prototype woven
geogrids.[Viswanadham and König (2004), Izawa
2 CENTRIFUGE MODEL TESTS and Kuwano (2010) and Balakrishnan and
Viswanadham (2017)]. Identical percentage open
2.1. Model Materials area of 97.4% was ensured for both model and
2.1.1 Model Soil prototype geogrids to avoid danger of losing contact
A locally available fine sand (SP as per USCS) and between soil-geogrid-soil and to prevent scale
commercially available kaolin (CL as per USCS) effects.The hybrid geosynthetic (G1N1) has an
were blended in a proportion of 4:1 (sand : kaolin) ultimate tensile load of 1.01 kN/m and an ultimate
by dry weight, which resulted in a model soil with tensile strain of 18.3%. The radial transmissivity
214
was obtained at a normal stress of 20 kPa as 1.85 x permanent markers with horizontal and vertical
10-6 m2/s. The 4.5 m radius large beam centrifuge distances of 350 mm and 200 mm respectively
facility at IIT Bombay was used to perform between them were used to trace the movement of
centrifuge tests. The specifications of the centrifuge discrete markers glued to geogrid layers and facing
are given in Chandrasekaran (2001). panels during the test. To achieve plane strain
condition, the inner side of Perspex glass of the front
2.1.3 Model Facing and the rear wall was lubricated and then covered
Facing is one of the most integral components of a with flexible polyethylene sheet strips of 100 mm
reinforced soil wall system and provides protection width which reduces friction.
against backfill sloughing and erosion. Rigid The geogrid reinforced soil wall models were
facings are usually preferred over a flexible facing instrumented with three miniature Pore water
due to their ability to reduce deformations and due Pressure Transducers (PPTs) placed along the base
to ease in transferring reinforcement loads. In this layer at a distance of 25 mm (PPT3), 125 mm
study, a precast concrete panel was selected as the (PPT2) and 225 mm (PPT1) from the seepage tank.
prototype wall facing. The prerequisites for Seepage tank was placed towards left hand side of
modelling the facing panel are that it should have model container to commence flow of water into the
low water absorption, light weight and high base layer to simulate an initial ground water table
stiffness. For model facing panels, Eq. (1) gives the up to the base of model reinforced soil walls. The
thickness as well as the type of the facing material seepage tank wall in contact with the soil was
required in centrifuge model tests. provided with perforated holes having a non-woven
1 geotextile covering in order to prevent migration of
(t f ) p  ( E f ) p  3 soil particles into the tank.
(t f ) m =   (1)
N  ( E f ) m 
Water
Watercontainer
container
where, tf is the thickness of the wall facing , Ef is
the elastic modulus of the facing material, N is the Solenoid valve
Solenoid valve
Nozzle
Nozzle assembly
assembly
gravity level and m, p represents model and attaching plate
attaching plate
prototype dimensions respectively. Nozzle withwith
Nozzles hanging
hanging rods
rods
Facing element is made from a marine plywood

sheet. Three layers of 3 mm thick marine plywood Seepage
Seepage tank for
tank for
sheets are joined together using a special wood glue inducing
inducing water table
water table
to get 10 mm (model dimension) thick facing panel.
Facing element consists of six panels, connecting
six layers of reinforcements adopted in centrifuge
model with a vertical spacing of 40 mm (model Toe
Toe drain for
drain for
collecting runoff
collecting runoff
dimensions). The facing panels are aligned to
provide a batter of 6º with vertical and the bottom Camera
Camera
most panel will be connected to an acrylic block to
simulate a strip foundation levelling pad block in the
field. Foundation block will be placed 25 mm below
Figure 1.View of model test package mounted on the swing
the base layer to attain a rigid toe condition and to
basket of centrifuge
create an embedment of 1 m in prototype
dimensions. The geogrid-facing connection should The same model soil was utilized for base layer as
be a hinged type connection and for this purpose, a well as for the reinforced wall portion. First of all,
stainless steel rod of 2.2 mm diameter and 198 mm the base was placed up to the level of foundation
length was used to attach to the panel with the help block. Then an acrylic block of length 200 mm,
of two screws at both side of the panel. width 50 mm and height 20 mm was placed at the
specified position. After that, a temporary wooden
2.2. Model Test Package frame was used to achieve facing inclination of 84°
Figure 1 shows the view of model test package with horizontal. The first facing panel was placed in
mounted on the swing basket of centrifuge. The the groove in the foundation block and construction
internal dimensions of model container used are 760 of base layer was proceeded up to 40 mm from
mm in length, 200 mm in width and 410 mm in container base. Subsequently, the wall was
height. The front wall of this container was made of constructed in seven soil layers compacted at OMC
100 mm thick transparent Perspex sheet for viewing and γdmax, with vertical spacing of 40 mm except at
front elevation of the model during the flight. The bottom and top layers (thickness of 30 mm each).
other three rigid walls are of mild steel plates. Four
215
Plastic markers were used to obtain displacement The deformation stages of the two centrifuge
field and strain along geogrid layers by tracing their models RSW1 and RSW2 are given in Figure 2 and
coordinates in images taken during various stages of Figure 3 respectively, thus indicating excessive
the test. Red food dye was placed on top of each deformation in wall reinforced with geogrid layers
layer to trace water flow within the wall. The (G1) alone.
temporary support was removed and a soil layer of
25 mm thickness was placed in front of the wall to
complete the base layer and to reach an embedment
depth of 25 mm (i.e. 1 m in prototype dimensions)
for the wall toe. PPT readings were regularly
recorded using data acquisition system to the right
of the basket.
2.3. Test Programme (a) At onset of rainfall at 40g (b) t = 1.2 min [1.33 days]
The wall height, H, was equal to 250 mm,
simulating a wall with 10 m height in prototype
dimensions. Wall face inclination, (β), was equal to
84° with horizontal. Six layers of geosynthetic
reinforcements of length of 200 mm (0.8H)) were
provided in the model walls with a vertical spacing
of 40 mm (Sv/H = 0.160). The present study
includes centrifuge tests on two models namely (c) t = 21.44 min[23.82 days] (d) t = 25 min [27.77 days]
RSW1 and RSW2. Model RSW1 was reinforced
with six layers of model geogrid G1. Model RSW2 Figure 2.Deformation stages of Model RSW1 [G1, ng = 6; Sv
= 1.6 m; I = 15 mm/h] at various stages during rainfall
was reinforced with six layers of hybrid
geosynthetic layers derived from non-woven
geotextile N1 and woven geogrid G1. The initial
ground water level was maintained to coincide with
the base layer. The centrifuge was rotated at
constant angular velocity of 92 rpm (revolutions per
minute) to attain 40g, after which the water pressure
(regulated by valve P1) and air pressure (regulated
by valve P2) were set at desired values as per
calibration to simulate rainfall intensity of 15 mm/h. (a) At onset of rainfall at 40g (b) t = 5.2 min [5.78 days]
As pressure values became constant, the solenoid
valve installed in the path of pressurized water flow
from the overhead water tank was activated
remotely to start the rainfall onto the soil wall and
was continued until wall failure occurred or
equilibrium conditions were achieved in terms of
pore water pressure development and wall
deformations. The rainfall was stopped by (c) t = 17.30 min [19.22 days] (d) t = 20.81 min [23.12 days]
switching off the solenoid valve.
Figure 3.Deformation stages of Model RSW2 [G1N1, nhg = 6;
Sv = 1.6 m; I = 15 mm/h] at various stages during rainfall
The surface settlements measured along the wall 3.1. Pore water pressure development during
crest, lateral displacements along the soil wall face rainfall
and strains in geosynthetic reinforcement layers Figures 4(a) and 4(b) depict the pore water pressure
with progress of rainfall were determined by variations with time for the three PPTs measured
performing image analysis on photographs captured within the geosynthetic reinforced soil wall models
in-flight during centrifuge testing using an on-board RSW1, and RSW2 respectively under a moderate
digital camera, while the positive pore water prototype rainfall intensity of 15 mm/h.
pressures generated during rainfall were measured
using pore pressure transducers (PPTs) placed
within the wall model.
216
Time of occurrence of rainfall
Time of occurance inmodel
of rainfall in modelscale,scale,
t m (min) tm (min) The pore pressure readings thus obtained were
25
25
0.0
0.0 4.5
4.5 9.0
9.0 13.5
13.5 18.0
18.0 22.5
22.5 27.0
27.0
used to determine the development of phreatic
surfaces within the soil wall models with the
(kPa)
Rainfall started at t = 0 days

All dimensions are
Rainfall started
progress of rainfall. Figure 5(a) and Figure 5(b)
in mm
20
20 Water table at wall base
I =at
15 tmm/h;
= 0 βdays
pressure (kPa)
25 = 84°
present the phreatic surfaces developed for different
waterPressure
100 100
Water table
PPT 1 PPT 2 PPT 3
15
15
5 PPT
PPT 1 1 at wall base time intervals during rainfall for the soil wall
10
10 0 PPT
PPT 2 2 I=15 mm/h; models RSW1 and RSW2 respectively. The
Water
0 PPT
PPT 3 23
β = 84° phreatic surfaces of hybrid geosynthetic reinforced
Pore Pore
55
wall Model RSW2 indicate lower phreatic level
00 with rainfall as compared to Model RSW1, thereby
00 55Time of occurance
10
10 15
15 20
20 25
25
of rainfall in prototype scale, t p (days)
30
30
highlighting the necessity of inclusion of suitable
Time of occurrence of rainfall in prototype scale, tp (days)
drainage layers within the reinforced wall. Also, it
(a) Model RSW1 [G1, ng = 6; Sv = 1.6 m; I = 15 mm/h] can be observed that irrespective of rainfall intensity
and wall inclination, the water head recorded by the
Time of occurrence of rainfall in model scale, tm (min) PPTs increased rapidly with time away from the toe.
0.0 4.5 9.0 13.5 18.0 22.5 27.0
25 In general, it can be observed that irrespective of
5 PPT 1
rainfall intensity and wall inclination, the water
Pore Water Pressure (kPa)
Rainfall started 0 PPT 2

20 head recorded by the PPTs increased rapidly with
at t = 0 days 0 PPT2 3
15 time away from the toe, towards the wall face for
Water table at wall base the entire duration of rainfall.
10 I=15 mm/h; β = 84°
5
3.2. Displacement vectors obtained in model
soil walls
0 The face movements and surface settlements
0 5 10 15 20 25 30
Time of occurrence of rainfall in prototype scale, tp (days) experienced by the model soil walls during rainfall
(b) Model RSW2 [G1N1, nhg = 6; Sv = 1.6 m; I = 15 mm/h] were analyzed in terms of displacement vectors
plotted on the original wall profile, obtained by
Figure 4.Variation of positive pore water pressure with
progress of rainfall
tracking the L-shaped plastic markers in horizontal
and vertical direction for each selected image.
12
12
t = days
t = 2.11 2.11 days 11 12
12
t = 6.55 days
t = 6.55 days Scale
Scale
8t = 9.88 days 10
10 11
t = 9.88
6tt == 15.44 days days 9 11 mm 10
10
Vertical
Head
21 days 9
t = 15.44
days days 88
Head of
4t = 25.44 I =Rainfall
15 mm/h intensity: 15 mm/h
Vertical distance,
t =water
Initial 21 table
days 7 Sf,max /H = 0.118 88
2 Sf,max /H =days 0.118
ofwater
t = 27.77
66 7
0 t = 25.44 days t= [Prototype
27.77 days scale]
Water
distance, Yp (m)
5 66
0Initial 2water table [Prototype scale]
(m) (m)
5
44
3 44 Yp (m)
22 3
Base layer
Base layer Initial Water Table
Initial water table 1 22
00 Base
Baselayer
layer 1
20
20 18
18 16
16 14
14 12
12 10
10 88 66 44 22 00 00
Distance (m)
Distance (m) 20
20 18
18 16
16 14
14 12
12 10
10 88 66
Horizontal distance, Xp (m)
44 22 00

(a) Model RSW1
12
12 (a) Model RSW1
t = 2.11
t = 2.11 days days 12
t = 6.55 days
11
Scale 12
t = 6.55 days
10t = 9.88 10
Scale
days
t = 9.88 10
t = 15.44 days days 11 mm 10
9 10
Vertical
Head
t =days
5t = 21 15.44 days 8
Head of
t = 25.44 days 8 I Rainfall

S=f,max
intensity: 15 mm/h
15/H mm/h
Vertical distance,
t =water
Initial 21 days
table 7 = 0.033
88
Stf,max /Hdays
= 23.12 = 0.033
ofwater
0 t = 25.44 days 66 [Prototype scale]

t= 23.12 days
distance Yp (m)
Water
0Initial 2water table 5 66

[Prototype scale]
(m) (m)
4
4
3 44
Yp (m)
22
Base layer
Base layer Initial Water Table
Initial water table 1 22
0 Base layer
Base layer
0
20
20 18
18 16
16 14
14 12
12 10
10 88 66 44 22 00 00
Distance (m)
Distance (m) 20
20 15
15 1010 55 00
Horizontal distance Xp (m)
(b) Model RSW2
(b) Model RSW2
Figure 5.Development of phreatic surfaces with rainfall within
Figure 6. Displacement vectors of reinforced soil walls
reinforced soil walls
subjected to rainfall
217
Figures 6(a) and 6(b) show the normalized As observed from Figure 7(b), the maximum
maximum face movement (Sf,max/H) and surface settlement at the penultimate stage of test
corresponding penultimate stage of each centrifuge corresponding to 20.81 min (23.12 days) of rainfall
test in addition to the displacement vectors obtained in Model RSW2 was 0.197 m, which is still
for models RSW1 and RSW2 respectively. Model substantial in magnitude, which is about 0.019 times
RSW2 experienced lesser distress as compared to wall height
Model RSW1. The displacement vectors for Model
RSW1 clearly depict formation of a global failure 3.4. Displacements along the wall face during
surface with the weakest zone clearly being the wall rainfall
facing and crest region of the soil wall, due to drop Figures 8(a) and 8(b) present the displacements
in soil matric suction as the wetting front descends observed along the model wall facing at various
gradually. time intervals, until failure occurred, or equilibrium
conditions were achieved. The face movements
3.3. Surface settlements induced during rainfall were plotted in prototype scale against normalized
The settlement values obtained from image analysis height of the wall (z/H), based on co-ordinates of L-
are plotted against horizontal distance measured shaped inclined plastic markers stuck at the wall
from the crest of the wall in prototype dimensions, facing. The maximum face movement in Model
corresponding to various intervals of time with the RSW1 was about 1.2 m in prototype scale at the end
progress of rainfall. The variation of surface of 25 min (27.8 days) as against a maximum of 0.34
settlement with horizontal distance from the crest of m only in Model RSW2 at the end of 20.81 min
the soil wall for the geogrid (G1) reinforced Model (23.1 days) of rainfall.
RSW1 subjected to 15 mm/h rainfall depicts a
steadily increasing trend with rainfall duration 1.0
1.0
(Figure 7a). The final crest settlement at the onset of Model

ModelRSW1 [G1]
(z/H)(z/H)
RSW1 [G1]
I =I =15
15 mm/h
mm/h
failure at the end of 25 min (27.8 days) was about 0.8
0.8
ofwallwall
1.2 m. t = 0t days
= 0 days
0.6
0.6 t = 1.34 days
t = 0.67 days
height of
height
WallWall
crest
crest Horizontal
Horizontaldistance from
distance from crestcrest of the
of the wall (m) wall (m)
t = 5.78 days
t = 1.33 days
00 22 44 66 88 1010 0.4
0.4 t = 9.93 days
t = 8.03 days
Normalized
00
Normalized
t = 15.71 days
t = 17.18 days
0.2
0.2 t = 23.82 days
t = 22.45 days
(m) (m)
0.3
0.3
t = 27.77 days
t = 23.12 days
settlement
Sc,max t = 0 days 0.0

0.0
0.6
0.6
t = 0 days
t = 1.34 days -1.5
-1.5 -1.0
-1.0 -0.5
-0.5 0.0
0.0 0.5
0.5 1.0
1.0 1.5
1.5
Surface settlement
Face movement (m) (prototype scale)

t = 0.67 days
t = 5.78 days Face movement (m) (prototype scale)
t = 1.33 days
0.9 t = t9.93 days (a) Model RSW1
Surface
0.9
= 8.03 days
Model RSW1
Model RSW1[G1]
[G1] t = t15.71
= 17.18 days
days
1.2
1.2 I = 15 mm/h
I = 15 mm/h
t = t22.45
= 23.82 days
days 1.0
1.0
t = t23.12
= 27.77 days Model RSW2 [G1N1]
(z/H)
days
Model RSW2 [G1N1]
1.5
1.5
0.8
0.8 I =I 15
= 15 mm/h
mm/h
wall (z/H)
of wall
(a) Model RSW1 t = 0 days

t = 0 days
WallWall
crest
crest Horizontal distance
distance from crest
of theof
wallthe
(m) wall (m) 0.6
0.6 t = 1.34 days
heightof
Horizontal from crest t = 1.34 days

00 22 44 66 88 10
10 8
t = 5.78 days
height
00 6 t = 5.78 days
Sc,max 0.4
0.4 t = 9.93
t = 9.93days
Normalized
4 days
Normalized
t = 15.71
t = 15.71days
settlement (m) (m)
0.3
0.3 2 days
0.2
0 t = 22.45
t = 22.45days
0.2
t = 0 days days
Surface settlement
0.6
0.6
t = 0 days
t = 1.34 days t 0= 23.122 days
t = 23.12 days
8 t = 1.34 days 0.0
0.0
t = 5.78 days
6 t = 5.78 days -1.5
-1.5 -1.0
-1 -0.5
-0.5 0.0
0 0.5
0.5
Face movement (m) (prototype scale)
1.0
1 1.5
1.5
0.9
0.9 t = 9.93 days Face movement (m) (prototype scale)
Model RSW2 [G1N1]
Surface
Model RSW2 [G1N1] 4 t = 9.93 days

I =I =15
15 mm/h
mm/h
2 t = 15.71
t = 15.71days
days
1.2
1.2
(b) Model RSW2
0 t = 22.45
t = 22.45days
days
t 0= 23.12 2 days
t = 23.12 days Figure 8. Face movements observed during rainfall in
1.5
1.5 reinforced soil walls
(b) Model RSW2
4 CONCLUSIONS
Figure 7. Surface settlements at various stages of rainfall for
reinforced soil walls In the present study, the potential of hybrid
geosynthetic layers was assessed for improving the
performance of soil walls with marginal backfill
218
subjected to rainfall. A total of two centrifuge model 6 REFERENCES
tests were conducted and based on analysis and AASHTO 2009. Standard specifications for highway
interpretation of the test results, the following bridges. American Association of State Highway and
inferences can be drawn: Transportation Officials. Washington D.C. USA.
i. The pore water pressure values recorded
ASTM-D4595, 2011. Standard Test Method for Tensile
indicated that the geogrid (G1) reinforced soil Properties of Geotextiles by the Wide-width Strip
wall registered increasing phreatic levels during Method. ASTM International, West Conshohocken,
rainfall with a maximum normalized value of PA, USA.
0.105 near the backfill side, thus indicating
Balakrishnan, S, Viswanadham, B. V. S. 2019.
build-up of positive pore water pressures within
Centrifuge model studies on the performance of soil
the soil wall constructed with low permeable walls reinforced with sand-cushioned geogrid layers.
backfill soil owing to absence of drainage Geotextiles and Geomembranes. 47(6):803-814.
elements. In comparison, lower values of pore
water pressures were observed in the hybrid Berg, R. R., Christopher, B. R. and Samtani, N. C. 2009.
Design and Construction of Mechanically Stabilized
geosynthetic (G1N1) reinforced soil wall model,
Earth Walls and Reinforced Soil Slopes. FHWANHI-
with a maximum normalized value of about 10-024 Report. National Highway Institute,
0.033, potentially attributed to the freely Washington, DC, 1:306.
draining conditions provided by geotextiles.
ii. The geogrid (G1) reinforced wall showed a Bhattacherjee, D. and Viswanadham, B.V.S. 2017,
gradual increase in surface settlements with Design and Performance of an Inflight Rainfall
Simulator in a Geotechnical Centrifuge, (ISSN: 0149-
rainfall of 15 mm/h, resulting in wall instability 6115, 5-year Impact Factor: 1.318), American
and face deformations with maximum settlement Society for Testing and Materials (ASTM), United
value as high as 1.18 m in prototype dimensions. States, 41(1): 72-91.
In comparison, the model reinforced with hybrid
geosynthetic layers (G1N1) was stable Bhattacherjee, D. and Viswanadham, B.V.S. 2018. Effect
of Geocomposite Layers on Slope Stability Under
throughout the rainfall event and recorded
Rainfall Condition. Indian Geotech J., 48(2):316-326.
reduced crest settlements of 0.197 m in
prototype dimensions, demonstrating benefits of Bhattacherjee, D., and Viswanadham, B. V. S. 2019.
coupling drainage and reinforcement functions Centrifuge Model Studies on Performance of Hybrid
offered by geogrid and non-woven geotextile Geosynthetic--Reinforced Slopes with Poorly
Draining Soil Subjected to Rainfall. Journal of
respectively for improving the performance of
Geotechnical and Geoenvironmental Engineering,
soil walls with marginal backfill when subjected ASCE.45(12).
to rainfall.
iii. The deformations observed at the wall facing Bui Van, D., Chinkulkijniwat, A., Horpibulsuk, S.,
reduced by ~72% within the hybrid geosynthetic Yubonchit, S., Limrat, I., Arulrajah, A.,
Jothityangkoon, C. 2017. Steady flow in
reinforced wall (G1N1) compared to the geogrid
mechanically stabilised earth walls using marginal
(G1) reinforced soil wall, which indicates the soils with geocomposites. Geosynth. Intl. 24 (6): 590-
effectiveness of hybrid geosynthetic layers as 606.
possible inclusions in improving performance of
soil walls with marginal backfill under rainwater Christopher, B. R., Zornberg, J. G and Mitchell, J. K.
infiltration. 1998. Design guidance for reinforced soil structures
with marginal soil backfills. In: Proceedings of the
Sixth International Conference on Geosynthetics,
Further studies are warranted by investigating the Atlanta, Georgia 2(6): 797-804.
effect of rainfall intensity, rainfall duration, wall
inclination, location of initial water table and Christopher, B. R. and Stuglis, R. S. 2005. Low
vertical spacing of hybrid geosynthetic layers permeable backfill soils in geosynthetic reinforced
soil wall: State of the Practice in North America, In:
placed within the wall cross-section.
Proceedings of North American Geosynthetics
Conference (NAGS), GRI-19, Cooperative
5 ACKNOWLEDGEMENTS Conference Proceedings, Las Vegas, Nevada, USA,
The authors would like to acknowledge the staff and 16.
team members of the National Geotechnical Chandrasekaran, V. 2001. Numerical and centrifuge
Centrifuge Facility (NGCF) in IIT Bombay for their modelling in soil structure interaction. Indian
active cooperation and for extending all support Geotechnical J., 31(1):30–59.
throughout the present study.
219
Craig, W, H., Bujang, B, K, H., and Merrifield, C, M. Olivares, L., Damiano, E., Greco, R., Zeni, L., Picarelli,
1991. Simulation of Climatic Conditions in L., Minardo, A., Guida, A., and Bernini, R., 2009. An
Centrifuge Model Tests, Geotechnical Testing Instrumented Flume to Investigate the Mechanics of
Journal, ASTM, 14(4), 406-412. Rainfall-Induced Landslides in Unsaturated Granular
Soil. Geotech. Test. J. 32(2): 108–118.
Garcia, E, F., Gallage, C, P, K., and Uchimura, T., 2007,
Function of permeable geosynthetics in unsaturated Porbaha, A., Goodings, D.J., 1996. Centrifuge modeling
embankments subjected to rainfall infiltration, of geotextile-reinforced cohesive soil retaining walls.
Geosynthetics International, 14(2): 89-99 J. Geotech. Eng, ASCE, 122(10): 840-848.
Holtz, R. D. and Kovacs, W. D. 1981. An Introduction to Portelinha, F, H, M., Bueno, B, S., Zornberg, J. G., 2012.
Geotechnical Engineering, Prentice-Hall, Englewood Performance of geotextile reinforced soil wall in
Cliffs, N.J, ISBN- 0-13- 484394-0. unsaturated poorly draining backfill soil conditions,
In: Proceedings of 5th European Geosynthetics
Izawa, J., Kuwano, J. 2010. Centrifuge modeling of
Congress, Soil Improvement and Reinforcement, 5:
geogrid reinforced soil walls subjected to pseudo-
455- 465.
static loading. Int. J. Phys. Model. Geotech, 10(6): 1-
18. Portelinha, F.H.M., Bueno, B.S., Zornberg, J.G. 2013.
Performance of nonwoven geotextile-reinforced
Koerner, R. M. and Koerner, G. R. 2011. The importance
walls under wetting conditions: laboratory and field
of drainage control for geosynthetic reinforced
investigations, Geosynth. Int, 20(2): 90-104.
mechanically stabilized earth wall. Journal of
GeoEngineering, 6(1): 3-13. Raisinghani, D. V., Viswanadham, B. V. S. 2011.
Centrifuge model study on low permeable slope
Koerner, R. M., Koerner, G. R. 2018. An extended data
reinforced by hybrid geosynthetics. Geotextiles and
base and recommendations regarding 320 failed
Geomembranes, 29(6): 567–580.
geosynthetic reinforced mechanically stabilized earth
(MSE) walls. Geotextiles and Geomembranes 46 Razeghi, H. R., Viswanadham, B. V. S, Mamaghanian, J.
(10): 904-912. 2019. Centrifuge and numerical model studies on the
behavior of geogrid reinforced soil walls with
Lee, L. M., Kassim, A., and Gofar, N., 2011.
marginal backfills with and without geocomposite
Performances of Two Instrumented Laboratory
layers. Geotextiles amdGeomembranes. 47(5):671-
Models for the Study of Rainfall Infiltration into
684
Unsaturated Soils” Eng. Geol., Vol. 117(1–2):78–89.
Stuglis, R.P. 2010. Fullescale MSE test walls. In: Proc.,
Ling, H, I., Wu, M, H., Leshchinsky, D and Leshchinsky,
GRI-19 Conference, Las Vegas. GII Publication,
B. 2009. Centrifuge Modeling of Slope Instability, J.
Folsom, PA, USA.
Geotech. Geoenviron. Eng., ASCE, 135(6): 758-767.
Tamate, S., Suemasa, N., and Katada, T. 2010.
Ling, H and Ling, H, I., 2012. Centrifuge Model
Simulating Shallow Failure in Slopes Due to Heavy
Simulations of Rainfall-Induced Slope Instability, J.
Precipitation. presented at the 7th International
Geotech. Geoenviron. Eng., ASCE, 138(9): 1151-
Conference in Physical Modelling in Geotechnics,
1157.
Zurich, Switzerland, CRC Press, Boca Raton, FL.
Mamaghanian, J., Razeghi, H. R., Viswanadham, B. V.
Tristancho, J., Caicedo, B., Thorel, L., and Obregon, N.
S, Manikumar, C. H. S. G. 2018. Behaviour of
2012. Climatic Chamber with Centrifuge to Simulate
geogrid reinforced soil walls with marginal backfills
Different Weather Conditions, Geotech. Test. J.
with and without chimney drain in a geotechnical
35(1): 159–171.
centrifuge. Proceedings of the 9th International
Conference on Physical Modelling in Geotechnics. Viswanadham, B, V, S and König, D. 2004. Studies on
17-20 July 2018, University of London, In: scaling and instrumentation of a Geogrid. Geotext.
McNamara, A, Divall, S, Goodey R, Taylor N, Geomembr. 22(5): 307-328.
Stallebrass S, Panchal, J (Eds). Taylor & Francis
Viswanadham, B. V. S., Razeghi, H. R., Mamaghanian,
Group (Pubs.), London, ISBN 978-1-138-34422-8. 1.
J., Manikumar, C. H. S. G. 2017. Centrifuge model
pp.1211-1216.
study on geogrid reinforced soil walls with marginal
Mamaghanian, J., Viswanadham, B. V. S., Razeghi, H. backfills with and without chimney sand drain.
R. 2019. Centrifuge model studies on geocomposite Geotextiles and Geomembranes. 45(5): 430-446.
reinforced soil walls subjected to seepage.
Yoo, C., Jang, D, W. 2013, Geosynthetic reinforced soil
Geosynthetics International. 26(4). 371-387.
wall performance under heavy rainfall. In: Delage, P.,
National Concrete Masonry Association. 1997. Design Desrues, J., Frank, R., Puech, A., Schlosser, F. (Eds.),
Manual for Segmental Retaining Walls, Second Proceedings of the 18th International Conference on
edition, Second printing, J. Collin (Editor), Herndon, Soil Mechanics and Geotechnical Engineering, Paris,
Virginia. pp. 2131-2134.
220
Centrifuge model studies on geogrid and hybrid geosynthetic reinforced slopes

subjected to rainfall
Dipankana Bhattacherjee1, B.V.S Viswanadham2
1
Department of Civil Engineering, Indian Institute of Engineering Science and Technology, Shibpur, Howrah -
711103, Assistant Professor, India
2
Department of Civil Engineering, Indian Institute of Technology, Bombay, Powai, Mumbai - 400076, Professor,
India
Corresponding author: B.V.S Viswanadham (viswam@civil.iitb.ac.in)
ABSTRACT: The objective of the study is to investigate the effect of inclusion of two varieties of
geosynthetic material, referred to as geogrids and hybrid geosynthetics on the seepage, deformation and
stability aspects of slopes subjected to rainfall. Geogrids function as reinforcing elements within slopes,
whereas hybrid geosynthetics integrate the drainage potential of non-woven geotextile with the reinforcing
property of woven geogrid. Centrifuge-based physical modelling was adopted in this study at 30 gravities
using the 4.5 m radius beam centrifuge facility available at IIT Bombay, India. The model slope
corresponded to a prototype height of 7.2 m and a crest width of 7.5 m, and was constructed with silty sand
having 20% fines content and a saturated permeability of 1.54 x 10-6 m/s. Rainfall was simulated using a
custom-designed rainfall simulating assembly, capable of replicating fine mist at high gravities with
intensities varying from 2 mm/h – 80 mm/h. The response of slope models reinforced with geogrid and
hybrid geosynthetic inclusions was monitored for a prototype rainfall intensity of 20 mm/h. It was observed
that the geogrid reinforced slope experienced considerable deformations and increasing phreatic levels with
rainfall due to insufficient drainage, accompanied with geogrid straining in the range of 48%. The inclusion
of hybrid geosynthetics resulted in reduction of pore water pressures by almost 40%, and the crest
settlements and slope face movements were reduced by 86% and 82% respectively. Further, negligible
straining in the order of 8% was observed in the hybrid geosynthetic layers at the end of rainfall event. The
above study thus highlights the importance of coupling reinforcement and drainage functions in the form of
hybrid geosynthetics within slopes subjected to rainfall.
Keywords: Slope Stability; Rainfall; Centrifuge Modelling; Geogrid; Hybrid Geosynthetic.
1 INTRODUCTION The present study explores the possibility of

