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

net/publication/362049808

The role of seismic intensity on the performance of caisson foundations

supporting bridge piers: preliminary results from dynamic centrifuge testing

Conference Paper · July 2022

CITATIONS READS

0 75

4 authors:

Domenico Gaudio Sebastiano Rampello

Sapienza University of Rome Sapienza University of Rome
28 PUBLICATIONS 101 CITATIONS 90 PUBLICATIONS 1,081 CITATIONS

SEE PROFILE SEE PROFILE

Gopal Madabhushi Giulia Viggiani

University of Cambridge University of Cambridge
307 PUBLICATIONS 5,357 CITATIONS 30 PUBLICATIONS 158 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

buildings View project

Advanced Technologies for Post-Disasters Reconnaissance, Forensic and Environmental Impact Studies-Geotechnical View project

All content following this page was uploaded by Domenico Gaudio on 16 July 2022.

The user has requested enhancement of the downloaded file.

 3rd International Conference on Natural Hazards & Infrastructure
5-7 July 2022, Athens, Greece

The role of seismic intensity on the performance of caisson foundations

supporting bridge piers: preliminary results from dynamic centrifuge
testing
Domenico Gaudio, Ph.D.1, Sebastiano Rampello, Ph.D.
Sapienza University of Rome

Gopal S.P. Madabhushi, Ph.D, Giulia M.B. Viggiani, Ph.D

University of Cambridge

ABSTRACT

Seismic performance of caisson foundations supporting bridge piers may take advantage of soil inelastic
response when subjected to strong seismic events, thanks to the soil nonlinear and hysteretic behaviour. This
can bring to a substantial optimisation in caisson design and major cost savings. In the framework of Capacity
Design extended to geotechnical systems, temporary attainment of plastic mechanisms may be permitted
provided that the resulting permanent displacements are lower than given threshold values, which in turn
depend on the considered limit state and performance level required to the structure. Clearly, this new design
approach needs to be validated against physically-sound numerical and experimental simulations. A campaign
of dynamic centrifuge tests was therefore recently carried out at the Schofield Centre, University of Cambridge,
where the seismic performance of caisson foundations was assessed. In this paper, a preliminary interpretation
of the experimental results is given, shedding some light on the interplay between seismic intensity and
mechanical soil properties. Specifically, the results obtained in two tests are discussed, where a caisson-pier-
deck system was subjected to earthquakes of increasing intensity. In the two tests, a soft and very soft clay
layer was reproduced, to either avoid or promote the plastic soil behaviour. It is shown that the highly nonlinear
and hysteretic response of the very soft clay limits the inertial forces transmitted to the superstructure, thus
validating the above-mentioned design approach. The beneficial effect of inelastic soil behaviour entailed
permanent displacements increasing with earthquake intensity, which should be checked against limit state
prescriptions.

Keywords: caisson foundations, bridge piers, earthquakes, centrifuge modelling

INTRODUCTION

In the presence of strong seismic shakings, inelastic soil behaviour is typically mobilised. However, current
code provisions do not allow the soil-foundation system to fall in the elastic-plastic regime, while development
of plastic hinges in the superstructure is allowed. Even if there are good reasons for this choice (e.g., post-
earthquake inspection of the foundation is usually unfeasible), this design approach requires some overstrength
factors to be applied in the foundation design, thus introducing a sort of overdesign (Gazetas, 2015).
Furthermore, this approach may be even dangerous for the superstructure, as this may be subjected to a high
ductility demand (Sakellariadis et al, 2020). Hence, plastic mechanisms occurring in the foundation soils may
be promoted to limit the inertial forces transmitted to the superstructure, leading to substantial cost savings
when either designing or retrofitting the structures at hand. Although attractive, this design procedure needs

