INTERNATIONAL SOCIETY FOR

SOIL MECHANICS AND

GEOTECHNICAL ENGINEERING

This paper was downloaded from the Online Library of

the International Society for Soil Mechanics and
Geotechnical Engineering (ISSMGE). The library is
available here:

https://www.issmge.org/publications/online-library

This is an open-access database that archives thousands

of papers published under the Auspices of the ISSMGE and
maintained by the Innovation and Development
Committee of ISSMGE.

The paper was published in the proceedings of the 10th

European Conference on Numerical Methods in
Geotechnical Engineering and was edited by Lidija
Zdravkovic, Stavroula Kontoe, Aikaterini Tsiampousi and
David Taborda. The conference was held from June 26th to
June 28th 2023 at the Imperial College London, United
Kingdom.
Proceedings 10th NUMGE 2023
10th European Conference on Numerical Methods in Geotechnical Engineering
Zdravkovic L, Kontoe S, Taborda DMG, Tsiampousi A (eds)

© Authors: All rights reserved, 2023

https://doi.org/10.53243/NUMGE2023-10

Numerical modelling of uplifting a surface foundation on clay

S. Mei, Y. Tian, M.J. Cassidy

Department of Infrastructure Engineering, The University of Melbourne, Melbourne, Australia

ABSTRACT: Offshore shallow foundations are required to be retrieved at the end of their life cycle. The required force is much
larger than their own submerged weight due to significant suction generated at the invert of the foundation. This paper reports
numerical modelling of uplifting a surface foundation whilst employing a hydro-mechanic interface that can capture both me-
chanic and hydraulic interactions between foundation and soil. The soil behaviour was modelled using the modified Cam-Clay
model. After verifying against physical modelling data from centrifuge tests, the numerical results are interpretated in terms of
pore pressure distribution at the interface, deformation mechanism and stress path of soils, which enhances the understanding of
uplift response of surface footings on clay seabed.

Keywords: uplift, shallow foundation, rate effect, interface, coupled analysis

1 INTRODUCTION opening gap. This hydraulic behaviour cannot be de-

scribed by a simply fully bonded interface (such as the
Shallow foundations are used offshore to support one adopted in Li et al., 2015) or purely mechanical in-
platforms, pipelines and manifolds and as anchoring terface models, which is not a problem in compression
systems (Bye et al., 1995; Bouwmeester et al., 2009; modelling.
Randolph and Gourvenec, 2010). They are required to This paper reports numerical modelling of uplifting a
be decommissioned at the end of their life cycle (Gour- surface footing on clay seabed using a hydro-mechanic
venec, 2018). The extraction force required often ex-
interface model that can capture both the mechanic be-
ceeds their own submerged weight, especially when the haviour and hydraulic behaviour (Cerfontaine et al,
footings sit on fine-grained soils. The excessed re- 2015; Peng et al., 2022; Tian et al., 2022). The soil be-
sistance emanates from the negative pore pressure (i.e., haviour of the seabed was modelled by the elastoplastic
suction) generated at the invert of the footings during modified Cam-Clay (MCC) model (Roscoe and Bur-
uplift. Accurately predicting the uplift resistance can land, 1968). The numerical model was verified against
provide an appropriate guide for vessel deployment and a series of physical geotechnical centrifuge tests re-
avoid severe safety consequences. ported in Mei et al. (2023). The numerical results are
Previous experimental studies on uplift of shallow discussed and interpreted to improve the understanding
foundations (e.g., Chen et al., 2012; Li et al., 2014; Mei of uplifting problems.
et al., 2023) have shown that uplift resistance is rate-de-
pendent. Under rapid uplift, the suction underneath the
footing is fully developed and has no time to dissipate. 2 NUMERICAL MODEL
Undrained behaviour with a reverse bearing mechanism
is reached with an uplift capacity comparable to com- The small strain finite-element analysis presented in this
pression capacity. Under slow uplift, drained conditions paper was carried out using commercial software
are achieved. The suction is dissipated and the uplift ca- ABAQUS (Dassault Systèmes, 2016).
pacity only consists of mobilized footing-soil friction.
Between these two extremes, an intermediate rate ends 2.1 Geometry and materials
up with partially drained conditions. An axisymmetric numerical model was established to
Compared to compression, numerical modelling of retrospectively simulate the centrifuge tests of Mei et al.
this rate-dependent uplifting process is far more chal- (2023) as shown in Fig. 1. The soil region was set as five
lenging. Firstly, a coupled analysis of soil and water is times the footing diameter D, which was far enough to
required within the effective-stress framework as excess eliminate boundary influence. The footing was mod-
pore pressure is involved. Secondly, appropriate model- elled as a rigid body and had a reference point at the top
ling of the interface between the footing and soil is com- centre. The footing and the soil were discretised as 4-
plex but essential. During uplift, negative pore pressure node elements named CAX4R and CAX4P in the
is generated at the foundation invert where a gap may be
formed. The fluid flow driven by the suction fills the