Instability of soil slopes and retaining walls mitigating rainfall induced slope instability through
followed by catastrophic landslides is a common provision of two varieties of geosynthetic material,
problem in many parts of the world, and cause referred to as geogrids and hybrid geosynthetics.
thousands of deaths and severe infrastructural Under the impact of load, strains in the soil mass
damage each year. Although some failures may induce strains in the geogrids, which in turn,
occur due to construction activities, many cases of generate tensile loads within these reinforcements.
failure have been reported in natural soil walls, Geogrids thus function as reinforcing elements
excavated slopes and road subgrades simply due to within slopes, whereas hybrid geosynthetics
rainwater infiltrating in an otherwise stable soil function as composite materials by integrating the
structure. As per the International Disaster Database drainage potential of non-woven geotextile with the
(EM-DAT), incidences of slope failures due to reinforcing property of woven geogrid.
hydro-meteorological disasters depict a rapidly Centrifuge-based physical modelling technique
increasing trend over the years, and have increased has been adopted in this study to replicate identical
by almost 70% on an average over the last decade. stress-strain response in model and prototype slopes
The mechanism that leads to slope failure is that the subjected to rainfall. Rainfall was simulated using a
negative pore water pressure starts decreasing when custom-designed actuator, capable of replicating
water infiltrates the unsaturated soil. This results in fine mist at high gravities with intensities varying
build-up of positive seepage forces, leading to from 2 mm/h – 80 mm/h. The model silty sand slope
excessive deformations, and in some cases, tested in centrifuge corresponded to a prototype
structural collapse. An effective technique of height of 7.2 m and a crest width of 7.5 m. The
providing stability of earth structures against response of geogrid and hybrid geosynthetic
rainwater infiltration is through the inclusion of stiff reinforced slopes was monitored in centrifuge for a
and durable geosynthetics [Koerner (2012)]. prototype rainfall intensity of 20 mm/h.
221
2 MATERIALS AND METHODS 2.2. Model soil and geosynthetic materials
Small-scale physical modelling of engineered earth The problem of rainfall induced slope instability
structures is widely adopted to provide insight into aggravates if the fill material has low permeability.
field performance. However, the stress levels of Design issues include drainage, deformations and
models scaled down under normal gravity condition reinforcement pull-out. Hence, there arise the
are much smaller than that of prototype structures necessity to investigate the behaviour of slopes
existent in the field. Although full-scale physical constructed with low-permeable fill material under
modelling can account for these complexities, it is rainwater infiltration. Accordingly, the model soil
expensive and time-consuming, and difficult to used in slope preparation was formulated in the
replicate with natural hazards like flooding and laboratory by blending sand and kaolin the ratio of
rainfall. In such situations, geotechnical centrifuge 4:1 by dry weight. The model silty sand has
modelling can be used as an effective tool percentage of fines equal to 20% and a saturated
[Schofield (1980), Taylor (1995)] to investigate the permeability (ksat) of 1.54 x 10-6 m/s, thereby
behaviour of earth structures as scaled-down models representing the properties of locally available
at high-gravity environment. Accordingly, a robust marginal soils. The ultimate tensile capacity (Tgu)
rainfall simulator was developed for studying the and ultimate tensile strain (εgu) of model geogrid
response of geosynthetic reinforced slopes based on (G1) and hybrid geosynthetic (G1N1) were
the principle of centrifuge modelling. Details of evaluated as per the wide-width tensile test
rainfall simulator, scale factors, model materials and procedure outlined in ASTM D 4595 (2005).
test procedure are discussed in subsequent sections. Further, the drainage potential of hybrid
geosynthetic was ascertained based on radial flow
2.1. Scale factors principle outlined in ASTM D 6574 (2006). The
Rainfall is characterized primarily by its intensity, properties of model geosynthetics as determined in
duration, and frequency, and can be scaled at high the laboratory are presented in Table 2.
gravities based on scaling factors derived for time
Table 1. Scale factors adopted in present study
and water flow under transient unsaturated
conditions. The scaling laws briefly discussed Parameter Unit Scale factor
below are derived based on Dell’Avanzi et al. _a
Unit weight of soil (γ) kN/m 3
N
(2004) and Bhattacherjee and Viswanadham Pore water pressure (u) kPa 1
(2018a). Let the notations p and m denote Suction (ψ) kPa 1
respectively the prototype and centrifuge model at Tensile load (Tg) kN/m 1/N
1/N scale, where N corresponds to the geometric
Transmissivity of geotextile (θ) m2/s 1
scale factor or gravity level. The intensity (r) and
Percentage open area (f) % 1
duration (t) of rainfall are scaled in centrifuge as per
Strain (ε) % 1
Eq. (1) and Eq. (2) respectively. Additionally, the
droplet size should be reduced at high gravities to Duration of rainfall (t) h 1/N2
ensure formation of mist, and occurrence of Rainfall intensity (r) mm/h N
equivalent impact pressure on model ground surface N: Gravity level; _a γm/γp = N; m = model; p = prototype
as that experienced by prototype.
Table 2. Properties of model geosynthetic materials
r m = N rp (1) Property Geogrid Geocomposite
tp = N t m
2
(2) Peak tensile load, Tgu (kN/m)
Wide-width 0.93 1.01
Further, standard scaling relationships have been
Zero-span 1.05 1.09
employed in this study to link the behaviour of
model geosynthetics with commercially available Peak tensile strain, εgu (%)
prototypes. The parent geotextile component of Wide-width 18.9 18.3
hybrid geosynthetic was scaled based on identical Zero-span 52.5 52.7
transmissivity requirements outlined in Raisinghani Secant modulus at 5% strain, Jg (kN/m)
and Viswanadham (2011), whereas the parent Wide-width 7.76 8.24
geogrid was modelled based on scaling Zero-span 6.01 7.60
considerations proposed by Viswanadham and Transmissivity 0.07 1.35
König (2004). Table 1 summarizes the relevant (x10-6), θg (m2/s)
scaling laws adopted during centrifuge modelling.
222
2.3. Rainfall Simulator monitored for a prototype rainfall intensity of 20
Reproducing rainfall in a geotechnical centrifuge is mm/h, which corresponds to a heavy rainfall event
possible using a set of nozzles [Kimura et al. as per standard global thresholds of Llasat (2001).
(1991)]. The nozzles used in the present study An initial water table was simulated in model slopes
(Figure 1) are specialized full cone air-atomizing up to the base before the onset of rainfall using the
brass nozzles, capable of producing rainfall in the perforated seepage tank mentioned previously. The
form of fine mist at high gravities by internal slopes were instrumented with four pore pressure
collision of pressurized air and pressurized water. transducers (PPTs) placed above the base layer at
The nozzles are capable of producing fine mist at a distances of 20 mm (PPT4), 125 mm (PPT3), 250
uniform rate in-flight condition for intensities mm (PPT2) and 350 mm (PPT1) from the perforated
ranging from 2 mm/h to as high as 80 mm/h. face of seepage tank in model dimensions. Further,
L-shaped plastic markers made from thin
transparency sheets of 20 mm x 10 mm dimensions
were embedded within the slope front elevation to
track slope displacements with progress of rainfall.
Additional L-shaped plastic markers were glued on
to the slope face to facilitate computation of slope
face movements with rainfall. The model geogrid
and hybrid geosynthetic (G1N1) were cut to a total
length of (LA + LF + LR) and to a width of 200 mm.
The anchorage length (LA) was equivalent to 0.25
times the model slope height H, LF represented the
length along the slope face and the reinforcement
length (LR) was 0.85H. The various stages involved
in model preparation are discussed elaborately in
Figure 1. Air-atomizing nozzle used in the study.
Bhattacherjee and Viswanadham (2018b).
Apart from spray nozzles, the rainfall simulating
assembly consist of nozzle hanging rods, nozzle 3 RESULTS AND DISCUSSION
assembly attaching plate, a water container The surface settlements, slope face movements,
assembly with support system and additional reinforcement strains and pore water pressures
components involving a solenoid valve, a seepage developed during rainfall were investigated in the
tank and run-off collector. The water container used study based on data recorded by pore water pressure
for supplying pressurized water to nozzles is transducers and through image analysis of selected
designed to act like a Modified Mariotte’s tube, images captured during centrifuge tests. The results
capable of maintaining a constant head of water of two centrifuge model tests (T1 and T2) are
with progress of rainfall at high gravities. Further, discussed, wherein Model T1 and Model T2
while designing the simulator, side slots were correspond to slopes reinforced with six layers of
provided at regular intervals along the entire length geogrid (G1) hybrid geosynthetic (G1N1)
of nozzle assembly attaching plate. This enabled respectively. The results are presented in this study
horizontal shift in position of the nozzles, depending up to a point of time beyond which the observed
upon direction of centrifugal rotation to nullify slope deformations and recorded pore water
effects arising due to Coriolis force at high gravity. pressure values depicted negligible variations with
progress of rainfall. The above duration has been
2.4. Model preparation and Test procedure referred to as the ultimate stage of centrifuge test
From here onwards, the values are mentioned in (Table 3) during rainfall simulation in centrifuge.
model dimensions, with corresponding prototypes
within parenthesis. Centrifuge modelling was 3.1. Deformed slope profile
performed at 30 gravities on a representative silty The geogrid (G1) reinforced slope (Model T1) was
sand slope of 240 mm (7.2 m) height, 60 mm (1.8 observed to experience excessive deformations until
m) base layer and 2V:1H inclination, having a crest a time period of 35 min (22 days) from the onset of
width of 250 mm (7.5 m). Tests were conducted rainfall, which has been referred to as the ultimate
using the 4.5 m radius beam centrifuge facility stage of test. Model T2 reinforced with hybrid
available at IIT Bombay, INDIA. A set of four geosynthetic layers (G1N1) was found to be stable
nozzles were placed at the slope crest, while another under 20 mm/h rainfall, and experienced minor
four were placed at the inclined face. The response slope movements close to the slope face up to 38
of geosynthetic reinforced slope models was min (23.5 days) of rainfall [ultimate stage].The
223
deformed profiles of Model T1 and Model T2 horizontal displacements registered in the
captured at respective ultimate stages of 35 min (22 geocomposite reinforced slope (Model T2) are
days) and 38 min (23.5 days) are presented in Figure found to be in the range of 7 mm (0.210 m) and
2(a)-2(b) respectively. 13.33 mm (0.40 m) respectively.
3.3. Reinforcement strains

The corresponding strains generated in geogrid (G1)
and geocomposite (G1N1) reinforcement layers
under 20 mm/h rainfall are presented in Figure 5 at
respective ultimate stage of tests. Figure 5 indicates
considerable straining of geogrids throughout the
reinforced zone. The maximum value of strain
among the peak strains mobilized in individual
reinforcement layers was identified as the maximum
peak strain value (εp,max), which was evaluated as
48% for Model T1 (G1). However, negligible
straining in the order of 8% was observed in hybrid
geosynthetic layers (G1N1) in case of Model T2 at
the end of rainfall event.
(a) Front-elevation of geogrid reinforced slope [Model T1; t = 3.4. Pore water pressure profile
22 days]
The pore water pressure profiles of geogrid (G1)
reinforced slope (Model T1) and geocomposite
(G1N1) reinforced slope (Model T2) captured
through pore pressure transducers are shown in
Figure 6. The pore water pressure values measured
by PPT3 (uPPT3/γH) have been normalized with
respect to the unit weight of model soil (γ)
multiplied by the slope height (H), and are
expressed in prototype dimensions, starting from
the time of occurrence of rainfall. As evident from
the development of pore pressure values with
rainfall shown in Figure 6, the pore water pressures
in hybrid geosynthetic reinforced slope (Model T2)
got reduced by almost 40% as compared to the
geogrid reinforced slope (Model T1). The above is
attributed to the permeable geotextile component of
(b) Front-elevation of geocomposite reinforced slope [Model hybrid geosynthetic, which provided alternate
T2; t = 23.5 days] drainage paths for the infiltrating rainwater. The
Figure 2. Front-elevation of reinforced slope models primary findings derived from the study, as
discussed above, are summarized in Table 3.
3.2. Crest settlements and face movements Table 3. Summary of centrifuge test results
The crest settlements and face movements within Parameter investigated Test T1 Test T2
slopes reinforced with geogrid (G1) and hybrid [G1] [G1N1]
geosynthetic inclusions (G1N1) were obtained by _a
Time, t (days) 22.0 23.5
image analysis. The resultant plots shown in Figure _b
Crest settlement, Sc,max/H 0.14 0.03
3 and Figure 4 respectively indicate that the peak _b
Slope face movement, Sf,max/H 0.18 0.06
vertical and horizontal displacements are substantial _b
Pore water pressure, umax/γH 0.51 0.21
in the geogrid reinforced slope (Model T1). The
Peak reinforcement strain, εp,max (%) 48 8
magnitudes are found to be in the order of about
33.3 mm (1.0 m) and 43.7 mm (1.31 m) respectively
a
Time in prototype scale (ultimate stage of centrifuge
test); bNormalized maximum values.
for Model T1. In comparison, the peak vertical and
224
Time of occurence of rainfall in model scale, t m (min)
0 4 8 12 16 20 24 28 32 36 40
0.0
0.5
Crest settlement, Sc (m) 1.0
1.5 Water table at slope base,

I = 20 mm/h, β = 63°, Sv = 1.2 m
2.0
2.5 Rainfall started at t = 0 days

3.0 Model T1 [G1]
Rainfall started
3.5 Model T9 [G1N1]
at t = 0 days
4.0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
Time of occurence of rainfall in prototype scale, t p (days)
Figure 3. Crest settlement of reinforced slopes with progress of rainfall
6.6
6.0 Water table at slope base
I = 20 mm/h, β = 63°, Sv = 1.2 m
5.4
4.8
Height of slope (m)
4.2
3.6
3.0
2.4
Model T1 [G1]
1.8
Model T2 [G1N1]
1.2
0.6
0.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Movement along slope face in prototype dimension (m)
Figure 4. Face movements of reinforced slopes with progress of rainfall
1.0
Water table at slope base
0.9 I = 20 mm/h, β = 63°, Sv = 1.2 m
0.8
0.7
Normalized height, z/H
0.6
0.5
0.4
Test Time Strain
0.3 t (days) εp,max (%)
0.2 T1[G1] 22.00 48.35
0.1 T2[G1N1] 23.50 8.01

0.0
0 5 10 15 20 25 30 35 40 45 50
Peak strain, ε (%)
Figure 5. Reinforcement strains at ultimate stage of test
225
Time of occurence of rainfall in model scale, t m (min)
0 4 8 12 16 20 24 28 32 36 40
0.6
PPT 3 is located at 7.35 m
distance from slope toe
0.5
0.4
Rainfall started at t = 0 days
uPPT 3 /γH 0.3 Model T1 [G1]

Model T2 [G1N1]
0.2
0.1
0.0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0
Time of occurence of rainfall in prototype scale, t p (days)
Figure 6. Pore water pressure development in reinforced slopes with progress of rainfall
4 CONCLUSIONS a geosynthetic by radial flow, American Society for

Testing and Materials, Pennsylvania, U.S.A.
In the present study, the effect of inclusion of two
varieties of geosynthetic material, referred to as Bhattacherjee, D., Viswanadham, B.V.S. 2018a. Design
geogrids and hybrid geosynthetics on the seepage, and Performance of an In-flight Rainfall Simulator in a
deformation and stability aspects of slopes Geotechnical Centrifuge. Geotechnical Testing
subjected to rainfall are investigated through Journal, ASTM, 41(1): 72-91.
centrifuge-based physical modelling technique.
Rainfall was simulated using a custom-designed Bhattacherjee D., Viswanadham B.V.S. 2018b. Effect of
rainfall simulating assembly, and the response of geocomposite layers on slope stability under rainfall
slope models was monitored for a prototype rainfall condition. Indian Geotechnical Jr., 48(2): 316-326.
intensity of 20 mm/h. The geogrid reinforced slope
Dell’Avanzi, E., Zornberg, J. G., Cabral, A. R. 2004.
was observed to experience considerable Suction Profiles and Scale Factors for Unsaturated
deformations and increasing phreatic levels with Flow Under Increased Gravitational Field. Soils and
rainfall due to insufficient drainage, accompanied Foundations, 44(3): 79-89.
with geogrid straining in the range of 48%. In
comparison, the crest settlements and slope face Kimura, T., Takemura, J., Suemasa, N., Hirooka, A.
movements were less by 86% and 82% respectively 1991. Failure of fills due to rain fall. In Proceedings of
in the hybrid geosynthetic reinforced slopes, along Centrifuge 91, H. Y. Ko and F. G. McLean (Eds.), A.A.
with 40% reduction in phreatic levels at the slope Balkema (Pub.), pp. 509-516.
base. Moreover, negligible peak strains of the order
Koerner, R. M. 2012. Designing with Geosynthetics, 6th
of 8% were registered along hybrid geosynthetics
edition, Xlibris Publishing.
throughout the rainfall event. The above
observations highlight the importance of coupling Llasat, M. C. 2001. An objective classification of rainfall
reinforcement and drainage functions within low- events on the basis of their convective features.
permeability soil slopes subjected to rainfall. International Jr. of Climatology, 21(1): 1385-1400.
5 ACKNOWLEDGEMENTS Raisinghani, D.V., Viswanadham, B.V.S. 2011.

Centrifuge model study on low permeable slope
The authors thank the staff at National Geotechnical
reinforced by hybrid geosynthetics. Geotextiles and
Centrifuge Facility, IIT Bombay, India, for their Geomembranes, 29(6): 567–580.
support throughout the present study.
Schofield, A. N. 1980. Cambridge geotechnical
6 REFERENCES operations. Geotechnique, 30(3): 227- 268.
ASTM D 4595 2005. Standard test method for tensile
properties of geotextile by wide-width strip method, Taylor, R. N. 1995. Centrifuges in modelling: principles
American Society for Testing and Materials, U.S.A. and scale effects, Geotechnical Centrifuge
Technology, Blackie Academic and Professional, U.K.
ASTM D 6574 2006. Standard test method for
determining the (in-plane) hydraulic transmissivity of Viswanadham, B. V. S., König, D. 2004. Studies on
scaling and instrumentation of a geogrid. Geotextiles
and Geomembranes, 22(5): 307-328.
226
Effect of fiber soil reinforcement on the bearing capacity of a surface strip

foundation
Maria Diakoumi1, Kevin Stone1, David Roddy2
1
Civil Engineering, University of Brighton, Brighton, U.K.
2
Amey, Kettering, U.K.
Corresponding author: Maria Diakoumi (m.diakoumi@brighton.ac.uk)
ABSTRACT: The use of randomly distributed fibers as a means of reinforcing the soil matrix has attracted
increasing attention in geotechnical engineering in recent years. Fiber soil reinforcement can be employed
in current practice to strengthen poor quality soil around foundations and/or limit excessive settlements.
This study presents an experimental investigation of the effect of discrete and randomly distributed fiber
reinforcement on the bearing capacity of a shallow foundation. To this aim, a series of centrifuge tests were
undertaken on a surface strip footing assuming plane strain conditions. The results are presented through
plots of bearing pressure versus settlement ratio. The inclusion of fiber reinforcement was found to improve
the foundation’s bearing capacity, reduce settlements significantly and increase the stiffness of the
reinforced soil-foundation system.
Keywords: Fiber Reinforcement; Reinforced Soil; Bearing Capacity; Footings/Foundations; Settlement.
1 INTRODUCTION the reinforcement orientation (Maher and Gray

Several methods and techniques have been explored 1990).
with the aim to improve the bearing capacity and A number of studies have focused on the effect of
limit the settlements of a shallow foundation resting fiber soil reinforcement on the soil’s loading
on poor quality soil. The use of randomly capacity and settlement characteristics. Consoli et
distributed synthetic or natural fibers as a means of al. (2003) conducted a series of plate load tests on
reinforcing the soil matrix is a relatively new soil silty sand reinforced with fibers at high bearing
improvement technique that has attracted increasing pressures and found that the fiber reinforced soil
attention in geotechnical engineering in recent exhibited a significantly stiffer response compared
years; this is due to its cost effectiveness, reliability to unreinforced soil. Consoli (2009) conducted
and bulk availability of fibers. further plate load tests on sand reinforced with 0.5%
Numerous published experimental studies (Gray by weight long polypropylene fibers at different
and Ohashi, 1983; Freitag, 1986, Gray and AI- relative densities. The results showed higher
Refeai, 1986; Maher and Gray 1990, Al-Refeai, bearing capacity and overall stiffness for all tests
1991; Fatani et al 1991; Maher and Ho, 1994; with these effects being more pronounced at higher
Ranjan et al., 1994; Michalowski and Zhao, 1996; values of relative density. This beneficial effect was
Michalowski and J. Čermák, 2002; Yetimoglu and attributed to a reduction in dilation due to the
Salbas, 2003) that involved the conduction of direct inclusion of fibers in the soil matrix. Kumar and
shear, triaxial or unfonfined compression tests have Kaur (2012) presented results of plate load tests on
shown that the inclusion of randomly distributed poorly graded sand reinforced with different
discrete fibers can increase the soil’s peak strength concentrations of randomly distributed fibers and
and stiffness and reduce the post peak shear strength different depths overlying an unreinforced soil
loss. The beneficial effect of the fiber reinforcement layer. A substantial increase in the ultimate bearing
on the soil has been shown to depend on the type, capacity was found; the beneficial effect of the fiber
length, aspect ratio (length to diameter), modulus of reinforcement was more pronounced at a relative
elasticity, concentration and orientation of the fibers density of 55% compared to a respective value of
in the soil matrix as well as the soil’s grain size and 25%. Kumar et al (2011) presented results of small
shape, density and confining stresses. The fiber scale 1g tests and numerical modelling on the
orientation is considered an important parameter bearing capacity of a strip footing founded on a
(Jewell and Wroth, 1987; Palmeira and Milligan, layer of sand reinforced with 1% randomly
1989). Random distribution of fibers within the soil distributed synthetic fibers for different depths of
matrix is usually preferred since it presents the the reinforced layer. The results showed a
advantage of strength isotropy and avoidance of the significant increase in the ultimate bearing capacity
development of a potential weak plane parallel to for depths of the reinforced layer equal to two times
the footing’s width, (B).
227
The published studies presented above have Table 1. Summary of centrifuge tests at 20g
investigated the effect of reinforcing the soil with Test Depth of fiber Fiber Fiber
randomly distributed discrete fibers on its load No reinforced length concentration
layer (mm) (%)
bearing capacity for a specific concentration and
1 0 - 0
length of fiber reinforcement. Therefore, the effect
2 3B 6 0.2
of varying the fiber concentration or length on the 3 3B 6 0.45
load bearing capacity in the case of a strip footing 4 3B 6 0.85
has not been explored widely. Clearly, the 5 3B 19 0.2
improvement in the bearing capacity can be more
realistically assessed under higher values of stresses 2.2. Material and equipment
that correspond to field conditions. To achieve a
The sand used for the tests is dry, poorly graded,
realistic simulation of soil stresses, small scale
Fraction C (300-600 microns) silica sand of rounded
physical modelling facilitated by a geotechnical
and sub-rounded particle shape with 98% silica
centrifuge is ideally suited. This study presents a
content supplied by David Ball Limited. The sand
series of centrifuge tests on a surface strip footing
has a specific gravity of 2.67, maximum and
founded on a two layer-soil; the top layer is
minimum void ratios of 1.06 and 0.61 respectively
reinforced with randomly distributed fibers
and a critical state angle of friction φcrit of 31°. The
extending to a depth equal to 3B, where B is the
fibers used as reinforcement are synthetic polyolefin
footing’s width, and the bottom layer is kept
fibers with a ridged, rough surface and two different
unreinforced for all tests. The effect of two different
cut lengths of 6mm and 19mm respectively; the
fiber lengths and three different fiber percentage
fibers’ physical and engineering properties are
concentrations on the response of the footing under
given in Table 2. The model strip footing was
applied load is explored.
manufactured from aluminium with dimensions
90mm (length) x 15mm (width) x 9mm (depth) and
2 MATERIALS AND METHODS can be seen in Figure 1. A recess was drilled on the
2.1. Experimental programme top surface of the footing to ensure that the vertical
The tests undertaken in this study and presented load is applied at its centre. A layer of sandpaper
herein are summarised in Table 1 below. Table 1 was attached to the bottom surface of the footing to
lists five tests which were conducted in a increase the friction in the interface between the soil
geotechnical balanced beam centrifuge at an and the foundation.
enhanced acceleration level of 20 gravities (20g) to All the tests were carried out in the balanced beam
investigate the influence of the length and geotechnical centrifuge at the University of
concentration of fiber reinforcement on the bearing Brighton. This machine is manufactured by
capacity of a surface strip footing assuming plane Thomas Broadbent & Sons Ltd. and is a 6 g-tonne
strain conditions. The first test investigated the machine (20kg payload to 300g) with a radius of
behaviour of a strip footing founded on 760mm. The tests were conducted in a rectangular
unreinforced dry sand and the results were used for steel strongbox of internal dimensions 300mm
comparison with subsequent tests (numbered 2 to 5 (length), 100mm (width) and 180mm (depth). The
in Table 1) in which two soil layers were introduced strongbox was equipped with clear, acrylic panels
below the footing: i) a layer of fiber reinforced sand placed on the long sides of the box that allowed the
below the footing and ii) a layer of unreinforced sample to be observed during testing. The strongbox
sand directly below the reinforced soil layer. The was placed in a cradle which was then mounted on
depth of the fiber reinforced soil layer was selected the centrifuge arm. A two degree of freedom
equal to three times the footing’s width, B, to ensure actuator was then mounted on the cradle; the
that the failure plane is well within the reinforced actuator in the vertical direction has a capacity of 2
soil zone. The second layer of unreinforced dry sand kN and was used in the tests presented herein.
had a depth of 6.3B. Two fiber cut lengths of 6mm
and 19mm were tested. The fibers were mixed Table 2. Physical and engineering properties of the synthetic
fibers used in the tests
randomly with the soil at different percentage
Property Fiber
concentrations as listed in Table 1.
Type Synthetic polyolefin
Fiber length 1 6mm
Fiber length 2 19mm
Diameter 0.5mm
Specific gravity 0.91
Tensile strength 590 N/mm2
Modulus of elasticity >11 GPa
228
where Vv is the volume of the voids. This method
of determining the mixing quantities of fiber
reinforcement is commonly used in previous studies
(e.g. Michalowski and Zhao, 1996).
(a) The friction between the soil and the acrylic
panels of the strongbox was considered negligible
since the panels appeared in good condition with no
obvious signs of abrasion. The model footing was
placed on the soil surface and then a two degree of
freedom actuator was mounted on the strongbox.
(b) Loads were applied via the vertical actuator so its
Figure 1. Aluminium model footing tested in the centrifuge: piston rod was lowered to the top of the centre (see
a) top surface b) bottom surface recess in Fig. 1) of the footing; a miniature load cell
was attached to the actuator to measure the applied
2.3. Model preparation and test procedures load and the footing’s settlement was determined
The unreinforced soil sample used in test 1 (see from the vertical displacement of the actuator. Two
Table 1) was placed in the rectangular steel Linear Variable Differential Transformers (LVDT)
strongbox and was vibrated to the target density via were positioned on the footing at a distance of
means of a vibrating table. The unreinforced soil 15mm from each edge to measure any potential
specimen had a bulk density of 1575 kg/m3 and a rotation. The model and actuator were set-up on the
sample height of 140mm. The soil sample in tests 2 laboratory bench and then loaded onto the
to 5 consisted of two soil layers; the bottom soil centrifuge. The actuator was connected to the power
layer was unreinforced, had a depth of 6.3B and control system; the actuator and measuring
(95mm), and its target bulk density was achieved gauges were controlled by the in-flight computer
via vibration. The second soil layer was placed which was remotely connected to a desktop
directly on top of the unreinforced soil layer and computer.
beneath the model footing; this layer had a depth of All tests were conducted at an acceleration level
3B (45mm) and was reinforced with randomly of 20g. Once the model had achieved the target test
distributed fibers with the lengths and percentage acceleration the actuator was run at a velocity of 3
concentrations listed in Table 1. The fibers were mm/minute for all tests and data from the load cell
mixed with the soil manually to ensure random and actuator encoders were recorded continuously
distribution and orientation and they were placed in until the end of each test. Figure 2 shows a view of
the strongbox in three layers using dry tamping to the assembled model package post testing. At the
achieve the required density. The average end of each test, the fibers were removed from the
concentration of fibers, fc, in the reinforced soil soil matrix and were inspected.
specimen was calculated as a percentage of the dry
weight of the sand:
fc = Wf / Ws x 100% (1)
where Wf is the weight of fibers and Ws is the

weight of dry sand. The weight concentration was
used as a useful guide for the mixing stage, but
actually the fiber concentration was calculated by
setting the void ratio of the reinforced soil specimen
equal to the respective value of the unreinforced
specimen. To achieve the target void ratio, it was
assumed that the fibers’ volume, Vf, is part of the
solid particles’ volume Vs; this is expressed with
Equation 2:
𝑉𝑉𝑣𝑣 𝑉𝑉𝑣𝑣
e= = (2)
𝑉𝑉𝑠𝑠 𝑉𝑉𝑠𝑠 +𝑉𝑉𝑓𝑓
Figure 2. Centrifuge model package assembled (photo taken
post-testing)
229
3 RESULTS AND DISCUSSION Bearing pressure [kPa]
The experimental results are presented through plots 0 200 400 600 800
of applied bearing pressure versus percentage 0
values of settlement ratio, s/B, where s is the 0%
measured settlement and B is the footing’s width, in 0.20%
order to investigate the behaviour of a strip footing 0.45%
resting on soil reinforced with different fiber 5 0.90%
concentrations and lengths. Figures 3 presents the
results for fiber lengths of 6mm for different fiber
percentage concentrations, namely 0%, 0.20%,
0.45% and 0.85%. Figure 4 shows the results for 10
s / B [%]
19mm long fibers at 0% and 0.2% fiber
concentrations. The results derived for the case of
unreinforced sand consisting of one uniform soil
layer are included in Figures 3 and 4 for 15
comparison.
All the plots presented in Figures 3 and 4
demonstrate a distinct peak value of applied bearing
pressure which is assumed to correspond to the 20
ultimate bearing capacity. For values beyond the
peak, the bearing pressure appears to gradually
reduce but it is still considerably higher than that for
unreinforced sand (fiber concentration equal to 0%). 25
It is apparent from Figures 3 and 4 that the
inclusion of a fiber reinforced soil layer directly Figure 3. Results for 6mm fiber length
below the footing increases the bearing pressure at
any settlement level and reduces the settlement at Bearing pressure (kPa)
any of the values of pressure applied in the 0 200 400 600 800
experiments presented herein. It can also be 0
observed that increasing the fiber percentage 0%
concentration will result in an increase in the 0.20%
capacity of the soil to carry the applied pressure at
the same level of settlement and a reduction in the 5
settlement compared to the case of unreinforced
soil. The increase in the bearing capacity can be
attributed to the mobilisation of additional frictional
resistance between the sand grains and fibers. It 10
s / B [%]
should be noted that for pre-peak values of bearing

pressure, the stiffness of the soil-footing when a
reinforced soil layer is included directly beneath the
footing is greater than that of the unreinforced soil. 15
Figure 5 shows the results for unreinforced and
reinforced soil with 0.2% fiber concentration for
two fiber lengths, 6mm and 19mm for comparison.
20
From Figure 5, it is obvious that the increase in fiber
length from 6mm to 19mm results in an increase in
the ultimate bearing capacity of the foundation.
Specifically, the ultimate bearing capacity in the
25
case of 19mm fiber length at 0.2% concentration is
1.3 times the respective value in the case of 6mm
Figure 4. Results for 19mm fiber length
fiber length.
230
Bearing pressure [kPa] compared to the case of unreinforced soil. As the
fiber concentration increases, the values of
0 200 400 600 800
0
settlement required to reach the ultimate bearing
unreinforced capacity reduce. It is also noticed that the reinforced
6mm soil samples can sustain greater values of bearing
19mm pressure at large post-peak displacements. At the
5
end of each test the fibers were removed from the
soil matrix and were checked; this inspection did not
reveal fiber elongation but a small amount of the
19mm long fibers appeared to be slightly bent.
10
s / B [%]
Table 3. Summary of results from all tests

Fiber Fiber Ultimate Settlement BCR
length concentration (peak) ratio (s/B)
(mm) (%) bearing at max
15 capacity pressure
(kPa) (%)
- 0 271.4 9.83 1
20 6 0.20 498.9 19.03 1.8
6 0.45 578.8 15.72 2.1
6 0.85 686.2 13.86 2.5
19 0.20 662.4 15.6 2.4
25
4 CONCLUSIONS
Figure 5. Results for unreinforced and reinforced soil with
0.2% fiber concentration for two fiber lengths, 6mm and This paper has presented a set of experimental
19mm. results derived from centrifuge tests to investigate
the bearing capacity of a strip foundation resting on
Table 3 lists the ultimate (peak) bearing capacity soil reinforced with synthetic rigid fibers of two
for each fiber percentage concentration and length different lengths and three different percentage
and the percentage values of the settlement ratio concentrations. The results were compared with
(s/B) that correspond to the peak values recorded. It those derived for a case of unreinforced soil. The
is common in the available literature on reinforced inclusion of fiber reinforcement improved the
soil to use the bearing capacity ratio to quantify the foundation’s bearing capacity and reduced the
beneficial effect of the inclusion of reinforcement in settlement significantly. The improvement in the
the soil matrix and it is thus useful to adopt it for the ultimate bearing capacity of the reinforced soil
results presented herein. The bearing capacity ratio compared to the respective value of the
(BCR) is defined in this study as the ultimate (peak) unreinforced soil was expressed by the bearing
bearing capacity of the fiber reinforced soil to the capacity ratio (BCR). For the fiber length and
ultimate bearing capacity (peak) of unreinforced concentrations tested in this study the bearing
soil; the values of BCR for the fiber lengths and capacity ratio (BCR) varied from 1.8 to 2.4 times.
percentage concentrations tested are listed in Table For pre-peak values of bearing pressure, the
3. stiffness of the soil-foundation is greater in the case
From Table 3, the increase in the ultimate bearing of fiber reinforced soil compared to unreinforced
capacity of the soil which has been reinforced with soil. As the fiber concentration increases, the values
6mm long fibers is between 1.8 to 2.5 times the of settlement required to reach the ultimate bearing
respective value of the unreinforced soil layer. For capacity reduce. The reinforced soil samples were
the 6mm fiber lengths, the relationship between found to sustain greater values of bearing pressure
fiber concentration and BCR appears to be non- at large post-peak displacements compared to
linear since an increase in fiber concentration from unreinforced soil.
0% to 0.2% results in a greater increase in BCR
compared to an increase in fiber concentration from 5 ACKNOWLEDGEMENTS
0.2% to 0.45% or 0.85%. From Figures 3 and 4 and
Table 3, it is apparent that for both fiber lengths, the The authors would like to express their gratitude to
ultimate bearing capacity is reached at greater Mr David Harker for his assistance in conduction of
settlement ratios in the case of fiber reinforced soil the centrifuge tests and fabrication of equipment.
231
6 REFERENCES behavior of a fiber-reinforced sand bed. International
Journal of Geotechnical Engineering, 5 (3): 353- 360.
Al-Refeai, T.O. 1991. Behaviour of granular soils
reinforced with discrete randomly oriented inclusions.
Kumar, A. and Kaur, A. 2012. Model tests of square
Geotextiles and Geomembranes, 10 (4): 319-333.
footing resting on fiber reinforced sand bed.
Geosynthetics International, 19 (5): 385-392.
Consoli, N. C., Casagrande, M. D. T., Prietto, P. D. M. &
Thome, A. 2003a. Plate load test on fibre-reinforced
Maher, M. H., and Gray, D.H. 1990. Static Response of
soil. Journal of Geotechnical and Geoenvironmental
Sands Reinforced with Randomly Distributed Fibers.
Engineering, ASCE 129 (10): 951–955.
Journal of Geotechnical Engineering 116 (11): 1661–
1677.
Consoli, N. C., Casagrande, M. D. T., Thome, A. Rosa,
F.D. and Fahey, M. 2009. Effect of Relative Density
Maher, M. H., and Y. C. Ho. 1993. Behaviour of Fiber
On Plate Loading Tests on Fibre Reinforced Sand.
Reinforced Cemented Sand under Static and Cyclic
Geotechnique 59 (5): 471–476.
Loads. Geotechnical Testing Journal 16 (3): 330–338.
Fatani, M.N., Bauer, G.E. and Al-Joulani N. 1991.
Micha1owski, R.L. and Zhao, A. 1996. Failure of fiber-
Reinforcing soil with aligned and randomly oriented
reinforced granular soils. Journal of Geotechnical
metallic fibers. Geotechnical Testing Journal, 14 (1):
Engineering, ASCE 122 (3): 226–234.
78-87.
Michalowski, R.L. and Cermak, J. 2002. Strength
Freitag, D.R. 1986. Soil randomly reinforced with fibers.
anisotropy of fiber-reinforced sand. Computers and
Journal of Geotechnical Engineering, ASCE, 112 (8):
Geotechnics, 29: 279–299.
823- 826.
Palmeira, E.M. and Milligan, G.W.E. 1989. Large scale
Gray, D.H. and AI-Refeai, T. 1986. Behavior of fabric-
direct shear tests on reinforced soil. Soils and
versus fiber-reinforced sand. Journal of Geotechnical
foundations, 29 (1): 18–30.
Engineering, ASCE 112 (8): 804-820.
Ranjan, G., Vasan, R. M. and Charan, H. D. 1994.
Gray, D. H., and Ohashi, H. 1983. Mechanics of Fiber
Behaviour of Plastic-Fibre-Reinforced Sand.
Reinforcement in Sand. Journal of Geotechnical
Geotextiles and Geomembranes, 13 (8): 555–565.
Engineering 109 (3): 335–353.
Yetimoglu, T. and Salbas, O. 2003. A study on shear
Jewell, R.A. and Wroth, C.P. 1987. Direct shear tests on
strength of sands reinforced with randomly distributed
reinforced sand. Geotechnique 37 (1): 53–68.
discrete fibers. Geotextiles and Geomembranes, 21 (2):
103-110.
Kumar, A., Bhatia, R., and Walia, B.S. 2011. An
experimental study on the load vertical settlement
232
5. Offshore geotechnics
Centrifuge tests of monopile foundations under one-way cyclic lateral loading

Kunkun Cheng1, Charles M. Heron1, Luke J. Prendergast1, Alec M. Marshall1
1
Corresponding author: Charles M. Heron (charles.heron@nottingham.ac.uk)
ABSTRACT: Monopile foundations are commonly used for offshore wind turbines due to their ease of
construction and comparatively low cost. These foundations are subjected to various loads from wind,
currents, waves and superstructure self-weight, resulting in a combination of cyclic lateral loads and vertical
dead loads. To ensure an offshore wind turbine meets serviceability requirements throughout the entire life
cycle, it is important to understand the behaviour of monopile foundations under this cyclic lateral loading.
A programme of centrifuge testing is ongoing at the University of Nottingham to investigate the pile-soil
interaction of monopiles subjected to large numbers of lateral cyclic loads. A newly developed load control
apparatus capable of applying both one-way and two-way cyclic lateral loads to a model pile is employed
to investigate foundation response under a large number of load cycles (greater than 50,000) and a sample
of the analyses is presented in this paper. The accumulation of lateral displacement and the change in pile-
soil interaction secant stiffness is discussed. The results may help enhance the understanding of monopile
foundation response to large numbers of cyclic lateral loads.
Keywords: Centrifuge; Monopile Foundations; Cyclic Loading.
1 INTRODUCTION O’Kelly (2017) conducted a series of 1-g cyclic

