Professional Documents
Culture Documents
net/publication/280181487
CITATIONS READS
78 1,355
2 authors, including:
C.W.W. Ng
The Hong Kong University of Science and Technology
603 PUBLICATIONS 19,545 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by C.W.W. Ng on 23 July 2015.
Subgrade soils are constantly subjected to daily variations of temperature and suction in the field.
Their deformation under cyclic traffic loads can greatly affect the performance of a pavement. Yet, the
behaviour of unsaturated subgrade soils under cyclic loads at various temperatures and suctions is not
fully understood. In this study, a new suction- and temperature-controlled cyclic triaxial apparatus was
developed. Using temperature-compensated Hall effect transducers, the accumulated plastic axial and
volumetric strains and resilient modulus of an unsaturated silt under cyclic loads were measured at
suctions ranging from 0 to 60 kPa and temperatures ranging from 20 to 608C. Experimental results
show that yield stress increases with increasing suction (suction-induced hardening), but decreases
with increasing temperature (thermal-induced softening). Consequently, axial plastic strain accumu-
lated during cyclic loading–unloading is larger at lower suctions and higher temperatures. At a given
temperature, measured resilient modulus significantly increases when suction increases. On the other
hand, the influence of temperature on resilient modulus depends on suction level. At zero suction,
resilient modulus is slightly affected by temperature. At suction larger than zero, an increase in
temperature induces an obvious decrease in resilient modulus. It is evident that thermal effects on
resilient modulus are more significant at unsaturated than saturated states.
KEYWORDS: partial saturation; pavements and roads; repeated loading; suction; temperature effects
INTRODUCTION contents result in differences not only in initial suction, but also
It is well recognised that deformation of subgrade soils in pore size distribution. It is difficult to separate the effects of
under cyclic traffic loads is crucial to pavement performance suction on cyclic behaviour from the effects of pore size
(Brown, 1996). In the field, subgrade soils are often unsatu- distribution. To address this problem, Yang et al. (2008) and Ng
rated and subjected to daily variations of water content and et al. (2013) carried out suction-controlled cyclic triaxial tests
hence pore-water pressure (or suction) and temperature (Jin to measure the resilient modulus of unsaturated soil. All soil
et al., 1994; McCartney & Khosravi, 2012). Many theor- specimens were prepared at the same initial water content.
etical and experimental studies have demonstrated that un- From the same initial state, soil specimens were dried or wetted
saturated soil behaviour is highly dependent on suction and to states with different suctions before the application of cyclic
temperature (Romero et al., 2003; François & Laloui, 2008; stresses. Yang et al. (2008) reported that when cyclic stress
Tang et al., 2008; Uchaipichat & Khalili, 2009). Most of increases, M R decreases at low matric suction but increases at
these studies, however, focused on soil behaviour under high matric suction. Ng et al. (2013) observed that M R in-
monotonic loading. As far as the authors are aware, there is creases by up to one order of magnitude when suction increases
no study of coupled effects of suction and temperature on from 0 to 250 kPa. Since these cyclic tests were conducted at
cyclic behaviour of unsaturated soils in the literature, which room temperature, however, the coupled effects of suction and
includes plastic strain accumulation and resilient modulus. temperature were not thoroughly investigated.
The resilient modulus (M R ) was defined by Seed et al. As for thermal effects on the cyclic behaviour of unsatu-
(1962) as the ratio of repeated deviator stress to axial rated soil, Zhou & Ng (2013) investigated thermal effects on
recoverable strain. Accumulated plastic strain and resilient the resilient modulus of a compacted silt at zero suction
modulus are widely used in pavement engineering to predict only. They found that resilient modulus changes by about 5%
irreversible and reversible deformations, respectively. when temperature is increased from 20 to 408C. Cekerevac
To study suction effects on the cyclic behaviour of unsatu- & Laloui (2010) carried out undrained cyclic triaxial tests on
rated soil, some researchers prepared soil specimens at different saturated consolidated kaolin (zero suction) at two different
initial water contents. The resilient modulus and accumulated temperatures (22 and 908C). Their experimental data showed
axial plastic strain of each specimen were then measured during that cyclic loads at 908C lead to lower axial plastic strain
cyclic loading–unloading (e.g. Khoury & Zaman, 2004; Siva- and smaller excess pore-water pressure per cycle than in
kumar et al., 2013). As shown by Tarantino & De Col (2008) unheated specimens. Although thermo-mechanical soil be-
and Sivakumar et al. (2013), different initial compaction water haviour depends on stress history (or overconsolidation ratio)
and suction (Baldi et al., 1988; Cekerevac & Laloui, 2004;
Tang et al., 2008), both Zhou & Ng (2013) and Cekerevac &
Laloui (2010) considered zero suction only, with the latter
Manuscript received 22 January 2014; revised manuscript accepted 10 testing only normally consolidated soil. Coupled effects of
July 2014. Published online ahead of print 22 September 2014. stress (overconsolidation ratio), suction and temperature un-
Discussion on this paper closes on 1 February 2015, for further details
see p. ii.
