Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/239410971
CITATIONS READS
33 243
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Konstantinos Georgiadis on 15 May 2014.
Finite element analyses of geotechnical problems that Les analyses delements finis des problemes geotechniques
involve partially saturated soils are commonly carried qui mettent en cause des sols partiellement satures sont
out using conventional fully saturated soil models. Such effectuees generalement en utilisant des modeles de sol
models, however, cannot successfully reproduce partially conventionnels entierement satures. Cependant, ces mod-
saturated soil behaviour. This paper investigates the influ- eles ne peuvent reproduire avec succes le comportement
ence of partial soil saturation on the behaviour of piles. dun sol partiellement sature. Cet expose etudie linflu-
A total stress constitutive model for partially saturated ence dune saturation partielle sur le comportement de
soils has been developed and implemented into the Im- piles. Nous avons developpe un modele constitutif de
perial College finite element Program (ICFEP). This contrainte totale pour des sols partiellement satures et
model, which is presented here, was used to perform a nous lavons mis en uvre dans le Programme delements
series of single pile analyses. The results of these analyses finis du College Imperial (ICFEP). Ce modele, que nous
show the effect of partial saturation and in particular of presentons ici, a ete utilise pour faire une serie danalyses
fluctuations of the groundwater table on the behaviour of sur des piles seules. Les resultats de ces analyses mon-
single piles. The effects of a rising groundwater table on trent leffet dune saturation partielle et en particulier
pile behaviour are also investigated in the case of a pile des fluctuations du niveau phreatique sur le comporte-
foundation of a high-rise building in London. The paper ment de piles seules. Nous etudions aussi les effets dune
indicates where prediction of pile capacity and displace- lelevation du niveau phreatique sur le comportement des
ment under different load and groundwater table fluctua- piles dans le cas dune fondation de pile dans une tour
tions may be very different when using fully or partially londonienne. Cet expose indique les endroits ou les previ-
saturated constitutive models. sions de la capacite et du deplacement de la pile sous
differentes charges ainsi que les fluctuations du niveau
phreatique peuvent etre tres differentes selon que lon
KEYWORDS: constitutive relations; numerical modelling and utilise des modeles constitutifs totalement satures ou
analysis; partial saturation; piles; suction partiellement satures.
11
Delivered by ICEVirtualLibrary.com to:
IP: 134.148.216.129
On: Wed, 12 Oct 2011 00:09:59
12 GEORGIADIS, POTTS AND ZDRAVKOVIC
2 s3
s
(1 ) 4 4(1 )5
K 1,2 1 1 (4)
2(1 ) (1 )2
J Cam Clay
Sinfonietta Classica
40 Single Hardening YS
Single Hardening PP
30
20
seq
10
p0* F2
0
0 20 40 60 80 100
Mean effective stress: kPa
p F1
Fig. 3. Examples of yield surface and plastic potential functions
Fig. 2. Primary and secondary yield surfaces reproduced from Lagioia et al. (1996)
Specific volume: v
seq2 > seq1
![(0)k]=[(seq )k]
c po 1
po p (11) (seq2)
pc
Typical soil 1
where pc is a characteristic pressure, (0) is the fully behaviour (seq1)
saturated compressibility coefficient, k is the compressibility 1
coefficient along elastic paths and is independent of suction, (0)
and (seq ) is the partially saturated compressibility coeffi- Equation (13)
cient, given by the following empirical equation (Alonso et
seq = 0
al., 1990):
(seq ) (0)[(1 r)eseq r] (12) Fig. 5. Assumed isotropic compression lines
50
Sinfonietta Classica tially increases with suction, reaches a peak at some value
Single Hardening YS of suction that depends on the adopted soilwater character-
Single Hardening PP
40
istic curve, and then reduces to zero at very high values of
suction (Fig. 6). The critical-state line in this case is given
30 by the following equation:
q M( p Sr seq ) (16)
20
d po po d p (20) Cam-clay model. The soil properties are for Lower Cromer
(0) k v till and are taken from the studies of Gens (1982), Gens &
Potts (1982) and Maswoswe (1985). A Poissons ratio, v, of
d so (so patm ) d p (21) 02 was used instead of a value for the shear modulus, G, in
s ks v
order to have a varying value of G with depth. For simpli-
where ks and s are the coefficients of compressibility along city a constant value of the parameter k was selected. None
constant mean total stress elastic and elastoplastic paths of the analyses involved depression of the water table (dry-
respectively, and G is the shear modulus. With appropriate ing), and consequently the secondary yield surface was not
selection of the model parameters (f 0:4, f 0:9 and relevant. A high value of the yield suction, so , was therefore
sair 0 kPa) and by selecting the option of a constant value selected.
