Computers and Geotechnics 91 (2017) 117130

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Numerical modelling of the effect of curing time on the creep behaviour

of a chemically stabilised soft soil
Paulo J. Venda Oliveira a,, Antnio A.S. Correia b, Lus J.L. Lemos c
a
ISISE, Department of Civil Engineering, University of Coimbra, Rua Lus Reis Santos, 3030-788 Coimbra, Portugal
b
CIEPQPF, Department of Civil Engineering, University of Coimbra, Rua Lus Reis Santos, 3030-788 Coimbra, Portugal
c
Department of Civil Engineering, University of Coimbra, Rua Lus Reis Santos, 3030-788 Coimbra, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: This work studies the effect of the curing time on the creep behaviour of a stabilised soft soil, using vol-
Received 16 March 2017 umetric and deviatoric creep laws associated with constitutive models. Results of unconfined compres-
Received in revised form 19 June 2017 sive strength tests for several curing times are used to define the time evolution of the mechanical and
Accepted 11 July 2017
creep properties. The models/creep laws are validated by oedometer and triaxial creep tests, for 28
and 90 days of curing. The long-term behaviour of an embankment built on a soft soil reinforced with
deep mixing columns shows that the effect of curing time decreases the settlement and increases the
Keywords:
improvement factor.
Creep
Curing time
2017 Elsevier Ltd. All rights reserved.
Stabilised soft soil
Deep mixing columns
Embankment on soft clay
Numerical modelling

1. Introduction
The reinforcement of soft soils with deep mixing columns
(DMCs) is widely used in several countries to allow the construc-
tion of embankments on such soils, mitigating their poor geotech-
nical properties (low strength, low permeability and high
deformability) [13]. Numerical analyses of the behaviour of
embankments (and shallow foundations) built on soft soil rein-
forced with DMCs have been performed in several works, most of
them with commercial software (Plaxis, Flac, Ansys, etc) and mod-
elling the behaviour of the DMCs with elastic-perfectly plastic laws
using MohrCoulomb failure criterion [413]. Recently, structured
models have been used to improve the numerical predictions of
the behaviour of stabilised soils [1418]. Moreover, the creep anal-
ysis of the behaviour of DMCs was performed with the Modified
Cam Clay model associated with two creep laws (volumetric and
deviatoric) [19].
Some works have shown that the inclusion of DMCs (with a
much higher stiffness) in the soft soil decreases the consolidation
time [4,20] and promotes the transfer of stress from the soil to
Corresponding author.
E-mail addresses: pjvo@dec.uc.pt (Paulo J. Venda Oliveira), aalberto@dec.uc.pt
(A.A.S. Correia), llemos@dec.uc.pt (L.J.L. Lemos).
the DMCs, due to the arching effect, which has a positive impact
on the reduction of the settlement of the embankment [4].
When the soft soils have a high organic matter (OM) content,
significant deformations occur in the long term (long after the
end of the primary consolidation process), due to the creep phe-
nomenon [21,22]. Recent research studies have shown that the
detrimental effect of the OM content, in terms of creep deforma-
tions, is still observed in soils chemically stabilised with binders
[23]; in fact, creep deformations also depend on the binder quan-
tity and composition, stress level and curing conditions [23,24].
Contrary to what is usually admitted by the geotechnical com-
munity, non-negligible creep strains can occur in embankments on
soft soils reinforced with DMCs [19], fundamentally induced by the
creep strain rate of the stabilised material and potentiated by the
arching effect which increases the stresses on the DMCs and conse-
quently the creep strains [19,24]. In fact, field data concerning
some of the embankments studied have confirmed the existence
of creep strains, since they show a constant settlement rate over
time [2527], which may be partially due to creep phenomena,
since a significant reduction of the consolidation time is obtained
when DMCs are used [4,20]. In another case, much higher settle-
ment (about 100%) than that predicted numerically was observed,
without any indication of the embankments instability [28], which
may be explained by the creep of the DMCs, potentiated by the rel-
atively low level of mechanical improvement of the DMCs.

http://dx.doi.org/10.1016/j.compgeo.2017.07.007
0266-352X/ 2017 Elsevier Ltd. All rights reserved.
118 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130

Nomenclature

D shear stress level (q/qult) p0 average (volumetric) effective stress (Pa)

e_ ax axial strain rate (%/min) PI plasticity index (%)
a Singh-Mitchells (creep deviatoric) parameter qu maximum unconfined compressive strength (Pa)
A Singh-Mitchells (creep deviatoric) parameter SCR stress concentration ratio
c0 cohesion (Mohr-Coulomb parameter) (Pa) tdi deviatoric reference time
CIU isotropically consolidated undrained triaxial test tvi volumetric reference time
CSL critical state line UCS unconfined compressive strength (Pa)
cu undrained shear strength (Pa) VCL virgin consolidation line
Ca e secondary consolidation index (De/Dlogt) wL Atterberg liquid limit (%)
DMC deep mixing column wP Atterberg plastic limit (%)
E0 Youngs modulus (Pa) w0 initial moisture content (%)
eo initial void ratio WT water table
Eu Secant effective Youngs modulus for 50% of peak eax axial strain (%)
strength (Pa) eax-creep creep axial strain (%)
eko void ratio for p0 = 1 u0 angle of shear stress (Mohr-Coulomb parameter) ()
FEM finite element method c unit weigh (kN/m3)
IF improvement factor j swell-recompression index (ln scale)
K0 coefficient of earth pressure at rest k virgin compression index (ln scale)
kh horizontal permeability coefficient (m/s) m0 Poisson ratio
kv vertical permeability coefficient (m/s) r0 creep vertical effective stress applied during the creep phase
M slope of critical state line on qp0 plane (Pa)
m Singh-Mitchell (creep deviatoric) parameter r0 v-DMC vertical effective stress on DMC (Pa)
OCR overconsolidation ratio r0 v-soil vertical effective stress on soil (Pa)
OM organic matter (%) r0 y vertical effective yield stress on DMC (Pa)

Moreover, a numerical study concerning the creep on the DMCs
shows that, with the consideration of creep phenomena, the DMCs
are less efficient, that is, the settlement over time increases, which
induces a decrease in the improvement factor, as well as augment-
ing the stresses on the DMCs [19].
Although several studies have shown the enhancement of the
mechanical properties (mainly strength and stiffness) of chemi-
cally stabilised soils over time, due to pozzolanic reactions
[1,23,29], the current design of structures with stabilised soils (as
usual in concrete engineering) makes use of the mechanical prop-
erties evaluated for 28 days of curing [4,19]. Considering the lack of
numerical studies concerning the enhancement of the mechanical
properties over longer periods of time, it is very pertinent to study
this effect in a long-term analysis, mainly when associated with
the consolidation and creep phenomena.
2. Objectives of the work
The novelty of this work is to joint together the effects of the
creep phenomenon with the curing time (tc) in order to improve
the long-term prediction of the behaviour of geotechnical struc-
tures. While the creep phenomenon increases the settlement over
time and reduces the efficacy of DMCs [19], the increase in
mechanical properties with curing time tends to induce the oppo-
site effect, attenuating the impact of creep.
Firstly, results of eodometer and triaxial creep tests for differ-
ent curing times [1] are used to validate the performance of the
constitutive models associated with the two creep laws used
(composed by deviatoric and a volumetric components) and tak-
ing the effect of the curing time into consideration. Finally, the
effect of the curing time on the long-term behaviour of an infi-
elastoplastic analyses with coupled consolidation and creep, is
used in these analyses. The constitutive models chosen for the sta-
bilised material is the Modified Cam Clay (MCC) combined with the
Von Mises (VM) model. The constitutive model is associated with
two creep laws, both of which depend on the curing time, thus
inducing different mechanical and creep characteristics of the sta-
bilised soil over time.
3. Description of the models/laws
3.1. Creep laws
The creep strains of a soil can be broken down into a volumetric
and a deviatoric component [3032]. Based on the secondary con-
solidation theory stated for oedometer tests [33], creep volumetric


strains ecv are calculated by Eq. (1):
Z
Cae
ecv dt 1
2:31 etv
where e is the void ratio, Cae (=De/Dlogt) the index of secondary
consolidation and tv the volumetric age in relation to a reference
time (tvi). Cae can be evaluated from oedometer creep tests, since
the general volumetric model in e-log r0  log t space states that
the deformation at constant deviatoric stress levels is represented
by a series of parallel planes, consequently, Cae is constant and
independent of the shear stress level [30]. For unstabilised soils, tv
is usually evaluated by Eq. (2), which is based on the difference
between the void ratio of the current state (e1) and the one on
the boundary surface state (e2) [3234]:
 
e2 e1

nite embankment built on a soft soil reinforced with DMCs is tv tvi :e Cae =2:3
2
analysed, considering elasto-plastic constitutive models and
Singh-Mitchells law [35] is used to calculate the creep devia-
creep laws.  
A 2-D finite element program with several constitutive models, toric strains eax-creep , which are based on the evolution of the
upgraded at the University of Coimbra and capable of carrying out axial strain from drained or undrained creep triaxial tests:
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 119