1
Department of Structural and Geotechnical Engineering, e-mail: [email protected];
formerly Department of Engineering, University of Cambridge, CB3 0EF, Cambridge, UK
the consequent displacements and rotations experienced by the structure to be checked, aiming at lying below
given threshold values, which in turn depend on the considered limit state and on the performance level
required to the structure.
Gaudio & Rampello (2019a, b; 2021) recently showed that the above-mentioned concept may be successfully
applied in the design of caisson foundations supporting bridge piers. Their results were obtained by performing
3D Finite Element numerical analyses in the time domain: hence, dynamic centrifuge tests on reduced-scale
models were recently carried out at the Schofield Centre, University of Cambridge, on the purpose of validating
the numerical results, understanding the physical phenomena underlying the seismic performance and
providing an experimental dataset for possible macro-element calibration.
In this paper, a preliminary interpretation of the experimental outcomes is given, stressing out the interplay
between the seismic intensity and mechanical soil characteristics. Figure 1 shows the problem layout, in which
a cylindrical caisson foundation of diameter D = 8 m and slenderness ratio H/D = 1 is embedded in a typical
alluvial deposit, made by a thin sand layer (H1 = 3 m) underlain by a soft or very soft clay deposit (H2 = 14 m).

ms

hs = 15 m ks, s

x
loose sand 3
H=8m z

14
D=8m
soft/very soft clay

direction of shaking

Figure 1. Schematic layout of the problem (prototype scale)

The water table is located at the sand-clay interface and the pore water pressure regime is hydrostatic. The pier
is represented by a Single Degree of Freedom (SDoF) system with a height hs = 15 m: its mechanical properties
(lumped mass ms and stiffness ks) were back-calculated to provide a given average contact pressure,
q = 140 kPa, and a fixed-based period, Ts = 0.50 s, both representative of a typical bridge configuration. The
seismic input is applied in terms of a horizontal acceleration time history at the bedrock depth (z = 17 m), here
assumed as infinitely rigid and therefore acting as a fully-reflecting boundary.
The results obtained in two out of three dynamic centrifuge tests are discussed in this paper, namely DG01 and
DG03, the former characterised by a caisson-pier system resting on a soft clay, the latter on a very soft clay
layer. Although the schematic layout considered in the centrifuge is not exactly the same as in the numerical
analyses, the experimental layout allowed to validate the applicability of the numerical results to different soil-
caisson-pier configurations.

CENTRIFUGE MODELLING

A reduced-scale model was produced to simulate the problem layout on the Turner beam centrifuge of the
Schofield Centre at University of Cambridge, UK. The centrifuge model was prepared and spun at a nominal
centrifugal acceleration of 60g.
The model container used was the most recent Equivalent Shear Beam (ESB) container (Brennan &
Madabhushi, 2002): reliability of this container in dynamic centrifuge testing has been recently shown and
discussed (Gaudio et al, 2022a). The caisson foundation was simulated through a hollow aluminium cylinder
topped by two circular aluminium plates, while the pier was modelled using an aluminium rod with a squat
brass cylinder at the top to reproduce the mass of the deck. At the model scale, the caisson is characterised by
a diameter of 133 mm, equal to its depth, while the pier height is equal to 250 mm (Fig. 2). The hole into the
caisson foundation was introduced to reproduce the mass of a concrete caisson (mcaisson = 4.74∙10-3 Mg,
corresponding to 1024.8 Mg at prototype scale), while the superstructure is reproduced by the brass mass
mdeck = 0.9∙10-3 Mg (194.4 Mg) and the pier, the latter characterised by a mass mpier = 0.665∙10-3 Mg (143.6 Mg)
and a bending stiffness EI = 2.37∙10-6 GN∙m2 (30.7 GN∙m2). Hostun sand HN31, available at the Schofield
Centre, was glued on the lateral surface of the caisson foundation to reproduce the roughness of the soil-
concrete contact (Fig. 3a).

Figure 2. Cross-sectional view of the reduced-scale model and miniaturized instrumentation – model
scale in mm (bracketed prototype scale in m)