Wind energy has attracted growing interest due to loading experiments with 6000 load cycles on a
increasing global energy demands, and the rigid pile instrumented with 8 strain gauges to study
requirement to adopt renewable energy systems, the effect of cyclic loading characteristics and the
leading to an emphasis on the development of evolution of lateral soil resistance profiles. It was
offshore wind turbines (OWTs). Monopile suggested that a logarithmic model between number
foundations have been widely adopted for of cycles and pile rotation yielded the best fit to the
supporting OWTs due to their relative ease of experimental data. In terms of lateral soil resistance,
construction and installation. OWT foundations are the authors suggested bending moment data should
subjected to cyclic loading from a large number of be fitted with a fourth order polynomial function
load cycles over their lifetime due to a combination and highlighted the sensitivity of results to the
of wind, currents and waves. However, the existing adopted polynomial function. The above small-
design codes for OWT foundations (API 2007 and scale 1-g tests commonly employed sands with low
DNV 2016) are based on a limited number of load values of relative density in an attempt to represent
cycles, which results in uncertainty about the safety the behavior of full-scale systems with soil of a
and projected lifespan of OWTs. higher density, however this approach does impose
A variety of research has been conducted to uncertainties caused by the stress dissimilarity
investigate the response of monopile foundations between the model and full-scale scenarios.
under cyclic loading. LeBlanc et al. (2010) An alternative to 1-g testing is geotechnical
performed a series of tests on stiff piles subjected to centrifuge modelling, wherein stress similarity
tens of thousands of lateral load cycles in drained between a small-scale model and a full-scale
sand at 1-gravity (1-g) and presented a model for prototype is achieved by increasing gravitational
predicting the pile rotation and pile stiffness under acceleration within the model, thereby achieving a
long-term cyclic loading. The model took into better simulation of the full-scale behavior of soil-
account the number of cycles, characteristics of pile systems. Klinkvort and Hededal (2012)
cyclic loading and sand relative density. Albiker et performed 12 centrifuge tests on monopiles in sand
al. (2017) also carried out 1-g model tests on stiff to investigate the effects of cyclic lateral loading
piles to investigate the soil relative density and characteristics on the change of pile stiffness and
characteristics of cyclic loading on the accumulation of pile head displacements. Empirical
accumulation of pile displacement, with results models relating the characteristics of cyclic loading
generally being in agreement with those from to pile head displacement and cyclic secant stiffness
LeBlanc et al. (2010). However, it was suggested were presented. It was suggested that one-way
that the sand relative density has limited influence cyclic loading is more critical than two-way
on the displacement accumulation. Arshad and loading, which contradicts LeBlanc et al. (2010).
235
Bayton et al. (2018) introduced a new method to separate stepper motor and gearbox system. In order
predict the rotation of a pile under cyclic lateral to apply cyclic loads to the pile, the AKM motor is
loading. The method predicts pile rotation operated to drive the circular disc in a rotary motion
diagrammatically using a cyclic accumulation and hence generate a periodic movement of the
contour diagram established based on experimental mass block. This movement of the mass block
centrifuge tests. The diagram is essentially a results in a varying force which is transferred to the
database of experimental test results, which were pile though the attached cable. The difference
based on a specific system (i.e. one type of sand, between forces in the cables attached to the dead
relative density, pile and load eccentricity). weight and the loading system provides the resultant
In terms of loading system, mechanical gear force applied to the pile. The movement range of the
driven devices have been widely adopted to impose mass block is determined by the diameter of the
cyclic loads on piles in 1-g tests (LeBlanc et al., circular disc. The centre point of movement of the
2010; Arshad and O’Kelly, 2017; Albiker et al., mass block is determined by the location of the
2017). These loading systems have the ability to AKM motor gantry, which is adjusted by the stepper
apply large numbers of cyclic loads with varying motor; this location controls the “bias” of the
load magnitude and frequency. Actuator loading applied pile load (i.e. varying from 1-way to 2-way
systems including servo and pneumatic systems loading).
have been deployed in centrifuge tests (Li et
al.,2010; Bayton et al., 2018). However, these
loading systems cannot achieve high loading (a)
frequencies at the load levels needed to replicate
full-scale OWT monopiles.
This paper describes the development of a new
mechanical system for centrifuge testing that is
capable of applying the large number of lateral
loading cycles needed to replicate the long term life-
cycle response of OWT monopiles. The paper
provides a description of the experimental
components, including a model monopile
instrumented with fibre Bragg grating (FBG) strain
sensors, the associated control systems, and results
from an experiment to illustrate the tested soil-pile
interaction behavior.
2 KIT DEVELOPMENT
2.1. General layout of load-control cyclic
loading system (b)
Figure 1 (a) and (b) show the components of the
newly developed system at the University of
Nottingham Centre for Geomechanics (NCG) used
for applying load-control cyclic loads to a model
pile. The pile and sand are contained in a steel tank
with an inner diameter of 0.49m and overall depth
of 0.5m. A cap is rigidly attached to the top of the
pile and connected by steel cables and pulleys to a
dead weight one side of the pile and the loading
system on the other side. The loading system is
comprised of a mass block sitting on a pin-
supported plate equipped with a carriage system and
an AKM 32E motor in conjunction with a 21.5:1
gearbox. The movement of the mass block is
controlled by the rotation of the AKM motor
through a circular disc and connection rod. The
AKM motor and gearbox are secured on a gantry Figure 1. Newly developed load-control cyclic loading system
and a conveyance system which is controlled by a for piles: (a) schematic drawing, and (b) image.
236
2.2. Loading principle 𝑀𝑀1 −𝑀𝑀𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝
𝑑𝑑2,𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 = � − 1� 𝑑𝑑1 (5)
𝑀𝑀2
The motivation of the system development was for
an experimental methodology that could generate For specific values of M1 and M2, d2,balance is fixed.
load-controlled cyclic loads in the centrifuge in a Through the movement of the mass block M2
way that would allow relatively high loading relative to this position, the kit can apply both one-
frequencies to be achieved alongside load way and two-way cyclic loading. The amplitude of
magnitudes that are representative of OWT ΔF can be controlled by changing the weight of the
monopile foundations. The developed system is mass block. Hence, the kit is capable of producing
capable of (i) applying relatively large loads by cyclic loads with different characteristics.
changing the weights of the mass block and dead
weight, and (ii) varying the loading frequency by
2.3. Control program
increasing the AKM motor speed (current
motor/gearbox system can reach 4Hz pile loading; The control program, written in LabVIEW,
higher frequencies can be achieved by altering the regulates the position of the AKM motor gantry.
motor/gearbox). Figure 2 shows a schematic of the This control program operates in two different
loading system. Since pile head displacement stages. The first stage ensures that forces F1 and F2
during tests are small, the plate containing mass M2 on each side of the pile remain balanced during
can be considered horizontal, and it is assumed that changes in gravity level (g-level) as the centrifuge
the mass block is at a constant g-level (which varies is spun up (any imbalance of these forces could
with distance from the central axis of the cause movement of the pile during spin-up,
centrifuge). Force F2 resulting from the mass block disturbing the soil sample and adversely influencing
and force F1 from the dead weight are calculated by test results). The stepper motor control system reads
Eq. 1-3; the resultant force on the pile ∆F is given in values of F1 and F2 and automatically adjusts the
by Eq. 4. position of the gantry to achieve a balance in the
forces (within a tolerance of 30N). The second stage
occurs when the target g-level is reached and the
system is ready for cyclic loading of the pile under
a prescribed loading condition (e.g. one-way or two-
way loading). In this stage, the stepper motor is
temporarily stopped and the AKM motor is set to
spin at a given speed. The stepper motor control is
then enabled to adjust the position of the AKM
motor gantry to achieve the desired cyclic loading
characteristic. To apply cyclic loading, the mass
block is always first positioned to the balance point
determined by M1 and M2. All the instruments were
logged at 250Hz in the test.
Figure 2. Schematic of pile loading system.
3 CASE STUDY: ONE-WAY CYCLIC
𝐹𝐹1 = 𝑁𝑁𝑀𝑀1 𝑔𝑔 (1) LOADING TEST
𝐹𝐹2 𝑑𝑑1 = 𝑁𝑁𝑀𝑀2 𝑔𝑔(𝑑𝑑1 + 𝑑𝑑2 ) + 𝑁𝑁𝑀𝑀𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔𝑑𝑑1 (2) A one-way cyclic loading test was conducted to
evaluate the performance of the new system. The
𝑑𝑑2 loading frequency was specified to be 3Hz with a
𝐹𝐹2 = 𝑁𝑁𝑀𝑀2 𝑔𝑔 �1 + �+ 𝑁𝑁𝑀𝑀𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝 𝑔𝑔 (3) load amplitude of 500N (2MN at prototype scale
𝑑𝑑1
given the applied scaling factor of N=63). A total of
∆𝐹𝐹 = 𝐹𝐹2 − 𝐹𝐹1 (4) 70,000 load cycles were applied to the pile.
where N is the centrifuge scaling factor (i.e. N times 3.1. Model pile
gravity 𝑔𝑔), Mplate is the mass of the plate supporting The model pile was made of aluminium with
mass M2, d1 is the distance from the pin support to properties shown in Table 1. The bottom of the pile
the cable attaching the plate to the pile (constant), d2 was sealed with a cap, making it a closed-ended
is the horizontal distance from the centre of mass M2 hollow pile. A mass of 2kg was mounted on the top
to the plate-cable attachment position (varies), and of the pile to simulate the dead-weight of a
mass M1 is the dead weight. The balance position superstructure. The pile was instrumented with 7
(i.e. ΔF = 0) of the mass block is calculated by Eq. evenly distributed FBG strain sensors from 20mm
5:
237
to 320mm at 50mm intervals relative to the pile different heights in order to evaluate pile rotation.
bottom. Laser 3 was mounted above the pile to monitor the
vertical displacement of the pile during testing.
Table 1. Properties of model pile.
Parameter Unit Pile
Pile length, L mm 500
Wall thickness, t mm 3
Outer diameter, d mm 60
Young's modulus, E, GPa 69
Moment of inertia, I, m4 2.2E-7
Flexural stiffness, EI MNm2 0.015
Cross section area, A m2 0.5E-3
3.2. Soil preparation

The sand used for this test was Congleton silica sand
(HST95) from Bent Farm in Congleton, Cheshire
(UK). The properties of the sand were evaluated by
the Norwegian Geotechnical Institute (NGI) and are
presented in Table 2. In this study, the ratio between
the pile diameter and the average grain size is 333,
which avoids scaling issues according to Garnier et
al. (2007). The air pluviation method was utilized to
prepare the sand sample. Sand was first air-
pluviated into the centrifuge container to form a
150mm thick layer. The model pile was then placed
on the sand and secured by a temporary frame. After
the placement of the pile, the sand was poured into Figure 3. Experimental setup.
the container to achieve a 300mm embedded depth.
This process represents a 'wished-in-place' closed- 4 DEMONSTRATION OF SYSTEM
ended pile and avoids the disturbance caused by PERFORMANCE
installation effects. All tests were performed in dry
dense sand with a relative density of 90%. This section provides an overview of the data
obtained from the cyclic loading tests. All results
are provided in model scale unless otherwise stated.
Table 2. Properties of HST95 Congleton sand.
Parameter Unit Sand
4.1 Spin-up and stabilisation cycles
Specific gravity of
- 2.63 Figure 4 shows forces F1 and F2 measured during
particles, Gs
Minimum void ratio, emin - 0.494 centrifuge spin-up to 63g and the subsequent 3
stabilisation cycles (from 63g-10g-63g) prior to
Maximum void ratio, emax - 0.765
cyclic loading of the pile. The stabilisation cycles
Average grain size, d50 mm 0.17
mm 1.9 are undertaken to encourage a uniform stress
Coefficient of uniformity
distribution within the soil body and improve
repeatability between tests. The difference between
3.3 Experimental setup F1 and F2 (i.e. the resultant force ∆𝐹𝐹 applied to the
Figure 3 shows a sketch of the test setup. The load pile) was generally controlled to less than 25N,
eccentricity and embedment depth for the test were preventing any significant disturbance to the soil.
200mm and 300mm, respectively. Laser sensors Upon reaching 40g during the initial spin-up, there
(Baumer OADM 20I6460/S14F) were utilised to was a peak of about 60N, which is believed to be
measure pile displacements. The sensing range of due to some reorientation of the sliding mass system
the lasers can be adjusted to a narrower range to caused by stretching of the cables; the force balance
improve accuracy. As shown in Figure 6, laser 1 and control system quickly corrected for this.
laser 2 were located on the side of the pile at
238
Figure 4. Measured forces F1 and F2 during centrifuge spin-up and stabilisation cycles.
800
n=1 2 3 4 n=15,000+
n=22 23 24 25 26
600 1 2 3 4 5
400
200
Force on the pile (N)
-200
0 2 4 6 13 13.5 14 14.5 15 5006 5007 5008
Time (s) Time (s) Time (s)
(a) (b) (c)

Figure 5. Resultant force on the pile for cycles: (a) n=1, 2, 3 and 4; (b) n=22, 23, 24, 25 and 26; (c) n=15001, 15002, 15003, 15004
and 15005.
4.2 Cyclic loading cycle number n=1, 2, 3, 10 and 100 where the pile
Figure 5 presents the resultant force applied to the is subjected to forces ranging from 0 to 600N.
pile for different cycle numbers, given by n. The Figure 6(b) highlights representative cycles when
specified loading frequency of 3Hz was reached the loads stabilized in the 0 to 500N range. From
after 22 cycles. The system requires a certain Figure 6(a), it is clear that residual displacement
number of cycles to converge on the specified load after each cycle accumulates but that maximum pile
amplitude (500N for this test); the load overshot to displacement remains approximately constant.
600N and then steadily decreased to 500N after After 400 cycles the pile’s residual displacement
cycle number 500, where it was subsequently fluctuates around 0.32mm; the maximum
maintained. displacement decreases steadily with subsequent
load number up to n= 70,000, which suggests the
4.3 Load-displacement curve sand experiences densification. Figure 7(a) and (b)
show the moment, which is the product of the lateral
Loads versus pile horizontal displacement curves force and the load eccentricity, against pile rotation
are shown in Figure 6 for different cycle numbers. for cycles before and after n=400, respectively. The
Figure 6(a) presents load-displacement data for pile rotation is calculated using readings from laser
239
1 and laser 2 and the pile above the sand is assumed
to be undeformed.
800 140
(a)
120
600
100
80
400
60
Moment (Nm)
200 40
n=1
n=2 n=1
20
n=2
n=10
0 n=10
n=100
0
n=100
n=400
n=400
-20
-200
-1 0 1 2 3 4
0 0.5 1
Pile rotation (rad) -3

10
Pile head displacement (mm)
(a) (a)
120
600
(b)
100
500
80
400
60
300
Moment (Nm)
n=400
40
n=400
200 n=800
n=800
n=2000
n=2000
20 n=10000
100 n=10000
n=20000
n=20000
n=70000
n=70000
0
0
0 1 2 3 4
0.3 0.4 0.5 0.6 0.7

Pile rotation (rad) -3
10
(b) (b)
Figure 6. Force versus pile horizontal displacement for
Figure 7. Moment versus pile rotation for cycles: (a) n=1, 2,
cycles: (a) n=1, 2, 10, 100 and 400; (b) n=400, 800, 2000,
10, 100 and 400, (b) n=400, 800, 2000, 10000, 20000, and
10000, 20000, and 70000.
70000.
240
4.4 Stiffness variation 4.5 Loading rate effect
Figure 8 shows the secant stiffness against number To evaluate the potential effect of loading rate, after
of cycles. Stiffness was calculated by curve-fitting completing 70,000 cycles, the test was paused and
the load and displacement hysteresis loop for each the loading frequency was varied between 4Hz and
cycle using linear regression. This was undertaken 2Hz. Figure 9(a) presents the stiffness variation for
to minimise the effect of signal noise in the data the different loading frequencies and shows that the
which has more of an effect when selecting the stiffness remained relatively constant across the
peaks/troughs of individual hysteresis loops. tested range of frequencies. As shown in
Figure 8 shows that stiffness increased rapidly at the Figure 9(b), the loading frequency had a limited
beginning of the test from 600N/mm at the first effect on the obtained load-displacement loop.
cycle to 1150N/mm at the 150th cycle. Stiffness 1260
subsequently increased slowly with additional cycle

number, with some fluctuations, which is consistent
with observations by Klinkvort and Hededal (2012). 1250
Figure 8 also presents the enclosed area of moment-

rotation curves for each cycle. The enclosed area is 1240
representative of the energy dissipated by the pile-

soil system. As shown in Figure 8, the enclosed area
continues to decrease with increasing load cycle.
1230
0.08 1300
Secant stiffness (N/mm)
1220
cycle
1200
Stiffness
0.06
1210
1100
3Hz 4Hz 2Hz

0.04
1000
1200
0 200 400 600
900
0.02 Cycle number
Stiffness (N/mm)
800
600
(a)
0
Area enclosed within moment-
Enclosed area of loop 700
500
-0.02 600
0 2 4 6 8
400
Cycle number 4
10
(a) 300
0.08 1300
Stiffness
200
cycle
1200
0.06
100
1100
3Hz
0.04
1000 0 4Hz
2Hz
900
0.02 -100
Stiffness (N/mm)
0.2 0.4 0.6 0.8

800

0
Area enclosed within moment-
Enclosed area of loop
(b)
700
Figure 9. (a) Stiffness for different loading frequency; (b)

force versus pile head displacement for different loading
-0.02 600
0 5
10 10
frequency.
Cycle number
(b) 5 CONCLUSIONS
Figure 8. Secant stiffness and area enclosed within moment- This paper described a newly developed centrifuge
rotation cycle versus cycle number: (a) cycle number in linear testing apparatus for applying large numbers of
scale; (b) cycle number in log scale.
cyclic loads and presented preliminary results
241
related to the response of OWT monopile Working Stress Design. Api Recommended Practice,
foundations. Results were provided to demonstrate 24-WSD (December 2000):242.
the successful operation of the experimental system,
including balancing the resultant force applied to Arshad, M. and O’Kelly, B. C. 2017. Model Studies on
Monopile Behavior under Long-Term Repeated
the pile during centrifuge spin-up to limit
Lateral Loading. International Journal of
disturbance to the pile, and imposing a specified Geomechanics, 17(1).
loading rate and amplitude (3Hz and 500N were
demonstrated in this paper). The system can be Bayton, S. M., Black, J. A., and Klinkvort, R. T. 2018.
adjusted to achieve different cyclic loading Centrifuge modelling of long term cyclic lateral
characteristics, ranging from one-way to two-way loading on monopiles. Physical Modelling in
loading and with different load amplitudes. The Geotechnics, 2:689–694.
current system has ability to apply a relatively large
number of load cycles at a frequency of up to 4Hz; DNV 2016. DNVGL-ST-0126: Support structures for
modifications to the motor and gearbox will allow wind turbines. Dnv Gl As.
future applications at higher frequencies. Data from
Klinkvort, R. T. and Hededal, O. 2012. Lateral response
a one-way cyclic load test were provided and of monopile supporting an offshore wind turbine.
showed a general increase in secant stiffness with Proceedings of the Institution of Civil Engineers -
load cycle, with some fluctuation, suggesting the Geotechnical Engineering, 166(2):147–158.
sand was continually densifying with load cycle. In
addition, it was shown that variation of loading LeBlanc, C., Byrne, B. W., and Houlsby, G. T. 2010.
frequency between 2Hz and 4Hz had a limited effect Response of stiff piles in sand to long-term cyclic
on the load-displacement curve and secant stiffness lateral loading. Gotechnique, 60(2):79–90.
evolution in the dry sand tests conducted.
Li, Z., Haigh, S. K., and Bolton, M. D. 2010. Centrifuge
modelling of mono-pile under cyclic lateral loads. 7th
6 REFERENCE International Conference on Physical Modelling in
Albiker, J., Arshad, M., Frick, D., and Flindt, F. 2017. 1g Geotechnics, 2:965–970.
Model Tests on the Displacement Accumulation of
Large-Diameter Piles Under Cyclic Lateral Loading. Garnier, J., Gaudin, C., Springman, S. M., Culligan, P. J.,
Geotechnical Testing Journal, 40(2):173–184. Goodings, D., Konig, D., and Thorel, L. 2007.
Catalogue of scaling laws and similitude questions in
API 2007. Recommended Practice for Planning, geotechnical centrifuge modelling. International
Designing and Constructing Fixed Offshore Platforms Journal of Physical Modelling in Geotechnics, 7(3), 1–
23.
242
Use of optical fibres to measure pore water pressure development during impact
pile diving: a geotechnical centrifuge study
Amin Askarinejad 1, Tristan O. Quinten 1,2, Mario Alvarez Grima3, Cornelis van ‘t Hof 3, Ken
Gavin1
1
Faculty of Civil Engineering and Geosciences, Delft University of Technology, the Netherlands
2
DEME group, Belgium
3
Royal IHC-MTI, Delft, the Netherlands
Corresponding author: Amin Askarinejad (A.Askarinejad@tudelft.nl)
ABSTRACT: Pile driveability is affected by, amongst other factors, the pore pressure changes in the vicinity
of the pile. However, accurate quantification of the generated pore pressure in the soil layer during impact
pile driving is a challenging task, especially in the small-scale models. An electro-mechanical miniature
impact pile driver has been designed and constructed for the geo-centrifuge at TU Delft. This hammer has
been used to study the pore pressure generations at various radial distances from the pile. An optical fibre
capable of measuring pore pressure changes was installed on the outer skin of the model pile to monitor the
pore pressure generation at the pile-soil interface. This data is complemented by standard miniature pore
pressure transducers embedded in the soil mass that allow the radial distribution of excess pore pressure to
be measured. The details of instrumentation, installation, performance and the results of this novel sensor
are presented and discussed in this paper.
Keywords: Pore Water Pressure; Impact Pile Driving; Optical Fibres; Geotechnical Centrifuge; Small-Scale Physical
Model.
1 INTRODUCTION based on the specifications of IHC Hydro-hammers

The vast majority of substructures for offshore wind S200, S280 and S500 with a blow rate of 45 blows
turbines require the installation of one (monopile) per minute. Considering the scaling requirements
or multiple open-ended steel piles. These offshore for the energy flux of the hammers an electro-
piles are mostly installed using costly operations mechanical lift system was adapted which is
involving large hydraulic hammers. Prior to composed of a 1.2 kW electric brushless motor, a
installation, drivability analyses are commissioned flywheel that picks up the ram mass at its lowest
which determine the required hammer capacity. point and releases it at a predefined point, and a set
Several full-scale experiments have demonstrated of gears. The gear system ensures that the motor can
that modification of driving parameters, operate at maximum efficiency while keeping the
specifically: (I) hammer weight; (II) driving flywheel at the correct speed for pile driving (Van
frequency; (III) falling height/impact velocity; can Zeben et al., 2018, Azua et al., 2018 and Azua et al.,
improve pile installation time. Not all of the 2019).
physical processes leading to the improved impact A modular system was utilised for the ram mass
pile driving efficiencies are yet fully understood. to provide flexibility in augments of the required
However, it is hypothesised that the generation and mass. The ram mass is composed of a central
dissipation of pore water pressure in the vicinity of bearing, which facilitates a smooth fall of the mass
the pile has potentially a major effect on the on to the pile head, a top cap and a connecting arm
driveability. A miniature pile driving hammer has to be picked up by the fly wheel (Figure 1). The
been developed to conduct impact pile driving on mass of the top cap can be altered and makes it
small scale models in the centrifuge in this study. possible to change the blow energy in between tests
The details of the hammer and instrumentation are and also accommodate different masses for future
discussed. Moreover, the results of pore pressure research without the necessity of major alternations.
measurements at various radial distances from the The fall height is determined by several factors and
model pile are presented. can be varied from 25 to 55 mm resulting in an
impact speed on the anvil up to 6.3 m/s at 50g
.
2 MINIATURE IMPACT PILE DRIVER
The miniature hammer developed at the Geo-
Engineering section of TU Delft was designed
243
Figure 1 Test setup, model pile and capillary tube
Figure 2 Pressure optical fibres (after Pinet 2009)
3 OPTICAL FIBRES TO MEASURE subsequently closing them off at the top with a
PORE PRESSURE water-resistant silicon adhesive, a closed chamber is
Optical fibres were installed at the interface of the obtained in which pressure fluctuations along the
pile and the soil and were housed by capillary tubes shaft can be monitored. At the bottom of the
(with inner diameter of 0.8 mm). The capillary tubes capillary, a small slit (behind which a filter is
were embedded into the outer periphery of the installed), allows for the transmittance of pressure
model piles. The pore pressure sensor is produced from the pore fluid at the interface to the sensor. An
by FISO and is made of a Micro-Opto Mechanical advantage of this installation method is that the
System (MOMS) bonded to the tip of a fibre optic sensors follow the movement of the pile. In total
(Pinet 2009). This system is composed of a three optical fibre pressure sensors (OFPS) were
deformable membrane (mechanical part) which is used, which were installed at different distances
assembled over a cylindrical cavity. The membrane with respect to the tip of the pile. The results of the
deflects towards the base of the cavity when OFPS2 installed at a distance of 0.7D from the pile
external fluid pressure is applied and this results in tip are presented in this paper (D = 42 mm: pile
a decrease in the length of the sensing cavity (Figure diameter) the initial embedment of the sensor before
2). By filling the capillaries with silicon oil, and pile driving is about 2D from the soil surface.
244
4 STATIC PORE PRESSURE SENSORS acceleration are presented in Figure 3. Both
To capture the development of excess pore fluid conventional and optical fibre pressure sensors were
pressure during the driving operation, static pore installed at similar depth with respect to the soil
fluid pressure sensors (SPS) are installed in the soil surface (~2D). Both sensors measure an increase in
layer at various radial distances from the pile and the hydro-static pressure due to the increase of
three different depths from the soil surface. The centrifugal acceleration. A very good agreement is
results of deepest sensor which is installed at about seen between the two measurements. Moreover, the
2D from the soil surface and a radial distance of decrease in the centrifugal acceleration and its effect
0.3D from the pile are presented in this paper. on the hydrostatic pore pressures is very well
captured by both sensors.
5 SAMPLE PREPARATION The optical fibre sensor shows higher values of
pore pressure after installation compared to pre-
Circular strongboxes with diameter of 315 mm and installation measurements. This increase in hydro-
an inner effective height of 175 mm were used. The static pore pressure indicates the added embedment
soil samples were prepared using submerged of the pile. Whereas, the static PPT measures equal
pluviation and subsequent densification to a relative values before and long after the hammering. The
density of 80%. Fine uniform silica sand with a D50 area confined by the rectangular zone in Figure 3
of 110 micrometers and internal friction angle of 34 shows the time span of pile driving and the details
degrees was used (Table 1). In order to scale the of the measurements in terms of excess pore
pore pressure generation and dissipation correctly in pressure (EPP) development are shown in Figure 4.
the centrifuge model the permeability of the soil has Optical sensor which is installed at the interface of
been reduced (Askarinejad et al., 2014) by the dense sand and the driven pile has measured the
saturating the samples with a viscous fluid with accumulation of negative pressures over a set of
dynamic viscosity of 50 cSt. consecutive blows. At its peak, the under-pressure
is about 5 kPa, equivalent to a decrease of about
Table 1 Properties of silica sand (De Jager et al. 2017,
Maghsoudloo et al. 2017) 10% with respect to hydrostatic conditions. Hence,
Parameter Silica sand it is cautiously stated that for dense sand samples as
Maximum void ratio 0.86 used for this research, the accumulation of positive
Minimum void ratio 0.625 excess pore water pressures in the vicinity of the
cu 1.38 pile is non-existent. Hence, a state of soil
D50 0.110 mm liquefaction seems to be ruled out in this zone.
cc 1.05 However, the limited resolution of data logging rate
Friction angle (cv) 34° inhibits the further exploration pore pressure
Permeability 8.41E-4 m/sec generation at the soil-pile interface. On the contrary,
Specific Gravity (-) 2.65 the pore pressure transducer installed with a radial
distance of about 0.3D measures 30% increase in
6 TEST RESULTS the pore pressure with respect to the initial hydro-
The results of the pore pressure measurements as a static values pre-installation.
function of the model time and centrifugal
60 60
Centrifuge accelaration (*g)
Static PPT
50 50
Pore pressure (kPa)
OFPS
40 Acceleration 40
30 30
20 20
10 10
0 0
0 100 200 300 400 500 600 700
Model time (s)
Figure 3 Results of pore pressure measurements as a function of model time. The rectangle shows the time span of pile driving.
245
15
Static PPT
10
Excess Pore pressure (kPa)
OF EPP
5
0
375 380 385 390 395 400 405 410
-5
-10
Model time (s)
Figure 4 Development of excess pore pressure in the vicinity of the pile.
7 CONCLUSIONS viscous pore fluid. Canadian Geotechnical

Journal, 52(6), pp.708-720
In this paper, the usage of optical fibre pressure
sensors for centrifuge models of offshore mono-pile Azúa-Gonzáleza, C.X., van Zebena, J.C.B., Grimab,
driving is described. From practical point of view, it M.A., van't Hof, C. and Askarinejada, A., 2019.
could be concluded that the largest challenge lies in Dynamic FE analysis of Soft Boundary (SB) effects
the area of sensor fixation and the saturation of the on impact pile driving response in centrifuge tests.
capillary tubes, which might tend to leak oil during Proceedings of the XVII ECSMGE-2019. doi:
the experiments. Moreover, it was observed that it 10.32075/17ECSMGE-2019-0821
would be beneficial to increase the measurement De Jager, R.R., Maghsoudloo, A., Askarinejad, A. and
frequency of these sensors. However, it should be Molenkamp, F., 2017. Preliminary results of
instrumented laboratory flow slides. Procedia
noted that the intrinsic data logging rate of the
Engineering, 175, pp.212-219
system used in this study is limited to 125 Hz. The
results of pore pressure measurements yield Maghsoudloo, A., Galavi, V., Hicks, M.A. and
valuable insights with regard to processes during Askarinejad, A. 2017, September. Finite element
impact pile driving. Slight negative development of simulation of static liquefaction of submerged sand
excess pore pressures was detected at the interface slopes using a multi-laminate model. In 19th
of the dense sand and the pile. Whereas, at larger International Conference on Soil Mechanics and
radial distances from the pile, the global trend of Geotechnical Engineering (Vol. 2, pp. 801-804).
measurements demonstrates a rather uniform
increase of the pore pressure with respect to Pinet, É., 2009. Fabry-Pérot fiber-optic sensors for
hydrostatic conditions in the soil body surrounding physical parameters measurement in challenging
conditions. Journal of sensors (2).
the pile. https://doi.org/10.1155/2009/720980
8 REFERENCES Van Zeben, J.C.B., Azúa-González, C.X., Alvarez

Grima, M., van’t Hof, C. and Askarinejad, A. 2018,
Askarinejad, A., Beck, A. and Springman, S.M., 2015.
July. Design and performance of an electro-
Scaling law of static liquefaction mechanism in
mechanical pile driving hammer for geo-centrifuge.
geocentrifuge and corresponding hydromechanical
In Proc. Int. Conf. Phys. Mod. Geotech. (ICPMG
characterization of an unsaturated silty sand having a
2018) (pp. 469-473).
246
The modelling of piles under multidirectional loading

Christophe Gaudin1, Conleth O’Loughlin, Manuel Herduin
1
Centre for Offshore Foundation Systems, The University of Western Australia, Australia
Corresponding author: Christophe Gaudin (christophe.gaudin@uwa.edu.au)
ABSTRACT: Offshore foundations are typically designed to resist mooring loads coming from a single
direction. This paper provides some insights into the geotechnical challenges associated with anchor sharing
(i.e. where mooring loads are coming from multiple directions to a single anchor) for Offshore Renewable
Energy (ORE) applications. This is relevant to arrays of floating renewable energy devices (waves or wind),
when multiple devices are anchored to a single foundation to generate economies of scale. The paper
presents the modelling methodology and the results of centrifuge tests investigating the behaviour and
performance of a rigid pile subjected to cyclic multidirectional loading regimes, as identified from
hydrodynamics analyses. Results highlight the importance of the magnitude of the vertical resultant of the
different loadings, as well as the variation of direction of the horizontal resultant.
Keywords: Pile, Centrifuge Modelling, Multidirectional Loading, Offshore Renewables
1 BACKGROUND installed cost of an offshore wind turbine (Willow

Offshore renewable energy is rapidly evolving as & Valpy, 2011), and up to 30% of the total installed
one of the highest potential sources of clean energy. cost of a wave energy converter (Martinelli et al.,
Fixed offshore wind has experienced rapid 2012). This is one order of magnitude higher than
development (+30% per year between 2010 and for oil and gas structures (Kost et al., 2013), and
2018, with a total capacity of 19 GW, IEA 2019), contributes significantly to the high levelised cost of
and has absorbed most of the research attention over offshore wind and wave energy. For marine
the last decade. More recently, floating technology renewable energy to be commercially affordable
has been increasingly considered to support the and competitive, a significant paradigm shift is
growth in demand for renewable energy. This the required both in foundation technology and in
case for wind, to access higher and more consistent foundation design.
wind resources, but also for wave and tidal, two Wind
technologies still in their infancies but with very turbine
large potential. v v v v v Mooring

line
Renewable energy generation from floating Mooring
systems has been proven, and includes for example:

line
v v v v v
(i) the Hywind spar floating wind turbine, which has Suction Wave energy
Suction
been in operation in 198 m of water off the
caisson converter
v v v v v
caisson
southwest coast of Norway since 2009, (ii) the

WindFloat semi-submersible floating wind turbine
that has been tested in 40-50 m water depths off the (a) (b)
coast of Portugal since 2011, (iii) the Ocean Power Figure 1. Example array of: (a) WECs and (b) floating wind
Technology floating wave energy device, which has turbines sharing anchoring points. Regardless of the pattern of
been tested off the coasts of Hawaii, USA and the array, the number of mooring points reduces linearly with
the number of mooring lines connected to it.
Scotland in water depths of up to 30 m since 2005,
and (iv) the Perth Wave Energy Project from
One of these paradigm shifts, either for wave
Carnegie Wave Energy, with three 240 kW wave
energy converters of floating wind turbines, is the
energy converters (WECs) operating over 12
sharing of foundation between multiple devices
months offshore Garden Island in Western
when assembled in an array (Karimirad et al., 2014)
Australia.
to significantly reduce the foundation engineering
Floating technology for offshore renewables still
costs. Different array patterns can be considered as
presents several significant challenges. It requires
a function of the power output and shadowing
reliable and economical anchoring systems that can
effects between devices (see Child & Venugopal,
perform in the type of seabed sediments 2010), as presented in Figure 1. But in all cases, the
encountered on the continental shelf, where marine
number of foundations can be reduced significantly,
renewables are expected to operate. Anchoring
such that for instance, in a three mooring point
systems can contribute up to 22% of the total
247
absorber in a honeycomb pattern (Figure 1a), 20 and highlighting some critical aspects of the pile
devices can be anchored by 30 foundations, instead response.
of 60 (see Figure 2).
300 2 HYDRODYNAMIC ANALYSIS
2.1. Numerical analysis
250
The characteristics of the multidirectional loading
acting on a foundation are complex. Each load does
200 not act independently of the other, but as a function
Total number of anchors
of the pattern of the array, the direction and period