Z  m
tdi shear strains and the expansion of the yield surface of the VM
eax-creep AeaD dt 3
model. Without the VM model, any stress path inside the yield sur-
td
face of the MCC model merely originates elastic strains. Thus, the
where A, a and m are material creep parameters, D (=q/qult) is the use of the VM model is intended to replicate the non-linear beha-
deviatoric stress level and td the deviatoric age stated in relation viour within the yield surface of the MCC model [3941]. When
to a reference time (tdi). Parameter A reproduces the effect of the both yield surfaces are activated, the vector of global plastic flow
structure and stress history [36], parameter a states the influence is the sum of the plastic flow vectors in each model.
of the shear stress level [37], while parameter m reflects the Both models (MCC and VM) assume a linear elastic behaviour
variation of creep strain rate with time [35]. Based on Borja and inside the yield surface and consider an associated plastic flow
Kavazanjian [32] and Singh and Mitchell [35], Eq. (3) is valid for rule, i.e., the vector of plastic strains is perpendicular to the yield
0.20.3 < D < 0.9. For unstabilised soils, td is evaluated by Eqs. (4a) surface.
and (4b) based on the actual deviatoric strain (c1) and that on the In brief, the MCC model is an elastoplastic soil model, with a
normally consolidated state (c2) [3234]: yield function described by Eq. (5) [42]:
" #1m   
c1  c2  1  m
1
g2
Fr0ij ; ek ; tv k  j  ln p0 1 2  hek ; tv ; tc 5
td m 1 4a M
A  eaD  tdi m
where k and j are, respectively, the slope of the virgin consolidation
 
c1 c2 line (VCL) and the slope of the overconsolidation line in the plot e-
td tdi  e AeaD m 1 4b ln p0 , M is the slope of the critical state line (CSL) and g is the ratio
q/p0 . The yield function is represented by an ellipse-shaped surface
In order to improve the numerical predictions for creep, tv and
oriented in line with the p0 axis, and its size changes with the isotro-
td coincide with the current time [19] for stabilised soils.
pic preconsolidation pressure, p0 c (Fig. 1a), which depends on the
A detailed description of the creep laws associated with the
void ratio (ek), tv and the curing time (tc). The hardening rule, h
MCC and MCC/VM models can be seen in some studies
(ek, tv, tc), depends on the plastic deformations, h(ej), the volumet-
[31,32,34,38].
ric creep law, Dh(tv), and the curing time, Dh(tc) (Fig. 1b), and are
described by:
3.2. Constitutive model associated with the creep laws and curing time  
Cae tv p0ct
hek ; tv ; tc eko  e j:lnp k  j  ln 0 c
0
 ln
The behaviour of the stabilised soil is simulated by a constitu- |{z} 2:3 tvi pct 28d
ej |{z
} |{z}
c
tive model with a double-yield surface, composed by the yield sur- |{z
} Dhtv Dht c
faces of the MCC and VM models (Fig. 1a). The two yield surfaces hej
(MCC/VM) can be activated independently or simultaneously. 6
When the stress path reaches the yield surface of the VM model
and is within the yield surface of the MCC model, the yielding where eko is the void ratio for p0 = 1 (Fig. 1b). In fact, the creep phe-
(plastification) of the VM model occurs, which originates plastic nomenon and the increase in the curing time promote the change of
h(ek, tv, tc), which induces the increases of the apparent p0 c (Fig. 1b).
In practical terms, this effect corresponds to the increase of the eko
through the term Dh(tc). The deduction of Dh(tc) can be found in
p
q =M Appendix A.
Yield surface of L,q The VM model is described by Eq. (7) [34]:
the VM model C S
 
Ellipsoid of the G r0ij ; cp ; epv ; tv ; td q  qc 7
MCC model |{z}
hcp ;epv ;tv ;td ;tc

where hcp ; epv ; tv ; td ; tc is the hardening rule, represented by the

p' cA p' cB parameter qc, which depends on the plastic deformations, the volu-
(a) p' metric and deviatoric creep time and the curing time. The harden-
pc(tv, t c) ing rule is evaluated considering that the trace of the yield
surface on the qc plane is a hyperbola [38,43], represented by:
e
cp0c
e qc Rf 8
a bc
h (e )
A where a and b are normalized hyperbolic parameters, and
Rf (=qfailure/qult) is the failure ratio [44].
ekA
h(tv, t c) 4. Parameters of the materials
eA
ekB
This work is performed with a Portuguese soft soil from Baixo
eB VCLisotropic
Mondego area, either in a reconstituted state or chemically
stabilised with binders.

CSL 4.1. Soft soil

p' = 1 p'
(b)
The soft soil used in this work comes from the Baixo Mondego
Fig. 1. Effect of creep and curing time on the MCC model: (a) yield surface (p0 q area (in Portugal) and was formed on the alluvial plane of the river
plane); (b) hardening rule (ep0 plane). Mondego during the last 20,000 years, presenting a variable
120 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130

thickness over a limestone bedrock. This soil is lightly overconsol-
idated (OCR < 4 5) near the surface (depth of 3.56.5 m), due to
desiccation and fluctuations in the water table due to agricultural
needs [2,3].
The natural soft soil used in this work is a silty soil (sand = 17
21%, silt = 71%, clay = 812%), classified as OH [45]. It has a high
organic matter content (9.3%), which is the main cause of the
low unit weight (14.6 kN/m3), high initial moisture content
(w = 80.9%), high void ratio (e0 = 2.1), high plasticity (wL = 71%,
wP = 42.8%, PI = 28.2%), low undrained shear strength (cu < 25
kPa) and high primary and secondary compressibility (k = 0.204,
j = 0.030, Cae = 0.0236) [2,3]. A detailed discussion of the proper-
ties of the soil can be seen in Correia [1], Coelho [2] and Venda Oli-
veira et al. [3].
Reconstituted samples were used in the experiments to miti-
gate the inherent/natural variability of these types of soils. To pre-
pare these samples from the natural soil, firstly, the shell residues
were removed with a 2.36 mm mesh sieve and, after that, the
moisture content of the soil was adjusted to obtain a 1.5 liquidity
index, in order to ensure the destruction of the structure of the nat-
ural soil [2].
Based on previous studies [13], the parameters of the soft soil
used in the numerical predictions to characterize the MCC model
and the two creep laws (volumetric and deviatoric) are presented
in Table 1.
4.2. Stabilised soft soil
The chemical stabilisation of the soft soil was performed with a
binder mixture composed by Portland cement Type I 42.5 R [46]
and blast furnace granulated slag (dry weight proportion of
75/25) with a binder quantity of 125 kg/m3 [1]. Due to their chem-
ical compositions, the Portland cement (CaO = 63.0%, SiO2 = 19.7%,
Al2O3 = 5.2%) reacts spontaneously with water (hydraulic material)
while the slag (CaO = 37.0%, SiO2 = 38.7%, Al2O3 = 11.6%) promotes
the long-term strengthening in consequence of the pozzolanic
reactions [47]. The results of the unconfined compressive strength
(UCS), oedometer and triaxial tests carried out with samples of sta-
bilised soil (Table 1) show that the stabilisation process has a great
impact on the improvement of their material properties. Thus,
even for a curing time of 28 days (Table 2), the stabilised soil pre-
sents a very high unconfined compressive strength (qu  1143 kPa)
and vertical yield stress (r0 y  2300 kPa), a low pre-yield
compression index (Cr = 0.032) and secondary compression index
(Cae  0.00102) [1].
As the creep properties of the stabilised soil tend to increase
over time, due to pozzolanic reactions [1,23], oedometer creep
tests and drained triaxial creep tests were carried out with differ-
ent values of D to characterize the creep behaviour of the stabilised
soil, for 28 and 90 days of curing time. The results of the oedometer
creep tests (Table 2) show that Cae decreases with the increase of
the curing time (tc) from 28 to 90 days [1]; furthermore, Cae
increases with the vertical stress applied during the creep phase
(r0 creep), as described by Eq. (9) [24].
Cae 1:67  106  r0creep 322:7  106 9
Fig. 2 depicts the evaluation of Singh-Mitchells creep parame-
ters from the drained triaxial creep tests. The plot e_ ax  D
(Fig. 2a) shows that the parameter a (0.671) can be considered
as independent of tc, while parameter A decreases significantly
(from 2.4  105%/min to 1.9  105%/min) with the increase of tc
from 28 and 90 days. The parameter m (0.280) is evaluated from
the plot e_ ax  t (Fig. 2b) and it is assumed to be independent of tc.
Table 1 shows the parameters of the MCC/VM model and for the
two creep laws (volumetric and deviatoric) used to predict the
behaviour of the stabilised soil with curing time.
4.3. Effect of curing time on stabilised soil
Several research studies have stated the enhancement of the
mechanical characteristics of a stabilised soil with curing time
[1,4854], mainly due to pozzolanic reactions. An increase in the
cementation bonds between the soil particles is expected to occur
over time, which promotes the hardness of the solid skeleton,
increasing the strength and decreasing the compressibility [23].