The soil deposit was produced in two steps. First, the clay slurry was prepared by mixing speswhite kaolin
clay powder and de-aired water in 1:1.3 ratio, and then the slurry was poured into the ESB box. Two different
profiles of undrained shear strength su were targeted, both slightly increasing with depth, as for soft clays, and
with average values equal to about 40 kPa (test DG01) and 10 kPa (DG03). The desired profiles were obtained
via a combination of 1D loading and hydraulic consolidation by suction-induced seepage (Garala &
Madabhushi, 2019) and are plotted in the following (Fig. 4b). After consolidating, the clay surface was
trimmed to obtain the desired depth of 233 mm (Fig. 2) and the instrumentation was inserted in the model. The
structure was then installed and the clay was covered with a layer of loose sand (relative density DR ≈ 50 %).
The instrumentation includes two far-field alignments of miniature Pore Pressure Transducers (PPTs, P1-4 in
Fig. 2) and piezo-electric accelerometers (A1-6): two piezo-electric accelerometers were also installed at the
outer base of the ESB container to record the applied seismic input (A8-9, with A9 not shown in the Figure as
located out of plane), while PPT P5 and piezo-electric accelerometer A7 were installed beneath the caisson
base. Horizontal and vertical acceleration time histories were recorded by Micro-Electro-Mechanical-Systems
(MEMs) glued on the brass mass (M1-4) and on the caisson (M5-8), whereas the average settlement and rigid
rotation of the caisson were measured using two Linear Variable Differential Transformers LVDTs (L2-3).
Another LVDT (L1) was installed at the clay surface along the far-field alignment to record vertical
displacements. The model was also equipped with the Air Hammer Device (AHD; Ghosh & Madabhushi,
2002) and T-bar penetrometer (Stewart & Randolph, 1984) to measure the shear wave velocity and the
undrained shear strength of the soil, respectively; the T-bar location is also given in the picture of the model
mounted on centrifuge reported in Figure 3b. Both the AH and T-bar tests were performed at 60g to obtain the
soil stiffness and strength right before the earthquakes were applied.

Figure 3. (a) system considered in the centrifuge tests DG01 and DG03; (b) model mounted in the
centrifuge

The experimental profiles of the small-strain shear modulus G0 and undrained shear strength su are plotted in
Figure 4, the former computed through the well-known equation for an elastic medium

G0  z  = ρ  VS,0  z  
2
(1)

where  is the mass density and VS,0 is the “small-strain” shear wave velocity. The undrained shear strength of
the clay layer was obtained from the following relation:

qc  z 
su  z   (2)
NT

where qc is the penetration resistance and NT = 12 is the T-bar bearing factor which was calibrated against the
results from miniaturised shear vane tests (Gaudio et al, 2022b). As for the small-strain shear modulus G0, the
average values of about 41 and 21 MPa were obtained for the soft (test DG01) and very soft (test DG03) tests,
respectively, while average values of the undrained shear strength su were equal to about 43 and 6 kPa (Tab. 1).
The experimental results were compared with values provided by some empirical relations available in the
literature, such as those by Azeiteiro et al (2017) and Hardin & Drnevich (1972) for the small-strain shear
modulus G0 and the one from Wroth (1984) for the undrained shear strength su. The best agreement with the
experimental results was obtained for the DG03 test in terms of soil stiffness, while a good accordance was
observed for both tests in terms of strength.
 DG01 DG03
Az. et al. (2017)
H. & D. (1972)
Wroth (1984)

(a) (b)

Figure 4. Experimental and empirical (a) small-strain shear modulus and (b) undrained shear strength
profiles
Table 1. Average mechanical properties of the far-field soil column in the two centrifuge tests
test G0 (MPa) su (kPa)
DG01 41.2 43.3

DG03 26.6 6.4

In both tests DG01 and DG03, the system was subjected to several real and sinusoidal seismic inputs. Here the
results obtained applying the seismic input recorded during the destructive Christchurch (2011) earthquake are
discussed; its amplitude was scaled to obtain a weak (axinpmax = 0.030g, Fig. 6d), a moderate (axinpmax = 0.100g,
Fig. 7d), and a strong (axinpmax = 0.212g, Fig. 8d) version of the seismic record. The first and the third base
excitations were applied to the base of the weak clay (DG03), while the second one to the base of the stronger
clay (DG01), so as to study the influence of the seismic intensity on the system performance. The ground
motions were imposed at the base of the ESB box through the servo-hydraulic shaker described by Madabhushi
et al (2012).

EXPERIMENTAL RESULTS

In this section, the results obtained applying the seismic inputs mentioned above are discussed, first referring
to the far-field soil response and then to the seismic behaviour of the whole soil-caisson-pier-deck system.