der cyclic loadings are not fully understood.
Department of Civil and Environmental Engineering, The Hong In this study, a new suction- and temperature-controlled
Kong University of Science and Technology, Clear Water Bay, triaxial system was developed. A series of cyclic triaxial
Kowloon, Hong Kong. tests on unsaturated soil were carried out using the new
† Guangzhou HKUST Fok Ying Tung Research Institute, Guangzhou, system. The scope of this study is to investigate influences
709
710 NG AND ZHOU
of number of cycles, cyclic stress, suction and temperature 1975). The volume of collected air was used to correct the
on the cyclic behaviour of unsaturated silt, including the measured change in soil water content.
accumulation of plastic strain and resilient modulus. In To apply thermal loading, a heating system was added to
particular, coupled effects of stress, suction and temperature the triaxial apparatus. The heating system consisted of a
under cyclic loadings are analysed. The measured and inter- thermostat, a heater and thermocouple A. As shown in Fig.
preted results would be useful for predicting settlement of 1, the heater and thermocouple A were installed inside the
subgrade soils and hence the performance of pavements. triaxial cell. Both of them were connected to the thermostat,
forming a closed-loop control and feedback system. During
testing, the thermostat adjusted the heater output based on
feedback from thermocouple A. When the dissipating energy
TESTING APPARATUS AND MEASURING DEVICE and the heater output were balanced, the air temperature
To test unsaturated soil at various suctions and tempera- inside the triaxial cell reached equilibrium. It is reasonable
tures, a suction-controlled triaxial apparatus (Ng & Yung, to assume that the soil temperature is equal to the air
2008) was modified to allow independent control of suction temperature in the triaxial cell at equilibrium state. Depend-
and temperature. Fig. 1 shows a schematic diagram of the ing on the target temperature, 1–3 h are required to achieve
new apparatus. Matric suction (ua uw), where ua and uw thermal equilibrium. The duration adopted is consistent with
are pore-air pressure and pore-water pressure respectively, that reported by Cekerevac et al. (2005). More discussion
was controlled using the axis-translation technique. This about temperature control is given in the later section ‘Test
technique imposes ua uw on the soil specimen by control- programme and procedures’.
ling ua and uw independently. ua was controlled through a To enhance the uniformity of temperature inside the
coarse, low air-entry value (AEV) porous stone sitting on triaxial cell, two small fans and a hollow aluminium cylinder
top of the soil specimen. uw was controlled through a were installed in the triaxial cell. The fan improves air
saturated, high AEV (i.e. 3 bar) ceramic disc sealed to the circulation, while the aluminium cylinder enhances heat
pedestal of the triaxial cell. The saturated ceramic disc transfer because of its high thermal conductivity. To check
permitted the exchange of water through it but prevented the the uniformity of temperature, two thermocouples (A and B)
passage of free air as long as ua uw was lower than its were installed at different locations inside the triaxial cell.
AEV. However, dissolved air in water may pass through the They measured air temperature at a distance of about 5 mm
ceramic disc and accumulate either underneath the ceramic from the soil specimen. When thermal equalisation was
disc or in the water drainage system. In this study, any reached, the readings of the two thermocouples were almost
accumulated air bubble was flushed out and collected once identical. They remained fairly constant with a maximum
every 24 h using a diffused air volume indicator (Fredlund, fluctuation of 0 .58C.