150 3
of the waves and the spacing between the floating
bodies, and the resultant of these loads follows
100 specific patterns defined by these input
characteristics.
50 Before engaging in a complex centrifuge
programme, it is necessary to identify clearly the
0 Wave
range of loading regimes that are applied on a multi-
0 20 40 60 80 100 moored foundation, to be able to identify which of
Number of floating structures these regimes are detrimental for the foundation
Figure 2. Example of the benefit of anchor sharing for arrays performance, and hence inform the optimal array
of renewable energy devices. A floating device is typically pattern and dimensions. To that end, a
anchored by three mooring points. By sharing these mooring comprehensive numerical analysis was undertaken
points between devices, it is possible to reduce significantly the with Orcaflex.
number of mooring points and hence foundations, reducing the
cost of foundation engineering (installation and infrastructure).
The benefit increases with the number of mooring lines on each
mooring points, but also the complexity of the loading applied
to the foundation.
A direct consequence of the foundation sharing

strategy is the complexity of the loading regime on
the foundation that can come from multiple
directions (Figure 2). Depending on the wave
spectra and period, and the spacing of the floating
devices, the foundation can be subjected to loads
coming from different directions that can be in
phase (resulting in alternate loading along each of
the mooring lines) or out of phase. This complex
multi-directional loading mode is fundamentally
different from typical design considerations for
floating oil and gas facilities, on which most of the
design methodologies are based.
This paper reports some of the outcomes obtained Figure 3. Case investigated via Orcaflex analysis to identify
from a research project undertaken to characterise multidirectional loading regimes on anchoring points. The
the response of rigid piles under multidirectional array consists of 6 floating bodies, with spacing of 95 m, in
mean water depth of 27 m. 6 sea states were considered,
loading. The first stage of the project entailed corresponding to field measurements made in Falmouth Bay,
defining precisely the loading regimes experienced UK, and provided by the University of Exeter. The wave
by the foundation as a function of the pattern of the direction ranged from 86° to 179°, while the wave period
array and the wave characteristics. This was ranged from 6.5 to 8.2 s and the significant wave height from
undertaken through a comprehensive hydrodynamic 0.19 to 2.31 m. The analysis provided load regimes on each
mooring line as a function of the sea state considered. Details
analysis, validated by in-situ measurements, with of the Orcaflex analysis, including all hydrodynamic
the main conclusions presented in the next section. parameters, are presented in Herduin (2019).
The second stage involved a series of centrifuge
tests modelling the identified loading regimes and The Orcaflex analysis was first validated against
investigating the response of a rigid pile embedded field data from a 3-point mooring floating structure,
in sand. This paper presents a limited number of mimicking the dynamic response of a scaled point
results, showcasing the modelling techniques used absorber in real open ocean conditions, and
provided by the University of Exeter (Herduin,
248
2019). The analysis was subsequently extended to Orcaflex analyses are demanding and time
an array of floating bodies at the sea surface, consuming. Based on the numerical analysis, an
arranged in a honeycomb pattern and attached to 3 analytical model was developed that enables rapid
mooring points. Each mooring point is connected to identification of the various loading regimes (see
2 (at the edge of the array) or 3 (inside the array) Herdouin, 2019 for full details). The analytical
catenary mooring lines (Figure 3). The array was model is a simplification of the real conditions,
subjected to 6 different sea states (established from assuming only regular waves, but it provides results
the field measurements provided by University of that are close enough to those generated by the
Exeter), which varied the direction and period of the Orcaflex analysis to be able to identify families of
wave, and its significant wave height. loading regimes. An example of output from the
An example of the results is provided in Figure 4, analytical model is presented in Figure 5 for the
which represents the time series of the load resultant three loading direction case. The model also defines
and its variation in direction. These two parameters the vertical resultant, which is an important
are important to define the loading regimes. Across parameter of the foundation behaviour as discussed
varying sea states, both the load resultant and its later in the paper.
direction may vary significantly in magnitude, or
only very slightly, with fundamentally different 3 CENTRIFUGE MODELLING
consequences on the anchor behaviour and 3.1. Models and experimental setup
performance.
Tests were performed in the 10 m diameter 240 g-
tonne beam centrifuge (Gaudin et al., 2018) at the
University of Western Australia. Sand samples were
reconstituted in a square strongbox, with breadth 1
m and depth 0.5 m.
The model piles are close-ended short rigid piles
with diameter D = 23 mm and length L = 155 mm
(L/D = 6.7). They were manufactured from a hollow
Figure 4. Example results from the Orcaflex analysis. The
figure relates to anchor 2 in Figure 3, with two loading
piece of aluminium and closed by screw caps at
directions spaced by 120°. It presents the time series of the either end. A drainage hole in the pile allowed the
force resultant FR, the magnitude of which ranges from 2.2 to pile to fill with water prior to the testing to avoid
3.2 kN, with a resultant loading direction in the horizontal unrealistic buoyancy. The fine silica sand used for
plane αR varying between 133° and 156°. the sample was glued to the outer shaft of the pile to
create a fully rough interface (Figure 6).
Commercial fine silica sand, with main
characteristics summarised in Table 1, was used to
create two soil samples using dry pluviation. The
sand was from a constant drop height to ensure that
the relative density remained constant with depth.
Pluviation was continued until a sample 150 mm
high was achieved, at which point the model piles
were suspended in the strong box from a customised
mesh, with the piles bases resting on the surface of
the sand. Pluviation was then resumed, such that the
piles were effectively pre-embedded in the sample.
The final sample height was 305 mm, with a pile
Figure 5. Example resultant load path in the horizontal plane embedment length of 155 mm (i.e. the full pile
for three loading directions spaced at 120°. The input on each length). Following pluviation, the samples were
line is established from a regular sinusoidal wave, with the
direction of the wave and the distance between floating bodies saturated with water from the base of the sample.
resulting in a phase angle between the loads on each line The two samples were characterised through in-
varying from 0 to 2π/3. The load path varies from a single line flight cone penetrometer tests (CPTs) performed
across the horizontal plane, indicating a large variation in before and after the pile tests, using a cone 7 mm in
magnitude but no variation in direction, to a circle, indicating
a large variation in direction and no variation in magnitude.
diameter. Tests were performed at 1 mm/s (under
This variation in loading regime results in different foundation drained conditions) to a target depth of 180 mm.
behaviour, as identified by the centrifuge testing campaign. Results (not presented in this paper for brevity)
indicated a good homogeneity and consistency
249
within and between samples. The relative density, DigiDAQ system developed in-house at UWA
estimated from the empirical relationship proposed (Gaudin et al, 2009).
by Jamiolkowski et al. (2003), was 82% for Sample
1 and 85% for Sample 2.
Table 1. UWA fine silica sand properties after Chow et al. 2019.
Peak and residual friction angles are established from drained
direct shear tests at a relative density of 88-98% and vertical
stresses in the range of 21-400 kPa.
Property
Sample relative density Dr 85%
Specific gravity, Gs 2.65
Maximum void ratios, emax 0.78
Minimum void ratios, emin 0.49
Average particle size, D50 0.19 mm
Peak friction angle, ϕp(DS) 40.1º
Residual friction angle, ϕr(DS) 33.1º
Singificant efforts were devoted to the design of

the centrifuge setup to enable application of loading Figure 6. Model pile and centrifuge setup. Adjustment of the
from 2 or 3 directions in the horizontal plane, and at bottom plane enables variation of the loading angle from the
varying angles from the horizontal in the vertical vertical and the direction of loading in the horizontal plane,
plane (see Figure 6). The setup consists of a while instrumentation measures the load and displacement on
each mooring line, the rotation of the pile (through
platform assembled around 2 I-beams, onto which 3 accelerometers) the horizontal displacement of the pile in two
linear electric actuators are attached vertically, directions and tis vertical displacement. The combination of the
allowing for loading along three lines connected line displacement, horizontal displacement and pile rotation
between each actuator and the pile head. A bottom gives the complete motion of the pile in the 6 degrees of
freedom.
plate located under the platform and featuring
multiple pulleys allowed the load inclination (to the
horizontal) to be vared in the range 10 to 60 degrees, The centrifuge setup is shown schematically in
and the load direction to be along three directions Figure 7 with all the relevant dimensions.
with spacing of 30, 60, 90, 120 and 180 degrees in
the horizontal plane.
Three in-line load cells, each with a measurement
range of 2 kN, were located on each loading line,
while linear displacement transducers (LDTs) were
mounted on each actuator to measure displacement
along the loading direction. A 3D-printed support
was fixed on the pile head (see Figure 4) to host two
accelerometers to measure the pile rotations in two
orthogonal directions (ωx, ωy) using the
methodology developed by Beemer et al. (2018).
Three additional LDTs were connected to a bar
sitting on top of the 3D-printed support to measure
horizontal (δx, δy) and vertical (δz) displacements.
The three linear actuators were controlled by the
Package Actuator Control System 2nd version
(PACS2), developed in-house at UWA. It provides
the user with four Proportional Integral Derivative
(PID) controller channels, which can be set up
individually in either displacement or load-control,
and synchronised using a built-in waveform and
sequence generator. This feature is particularly
critical to model accurately the load regime Figure 7. Centrifuge setup with main dimensions. The loading
rig can be displaced as a single block to load up to 16 piles in
generated by given wave conditions. All data were the 1× 1 m sample, while maintaining a spacing between piles
acquired at a sampling rate of 10 Hz using the and boundaries of at least 4D.
250
4.2 Testing programme and procedure This was undertaken simultaneously on all lines.
A total of 24 tests were performed in this research The pre-tension applied was very small (~0.05Fmono)
programme, with 3 reported in this paper. The to avoid any significant vertical (in the case of the 2
programme included unidirectional monotonic and 3 loading lines case) and horizontal (in the case
vertical and inclined (40°) tests used as reference for of the 2 loading line case) displacements.
ultimate capacity, and multidirectional cyclic
sinusoidal tests in 2 and 3 directions with two
different types of loading regimes. The first type,
defined as alternate loading, involved loading in 2
or 3 directions one after another. While this does not
model specific wave regimes, it does provide
insights on the effect of multidirectional loading on
the pile performance. The second type, defined as
out of phase loading, aims at replicating loading
regimes as defined by the hydrodynamic analysis.
Figure 8. Multidirectional cyclic loading tests includes loading

along 2 or three directions, with spacing α in the horizontal
plane varying as 60, 90, 120, 180 and 180 degrees. During
phase loading, the phase angle between the two or three
sinusoidal loading signals varies as π/9, π/6, π/4, π/3, 7π/18 Figure 9. Loading regimes for the 3 multidirectional cyclic
tests reported in this paper. The figures on the right shows the
and π/2. The resulting loading path in the horizontal plane
load applied along loading lines as a ratio of the ultimate
encompasses all the loading regimes identified in the
unidirectional monotonic capacity. Test cyc_04_01 and cyc-06-
hydrodynamic analysis.
01 are 2 directions and 3 directions alternate loading tests,
where the loads were applied one after another. Test cyc_06_02
Figure 8 presents the different parameters that is a phase loading test, with the phase angle between the second
were varied during the multidirectional cyclic tests, and third loading being π/6 and π/2, respectively. The resulting
while Figure 9 presents the details of the 3 tests loading path in the horizontal plane is presented in the middle
figure and represent an ellipsoid. For the three tests, the figures
reported here, with parameters listed in Table 2. on the left shows the trace of the vertical resultant. Note that in
For all tests, the magnitude of the load was set as some cases (for tests not reported here), the vertical resultant
a ratio of the ultimate capacity established from the may be a monotonic load. This has consequences on the
unidirectional monotonic inclined loading tests, performance of the pile.
Fmono. The load was increased by increments of 0.25
or 0.05Fmono, with number of cycles varying from 50 4.3 Test results and discussion
to 100 depending on the displacement response Results for the three tests reported in this paper are
during the tests. presented in Figures 10, 11 and 12, respectively.
The figures include (i) the load displacement curves
Table 2. Summary of the multidirectional cyclic tests tests along the loading lines, (ii) the accumulation of
reported in the paper
displacement with number of cycles along the
Test
Test ID α (º) ω1−2 (º) ω1−3 (º) loading lines, (iii) the resultant vertical load
type
cyc_04_1 A-2L 120 - - displacement curve, and (iv) the 3D motion of the
-
pile in x, y, z space. The figures need to be examined
cyc_06_1 A-3L 120 -
with respect to the load regimes presented in Figure
cyc_06_2 P-3L 120 π/6 π/2
7 and notably the resultant vertical load input.
For all three tests, failure occurred under vertical
The test procedure involved an important step motion, with the pile being pulled out of the soil.
prior to the start of the test in order to calibrate The kinematics of the pile is however strongly
PACS2. For PACS2 to be able to apply influenced by the load path followed, and notably
synchronised loading in three directions, it was the vertical resultant. Under the first packet of
necessary to slightly pre-tension each line equally. cycles (at 0.25Fmono), horizontal displacements are
251
<0.1D and of the same amplitude as the vertical difference in peak vertical resultant load, as
displacement for both test cyc_04_01 (2 loading illustrated in Figure 9. For test cyc_04_01, the
directions) and test cyc_06_01 (3 loading vertical load resultant peaks at 0.25Fmono, while it
directions). Test cyc_04_01 is able to sustain peaks at 0.3Fmono for test cyc_06_01. This slight
0.25Fmono, as well as a higher 0.5Fmono, before failing difference may be sufficient to explain the
at 0.75Fmono, at which point horizontal and vertical predominance of vertical failure for test cyc_06_01.
displacements are still of the same magnitude This is discussed in more details later in the paper.
(0.6D-0.8D). In contrast, test cyc_06_01 is unable Comparison of test cyc_06_01 and cyc_06_02
to sustain 0.25Fmono for more than 45 cycles. Failure provides insights into the effect of alternate and out
occurs rapidly, essentially through an acceleration of phase loadings on the pile behaviour. For the
of the accumulation of vertical displacement (0.5D), latter case, the pile experiences an ellipsoid motion,
while horizontal displacement remains limited to with horizontal and vertical displacement of very
0.1D. The different behaviour in the two tests limited amplitude (<0.05D) within the first packet
cannot be attributed to the different multidirectional of cycles, increasing to 0.1D over the second packet,
loading regimes, but is most likely explained by the before reaching 0.15-0.25D in the horizontal plane
and 0.5D in the vertical plane at failure over the
third packet. Interestingly, the peak vertical
resultant for test cyc_06_01 is 0.5Fmono for the first
Figure 10. Results of test cyc_04_01. It is an alternate two lines

loading test. The test was conducted with 3 increasing cyclic
packets, 50 cycles at 0.25Fmono, 100 cycles at 0.5Fmono and 18
cycles at 0.6Fmono as failure was reached. Displacement along
the two loading lines show similarities, with no accumulated
displacements over the first packet, a more significant
accumulation over the second packet, and an acceleration over
the third, leading to failure. Displacement reached at failure is
0.9D along the first direction of loading and 0.8D along the
second direction, showing a dissymmetry of motion and the
preponderance of the first loading direction on the pile
behaviour. The bottom right plot shows the evolution of the
vertical resultant with vertical displacement. The trend is Figure 11. Results of test cyc_06_01. It is an alternate 3 lines
identical to that along the two loading lines, with accumulated loading test. 80 cycles at 0.25Fmono were performed and led to
vertical displacement reaching 0.65D at failure. The bottom left failure. The accumulation of displacement along the three
plots shows the 3D pile motion in x, y, and z space, and confirms loading lines are identical, with a significant acceleration of
the observation made from the previous plots. Failure is displacement after about 45 cycles. As seen on the 3D pile
reached by the pile being pulled out of the ground, while motion, horizontal displacements are <0.1D at failure, while
experiencing horizontal displacements of similar magnitude as the vertical displacement goes beyond 0.5D, indicating that
the vertical displacements, with a slight dissymmetry towards failure is essential a shearing failure at the pile soil interface.
the first direction of loading (the resultant horizontal Closer examination of the pile displacements on the horizontal
displacement is not along the resultant load, but slightly biased plane (figure not shown here) indicates that pile drifted rapidly
towards the first direction of leading). towards the 3rd loading direction as the vertical displacement
accelerated.
252
packet of cycles, i.e. higher than for test cyc_04_01, component, depend essentially on the nature of the
which experienced failure after only 45 cycles. In multidirectional loading, alternate or out of phase,
this case, the multidirectional loading regime is the which itself is governed by the wave characteristics
main contributor to failure, with alternate loading, and the geometry of the array.
generating abrupt loading reversal and higher In all cases, the piles fail by behind pulled out
vertical resistance degradation. from the soil, at vertical peak load not exceeding 0.5
times the monotonic vertical capacity. Such
reduction cannot be only explained by cyclic
friction degradation at the interface from ϕp to ϕr.
Additional mechanisms must take place and they
include (i) a reduction of the stress state around the
pile from K0 conditions to Ka conditions, as the pile
is loaded in multiple directions and the soil in the
vicinity moves to active state, and (ii) a reduction of
the soil density in the vicinity of the pile. In all
cases, the horizontal displacements prior to failure
were in the range 0.05-0.1D leading to strain levels
in the soil that are expected to be of the order of 5-
10%. At these strain levels the density is reduced
significantly, as observed from direct shear tests
reported by Chow et al. (2019). Further analysis is
required to understand how each of these factors
contribute to the degradation of pile performance
under multidirectional loading.
6 CONCLUSIONS
Centrifuge tests were performed to investigate the
performance of piles under multidirectional
loading, as relevant for anchor sharing within arrays
of floating renewable energy devices (wind turbines
or wave energy converters). The load paths
experienced by the piles were identified from
numerical hydrodynamic analyses. 3 tests are
reported in this paper that illustrate the reduction in
Figure 12. Results of test cyc_06_02. It is a 3 lines phase pile capacity due to multidirectional loading. In all
loading test. The testing sequence included 90 cycles at cases, the pile fails by being pulled out vertically of
0.05Fmono, 170 cycles at 0.15Fmono, and 10 cycles at 0.2 Fmono at
which point failure was reached. Load displacement curves
the soil, while experiencing horizontal
along the three loading direction indicate no accumulated displacements, of a magnitude that was a function
displacement over the first packet of cycles, a moderate of the type of multidirectional loading experienced.
accumulation of displacement over the second packet (about Failure occurs at resultant vertical peak loads lower
0.04D over 170 cycles), and a very large acceleration of than 0.5 times the monotonic vertical capacity. This
accumulation of displacement, leading to failure after less than
10 cycles in the third packet. Examination of the 3D reduction of vertical capacity is likely to be due to a
displacement motion, and displacements on the horizontal combination of cyclic friction degradation, change
plane indicates that the pile experienced an ellipsoid motion of stress state and reduction of density around the
around its centre, with its major axis oriented 90° from the pile due to the multidirectional loading. The
major axis of the ellipsoid load path defined in Figure 7.
contribution of each of these factors, as a function
Displacements were of very limited amplitude in the x, y and z
directions (<0.05D). Upon application of the third packet, the of the type of multidirectional loading (alternate or
orientation of the ellipsoid remained unchanged, but out of phase) and the magnitude the cyclic vertical
displacement increased significantly in amplitude, reaching resultant requires further analysis.
0.15D in the x direction, 0.25D in the y direction and over 0.5D
in the z direction. Failure occurs with the pile being pulled out
of the ground with significant horizontal displacements. 7 ACKNOWLEDGEMENTS
The research presented in this paper was supported
These 3 tests illustrate that the pile performance is by the Australian Research Council Discovery
affected by multidirectional loading, but that the Project programme DP35402000, the Lloyd’s
parameters governing failure, whether the change in Register Foundation (LRF) Centre of Excellence in
loading direction, or the magnitude of the vertical Offshore Foundation Engineering and the Western
253
Australian State Government via the Wave Energy Herduin, M. 2019. Multidirectional loading on shared
Research Centre. In-situ wave and hydrodynamic anchors for offshore renewable energy: definition and
data were provided by the Renewable Energy preliminary investigation into soil behaviour and
Research Group of the University of Exeter. Their anchor performance. PhD Thesis, the University of
Western Australia.
contribution is gratefully acknowledged.
IEA 2019. Offshore wind outlook 2019. World energy
8 REFERENCES outlook special report. Technology report 2019.
Beemer, R.D., Murali, M., Aubeny, C.P., and Biscontin,
G. 2018. Use of a MEMS accelerometer to measure Jamiolkowski, M., Presti, D.C.F. Lo, and Manassero, M.
orientation in a geotechnical centrifuge. International 2003. Evaluation of relative density and shear strength
Journal of Physical Modelling in Geotechnics. 15(5): of sands from CPT and DMT. ASCE Geotechnical
253-265. Special Publication. 119: 201–238.
Child, B. F. M. and Venugopal, V., 2010. Optimal Karimirad, M., Koushan, K., Weller, S.D., Hardwick, J.,
configurations of wave energy device arrays. Ocean Johaning, L., 2014. Applicability of offshore mooring
Eng., 37(16): 1402–1417. and foundation technologies for marine renewable
energy (MRE) device arrays. Proc. 1st International
Chow, S., Roy, A., Herduin, M., Heins, H., King, L., Conference on Renewable Energies Offshore, Lisbon,
Bienen, B., O’Loughlin, C.D., Gaudin, C., Cassidy, Portugal.
M.J. 2018. Characterisation of UWA superfine silica
sand. GEO report 18844. Centre for Offshore Kost, C., Mayer, N.J., Thomsen, J., Hartmann, N.,
Foundation Systems, The University of Western Senkpiel, C., Philipps, S., Nold, S., Lude, S., Saad, N.,
Australia Schlegl, T., 2013. Levelized Cost of Electricity. PV
and CPV in Comparison to Other Technologies. Proc.
Gaudin, C., O’Loughlin, C.D., Breen, J. 2018. 29th International Photovoltaic Solar Energy
Commissioning of a 240 g-tonne geotechnical Conference, Amsterdam, The Netherlands.
centrifuge at the University of Western Australia.
Proceedings of the 9th International Conference on Martinelli, L., Ruol, P., Coretelazzo, G., 2012. On
Physical Modelling in Geotechnics, London, UK. mooring design of wave energy converters: The
Seabreath application. Proceedings of the 33rd
Gaudin, C., White, D.J., Boylan, N., Breen, J., Brown, International Conference on Coastal Engineering,
T.A., De Catania, S., Hortin, P. 2009. A wireless data Santander, Spain.
acquisition system for centrifuge model testing.
Measurement Science and Technology. 20: 095709. Willow, C., and Valpy, B. 2011. Wave and tidal energy
in the UK, state of the industry report. Renewable UK.
254
Suction caisson response under tensile and compressional vertical cyclic loading in
sand over clay
Marc Stapelfeldt1, Jürgen Grabe1
1
Institute of Geotechnical Engineering and Construction Management, Hamburg University of Technology,
Germany
Corresponding author: Marc Stapelfeldt (marc.stapelfeldt@tuhh.de)
ABSTRACT: Offshore wind turbines supported by suction caisson jackets transfer the loading to the
foundation primarily through a vertical push-pull mechanism. This leads to the foundation performance
under vertical cyclic loading being a critical design aspect. Recent research of suction caissons under cyclic
vertical loading into tension in uniform soil profiles has revealed the complexity of the load transfer
mechanisms and the importance of understanding the soil permeability and shear strength governing the in-
service foundation performance. Soil stratification, with clay and sand layers, adds further complexity.
Numerical modelling of the suction caisson under vertical cyclic loading is complex, challenging and the
consideration of individual aspects may not necessarily be required. Simplified numerical models, once
validated, can provide insights into key aspects of the load bearing behaviour. However, the identification
of those aspects and the validation of a numerical model requires an experimental database that does not
exist to date. The current paucity of data of suction caisson in-service performance in sand over clay under
loading conditions relevant to offshore wind turbines calls for physical evidence to be gathered. This is
addressed here through centrifuge tests on vertical cyclic loading, which unravel the actual impact of tensile
loads below and above the recommended design capacities. The vertical cyclic loading test results indicate
that not just the development of suction pressure inside the caisson foundation, but also the interplay
between tension and compression have a significant effect on the overall response of suction caissons under
cyclic vertical loading in sand over clay.
Keywords: Offshore Wind Turbine; Suction Caisson; Layered Soil; Cyclic Loading; Numerical Modelling; Centrifuge
Modelling.
1 INTRODUCTION model that is validated through those can provide

Offshore wind farm locations are moving from valuable insights into the suction caisson
shallow water depths towards deeper water. This performance. This approach is taken here. This
development often implies sea beds characterised paper aims to investigate suction caisson
by clay and layered soils rather than sand. Suction performance under vertical cyclic loading in SoC
caisson jackets are a suitable foundation system for through complementary centrifuge tests and
offshore wind turbines (OWTs) in such conditions numerical analyses.
(Tjelta, 2015). OWT supporting jackets are Investigations in sand showed that the average
typically equipped with three or four suction load has a major influence on the response under
caissons (also called suction buckets). The aspect vertical cyclic load (Bienen et al, 2018a,b). Similar
ratio of length L to diameter D usually ranges behaviour might be present in layered soil profiles,
between L/D = 0.5 and L/D = 1. Steel suction such as SoC. Hence, two different average loads are
caissons for OWTs are relatively thin-walled (Tjelta considered in order to provide insights into the
et al., 1990; Masui et al., 2001), with wall interplay between tension and compression.
thicknesses t ranging from t ≈ D/150 to t ≈ D/250.
OWTs are significantly lighter than most other 2 METHODS
offshore platforms. Wind and wave action 2.1. Numerical simulations
transferred to the individual footings is primarily The numerical simulations were performed with
withstood by a vertical push-pull mechanism. Abaqus/Standard 2018. Undrained caisson response
Hence, the windward caisson is subjected to vertical is expected in SoC due to the low effective
cyclic loading, which might extend into tension permeability of the clay main layer. This also holds
(Bienen et al, 2018a,b). for the sand part of the soil plug, because it is
The behaviour of suction caissons embedded in hydraulically encapsulated inside the caisson skirt.
layered soil profiles under vertical cyclic loading – The free top layer is not expected to be subject to
particularly in sand over clay (SoC) – is not fully considerable soil state changes besides the frictional
understood. Experimental data and a numerical
255
interaction with the caisson skirt. Hence, undrained The utilised model geometry is shown in Figure 1.
one-phase numerical analyses are conducted. The given dimensions were found to minimise
boundary effects in the numerical simulations. The
Axis of symmetry
0.5D two-dimensional axisymmetric model geometry
was meshed with high order rectangular elements
(CAX8). The suction caisson was modelled as a
rigid body meshed with triangular elements
L (CAX6). A half-circular shaped skirt tip was found
to reduce artificial stress peaks in that area. The
discretisation of the computational mesh is shown
V
Caisson
t in Figure 1b.
Horizontal displacements were prevented at the
0.5L
axis of symmetry and outside vertical edge of the
model geometry, while vertical displacements were
permitted. The base featured zero vertical
displacements boundary conditions. The
displacements at the soil surface were free. The
Top layer vertical load V was applied to the centre of the
caisson (see Figure 1a).
6L
The parameters required for the utilised
Main layer
constitutive models are summarised in Table 1. The
top layer consists of dense fine silica sand and the
kaolin clay main layer is considerably
overconsolidated. The parameters found suitable for
fine silica sand are determined under consideration
of the laboratory test discussed in Tran (2005),
Pucker et al. (2013), and Chow et al. (2018). The
2D
laboratory experiment results discussed in Qui
(2012) and Mašín (2019) form the basis of the
determination of the parameters for kaolin clay
listed in Table 1.
Table 1. Model parameters for FEM simulations featuring

silica sand and kaolin clay (Pucker et al., 2013, Chow et al.,
2018).
Parameter Unit FSS KC
Undrained shear strength su kPa - 25
Young’s modulus E MPa 60 1.2
Poisson’s ratio ν - 0.475 0.475
Specific weight soil γ’ kN/m³ 10.5 7.1
FSS = fine silica sand; KC = Kaolin clay
The soil-structure interaction is modelled through

hard normal contact. Separation of the contact
surfaces is prevented. Tangential friction is
modelled by means of the Coulomb friction model
utilising a frictional coefficient of µs = 0.15 in sand
and µc = 0.4 in clay, a shear stress limit of
τmax,s = 12 kPa and τmax,c = 20 kPa and an elastic slip
limit of γi = 0.005.
The simulations are conducted according to the
following procedure: First, the vertical and
Figure 1. Dimensions (a) and discretization (b) of the horizontal stresses are initialised by applying a K0
numerical model utilised for FEM analyses of the suction initial condition. A minimum load is applied at the
caisson response to vertical cyclic loading in a layered soil. caisson in order to minimise the effect on the K0 soil
state. The installation process is neglected. Load
256
controlled caisson cyclic loading commences once characterisation test results are presented and
stable K0 conditions are reached. The loading discussed in detail in Stapelfeldt et al. (2020b).
amplitudes are given in Table 2 (see section 2.2.3).
2.2.2 Caisson model, test arrangement and
2.2. Centrifuge tests instrumentation
The experiments are performed at 100g in the The tests were performed with the same centrifuge
Acutronic Model 661 centrifuge (Randolph et al., arrangement as utilised by Stapelfeldt et al.
1991; Randolph and Gaudin, 2017) at the University (2020a,b). The anodised aluminium model caisson
of Western Australia. The presented test results has a diameter of D = 80 mm and a skirt length of
from part from a comprehensive series of centrifuge L = 40 mm, which when tested at 100g represents a
test discussed in Stapelfeldt et al. (2020b). diameter D = 8 m and a skirt length of L = 4 m in
prototype scale. The skirt wall thickness is
2.2.1 Soil sample and fluid preparation t = 0.5 mm, which corresponds to a prototype wall
Overconsolidated stiff clay and dense sand were thickness of t = 50 mm.
used in the centrifuge tests. The layered soil sample
with a thickness of the top layer of 20 mm was
prepared to create realistic conditions for sites that Actuator
allow the utilisation of suction caisson foundations.
The clay was prepared from slurry at a water content Camera
of approximately 120 %, mixed under vacuum,
filled in a strong box and consolidated in a hydraulic
press under a maximum vertical pressure of Load cell
σv = 350 kPa. The height of this clay layer was
LDT
approximately 140 mm.
The sand top layer was prepared by pluviation of
dry sand into the strongbox from a height of
approximately 1.2 m by means of an automatic Three way valve
pluviator. The sand layer was vacuum levelled to a Pulley chord
target total sample height of 160 mm above the base
Hose
of the strongbox. The sand layer was saturated from
the layer boundary through tubes – one located in
each corner of the strongbox. A total sample height
of at least 4L was targeted in order to minimise
boundary effects (Tran and Randolph, 2008). The
sand unity weight was determined from weight and Rod
CoS
volume measurements. The sand relative density
was thus determined to be Dr ≈ 83 %.
Under consideration of scaling laws governing the Caisson
suction caisson response, the viscosity η of the sand Filter layer
top layer pore fluid was increased from
approximately ηw = 1 cSt for water to a target value
of η = 100 cSt (Taylor, 2011). Hereby an effective
Figure 2. Centrifuge test set-up for cyclic loading model test
permeability of the sand layer – i. e. kf ≈ 1∙10-4 m/s in layered soil.
– at 100g was targeted, which is representative of
typical North Sea sands (Tan and Scott, 1985). The The model caisson has been equipped with two
viscosity was increased through the addition of total pressure transducers (TTP) – one at the lid
methyl cellulose ether (DOW, 2002). The targeted invert and another one at the top of the lid, and a
pore fluid viscosity is confirmed by measurements. pore pressure transducer (PPT) at the lid invert. The
The clay layer was water saturated. caisson model shown in Fig. 2 is connected to a load
The targeted soil sample characteristics were cell with a maximum capacity of 3 kN by means of
confirmed through in-flight cone penetration and a rigid rod. This assembly is carried by an actuator,
T-bar tests conducted before and after the suction which is mounted on top of the strongbox. The
caisson tests. A T-bar test was also conducted three-way valve on the caisson lid was operated by
immediately after the clay consolidation in order to an electric winch and was able to vent to ambient,
verify the targeted undrained shear strength of to connect the caisson to the syringe pump (House,
su ≈ 25 kPa. The centrifuge soil sample
257
2002) and to be sealed. The fluid table was at least is tensile in each cycle, except the first loading
hw = 100 mm above the soil surface to ensure that sequence (LSQ 1) of LC 21. The average stresses
the model caisson including the three-way valve differ: An increased compressive average stress is
were submerged to minimise changes in buoyancy applied in LC 21. The calculated displacement
during the installation. The self-weight and the response of the suction caisson model is presented
cyclic loading of the model caisson were applied in Figure 3.
with the actuator, which was controlled through a
feedback loop from the load cell. The displacement -150 -100
Vertical load
-50 0
V/A (kPa)
50 100 150
of the caisson was recorded by means of a linear -0.01
displacement transducer (LDT) that was -0.005

net settlement
independently mounted.
z/L
0
2.2.3 Testing procedure 0.005
Relative vertical displacement

The model caisson was installed in-flight through 0.01
self-weight penetration and subsequent suction

installation. The self-weight was applied with the
0.015
actuator. The syringe pump provided suction 0.02

net settlement
pressure through the three-way-valve. Once the 0.025
installation was completed, a static vertical pressure SoC-0.5-FEM-21N
SoC-0.5-FEM-20N
of V/A = 116 kPa representing the full self-weight 0.03
of the OWT was applied. This load was maintained

until the internal excess pore pressure dissipated, -0.01
SoC-0.5-FEM-21N
which was monitored by the pore pressure SoC-0.5-FEM-20N
z/L
-0.005
transducer at the lid invert. Unloading to the low LSQ 4
magnitude compressive average load (see Table 2) 0 LSQ 2
including another waiting period simulated a 0.005
reduction of the vertical load on the critical

0.01
windward caisson due to wind and wave loading,
before the cyclic loading amplitude was applied. 0.015
0.02
Table 2. Load cases (LCs) and specifications of the four cyclic LSQ 1
loading sequences (LSQs). 0.025 LSQ 3

LSQ Cycles Load LC 20 Load LC 21
0.03
n V/A V/A 30 35 40 45 50 55 60 65 70
- - kPa kPa Number of cycles n
1 1000 (18) 8±16 32±24 Figure 3 Calculated normalised vertical displacements during
2 100 (5) 8±30 32±54 vertical cyclic loading.
3 10 (3) 8±60 32±84
The calculated vertical displacements show net
4 1 (1) 8±90 32±114
settlements in both LCs. The shift of the average
Number of cycles for FEM analysis in brackets
load towards compression yields increased net
settlements in LC 21. The calculated net settlements
3 RESULTS AND DISCUSSION always occur during the first cycle of LSQ within
The experimental and numerical results are the first loading package –i. e. every time a new
presented in non-dimensional form, unless stated maximum compressive load is reached. Figure 3b
otherwise. Positive values indicate compression and indicates that almost elastic behaviour results from
downward displacements, whereas negative values the application of the second loading package. This
represent tension and upward displacements. is consistent due to the utilisation of the elastic-
plastic constitutive model, which is a fundamental
3.1. Numerical simulations simplification. Hence, a more advanced constitutive
model is required. However, determining a suitable
The presented numerical model is utilised in order constitutive model requires fundamental knowledge
to investigate the response of a suction caisson of the underlying load transfer mechanisms, which
foundation subjected to vertical cyclic loading. The currently does not exist. Hence, the performance of
two applied load cases (LCs) featured cyclic vertical experiments is required in order to evaluate and
loading to the same minimum vertical load, which improve the validity of numerical simulations.
258
3.2. Centrifuge tests LC 21. Consequently, considerable net settlements
The centrifuge results presented in Figure 4 reveal result instead of minimal net uplift, which occurs in
that LC 21 results in increased relative vertical test LC 20 (see Figure5b).
displacement amplitudes compared to LC 20. This
is consistent, because larger cyclic loading
-100
V/A
LSQ 4
amplitudes are required to reach the same tensile -75 u
peak load starting from an increased compressive LSQ 2
u (kPa)
-50
average load. Furthermore, the results obtained
(kPa)
from test SoC-0.5-150J-21 visualise the occurrence -25
V/A
of net settlement in response to load sequences that 0
Excess pore pressure

feature significant excursions into tension. The
Vertical stress
25
results of test SoC-0.5-125SI-20 show minor net
uplift. This holds for each of the three loading 50
packages evaluated in Figure 4a. 75 LSQ 3