Far-field soil response

The far-field soil response is given in Figure 5, where the shear stress-shear strain loops obtained at depth
z = 6.5 m, which is close to the caisson base, is given for all cases considered in this paper, namely the weak
input + weak soil (i.e. very soft clay, DG03 test, Fig. 5a), the moderate input + strong soil (i.e. soft clay, DG01,
Fig. 5b), and the strong input + weak soil (i.e. very soft clay, DG03, Fig. 5c). In the first case, low seismic-
induced shear strains (max = 2.5∙10-3) were mobilised, which were accompanied by quite a narrow loop area
and consequent nearly null energy dissipation: in this case, the induced shear stress  in the soil is far from the
limit shear stress, max = su. Vice versa, the third case (Fig. 5c), ruled by the high seismic intensity travelling
into a weak clay deposit, shows an almost tripled peak shear strain (max = 7.0∙10-3) together with a much
broader loop area, which indicates a remarkable soil hysteretic damping: this time, the undrained shear strength
is attained, limiting the shear stress  that can be transmitted to the upper portion of soil deposit. The moderate
input + strong soil (Fig. 5b) locates in the middle between the two extreme cases, showing a peak shear strain
max which is comparable to that mobilised in the strong input case, although not triggering soil shear strength.
Therefore, the first case (Fig. 5a) can be seen as representative of an elastic case, whereas the second (Fig. 5b)
and the third ones (Fig. 5c) can be viewed as cases where soil inelastic behaviour is slightly and strongly
mobilised, respectively, mainly due to the interaction between the increasing amplitude of the seismic shaking
and the varying shear strength of the soil. This interplay will be influencing the seismic performance of the
entire soil-caisson-pier-deck system, as discussed in the next section.

(a) (b) (c)

Figure 5. Far-field shear stress-shear strain loops (depth z = 6.5 m) for increasing amplitude of the
scaled Christchurch motion: (a) weak input on weak soil (test DG03); (b) moderate input on strong soil
(test DG01); (c) strong input on weak soil (test DG03)

Seismic behaviour of the system

The seismic behaviour of the system at hand was quantified in terms of the nondimensional deck drift urel/hs
and of the bending moment acting at the base of the pier, Ms. The deck drift is defined as

urel  udeck  ucaisson head  tan   hs  uflex  urel, rigid  uflex (3)

where udeck and ucaisson head are the horizontal displacements of the deck and the top of the caisson,  is the rigid
caisson rotation, urel, rigid is the rigid component of the drift and uflex is the elastic bending of the pier, that is the
one directly related to the bending moment Ms. The latter has been computed as

Ms  meff  adeck  hs (4)

where meff = 270.43 Mg is the effective mass related to the first mode of the fixed-base pier. The time histories
of horizontal displacement were computed through double integration of the ones of average acceleration as
obtained from MEMS M1-2 and M5-6, while the time history of the rigid rotation  was computed as  = [wL3–
wL2]/d, where w indicates the caisson settlement and d is the distance between LVDTs L2-3. All results presented
hereafter are at the prototype scale, unless otherwise specified.
The observed time histories of drift and bending moment are plotted in Figures 6, 7, and 8 for the three
increasing amplitudes of the ground motion. In the weak input + weak soil case (Fig. 6), no permanent drift is
obtained (urel, rigid/hs = 0) due to the very low intensity of the input motion. This result is consistent with the far-
field shear stress-shear strain loop computed at a depth z = 6.5 m (Fig. 5a). This outcome also allows to state
that soil response is about elastic when subjected to the weak ground motion.
With increasing seismic amplitude (test DG01), the system starts accumulating a permanent rigid rotation
(Fig. 7b). However, this is still relatively small (urel, rigid/hs = -0.18 ‰) as a consequence of the higher undrained
shear strength measured along the far-field alignment. Hence, in this test, irreversible soil behaviour influenced
the system response only very slightly, while still limiting the inertial forces transmitted to the superstructure;
indeed, despite the input amplitude increased more than three times (ratio between axinpmax equal to
0.102/0.030 = 3.4), the increment obtained in the peak values of the deck drift, urel, was only 33 % and the
maximum bending moment, Ms, increased by 70 %.
Conversely, in the case of the strong seismic input applied to the weak soil, a strong increase of the permanent
rotation was accumulated by the system, which is tripled (= 0.60 ‰, Fig. 8b). Nonetheless, the bending
moment turned out to be lower than that attained by the moderate input on the strong soil case (Fig. 8c), which
decreased by about 32 %, despite the input amplitude doubled. The peak bending moment attained in this case
(Ms, max = 11.37 MNꞏm) is limited by the soil-caisson capacity, as discussed in the following section. It is
therefore apparent that the irreversible and hysteretic soil behaviour is contributing to reduce the inertial forces
transmitted to the superstructure, governing the system seismic performance. This is also clear from the
noticeable period elongation of both the time histories of deck drift and bending moment plotted in Figure 8.