Load cell
LVDT
Thermostat
Porous stone
Suction probe
Axial Hall effect transducer
Pressure: kPa
deformation at the mid-height of each specimen. One of 60°C – Unloading
200
these Hall effect transducers was used to measure radial
deformation, while the other two were used to measure axial
deformation independently. After calibration, the resolution
100
and accuracy of each Hall effect transducer were about 1
and 3 ìm, respectively, corresponding to vertical strains of
about 0 .001% and 0 .003% (Ng & Xu, 2012). As expected,
0
axial strain obtained using an LVDT (external device) was 0 10 20 30 40
generally larger than that obtained using a Hall effect Output voltage: mv
transducer (local device). This is because the former method (a)
measures the overall deformation of a soil specimen together 5 20°C – Loading
with bedding errors and compliance of the system (Jardine 20°C – Unloading
et al., 1984), whereas the latter records the actual displace- 40°C – Loading
4
ment of the soil specimen. For greater reliability, axial strain 40°C – Unloading
measured using Hall effect transducers was used and re- 60°C – First loading
Displacement: mm
ported. 3 60°C – First unloading
Since the Hall effect transducers measured local deforma- 60°C – Second loading
tion at the mid-height of a specimen, it was sensible to 60°C – Second unloading
2
measure pore-water pressure also at this level. A suction
probe was employed to monitor pore-water pressure at the
mid-height of a specimen during cyclic loading–unloading, 1
although a conventional pore-water pressure transducer was
still used to measure pore-water pressure at the bottom. For 0
unsaturated soil testing, pore-water pressure measured at the ⫺8 ⫺4 0 4 8
mid-height also provides guidance for suction equalisation. Output voltage: mv
(b)
The suction probe was modified from a Druck PDCR-81
miniature pore-water pressure transducer by replacing the Fig. 2. Calibration of transducers at different temperatures:
low AEV ceramic tip with a high AEV ceramic tip (a) suction probe; (b) Hall effect transducer
(500 kPa) (Ng & Xu, 2012). Thus it was able to measure
negative pore-water pressure of up to 480 kPa, close to the
AEV of the ceramic tip. 100
Considering that the Hall effect transducers and the
suction probe were exposed to elevated temperatures, they
80
were carefully calibrated in a temperature-controlled oven at
Percentage passing: %
SOIL TYPE AND SPECIMEN PREPARATION Table 1. Index properties of tested silt
The material investigated in this study is a completely
Index test Measured value
decomposed tuff (CDT) sampled from Hong Kong. It is
yellowish-brown, slightly plastic and contains very small Standard compaction tests
percentages of fine sand and coarse sand. Fig. 3 shows the Maximum dry density: kg/m3 1760
particle size distribution as determined by sieve and hydro- Optimum water content: % 16 .3
meter analyses (BSI, 1990). In addition, the physical proper- Grain size distribution
ties are summarised in Table 1. According to the Unified Percentage of sand: % 24
Soil Classification System, CDT is classified as silt (ML) Percentage of silt: % 72
(ASTM, 2006). Percentage of clay: % 4
Previous studies have shown that the mechanical and Specific gravity 2 .73
Atterberg limits (grain size , 425 ìm)
hydraulic behaviours of unsaturated soils are fully coupled
Liquid limit: % 43
(e.g. Wheeler et al., 2003). To study the hydraulic behaviour Plastic limit: % 29
of CDT, the water retention curve of compacted CDT was Plasticity index: % 14
measured using a triaxial pressure plate system (Ng et al.,
712 NG AND ZHOU
2013). The soil specimen tested was prepared using the respectively. The average initial suction of the specimens
identical method adopted in this current study. It was 70 mm after compaction was 95 kPa as measured by a high-capacity
in diameter and 35 mm high. The initial water content and suction probe (Ng & Xu, 2012). The variations in measured
initial dry density prepared were 16 .3% and 1 .73 g/cm3, initial suction at different heights of each specimen and the
respectively. After compaction and set-up in the pressure variations in measured initial suction among different speci-
plate system, the soil specimen was first wetted to zero mens were all less than 2 kPa.