LSQ 1
Vertical load V/A (kPa) 100

2175 2200 2225 2250
-150 -100 -50 0 50 100 150 Number of cycles n
-0.02
-100
net uplift V/A
z/L
LSQ 4
-75
LSQ 2 u
-0.01
-50
u (kPa)
-25
(kPa)
0
V/A
0
25
Excess pore pressure
Vertical stress
50
75
0.01
net settlement 100 LSQ 1
125 LSQ 3
SoC-0.5-150J-21
SoC-0.5-125SI-20 150
0.02 2175 2200 2225 2250
Number of cycles n
-0.02
SoC-0.5-150J-21
LSQ 4
Figure 5 History of vertical stresses and excess pore pressures
SoC-0.5-125SI-20
at the lid invert during SoC-0.5-125SI-20 (a) and SoC-0.5150J-
z/L
LSQ 3
LSQ 1
21 (b).
-0.01 LSQ 2
The transfer mechanisms in tension differs in the

centrifuge experiments: Tensile loads are
LSQ 4
0 completely transferred through suction pressure at
the lid invert in LC 21 (Figure 5a). In contrast,
tensile loads are carried by a combination of skirt
0.01 friction and suction pressure in test SoC-0.5-125SI-
LSQ 2
LSQ 1
20 (Figure 5b). Excess pore pressures, which are
LSQ 3
present in compression, dissipate before tensile
0.02 loading commences at an increased average vertical
2175 2200 2225 2250
stress in LC 21. Internal suction pressure develops
Number of cycles n
immediately – i. e. before the skirt friction is
Figure 4 History of the normalised relative vertical mobilised – and provides sufficient resistance.
displacements during three cyclic loading packages (a) and Excess pore pressures only partially dissipate before
during load package two (b) in centrifuge tests. tensile loading is reached, if low compressive loads
are present in LC 20. Hence, internal suction
The data shown in Figure 5 indicates that pressure mobilisation is postponed in this case and
compressive loads are transferred by the skirt a considerable portion of the tensile load is
friction, tip resistance and the lid similarly in LC 20 transferred by the skirt resistance. The results
and LC 21. Figure 5 also shows that the estimated presented in Figure 4b indicate that this leads to
sum of the skirt and the tip resistance increased uplift, if the applied tensile load exceeds
(V/A)S+T ≈ 74 kPa is arguably fully mobilised in the skirt resistance considerably in LSQ 3 and 4.
259
Thus, the presented results suggest that the average net settlements or net uplift results from vertical
compressive load determines whether cyclic cyclic loading featuring significant excursions into
loading with similar excursions into tension leads to tension. This is crucial for the in-service
net uplift or net settlement. performance of offshore wind turbines with low
magnitude compressive loading determining the
3.3. Evaluation of the numerical analyses response of the windward caisson foundation.
The comparison between Figure 3 and 4 shows that
the calculated displacements amplitudes are within 5 ACKNOWLEDGEMENTS
a similar magnitude with the centrifuge test results. The centrifuge experiments were performed at the
However, the net displacement response differs National Geotechnical Centrifuge Facility (NGCF)
considerably: The numerical model yields plastic in Perth, Western Australia, and in cooperation
and therefore irreversible soil deformations as soon between the Hamburg University of Technology
as the shear strength is exceeded locally. A recovery (TUHH) and the Centre for Offshore Foundation
of the soil shear strength or a certain residual soil Systems (COFS) at the University of Western
strength is not considered in the utilised constitutive Australia (UWA).
model. This is certainly different from the behaviour We acknowledge the support from the Deutsche
observed in the centrifuge experiments: Effects Forschungsgemeinschaft (DFG) for our research
resulting from repetitive loading occurring during project GR 1024/26-2 “Sauginstallation Maritimer
LSQ 3 (Figure 4b) are not considered. These effects Strukturen” (SIMS). This work forms part of the
are considered in advanced constitutive models – activities of the Centre for Offshore Foundation
e. g. the ISA plasticity-based model for clay Systems (COFS), which is currently supported as a
discussed in Fuentes et al. (2018). Centre of Excellence by the Lloyd’s Register
The experimental results show an accumulation of Foundation. Lloyd’s Register Foundation helps to
displacements that is accompanied by an increasing protect life and property by supporting engineering-
amplitude, which appears to reach a certain related education, public engagement and the
maximum in tension (LC 20), but continues until application of research. This support is gratefully
the end of LSQ 3 in compression (LC 21). Hence, acknowledged.
the experimental results indicate that the utilisation
of an advanced constitutive model like the ISA 6 REFERENCES
plasticity-based model for clay is required in order Bienen, B., Klinkvort, R.T., O'Loughlin, C.D., Zhu, F.
to improve numerical investigations. However, the and Byrne, B.W., 2018a. Suction caissons in dense
ISA plasticity-based model for clay is still under sand, part I: Installation, limiting capacity and
development and remains unavailable for further drainage. Géotechnique, 68(11): 937-952.
investigations on the response of suction caissons doi:10.1680/jgeot.16.P.281.
subjected to vertical cyclic loading.
Bienen, B., Klinkvort, R.T., O'Loughlin, C.D., Zhu, F.
and Byrne, B.W., 2018b. Suction caissons in dense
4 CONCLUSIONS sand, part II: Vertical cyclic loading into tension.
Numerical investigation of suction caissons Géotechnique 68(11), 953-967.
embedded in sand over clay and the performance of doi:10.1680/jgeot.16.P.282.
centrifuge tests featuring vertical cyclic loading into
tension support the following findings: Chow, S.H., O’Loughlin, C.D., Gaudin, C., and Lieng,
Numerical methods are capable of the simulation J.T., 2018. Drained monotonic and cyclic capacity of a
of suction caissons subjected to vertical cyclic dynamically installed plate anchor in sand. Ocean
Engineering 148, pp. 588-601.
loading in general. The comparison to experimental
doi:10.1016/j.oceaneng.2017.11.051.
results shows that advanced constitutive models that
include hardening and account for cyclic loading DOW, 2002. Methocel Cellulose Ethers: Technical
effects are required in order to obtain appropriate handbook, United States of America.
results.
The pivotal mechanisms underlying the vertical Fuentes, W., Tafili, M., and Triantafyllidis, T., 2018. An
cyclic load transfer mechanisms in sand over clay ISA-plasticity-based model for viscous and non-
are pointed out through the centrifuge test results. viscous clays. In: Acta Geotechnica 13.2, p. 367-386.
Hence, the presented experiments form a basis for doi:10.1007/s11440-017-0548-y.
future developments of numerical models.
House, A.R., 2002. Suction Caisson Foundations for
The centrifuge test results furthermore show that
Buoyant Offshore Facilities, PhD thesis, The
the average compressive load determines whether University of Western Australia, Perth, Australia.
260
Mašín, D., 2019. Modelling of Soil Behaviour with caissons. Journal of Geotechnical and
Hypoplasticity: Another Approach to Soil Constitutive Geoenvironmental Engineering 0(0), pp. 1-20. (under
Modelling, Springer Series in Geomechanics and review).
Geoengineering, Springer International Publishing.
doi:10.1007/978-3-030-03976-9. Tan, T.S. and Scott, R.F., 1985. Centrifuge scaling
considerations for fluid-particle systems: Centrifuge
Masui, N., Yoneda, H., Zenda, Y., Ito, M., Iida, Y. and scaling considerations for fluid-particle systems.
Hermstad, J., 2001. Installation of offshore concrete Géotechnique 35(4), 461-470.
structure with skirt foundation. In Proceedings of the doi:10.1680/geot.1985.35.4.461.
11th International on Offshore and Polar Engineering
Conference, pp. 626-630. Taylor, R.N., 2011. Geotechnical centrifuge technology.
Taylor & Francis, London.
Pucker, T., Bienen, B., and Henke. S., 2013. CPT based
prediction of foundation penetration in siliceous sand. Tjelta, T.I., 2015. The suction foundation technology,
Applied Ocean Research, 41, pp. 9-18, Meyer, V. (Ed.), In Frontiers in Offshore Geotechnics
doi:10.1016/j.apor.2013.01.005. III. Balkema, pp. 85-93. doi:10.1201/b18442-6.
Qui, G., 2012. Coupled Eulerian Lagrangian Simulations Tjelta, T.I., Aas, P.M., Hermstad, J., and Andenaes, E.
of Selected Soil-Structure Interaction Problems, PhD 1990. The skirt piled Gullfaks C platform installation,
thesis, Hamburg University of Technology, Hamburg, In Offshore Technology Conference 1990, pp. 453-
Germany. 462. doi:10.4043/6473-MS.
Stapelfeldt, M., Bienen, B. and Grabe, J. 2020a. The Tran, M.N., 2005. Installation of Suction Caissons in
influence of the drainage regime on the installation and Dense Sand and the Influence of Silt and Cemented
the response to vertical cyclic loading of suction Layers, PhD thesis, The University of Sydney, Sydney,
caissons in dense sand”. Ocean Engineering 0(0), pp. Australia.
1-29. (under review).
Tran, M.N. and Randolph, M.F., 2008. Variation of suction
Stapelfeldt, M., Bienen, B. and Grabe, J. 2020b. The pressure during caisson installation in sand.
influence low permeability layers on the installation Géotechnique 58(1), 1-11. doi:10.1680/geot.2008.58.1.1.
and the response to vertical cyclic loading of suction
261
Centrifuge Modelling of Suction Caissons Subjected to Cyclic Loading in Tension
Yuen Zhang1, Kamaleshwar Sudhakaran1,2, Amin Askarinejad1

1
Faculty of Civil Engineering and Geosciences, Delft University of Technology, the Netherlands
2
Heerema Marine Contractors, the Netherlands
Corresponding author: Amin Askarinejad (A.Askarinejad@tudelft.nl)
ABSTRACT: Wind is a major source of renewable energy in Northern Europe. Suction caissons are
increasingly considered as an alternative foundation solution for offshore wind turbines (OWT) in the last
two decades because of various advantages such as low installation noise, low cost, ease of installation
and removal as well as high bearing capacity under extreme environmental conditions (Hung et al., 2018;
Jin et al., 2012). For a jacket structure installed on three suction caissons, the response of the foundation
under cyclic vertical loading in tension is of primary interest in geotechnical design. In practice, horizontal
loads induced by wind and waves are relatively large in comparison to the self-weight of the jacket
structure. This leads to relatively large overturning moments and large pull-out loads on the windward leg
of the tripod jacket structure. During this study, a series of centrifuge tests has been conducted to
investigate the response of suction caissons to two types of vertical cyclic loading, i.e. one-way and two-
way asymmetric loading. Test results in terms of cumulative settlement response of the caisson along with
damping and stiffness behaviour are analysed.
Keywords: Suction Caissons; Cyclic Loading; Storm Loading; Unloading Stiffness.
INTRODUCTION Randolph, 2009). It is a straight forward and

As countries around the world are in a clean accurate method as very few parameters are
energy race, various renewable energy sources are needed for the calculation. Houlsby and Byrne
harvested to meet the energy needs. In Western (Houlsby and Byrne, 2005) prediction model
Europe, wind is the primary source of renewable follows a beta approach and is used widely when
energy. To meet the energy demand, higher CPT data for the site or sample is not available.
capacity wind turbines are being installed in deeper However, estimation of key input parameters such
waters. In deep waters, monopile foundations are as Nq, N and pore pressure factor creates
uneconomic and ineffective due to its inability to uncertainty in the calculation for prediction
withstand large overturning moments. Thus, multi- Suction caissons had been used in the foundation
caisson foundations are preferred in deeper waters for oil and gas platforms in deep waters due to
(Bo Ibsen et al., 2005). their economic advantages over traditional pile
Typically jacket or tripod structures are preferred driving in deep waters (Houlsby and Byrne, 2005).
as multi-footing foundation and these foundations Due to the very high vertical load provided by the
are supported by either piles or suction caissons. platform, overturning moments are negligible in
Suction caissons are preferred due to the ease of the oil rigs. In case of an offshore wind turbine
installation and removal. Installation of suction (OWT), since the structures are relatively light,
caisson takes place in two phases. First phase is the horizontal loads and overturning moments are
self-weight penetration phase (SWP), where the larger than the vertical loads (Houlsby and Byrne,
caisson is allowed to penetrate the soil under its 2000). The large overturning moments are
own weight. In the second phase, penetration is balanced using a ‘push-pull’ system (Figure 1),
achieved using suction, which creates a differential where when the windward caisson experiences
force on caisson lid resulting in downward tension, the movement is compensated by the other
movement. During the second phase, seepage flow two leeward caissons. Due to the strict permanent
occurs, and this helps in reducing the soil rotational limit of 0.25 in serviceability conditions
resistance at the tip of the caisson (Houlsby and for OWTs, ‘pull behaviour’ or tensile capacity
Byrne, 2005). Prediction models are also available plays a major role in the design of an OWT (Peire
to predict the self-weight penetration depth and et al., 2009).
suction required for penetration. Different Various investigations have been done in
prediction methods use different set of parameters laboratory conditions and enhanced acceleration
for the calculation. Senders and Randolph method levels in centrifuge to study the tensile response of
uses the cone penetration resistance data to predict suction caissons (Kelly, Houlsby, & Byrne, 2006;
the suction required for installation (Senders & Kim, Lee, & Kim, 2016). Response of suction
263
caissons under cyclic loading in sand depends 2.1 Model suction caisson
mainly on the drainage conditions. Drainage Cyclic loading tests are done at an acceleration
conditions depend on several factors such as level of 100g. To study the worst-case scenario of
density, permeability of sand and frequency of the a caisson in tension, an embedment ratio (L/D) of
applied cyclic loads. In an offshore scenario, a 0.5 is used, where diameter of model caisson (D) is
partially drained condition is usually witnessed. 60 mm and skirt length (L) is 30 mm equivalent to
Thus, a wide knowledge on the settlement and prototype diameter of 6 m and skirt length of 3 m.
stiffness response of caisson under various The model caisson has a thickness of 1 mm (skirt
drainage conditions is necessary, as OWTs are and lid) and is fabricated out of brass. A fine filter
designed as a soft-stiff system. Bienen et al. layer with a thickness of 0.5 mm is placed inside
(Bienen et al., 2018a) have conducted several tests the caisson beneath the lid to avoid damage to the
in almost undrained condition in sand. They pore water pressure sensor and to prevent sand
observed that foundation stiffness changed when outflow during suction application.
the loading changed from compression to tension, Both phases of the installation of suction
which indicates that serviceability under tensile caissons (SWP and SAP) are simulated in these
loading is vital for the design of foundation. tests. The caisson lid has two openings for water
Undrained behaviour will only occur locally or outlet during each phase. The SWP outlet is closed
when the frequency of the load cycle is very high. after self-weight installation is completed and the
SAP outlet is connected to the pump or reservoir to
facilitate suction during the main phase of
installation using suction application. Two pore
1 - Windward caisson
2, 3 - Leeward caisson 2 pressure transducers are mounted on top of the
caisson lid to measure the water pressure both
outside and inside the caisson (Figure 2).
Loading direction Pressure sensor on top of lid

1
Guide rod screw
1.5
Filter
30
Pressure sensor
Sp a beneath caisson lid
c in 3
gb
e tw 60 1
een
cai
sso
ns Self weight penetration outlet
Suction assisted penetration outlet Unit : mm
a) b)
Figure 1. 'Push-Pull' behaviour of offshore tripod foundation
Figure 2. Schematics of instrumented suction caisson a) Top
view b) Side view
In this paper, installation of caisson is done at
laboratory conditions and effect of installation
method is studied. Extraction of caisson is also 2.2 Installation of suction caissons
done in a very slow manner to obtain the drained Two setups are designed for the application of
tensile capacity of the caisson. Loading scenario is suction during installation. The first setup uses
designed based on the drained tensile capacity of gravity flow to apply suction during suction
the caisson. The main goal of this paper is to study assisted penetration phase (Figure 3). Gravity flow
the drained response of suction caisson under one- is established through the head difference between
way and two-way asymmetric loading. strong box (h1) and the reservoir (h2). Using a
pressure sensor at the bottom of the reservoir, flow
EXPERIMENTAL METHODS rate can be measured. A very low flow rate of 7.52
ml/sec is attained during this set-up. All the pipe
All the tests are done in the geo-centrifuge at Delft
connections are kept beneath the water level in the
University of Technology. Installation of suction
strong box to avoid air bubbles. Flow is controlled
caisson is made possible using an apparatus
using a pneumatic valve which is operated
developed in-house which constitutes of a strong
remotely.
box, reservoir, pressure sensors and appliances for
In the second installation setup, a gear wheel
the application of suction.
pump is used to apply the necessary suction
(Figure 4). Flow rate in this setup is controlled
with the help of a needle valve manually.
264
2.3 Soil description and sample preparation

G
R - Reservoir
P1 - Pressure sensor for reservoir Medium sized silica sand is used to in this research.
P2 - Pressure sensor on top of caisson lid
P3 - Pressure sensor beneath caisson lid
G - Guide rod
Geotechnical properties of silica sand types are
Pneumatic
valve obtained from laboratory tests (Table 1). Soil
P2
P3 profile characteristic is illustrated (Figure 5).
Water Table 1. Properties of silica sand (De Jager et al., 2017)

Parameter Silica sand
P1 Maximum void ratio 0.86
R h1
Soil Minimum void ratio 0.625
cu 1.38
h2
D50 0.39 mm
cc 1.05
Figure 3. Setup for installation using gravity flow
Friction angle (cv) 34.1°
Cohesion 0 kPa
Permeability 8.41E-4 m/sec
Specific Gravity (-) 2.65
In this paper, de-aired water is used as a pore

fluid for preparing the samples in order to study an
almost drained behaviour of sand. Homogeneous
sand samples are prepared using dry pluviation
technique. First, dry silica sand is pluviated into
the strong box filled with water, and vibration is
applied during this process to obtain a dense
sample. Once the sand reached a depth of 20 mm,
further densification is done by dropping the
sample onto the floor from a height of 5 mm. This
process is repeated till the desired sample height is
achieved. After the desired height is reached,
vibration is applied for a period of 15 minutes.
This created a sample with a relative density of
752%.
Figure 4. Setup for installation using pump
In order to avoid tilting of suction caisson during

installation, a guiding system is introduced for both Figure 5. Soil profile characteristic in homogeneous sand
setups. It comprised of a guide rod and a few guide
screws. Guide rod is a cylindrical steel rod 2.4. Test procedure
connected to the caisson. Strain gauges are For all the tests, installation of suction caisson is
mounted on the actuator to measure the load conducted at 1g laboratory conditions. The
applied during cyclic loading and extraction phase. installation is followed by cyclic loading tests or
The vertical displacement of caisson during
extraction tests which are performed at 100g. A
installation and cyclic loading tests are measured detailed procedure of the entire process is given
using a linear variable displacement transducer below:
(LVDT), which is connected to the other end of
(a) Saturated sand sample prepared in the strong
guide rod.
box is placed in the centrifuge carrier which is
fitted with a loading actuator.
265
Table 2. Overview of all the centrifuge cyclic loading tests

Test mode Soil type Average Max. Max. Initial Final Load
vertical load compressive tensile cycle cycle frequency (at
(kN) load (kN) load (kN) No. No. model scale)
One-way Homogeneous 250 500 0 0 310
-100 500 -700 0 150
0.028
Two-way Homogeneous -175 500 -850 151 250
-300 500 -1100 251 310
(b) Caisson is slowly and completely submerged compressive loading and two-way asymmetric
into the de-aired water in a diagonal manner to get storm loading. Bienen et al. observed that
rid of any air bubbles underneath the caisson lid. significant uplift of the caisson is witnessed when
The SWP valve is open during this process. tensile loads exceed 30~50% of the drained tensile
Caisson is then fixed to the guide rod before capacity (Bienen et al., 2018b). Thus, 30% of the
placing it gently on the soil surface. drained tensile capacity (500 kN) is fixed as
maximum tensile amplitude in serviceability limit
(c) Using the loading actuator, self-weight
state. For two-way asymmetric storm loading,
installation is done to a depth of 10 mm using the
tensile loads of more than 50% of the drained
actuator with a displacement rate of 0.05 mm/sec. tensile capacity is used. Details of each test and
During this phase, the SWP outlet is kept open. amplitudes in prototype scale are provided (Table
Once self-weight installation is done, SWP outlet 2). The loading history curve is presented in
is sealed. Figure .
(d) Suction assisted penetration is conducted by 800
One-way
Two-way
applying suction either using a pump or with the 400
help of gravity flow up to a depth that the inner
Load (kN)
0
side of caisson lid came in contact with the soil
-400
surface. This is realised when there is a sudden
-800
spike in the pressure sensor reading underneath the
-1200
caisson lid and no displacement in LVDT is
recorded. 0 50 100 150 200 250 300
Number of cycle, n
(e) Once installation is complete, centrifuge is Figure 6. Loading history corresponding to cyclic load
spun to 100g, and cyclic loads are applied in a packets
load-controlled manner. Maximum and minimum
amplitudes, along with the number of cycles are 3 RESULTS AND DISCUSSION
pre-designed. 3.1. Installation response
(f) Extraction of the caisson is also done at In offshore suction caisson installation, a pump is
100g after all the tests using the loading actuator. It always used for pumping water out of the caisson.
is performed at a very slow rate of 0.001 mm/sec In laboratory conditions, various methods are
to determine the drained tensile capacity of the available for the application of suction required for
caisson. installation. Figure 6 compares the results obtained
A very low frequency of 0.028 Hz (at model from installation done using pump and gravity
scale) is used in this research for all the tests. This flow. Although the overall flow rate of gravity
led to the drainage conditions being predominantly flow is low, initial flow rate in gravity flow is
drained, which is determined using pore pressure 25.17 ml/sec, which is higher than the rate
measurements beneath the caisson. Drained tensile obtained in slow installation using pump. Thus,
capacity of almost 1700 kN in prototype scale is suction required for the onset of seepage is higher
obtained from all the extraction tests. in gravity flow than the suction required in slow
Compressive load amplitude is fixed based on installation. Therefore, as in the previous case,
the buckling capacity of the guide rod, using which required suction increases with increase in
load is applied on the caisson. Since compressive penetration depth. Thus, it can be concluded that
loads are not the main concern in this paper, higher the method of suction does not affect the
compressive loads are not used and 500 kN is fixed installation process as long as the necessary
as maximum compressive amplitude in one-way suction is applied.
266
3.3. Unloading stiffness

The unloading stiffness is by obtaining a slope
between the maximum load point of the previous
cycle and the minimum load point of the next cycle
as shown in Figure 9.
Figure 6. Normalized suction pressure acquired against the

normalized depth for the two installation methods
3.2. Vertical displacement

Under one-way compressive loading, no
significant settlement of the caisson is witnessed as
the compressive load applied was much smaller
compared to the bearing capacity of the foundation
(for this case the estimate is in excess of 1.5 MPa).
Under two-way asymmetric loading, compressive Figure 9. Load-displacement curves for the definition of
amplitude load is kept constant at 500 kN and the unloading stiffness
tensile load amplitude is increased. When tensile
loads are applied to the caisson, resistance to the 800
uplift of caisson is generated by the internal and
400
external skin friction. Uplift of caisson causes
shearing of sand at the caisson-sand interface, One-way
Load (kN)
0
Two-way
which further leads to loosening or restructuring of -400
sand. Thus, in the next compressive cycle,
-800
significant settlement of caisson is witnessed.
When the maximum tensile load is less than or -1200
equal to 50% of the drained tensile capacity (850
-30 -20 -10 0 10 20 30 40 50
kN), caisson uplifted up to 10 mm. Whereas, when Displacement (mm)
the tensile load exceeds 50% of the drained tensile Figure 10. Displacement behaviour under cyclic loading
capacity (1100 kN), larger uplift of caisson is (prototype scale)
witnessed, and the uplift increases by a factor of 2
(Figure 8). In all the results, positive denotes The normalized stiffness of the caisson is
settlement and negative denotes uplift of the obtained as shown in Figure 11. K1 is the reference
caisson. unloading stiffness for the first loading circle, and
Ku is the unloading stiffness for each loading circle.
60
One-way For one-way compressive loading, the unloading
Two-way
40
stiffness remains constant during the entire loading
duration. However, for two-way asymmetric
Displacement (mm)
20 loading, the unloading stiffness decreases with a

decreasing average load in general. Within each
0 cyclic amplitude packet, there is a slight increase
-20
in the unloading stiffness. When the average load
decreases, the stiffness behaviour shifts to a lower
-40 magnitude instantly and remains constant or shows
0 50 100 150 200 250 300
a slight increase in that packet. It can be noted that,
Number of cycle, n
at 250 cycles, there is a tremendous drop in the
Figure 8. Displacement behaviour under cyclic load packets
stiffness. This is attributed to the introduction of
tensile load which is higher than 50% of the
drained tensile capacity.
267
One-way
author has been funded by the China Scholarship
1.3
Two-way Council (CSC).
1.2
6 REFERENCES
1.1
Ku/K1
Bienen, B., Klinkvort, R.T., O’Loughlin, C.D., Zhu, F.,

1.0 Byrne, B.W., 2018a. Suction caissons in dense sand,
part I: Installation, limiting capacity and drainage.
0.9 Geotechnique 68, 937–952.
https://doi.org/10.1680/jgeot.16.P.281
0.8
0 50 100 150 200 250 300 350 Bienen, B., Klinkvort, R.T., O’Loughlin, C.D., Zhu, F.,
Number of cycle, n Byrne, B.W., 2018b. Suction caissons in dense sand,
Figure 11. Unloading stiffness behaviour under cyclic loading part II: vertical cyclic loading into tension.
Géotechnique 68, 953–967.
4 CONCLUSIONS https://doi.org/10.1680/jgeot.16.P.282
The results of an experimental study on cyclic
Bo Ibsen, L., Houlsby, G., Byrne, B., 2005. Suction
loading of suction caisson foundations in caissons for wind turbines, in: Frontiers in Offshore
homogenous sand have been described. The main Geotechnics. Taylor & Francis, p. 1110.
results can be summerised as follows: https://doi.org/10.1201/NOE0415390637.ch4
(a) Although suction installation in a real
scenario is done only using a pump, various De Jager, R.R., Maghsoudloo, A., Askarinejad, A.,
methods are available for the application of suction Molenkamp, F., 2017. Preliminary Results of
at laboratory scale. It is noted that change in the Instrumented Laboratory Flow Slides. Procedia
Engineering 175, 212–219.
installation method has no significant effect on the https://doi.org/10.1016/j.proeng.2017.01.012
installation process, as long as sufficient flow rate
is obtained. Houlsby, G.T., Byrne, B.W., 2005. Design procedures
(b) Tensile loads lead to an increase in the for installation of suction caissons in sand.
settlement of the suction caisson compared to the Proceedings of the Institution of Civil Engineers:
Geotechnical Engineering 158, 135–144.
settlement under one-way compressive load. Also, https://doi.org/10.1680/geng.2005.158.3.135
unloading stiffness decreases with a decreasing
average load and further remains either constant or Houlsby, G.T., Byrne, B.W., 2000. Suction Caisson
increased slightly throughout each cyclic load Foundations for Offshore Wind Turbines and
packet. Anemometer Masts. Wind Engineering 24, 249–255.
https://doi.org/10.1260/0309524001495611
(c) Significant uplift of caisson is witnessed
when the tensile load amplitude equals to 50% of Hung, L.C., Lee, S.H., Vicent, S., Kim, S.R., 2018. An
the drained tensile capacity after which, double the experimental investigation of the cyclic response of
amount of uplift displacement is observed when bucket foundations in soft clay under one-way cyclic
tensile load is larger than 50% of the tensile horizontal loads. Applied Ocean Research 71, 59–68.
capacity. Stiffness of the caisson instantly https://doi.org/10.1016/j.apor.2017.11.010
decreased with increasing tensile load. Jin, S.C., Zhang, Y.T., Wu, Q.H., 2012. A Study on the
(d) To understand the role of pore pressure in Failure Mechanism of Suction Caisson under Vertical
the response of the foundation, further centrifuge Load. Applied Mechanics and Materials 256–259,
tests can be done with samples prepared using 1985–1989.
viscous fluid to study the undrained or partially https://doi.org/10.4028/www.scientific.net/AMM.256
-259.1985
drained behaviour of sand.
Peire, K., Nonneman, H., Bosschem, E., 2009. Gravity
5 ACKNOWLEDGEMENTS Base Foundations for the Thornton Bank Offshore
This work is funded by the Section of Geo- Wind Farm. Terra et Aqua 115, 19–29.
Engineering, Delft University of Technology. One
268
An investigation into the performance of helical piles with fins under lateral
loading
Kevin Stone1, Maria Diakoumi1, Oliver Jeffcock1
Civil Engineering, University of Brighton, Brighton, U.K.
1
Corresponding author: Kevin Stone (kevin.stone@brighton.ac.uk)
ABSTRACT: Helical or screw piles have been used extensively with applications in a variety of
geotechnical foundations with renewed interest in their potential for offshore applications. This paper
presents a study of the performance of a helical pile with enhanced lateral capacity provided by fins
(‘screwfin’ pile). It is proposed the installation process of the screw pile is utilised to embed the fins into
the soil. In the model tests undertaken in this study the installation of the model piles is carried out prior to
conducting centrifuge tests on the lateral capacity of the pile. Consequently, full-installation effects are not
modelled, although there will be disturbance to the soil from the installation procedure carried out prior to
testing. These tests therefore investigate primarily the post-installation performance of model screw piles
with fins.
Keywords: Helical Piles; Screw Piles; Offshore Foundations; Lateral Capacity.
1 INTRODUCTION
This paper presents a centrifuge model study of the
performance of a helical or screw pile with
enhanced lateral capacity provided by fins (screwfin
pile). The screw pile potential for the Offshore
Wind industry is currently the focus of recent
research, see for example Al-Baghdadi et. al.
(2015), Davisson et. al. (2018), Richards et. al.
(2018). A potential disadvantage of a conventional
screw pile is the small cross section of the shaft.
Since the pile is installed by rotation, then the
smaller the pile shaft the less the shear resistance
developed as the shaft rotates against the soil during
installation which would result in less installation
torque and associated equipment. However, the
small shaft diameter means that the installed screw
pile is unable to develop any significant lateral
resistance. To overcome this limitation commercial
systems have been developed to retro-fit screw piles
with a system of fins or other enhancements to
provide additional lateral resistance (Stone and -a- -b-
Figure1. (a) Rotation of shaft and auger embeds the pile
Diakoumi, 2019). and (b) engagement of shaft collar embeds fin on
The screwfin pile is designed, that as the screw continued installation of pile.
pile penetrates the soil the pile shaft engages and
drags into the soil a set of fins to a predetermined The screwfin pile once installed is likely to behave
embedment depth. The fins are not rotated during in a similar manner to a finned-pile, a system which
installation and are installed as part of the screw pile has been reasonably well reported in the literature.
installation and not retrospectively. The screwfin Early studies using tapered fins reported by Lee and
pile is designed so that the fins are not rotated during Gilbert (1980) followed by later research has clearly
installation. The fins are not rotated during demonstrated the enhancement of pile lateral
installation An example of the installation process capacity through the use of fins under both
is illustrated schematically in Figure 1. However, it monotonic and cyclic loading, see for example
is also noted that the fins could be installed by (Sayles et al 2018, Bienen et al 2012, Duhrkop et al
jacking against the tensile capacity of the installed 2010, Nasr 2014, Rudolph & Grabe 2013, Duhrkop
screw pile. & Grabe 2008, Peng 2006, DTI 2006, Peng et al
269
2010). Several of these studies focused on the Figure 2 shows the model screwfin pile used in the
geometrical arrangement of the fins, (Sayles et al centrifuge tests.
2018, Peng 2006), in terms of length and projection
from the pile shaft. 2.2. Centrifuge Test Package
These studies have generally concluded that piles The tests were carried out on the balanced beam
with fins can provide a considerably higher lateral geotechnical centrifuge at the University of
resistance and stiffer behaviour than straight shafted Brighton. This machine is manufactured by
piles. A simple approach to the design of a finned Thomas Broadbent & Sons Ltd. and is a 6 g-tonne
pile is to consider the lateral load development as machine (20kg payload to 300g) with a 750mm
the sum of the lateral capacity of the pile plus an radius. The tests were conducted in a rectangular
additional contribution from the fin in terms of steel strongbox with dimensions 300mm (length),
passive soil resistance and to a lesser extent possible 155mm (width) and 180mm (depth). The steel
shear stress developed on the fin surface. The extent strongbox was placed in a cradle which was then
of these contributions will be a function of the fin mounted on the centrifuge arm. A two degree of
geometry and its location on the pile. For example, freedom actuator is then mounted on the cradle.
a centrifuge model study reported by Sayles et. al. Lateral loading is applied to the pile through the use
(2018) on monotonic lateral load performance of fin of a rod push mechanism attached to the horizontal
piles in dry sand, investigated only the effect of the actuator as shown in Figure 4. The rod is attached to
fin length on the pile performance, and whilst there a load cell fixed to the actuator traveller plate. The
is not complete agreement in the literature, it is horizontal displacement at the point of loading is
generally observed that the relationship between determined from the horizontal displacement of the
lateral load and fin length tends to decrease linearly actuator, and is also measured via a Linear Variable
up to a fin to pile length ratio of about 0.3, after Differential Transformer (LVDT) which is attached
which the efficiency reduces. to the pile at the point of loading.
2 MATERIALS AND MODEL

PREPARATION
The sand used for the model tests was uniformly
graded Fraction C (300-600 micron) silica sand
supplied by David Ball Limited. The sand has
maximum and minimum void ratios of 1.06 and
0.61 respectively, with a critical state angle of
friction φcrit of 31°. The models were prepared
though a combination of dry pluviation and
vibration using a vibrating table. This method
produced consistent soil specimens with a bulk
density of 1725kg/m3.
2.1. Model Piles

All the model pile shafts were formed from a solid
9.5mm diameter aluminium rod. The auger flights
were made from aluminium plate curved into a
helical shape and then glued onto a circular section
which can be screwed into the pile tip. For the Figure 2. Photograph of model screwfin pile. The fins are
screwfin pile, the pile shaft is reduced to a 4mm free to rotate on the pile shaft and are pushed into the soil
diameter section and passed through a hollow by the widened section of shaft.
section of 9.5mm rod with the fin attached to this
section. Above this section the shaft widens to form The model screw pile and screwfin piles were
a shoulder. As the main shaft rotates and embeds installed into the soil approximately 55mm from the
itself into the soil the shoulder of the shaft extending edge of the container. A special former was utilised
contacts the finned section and pulls it into the soil to allow the pile to rotate and embed itself without
as the pile continues to embed itself, refer to Figure the fins rotating and hence overly disturbing the
1. Consequently, the fins are installed as part of the soil. Installation was undertaken by hand prior to
screw pile installation and not retrospectively. mounting the actuator onto the strong box. Great
270
care was taken during installation to ensure
verticality of the model piles and since the pile is
displaced away from the tub walls and towards the
centre of the sample, boundary effects are not
considered significant. The tests listed in Table 1
were conducted at an acceleration level of 25g to
investigate performance of a screwfin pile over that
of a monopile or simple screw pile. The pile
embedment depth was maintained for all the tests at
90mm and the load was applied at an elevation of
32mm above the soil surface.
(a) (b)
Figure 3a) Open fin arrangement and b) closed fin Figure 4. Model package (post-test).
arrangement; the arrow indicates the direction of the screw-
fin pile movement.
Table 1. Summary of centrifuge tests at 25g The experimental results are presented through plots
Test ID Pile type Shaft Embedded of lateral load versus lateral displacement measured
Diameter Length at the point of load application. Figure 5 presents a
D (mm) L (mm) summary of the centrifuge model tests where the
MP Pile only 9.5 90 response of the monopile (MP) and simple screw
SP Screwpile only 9.5 90 pile (SP) is compared to the lateral response of the
SFP90OV Screwfin open 9.5 90 screwfin pile (SFP) for two different orientations of
SFP90CO Screwfin closed 9.5 90
the fin arrangement. The two fin orientations to the
SFP110OV Screwfin open 9.5 110
SFP110CV Screwfin closed 9.5 110 direction of movement were with the fins forming
either an open or closed ‘V’ in the direction of
movement into the soil, refer to Figure 3. These fin
2.2.1 Centrifuge Test Procedure
orientations are denoted respectively OV and CV
After the model and actuator assemblage have been representing either the open or closed arrangement.
set-up on the laboratory bench the package was then The results presented in Figure 5 indicate that the
loaded onto the centrifuge. The actuator was addition of the auger plate provides some
connected to the power and control system and a enhancement of the lateral capacity of the pile
digital video camera was mounted on the top of the which can be associated with the bearing resistance
cradle to observe the model during the test, refer to of the auger which is analogous to a buried
Figure 4. All the tests were conducted at an horizontal plate fixed to the pile shaft. For the pile
acceleration level of 25g. to rotate in the vertical plane the plate must also
Once the model had achieved the target test rotate in the horizontal plane and thus develops a
acceleration the actuator was run at a velocity of 3 resisting moment and point of fixity to the pile. It is
mm/minute and data from the load cell and actuator thus apparent that this additional resistance would
encoders were recorded continuously until the end be a function of the plate diameter (auger diameter)
of the test. and its location (depth) on the shaft. For all the tests
reported herein the same configuration was used
with regard to screwplate size and location.
271
For the screwfin pile tests with the 90mm long
embedment (from base of pile to rear of fins) an 160 MP
additional increase in lateral capacity is observed. It SP
is interesting to note that the load-displacement 140 SFP90OV
plots for these tests are very similar and suggests SFP90CV
120 SFP110OV
that the orientation of the fins to the direction of
travel does not have a significant effect. However SFP110CV
Lateral load (N)