Table 1
Parameters of the soft soil and the stabilised soil (i.e. DMCs) used in the numerical analyses [1,19].

Soil type Soft soil Stabilised soil/DMC

Constitutive model MCC MCC/VM
Curing time tc (days) 28 90 360
Elastic parameters E0 (MPa) *
164.7* 189.0# 225.2#
v 0.3 0.3
MCC model ek0 2.315 5.070 5.204# 5.349#
* * * *
e0
k 0.204 0.435 0.435 0.435
j 0.030 0.0074 0.0074 0.0074
M 1.50 1.50 1.50 1.50
VM model a 0.0013 0.0013 0.0013
b 1.683 1.683 1.683
Rf 1.0 1.0 1.0
Volumetric creep law C ae 0.0236 0.00102* 0.00075*# 0.00053*#
** ** **
tvi (min) 1.0 min
#
Deviatoric creep law A (%/min) 0.001148 0.0000244 0.000018 0.000013#
m 0.978 0.28 0.28 0.28
a 0.734 0.671 0.671 0.671
** ** **
tdi (min) 1.0
a 0.007 0.0013 0.0013 0.0013
b 1.1033 1.683 1.683 1.683
Rf 0.95 1.0 1.0 1.0
*
Variable, depends on the stress level.
**
The creep time is equal to the current time.
#
Evaluated from the Eqs. (11)(14).
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 121

Table 2
Experimental results for the stabilised soil. Effect of curing time [1].

Test Property Curing time (days)

28 60 90 180 360
UCS test qu (kPa) 1143 1433 1586 2094 1948
Eu-50% (MPa) 274 292 324 356 370
Oedometer classical test r0 y (kPa) 2300 2900 3000
Oedometer creep test C ae 0.00102 0.00076
Triaxial creep test A (%/min) 2.4  105 1.9  105

Fig. 3 plots the results of the UCS tests in terms of the evolution
0.00010
Stabilised soil of the unconfined compressive strength ratio (qu/qu-28d) against
curing time (from 28 to 360 days). As observed in several studies
= 0.671 [50,52,53] the value of qu grows faster for shorter curing times,
Strain rate (%/min)

tending to reach a constant value for higher values of tc. Thus, in

the present study, the increase of qu when the curing time varies
between 28 and 360 days is replicated by the power function
tc=28 days, t=1 min
(10), having as a reference the unconfined compressive strength
for the reference curing time of 28 days (qu-28d).
A= 0.0000244 %/min tc=28 days t=1000 min
(tc=28 days)
tc=90 days, t=1 min qutc qu28d  0:4568  tc0:2438  tc 6 360 days 10
A= 0.000019 %/min
(tc=90 days) tc=90 days, t= 1000 min
For curing times higher than 360 days, it is assumed that there
0.00001
is no increase of the cementation bonds, that is, qu is constant and
0.00 0.20 0.40 0.60 0.80 1.00
(a) equal to 2193 kPa.
Deviator stress level, D
As Eq. (10) reflects the effect of increase in the cementation
bonds on the mechanical behaviour, the same function can be used
tc=28 days, D= 0.27
to predict the evolution of the r0 y over tc with good agreement
0.00010
Stabilised soil tc=28 days, D= 0.45
(Fig. 4a):

r0ytc r0y28d  0:4568  t0:2438

tc=28 days, D= 0.62
c  tc 6 360 days 11
tc=28 days, D= 0.73
0
tc=90 days, D=0.40 Eq. (11) is also used to evaluate the increase of p c over time.
Strain rate (%/min)

tc=90 days, D=0.60 The prediction of the secant undrained Youngs modulus for 50%
of peak strength (Eu50) over tc, (Fig. 4b) is also based on Eq. (10), but
in order to match the experimental results it is necessary to use a
reduction factor (Fred = 0.4), taking the form:
m= 0.280  
Eu50tc Eu5028d  Eu5028d  0:4568  t0:2438
c  Eu5028d
1
 |{z}
0:4 tc 6 360 days 12
0.00001
Fred
1 10 100 1000 10000 100000
(b) Time (min) Naturally, the increase in the cementation bonds over time
also has a positive impact on the creep compressibility of the sta-
Fig. 2. Evaluation of Singh-Mitchells creep parameters for the chemically stabilised
bilised material, decreasing the creep deformations. Thus, the
soil from the drained triaxial creep tests for a curing time of 28 and 90 days [1]: (a)
parameters A and a; (b) parameter m. functions used to predict the parameters Cae (Fig. 4c) and A
(Fig. 4d) are based on Eq. (10), but using the inverse of the func-
tion (10), that is:

1
2.0 Caetc Cae28d   t0:2438
c tc 6 360days 13
0:4568
1.8

y = 0.4568x0.2438 1
1.6
R = 0.9753
Atc A28d   t0:2438
c tc 6 360 days 14
0:4568
1.4
qu/qu-28d

In terms of numerical predictions, Eqs. (11)(14) are included

1.2 UCS tests
in the finite element method (FEM) code in order to take into
Power (UCS tests) account the effect of the curing time on the stiffness (E0 ), the
1.0
size of the yield surface of the MCC model (p0 c) and the creep
0.8 strains (Cae and A); E0 is estimated using the same relationship
for Eu50 (i.e. Eq. (12)), ek0 is evaluated from Eqs. (A.7) and (11),
0.6
0 100 200 300 400 and Cae and A are calculated from the power functions (13)
Curing time (days) and (14), respectively. Table 1 shows the estimated properties
of the stabilised soil for a curing time of 28, 90 and 360 (or
Fig. 3. Effect of curing time on the unconfined compressive strength [1]. longer) days.
122 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130

5000 400
'y (tc> 360 d) = 4410 kPa u (tc> 360 d) = 378 MPa

4000 350

tc = 360 days tc = 360 days

'y (kPa)

Eu (MPa)
3000 300
UCS test
Oedometer test
2000 250 Prediction - Power function
Prediction - Power function

1000 200
0 100 200 300 400 500 600 0 100 200 300 400 500 600
Curing time (days) Curing time (days)
(a) (b)

0.0012 tc = 28 days Triaxial creep tests

Oedometer creep tests 0.000030 tc = 28 days
Prediction - Power function Prediction - Power function

0.0010 0.000025

0.0008 0.000020
A
e

tc = 360 days
C

tc = 360 days

0.0006 0.000015

C (tc>360d) = 0.00053 A(tc>360d) = 0.000013%/min

0.0004 0.000010
0 100 200 300 400 500 600 0 100 200 300 400 500 600
(c) Curing time (days) (d) Curing time (days)

Fig. 4. Effect of curing time: (a) yield strength, r0 y; (b) secant Youngs modulus for 50% of peak strength, Eu; (c) secondary compression index, Ca; (d) parameter A of Singh-
Mitchells law.