suction along a scanning path. Then, the specimen was
subjected to a cycle of drying and wetting. Fig 4 shows the
measured drying and wetting curves. On drying the soil TEST PROGRAMME AND PROCEDURES
specimen, it is expected to follow a primary drying path. Seven suction- and temperature-controlled cyclic triaxial
Owing to the limited suction range applied, however, it is tests were carried out in this study. Three of them (S0T20,
very likely that the residual condition was not reached. On S30T20, S60T20) were carried out at 208C, but at different
the subsequent wetting, the soil should follow a scanning suctions (0, 30 and 60 kPa). The other four tests (S0T40,
wetting path, not the primary one. The figure reveals that S0T60, S30T40 and S60T40) were conducted at elevated
along the drying path, volumetric water content decreases temperatures (408C and 608C). Each test consisted of three
non-linearly as suction increases. The AEV is estimated to stages: suction equalisation, thermal equalisation and cyclic
be 60 kPa. Along the wetting path, adsorption rate changes loading–unloading. Suction equalisation took 1–2 weeks in
abruptly at a suction of about 30 kPa. Although the water the suction range from 0 to 60 kPa. The duration of thermal
retention curve was measured only at 208C, thermal effects equalisation was 3–5 days, depending on suction and tem-
on the curve may be estimated based on previous experi- perature conditions. Details of the test programme are given
mental studies, where it has been found that the water in Table 2. It should be noted that there is no test at 608C at
retention capacity of unsaturated soil reduces slightly at suctions of 30 and 60 kPa in the current test programme.
higher temperatures (Romero et al., 2001). This is because according to previous experimental studies
Each specimen, about 76 mm in diameter and 152 mm on unsaturated soils (for example, Tang et al., 2008; Uchai-
high, was statically compacted in ten layers in the current pichat & Khalili, 2009), thermal effects on mechanical be-
study. The compaction water content was about 16 .3%, haviour such as yield stress and shear stress–strain
corresponding to the optimum water content as determined relationship follow the same trend in the temperature range
in the standard Proctor compaction test (BSI, 1990). To from 20 to 608C. It is expected that testing soil specimens at
minimise excessive compaction of the lower layers during 608C with suctions of 30 and 60 kPa would not alter the key
compaction of the upper layers, the process of compaction conclusions of the current study. Given that suction- and
was stress-controlled with a maximum compaction pressure temperature-controlled tests are very time consuming. Each
of 1200 kPa. After sample compaction, the initial dry density test takes about 20 days, as shown in Table 2. It was
of each soil specimen was found to be about 1 .73 g/cm3, therefore decided that tests at 608C with suctions of 30 and
which corresponds to a dry density ratio of 98%. The height 60 kPa would not be carried out for the sake of complete-
and diameter of the specimen were measured by a calliper ness in this study.
(readable to 0 .01 mm) and a Pi tape (readable to 0 .01 mm), Figure 5 shows the thermo-hydro-mechanical path of each
specimen in the first two stages. After compaction, each
specimen was set up in the triaxial system. The initial stress
0·45
AEV: 60 kPa
state of each specimen was controlled at point A. Each
specimen was first isotropically compressed to a net confin-
ing stress of 30 kPa at constant water content (A!B).
Volumetric water content, θ
0·35
According to AASHTO (2003) for resilient modulus tests on
Drying path subgrade soil, the confining pressure applied should be in
the range from 2 to 6 psi (13 .8 to 41 .4 kPa). A single
0·25
pressure of 30 kPa, which is close to the mean value (4 psi),
is therefore considered and adopted. It should be pointed out
30 kPa
that effects of net confining pressure on cyclic behaviour are
0·15 Wetting path
not investigated in this study. This is because they have been
widely investigated in previous studies (for example, Sivaku-
mar et al., 2013). After isotropic compression, each soil
0·05 specimen was then wetted by decreasing suction from
0·1 1 10 100 1000
Suction, s: kPa 95 kPa to 0, 30 and 60 kPa (B!E, B!D and B!C). To
control soil suction, predefined ua and uw were applied from
Fig. 4. Water retention curve of a recompacted specimen at 208C the top and bottom, respectively. Then the soil specimen was
Specimen identity Matric suction: Temperature: Initial water Initial void S r (ua uw ) after Suction Thermal
kPa 8C content: % ratio suction and thermal equalisation equalisation
equalisation duration: days duration: days
S0T20 0 20 16 .6 0 .57 0 12 0
S30T20 30 20 16 .6 0 .59 26 7 0
S60T20 60 20 16 .5 0 .56 50 4 0
S0T40 0 40 16 .0 0 .56 0 13 5
S0T60 0 60 16 .4 0 .56 0 14 6
S30T40 30 40 16 .4 0 .58 26 8 4
S60T40 60 40 16 .4 0 .58 50 14 3
CYCLIC BEHAVIOUR OF AN UNSATURATED SILT AT VARIOUS SUCTIONS AND TEMPERATURES 713
Test Suction and
ID thermal paths
S0T20 A→B→E
S30T20 A→B→D
Application of cyclic
loading-unloading S60T20 A→B→C
60
C1
S60T40 A→B→C→C1
95 A B
LC curve
(40°C)
D1
40 E1
60 C
LY curve
LC curve
30 (s ⫽ 0)
D (20°C)
E
20
0 30 105 120
Net confining pressure: kPa
subjected to suction equalisation. The equalisation was con- surface with temperature is described by the position of the
sidered to be completed when the water flow of the soil loading yield (LY) curve. It is obvious that stress paths
specimen was less than 0 .5 cm3 within 24 h, corresponding during the first two stages are all inside the LC curves.