for the cohesion increase parameter, k, and equation (11) for A constant value with depth of 15 was assumed for the
the shape of the yield surface in the isotropic-stresssuction OCR for the fully saturated case. In the partially saturated
( p seq ) space, the model becomes equivalent to the Barce- cases OCR has no practical significance. Instead the YSR
lona basic model. These parameters and options were used (yield stress ratio) po = p is introduced. YSR is expressed
for the initial validation of the model implementation. The in terms of total stresses, and varies with depth. The yield
model predictions for a variety of stress paths presented in stress increases with suction, giving high YSR values close
Alonso et al. (1990) were compared with analytical solutions to the ground surface. The variation of YSR with depth
and experimental data and were found to be in excellent depends on the relationship adopted between the current
agreement. The model predictions can be seen in Alonso et yield stress, po , and the equivalent fully saturated yield
al. (1990). Further validation exercises were performed using stress, po . For the partially saturated analyses at which small
a variety of model parameters and all model options (k equal depths of the groundwater table were considered (D 0
to the degree of saturation, Sr , and equation (13) for the 10 m) equation (11) was used, and a value of 12 KPa was
shape of the yield surface in the isotropic stress space). taken for the characteristic pressure, pc . However, use of
this equation in the analyses with D 25 m, where the
partially saturated zone extended to high stress regions,
SINGLE PILE ANALYSES
As mentioned in the introduction, in many cases a capil-
lary zone exists above the groundwater table where the soil
is partially saturated. The presence of such a zone has been Conventional Partially saturated
found to be of significant importance in the case of shallow analyses: analyses:
Profile 1 Profile 2
foundations (Nesnas, 1995; Georgiadis et al., 2002), giving 1m
uw = 0 kPa uw = Dw
high values of bearing capacity, but also the danger of
collapse upon wetting. This section investigates the influence
of the presence of a capillary zone above the groundwater D
table and of fluctuations of the groundwater table on the GWT
behaviour of a single pile in a uniform soil.
20 m
Details of analyses
All analyses involved a wished-in-place single pile in
uniform soil. The pile was 20 m long and 1 m in diameter. pwp
No interface elements were used: therefore the concretesoil
angle of friction was taken to be equal to the angle of Fig. 7. Single pile analysis: pore water pressure profiles (pwp)
10
Depth: m
15
20
10
Results
Figure 11 shows the loadsettlement curves for some
cases in the first set of analyses, which involved drained
Depth: m
loading of the pile. Curves P2, P6 and P10 are for the
analyses with the groundwater table at 2 m, 6 m and 15
10 m respectively and pore pressure profile 2, curves D2,
D6 and D10 are for the groundwater table at 2 m, 6 m
and 10 m respectively and pore pressure profile 1, and
curve S0 is for the fully saturated analysis (groundwater 20
table at 0 m). As can be seen in this figure the bearing
capacity of the pile increases with increase of the depth of
the groundwater table. This increase is due mainly to the
increase of the shaft resistance. Once the shaft resistance is
exceeded, at approximately 20 mm displacement, the load 25
displacement curves for all analyses are close to parallel.
Although the shaft resistance for all analyses is reached at
small displacements (less than 2% of the pile diameter), the
base resistance continues to increase even at high values of 30
vertical displacement. As stated in Potts & Zdravkovic
(2001), it can be assumed that the ultimate load for practical Fig. 10. Predicted yield stress ratio profile for D 25 m from
purposes is reached at vertical displacements of 10% of the equation (13)
8000
7000
6000 S0
D2
Load: kN
5000
P2
4000
D6
3000
P6
2000 D10
P10
1000
0
0 20 40 60 80 100 120 140 160 180 200
Vertical displacement: mm
pile diameter. This definition of ultimate load will be increase of the ultimate load with the depth of the ground-
adopted for this study. water table and, more importantly, the increase of the effect
As seen in Fig. 11 the influence of partial soil saturation of partial soil saturation with the increase of the depth of
is insignificant for small depths of the groundwater table the groundwater table can be seen. The ultimate loads
(curves P2 and D2 are almost identical), and becomes more predicted by the partially saturated analyses depend on the
important at larger depths of the groundwater table. For the choice of the loading-collapse yield surface and the variation
groundwater table at 10 m from the ground surface (curves of the cohesion increase parameter, k, with suction. In the
P10 and D10) the conventional analysis underestimates the extreme case where the loading-collapse yield surface is a
load-carrying capacity by approximately 17%. However, for straight vertical line in p seq stress space ( po po ) and
the factors of safety used in practice (usually FOS > 2) the k 0, the partially saturated and conventional predictions
predictions of the partially saturated and conventional ana- are identical.