5. Numerical prediction of the behaviour observed in the
laboratory tests
Three types of laboratory tests were used: triaxial CIU (isotropic
consolidation followed by an undrained shear phase) tests,
oedometer creep tests and multi-stage triaxial tests. The shear
phase of the multi-stage triaxial tests is composed by three stages:
(i) an undrained initial stage; (ii) a consolidation and creep stage
performed for a specific shear stress level; (iii) an undrained stage
until the ultimate conditions are reached.
The numerical predictions were made according to the follow-
ing conditions: (i) axisymmetric finite element (FE) analyses were
carried out to simulate the conditions imposed by the tests; (ii) the
oedometer creep tests were simulated by a mesh with 13 FE ele-
ments and 68 nodes; (iii) the triaxial CIU tests and multi-stage tri-
axial tests were simulated with only 1 FE element and 8 nodes; (iv)
eight-noded isoparametric quadrilateral elements were used in the
analyses; these elements have 20 nodal degrees of freedom, allow-
ing the evaluation of the displacement at eight nodes and the
excess pore pressure at the four corner nodes.
Table 1 shows the parameters for the MCC/VM model and the
creep laws (volumetric and deviatoric) used to simulate the beha-
viour of the stabilised soil numerically both without and under
creep conditions.
5.1. CIU triaxial test
Fig. 5 compares the numerical predictions with the experimen-
tal results of the CIU triaxial tests carried out with samples of the
stabilised soil, isotropically consolidated with a confining pressure
(p0 0) of 50 kPa, for a curing time of 28 and 90 days. As expected, the
results show that the increase in the curing time promotes an
increase in the cementation bonds due to pozzolanic reactions
and, consequently, the strength of the material increases over
time; the numeric model considered is able to replicate this beha-
viour. In fact, the stress-strain behaviour (Fig. 5a) obtained exper-
imentally matches the numerical results very well until the peak
strength; however, after that, the MCC/VM model is not able to
simulate the loss of strength after the peak strength, which is
due to the degradation of the structure of the stabilised soil. In fact,
this is an important weakness of the MCC/VM model.
In terms of stress paths (Fig. 5b), the final state is predicted by
the numerical analyses for both curing times reasonably well,
although a poor prediction is obtained for the intermediate stres-
ses, since more flattened stress paths are observed in the experi-
ments. These differences are mainly due to the characteristics of
the model used, since it assumes a linear elastic behaviour inside
the yield surface of the MCC model, which originates a vertical
effective stress path. As the effective stress path is located on the
dry side of the p0 q plane, when the yield surface of the MCC model
is reached, the flow rule of this model induces a slight decrease in
the size of the yield surface (softening), consequently the effective
stress path moves to the right in the direction of the curve obtained
experimentally.
The evolution of the excess pore pressure (Du) against axial
strain (eax) obtained numerically is, in qualitative terms, similar
to the experimental results (Fig. 5c), although, with significant
numerical discrepancies. The greatest disparities (in terms of
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 123

2000 5.2. Oedometer creep tests

p'o = 50 kPa
1800
Fig. 6 compares the time evolution of the creep axial strain
1600
(eax,creep) obtained numerically from the experimental oedometer
1400 creep tests, carried out with a stabilised soil under a creep vertical
1200 pressure of 448 kPa for a curing time of 28 and 90 days.
Independently of the curing time, the creep strain rate is ini-
1000
tially almost constant, only increasing slightly after a time of
q (kPa)

800 1000 min. As expected, when the curing time increases from 28
600 CIU test, tc= 28 days to 90 days, the creep axial strain rate decreases due to the higher
CIU test, tc=90 days effectiveness of the cementation bonds. Notwithstanding the low
400
Num. analysis, tc=28 days values of the eax,creep measured, which makes it harder to obtain
200 good numerical predictions, they match the experimental results
Num. analyis, tc=90 days
0 very well, even when the creep strain rate increases. Thus, these
0.0 0.5 1.0 1.5 2.0 2.5 3.0 results show that the constitutive models associated with the
(a) ax (%) two creep laws and incorporating the effect of curing time can be
used for predicting the oedometer creep behaviour for different
curing times.
Yield surface MCC,
2000 tc=90 days
p'o = 50 kPa
5.3. Multi-stage triaxial tests

1600 Fig. 7 compares the experimental and computed results of two

Yield surface MCC, multi-stage triaxial tests, both carried out with samples of sta-
tc=28 days bilised soil isotropically consolidated for p0 0 of 50 kPa. One of the
1200
tests was performed with a sample cured for 28 days and submit-
q (kPa)

ted to a shear stress level (D) during the creep phase of 0.27, while
800 in the other test a sample with a curing time of 90 days was used
CIU test, tc= 28 days and submitted, during the creep phase, to a value of D equal to
CIU test, tc=90 days 0.40.
400
Num. analysis, tc=28 days The computed results of the first undrained stage shows an
Num. analysis, tc=90 days excellent agreement with the stress-strain behaviour (Fig. 7a)
0 obtained experimentally, which proves that the behaviour is,
0 200 400 600 800 approximately, linearly elastic at this stage independently of the
(b) p' (kPa) curing time. After the consolidation/creep stage, the experimental
data show a slight increase in the stiffness which is due to the
250 CIU test, tc=28 days
hardening of the soil as result of the reduction of the void ratio dur-
ing the consolidation and creep phase (i.e., second stage). The
CIU test, tc=90 days numerical predictions are not able to reproduce this slight increase
200
Num. analysis, tc=28 days of stiffness, since constant elastic parameters (evaluated for the
150 first stage) are used (in this case) to simulate the three stages of
Num. analysis, tc=90 days
the multi-stage triaxial tests. As observed in Fig. 5a, the numerical
100 model replicates the increase in strength due to curing time, but is
not able to simulate the loss of strength observed after the peak
u (kPa)

50 strength, due to the destructuration of stabilised soils. In fact, the

ax (%)
MCC model is unable to replicate the degradation of the structure
0 observed in the generality of the stabilised soils.
0.0 0.5 1.0 1.5 2.0 2.5 3.0
-50 p'o = 50 kPa
Time (min)
1 10 100 1000 10000
-100 0.00
(c)
'v,creep = 448 kPa
Fig. 5. Numerical prediction of the CIU tests for a chemically stabilised soil for a
curing time of 28 and 90 days: (a) qeax curves; (b) stress-path curves; (c) Du-eax
Creep axial strain (%)

0.05
curves.

excess pore pressures) between the numerical predictions and the 0.10
experimental results occur fundamentally when the effective Stabilised soil
stress paths are within the yield surface of the MCC model with
a vertical path (Fig. 5b). This behaviour is a consequence of the Oedometer creep test, tc=28 days
0.15
associated flow rule used for the VM model, which does not predict Numerical analysis, tc=28 days
volumetric plastic strains. After the yielding of the MCC model, as Oedometer creep test, tc= 90 days
the effective stress path moves to the right, the excess pore pres- Numerical analysis, tc=90 days
0.20
sure tends to decrease, thus maintaining an approximately con-
stant difference between the values predicted numerically and Fig. 6. Numerical prediction of an oedometer creep test carried out with the
those obtained experimentally. chemically stabilised soil for a curing time of 28 and 90 days.
124 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130

2000 pation of the excess pore pressure generated during the previous
stage. After the yielding, the flow rule of the MCC model induces
1750 a slight contraction of the yield surface, which promotes the lean-
ing of the stress path towards the experimental results, as seen in
1500
Fig. 5b.
1250 The time evolution of the axial strain during the creep stage
(eax-creep), depicted in Fig. 7c, shows an exponential variation of
q (kPa)