(a) (c)

(b) (d)

Figure 6. Time histories of the (a) total and (b) rigid component of the deck drift, (c) bending moment and
(d) horizontal acceleration of the weak input motion on weak soil (DG03)

(a) (c)

(b) (d)

0 5 10 15 20 25
t (s)
Figure 7. Time histories of the (a) total and (b) rigid component of the deck drift, (c) bending moment and
(d) horizontal acceleration of the moderate input motion on strong soil (DG01)
 (a) (c)

(b) (d)

0 5 10 15 20 25
t (s)
Figure 8. Time histories of the (a) total and (b) rigid component of the deck drift, (c) bending moment and
(d) horizontal acceleration of the strong input motion on weak soil (DG03)

ASSESSMENT OF SEISMIC PERFORMANCE

From the seismic behaviour discussed in the previous section, it follows that the system response was assessed
referring to some performance indexes, such as the peak and permanent values of the nondimensional deck
drift, urel, max/hs and urel, perm/hs, and the peak bending moment Ms, max, which are plotted in Figure 9 against the
peak acceleration of the seismic shaking, axinpmax. This shows that all quantities increase in the low-intensity
range (i.e. axinpmax ≤ 0.1 g) for increasing amplitudes, as expected. Conversely, the peak deck drift almost
stabilises while the peak bending moment even decreases in the moderate-to-high intensity range
(axinpmax > 0.1 g), being limited by the attainment of the soil-caisson resistance. The latter was estimated using
the empirical relations provided for clays by Gerolymos et al (2015) and Rosati et al (2022), the first developed
through a total stress approach, while the second with an effective stress approach. Their evaluation of the
limit bending moment acting at the base of the pier, causing the bearing capacity mechanism to occur, Ms,f, is
equal to 12.5 and 14.9 MN∙m, respectively, being in a good agreement with the value obtained in the centrifuge
(Ms,max = 11.37 MNm) for the test DG03 with axinpmax = 0.212 g. Higher values of Ms,max, lying in between the
two limit values Ms,f, have been obtained when applying other seismic inputs (not shown in this paper for the
sake of space), which confirms that the actual limit value might be located in between them (Gaudio et al.,
2022b).
Opposite to what observed for the peak values, the permanent deck drift shows an almost linear increase with
the seismic intensity, which confirms the growing role of inelastic soil behaviour on the seismic performance
of the system as the seismic amplitude increases.

(a) (b) (c)

Ms = Ms,f

Figure 9. Seismic performance against the peak acceleration of the seismic input: (a) peak and (b)
permanent nondimensional deck drift; (c) peak bending moment
 CONCLUDING REMARKS

This paper illustrated a preliminary interpretation of the results obtained from dynamic centrifuge tests of
caisson foundations supporting bridge piers subjected to weak, moderate and strong ground motions. The
discussion mainly showed that the inelastic and hysteretic behaviour of the foundation soil can play a
fundamental role in the seismic performance of the whole system; this is particularly true for strong seismic
events, for which the soil acts as a “fuse”, thus limiting the inertial forces transmitted to the superstructure.
Further elaboration of the discussed results is of course needed, e.g., to understand and quantify the relative
contribution of the free-field and near-field hysteretic soil behaviour, considering also the remaining seismic
inputs applied in the centrifuge tests.