1 NUMGE 2023 - Proceedings

 Offshore geotechnics

ABAQUS element library, respectively. The soil behav- The initial void ratio of a soil element is given as
iour was defined by the MCC model, which uses an as-
sociated flow. The MCC parameters are listed in Table 𝑒0 = 𝑒𝑁 − (𝜆 − 𝜅) 𝑙𝑛 𝑝𝑐′ − 𝜅𝑝0′ (4)
1, which are fit from laboratory test on the same type of
kaolin clay used in the centrifuge tests (O'Loughlin, where eN is the void ratio at p´ = 1 kPa on normally
2022). Among them, the permeability was deduced from isotropic consolidation line.
the consolidation coefficient obtained from the Rowe The undrained shear strength of overconsolidated
cell test at the stress level at depth of D/6. The permea- soils can be deduced based on the MCC model
bility was assumed to be isotropic in the numerical sim- parameters as below (Potts and Zdravkovic, 1999):
ulation of lightly overconsolidated clay with OCR = 2,
as the soil horizontal and vertical permeabilities are re- ′ (1+2𝐾0nc )
𝑠𝑢 = 𝜎v0 [OCR𝑔(𝜃) 𝑐𝑜𝑠 𝜃 6
(1 + 𝐴2 )] ×
ported to be close in the reconstituted NC clay samples
2(1+2𝐾oc ) 𝜅/𝜆
in the centrifuge (Vessia et al., 2012). 0
[(1+2𝐾nc )OCR(1+𝐴2 )] (5)
0

Table 1. Parameters of the numerical model

𝑠𝑖𝑛 𝜑′
Parameter Value where 𝑔(𝜃) = 𝑐𝑜𝑠 𝜃+(1/
√3) 𝑠𝑖𝑛 𝜃 𝑠𝑖𝑛 𝜑′
Soil √3(1 − 𝐾0nc )
Angle of internal friction, φ´ 𝐴=
23° 𝑔(−30°)(1 + 2𝐾0nc )
Void ratio at p´=1 kPa on virgin 3.96 with θ being the Lode’s angle (taken as 0 for plain strain
consolidation line, eN strength). To match the su profile in the centrifuge tests,
Slope of normal consolidation line, λ 0.435 the submerged density of soil was reduced from 400
Slope of swelling line, κ 0.063 (measured in the centrifuge tests) to 260 kg/m3. Then,
Poisson ratio, ν´ 0.3 the obtained undrained shear strength profile (in unit of
Submerged density, ρ´: kg/m3 260 kPa) was su = 0.8 + 52.6z, where z is the depth in unit of
Permeability, k: m/s 5×10-8 m.
To improve the numerical stability, a thin layer of soil
Interface underneath the interface was modelled by the elastic
Threshold gap size, δ: m 1×10-3 model with E = 3000 kPa and v´ = 0, as recommended
Leak-off coefficient, c: m/(s·kPa) 1×10-3 in Cerfontaine et al. (2016).
Tangential permeability, kx: m/s 5×10-8 The interface was located between the footing and the
Fluid dynamic viscosity, µ: kPa·s 1×10-6 soil, and its upper and lower boundaries were tied to the
footing and the soil, respectively. The footing, the inter-
The numerical simulations were conducted in terms face and the underneath soil had consistent element sizes
of model surface footing with D = 0.06 m in the centri- in horizontal direction, as shown in the close-up view of
fuge at an acceleration of 40 g in order to make straight- Fig. 1 and Fig. 2. This is helpful for the interface to be
forward comparisons. In the numerical analysis, the functioning well.
earth pressure coefficient in the lightly overconsolidated
soil is