1000 the eax,creep with time, which increases with the shear stress level
LAB data, tc= 28 days (D). The results also show a higher axial creep strain rate for a cur-
750
LAB data, tc=90 days ing time of 90 days than for 28 days, which is apparently incoher-
500 ent. However, the higher axial creep rate obtained for tc of 90 days
Num. analysis, tc=28 days
is justified by the higher D (=0.40) used in the test of this sample.
Consolidation and
250 creep phases Num. analysis, tc=90 days The results show good agreement between the numerical and the
experimental results for both curing times; indeed, the numerical
0
results even make it possible to predict the exponential variation
0.0 0.5 1.0 1.5 2.0 2.5 3.0
of the eax,creep with time.
(a) ax(%)
Regarding the results presented in Figs. 57, it can be concluded
Yield surface MCC that the use of the MCC/VM models associated with both creep
2000
tc= 90 days laws (volumetric and deviatoric) and the effect of curing time are
1750 capable of reproducing the behaviour of the stabilised soil in the
long-term, that is, considering the effect of creep and curing time.
1500 Yield surface MCC
tc= 28 days
1250 6. Numerical prediction of the long-term behaviour of a large-
q (kPa)

scale embankment
1000
6.1. General characteristics
750
LAB data, tc = 28 days

500 Num. analysis, tc=28 days Considering a previous research study [19] where the MCC
LAB data, tc= 90 days model associated with creep laws (volumetric and deviatoric)
250
Num. analysis, tc=90 days
was used to replicate the behaviour of the soft soil of Baixo Mon-
Consolidation and
creep phases dego and the good predictions of the curing time behaviour of the
0 stabilised soil with the models/laws presented in the previous sec-
0 100 200 300 400 500 600 700 800
p' (kPa)
tion of this work, these models/laws are now used to analyse the
(b)
long-term behaviour of a large-scale embankment built on soft
soils reinforced with Deep Mixing Columns (DMCs).
0.07 As may be seen in Fig. 8, the soil foundation is composed by
Triaxial creep test, tc= 28 days 7.5 m of soft soil placed under a 0.5 m thick layer of sand. The
0.06 D= 0.40 water table is on the top of the layer of sand, i.e., no water flow
Triaxial creep data, tc=90 days
was considered above the water table. The construction of the
0.05 Num. analysis, tc= 28 days
embankment consists of three sub-layers, each one with a thick-
Num. analysis, tc=90 days ness of 1.0 m applied with a time delay of one day. The DMC-soil
(%)

0.04
D= 0.27 system is simulated by an axisymmetric cylindrical unit cell
ax-creep

0.03
0.02
0.01
0.00
1.000 10.000 100.000 1000.000
(c)
Fig. 7. Numerical prediction with the MCC/VM model of a multi-stage triaxial test
(with creep stage) carried out with the chemically stabilised material for a curing
time of 28 and 90 days: (a) q-eax curves; (b) stress-path curves; (c) axial strain
during creep phase.
The comparison of the stress paths obtained in the experiments
with the numerical predictions (Fig. 7b) reveals that the peak
strength is predicted reasonably well by the numerical model.
The main differences are seen for the intermediate stress states,
since vertical effective stress paths are obtained numerically inside
the yield surface, which is due to the use a linear elastic law, while
more flattened stress paths are obtained experimentally.
The consolidation/creep stage corresponds to a horizontal (the
deviatoric stress is constant) and long stress path due to the dissi-
(Fig. 8), consisting of half of one DMC (with a radius of 0.4) metre
and its radius of influence (with 1.0 m).
The finite elements (FE) mesh used in the axisymmetric analysis
consists of 130 eight-noded isoparametric quadrilateral elements
and 355 nodal points (Fig. 8). FE with twenty nodal degrees of free-
dom were used below the water table, providing a quadratic inter-
10000.000 polation of displacement and a linear interpolation of pore
Time (min) pressures, making it possible to simulate the consolidation phe-
nomenon. No element joints were used to simulate the interface
between the DMC and the soft soil, since this, in general, shows
great irregularity [4,19]. In terms of the boundary conditions of
the FE mesh, the bottom boundary was restrained from moving
in both directions, both lateral vertical sides were restrained from
moving horizontally, the bottom and top boundaries of the soil
foundation are permeable, while the lateral boundaries are imper-
meable (i.e., no flow of water through these boundaries).
Table 1 shows the parameters of the constitutive models, creep
laws and the effect of the curing time of the soft soil and the sta-
bilised soil, used in the numerical analyses of the embankment.
It is considered that the construction of the embankment only
starts after a minimum curing time of 28 days for the DMCs.
Table 3 presents the initial state, elastic, strength and perme-
ability parameters used to characterize the embankment, sand
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 125

The embankment material (c = 22.0 kN/m3) is characterized by

the Mohr-Coulomb failure criterion (c0 = 10, u0 = 35) associated
with an isotropic linear elastic law, where E varies between 1
and 15 MPa, from the upper to the bottom sub-layer of the
embankment, in order to simulate the greater confinement of the
bottom sub-layers.
Table 4 shows the eight numerical cases carried out in this
work. The first two cases (A and B) analyse the behaviour of the
embankment on the soft soil without DMCs, one without creep
(case A) and the other considering the creep of the soft soil (case
B). The six remaining cases (C to H) simulate the reinforcement
of the soil foundation with DMCs, considering the existence of
creep for both the soft soil and the DMCs (cases D, F and H) or
ignoring the creep phenomenon (Cases C, E and G). Additionally,
the analysis of the repercussions of the effect of the creep phe-
nomenon is performed for the properties of the DMCs evaluated
for 28 days (Cases C and D), 90 days (cases E and F) and considering
the enhancement of the properties of the DMCs with the curing
time, as stated in Section 4.3 (cases G and H).

6.2. Settlement analysis

Fig. 9 depicts the time evolution of the settlement at the top of

Fig. 8. FEM mesh used in the numerical analyses of the embankment.
layer, soft soil and DMCs. An anisotropy of permeability with kh/kv
of 3 was used for the soft soil in accordance with the range
observed in similar sedimentary soils [55]. Eq. (15) is used to con-
sider the variation of the coefficients of the permeability with the
void ratio [33]:
ee0
k k0  10 Ck
where e0 is the initial void ratio, k0 the coefficient of permeability
corresponding to e0, k is the corrected coefficient of permeability
relative to the actual void ratio e, and Ck is a constant equal to
e0/2 [56].
The behaviour of the soft soil inside the yield surface of the MCC
model was characterized by an isotropic linear elastic law, defined
by Eq. (16) [59]:
31 e0 1  2t0
E0 p00
j from 14.8 cm (case D) to 12.6 cm (case H), i.e. about 15%. Another
where p0 0 is the initial volumetric effective stress.
the soil foundation for the vertical axis of the unit cell. Indepen-
dently of taking the creep phenomenon into consideration or not,
the results clearly show that the reinforcement of the soft soil with
DMCs has a great impact on the settlement of the embankment,
inducing, for a time of 20 years, a decrease in the settlements from
about 90 cm (without DMCs) to values lower than 15 cm (with
DMCs).
In relative terms, the results also show that the effect of the
creep phenomenon is more significant when the soil foundation
of the embankment is reinforced by DMCs. In fact, comparing cases
A and B, where the effect of creep phenomenon on the behaviour of
the unreinforced (no DMCs) soft soil is studied, the results show
that taking the creep phenomenon into consideration increases
the settlement slightly, from 84.8 to 88.7 cm (i.e. about 5%) for a
time of 20 years. However, when the soft soil is reinforced with
15
DMCs, the consideration of creep has a much greater impact on
the settlements; for instance, for a curing time of 28 days the con-
sideration of the creep increases the settlement about two times,
from 7.4 cm to 14.8 cm (cases C and D).
The results of Fig. 9 also reveal that the inclusion of the effect of
the curing time in the numerical calculations (cases G and H) only
has a visible impact in the long-term predictions when the creep
phenomenon is considered (case H). Indeed, the increase of the
cementation bonds over time (i.e. the effect of the curing time)
16 induces, for a time of 20 years, a slight reduction in the settlement
fact observed, is that the numerical results obtained during the

Table 3
Properties of the materials used to predict the behaviour of an embankment (based on [13]).

Material Initial state Elastic parameters Permeability Mohr-Coulomb

parameters
K0 c OCR m0 E0 kv kh/kv c0 /0
(kN/m3) (MPa) (m/s) (kPa) ()
Soft soil 0.4a 15.0 1.0 0.3 c
109 3.0 MCC model
DMCs 0.8b 16.0 0.3 164.7 3  1010 1.0 MCC/VM model
Sand layer 0.4 15.0 0.3 2.0 104 1.0
Embankment 22.0 0.3 1.0/7.5/15.0d 10.0 35.0
a
Evaluated from triaxial test [57].
b
Established from hnberg [58].
Calculated by [59]: E0 31e0 j12t p00 .
0
c

d
Increases from the top to the bottom sub-layer.
126 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130

Table 4
Numerical predictions. Cases analysed.