100
comparison between the 90mm and 110mm
embedment lengths as shown in Figure 6 suggests 80
that for this case the closed V arrangement provides
a greater lateral resistance than the open V 60
configuration. The reasons for this are not entirely
clear and more experimental work is required to 40
verify this result.
With regard to the performance of the system at 20
serviceability loads the data has been plotted up to
0
4mm of lateral displacement to observe the
performance at smaller displacements. Referring to 0 1 2 3
Figure 7 it is apparent that the lateral stiffness of the Lateral displacement (mm)
monopile is initially enhanced by the presence of the
Figure 7. Initial lateral load versus displacement plots.
screw plate and then further enhanced by the action
of the fins. For all the screwfin pile tests relatively
The results and observations are consistent with
similar initial lateral stiffness responses are
those associated with the response of piles with fins
observed although it is generally observed that the
or wings. Therefore, once the screwfin pile is
longer 110mm embedment does provides an
installed, it can be treated as a finned pile. There are
enhanced stiffness over the 90mm depth of
also other issues that would require to be addressed
embedment.
which are beyond the scope of this study, for
The results presented herein represent an initial
example the disturbance of the soil resulting from
set of experimental results as a ‘proof of concept’
the installation process, and the optimisation of fin
study of screwfin piles.
geometry to allow them to be carried into the soil by
the action of the screw pile installation.
400
SFP90CV
4 CONCLUSIONS
350 SFP90OV
Based on the relatively limited information
SFP110OV presented, it can nevertheless be concluded that a
300 SFP110CV ‘finned’ screw pile (screwfin pile) where the
installation process of the screw pile is utilised to
Lateral load (N)
250
install a set of fins integral with the pile shaft, can
200 significantly improve the lateral capacity of the
screw pile. It is apparent that there remains much to
150 investigate with regard to the application of the
system and in particular the limitations based on the
100 ability to actually install the system and the effects
of the installation process on the ground properties
50 and lateral performance. Considerably more
research is required to investigate these issues and
0 determine the viability of the system.
0 5 10
Lateral displacement (mm) 5 ACKNOWLEDGEMENTS
The authors would like to thank Mr. David Harker
Figure 6. Lateral load versus displacement plots for 90mm for fabricating the models and assisting in running
and 100mm embedded screwfin piles.
the centrifuge.
272
6 REFERENCES
Lee, P.Y. and Gilbert, L. W. 1980. The Behaviour of
Al-Baghdadi, T., Brown, M.J., Knappett, J. A. &
Steel Rocket Shaped Pile. In Proceedings of
Ishikura, R. 2015. Modelling of laterally loaded screw
Symposium of Deep Foundations. Atlanta, U.S.A.,
piles with large helical plates in sand. In Proceedings
October, 1979, ASCE, New York, pp. 253-273.
of the 3rd International Symposium on Frontiers in
Offshore Geotechnics, Oslo, Norway, 10-12 June
Nasr, A. 2014. Experimental and theoretical studies of
2015. Taylor & Francis Group, London, pp. 503-508.
laterally loaded finned piles in sand. Canadian
Geotechnical Journal, 51(4):381-393.
Bienen B, Dührkop J, Grabe J, Randolph MF, White D.
2012. Response of piles with wings to monotonic and
Peng, J. 2006. Behaviour of Finned Piles in Sand under
cyclic lateral loading in sand. Journal of Geotechnical
Lateral Loading. PhD thesis. Newcastle University,
and Geoenvironmental Engineering 138(3): 364-375.
U.K., 2006
Davidson, C., Al-Baghadadi, T., Brown, M.J., Knappett,
Peng, J., Rouainia, M. and Clarke, B. G. 2010. Finite
J., Brennan, Augarde, C.E., A., Wang, L., Coombs,
element analysis of laterally loaded fin piles,
W.M., Richards, D. & Blake, A. & Ball, J. 2018. A
Computers and Structures. 88(21-22): pp. 1239–1247.
modified CPT based installation torque prediction for
large screw piles in sand. In Proceedings of the 4th
Richards, D.J, Blake, A.P., White, D.J, Bittar, E.M, &
International Symposium on Cone Penetration
Lehane, B.M. 2018. Field tests assessing the
Testing, 21-22 June 2018, Delft, Netherlands, pp. 255-
installation performance of screw pile geometries
261.
optimised for offshore wind applications. In
Proceedings of the 1st International Symposium on
Davidson, C., Al-Baghadadi, T., Brown, M.J., Knappett,
Screw Piles for Energy Applications, Dundee, United
J., Brennan, Augarde, C.E., A., Wang, L., Coombs,
Kingdom, 27-28 May 2019, pp. 47-54.
W.M., Richards, D. & Blake, A. 2018. Centrifuge
modelling of optimised screw piles for offshore wind
Rudolph C. & Grabe J. 2013. Laterally loaded piles with
energy foundations. In Proceedings of the 9th
wings: In situ testing with cyclic loading from varying
International Conference on Physical Modelling in
directions. In Proceedings of the ASME 32nd
Geotechnics, London, U.K., 17-20 July 2018. Taylor
International Conference on Ocean, Offshore and
Francis Group, London, pp. 695-700.
Arctic Engineering, Nantes, France, 9-14 June 2013.
ASME, New York, pp.
Department of Trade and Industry (DTI). 2006. “Finpile
Project: Final Report,” London: Department of Trade
Sayles S., Stone K.J., Diakoumi M. and Richards D.
and Industry.
2018. Centrifuge model testing of Fin Piles in sand . In
Proceedings of the 9th International Conference on
Duhrkop, J. & Grabe, J. 2008. Laterally Loaded Piles
Physical Modelling in Geotechnics, City University
with Bulge. Journal of Offshore Mechanics and Arctic
London, U.K., 17-20 July 2018. Taylor Francis Group,
Engineering, 130(4): 041602.
London.
Duhrkop, J., Grabe, J., Bienen, B., White,
Stone K.J. & Diakoumi M. (2019) Performance of screw
D. & Randolph, M. 2010. Centrifuge experiments on
piles enhanced for lateral load capacity. In Proceedings
laterally loaded piles with wings. In Proceedings of the
of the XVII European Conference on. Soil Mechanics
7th International Conference on Physical Modelling in
and Geotechnical Engineering, Reykjavik, Iceland, 1-6
Geotechnics, Zurich, Switzerland, 28 June-1 July
September 2019, pp. 551-556.
2010. Taylor & Francis Group, London pp. 919-924
273
Pile test database on centrifuged models: tension loading for floating wind turbine
anchor
Luc Thorel1, Ismat El Haffar1,2, Matthieu Blanc1
1
GERS-GMG, Univ. Gustave Eiffel, IFSTTAR, F-44344 Bouguenais, France
2
Fugro GeoConsulting, Bruxelles, Belgique
Corresponding author: Luc Thorel (luc.thorel@ifsttar.fr)
ABSTRACT: The memory of series of centrifuge tests performed in Nantes is being to come accessible
through an open database. The main parameters of the database (piles-anchors in sand) which includes 190
tests on small-scale centrifuged models of piles subjected to tension loading are presented: density,
saturation, installation, embedment and loadings (monotonic, cyclic or combined).
Keywords: Close-Ended Pile; Helical Pile; Tension; Sand; Centrifuge Models.
1 INTRODUCTION 2 MATERIALS
As part of the energy transition and the development The experiments were conducted in the
of marine renewable energies, the Ifsttar began in geotechnical centrifuge of the Univ. Gustave Eiffel.
2017, in response to a call for proposals from the These are models of smooth piles installed in sand
Weamec, the REDENV-EOL research programme, that is usually very dense, in flight (that is, under
which deals with the Reduction of the conditions of stress similar to those observed in real
Environmental Impact of Floating Wind Turbine size), and then subjected to a coaxial tensile load.
anchored with Pile Foundations. The idea here is to Figure 2 shows the example of a test on a closed
reduce the length of the anchor lines, favouring the cylindrical pile (D=18mm) in saturated dense sand,
tension lines that make the anchor pile work in including the driving phase (with a hydraulic jack in
traction, mobilizing the vertical resistance of the flight) and then the loading phase in tension. The
pile-soil interaction. procedure is detailed in El Haffar et al. (2017).
190 tensile tests on close-ended cylindrical piles
or helical piles have been carried out and grouped in Head Force (without pile self weight) [N]
a database, which is presented here. Four smooth -2000 3000 8000
piles (Figure 1) installed in flight at 100g in 0
Fontainebleau Sand NE34, dense or loose, dry or Fontainebleau sand NE34
saturated, are then subjected to vertical tension. Five 50 very dense (ID=99%), saturated
Displacement [mm]
single helix helical piles (Figure 1) of different

100
diameter are installed in flight by screw driving at
10 different relative depths, in Hostun sand HN38, 150
dense or loose, dry or saturated. The installation
forces are measured. The helical piles are subjected 200
to monotonous, cyclic or even monotone post-cyclic unloading
loads. All these tests are grouped in a database 250
whose detailed characteristics are presented here. tension
IELT-08-P4S16C-100-NE34VD-M OV
Figure 2 Displacement stress curve (model sizes) of a closed
cylindrical smooth pile installed in very dense saturated sand
and then loaded in traction (D=18mm, according to El
Haffar, 2018).
For helical piles, the procedure, described by

Schiavon et al. (2018), includes a controlled pole
movement and rotation installation phase, followed
by monotonous or cyclic loading phases (Figure 3).
Table 1 presents the parameters of geometry, scale
(a scale model 1/N is tested under a centrifugal
Figure 1 Model piles. Example of single helical steel helical acceleration of N times the acceleration of earth
pile (left, Schiavon et al. 2018) and closed cylindrical gravity), and loading applied. Some tests were
aluminum pile (D=18mm, length 250mm, El Haffar, 2018). carried out in dry sand and others in saturated sand.
275
Figure 3 Example of the evolution of axial force versus time of a helical pile test, including installation, cyclic loading, and post-
cyclic monotone phases (Schiavon et al., 2018).
Table 1 Distribution of geometries, densities and loading types of the 190 tests.
Pile Number of tests Loading

Relative
D Scale Sand
Type depth dry saturated m c m-c m-c-pc c-pc
[mm]
z/D
12 20.83 1/100 NE34 1 MD 1+1 MD 3 0 0 0 0
14 17.86 1/100 NE34 1+1 MD 1+1 MD 4 0 0 0 0
Close ended
16 15.63 1/100 NE34 1+1 MD 1+1 MD 4 0 0 0 0
18 13,89 1/100 NE34 1+1 MD 1+1 MD 4 0 0 0 0
40 6 1/10 HN38 4 9 0 0 0 6 7
6 - 7.4 - 7.8
33 1/10 HN38 20 41 5 10 1 7 38
-8
Mono-helix 6 - 8 - 9.6 -
25 1/10 HN38 12 36 3 0 1 17 27
d=10mm 10 - 12 - 14
6 - 8 - 10 -
18 1/10 HN38 15 17 10 0 7 11 4
10.5 - 13.3
13 6 - 8 - 18.5 1/10 HN38 5 3 4 0 3 1 0
Mono-helix
instrumented 33 7.4 1/10 HN38 13 0 0 0 0 0 13
d=10mm
MD : medium dense / D : helix or cylindrical pile diameter / d : diameter of the helical pile’s shaft
m : monotonic initial / c : cyclic / pc : monotonic post-cyclic
The helix geometry is specified in Table 2 Table 2 Helical pile’s geometry

(Schiavon, 2018). The relative depth of the tip of the Helical pile D/d D (mm) d (mm) th (mm) φh (º)
closed cylindrical piles takes four different values HA13 1.3 13 10 1.0 6
shown in a balanced way, while 11 helix depths HA18 1.8 18 10 1.0 6
HA25 2.5 25 10 1.5 6
have been tested, four of them representing almost
HA33 3.3 33 10 1.5 6
the entire population (Figure 4). The relative depths HA40 4.0 40 10 2.0 6
tested make it possible to scan several D : Helix’s diametre / d : shaft’s diametre / th :
configurations, but also, for some of them to test Helix’s thickness / φh : Helix’s angle
different combinations and loading intensities.
276
Close-ended cylindrical pile intensity of the load, between 3 and 10,000 cycles.
5
The maximum observed displacement reaches 1.5D
and the SLS criterion used corresponds to a
4 4 4
4 cumulative displacement of 0.1D. The frequency of
the sine signal applied is not the parameter studied
3
3 in priority, the idea being to be within the
n test
framework of quasi-static loads. Only 3 frequencies

2 (model size) were tested: 0.3Hz (56% of cases), 0.4
Hz (1%) in saturated sand and 1Hz (43%) in dry
1 sand.
V
0 Vmax Vmax = Vm+Vc
13,89
13.89 15,63 17,86
15.63 17.86 20,83
20.53 Vmin = Vm-Vc
z/D [-]
Vc
Helical pile
60 Vm
52
50 45
39 t
40 Vmin
n test
30
20 (a)
20
Vc/Vut
100%
10 5 5
3 1 1 2 2 Cyclic loading
0 80%
7,4
7,8
9,6
66
88
1212
1414
1010
10,5
13,3
18,5
7.4
7.8
9.6
10;5
13.3
18;5
60%
z/D [-]
Figure 4 Relative depth: Close-ended cylindrical piles (top) 40%
and buried helical pile’s helix (bottom).
20%
3 CYCLIC LOADING
Cyclic loads are controlled in force and take the 0%
0% 20% 40% 60% 80% 100%
form of a temporal sinus. Thus, the load is
characterized by its frequency and by two Vm/Vut
(b)
parameters: the mean vertical load Vm and the cyclic Figure 5 Cyclic loads: shape of sinusoidal loading (a) and
half-amplitude Vc (Figure 5a). By introducing the combinations of the different loads applied (b).
ultimate tensile Vut load, we can represent in an
dimensionless diagram all the cyclic loading 40
37
conditions of the experiments (Figure 5b). 35
Most cyclic tests (92%) follow tension loads
(Vm>Vc) and some alternate loads are also included.
30
26
The distribution of the stress types according to the 25
N tests
Vc/Vm ratio (Figure 6) shows a wide variety, as well 20
as 3 predominant values which correspond to an 15 14

alignment of the test points in Figure 5 according to 10
a linear function (for example, with a slope of 0.5). 10
One of the objectives of studying the behaviour of 3 4 4 5 54

2 321111 11211211322111111
5
1 111 2 2 2 2
1 1 2 2
a cyclically charged pile is to identify the 0
movements accumulated during the cycles that

could potentially lead to failure. The number of Vc/Vm [-]
cycles applied varies greatly depending on the Figure 6 Distribution of the cyclic load to average load ratio
277
4 DATABASE STRUCTURE AND - Centrifuge acceleration of 100g
NUMERICAL DATA corresponding to a test at scale 1/100
The structuring of the base (Figure 7) uses other - Fontainebleau sand NE34 Very Dense
files containing detailed information which are not - Monotonic Vertical Loading
in the data file stricto sensu, such as the The data for each test of the database will be
characteristics of the soil used, the nature of the recorded in a « dataverse » system, with a free
experimental campaign, the properties of the pile, access for tests already published, and identified
the experimental assembly or loading conditions. with a DOI number: https://research-
The file of each test follows a particular data.ifsttar.fr/dataverse/piles-anchors_in_sand.
organization, starting with its name, which follows
a specific nomenclature (Figure 8). The termination 5 CONCLUSIONS
depends on the type of software used to edit the data An overview of a new database of experiments
(e.g. .txt or .xls). Each file contains in the first carried out in the geotechnical centrifuge of the
column the time, then in the other columns the Université Gustave Eiffel, Campus de Nantes, has
recorded data. been presented here. All the tests carried out relate
to traction loading. This base includes closed
cylindrical piles under monotonous load and mono-
helical piles under monotonous or cyclic loads. All
model piles are installed in macro gravity.
6 ACKNOWLEDGEMENTS
The authors would like to thank WEAMEC
(Projects REDENV-EOL and JE-CORECT), the
Pays de la Loire Region, Nantes Metropole and
CARENE for their financial support, which enabled
the study to be carried out. Special thanks go to the
Gustave Eiffel University centrifuge team for their
technical support and assistance during these
experimental campaigns.
Figure 7 Database structure 7 REFERENCES
First name’s initial El Haffar I., Blanc M., Thorel L. 2017. Impact of pile

installation method on the axial capacity in sand.
 Name’s initial of the researcher
- Géotechnique Letters, 7, 260-265.
 http://dx.doi.org/10.1680/jgele.17.00036
- N° container
 Type of pile El Haffar 2018. Physical Modelling and study of the
- behavior of deep foundations of offshore wind
 g-level turbines. PhD thesis Univ. Bretagne Loire, Sciences
- pour l’Ingénieur.
 Soil
- Schiavon J.A., Tsuha C., Neel A., Thorel L. 2018.
 Loading Centrifuge modelling of a helical anchor under
different cyclic loading conditions in sand.
Figure 8 Nomenclature of the files’ name. Each rectangle  International journal on physical modelling in
corresponds to a character
geotechnics; 19(2): 72–88.
https://doi.org/10.1680/jphmg.17.00054
The file corresponding to the test presented on
figure 2 is named IELT-08-P4S16C-100-NE34VD- Schiavon J.A. 2018. Effect of the helix-to-shaft diameter
MOV corresponding to the following informations: ratio on the behaviour of single-helix anchors
- Ismat EL Haffar Thèse subjected to cyclic loading. Scientific mission report.
- Container (or strongbox) n°08 IFSTTAR-USP Partenariat structurant international
- Pile n°P4, Smooth, 16mm in diameter, HELICAL PILE. 108p.
Closed at the tip
278
6. Extended abstracts
Experimental study of the interaction of shallow foundations with an emerging

strike-slip fault rupture
Athanasios Agalianos1, Evangelia Korre2, Tarek H. Abdoun2, Ioannis Anastasopoulos1
Institute of Geotechnical Engineering, ETH Zurich, Switzerland
1
Dept. of Civil and Environmental Engineering, Rensselaer Polytechnic Institute, Troy, NY, USA
2
Corresponding author: Athanasios Agalianos (agaliana@ethz.ch)
Keywords: Strike-Slip Fault; Shallow Foundations; Centrifuge Testing; Fault-Rupture-Soil-Foundation Interaction.
1 INTRODUCTION
Several case histories of structures such as 0.2 m
buildings, bridges, dams, tunnels, or pipelines
subjected to tectonic deformation have been
recorded during past seismic events, most
commonly associated with severe damage or Rupture
collapse. Nevertheless, there have been also cases
of satisfactory performance. Foundation continuity
and rigidity have been shown to play a crucial role
in the survival of structures, while diversion of the
fault rupture is possible due to the presence of a
heavily loaded structure lying on compliant soil.
Several researchers have studied experimentally
Figure 1. 3-section split container at RPI centrifuge lab.
the interaction of propagating dip-slip fault ruptures
with foundation–structure systems. A significant
Initially, the free-field fault rupture propagation
amount of centrifuge model tests has been
through the dense sand layer was examined (in the
performed, focusing on the interaction with shallow
absence of the footings) and used as reference case.
foundations (Bransby et al. 2008), embedded
Subsequently, 2 footings were placed at the ground
caisson foundations (Loli et al. 2012) and piled
surface to examine their interaction with the
foundations (Cai and Ng 2016). With respect to
outcropping faults. The parameters that were
strike-slip faulting, most studies have examined the
parametrically varied in this study to examine their
response of buried pipelines subjected to earthquake
effect on the system response were: (a) the relative
faulting (Abdoun et al. 2009). Aiming to bridge the
footing location with respect to the fault trace; (b)
apparent gap in the literature, this paper studies
the foundation surcharge load; (c) the foundation
experimentally the interaction of slab foundations
aspect ratio; and (d) the foundation stiffness. A
with emerging strike–slip faults on dense sand.
schematic illustration of the conducted centrifuge
tests is shown in Figure 2, while the key parameters
2 EXPERIMENTAL SETUP of each test are collected in Table 1.
Five centrifuge model tests were conducted at the
Rensselaer Polytechnic Institute (RPI) geotechnical 1.14
centrifuge facility, utilizing a 3-section split 0.38 0.38 0.38
container (Figure 1). The latter is capable of

reproducing 2 parallel strike-slip fault ruptures
simultaneously, resulting in testing of 10 different
0.33
b = 0.05
prototype problems. All centrifuge tests were
l = 0.05/0.25
performed at 75g centrifugal acceleration. Dry 0.33 0.355

0.71
Nevada 120 sand was used, which was deposited in

the split container using the air pluviation method.
In this way, a uniform 7.5 m (in prototype scale) s = 0.025/0.035
Fault rupture
Fault rupture
0.33
thick sand layer of dense sand (Dr ≈ 85%) was

y
achieved. The middle section of the split container
x
was then subjected in a quasi-static manner to a total
horizontal displacement of 1.5 m (prototype scale) Total applied offset: 0.02 m
with respect to the two stationary edge sections. Figure 2. Schematic illustration of the centrifuge tests.
281
Table 1. Key foundation parameters of the 4 centrifuge tests. test where the square footings were positioned with
Test 1 2 3 4 an eccentricity with respect to the fault parallel axis
Fault location, (s/b = 0.7) and for the test with the longer footings
0.5 0.7 0.5 0.7
s/b (-) (b/l = 0.2). In both cases the rotational resistance of
Surcharge load,
27/105 27/105 21/98 21/94 the practically rigid footings was large enough to
q(kPa)
Aspect ratio, facilitate the diversion of the fault rupture, which
1 1 0.2 1 always follows the minimum energy path. Each
b/l (-)
Stiffness (-) Rigid Rigid Rigid Flexible mechanism leads to different foundation distress
(not shown herein).
An additional test was then conducted,
3 KEY RESULTS AND DISCUSSION considering two idealized flexible foundations (by
Regarding the deformation field at the ground cutting them in multiple small pieces), to examine
surface under free-field conditions derived from the influence of foundation rigidity. The flexible
image analysis (not presented herein), there is square footings were placed eccentrically with
diffusion of the fault-induced deformations, without respect to the fault parallel axis (s/b = 0.7). As
evidence of significant strain localization at the 2 opposed to the corresponding rigid foundations test,
fault rupture locations. As also confirmed through where fault diversion was observed, in this case the
Finite Element analyses, the surface deformation fault rupture propagated through the foundations.
field due to an emerging strike-slip fault depends This confirmed the importance of foundation
strongly on the test boundary conditions. The rigidity, which affects its rotational resistance.
interface friction at the base and sides of the split-
box can alter significantly the results. In general, 4 CONCLUSIONS
rough base and side interfaces lead to more realistic
The present study examined experimentally the
results, with vertical propagation of the fault rupture
interaction of slab foundations with an emerging
and development of a shear band parallel to the fault
strike-slip fault. The split container boundary
trace due to strain localization of dense sand.
conditions was shown to affect significantly the
The fault rupture-soil-foundation interaction (FR-
strike-slip fault rupture propagation under free-field
SFI) tests revealed the existence of two distinct
conditions. Regarding the FR-SFI tests, the system
kinematic mechanisms (Figure 3): a rotational,
response was mainly controlled by the foundation,
where the foundation rotation around the vertical z
acting as a kinematic constraint, with the surcharge
axis θz increases proportionally to the fault offset h;
load playing a secondary role. Such investigations
and a translational, where θz stops increasing and
provide useful insights on the response of surface
the foundation either displaces horizontally parallel
foundations subjected to strike-slip faulting,
to the fault trace or remains almost stationary,
facilitating thus the development of appropriate
depending on whether most of its footprint lies on
design methods against large tectonic deformations.
the moving or the stationary block.
5 5 REFERENCES
s/b = 0.5
Abdoun, T.H., Ha, D., O’Rourke, M.J., Symans, M.D.,
4 s/b = 0.7
O’Rourke, T.D., Palmer, M.C., Stewart, H.E. 2009.
3 Factors influencing the behavior of buried pipelines
θz (deg)
subjected to earthquake faulting. Soil Dynamics and

2 Earthquake Engineering, 29(3): 415-427.
1 Bransby, M.F., Davies, M.C.R. and El Nahas, A. 2008.
0
Centrifuge modelling of normal fault–foundation
0 0.3 0.6 0.9 1.2 1.5 interaction. Bulletin of Earthquake Eng, 6(4): 585-605.
h (m)
Figure 3. Foundation rotation θz as function of fault offset h. Cai, Q.P. and Ng, C.W.W. 2016. Centrifuge Modeling of
Pile-Sand Interaction Induced by Normal Faulting.
The rotational mechanism develops when the fault Journal of Geotechnical and Geoenvironmental
Engineering, 142(10): 04016046.
rupture emerges vertically at the ground surface, as
observed in the test where the square footings (b/l = Loli, M., Bransby, M.F., Anastasopoulos, I., Gazetas, G.
1) were centrally placed relative to the free-field 2012. Interaction of Caisson Foundations with a
fault trace (s/b = 0.5). In contrast, the translational Seismically Rupturing Normal Fault: Centrifuge
mechanism develops when the fault rupture deviates Testing versus Numerical Simulation. Géotechnique,
from its free-field trace. This was observed for the 62(1): 29-43.
282
Investigation of the dynamic behavior of gravity-type quay wall under saturated

condition
Yeon-Sam Kim1, Kil-Wan Ko1, Dong-Soo Kim1

Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology,
1
Corresponding author: Dong-Soo Kim (dskim@kaist.ac.kr)
Keywords: Quay wall; Soil-Structure Interaction; Saturation System; Liquefaction.
1 INTRODUCTION Physical properties of the silica sand are presented

in Table 1.
After the port of Kobe was severely damaged during
the 1995 Kobe Earthquake, interest in the behavior Table 1. Physical properties of the silica sand
of the port structures has increased (Inagaki et al. Property Value
1996). Damages of port and harbor structures were Maximum unit weight, 𝛾𝛾𝑚𝑚𝑚𝑚𝑚𝑚 (t/m3) 1.645
observed during 2017 Pohang earthquake. Minimum unit weight, 𝛾𝛾𝑚𝑚𝑚𝑚𝑚𝑚 (t/m3) 1.244
Therefore, an increased need for evaluation of the
Specific gravity, 𝐺𝐺𝑠𝑠 2.65
dynamic behavior of the port and harbor structures
Coefficient of curvature, 𝐶𝐶𝑐𝑐 1.11
has arisen in Korea. Since one of the most common
type of port and harbor structures in Korea is a Uniformity coefficient, 𝐶𝐶𝑢𝑢 1.96
gravity-type quay wall, this study focused on the
quay wall. Complex interaction between water, soil, 2.1.2 Structure modeling
and quay wall results in the dynamic response of the
The gravity-type quay wall was modelled in a
soil-structure system, so that a dynamic centrifuge
cuboid shape to simplify its shape. The dimension
test is one of the most useful method (Lee 2005). To
and weight of the structure was determined to
evaluate its complicated dynamic behavior,
achieve kh value of 0.13, based on the design
dynamic centrifuge test was performed under fully
standard provided by the Korean Ministry of
saturated soil. The objective of this study is to
Oceans and Fisheries (2014).
evaluate the dynamic behavior of a gravity-type
quay wall under a strong earthquake. Liquefaction
2.2. Model saturation
occurrence, permanent settlement caused by soil-
structure interaction will be evaluated and the After preparing a dry testing model, viscous fluid
sliding mechanism of the quay wall will be which is a mixture of methylcellulose and water was
investigated throughout the dynamic centrifuge test. used to saturate the model. 2 % mass ratio of water
and methylcellulose was used to achieve a target
2 CENTRIFUGE TESTING PROGRAM viscosity of 60 cSt, according to the scaling law.
The degree of saturation of the testing model was
Dynamic centrifuge test was conducted at the over than 99 % by applying Okamura method
KOCED Geotechnical Centrifuge Testing Center at
(Okamura and Inoue 2014).
KAIST. The centrifugal acceleration was 60 g-level.
Further descriptions and test results are expressed in
2.3. Dynamic loading
prototype scale.
To investigate a dynamic behavior of the quay wall,
2.1. Testing model a strong 2 Hz sinusoidal wave was applied to the
2.1.1 Ground modeling testing model, which had a duration of 15 seconds
and 0.3 g of peak acceleration at bedrock.
The testing model was constructed in a rigid model
container. Silica sand (SP) was used to construct the 3 RESULTS AND DISCUSSION
whole testing model. Sand pluviation method was
used to achieve a target relative density of 50 %. The testing results will be described in three parts:
liquefaction, settlement, and sliding behavior of the
quay wall.
283
3.1. Liquefaction water pressure remained its positive value.
Consequently, inertial force of quay wall, which
Excess pore water pressure time history at the top was generated by the dynamic loading, governed the
layer of far-field is shown in Figure 1. In the upper failure.
layer of far-field backfill soil, liquefaction was
observed. Maximum value of excess pore water
pressure remained for about 75 seconds. Total
dissipation time was about 8 minutes. Minus peaks
were observed in acceleration records.
Figure 1. Excess pore water pressure time history at the top

layer of far-field soil.
3.2. Soil settlement Figure 3. Acceleration time histories of bedrock and quay -
wall with its cumulative displacement.
In near-field, the interaction between the quay wall
and backfill soil played an important role of causing
soil settlement. Figure. 2 shows the interaction 4 CONCLUSION
between near-field soil and quay wall resulting in In this paper, dynamic behavior of quay wall was
soil settlement. When the displacement time evaluated by the centrifuge test. Important findings
histories of quay wall and near-field soil were out of in this study are summarized as follow.
phase, soil and the quay wall repetitively moved
(1) Liquefaction occurred only in the upper layer
against and towards each other. During the
of far-field backfill soil.
repetitive interaction, excess pore water pressure
value fluctuated, making water to move inside and (2) Settlement in the near-field ground was due to
to be pushed out. This interaction resulted in the the interaction between water, soil, and quay
cumulative settlement of near-field ground. wall structure.
(3) Sliding failure of the quay wall was mainly
due to the inertial force of quay wall.
5 ACKNOWLEDGEMENTS
This research was supported by a grant (18CTAP-
B139287-02) from Technology Advancement
Research Program (TARP) funded by Land,
Infrastructure and Transport of Korean government.
6 REFERENCES
Inagaki H, Iai S, Sugano T, Yamazaki H, and Inatomi T.
1996. Performance of caisson type quay walls at Kobe
Port, Soils and Foundations, Special Issue, pp. 119-
136.
Figure 2. Interaction between near-field soil and quay wall
resulting in soil settlement.
Lee, C. J. 2005. Centrifuge modeling of the behavior of
caisson-type quay walls during earthquakes. Soil
3.3. Sliding failure Dynamics and Earthquake Engineering, 25(2), 117–
Figure. 3 shows acceleration time histories of the 131.
bedrock and quay wall, with its cumulative
Okamura, M., & Inoue, T. 2012. Preparation of fully
displacement records. Sliding of the quay wall saturated models for liquefaction study. International
occurred with 45 cm displacement, which lead to Journal of Physical Modelling in Geotechnics, 12(1),
failure. The acceleration of quay wall stopped after 39–46.
the end of the shake, even though the excess pore
284
Design of a Miniature Cone Penetrometer for the Geotechnical Centrifuge of Delft

University of Technology
Dirk de Lange1,2, Siavash Honardar1, Amin Askarinejad1
Geo-engineering, Delft University of Technology, The Netherlands
1
Geo-engineering, Deltares, The Netherlands

2
Corresponding author: Dirk de Lange (Dirk.deLange@deltares.nl)
Keywords: Miniature Cone Penetrometer; Design Considerations; Geotechnical Centrifuge.
1 INTRODUCTION samples. Considering such a limiting ratio, the