𝐾0𝑜𝑐 = 𝐾0𝑛𝑐 OCR𝑠𝑖𝑛 𝜑′ = (1 − 𝑠𝑖𝑛 𝜑 ′)OCR𝑠𝑖𝑛 𝜑′ (1)

where K0nc is the earth pressure coefficient of normally

consolidated soil, φ´ is the effective internal friction
angle. The OCR is the ratio between the maximum
vertical effective stress experienced by the soil σvm´ and
the initial vertical effective stress σv0´:

′ ′
𝜎vm = 𝜎v0 OCR (2)

The size of the yield envelope is linked to the maximum

mean effective stress pm´ and deviatoric stress qm:

𝑞2 ′
𝑝𝑐′ = 𝛭2𝑚𝑝′ + 𝑝𝑚 (3)
𝑚

6 𝑠𝑖𝑛 𝜑′
with 𝛭 = 3−𝑠𝑖𝑛 𝜑′, 𝑞 = 𝑠𝑖𝑔𝑛(𝜎𝑣´ − 𝜎ℎ́ )√2𝐽2. Figure 1. Finite-element mesh

2 NUMGE 2023 - Proceedings

 Numerical modelling of uplifting a surface foundation on clay

tangential flow as either Poiseuille flow or Darcy flow.

A more appropriate approximation of tangential flow is
supposed to transition from Darcy flow to Poiseuille
flow with an increase in gap opening. When the
foundation initially sits on the soil, the gap is closed (gN
= 0) and thus the tangential flow is deemed obeying
Darcy flow. When the footing moves apart from the soil
enough (e.g., the gap size is larger than a threshold gap
size (gN ≥ δ)), the tangential flow is assumed to obey
Figure 2. Interface model
Poiseuille flow. Before the gap reaches the threshold (0
< gN < δ), the flow is in a transition stage and
2.2 Interface model interpolated between Darcy flow and Poiseuille flow:
An appropriate interface between the foundation
𝑔 2 𝑘 𝛿 𝑔 2 𝑔3
invert and the soil is the key to modelling uplifting. The 𝑓𝑇 = {[1 − ( 𝛿𝑁 ) ] 𝑥
+ ( 𝛿𝑁 ) 𝑁
} 𝛻𝑝 (9)
hydro-mechanic interface model applied in this paper 𝛾𝑤 12𝜇
was a layer of 6-node elements, labelled as CODAX4P
in the ABAQUS element library as illustrated in Fig. 2. The threshold gap size δ was taken as 1 mm (1/60D)
The top and bottom four nodes have degree of freedom in this study, which is shown to be a reasonable
(DoF) of displacement and pore pressure, while the two magnitude for the transition (see the comparison of
middle nodes only have pore pressure DoF. physical and numerical results in section 3).
The fluid flow within an interface element was The fluid within the interface elements obeys mass
simplified as normal flow from the soil and tangential conservation with consideration of normal and
flow along the interface (Fig. 2). The normal flow was tangential flow and the rate of gap opening. The
defined by a ‘leak-off model’ (Gerke and Genuchten, interface papameters are listed in Table 1.
1993) as
2.3 Boundary conditions and loading
𝑓𝑁 = 𝐿𝑐𝛥𝑝𝑤 (6) As shown in Fig.1, the bottom of the soil domain was
fixed both horizontally and vertically, while the right
where fN is the normal flow rate, L is the element size, c boundary of the soil was constrained only horizontally.
is the leak-off coefficient, and ∆pw is the differential The soil surface, except for along the footing, was set as
pressure between the middle and bottom nodes of the a zero (excess) pore pressure boundary. The edge of the
interface element (noting the bottom nodes have the interface element was exposed to the ambient free water,
same pore pressure as the soil surface). and thus it was also defined as a drainage boundary, i.e.,
For tangential flow along the interface, there are two zero (excess) pore pressure. In the numerical modelling,
flow modes that can be considered. If the gap is small the initial stress and pore pressure of the numerical
enough that the gap flow is amid soil particles, the flow model was first self-balanced in a geostatic step. Then,
can be deemed to obey Darcy’s law (termed as Darcy in the subsequent loading step, the reference point of the
flow here). The tangential flow rate is footing was vertical uplifted at varying uplift rates rang-
ing from 1×10-7 to 1×108 mm/s.
𝑘𝑥 𝛿
𝑓𝑇 = 𝛾𝑤
𝛻𝑝 (7)