Case Soft soil DMC

No creep Creep No creep Creep tc (days)
A d
B d
C d d 28
D d d 28
E d d 90
F d d 90
G d d Variablea
H d d Variablea

d Condition used.
a
The properties depend on the tc (28360 days).

0.0
-0.050

-0.1
Case C Case E Case G
-0.2 -0.075
Soil (No creep) - A

Settlement (m)
-0.3 Soil (Creep) - B
Settlement (m)

Soil (No creep), DMC (No creep), tc=28 days - C -0.100

-0.4
Soil (Creep), DMC (Creep), tc=28 days - D
-0.5 Soil (No creep), DMC (No creep), tc=90 days - E
-0.125
Soil (Creep), DMC (Creep), tc=90 days - F
-0.6
Soil (No creep), DMC (No creep), tc=var - G
-0.7 Soil (Creep), DMC (Creep), tc=Var - H -0.150
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
-0.8 Time (years)

-0.9
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0

Time (years)

Fig. 9. Numerical predictions. Evolution over time of the settlement at the surface of the soil foundation.

first 20 years with the parameters evaluated for a curing time of
90 days (case F) are similar to those obtained when considering
the effect of curing time (case H). Thus, the use of the properties
evaluated for 90 days of curing can be used to simulate the long-
term behaviour with reasonable accuracy.
Independently of the curing time considered, the evolution of
the improvement factor (IF, the ratio between the settlement
obtained with unreinforced soil and the settlement with DMCs)
over time (Fig. 10) reveals a significant decrease in the IF when
the creep phenomenon is considered in the analysis. Thus, for tc
of 28 days the decrease in the IF is about 46%, i.e., from 10.7 (case
C) to 5.8 (case D), while taking the effect of the tc into consideration
in the calculations has a lower impact on the IF, since a reduction
of about 38% is observed, i.e., from 10.9 (case G) to 6.8 (case H).
These results clearly show that the effectiveness of the reinforce-
ment with DMCs decreases significantly when creep is included
in the numerical analysis, although the consideration of the effect
of tc attenuates this influence slightly, i.e., the IF increases from 5.8
(Case D) to 6.8 (Case H). As seen in Fig. 9, these results show that
the use of properties for tc of 90 days (case F) can be used instead
of the use of the variation of properties with curing time (case H).
The previously described behaviour is due to the arching
effect, which is the transference of stresses from the soft soil, a
more deformable material, to the DMCs, a stiffer material. Thus,
the increase in the vertical stresses on the DMCs due to the arch-
ing effect [4,19] potentiates the creep strains, since Cae increases
with the creep stress (Eq. (9)) and the axial creep strain rate, as sta-
ted by Singh-Mitchells law, increases exponentially with the stress
level (Eq. (3)). Furthermore, the consideration of the enhancement
of the mechanical and creep properties of DMCs over time (i.e.
effect of tc), attenuates the effect of creep slightly.
6.3. Analysis of the stresses
Fig. 11 illustrates the evolution over time of the vertical
effective stress for the DMCs, r0 v-DMC, (Fig. 11a) and for the soil,
r0 v-soil, (Fig. 11b) respectively, for a depth of 2.5 m. The results
show a huge increase in r0 v-DMC and a significant reduction in
r0 v-soil in the surrounding soil (cases C to H), which confirms the
existence of the arching effect. Without the creep phenomenon
(cases C, E and G), the effect of tc in terms of distribution of
stresses, r0 v-DMC and r0 v-soil, is negligible.
The results also reveal that taking the creep phenomenon into
consideration (cases D, F and H), potentiates the impact of the
arching effect, since a greater increase of the r0 v-DMC is observed.
This behaviour is explained by the greater creep strains of the soft
soil in relation to the DMCs, which promotes a higher transference
of stresses from the soft soil to the DMCs, therefore r0 v-DMC
increases in relation to the cases without creep (cases C, E and
G). Fig. 11a also shows that, with creep, the consideration of the
enhancement of the properties over time (case H) originates less
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 127

12
Soil (No creep), DMC (No
creep), tc=28days - C

10 Soil (Creep), DMC

Improvement factor, IF
(Creep), tc=28 days - D

8 Soil (No creep), DMC (No

creep), tc=90 days - E

6 Soil (Creep), DMC

(Creep), tc=90days - F

Soil (No creep), DMC (No

4 creep), tc=Var - G

Soil(Creep), DMC(Creep),
2 tc=Var - H
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Time (years)

Fig. 10. Numerical predictions. Evolution over time of the improvement factor.

500 440

430
400
'v-DMC (kPa)

420
z = 2.5 m
'v-DMC (kPa)

410
300 Soil (No creep), DMC (No creep), tc=28 days - C

400
Soil (Creep), DMC (Creep), tc=28days - D

200 Soil (No creep), DMC (No creep), tc=90 days - E 390

Soil (Creep), DSM (Creep), tc=90 days - F

380
0.0 1.0 2.0 3.0 4.0 5.0
100 Soil (No creep), DMC (No creep), tc=Var - G
Time (years)
Soil (Creep), DMC (Creep), tc=Var - H

0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
(a) Time (years)

80
z = 2.5 m

16
60
Soil (No creep) - A / Soil (Creep) - B
'v-soil (kPa)

Soil (No creep), DMC (No creep), tc=28 days - C Case C Case E Case G
12
Soil (Creep), DMC (Creep), tc=28days - D
40
'v-Soil (kPa)

Soil (No creep), DMC (No creep), tc=90 days - E

Soil (Creep), DSM (Creep), tc=90 days - F 8

Soil (No creep), DMC (No creep), tc=Var - G
20 Soil (Creep), DMC (Creep), tc=Var - H
4

0 0
0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 0.0 1.0 2.0 3.0 4.0 5.0
Time (years)
(b) Time (years)