Cone penetration testing (CPT) has proven to be one design proposal depends heavily on the dimensions
of the most significant forms of in-situ geotechnical of the testing containers and the centrifuge platform.
site investigation techniques. Different types of Two types of containers are taken into consideration
cone penetrometers have been designed. It has been with an inner width or diameter of 150 and 300 mm,
recognized that several factors affect the respectively.
measurements conducted by a CPT (Lunne et al, Another factor that is considered is the actuator
1997). The most general resultant measures from a and loading frame utilized in the DUT centrifuge.
CPT test consist of the cone resistance qc, the sleeve The current maximum loading range of this actuator
friction fs and often the pore pressure u. Even though is approximately 3.5 kN and the maximum
CPTs are one of the most effective geotechnical displacement rate is 0.1 mm/s.
investigation techniques, it is undeniable that
correlation of certain properties based on 3 GRAIN SIZE EFFECTS
measurements obtained through CPT testing, relies Another issue is the grain size of the testing
on a series of empirical factors that are derived material. Results from various experimental studies
through calibration and experimentation (Gui & suggest requirements for the CPT diameter (B) to
Bolton, 1998). The use of miniature CPTs in small median grain size (d50) ratio. Research has shown
scale physical models can be helpful in a better that for CPTs measuring only the cone resistance, a
interpretation of experimental results and relating minimum B/d50 ratio of 20 is required (Gui &
measurements to full scale prototypes. Bolton, 1998). If the device is to measure sleeve
The objective of this paper is to illustrate the main friction as well as cone resistance, a minimum ratio
steps taken in designing a miniature penetrometer of 28 is required (Fioravante, 2002).
for a geotechnical centrifuge. The geotechnical
centrifuge of Delft University of Technology (DUT) 4 CONSIDERATIONS AND DESIGNS
and the available actuator define physical boundary Designs of various standard and miniature cone
conditions which should be respected in the design. penetrometers were reviewed in order to derive a
Boundary conditions and boundary effects act as basis for the proposal of a new design. The obtained
limiting features and create constraints to the insight was then considered with regard to the
application of a standard cone penetrometer. aforementioned boundary conditions. The ratio of
container width to CPT diameter was a defining
2 BOUNDARY CONDITIONS factor resulting in a maximum probe diameter of 4
Consideration of boundaries and physical mm for the container of 150 mm width and a
specifications of the test apparatus is crucial. Firstly, maximum CPT diameter of 7.5 for the container
the inner width or diameter of the sample container with 300 mm diameter. For each of the two
is a limiting factor as the rigid boundary walls may proposed dimensions, different load cells needed to
affect the readings from a CPT. Research has shown be designed.
that the ratio of container width (B) to CPT diameter The design of the loadcells revolved around the
(D) should exceed a minimum measure of 40 in measurement of a minimum threshold of 500 micro-
order to negate all boundary effects. (D / B ≥ 40) strains by the strain gauges of each loadcell. Three
(Gui & Bolton, 1998). This ratio was derived categories of cone resistances were defined: (I) 0.5
through tests conducted on sand samples with a – 2 MPa, (II) 5 – 25 MPa and (III) 30 – 70 MPa.
density index varying between 74% and 81%. It has Based on the ranges defined in these categories, the
been recognized by the authors that such a ratio will geometry of the load cell unit and the material type
be clearly dependent on the relative density of the
285
was chosen so that it would register the minimum Another option that is considered is the usage of a
threshold strain. bending load cell for the second design and the
The first penetrometer design is illustrated in replacement of the load cell with a hollow tube. This
Figure 1 (left). This design is 7.5 mm in diameter, tube extends and is attached to the load cell mounted
and consists of a hollow tube load cell with a on the loading frame. This allows for more accurate
diameter of 4 mm and a wall thickness of 0.125 mm. measurements and simplicity during the design and
This penetrometer is capable of measuring both the manufacturing processes. This design also enables
cone resistance and the sleeve friction. The hollow the users to conduct measurements in soft soils with
chamber outlined at the cone creates opportunity for lower cone resistance values. An illustration of this
the usage of pore-pressure transducers if need be. design is illustrated in Figure 2.
This device can be used for soil samples of a
maximum average grain size of 270 µm.
Figure 1 (right) illustrates the second design with
a diameter of 4 mm. This device consists of a
modular load cell with a tube diameter of 2.5 mm
and a wall thickness of 0.1 mm. This design utilizes
bolt strain gauges which are inserted and glued
inside the load cell chamber. This device is only
capable of measuring the cone resistance (qc) of a
sample. This penetrometer is to be used in any soil
sample with a maximum average grain size of 200
µm.
Load Cell
Load Cell
Figure 2. Bending load cell option
5 CONCLUSIONS
It has been concluded that an extended inner tube,
Strain connected to the cone, attached to a load cell, which
Gauge mounted on the loading frame is the most feasible
option, since the boundary conditions only allow
cone diameters up to 7.5 mm.
Sleeve 6 ACKNOWLEDGEMENTS
The authors would like to acknowledge Karel Heller
and Kees van Beek for their practical input.
7 REFERENCES
Fioravante, V. 2002. On the shaft friction modelling of
non-displacement piles in sand. Soils and Foundations,
42, 23–33. doi:10.3208/sandf.42.2 23
Gui, M., and Bolton, M.D. 1998. Geometry and scale

effects in CPT and pile design. Proc. 1st International
Conference of Site Characterisation 1063-1068.
Lunne, T., Robertson, P., and Powell, J. 1997. Cone

penetration testing in geotechnical practice (1st ed.).
Figure 1. 7.5 mm Ø design (left) and 4.0 mm Ø design (right)
Blackie Academic and Professional, London, England.
286
Behaviour of prefabricated concrete screw piles under axial loading

André Arnold1, Amin Askarinejad2
1
Lucerne University of Applied Sciences and Arts, Engineering and Architecture, Institute of Civil Engineering,
Switzerland
2
TU Delft, Geo-Engineering Section, The Netherlands
Corresponding author: André Arnold (andre.arnold@hslu.ch)
Keywords: Concrete Screw Piles, Loading Behaviour, Centrifuge Modelling.
1 INTRODUCTION 2 CENTRIFUGE MODELLING

A newly developed prefabricated concrete screw The model piles were made using advanced 3D
pile system (Figure 1) is being used in Switzerland printing techniques. Vingerling Clay K122,
for a wide range of applications such as pile described in Hiemstra and Rijsdijk (2003), available
foundations for masonry buildings and offshore in blocks has been used for the soil model (Fig. 2).
test-foundations for the cable car project across Its strength and stiffness have been investigated
Lake Zurich (Arnold et al. to be published 2020). using geotechnical standard tests such as
Static and dynamic load tests on the piles have Oedometer-, Direct Shear-, Vane Shear- and
shown that the shaft resistance of the piles is higher Triaxial tests. A list of approximate soil parameters
compared to conventional displacement piles, is given in Table 1. The outer diameter of the 3D
whereas the exact loading behaviour of this screw printed model screw piles is 5.5 mm (equivalent to
pile system is mostly unknown due to the new screw 550 mm at prototype scale). The centrifuge tests
geometry. Therefore, the screw piles have been were conducted at 100g.
investigated preliminary with the use of centrifuge
model tests in the beam centrifuge at TU Delft Table 1. Soil parameters of Vingerling Clay
(Allersma 1994; Li et al. 2020). The testing Parameter Unit
procedure in the centrifuge and first results on the Unit weight, γ kN/m3 19.5
loading behaviour are presented in this contribution. Water content, w % 29
Pre-consolidation stress, σ’0 kN/m2 100
Friction angle, ϕ’crit º 24-26
Undrained shear strength, su kN/m2 35-37
Water permeability, k m/s 7∙10-11
Degree of Saturation, Sr % 96
2.1. Testing procedure

A screw model pile and a conventional smooth
model pile both with an outer diameter of 5.5 mm
have been placed in the same clay block to directly
compare the loading behaviour of the two pile
systems in the same soil model. The model piles
have been installed at 1g conditions.
The soil model including the two model piles has
been consolidated at 100g for approximately 40 min
to ensure dissipation of excess pore water pressures.
Both piles have been tested axially at a constant
axial settlement-rate of 0.005 mm/s. This
settlement-rate is lower than recommended by
Knappett and Craig (2012) who recommended a
rate of 0.0083 – 0.03 mm/s. The centrifuge needed
to be stopped in between the two pile tests to
Figure 1. Prefabricated concrete screw piles “KIDRILL” for
the test foundation of the cable car project across Lake Zurich. reposition the actuator. The soil model was again
Pile-sections length 8.0 m. consolidated at 100g for at least 5 minutes before
the second pile test has been conducted.
287
in terms of some hardening effects with a more
gradual decrease of stiffness compared to the
smooth pile, which tend to suddenly reach the
ultimate limit state showing an elastic – perfectly
plastic behaviour. The tests given here allow for a
comparison of two different pile systems in the
same soil model. More centrifuge tests would be
needed to investigate the loading behaviour of the
screw piles in more detail.
Figure 2. Vingerling Clay-block after centrifuge test:

Backwards: smooth pile; in front: screw pile with screw profile
visible in the pile-soil interface. Model width: 133 mm; model Figure 3. Load-settlement curves of the screw- and smooth
length: 200mm; model depth: 120 mm; pile-embedment depth: piles tested in the same Vingerling-Clay soil model.
80 mm.
5 ACKNOWLEDGEMENTS
3 RESULTS AND DISCUSSION We are most grateful to Alphabeton AG,
The load-settlement-curves of both model piles are Switzerland, KIBAG AG, Switzerland and
given in Fig. 3. The screw pile can sustain a Innosuisse, which funded the research project and
prototype axial load of approx. 390 kN at an axial section of Geo-Engineering at TU Delft, which
settlement of approx. 60 mm. Whereas, the smooth partly funded the centrifuge model tests.
pile can sustain a prototype axial load of approx.
270 kN at an axial settlement of approx. 26 mm. The 6 REFERENCES
ultimate limit state of the screw pile is approx. 40 % Allersma, H.G.B. 1994. The University of Delft
higher than the ultimate limit state of the smooth geotechnical centrifuge. In Centrifuge ’94:
pile in the two given tests. The settlement at ultimate Proceedings of the International Conference (Leung,
limit state of the screw pile fits well to 0.1∙D, which C.F. and Lee, F.H. (eds)), Balkema, Rottderdam, the
would be 55 mm. The smooth pile reaches its Netherlands, pp. 47-52.
ultimate limit state earlier at approx. 0.05∙D.
The curves do also show, that for small Arnold, A., Espinosa, T., Felder, H.P., Zürcher, E. and
settlements both pile systems show the same Askarinejad, A. to be published 2020. Seilbahnstützen-
stiffness as a load-transferring system, whereas the Fundamente für die ZüriBahn. In 35. Christian Veder
Kolloquium Graz, Austria, 16–17 April 2020.
smooth pile seems to reach its ultimate limit state
more or less suddenly by reaching the maximum Hiemstra, J.F. and Rijsdijk, K.F. 2003. Observing
axial load without any hardening effects. One could artificially induced strain: implications for subglacial
say that the pile system seems to have an elastic, deformation. Journal of Quaternary Science, 18(5):
perfectly plastic behaviour. The screw pile shows 373-383. doi: 10.1002/jqs.769.
not only a greater ultimate limit state but also some
hardening with a distinct decrease of stiffness Knappett, J.A. and Craig, R.F. 2012. Craig’s Soil
before reaching the ultimate limit state. Mechanics. Spon Press, an imprint of Taylor &
Francis, London and New York.
4 CONCLUSION Li, Q., Prendergast, L., Askarinejad, A., Chortis, G. and
The two tests, among more tests in this series, show Gavin, K. 2020. Centrifuge Modeling of the Impact of
that the screw pile can sustain larger axial load than Local and Global Scour Erosion on the Monotonic
a smooth pile with the same outer diameter. The Lateral Response of a Monopile in Sand. Geotechnical
loading behaviour of the screw pile is more robust Testing Journal (in-press). doi:10.1520/GTJ20180322.
288
Validation of ESB Box and Estimation of Shear Wave Velocity in Partially

Saturated Soil
Yoon-Ah Kim1, Hae-In Lee2, Satish Manandhar1, Dong-Soo Kim1
1
Department of Civil and Environmental Engineering, Korea Advanced Institute of Science and Technology
(KAIST), Korea
2
The 4th R&D Institute, Agency for Defense Developement, Korea
Corresponding author: Dong-Soo Kim (dongsookim@kaist.ac.kr)
Keywords: Dynamic Centrifuge Test; ESB Box; Shear Wave Velocity; Partially Saturated Soil.
1 INTRODUCTION Following the procedure from Lee at al. (2012), the

In this study, dynamic centrifuge tests were empty ESB box was shaken again to see if the
conducted to simulate and verify the 1D site natural frequency of the box has changed over the
response analysis of level-ground during years. A summary of centrifuge tests cases are
earthquakes. The dynamic centrifuge test was shown in Table 1.
performed using a uniform layer of partially
Table 1. Summary of centrifuge test cases (in prototype scale)
saturated sand model inside an equivalent shear
beam (ESB) box, which is known to minimize the Test g-level Max. input
boundary effect during earthquake shaking (Lee et Input Signals
Cases (g) acceleration (g)
al., 2012). Strata was used as the program for sweep, resweep,
Filled
performing the 1D site response analysis, and the VS 50 artificial, sine 0.022-0.076
ESB
profile obtained from the bender elements (BE) 3Hz,
placed at different depths during the centrifuge test Empty sweep, resweep,
20, 40 0.041-0.241
was set as an input parameter to Strata. Then ratio ESB artificial
of response spectra (RRS), ratio of response spectra
at the sand surface and of the input motion, obtained 3 METHODS AND RESULTS
from the centrifuge test and strata were compared. 3.1. Validation of ESB Box
In RRS from the centrifuge test, greatest According to the RRS of the empty ESB box (ratio
amplification occurred at 0.5 second, while the RRS of response spectra at the top and the bottom of the
from Strata showed largest amplification at around box), the natural period of the empty ESB box was
0.75 second. This discrepancy could be due to two estimated to be 0.17 ~ 0.18 s at 20g and 0.3 ~ 0.4 s
main reasons; boundary effect from the container at 40g. According to Lee et al. (2012), the natural
box and incorrect estimation of VS from BE test in period of the ESB box was 0.2 second at 20 g and
the centrifuge test. Therefore, validation of the ESB 0.4 second at 40 g. Hence, it can be inferred that the
box used in the centrifuge test was done, and the VS ESB box has become stiffer. However, these
was obtained by other various methods. changes in the natural period of the empty ESB box
did not appear to have a significant impact on the
2 EXPERIMENTAL CONFIGURATIONS performance of the filled ESB box during shakings,
The geotechnical centrifuge facility and the ESB because the accelerometers placed at the same
box at KAIST were used to perform the tests. The height (centre of the box and near the box wall) in
ESB Box was initially filled with landfill soil from the soil model have the same response. This means
Ulsan, and it is classified as silty-sand (SM). The there were almost no boundary effect.
water content of the soil was 11%, and it was
compacted with to a relative density of 40%. As the 3.2. Estimation of Shear Wave Velocity
results from the centrifuge test and Strata results In Section 3.1, it was shown that the performance of
differed significantly, performance of ESB was the ESB box was intact, so it is necessary to check
revalidated. the reliability of VS obtained by the BE test.
Lee et al. (2012) validated the performance of Therefore, the VS was calculated by using other
KAIST ESB box by shaking an empty ESB box, a methods.
fully-filled ESB box, and a partially-filled ESB box
in 20g and 40g centrifugal acceleration levels. Since
3.2.1 Method 1: Resonant Column Test
the ESB box have aged for over 8 years, the initial
performance of the ESB box may have deteriorated. The soil samples used in the resonant column test
(RC test) were the same as the soil used in the
289
centrifuge test, and were performed under three amplification function, there is a disadvantage that
initial confining pressures; 60 kPa, 120 kPa, and 240 only the average VS of the entire soil model is
kPa. The VS can be obtained by using Equation 1, obtained, not the VS of each layer of the model.
and VS profile is shown in Figure 1.
𝑉𝑉𝑆𝑆 = �𝐺𝐺𝑚𝑚𝑚𝑚𝑚𝑚 /𝜌𝜌 (1)
3.2.2 Method 2: Stress-Strain Loop

The shear modulus (𝐺𝐺) can be determined using the
density of the soil and the VS. The VS of the ground
can thus be calculated from the shear modulus
obtained from the acceleration responses at different
depths. The shear modulus is equal to the slope of
the stress-strain loop, and a method for calculating
the shear stress and the shear strain using the
acceleration time history is described in Zeghal et al Figure 2. Line: fitted amplification function, dots:
(2018). VS from stress-strain loop is shown in amplification ratio from the centrifuge test
Figure 1. The reason for the Vs obtained by the
stress-strain loop being smaller than the Vs from the 4 CONCLUSIONS
RC test is because of higher shear strain in In this study, the performance of the ESB box after
centrifuge test than the RC test. a period of 8 years was revalidated and the shear
wave velocity of unsaturated soil was estimated by
various methods. The ESB box became a little
stiffer, but the performance during the centrifuge
test was intact. By comparing the VS obtained from
the BE test, the RC test, stress-strain loops, and
amplification functions, it was found that the BE
test underestimates VS in partially saturated soil.
Therefore, it is suggested to estimate the shear wave
velocity of the model using stress-strain loops or
amplification functions or the RC test rather than
using the BE test if unsaturated soil is used in
centrifuge model tests.
5 ACKNOWLEDGEMENTS
This research was supported by a grant (18CTAP-
B139287-02) from Technology Advancement
Figure 1. Vs profiles from BE test, RC test, stress-strain loops. Research Program (TARP) funded by Land,
The dashed line is a fitted line of data points at three confining Infrastructure and Transport of Korean government.
pressures.
6 REFERENCES
3.2.3 Method 3: Amplification Function
Lee, S.H., Choo, Y.W., Kim, D.S. 2013. Performance of
The shape of the amplification function is equivalent shear beam (ESB) model container for
determined by the shear wave velocity and the dynamic geotechnical centrifuge tests. Journal of Soil
damping ratio of the ground. In centrifuge tests, the Dynamics and Earthquake Engineering, 44(2013):
amplification ratio can be obtained by Fourier 102–114.
transform of the acceleration time history. As shown
in Figure 2, the best-fit function was obtained by Conti, R., Viggiani, M.B.G. 2012. Evaluation of Soil
fitting the amplification ratio with an amplification Dynamic Properties in Centrifuge Tests. Jr. Geo. and
function. The Vs and damping ratio obtained from Geoenv. Eng., 138(7): 850-859.
the amplification function is the average VS and
Zeghal M, Goswami N, Kutter B, Manzari MT, Abdoun
damping ratio of the ground (Conti et al. 2012). The T, Arduino P, et al. 2018. Stress-strain response of the
estimated average VS is 249.7 𝑚𝑚/𝑠𝑠 , which is very LEAP-2015 centrifuge tests and numerical predictions.
close to the average value of 248.7 𝑚𝑚/𝑠𝑠 obtained J Journal of Soil Dynamics and Earthquake
from the RC test. However, when using the Engineering, 113(2018): 804-818.
290
Retrofit, redesign, and restructuring of a large geotechnical centrifuge research

center
April J. Bowman1, Anthony F. Tessari1, Christopher Price1
Centrifuge Research Center, Engineer, Research & Development Center, United States Corps of Engineers, USA
1
Corresponding author: April J. Bowman (april.j.bowman@erdc.dren.mil)
Keywords: Centrifuge; Business practices; US Army Corps of Engineers.
1 INTRODUCTION complex problem areas, and validating numerical

The US Army Corps of Engineers (USACE) methods with instrumented physical models. The
Engineer Research and Development Center Centrifuge Research Center (CRC) was originally
(ERDC) solves the United States’ most challenging commissioned for operation in 1995 (Schofield and
problems in civil and military engineering, Steedman 1992) and was designed and built by
geospatial sciences, water resources, and Actidyn Systems of Elancourt, France. The original
environmental sciences for the US Army, price of construction for the facility was
Department of Defense, government agencies, and approximately US$15M. Estimated cost of
other research institutions. With seven laboratories construction in 2020 dollars is about US$26M. The
and multiple field sites, over 2,100 employees with CRC represents a significant investment by the
incredible breadth and depth in science and ERDC into a highly specialized and unique
engineering knowledge, and a US$2 billion research scientific capability. Efforts to modernize the
program means the ERDC's capabilities are far- centrifuge and support facilities, and updates to the
reaching. Research and development is conducted business model are ongoing.
in five primary business areas.
• Engineered Resilient Systems 2 RETROFIT AND REDESIGN
• Geospatial Research and Engineering The implementation of major procedural and
• Military Engineering technological advances in a centrifuge research
• Environmental Quality and Installations center presents a significant challenge, especially if
• Water Resources enacted all at once, and raises an interesting
The US Army Centrifuge Research Center (CRC), question, “how does one revitalize the infrastructure
located at the ERDC's Geotechnical and Structures and research objectives of an entire centrifuge
Laboratory (GSL) in Vicksburg, Mississippi, operation?” This is the challenge currently facing
provides researchers an economical approach for the Centrifuge Research Center.
evaluating alternative designs, investigating
Figure 1. Facilities associated with the ERDC's Centrifuge Research Center
291
The first step has been to upgrade the existing • Replacement of variable frequency drive
infrastructure and testing capabilities followed by a cabinets, fluid rotary joint, optical rotary
program of new construction. Approximately joint, and aerodynamic cowlings,
US$5.3M has been invested into this modernization • Upgrading the self-balancing system and
effort with an additional US$5.2M projected to data acquisition system, and the 1-D shaker
complete all phases. Figure 1 shows the new CRC control systems to accommodate waveform
complex with indicators of how this investment will generation, and
be phased. • Development and implementation of novel
instrumentation including fibre-optics.
2.1. Phase 1: Upgrade existing Centrifuge
Control Facility 2.3. Phase 3: Centrifuge Storage Facility
The original control facility was built in 1989 and In 2022, a new 298 sq. m. storage facility is planned
modified in 1993 to support centrifuge operations. to house large items supporting centrifuge
Demolition and renovation of this facility started in operations. In addition to warehouse-type shelving,
November of 2019 with scheduled completion in this facility will house the ERDC’s large scale
October 2020. The new building will have 10 consolidometer for clay model preparation.
offices for centrifuge personnel and visiting
scholars, a state-of-the-art control room, and a 3 RESTRUCTURING
secure conference room capable of supporting The facility has a long tradition of working with
classified briefings, training, and test protocols. other Army Corps of Engineers Districts in addition
Data transfers between office work stations, the to other government and civilian institutions. The
control room, and the centrifuge will be conducted business plan is being restructured to allow
over a stand-alone fiber-optic network system. flexibility to venture outside this sphere and engage
Additionally, the existing walkway connecting this with academic and civilian institutions, both US-
facility and the centrifuge will be enclosed and based and international. Further, the business model
house historical exhibits chronicling the ERDC’s 25 is being modified to reflect recent upgrades in
year history with the centrifuge including our work capabilities and captive markets in the centrifuge
with blast effects (Charlie et al. 2005) and on the modelling arena. Many high-performing facilities
New Orleans Levees following Hurricane Katrina around the world have carved out productive niche
(Ubilla et al. 2008). markets based on their production, capabilities,
reputation, and geographic location. The facility is
2.2. Phase 2: Centrifuge Upgrades similarly tailoring their approach to generate new
At the time of its commissioning the CRC was the strategic collaborations in what it has identified as
only centrifuge facility in the world that could underserved markets that align with key capabilities
accommodate large model sizes and high g-levels. of the centrifuge and the ERDC. The CRC is
Presently, only the Deltares GeoCentrifuge in Delft, actively seeking to increase engagement with the
Netherlands matches the U.S. Army centrifuge in international physical modelling community and to
size and power. Table 1 outlines some of the key forge mutually-beneficial partnerships that increase
specifications of the ERDC centrifuge. research output in vital areas, support graduate
students, and increase visibility of said research
Table 1. The ERDC’s Centrifuge Specifications efforts through widespread dissemination of results.
Parameter Unit Specifications
Payload tons 8.8 4 REFERENCES
Platform Dimensions Charlie, W.A., Dowden, N.A., Villano, E.J., Veyera,
m 1.4x1.4x1.8
(l x w x h) G.E. and Doehring, D.O., 2005. Blast-induced stress
Arm Radius m 6.5 wave propagation and attenuation: centrifuge model
Max g-force g 350 versus prototype tests. Geo. Test. Jr., 28(2): 207-216.
To continue to carry out fundamental research in Schofield, A., and Steedman, R. 1992. Development of a
geotechnical and structural engineering several WES Centrifuge. Report No. 26-02-R-004.
upgrades to the centrifuge itself and its equipment
are ongoing. In addition to infrastructure Ubilla, J., Abdoun, T., Sasanakul, I., Sharp, M.,
Steedman, S., Vanadit-Ellis, W. and Zimmie, T., 2008.
improvements to the facility, centrifuge upgrades New Orleans levee system performance during
include hurricane Katrina: London Avenue and Orleans canal
south. Jr. Geo. Geoenv. Eng., 134(5): 668-680.
292
Using Fibre Bragg Grating sensors to estimate the horizontal monopile behaviour
in centrifuge
Matthieu Blanc1, Zhong-Sen Li2, Luc Thorel1
Geomaterial and Geotechnical modelling laboratory, University Gustave Eiffel, France
1
Department of Civil Engineering, Aalto University, Finland

2
Corresponding author: Matthieu Blanc (matthieu.blanc@univ-eiffel.fr)
Keywords: Monopile; Horizontal Loading; Fibre Bragg Grating Sensors; Sand.
1 INTRODUCTION Two identical optical fibres with a diameter of 180

In the literature, reduced-scale centrifuge tests have µm are glued in the two grooves. On each fibre, 10
been used to characterise the response of pile or Fibre Bragg Grating sensors (FBGs) have been
monopile under horizontal loads (e.g. Verdure et al. manufactured every D/2 (i.e. 25 mm) from the
2003; Rosquoet et al. 2007; Li et al. 2010; Klinkvort ground level to the pile base. These FBGs are
and Hededal 2014; Klinkvort et al. 2019; Choo and protected by a Teflon tube of 200 µm in diameter.
Kim 2015; Truong et al. 2018). Through carrying These FBGs measure the external deformations at
out centrifuge tests, researchers were able to test 10 different cross sections of the pile and then, after
monopiles as large as 50 mm in diameter (e.g. calibration, give the bending moments of the pile at
Klinkvort et al. 2019). these levels.
The present study is part of the SOLCYP+ project,
which is an extension of the SOLCYP project 3 EXPERIMENTAL SET-UP
(Puech and Garnier 2017). The main objective was Concerning the pile installation, the instrumented
to experimentally investigate the response of the model pile was pushed into the sand at 1g level by a
model piles under horizontal load in geotechnical hydraulic jack. The jacking speed was 1 mm/s and
centrifuge and then to extrapolate the prototype pile the attained depth is 450 mm, i.e., 9D. The sand
behaviour. Fibre Bragg Grating sensors (FBGS) level inside the pile was checked and there was no
were used to measure the local strains and to plug generated.
determine the moment profile of the pile (Li et al., After pile installation, an electric actuator was
2020). mounted on two supporting beams that
perpendicularly placed on the longitudinal edges of
2 MODEL PILE the strongbox (Figure 1).
In previous centrifuge tests, piles were in- Centrifuge test was performed at 100g level.
strumented with strain gages glued on the external Horizontal load was applied through the centre of
pile shaft (Roesquoët et al. 2007) or even inside the the cross section by pushing the steel rod that
tube (El Haffar 2018), with the connection wires crosses the monopile perpendicularly to the
embedded inside the tube. These methods, however, instrumentation plan. The loading point situates at
alter the pile dimension and surface roughness and 500 mm (i.e. 10D) above the ground level. The
as a result may lead to erroneous measurements. To loading was displacement-controlled with a rate of
accurately model the monopile behaviour, the pile Piston
Load sensor
must be open-ended with sand inside the pile (i.e. Fork Steel rod
Laser
in-situ situation). For this purpose, the model Electric actuator -500
monopile in this study has been instrumented with -400
Fibre Bragg Grating sensors (FBGs). -300
The model monopile is an open-ended aluminium

-200
160 mm
2017A tube with the external diameter D = 50 mm,

-100
Water level
the embedded length L = 450 mm, the load 0 Ground level

100
eccentricity, le = 500 mm and the wall thickness t =
200
2.5 mm.
560 mm
300
Two semi-cylindrical grooves, diametrically 400
opposed and with a radius of 0.5 mm, have been 50
machined along the pile shaft in its embedded part. z (mm) Sand
In this way, the depth of the grooves is 0.5 mm, i.e. 1200 mm
1/5 of the wall thickness. Figure 1. Schematic drawing of the horizontal loading
system in the centrifuge at 100g.
293
Strains (mm/m) Moment (kN.m) Shear force (kN) Soil reaction (kN/m)
-4 0 4 0 0.5 1 -2 0 2 4 -50 0 50
0 0 0 0
50 50 50 50
100 100 100 100
150 150 150 150

Depth (mm)
200 200 200 200
250 250 250 250

Load increases
300 300 300 300
350 350 350 350
400 400 400 400

Compressive Tensile
450 450 450 450
Figure 2. Profile of (a) the measured strains, (b) the bending moment, (c) the shear force and (d) the soil reaction (i.e., the
distributed load) for seven horizontal loads (every 0.25 kN until 1.75 kN).
1 mm/s at the actuator level and terminates at 50 mm Monopiles in Sands: Centrifuge Tests. J. Geotech.
(i.e. 1D). The measurements of the FBGS, the laser Geoenviron. Eng., 142(1), 04015058.
and force transducer were recorded and registered
every 0.02 second (i.e., 50 Hz). El Haffar, I. 2018. Physical modelling and study of the
behaviour of deep foundations of offshore wind
turbines in sand (Doctoral dissertation, IFSTTAR).
4 CONCLUSION
The experimental results (Figure 2) show that: Klinkvort, R.T., Hededal, O. 2014. Effect of load
• All the 26 FBGS survived in the 100g centrifuge eccentricity and stress level on monopile support for
test. offshore wind turbines. Can. Geotech. J., 51(9), 966-
• At the ground level, the moments determined by 974.
FBGS are in accordance with the moments
Klinkvort, R.T., Britta Bienen B., Fan, S., Black, J.,
calculated from the transversal force; Bayton, S., Thorel, L., Blanc, M., Madabhushi, G.,
• Shear forces at the ground level are slightly larger Haigh, S., Broad, T., Zania, V., Askarinejad, A., Li, Q.,
than the applied transversal forces. Such Kim, D. S., Park, S. 2019. Centrifuge modelling
difference mainly results from the difficulty of the considerations of laterally loaded monopiles in sand.
accurate derivation operation at the ground level. Géotechnique (submitted).
• The pile local behaviour is characterized: under
the effect of the increasing horizontal load, pile Li, Z.S., Blanc, M., Thorel, L. 2020. Using FBGS to
rotates at the depth 210 - 275 mm (i.e., 4.2 - 5.5D) estimate the horizontal response of a monopile in
below the ground level. Due to the large geotechnical centrifuge. Int. J. Phys. Model Geo.
embedment depth, the pile deflection below the https://doi.org/10.1680/jphmg.19.00022.
rotation centre was close to zero. The P-y curves
Puech, A. and Garnier, J. 2017. Design of piles under
at different depths were also determined. cyclic loading: SOLCYP recommendations. Wiley-
ISTE, 454 pages.
5 ACKNOWLEDGEMENTS
This work benefited from France Energies Marines Rosquoet, F., Thorel, L., Garnier, J., and Canepa, Y.
and State finance managed by the National 2007. Lateral cyclic loading of sand-installed piles.
Soils and Foundations, 47(5), 821-832.
Research Agency under the Investments for the
Future program bearing the reference SOLCYP+ Truong, P., Lehane, B.M., Zania, V., and Klinkvort, R.
ANR-10-IEED-0006-18. Assistances from the T. 2018. Empirical approach based on centrifuge
technical staffs of the centrifuge group are testing for cyclic deformations of laterally loaded piles
acknowledged. in sand. Géotechnique, 69(2), 133-145.
Verdure, L., Garnier, J., and Levacher, D. 2003. Lateral

6 REFERENCES
cyclic loading of single piles in sand. Int. J. Phys.
Choo, Y.W., Kim, D. 2015. Experimental Development Model Geo., 3(3), 17-28.
of the p-y Relationship for Large-Diameter Offshore
294
High density soil instrumentation for a thermal pile test

Anders Bergström1, Saqib Javed2,3, Jelke Dijkstra2
1
Infrastructure, NCC, Sweden
2
Architecture and Civil Engineering, Chalmers University of Technology, Sweden
3
Building and Environmental Technology, Lund University, Sweden
Corresponding author: Jelke Dijkstra (jelke.dijkstra@chalmers.se)
Keywords: Field Test; Thermal Piles; Sensitive Clay; Instrumentation.
1 INTRODUCTION 2.2. Soil instrumentation

Whilst physical modelling at 1-g and N-g in the The number of instruments used in a full-scale field
laboratory has proven to be a very effective tool for test is often rather limited. This not only is a cost
systematic quantification of the response of deep consideration, as instruments are often lost during
foundations in various soils (e.g. Leung et al. 2004, the test, it is also related to the discipline specific
Lehane and White 2005) including thermal piles background. Most field instrumentation is designed
(Stewart and McCartney 2013, Ng et al. 2014, by geotechnical engineers with a focus on assessing
Yavari et al. 2016), there are scaling mechanisms the occurrence of well-established mechanisms
remaining. The scaling issues are most pronounced (Dunnicliff 2012). Hence, only few instruments at
in case creep is considered, i.e., one of the carefully chosen locations are sufficient. In contrast
governing mechanisms that does not scale at this particular project aimed to provide a detailed
elevated acceleration levels (Taylor 1995). In the monitoring of the soil surrounding the pile during
case considered here, the high sensitivity of the clay thermal loading.
cannot be easily replicated in the laboratory. A combination of manual measurements with
The project sets out to investigate the impact of classic extensometers using a bellow-hose
thermal loading on the long-term soil response of a (Bozozuk and Fellenius 1978) and 17 retrievable
floating pile in sensitive clay by performing a scaled piezometers (Torstensson 1984) have been used.
in-situ test at a well characterised and densely The standpipes and the filter tips of the BAT®
instrumented test site with sensitive clay. For a full system were made out of plastic, to minimise
overview of the test setup and results, see Bergström thermal conduction paths from the instrumentation.
et al. (2020) A Sensornet fibre based Distributed Temperature
Sensing (DTS) system was used to measure the
2 DESIGN CONSIDERATIONS temperatures in the heat exchanger system and on
2.1. Thermo-mechanical design the outer pile wall.
The installation of the plastic standpipes proved to
The design approach of the instrumented field test be challenging, even in the slightly over-
on a 28 m long floating pile in sensitive clay fitted consolidated deposit of soft sensitive clay. The soil
with a heat exchanger followed the following at the site has a linear increasing undrained shear
guiding design principles: strength profile: 15 kPa below the dry crust near the
1) The mechanical and thermal boundary surface and 50 kPa at the depth of the pile base.
conditions should reflect those of a loaded
floating thermal pile in soft soil. 3 RESULTS
2) The soil response during heating and After pile installation, load application of the
cooling needs to be captured with high Thermal Response Test (TRT) is performed, during
spatial and temporal resolution. which the distribution of excess pore water
These requirements led to a long small diameter pressures and temperature in the soil is assessed.
slender steel tubular pile. The small diameter and The results of 17 pore water pressure and
steel section reduced the total test duration for temperature sensors are spatially integrated by
thermal loading, whilst the long pile length still interpolating the point measurement on an
accessed the deep soil layers between 7 and 28 m. unstructured mesh for several instances in time.
Furthermore, the long-term mechanical load This in turn allows to process full-field information
remained manageable at 80-90 kN and the thermal of pore water pressures and temperature fields in a
capacity of the mobile heat exchanger system. ‘virtual’ cross-section of ca. 7 m width and 21 m
height. Figure. 1 & 2 show an example result for
295
Figure 1. Temperature field in the soil after 504 hrs of heating Figure 2. Excess pore pressure field in the soil after 504 hrs of
from the thermal pile. The black dots denote the location of the heating from the thermal pile. The black dots denote the
sensors. Lpile = 28 m; Rpile = 0.0575 m. location of the sensors. Lpile = 28 m; Rpile = 0.0575 m.
respectively the temperature and excess pore water 6 REFERENCES

pressures at the end of the TRT test. Despite the Bergström, A., Javed, S. & Dijkstra, J. 2020. Field test of
reasonably small number of measurement points the a floating thermal pile in sensitive clay. Géotechnique,
uniform heating of the thermal pile and the excess ahead of print: doi.org/10.1680/jgeot.19.P.094
pore water pressures it triggers can be captured with
high fidelity. One of the main findings is the Bozozuk, M., & Fellenius, B. H. 1979. The bellow-hose
relatively benign magnitude of excess pore water settlement gauge. Canadian Geotechnical Journal,
pressures (∆u) during heating. ∆u = +/-3 kPa after 16(1), 233-235.
20 days of heating and cooling respectively. The
Dunnicliff, J. 2012. Chapter 95 Types of geotechnical
temperature for the 50W/m heat exchanger pile instrumentation and their usage. In ICE manual of
reached in excess of 25 oC at the pile wall, quickly geotechnical engineering (pp. 1379-1403). Thomas
reducing to 13 at 8.5Rpile distance. Telford Ltd.
4 CONCLUSION Lehane, B. M., & White, D. J. 2005. Lateral stress

changes and shaft friction for model displacement piles
The densely instrumented field test enabled new
in sand. Can. Geot. Journal, 42(4), 1039-1052.
insights in the spatial and temporal evolution of the
pore water pressure and temperature fields. The Leung, C. F., Liao, B. K., Chow, Y. K., Shen, R. F., &
thermo-mechanical design and instrumentation plan Kog, Y. C. (2004). Behavior of pile subject to negative
were developed with a physical modelling approach skin friction and axial load. Soils and Foundations,
in mind. The first-time combination of plastic filter 44(6), 17-26.
tips and standpipes and retrievable BAT probes
proved to be attainable in soft clays. Ng, C. W. W., Shi, C., Gunawan, A., & Laloui, L. 2014.
Centrifuge modelling of energy piles subjected to
5 ACKNOWLEDGEMENTS heating and cooling cycles in clay. Géotechnique
letters, 4(4), 310-316.
The Authors are grateful for the excellent support
from the research engineers Håkan Larsson and Stewart, M. A., & McCartney, J. S. 2013. Centrifuge
Aaro Pirhonen at Chalmers. The financial support modeling of soil-structure interaction in energy
from SBUF (grant no. 12904), FORMAS (grant no. foundations. Journal of Geot. and Geoenvironmental
243-2013-771), Swedish Energy Agency (grant no. Engineering, 140(4), 04013044.
P37307-1), NCC, Hercules and Chalmers
University of Technology is gratefully Torstensson, B.A. (1984). A new system for ground
water monitoring. Groundwater Monitoring &
acknowledged. SAB/Ruukki kindly donated the
Remediation, 4(4), 131-138.
steel piles used in this research project.
Yavari, N., Tang, A. M., Pereira, J. M., & Hassen, G.
(2014). Experimental study on the mechanical
behaviour of a heat exchanger pile using physical
modelling. Acta Geotechnica, 9(3), 385-398.
296
Physical Modelling of Geotechnical structure subjected to Hydraulic Scour

Liam Jones1, Ioannis Anastasopoulos1
1
Institute for Geotechnics, ETH Zürich, Switzerland
Corresponding author: Liam Jones (jonesli@ethz.ch)
Keywords: Scour; PIV; Tsunami; Flooding.
1 INTRODUCTION
Increases in the frequency of freak weather events,
rising sea levels and aging infrastructure are
speculated to increase the demands on coastal or
near-water infrastructure (Steffen, Hunter and
Hughes, 2014). In order to make sure that these
structures are adequate, and in order to construct so-
called “resilient” societies, we must first understand
how structures behave during these events.
Scouring is the process of removal of soil by a flow Figure 1. Planar model dimensions, the incoming wave
travels from right to left.
of water, and is often seen at the bases of submerged
bridge piers and other foundation structures placed
within flowing water. Scour is known to 3 RESULTS AND DISCUSSION
significantly alter the geometry of foundation soil Initially, as the breakwater is overtopped, scouring
systems, possibly affecting their performance occurs and rapidly begins to create a hole of
(Yilmaz, Banerjee and Johnson, 2016). There exists increasing depth (Figure 2). During this phase the
much physical experimental data which is used to block exhibits some small (apparently negative,
predict various scour for various configurations (eg. clockwise) rotation, towards the wave. At some
Richardson & Davis, 1995). However, there is time (A), the scour depth appears to reach a steady
considerably less knowledge of how the process of state, and almost immediately we begin to see the
scouring affects the performance of real structures. block rotating anticlockwise.