where kx is the tangential permeability (in dimension of 3 COMPARISON WITH EXPERIMENT

L/T), δ is a threshold gap size for Darcy flow, γw is the RESULTS
unit weight of water, ∇𝑝 is the gradient of pore pressure The constitutive model of the soil used in the numerical
along the element. modelling is a non-viscous model ignoring strain rate ef-
If the gap is large engough that the pure fluid flow is fects. To facilitate comparison between the numerical
along the interface, the flow can be described by the and the experimental results, the experimental results
Navier-Stokes equations. The Poiseuille flow presented below have strain rate effects subtracted (as
(Schlichting and Kestin, 1961) can be deduced by discussed in Mei et al. (2023)).
ignoring the acceleration terms: Fig. 3(a) shows typical uplift curves under varying
3
uplift rates from the numerical simulation, and Fig. 3(b)
𝑔𝑁
𝑓𝑇 = 𝛻𝑝 (8) illustrates their comparison with the experiment results
12𝜇
under three uplift rates, where the horizontal axis is the
dimensionless displacement w/D and the vertical axis is
where µ is the fluid dynamic viscosity. the dimensionless uplift resistance F/Asum, where A is
Previous studies (e.g., Cerfontaine et al., 2015; Tian the foundation plate cross area and sum is the undrained
et al., 2022; Peng et al., 2022) solely treated the

3 NUMGE 2023 - Proceedings

 Offshore geotechnics

shear strength at the surface. The uplift velocity is nor- are achieved, and the uplift capacity is equal to the un-
malized based on V = vD/cop, where cop is the operative drained compression capacity. V < 2.4×10-3 reaches
consolidation coefficient and taken as 0.25 mm2/s (Mei drained conditions where the uplift capacity is near zero.
et al., 2023). The uplift resistance increased with the up- Between these two limits are partially drained condi-
lift displacement until reaching a peak, and then dropped tions.
to zero. Higher uplift rates result in larger peak re-
sistance and gentler post-peak drop, which is consistent
with the experimental observations. The peak re-
sistances on the uplift curves are taken as the uplift ca-
pacity for each case. The uplift capacities in the numer-
ical simulation are consistent with those in the
experiment, although the uplift stiffness in numerical re-
sults is larger. This might be due to elastic deformation
of soil within the initial yield surface which exaggerates
the stiffness of soils. This paper focuses on the uplift ca-
pacity, and the stiffness of the uplift response is not in-
tended to be intensively explored.