Fig. 11. Numerical predictions. Time evolution of the vertical effective stress at 2.5 m depth. (a) DMC; (b) soft soil.
128
r0 v-DMC in the short-term (t < 4 years) than in the cases D (tc = 28 -
days) and F (tc = 90 days).
The evolution of the stress concentration ratio (SCR, the ratio
between r0 v-DMC and r0 v-Soil) over time clarifies some aspects of
the behaviour (Fig. 12). Without creep, the SCR is practically not
dependent on the tc considered (cases C, E and G), and stabilises
at a value of about 30. With the consideration of the creep phe-
nomenon, the SCR is much higher than the values obtained with-
out creep and the SCR grows when the mechanical and creep
properties of the DMCs are enhanced over time; thus, from case
D to case H, the SCR increases from 45.4 to 54.1 (i.e. about 19%)
for a time of 20 years. Another aspect observed in Fig. 12 is the sig-
nificant decrease in the SCR over time when creep is considered,
which is a direct consequence of the increase of settlement on
the DMCs over time due to creep (Fig. 9), which decreases the
arching effect. In fact, the decrease in the SCR over time was
observed in the field data of an embankment built on soft soils
improved with DMCs [28], predicted in numerical analyses
[19,60] and confirmed by physical model tests [20]. Although this
behaviour has been attributed to the softening/yielding of DMCs
[20,28,60], the effect of creep is an alternative justification of this
phenomenon.
7. Conclusions
This work studies the ability of the Modified Cam Clay (MCC)
model associated with the Von Mises (VM) model, with volumetric
and deviatoric creep laws and considering the effect of curing time
on the mechanical and creep properties to simulate the long-term
behaviour of a chemically stabilised soft soil. The soil was sta-
bilised with 125 kg/m3 of a binder mixture composed by Portland
cement Type I 42.5 R and blast furnace granulated slag with a pro-
portion of 75/25.
The relationship between the curing time and the undrained
shear strength obtained in UCS tests was approximated by a power
function. This function is shown to be suitable to predict the evo-
lution over time of the effective yield stress and the undrained
Youngs modulus (with a correction factor of 0.4), while the inverse
of this function is suitable to predict the creep parameters Ca and A
over the curing time.
The results of the CIU triaxial tests, oedometer creep tests and
multi-stage triaxial tests, carried out for curing times of 28 and
90 days, were used to validate the constitutive model, the creep
laws and the effect of curing time. Some conclusions can be drawn
from this study:
100
80
Stress concentration ratio, SCR
60
40
20
0 tc=Var - H
0.0
Fig. 12. Numerical predictions. Time evolution of the stress concentration ratio at 2.5 m depth.
Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130
The MCC/VM model associated with the effect of curing time is
suitable to predict the stress-strain behaviour until failure of
samples of the stabilised soil, for two curing times (28 and
90 days). However, the model is unable to predict the softening
observed after the peak strength.
The models/laws used predict the qualitative evolution of the
excess pore pressure and the state of stress at the peak strength
reasonably well. However, they are not able to replicate the
evolution of the stress paths, since vertical stress paths inside
the yield surface of the MCC model are predicted, which are dif-
ferent from those obtained in the CIU triaxial tests.
The MCC/VM model associated with the creep laws and the
effect of curing time predict axial creep strains similar to those
observed in the oedometer creep tests and multi-stage triaxial
tests performed for the two curing times.
The numerical predictions of the multi-stage triaxial tests (with
an intermediate creep stage) originate stress paths and stress-
strain curves coherent with the theoretical frameworks,
although with some discrepancies in relation to the experimen-
tal results.
The models/laws previously tested were used to predict the
long-term behaviour of an embankment built on a soft soil rein-
forced with Deep Mixing Columns (DMCs). In these analyses, the
soft soil was modelled by the MCC model associated with two
creep laws (volumetric and deviatoric). The numerical results of
eight analyses, with and without creep, for different curing times,
were analysed in terms of settlement and vertical effective stres-
ses. Some conclusions can be drawn from the long-term analysis
of the embankment built on a soft soil reinforced with DMCs:
The consideration of the creep phenomenon for the soft soil and
DMCs, significantly increases the settlement of the embank-
ment, which leads to a decrease in the improvement factor, that
is, the reinforcement of the soil foundation with DMCs is less
efficient when creep in considered.
In relation to the cases without creep, the consideration of this
phenomenon induces higher effective vertical stresses on the
DMCs, and consequently a greater stress concentration ratio.
The results with creep, also show that the stress concentration
ratio tends to decrease over time, which means that the trans-
ference of the stresses from the soft soil to the DMCs (i.e., the
arching effect) becomes less important with the elapsed time.
The consideration of the effect of curing time on the enhance-
ment of the mechanical and creep properties of the DMCs,
Soil (No creep), DMC (No
creep), tc=28days - C
Soil (Creep), DMC
(Creep), tc=28 days - D
Soil (No creep), DMC (No
creep), tc=90 days - E
Soil (Creep), DMC
(Creep), tc=90days - F
Case C Case E Case G Soil (No creep), DMC (No
creep), tc=Var - G
Soil(Creep), DMC(Creep),
2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0
Time (years)
 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130 129

due to pozzolanic reactions, induces a reduction in the settle- [4] Venda Oliveira PJ, Pinheiro JLP, Correia AAS. Numerical analysis of an
embankment built on soft soil reinforced with deep mixing columns:
ment in the long-term, as well as a slight increase in the
parametric study. Comput Geotech 2011;38(4):56676.
improvement factor and the stress concentration ratio. [5] Jamsawang P, Yoobanpot N, Thanasisathit N, Voottipruex P, Jonpradist P.
From a practical point of view, the results with creep also show Three-dimensional numerical analysis of a DCM column-supported highway
that the use of the mechanical and creep properties of the DMCs embankment. Comput Geotech 2016;72:4256.
[6] Poorooshasb HB, Meyerhof GG. Analysis of behaviour of stone columns and
for a curing time of 90 days is similar to considering the varia- lime columns. Comput Geotech 1997;20(1):4770.
tion of these properties over time in the numerical calculations. [7] Rampello S, Callisto L. Predicted and observed performance of an oil tank
founded on soilcement columns in clayey soils. Soils Found 2003;43
Finally it should be emphasised that, although the constitutive (4):22941.
model used (MCC/VM) does not simulate the degradation of the [8] Han J, Oztoprak S, Parsons RL, Huang J. Numerical analysis of foundation
columns to support widening of embankments. Comput Geotech 2007;34
structure observed in stabilised soils, the association of the MCC/ (6):43548.
VM model with volumetric and deviatoric creep laws and the effect [9] Huang J, Han J, Oztoprak S. Coupled mechanical and hydraulic modeling of
of curing time provides reliable long-term numerical predictions geosynthetic-reinforced column-supported embankments. J Geotech
Geoenviron Eng (ASCE) 2009;135(8):101121.
for the behaviour of stabilised soils. Indeed, this combination of [10] Huang J, Han J. 3D coupled mechanical and hydraulic modeling of a
features is not seen in commercial codes, which is the novelty of geosynthetic-reinforced deep mixed column-supported embankment.
this work. Geotext Geomembr 2009;27(4):27280.
[11] Abusharar SW, Zheng JJ, Chen BG. Finite element modeling of the consolidation
behavior of multi-column supported road embankment. Comput Geotech
Acknowledgements 2009;36(4):67685.
[12] Zheng JJ, Abusharar SW, Wang XZ. Three-dimensional nonlinear finite element
modeling of composite foundation formed by CFGlime piles. Comput Geotech
The authors would like to express their thanks to the institu- 2008;35(4):63743.
tions that financially supported the research: CIEPQPF (project [13] Liu K-W, Rowe RK. Numerical study of the effects of geosynthetic
EQB/UI0102/2014) and ISISE (project UID/ECI/04029/2013). reinforcement viscosity on behaviour of embankments supported by deep-
mixing-method columns. Geotext Geomembr 2015;43(4):567678.
[14] Suebsuk J, Horpibulsuk S, Liu MD. Modified structured Cam Clay: a generalized
critical state model for destructured, naturally structured and artificially
Appendix A. Deduction of the hardening rule due to the effect of
structured clays. Comput Geotech 2010;37:95668.
the curing time [15] Robin V, Javadi AA, Cuisinier O, Masrouri. An effective constitutive model for
lime treated soils. Comput Geotech 2015;66:189202.
Considering Fig. 1, Dh(tc) can be expressed by: [16] Horpibulsuk S, Liu MD, Liyanapathirana DS, Suebsuk J. Behaviour of cemented
clay simulated via the theoretical framework of the structured Cam Clay
model. Comput Geotech 2010;37:19.
Dhtc eAj  eBj A:1 [17] Nguyen LD, Fatahi B, Khabbaz H. A constitutive model for cemented clays
capturing cementation degrdation. Int J Plast 2014;56:118.
A B
where ej and ej are represented by the following equations: [18] Nguyen L, Fatahi B. Behaviour of clay treated with cement & fibre while
capturing cementation degradation and fibre failure C3 F model. Int J Plast
eAj eA j  ln p0A
c A:2 2016;81:16895.
[19] Venda Oliveira PJ, Correia AAS, Lemos LJL. Numerical prediction of the creep
behaviour of an unstabilised and a chemically stabilised soft soil. Comput
eBj eB j  ln p0B
c A:3 Geotech 2017;87:2031.
[20] Horpibulsuk S, Chinkulkijniwat A, Cholphatsorn A, Suebsuk J, Liu M.
Substituting Eqs. (A.2) and (A.3) in Eq. (A.1), it is obtained: Consolidation behavior of soil-cement column improved ground. Comput
Geotech 2012;43:3750.