REFERENCES

Azeiteiro RJN, Coelho PALF, Taborda DMG, Grazina JCD (2017). Critical State-Based Interpretation of the Monotonic
Behaviour of Hostun Sand. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 143(5), 04017004.
Brennan AJ & Madabhushi SPG (2002). Design and performance of a new deep model container for dynamic centrifuge
testing. Proc. Int. Conf. on Phys. Mod. Geotech., Balkema, Rotterdam, The Netherlands, St Johns NF, Canada, (pp. 183-
188).
Garala TK & Madabhushi SPG (2019). Seismic behaviour of soft clay and its influence on the response of friction pile
foundations. Bulletin of Earthquake Engineering 17(4), pp. 1919-1939.
Gaudio D, Madabhushi SPG, Rampello S & Viggiani GMB (2022b). Experimental investigation of the seismic
performance of caisson foundations supporting bridge piers, Géotechnique (submitted for publication)
Gaudio D & Rampello S (2019a). The influence of soil plasticity on the seismic performance of bridge piers on caisson
foundations. Soil Dynamics and Earthquake Engineering, 118, pp. 120-133.
https://doi.org/10.1016/j.soildyn.2018.12.007
Gaudio D & Rampello S. (2019b). The role of soil constitutive modelling on the assessment of seismic performance of
caisson foundations. Proc. 7th Int. Conf. Geot. Earthq. Eng., Rome, Italy, (pp. 2574-2582)
https://doi.org/10.1201/9780429031274
Gaudio D & Rampello S (2021). On the assessment of seismic performance of bridge piers on caisson foundations
subjected to strong ground motions. Earthquake Engineering & Structural Dynamics, 50(5), pp. 1429-1450.
https://doi.org/10.1002/eqe.3407
Gaudio D, Seong J, Haigh SK, Viggiani GMB, Madabhushi SPG, Shrivatsava R, Veluvolu R, Padhy P (2022a). Boundary
effects on dynamic centrifuge modelling of onshore wind turbines on liquefiable soils, International Journal of Physical
Modelling in Geotechnics (accepted for publication)
Gazetas G (2015). 4th Ishihara lecture: soil–foundation–structure systems beyond conventional seismic failure thresholds.
Soil Dynamics and Earthquake Engineering 68, pp. 23-39.
Gerolymos N, Zafeirakos A, Karapiperis K (2015). Generalized failure envelope for caisson foundations in cohesive soil:
Static and dynamic loading. Soil Dynamics and Earthquake Engineering 78, pp. 154-174.
Ghosh B & Madabhushi SPG (2002). An efficient tool for measuring shear wave velocity in the centrifuge. Proc. Int.
Conf. Phys. Mod. Geotech. (eds R. Phillips, P. J. Guo and R. Popescu), Balkema, Rotterdam, The Netherlands, (pp. 119–
124).
Hardin BO & Drnevich VP (1972). Shear modulus and damping in soils: design equation and curves. J. Soil Mech. Found.
Eng. Div. 98(SM7), pp. 667–692.
Madabhushi SPG, Haigh SK, Houghton NE Gould E (2012). Development of a Servo-Hydraulic Earthquake Actuator for
the Cambridge Turner Beam Centrifuge. International Journal of Physical Modelling in Geotechnics 12, No. 2, 77-88.
Rosati A, Gaudio D, di Prisco CG, Rampello S (2022). Use of interaction domains for a displacement-based design of
caisson foundations, Acta Geotechnica (under revision)
Sakellariadis L, Anastasopoulos I, Gazetas G (2020). Fukae bridge collapse (Kobe 1995) revisited: New insights. Soils
and Foundations 60 (6), pp. 1450-1467.
Stewart DP & Randolph MF (1991). A New Site Investigation Tool for the Centrifuge. Proc. Int. Conf. Centr. Mod.,
Centrifuge ’91, Boulder, Colorado, 13-14 June 1991, Balkema, Rotterdam, the Netherlands, (pp. 531-538).
Wroth CP (1984). The Interpretation of In Situ Soil Tests. Geotechnique 34(4), pp. 449–489.

View publication stats