lyses are very similar. The results of the analyses with the Figure 13 shows the loaddisplacement curves predicted
groundwater table at 4 m and 8 m have been omitted by the conventional (curve D25) and partially saturated
from this figure for clarity, but gave similar results. (curve P25) analyses for the groundwater table at 25 m
The ultimate loads for different depths of the groundwater below ground level. The partially saturated analysis predicts
table predicted by the conventional (profile 1) and partially approximately 75% higher ultimate load than the conven-
saturated (profile 2) analyses are shown in Fig. 12. The tional analysis. Although these analyses appear to represent
8000
Conventional
7000 analyses
Partial saturated
6000 analyses
Ultimate load: kN
5000
4000
3000
2000
1000
0
0 1 2 3 4 5 6 7 8 9 10
14 000
12 000
10 000
Load: kN
8000
6000
Partially saturated
4000 analyses
conventional
analyses
2000
0
0 20 40 60 80 100 120 140
Displacement: mm
an extreme case, such as situation may occur when dewater- changes in the groundwater table depth. These analyses
ing or underdrainage takes place, causing suppression of the involved loading the pile to different loads for a range of
groundwater level to great depths. In these cases it may be groundwater table depths and then raising the water table to
important to understand the likely overprediction of capacity ground level. None of the analyses involving small depths of
that a pile load test may show, compared with the capacity the groundwater table (D 0 10 m) predicted additional
when the groundwater table returns to a smaller depth. In settlement of the pile due to wetting. In fact, in all cases
this particular case an overprediction of approximately 320% upward movement of the pile was observed. The response of
would be made if the groundwater table was to return to the soil to wetting was mostly elastic. The reduction of the
ground level. Depending on the pile load at which the rise effective stresses in the fully saturated region caused only
of the groundwater table takes place large settlements may elastic swelling, whereas in the partially saturated region the
also occur. wetting paths at constant total stress intersected the loading-
Conventional fully saturated analyses accounting for suc- collapse yield surface only in a small region close to the
tions by using pore pressure profile 2 predict higher ultimate soilpile interface. A typical pile response to wetting is
loads than do partially saturated analyses (with the same shown in Fig. 14 for a partially saturated analysis with an
pore pressure profile). This is clearly not conservative, and initial D 10 m and a constant load of 4 MN (approxi-
is due to the fact that, while for fully saturated soils strength mately 50% of ultimate load).
increases linearly with suction, this increase is non-linear for A different response was predicted, however, for the case
partially saturated soils. where the groundwater table was initially at 25 m. Raising
The second set of analyses was performed in order to the water table at a constant load of 10 MN (approximately
investigate the potential for pile movements associated with two thirds of the ultimate load) caused the pile to collapse,
16
14
12
Vertical displacement: mm
10
0
0 1 2 3 4 5 6 7
Rise of water table: m
Fig. 14. Progression of vertical movement of pile during rise of groundwater table (D 10 m, L 4 MN)
0
Vertical displacement: mm
50
Partially saturated
100 analyses
Conventional
analyses
150
200
250
Fig. 15. Progression of vertical movement of pile during rise of groundwater table (D 25 m, d
14 mm)
20
study is for illustrative purposes it was thought reasonable to
proceed with these values. The compressibility coefficient
along constant suction elastic paths, k, was calculated from
25 Cessation of
the assumed Poissons ratio, , and Youngs modulus, E, as
dewatering
0005. The fully saturated compressibility coefficient, (0),
was assumed to be 12 times larger than k, equal to 006.
30
Pile construction
Initial hardening parameters
35 The specific volume at unit confining stress, v1 , was
assumed equal to 1826 for Lambeth sand and 1872 for
Thanet sand. An estimate of the OCR value was obtained
40 from the relationships among OCR, K0 and PI (plasticity
20 15 10 5 0 5 10 15 20 25 index) presented by Brooker & Ireland (1965). For a very
Piezometric level (m) low plasticity soil with a K0 value of 115 a value of 6 was
selected for the OCR. This value refers to the original fully
Fig. 18. Assumed pore water pressure profiles
saturated sate.