In order to shed some new light on these problems;
a new physical experimental apparatus is under
development at ETH Zürich. Consisting of a
“Miniaturised Tidal Generator” (MTG), this set-up
aims to combine geotechnical physical modelling
with the ability to model a wide range of hydraulic
events such as Tsunamis and flooding (Jones and
Anastasopoulos, 2018, 2019). A simple experiment,
consisting of a 50mm by 50mm aluminium
“breakwater” placed on a pluviated layer of 50mm
of cohesionless marine sand (Dr = 65%) (“Perth
Sand”), was constructed inside the MTG (Figure 1). Figure 2. Plot showing the evolution of scour depth,
breakwater rotation and breakwater velocity with time.
This structure was then subjected to a “Tsunami-
like” pulse wave, period 10s, and maximum
The third phase (beginning point B) shows how as
amplitude of 20mm, targeting a model scale of N =
the block begins to undergo significant movement
100. The resulting failure is then recorded with high
(rotation), and the scour hole begins to re-fill (the
speed photography, through the Perspex side of the
scour depth decreases). Here, we see large spikes in
model box. The camera used is a Mikrotron Cube
the velocity of the block while the structure is
7, shooting at 300fps. Images are then processed
sliding over the soil, rather than simply moving in
using the particle image velocimetry method (PIV)
response to soil loss. Important observations can be
using the code GeoPIV_RG (Stanier et al., 2016).
made by viewing the strain-field within the soil
297
Figure 3. Computed strain field within the soil from PIV analysis using GeoPIV_RG, volumetric (top) and shear (Bottom)
(Figure 3). Generally, a trend can be seen where the stress similitude on such problems (especially to the
soil in the immediate vicinity of the scour hole observed loosening). These questions are the
develops shear and volumetric strains. subject of ongoing work; adapting the entire scheme
Interestingly, the soil on the breakwater side appears to the new ETH IGT Geotechnical Centrifuge.
to loosen. Shear planes, in the form of a partial slip
surface can also be seen forming under the 5 REFERENCES
foundation before we observe significant rotation. Chock, G. Y. K. 2016. Design for Tsunami Loads and
These shear planes may be the source of the Effects in the ASCE 7-16 Standard, JoSE (ASCE),
apparent loosening observed, caused by soil 142(11), p. 04016093. doi: 10.1061/(asce)st.1943-
dilation. 541x.0001565.
4 CONCLUSION Jones, L. and Anastasopoulos, I. 2018. Development of

an Apparatus for the Simulation of Coastal Structures
Although only one short test has been presented Subjected to Tsunami, 16ECEE.
here, some important observations can already be
made about such soil-structure-hydraulic systems. Jones, L. and Anastasopoulos, I. 2019. Design and
Firstly, after observing the complex, multi-phase Characterisation of a 1-g Testing Apparatus for
deformation/failure behaviour it seems unlikely that Simulating the interaction between Tsunamis and
the response of such systems can be captured by Geotechnical Systems, ICONHIC 2019.
applying an equivalent static force to a simple block
foundation as is often used in practice (Chock, Richardson, E. V and Davis, S. R. 1995. Evaluating scour
at bridges - Third edition.
2016). Secondly, even if the scoured geometry of
the soil could be approximated and simulated, the Stanier, S. A. et al. 2016. Improved image-based
changes in soil state, as well as the interaction deformation measurement for geotechnical
between scour evolution and block rotation, mean applications, Canadian Geotechnical Journal, 53(5), pp.
that using approximate scour geometries for such 727–739. doi: 10.1139/cgj-2015-0253.
systems may not yield satisfactory results. Further
questions include: how can we first explain, and Steffen, W., Hunter, J. and Hughes, L. 2014. Counting
further predict what is happening within the body of the Costs : Climate Change and Coastal Flooding.
the soil in terms of changes in state (loosening), and
what is the effect of this observed re-filling. Lastly, Yilmaz, T., Banerjee, S. and Johnson, P. A. 2016.
Performance of Two Real-Life California Bridges
one critical question that must addressed is: how
under Regional Natural Hazards, JoBE, 21(3), p.
much of this observed behaviour is a function of 04015063. doi: 10.1061/(ASCE)BE.1943-
boundary conditions i.e. what is the influence of 5592.0000827.
298
Distribution of temperature in culverts in cold climate
Karina Tommik1, Jan Laue1, Sven Knutsson1, Nina Lintzén1

1
Departament of Civil, Environmental and Natural resources Engineering, Luleå University of Technology, Sweden
Keywords: Culvert, Frost Heave, Cold Climate, Convective Heat Transfer.
1 INTRODUCTION depth and consequently predicting road damages

Frost heave and thaw settlement in roads and that occur around culverts.
railways is a persistent problem in cold climates
leading to degradation of the road structure over 2 METHOD
time. The presence of a culvert in the road structure As measurements in the field provide insight into
changes the temperature distribution in that the topic, it has to be studied in ideal conditions to
particular area considerably, as cold air is allowed investigate the various contributing parameters
passes through the structure. Therefore the frost separately under defined conditions. Thus,
depth around the culvert is increased compared to laboratory investigations will be used to better
the rest of the road. If frost susceptible soil is present understand the temperature distribution inside
under the road or railway structure, this might lead culverts.
to a larger frost heave in areas of culverts.
The temperature inside the culvert usually is
either considered equal to the ambient outside
temperature or a function of it (Hua et al. 2014).
There are no studies incorporating temperature
change along the axis of the culvert or the
distribution inside the culvert and the factors
affecting it.
Considering the placement of the culvert in the
embankment, it is reasonable to assume that there
will be airflow through the culvert if the ends are
unobstructed. There are two factors contributing to
the air flow. Firstly, stack effect that causes air to
flow from higher pressure area to lower pressure
area. Stack effect will be present if the culvert is
inclined and there is a temperature difference
between the ends of the culvert. Secondly, air flow
inside the culvert will be caused by wind (Zeng et
al. 2017). Increased cooling effect due to air flow
has previously been studied in relation to road
construction on permafrost, where it has been noted
that frost depth under air ducts placed in road
embankments increases during cold periods
(Coulombe et al. 2012).
Presence of air flow inside the culvert indicates
that the heat transfer mechanism is convection.
Therefore, heat transfer between the culvert and the
soil will be governed by temperature and air
velocity inside the culvert as well as the thermal
properties of soil. Additional factors contributing to
the temperature distribution are solar radiation,
snow cover, geometry and material of the culvert.
Developing a model that describes temperature
distribution inside a culvert accurately will result in
a higher degree of precision when estimating frost
Figure 1. Schematic of the experiment setup
299
In order to investigate the effect of air velocity and measurements of actual culverts will be used as
temperature on the temperature distribution inside basis for developing of a numerical model
the culvert, a simplified model is constructed. This describing the temperature distribution in full-scale
first laboratory investigation allows control of culverts and subsequently the distribution in the
environmental factors in order to focus on the surrounding soil. The purpose of the experiments is
convection process. to identify to what extent the heat extraction rate
A schematic setup of the experiment can be seen will increase with increasing air velocity. The more
in Figure 1. The experiment setup is constructed heat extracted per unit of time, the greater the
inside an insulated pallet with dimensions of temperature change in the soil will be. A larger heat
116x76x29 cm (length:width:height). Homogenous extraction rate in situ would lead to increased frost
dry sand with known thermal properties is used as depth and thus possible frost heave.
the soil surrounding a steel corrugated pipe with a It is also of interest how the temperature
diameter of 10cm. Homogeneity of the sand will distribution changes along the longitudinal axis of
allow for a simplified verification of the model the pipe. Temperature differences along the axis of
using analytical methods. The steel pipe has a high a culvert would lead to varying frost depth under the
thermal conductivity which will allow for faster length of the culvert which in turn could lead to
heat transfer between air and soil. The corrugated uneven frost heave in the perpendicular cross-
surface of the pipe will introduce some turbulence section of the road.
to the model in order to compensate for the reduced Additional studies looking into thermal properties
diameter of the pipe compared to an actual outdoor of potential culvert materials as well as insulating
culvert. A fan with adjustable rotation speed is used properties of snow cover and solar radiation will
to push air through a radiator unit and further into have to be performed and taken into account when
the steel pipe. The air velocity can be regulated by addressing temperature distribution in culverts.
altering the rotational speed of the fan. A radiator The initial laboratory experiment described above
connected to a cryobath is used to cool the air. A will serve as initial basis for development of a
large surface area of the radiator allows for rapid numerical study and future field tests. The
cooling of the air while cryobath allows for simple experiment will determine the effect of air velocity
temperature adjustments. Thermocouples are on heat transfer within culverts thus aiding in
embedded into the soil around the pipe as well as estimation of frost depth around culverts.
placed inside the pipe as it is arranged in the field.
A thermal anemometer is placed at the inlet of the 4 REFERENCES
pipe to verify the air velocity at the entrance. The Coulombe, S., Fortier, D., Stephani, E. 2012. Using air
experiment is conducted in a temperature controlled convection ducts to control permafrost degradation
room to maintain a constant temperature of the inlet under road infrastructure: Beaver Creek experiment
air. site, Yukon, Canada. In Proceedings of the Cold
Experiments will be performed with different air Regions Engineering 2012: Sustainable Infrastructure
velocities and temperatures. Between each test soil Development in a Changing Cold Environment,
and the pipe temperature will be allowed to return Quebec City, Canada, 19-22 August 2012. American
to initial conditions at room temperature. Recorded Society of Engineers (ASCE), pp. 21-31
temperatures inside and around the pipe will be used
Hua, L., Fujun, N., Yonghong, N., Xifebg, Y. 2014.
to estimate the effect of air velocity on the Study on thermal regime of roadbed-culvert transition
temperature distribution inside the experiment setup section along a high speed railway in seasonally frozen
in the direction of air flow. ground. Cold Regions Science and Technology, 106-
107: 216-231
3 DISCUSSION
Implementation of the experiment described above Zeng, Y., Liu, K., Zhou, X., Fan, L. 2017. Tunnel
Temperature fields analysis under the couple effect of
will aid in describing convective heat transfer in
convection-conduction in cold regions. Applied
soil. The results of the experiment along with field Thermal Engineering, 120: 378-392
300
Geosynthetic reinforcement in lightly piled embankments: Laboratory model

development
Per Gunnvard1, Jan Laue1, Qi Jia1
Department of Civil, Environmental and Natural Resources Engineering, Luleå University of Technology, Luleå
1
Corresponding author: Per Gunnvard (per.gunnvard@ltu.se)
Keywords: Piled Embankment, Geosynthetics, Arching, Beam Theory, Catenary Theory
1 INTRODUCTION efficient design. Based on this, a more detailed

This contribution describes part of a study with the analysis in which the development of the arches and
aim of optimising the use of geosynthetic the need and placement of the GR will be studied
reinforcement (GR) in the method of geosynthetic physically and numerically.
reinforced pile-supported embankments (GRPSE)
with timber piles, also called light embankment
piling (Gunnvard et al. 2017). GRPSE is an
effective foundation method, which utilises arching
(Terzaghi 1943) and membrane effect (Villard et al.
2000) to transfer the embankment weight and traffic
load onto the piles instead of the soft subsoil. Light
embankment piling used in Sweden is a low-cost
alternative to GRPSE with steel or concrete piles, as
the material and transport cost of timber often is far
less. More importantly, timber piling allows for a
solution with lower carbon footprint than concrete
or steel piling.
The use of timber piles and the exclusion of pile
caps in lightly piled embankments result in a design
with very low pile coverage ratio, which could lead
to punching failure and unwanted settlements. To
counter this, the lightly piled embankment should,
according to Swedish standard TK Geo 13 (STA
2016), be reinforced with two layers of GR (Case 3
in Figure 1) and the centre-to-centre pile spacing (𝑠𝑠)
is limited to 0.8 ≤ 𝑠𝑠 ≤ 1.2m. This requires a lot of
timber piles and GR. A layer of dense soil is added
on top of the pile heads in order to distribute the load
better and to keep the soil around the timber piles
fully saturated, as the timber will otherwise rot over
time. The layer thicknesses 𝑡𝑡1 , 𝑡𝑡2 and 𝑡𝑡3 are the
minimum possible from practical point of view but
means that the height of the arch above the GR (ℎ′ )
is less than half the arch height (ℎ) for the beam
theory design if 𝑠𝑠 ≤ 1.2𝑚𝑚, assuming semi-circular
arch formation. Satibi (2009) showed using
numerical modelling that the GR layers need to be
placed within the lower half of the arch height (ℎ′ ≥
ℎ/2) to efficiently reduce subsoil settlements (𝑑𝑑).
Thus, it needs to be evaluated whether increasing 𝑠𝑠
or adjusting the placement and stiffness of the two
GR layers is the most resource efficient or if a single
GR layer (Case 2 in Figure 1) could improve Figure 1. Illustration of a semi-circular arch formation
enough, as well as if an exclusion of GR entirely between two piles in relation to the three different cases of
(Case 1 in Figure 1) could yield a more resource geosynthetic reinforcement (GR). Dimensions in case 3 are
according to TK Geo 13 (STA 2016) with 𝑠𝑠 = 1.2m.
301
The aim of the analyses also includes the potential GRPSE design as the piles will be modelled as short
of reducing the minimum embankment height (𝐻𝐻) rigid blocks.
and increasing the maximum allowed 𝑠𝑠 above 1.2m
for lightly piled embankments in TK Geo 13, further 4 ACKNOWLEDGEMENTS
saving resources. This study has been supported by the Swedish
transport administration (STA) and the Swedish
2 LABORATORY MODEL joint research program for road and railway
A 1-g physical experiment in scale 1:1 or 1:2 will geotechnology Branschsamverkan i grunden (BIG).
be developed at Luleå University of Technology to The authors would like to express their gratitude for
study the load distribution between two adjacent the financial support.
piles while simulating the cross section of an
embankment and load transfer platform. The test 5 REFERENCES
setup is the part of the pile-supported embankment Briançon, L., and Simon, B. 2012. Performance of pile-
shared by two adjacent piles with centre-to-centre supported embankment over soft soil: Full-scale
pile spacing 𝑠𝑠 (see Figure 1). The width of the model experiment. Journal of Geotechnical and
is 𝑠𝑠 + 𝑎𝑎, where 𝑎𝑎 is the equivalent square width of Geoenvironmental Engineering, 138(4): 551–561.
a 0.25m diameter timber pile, and the height from doi.org/10.1061/(ASCE)GT.1943-5606.0000561.
the pile heads equals 𝐻𝐻. Plane strain is chosen to
enable optical measurements of soil and GR da Silva, T. S., Elshafie, M. Z. E. B., and Madabhushi, G.
S. P. 2016. Physical modelling of geosynthetic-
displacements using Particle Image Velocimetry
reinforced soils spanning voids. In Proceedings of the
(PIV), captured through sheets of plexiglass on 6th European Geosynthetics Congress, Ljubljana, 25-
either side of the model setup, drawing inspiration 28 September 2016. Turkish chapter of IGS, Istanbul,
from the trapdoor geotechnical centrifuge tests by pp.774–781.
da Silva et al. (2016) and Rui et al. (2016). This
would allow the observation of formation of arches Gunnvard, P., Mattsson, H., and Laue, J. 2017.
in real time while measuring pile loads as well as Numerical analysis of the mechanical behaviour of
deformation (settlements) between the piles. It will light embankment piling. In Proceedings of the
also allow observation of the development of voids 19thICSMGE, Seoul 17-22 September 2017. Korean
below the GR (or the dense soil). Traffic load will Geotechnical Society, Seoul, pp.749–752.
be added via a hydraulic piston from the top.
Rui, R., van Tol, F., Xia, X.-L., van Eekelen, S., Hu, G.,
The setup of the physical experiment will be based and Xia, Y.-y. 2016. Evolution of soil arching; 2D
on the current design criteria in TK Geo 13 for DEM simulations. Computers and Geotechics, 73:
lightly piled embankments, i.e. Case 3, as well as 199–209. doi.org/10.1016/j.compgeo.2015.12.006.
Case 1 and 2 in Figure 1. Only the upper-most part
of the piles will be included and made from steel to Satibi, S. 2009. Numerical analysis and design criteria of
exclude the stiffness of the timber. The subsoil will embankments on floating piles. Mitteilung 62, Institute
be modelled as a cushion filled with water, where of Geotechnical Engineering, University of Stuttgart,
the subsoil settlements are controlled via drainage Stuttgart.
during the test through a tap (van Eekelen et al.
STA. 2016. Trafikverkets tekniska krav för
2012). Load cells will be placed on the two piles to
geokonstruktioner - TK Geo 13 (TDOK 2013:0667),
measure arching and the GR strain will be measured Swedish Transport Administration (STA), Borlänge,
by fibre optics (Briançon & Simon, 2012) or PIV Sweden. (In Swedish)
analysis.
Terzaghi, K. 1943. Theoretical Soil Mechanics. In Karl
3 OUTLOOK Terzaghi. John Wiley & Sons.
The analyses of the physical modelling and
van Eekelen, S. J. M., Bezuijen, A., Lodder, H. J., and
complementary numerical modelling will allow the van Tol, A. F. 2012. Model experiments on piled
revision of the design criteria in TK Geo 13 for embankments. Part I. Geotextiles and Geomembranes,
lightly piled embankments. A relationship will be 32: 69–81. doi.org/10.1016/j.geotexmem.2011.11.002.
developed between 𝐻𝐻, clear spacing between piles
(𝑠𝑠 − 𝑎𝑎) and GR placement and stiffness. The results Villard, P., Gourc, J. P., and Giraud, H. 2000. A
will improve timber piling as an environmental and geosynthetic reinforcement solution to prevent the
cost-efficient alternative in GRPSE design. formation of localized sinkholes. Canadian
However, the results are applicable on general Geotechnical Journal, 37: 987–999. doi.org/
10.1139/cgj-37-5-987.
302
Co-seismic response of a mat-founded structure on deposits with silty-sand layers.

April J. Bowman1, Orestis Adamidis2, Shideh Dashti3, Ioannis Anastasopoulos2
1
Centrifuge Research Center, Engineer, Research & Development Center, United States Corps of Engineers, USA
2
Institute for Geotechnical Engineering, ETH Zurich, Switzerland
3
College of Engineering and Applied Science, University of Colorado Boulder, USA
Corresponding author: April J. Bowman (april.j.bowman@erdc.dren.mil)
Keywords: Silty-Sand; Liquefaction; Centrifuge; Settlement; Rotation.
1 INTRODUCTION response of a mat-founded structure. A sketch of the

Deposits of loose, non-plastic silt and silty-sand models is given in Figure 1. The soil layers were
liquefied during the 1999 Kocaeli earthquakes and prepared by dry pluviation into a flexible-shear-
the 2010-2011 Canterbury earthquake sequence, beam container, using an automated sand-pourer
causing significant damage to structures with described by Kirkwood and Dashti (2018a). The
shallow foundations and their connected lifelines same soil profile was used in each model consisting
(e.g. Bray et al., 2004; Cubrinovski et al., 2011). of a lower 10m-thick dense, made of clean, Ottawa
Both in Adapazari and in Christchurch, sand (Dr = 90%), a 6m-thick layer of liquefiable
discrepancies between liquefaction triggering silty-sand whose properties varied for each test as
criteria and field observations existed, which were described in Table 1, and a 2m-thick surface crust
attributed to the influence of fine grained soils layer of dense Monterey sand (Dr = 90%). Three
(Sancio et al. 2002; Beyzaei et al. 2018). Element levels of non-plastic fines content were examined,
tests performed on high-quality extraction samples 0%, 5%, and 15%. In all cases, the relative density
showed that the content of clay and the stratigraphy was kept nearly the same, calculated using the
must be taken into account for the evaluation of relevant maximum and minimum void ratio for each
liquefaction triggering. In Adapazari, a set of layer. Soil parameters for the lower layer and the
updated criteria was developed to account for the surface layer can be found in Kirkwood and Dashti
influence of fines (Bol et al. 2008). In Christchurch, (2018b). The structure examined was placed at the
extensive testing showed that the key problem in centre of the model, as shown in Figure 1, and was
evaluating the influence of fines is their effect on instrumented using LVDTs and MEMS
both density and penetration resistance. This double accelerometers.
effect can be reconciled by using appropriate
Table 1. Material parameters for the liquefiable layers of
density measures for comparison (Cubrinovski et al. silty-sand used in each test.
2010). Parameter Unit Test 5 Test 3 Test 4
Despite significant advances in our understanding Fines content, fc % 0 5 15
of the influence of fines on the element response of Relative density,
saturated granular soils, uncertainty remains % 40 36 42
Dr
regarding boundary-value problems involving soil- Minimum void
structure interaction, which include layers of such - 0.530 0.420 0.216
ratio, emin
intermediate materials (Sancio et al. 2002). In order Maximum void
- 0.810 0.700 0.697
to investigate this problem, three dynamic ratio, emax
centrifuge experiments were performed, modelling Mean grain size,
mm 0.18 0.22 0.21
a mat-founded, single-degree-of-freedom structure d50
resting on a soil profile that included a layer of
loose, saturated silty-sand, whose fines content was
altered for each experiment. The goal was to
investigate which density measure method might be
appropriate for the silty-sand layers, in order to
generate an equivalent response for the structures.

Three centrifuge experiments were performed at the
University of Colorado Boulder, examining the Figure 1. Layout of centrifuge models. Dimensions in
prototype scale.
303
The models were saturated using a pore fluid non-plastic fines content was examined in each test:
whose viscosity was appropriate (Adamidis and 0%, 5%, and 15%. In all cases, the relative density
Madabhushi, 2015). A centrifugal acceleration of was kept about the same, calculated using the
70g was used. A major 1-D horizontal earthquake relevant maximum and minimum void ratio. A
motion was applied, called “Kobe-L”, full details of similar relative density led to a similar structural
which can be found in Kirkwood and Dashti response in terms of settlement and rotation,
(2018b). irrespective of the increase in fines, up to FC = 15%.
3 RESULTS AND DISCUSSION 5 REFERENCES

Figure 2 displays the combined results from LVDTs Adamidis, O. and Madabhushi, G.S.P., 2015. Use of
1 through 4 positioned on the top four corners of the viscous pore fluids in dynamic centrifuge modelling.
structure (Figure 1) for settlement, plotted versus International Journal of Physical Modelling in
rotation, which was calculated using results from Geotechnics, 15(3), 141-149.
MEMS accelerometers for the high-frequency
Beyzaei, C.Z., Bray, J.D., Cubrinovski, M., Riemer, M.
rotation component and LVDTs for the low and Stringer, M., 2018. Laboratory-based
frequency rotation trend. The strong pulses in the- characterization of shallow silty soils in southwest
earthquake led to significant settlement and Christchurch. Soil Dynamics and Earthquake
rotations. Settlement of more than 0.35 m was Engineering, 110, pp.93-109.
accumulated at the end of the earthquake for all
three experiments. However, not much residual Bol, E., Önalp, A., Arel, E., Sert, S. and Özocak, A.,
rotation was recorded at the end of the excitation, 2010. Liquefaction of silts: the Adapazari criteria.
for any of the three experiments. In Test 3, where Bulletin of Earthquake Engineering, 8(4): 859-873.
fines content, FC = 5%, a re-centering response was
Bray, J. D., Sancio, R. B., Durgunoglu, T., Onalp, A.,
observed towards the end of shaking.
Youd, T. L., Stewart, J. P., Seed, R. B., Cetin, O. K.,
Overall, the settlement-rotation response was Bol, E., Baturay, M. B., Christensen, C., and
remarkably similar for all tests, showing that Karadayilar, T., 2004. Subsurface Characterization at
relative density, calculated using the relevant values Ground Failure Sites in Adapazari, Turkey. Journal of
of emin and emax for each layer, is a good indicator of Geotechnical and Geoenvironmental Engineering,
equivalency in terms of structural response, when 130(7): 673–685.
considering layers with increasing fines content.
Cubrinovski, M., Bray, J. D., Taylor, M., Giorgini, S.,
-0.1 Bradley, B., Wotherspoon, L., and Zupan, J. 2011. Soil
Liquefaction Effects in the Central Business District
-0.05
during the February 2011 Christchurch Earthquake.
0
Seismological Research Letters, 82(6): 893–904.
0.05
Cubrinovski, M., Rees, S. and Bowman, E., 2010. Effects
0.1
of non-plastic fines on liquefaction resistance of sandy
soils. In Earthquake Engineering in Europe (125-144).
0.15
Springer, Dordrecht.
0.2
Kirkwood, P. and Dashti, S., 2018a. A centrifuge study
0.25 of seismic structure-soil-structure interaction on
0.3
liquefiable ground and implications for design in dense
urban areas. Earthquake Spectra, 34(3), 1113-1134.
0.35
0.4
Kirkwood, P. and Dashti, S., 2018b. Considerations for
-0.02 -0.01 0 0.01 0.02 the Mitigation of Earthquake-Induced Soil
Liquefaction in Urban Environments. Journal of
Figure 2. Rotations and settlements of structure during the Geotechnical and Geoenvironmental Engineering,
earthquake. 144(10).
4 CONCLUSIONS Sancio, R.B., Bray, J.D., Stewart, J.P., Youd, T.L.,

Durgunoǧlu, H.T., Önalp, A., Seed, R.B., Christensen,
Three dynamic centrifuge experiments were C., Baturay, M.B. and Karadayılar, T., 2002.
presented, examining the response of a mat-founded Correlation between ground failure and soil conditions
structure resting on a saturated soil profile that in Adapazari, Turkey. Soil Dynamics and Earthquake
includes a layer of silty-sand. A different level of Engineering, 22(9-12): 1093-1102.
304
Author Index
Abdoun T. H. 281 Hashemi M. A. 109 Pajalić S. 115

Abrashitov A. 95 Henke S. 125 Peranić J. 115
Adamidis O. 303 Herduin M. 247 Pettey A. 57
Agalianos A. 281 Heron C. M. 57, 109, 139, 235 Prendergast L. J. 235
Alén C. 103 Honardar S. 285 Price C. 291
Anastasopoulos I. 281, 297, 303 Prodan M. V. 115
Arbanas Ž. 115 Jafari M. K. 201
Arnold A. 287 Jagodnik V. 115 Quinten T. O. 243
Askarinejad A. 243, 263, 285, 287 Javed S. 295
Azad A. 205 Jayanandan M. 213 Ritchie. E. 25
Jeffcock O. 269 Robinson S. 185
Bergström A. 295 Jensen M. R. 17 Rocchi I. 9
Bezuijen A. 75 Jia Qi 301 Roddy D. 227
Bhattacherjee D. 213, 221 Jimenez R. 139 Rotisciani G. M. 49
Blanc M. 89, 275, 293 Jones L. 297
Bowman A. J. 291, 303 Savin A. 95
Brennan A. 185 Kim Dong-Soo 283, 289 Schiavon J. A. 151
Breure W. 163 Kim Yeon-Sam 283 Sidrakov A. 95
Kim Yoon-Ah 289 Silva da D. M. 143
Casini F. 49 Knutsson S. 299 Song Geyang 139
Ćeh N. 115 Ko Kil-Wan 283 Stallebrass S. 157
Cheng Kunkun 235 Korre E. 281 Stapelfeldt M. 255
Christensen M. S. 9 Kumar P. 213 Stone K. 227, 269
Coelho P. 31 Kumar V. 177 Støren A. 83
Connolly D. 185 Kundu S. 43 Sturm H. 83
Sudhakaran K. 263
Dashti S. 303 Lagerlund J. 171
de Lange D. 65, 75, 133, 285 Lagrouche O. 185 Terwindt J. 65, 75, 133
de Sauvage J. 3 Lalicata L. M. 25, 49, 157 Tessari A. F. 291
Desideri A. 49 Laue J. 171, 299, 301 Thorel L. 49, 89, 151, 275, 293
Diakoumi M. 227, 269 Lee Hae-In 289 Tohidifar H. 201
Dijkstra J. 103, 295 Lehane B. M. 17 Tommik K. 299
Divall S. 25 Li Zheng 39 Toromanovic J. 171
Dubreucq T. 3 Li Zhong-Sen 293 Tsuha C. H. C. 143, 151
Lintzén N. 299
Escoffier S. 39 Lundberg A. B. 163 van der Star W. 65
Esen A. 185 van’t Hof C. 243
Maatouk S. 89 Viklander P. 171
Franza A. 193 Madabhushi S. P. G. 31, 177 Viswanadham B. V. S. 43, 213, 221
Mahmoudi M. Z. A. 205
Gaudin C. 247 Manandhar S. 289 Woodward P. 185
Gavin K. 243 Marques A. 31
Goodey R. 25 Marshall A. M. 109, 139, 193, 235 Xu Jingmin 109, 193
Gorlov A. 95 McNamara A. 157
Grabe J. 255 Moosavi M. 201 Yannie J. 103
Grima M. A. 243 Munro C. 185
Gunnvard P. 301 Zania V. 9, 17
Gutiérrez-Ch J. G. 139 Naghibi M. 205 Zarandi A. E. 205
Zaytsev A. 95
Haffar El I. 275 O’Loughlin C. 247 Zhang Yuen 263
Haigh S. 31 Zwaan R. 75
305

Physical Modelling

Uploaded by

Copyright:

Available Formats

Physical Modelling

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Physical Modelling

Uploaded by

Copyright:

Available Formats

ECPMG 2020

4th European Conference on

4th European Conference on

Jan Laue & Tarun Bansal

ISBN (print): 978-91-7790-542-4

Silty sand characterization in the geotechnical centrifuge 9

The effects of sand grading on the bearing capacity of surface foundations 17

Tests of varied sample preparation methods for centrifuge modelling 25

Accelerations in dynamic centrifuge modelling of seismic liquefaction problems 31

Boundary effects in centrifuge modelling of soil liquefaction 39

Model preparation for unsaturated soil testing in a centrifuge environment 49

Effect of Crushed Particles on Soil-Structure Interface Behaviour 57

Carboxymethylcellulose as a Newtonian viscous fluid for centrifuge modelling 65

2 Physical modelling techniques and facilities

A new geotechnical centrifuge at Deltares, Delft, the Netherlands 75

A Novel Small-Scale Model Testing Device for Laterally Loaded Piles 83

Development of a Hammer to drive monopiles of offshore wind turbines in Centrifuge 89

3 Piles: installation and capacity

Centrifuge modelling of shaft resistance of a rough rock-socketed pile 139

Geotechnical instrumentation of an experimental embankment dam 171

Centrifuge modelling of train passage over clay subgrades 185

The modelling of piles under multidirectional loading 247

Centrifuge Modelling of Suction Caissons Subjected to Cyclic Loading in Tension 263

Behaviour of prefabricated concrete screw piles under axial loading 287

High density soil instrumentation for a thermal pile test 295

Physical Modelling of Geotechnical structure subjected to Hydraulic Scour 297

Distribution of temperature in culverts in cold climate 299

Geosynthetic reinforcement in lightly piled embankments: Laboratory model development 301

Co-seismic response of a mat-founded structure on deposits with silty-sand layers 303

Author Index 305

Dr.-Ing. Jan Laue

Gustave Eiffel, France

Keywords: Friction; Scaling Law; Centrifuge Modelling; Reinforced Soil.

1 INTRODUCTION tests were carried out as well as numerical

Figure 1. Lateral friction against the reduced displacement

The peak shear strength, qs, is defined as the

Chareyre, B. and Villard, P. 2002. Discrete element

Cundall, P. A. and Strack, O. D. L. 1979. A discrete

Dubreucq, T. 1999. Renforcement des fondations

Garnier, J. et al 2007. Catalogue of scaling laws and

Schlosser, F. et al, 1993. Recommendations Clouterre

Silty sand characterization in the geotechnical centrifuge

Corresponding author: Varvara Zania (vaza@byg.dtu.dk)

1 INTRODUCTION Mayne,1990, and Robertson and Cabal, 2015). As

Cone penetration depth [mm]

The sleeve friction for the three samples is Pore pressure, u2

sleeve friction. Nevertheless, the sleeve friction at

4.1. Effect of stress level qc/σ'v [-] 40

The effect of the stress on the cone penetration has 30

Normalized cone resistance Qtn

Normalized friction ratio Fr

The soil behaviour type index Ic which is

10 where Qt is dimensionless cone resistance, and Fr is

The effects of sand grading on the bearing capacity of surface foundations

Civil Engineering Department, University of Western Australia, Australia

Corresponding author: Varvara Zania (vaza@byg.dtu.dk)

Keywords: Footings; CPT; Centrifuge Modelling; Sands; Sand Grading.

1 INTRODUCTION 𝑞𝑞 = 𝑟𝑟𝑠𝑠 · �𝑠𝑠/𝐵𝐵 (1)

Percentage passing (%)

2 SOIL PROPERTIES 2.1. Basic indices