Figure 4. Backbone curve of uplifting

4 INTERPRETATION OF NUMERICAL
RESULTS
Numerical modelling provides an effective way to visu-
alize the pore pressure distribution, deformation mecha-
nisms and stress path of the soil during uplift.
Fig.5 illustrates the excess pore pressure distribution
along the footing radius under various uplift rates at the
peak resistance. The generated excess pore pressure was
(a) negative in all cases (except a spike near the edge in
some cases, which is believed to be a numerical issue).
Higher uplift rate results in larger negative pore pressure
generation. For each case, the negative pore pressure
was generally larger within the central area than near the
edge where some oscillation exists. Integration of the
pore pressure with the area underneath the footing is
shown to be equal to the uplift resistance. This means
that the uplift resistance was completely contributed
from negative pore pressure and that the soil particles do
not provide ‘effective’ tension.
Fig. 6 shows the soil displacement vector plot at the
peak resistance representing the soil flow mechanism.
Under drained conditions, the soil is hardly mobilized.
With the increase of uplift velocity, the soil deformation
(b) increases. When the undrained conditions are achieved,
Figure 3. Uplift curves under various uplift rates: (a) numer- a reverse bearing mechanism is fully mobilized.
ical results; (b) comparison with experiment results Fig. 7 illustrates the gap opening size gN along the
footing radius at the peak normalized by the correspond-
As shown in Fig. 4, a backbone curve is established ing upward displacement w under different rate uplifts.
by plotting uplift capacity factors Nup (i.e., the dimen- At the rapidest uplift rate (V = 2.4×1010), the gap hardly
sionless peak resistance) against dimensionless uplift opened except near the interface edge where the drain-
velocities V. The undrained capacity factor from com- age boundary existed. With a decrease in uplift velocity,
pression test conducted in Mei et al. (2023) is plotted as the gap opening increased with larger gap near the edge
a dotted line. When V > 2.4×107, undrained conditions than at the centre. This causes more rapid decrease in the
uplift resistance during the post-peak period (Fig. 3). At

4 NUMGE 2023 - Proceedings

 Numerical modelling of uplifting a surface foundation on clay

a low uplift rate (i.e., V = 0.0024), the gap opening was stress changed with effective mean stress p´ constant
almost equal to the upward displacement. within the initial yield surface. When it started yielding,
the stress moved toward the critical state line (CSL) with
the yield surface shrinking until reaching critical state.
Under partially drained conditions, only some soil el-
ements (e.g., a and b) were yielded, while other soil el-
ements (e.g., c, d and e) was in elastic state within the
initial yield surface.

Figure 5. Excess pore pressure distribution at the interface

(a)

(a) (b)
Fig. 6. Deformation mechanism under partially drained (a)
and undrained (b) conditions

(b)

Fig. 7. Gap opening at the interface

Fig. 8. illustrates the stress path of typical soil ele-

ments labelled as a~e in Fig. 6(a) under drained, par-
tially drained and undrained conditions in p´-q plane.
The soil was initially under K0 state with OCR = 2. (c)
Thus, the initial stress on the K0 state line (K0-SL) was Fig. 8. Stress path of soil elements under drained (a), partially
inside the initial yield surface and on the ‘dry side’ of drained (b) and undrained (c) conditions
the MCC model. Under undrained conditions, the soil