B 
Dhtc eA j  ln p0A
c  e j  ln p0B
c
[21] Venda Oliveira PJ, Correia AAS, Mira ESP. Mitigation of creep deformations by
preloading: laboratory study. Proc Inst Civ Eng (ICE) Geotech Eng 2012;166
p0B (6):594600.
eA  eB  j  ln c
A:4 [22] Venda Oliveira PJ, Vieira AFV, Correia AAS. Effect of organic matter content on
p0A
c creep mitigation by preloading. ICE Geotech Eng 2017. http://dx.doi.org/
10.1680/jgeen.16.00082.
Substituting the following equation: [23] Venda Oliveira PJ, Correia AAS, Garcia MR. Effect of organic matter content and
curing conditions on the creep behavior of an artificially stabilized soil. J Mater
eA  eB p0B Civ Eng 2012;24(7):86875.
k p0B
) eA  eB k  ln c
A:5 [24] Venda Oliveira PJ, Correia AAS, Garcia MR. Effect of stress level and binder
ln c p0A
c
p0A
c
composition on secondary compression of an artificially stabilized soil. J
Geotech Geoenviron Eng 2013;139(5):81020.
in Eq. (A.4), Dhtc assumes the form: [25] Voottipruex P, Bergado DT, Suksawat T, Jamsawang P, Cheng W. Behavior and
simulation of deep cement mixing (DCM) and stiffened deep cement mixing
(SDCM) piles under full scale loading. Soils Found 2011;51(2):30720.
p0B p0B p0B
Dhtc k  ln c
0A
 j  ln 0A
c
k  j  ln 0A
c
A:6 [26] Chai JC, Shrestha S, Hino T, Ding WQ, Kamo Y, Carter J. 2D and 3D analyses of
pc pc pc an embankment on clay improved by soil-cement columns. Comput Geotech
2015;68:2837.
which can be expressed by Eq. (A.7) for a reference curing time of [27] Bergado DT, Ruenkrairergsa T, Taesiri Y, Balasubramaniam AS. Deep soil
mixing used to reduce embankment settlement. Ground Improve 1999;3
28 days.
(4):14562.
[28] Yapage NNS, Liyanapathirana DS, Kelly RB, Poulos HG, Leo CJ. Numerical
p0ctc P28d
Dhtc k  j  ln A:7 modeling of an embankment over soft ground improved with deep cement
p0ctc 28d mixed columns: case history. J Geotech Geoenviron Eng 2014;140
(11):04014062.
[29] Harichane K, Ghrici M, Kenai S. Effect of curing time on shear strength of
cohesive soils stabilized with combination of lime and natural pozzolana. Int J
References Civ Eng 2011;9(2):906.
[30] Kavazanjian E, Mitchell JK. Time-dependent deformation behaviour of clays. J
[1] Correia AAS. Applicability of deep mixing technique to the soft soil of Baixo Geotech Eng Div (ASCE) 1980;106(GT6):61130.
Mondego. Ph.D. dissertation. Coimbra (Portugal): University of Coimbra; 2011 [31] Borja RI. Generalized creep and stress relaxation model for clays. J Geotech Eng
[in Portuguese]. (ASCE) 1992;118(11):176586.
[2] Coelho PALF. Geotechnical characterization of soft soils. Study of the [32] Borja RI, Kavazanjian Jr E. A constitutive model for the stress-strain-time
experimental site of Quinta do Foja. MSc dissertation. Portugal: University of behavior for wet clays. Gotechnique 1985;35(3):18398.
Coimbra; 2000 [in Portuguese]. [33] Taylor DW. Fundamentals of soil mechanics. New York: John Wiley and Sons
[3] Venda Oliveira PJ, Lemos LJL, Coelho PALP. Behavior of an atypical Inc; 1948.
embankment on soft soil: field observations and numerical simulation. J [34] Hsieh HS, Kavazanjian Jr E, Borja RI. Double yield surface Cam-Clay plasticity
Geotech Geoenviron Eng 2010;136(1):3547. model I: Theory. J Geotech Eng (ASCE) 1990;116(9):1381401.
130 Paulo J. Venda Oliveira et al. / Computers and Geotechnics 91 (2017) 117130
[35] Singh A, Mitchell JK. General stress-strain-time functions for soils. J Soil Mech
Found Div (ASCE) 1968;94(SM1):2146.
[36] Morsy MM, Morgenstern NR, Chan DH. Simulation of creep deformation in the
foundation of Tar Island Dyke. Can Geotech J 1995;32(6):100210023.
[37] Kuhn MR, Mitchell JK. New perspectives on soil creep. J Geotech Eng 1993;119
(3):50723.
[38] Morsy MM, Chan DH, Morgenstern NR. An effective stress model for creep of
clay. Can Geotech J 1995;32(5):81934.
[39] Jardine RJ, Symes MJ, Burland JB. The measurement of soil stiffness in the
triaxial apparatus. Gotechnique 1984;34(3):32340.
[40] Atkinson JH, Richardson D, Stallebrass SE. Effect of recent stress
history on the stiffness of over consolidated soil. Gotechnique 1990;40(4):
53140.
[41] Smith JR, Jardine RJ, Hight DW. The yielding of Bothkennar clay. Gotechnique
1992;42(2):25774.
[42] Troung DM, Magnan JP. Application des modles lastoplastiques de
lUniversit de Cambridge au calcul du comportement dun remblai
exprimental sur sols mous. Rapport de recherche LPC no. 74. Paris:
Laboratoire Central des Ponts et Chausses; 1977.
[43] Kondner RL. Hyperbolic stress-strain response: cohesive soils. J Soil Mech
Found Div, ASCE 1963;106(6):61130.
[44] Duncan JM, Chang CY. Nonlinear analysis of stress and strain in soils. J Soil
Mech Found Div, ASCE 1970;96(SM 5):162953.
[45] ASTM D2487. Standard classification of soils for engineering purposes (unified
soil classification system). West Conshohoken (PA); 1998.
[46] EN 197-1. Cement e Part 1: composition, specifications and conformity criteria
for common cements. European Committee for Standardization, June 2010;
2000.
[47] Janz M, Johansson S-E. The function of different binding agents in deep
stabilization. Report 9. Linkping (Sweden): Swedish Deep Stabilization
Research Centre; 2002.
[48] Kawasaki T, Niina A, Saitoh S, Honjyo Y. Deep mixing method using cement
hardening agent. In: Proceedings of the 10th international conference on soil
mechanics and foundation engineering, vol. 3; 1981. p. 7214.
[49] Terashi M. Theme lecture: deep mixing method brief state of the art. In: 14th
international conference on soil mechanics and foundation engineering,
Hamburg, Germany, vol. 4; 1997. p. 24758.
[50] Kitazume M, Terashi M. The deep mixing method principle, design and
construction. Japan (Balkema): Coastal Development Institute of Technology;
2002.
[51] Horpibulsuk S, Miura N, Nagaraj TS. Assessment of strength development in
cement-admixed high water content clays with Abrams law as a basis.
Gotechnique 2003;53(4):43944.
[52] Kalantari B, Prasad A. A study of the effect of various curing techniques on the
strength of stabilized peat. Transp Geotech 2014;1(3):11928.
[53] Zhang RJ, Lu YT, Tan TS, Phoon KK, Santoso AM. Long-term effect of curing
temperature on the strength behavior of cement-stabilized clay. J Geotech
Geoenviron Eng 2014;140(8):04014045.
[54] Wong LS, Hashim R, Ali F. Utilization of sodium bentonite to maximize the
filler and pozzolanic effects of stabilized peat. Eng Geol 2013;152(1):5666.
[55] Jamiolkowski M, Ladd CC, Germaine JT, Lancellotta R. New developments in
field and laboratory testing of soils. In: Proc 11th ICSMGE, San Francisco, vol. 1;
1985. p. 57154.
[56] Tavenas F, Jean P, Leblond P, Leroueil S. The permeability of natural soft, part II:
permeability characteristics. Can Geotech J 1983;20(4):64560.
[57] Correia AAS, Venda Oliveira PJ, Lemos LJL, Mira ESP. Creep deformations of
deep mixing columns. Evaluation from laboratorial tests. In: Proc int symp of
deep mixing and admixture stabilization, Port and Airport Research Institute,
Yokosuka City, Japan; 2009.
[58] hnberg H. Strength of stabilised soils a laboratory study on clays and
organic soils stabilised with different types of binder. Ph.D. thesis. Sweden:
University of Lund; 2006.
[59] Atkinson JH, Bransby PL. The mechanics of soils an introduction to critical
state soil mechanics. McGraw-Hill: University Series in Civil Engineering;
1978.
[60] Jiang Y, Han J, Zheng G. Influence of column yielding on degree of
consolidation of soft foundations improved by deep mixed columns.
Geomech Eng 2014;6(2):17394.