to a rate of water content change of less than 0 .04% per day Consequently, the volume change induced by wetting and
(Sivakumar, 1993). After suction equalisation, suction heating was quite small (less than 0 .1%). On the contrary,
throughout the soil specimen was equal to the difference Cekerevac & Laloui (2010) heated soil specimens at nor-
between predefined ua and uw. mally consolidated state prior to application of cyclic load-
The second stage of each test was thermal equalisation. ing–unloading. Contractive volume change of up to 1% was
For the three tests carried out at room temperature (208C) measured under thermal loadings. This is one of the key
(S0T20, S30T20, S60T20), it was not necessary to change differences between their study and the present one. The
the soil temperature. For the four tests carried out at influence of this difference on cyclic soil behaviour is
elevated temperatures, however, three specimens (S0T40, discussed later.
S30T40 and S60T40) were heated to 408C (E!E1, D!D1 After equalisation of both suction and temperature, a
and C!C1), while a fourth (S0T60) was heated to 608C cyclic triaxial test was carried out on each specimen to
(E!E2). It should be noted that thermal loading may induce investigate soil deformation under cyclic loads. Cyclic devia-
excess pore-water pressure. In this study, 3–5 days were tor stress in haversine form was applied, while net confining
needed for the dissipation of excess pore-water pressure, pressure was maintained constant at 30 kPa. At each suction
depending on suction and temperature conditions. With such and temperature condition, 100 cycles were applied at a
a long duration, it was sufficient for each soil specimen to frequency of 1 Hz. Fig. 6 shows variations in deviator stress
reach the predefined temperature and suction conditions. during the first and last ten cycles of a typical test. The
Ng & Yung (2008) carried out a series of isotropic com- difference between maximum and minimum applied deviator
pression tests on compacted CDT specimens. The initial dry stresses is defined as cyclic stress qcyc : With references to
density of soil specimens in their study is 1 .68 g/cm3, which AASHTO (2003) standard for resilient modulus tests, four
is about 3% smaller than that used in the current test levels of cyclic stress (30, 40, 55 and 70 kPa) were applied
programme. The compaction water contents in both studies to each specimen in succession. Although the standard
are 16 .3%. Considering that these two studies adopt very procedure was adopted, it is possible that soil response
close initial density and water content, experimental results would be affected by previous cyclic stress levels. This is
reported by Ng & Yung are used to assist the interpretations because soil behaviour is dependent on stress history. Brown
of current data. At suctions of 0, 50 and 100 kPa, the yield & Hyde (1975) investigated effects of stress history on
stresses of compacted specimens were found to be about resilient modulus and plastic axial strain accumulation. They
120, 170 and 200 kPa, respectively. Based on these results, found that resilient modulus is insensitive to previous cyclic
the loading collapse (LC) curve at 208C was determined and stress level, provided that applied stress is low enough to
is shown in Fig. 5. It can be seen that yield stress increases prevent substantial axial plastic strain. The rate of plastic
with increasing suction (suction hardening). In addition, an strain accumulation is reduced by the application of previous
isotropic compression test was carried out at zero suction cyclic stress. To minimise effects of stress history on meas-
and 408C in this study. The corresponding yield stress was ured cyclic behaviour, the magnitude of cyclic stress is
about 105 kPa. Assuming that the shape of the LC curve is larger than previous cyclic stress levels. During cyclic load-
independent of temperature, the LC curve at 408C was ing–unloading, the condition of constant water content was
deduced and is also shown in the figure. It is revealed that simulated for subgrade soil because the dissipation rate of
yield stress decreases with an increase in temperature at a excess pore-water pressure was relatively low compared to
given suction (thermal softening). The evolution of yield the rate of repeated traffic loads in the field. In each test,
714 NG AND ZHOU
120 100 Deviator stress 65 70
0·5 s Pore-water pressure (base)
Pore-water pressure (mid-height)
Deviator stress 80 0·6 s 60
0·7 s
uw: kPa
60 55
q: kPa
Pore-water pressure (base)
100 65
40 50
Pore-water pressure (mid-height)
20 45
...