10
05
0
0 05 10 15 20 25 30 35 40
Net mean stress normalised by critical state stress
Table 3. Soil properties for Lambeth Sand the first stage of each analysis to simulate the excavation of
f 008 002 kPa 1 the cofferdam. The pile analysed was 15 m in diameter and
f 20 ks 0001 205 m long, and was wished in place. The finite element
Mf 09 v1 1826 mesh used was similar to that used in the previous study
g 001
02 (Fig. 8). Two types of analysis were performed: the first type
g 057 sair 150 kPa involved vertical loading of the pile to failure, and the
Mg 132 003 kPa1 second type loading of the pile to a certain load and then
c 1667 m 035 raising of the water table in the lower aquifer. Both conven-
(0) 006 n 45 tional and partially saturated analyses were carried out, using
k 0005 Sro 015 the best estimate for the model parameters from the avail-
r 025
able data. However, partially saturated analyses were also
performed using different values for some parameters in
order to investigate their influence on the pile behaviour.
The parameters that were varied were the overconsolidation
aquifers at 55 m OD. Before the construction of the pile ratio, OCR, and the parameters r and , which control the
the piezometric level in the lower aquifer was reduced to expansion of the primary yield surface with suction. The set
280 m OD and the top 10 m of fill was excavated. The fill of parameters given in Tables 3 and 4 will be referred to as
layer was simulated as a uniform surcharge acting through- the reference set in this paper. Note that, because of the
out the top of the finite element mesh and was removed at relatively high stresses in the partially saturated region
100
90
Thanet sand
80
Lambeth sand
70
Degree of saturation: %
60
50
40
30
20
10
0
1 10 102 103 104 105
Suction: kPa
Fig. 20. Soilwater characteristic curves for Thanet sands and Lambeth sands
45
40
35
30
Load: MN
25
20
15
Measured
10 Partially saturated
(OCR = 6)
Conventional
5 (OCR = 6)
0
0 20 40 60 80 100
Settlement: mm
10
20
Vertical displacement: mm
30
40
Partially saturated
50
(L = 19 MN)
Partially saturated
(L = 26 MN)
60 Partially saturated
(L = 32 MN)
Conventional
70 (L = 261 MN)
80
45
40
35
30
Load: MN
25
20
15
Partially saturated
10 (OCR = 6)
Partially saturated
(OCR = 2)
5 Partially saturated
(OCR = 15)
0
0 20 40 60 80 100
Settlement: mm
(reference set of parameters), 2 and 15. It can be seen that, ences large additional settlements due to wetting only at
by reducing the value of OCR, much lower ultimate loads high loads, close to or after the shaft resistance has been
are predicted. Fig. 24 shows the pile response to the rise of exceeded.
the groundwater table at a constant load of approximately Loading and wetting partially saturated analyses were also
19 MN for OCR 1:5 and OCR 6. Whereas for OCR performed using different values for the parameters r and ,
6 the pile moved upwards, as mentioned before, for the which control the compressibility coefficient (seq ) and con-
lower OCR case very large settlements were predicted, more sequently the isotropic yield stress, po , through equation
than 10 times larger than those induced by the pile loading. (13). Fig. 25 shows the loaddisplacement curve predicted
Also shown in this figure are the responses of the pile to for r 0:35 and 0:01, compared with the curve pre-
wetting for OCR 1:5 and for loads of 135 MN and dicted with the reference parameter set. This choice of r and
202 MN. As can be seen, the response to wetting is highly gives a smaller increase of yield stress with suction than
dependent on the load at which it takes place. Similar results the reference values (Fig. 26) and, as expected, predicts a
were obtained from the OCR 2 analyses. The pile experi- lower ultimate load of the pile. It is interesting to see that a
50
100
Vertical displacement: mm
150
200
Partially saturated
250 (OCR = 6, L = 19 MN)
Partially saturated
(OCR = 15, L = 135 MN)
300 Partially saturated
(OCR = 15, L = 19 MN)
Partially saturated
350 (OCR = 15, L = 202 MN)
400
45
40
35
30
Load: MN
25
20
10 Partially saturated
(modified parameters)
0
0 20 40 60 80 100
Settlement: mm
small change of these two parameters (from 025 to 035 for 600
r and from 002 to 001 for ) affects the predictions so
significantly (the reference set gives an ultimate value 500
approximately 16% higher).
400
Suction: kPa
300
CONCLUSIONS
A generalised constitutive model for partially and fully 200 Reference
saturated soils has been presented in this paper. The model parameters
100 Modified
incorporates a versatile equation for the yield and plastic Air entry suction parameters
potential surfaces, which allows the behaviour of a wide
0
variety of soils to be modelled. Basic features of partially 0 5000 10000 15000 20000 25000
saturated soil behaviour such as collapse or swelling upon Mean total stress: kPa
wetting can be reproduced. The model also predicts realistic
values of yield stress at high stresses and a realistic variation Fig. 26. Primary yield surfaces in isotropic stress space for
of shear strength with suction. different partially saturated parameters