5 NUMGE 2023 - Proceedings

 Offshore geotechnics

Under drained conditions, the stress had negligible Chen, R., Gaudin, C., Cassidy, M.J. 2012. Investigation of the
changes as the soil was hardly mobilized. In Li et al. vertical uplift capacity of deep water mudmats in clay.
(2015) where the foundation and soil were bonded with- Canadian Geotechnical Journal, 49(7), 853-865.
out using a hydro-mechanic interface, even under very Dassault Systèmes. 2016. Abaqus Analysis Users' Manual.
Simulia Corp, Providence, RI, USA.
low uplift rate the soil experienced unloading and pro-
Gerke, H. H., & Van Genuchten, M. T. 1993. A dual ‐
vided unrealistic effective tension force, which resulted
porosity model for simulating the preferential movement
in relatively large uplift resistance. This illustrates the of water and solutes in structured porous media. Water
effectiveness of employing hydro-mechanic interface resources research, 29(2), 305-319.
when modelling uplifting problems. Gourvenec, S. 2018. Shaping the offshore decommissioning
agenda and next-generation design of offshore
infrastructure. Proceedings of the Institution of Civil
5 CONCLUSIONS Engineers-Smart Infrastructure and Construction, 171(2),
54-66.
Numerical modelling of uplifting a surface footing un- Li, X., Gaudin, C., Tian, Y., Cassidy, M.J. 2014. Effect of
der varying uplift rates was achieved by utilizing a hy- perforations on uplift capacity of skirted foundations on
dro-mechanic interface. The numerical model was veri- clay. Canadian Geotechnical Journal, 51(3), 322-331.
fied against the centrifuge test results. Uplift curves and Li, X., Tian, Y., Gaudin, C., Cassidy, M.J. 2015. Comparative
uplift capacities are consistent between the numerical study of the compression and uplift of shallow foundations.
and experimental results. The mobilized capacity ranges Computers and Geotechnics, 69, 38-45.
from undrained capacity under rapid uplift to zero under Mei, S., Tian, Y., Cassidy, M.J., O'Loughlin, C. 2023.
slow uplift. The numerical results were interpreted to Experimental investigation of rate effects on the uplift
behaviour of shallow foundations. To be submitted.
uncover the mechanism of uplift. During uplift, the up-
O'Loughlin, C. 2022. Private communication.
lift resistance is found to derive from negative pore pres- Peng, M., Tian, Y., Gaudin, C., Zhang, L., Sheng, D. 2022.
sure (suction) generated underneath the foundation plate Application of a coupled hydro-mechanical interface
without any effective mechanical tension force. Rapid model in simulating uplifting problems. International
uplift results in mobilization of the reverse undrained Journal for Numerical and Analytical Methods in
bearing mechanism without separation between soil and Geomechanics, 46(17), 3256-3280.
foundations and soil elements experienced yielding with Potts, D.M., Zdravkovic, L. 1999. Finite element analysis in
no volumetric strain. Under an intermediate rate soil was geotechnical engineering – theory. London, UK: Thomas
partly mobilized and the gap at the interface gradually Tel-ford.
opened from the interface side to the central. Under slow Randolph, M.F., Gourvenec, S.M. 2010. Offshore
Geotechnical Engineering. Taylor & Francis, London,
uplift, the soil was hardly mobilized as little suction was
ISBN: 978-0-415-47744-4.
generated with gap nearly fully opened. Roscoe, K.H., Burland, J. 1968. On the generalized stress-
strain behaviour of wet clay. Engineering Plasticity.
Cambridge, UK: Cambridge University Press. p. 535–609.
6 ACKNOWLEDGEMENTS Schlichting, H., Kestin, J. 1961. Boundary layer theory (Vol.
121). New York: McGraw-Hill.
This research was undertaken with support from Aus-
Tian, Y., Ren, J., Zhou, T., Peng, M., Cassidy, M.J. 2022.
tralian Research Council Discovery Projects Coupled hydro-mechanical interfaces to enable uplift
(DP190103315), Future Fellowship (FT200100457). modelling in offshore engineering. Ocean Engineering,
245, 110570.
Vessia, G., Casini, F. & Springman, S. M. (2012). Discussion
7 REFERENCES on ‘Estimating hydraulic conductivity from piezocone
soundings. Géotechnique, 62, No. 10, 955–956.
Bouwmeester, D., Peuchen, J., Van der Wal, T., Sarata, B.,
Willemse, C. A., VanBaars, S., Peelen, R. 2009. Prediction
of breakout force for deep water seafloor objects. In Pro-
ceedings of the Offshore Technology Conference, Houston,
Tex. OTC 19925.
Bye, A., Erbrich, C., Rognlien, B., Tjelta, T.I. 1995.
Geotechnical design of bucket foundations. In Proceedings
of the Annual Offshore Technology Conference, Houston,
Texas, Paper OTC 7793.
Cerfontaine, B., Dieudonné, A. C., Radu, J. P., Collin, F.,
Charlier, R. 2015. 3D zero-thickness coupled interface
finite element: Formulation and application. Computers
and Geotechnics, 69, 124-140.
Cerfontaine, B., Collin, F., Charlier, R. 2016. Numerical
modelling of transient cyclic vertical loading of suction
caissons in sand. Géotechnique, 66(2), 121-136.

6 NUMGE 2023 - Proceedings