60 55
40 50
20 45
...
0 40
0 2 4 6 8 10 ... 90 92 94 96 98 100
Number of cycles
Fig. 6. Applied deviator stress and pore-water pressure measured at the base and mid-height of a
specimen during a cyclic triaxial test
the drainage valve for water was closed and pore-water loads, two parameters are considered and used, as shown in
pressure was measured at the base and mid-height of soil the figure. The first parameter is axial plastic strain (åpa )
specimen. accumulated during 100 cycles of loading–unloading. For
the test shown in the figure, åpa is equal to about 0 .3%. The
other parameter is resilient modulus (M R ), which is equiva-
INTERPRETATION OF EXPERIMENTAL RESULTS lent to secant Young’s modulus along the unloading path. As
Typical cyclic behaviour of unsaturated soil summarised by Brown (1996), åpa and M R are important
Figure 6 shows measured variations in pore-water pressure parameters for analysing cracking and rutting failures of a
at the base and mid-height of a specimen during a typical pavement.
cyclic loading–unloading test. For clarity, only the first and Figure 7(b) shows volumetric strain measured during
last ten cycles are illustrated. It can be seen that pore-water cyclic loading–unloading. It can be seen that during the first
pressures measured at the base and mid-height both vary ten cycles, contractive volumetric strain accumulates as the
with the applied deviator stress but with different magni- number of cycles increases, but at a decreasing rate. The
tudes. The magnitude of variation is about 10 kPa and 5 kPa contractive volume change during cyclic loading–unloading
at the base and at the mid-height, respectively. The larger induces soil densification and strain hardening. According to
oscillation of pore-water pressures at the base can likely be the theory of elastoplasticity (Alonso et al., 1990), the
attributed to the cyclic stress applied there. It is thus preconsolidation pressure of soil specimen would increase
expected that the pore-water pressure response should be owing to the occurrence of strain hardening. The LC curve
faster and larger at the base than that at the mid-height. This shown in Fig. 5 would shift to the right. During the last ten
is illustrated by pore-water pressure responses measured cycles, the soil response becomes almost reversible. This is
during the first two cycles as shown in an insert of the consistent with the measurement of axial strain, as shown in
figure. Compared to the applied cyclic stress, pore-water Fig. 7(a). Lackenby et al. (2007) carried out a series of
pressure measurement delays for about 0 .1 s and 0 .2 s at the cyclic triaxial tests on dry sand. They found that accumula-
base and mid-height, respectively. The relatively small varia- tion of axial strain is not obvious after 10 000 cycles, while
tions in pore-water pressure suggest that changes in suction volumetric strain accumulates continuously even after
under cyclic loading–unloading may be negligible in this 1 000 000 cycles. Through drained cyclic triaxial tests on
study. loose, saturated Dogs Bay sand, Lopez-Querol & Coop
Figure 7(a) shows the stress–strain relationship measured (2012) found that soil volume reaches equilibrium within
during a typical 100-cycle triaxial test (S0T40: ó 3 ua ¼ 100–1000 cycles. Compared to these previous studies, the
30 kPa, ua uw ¼ 0, qcyc ¼ 30 kPa and T ¼ 40 8C). For number of cycles required to reach a stable response in this
clarity, only results from the first and last ten cycles are study is much smaller. This may be because the CDT speci-
shown. During the first ten cycles, axial plastic strain mens in this study were overconsolidated to a large extent,
accumulates as the number of cycles increases, but at a as shown in Fig. 5, and hence more resistant to cyclic
decreasing rate. Furthermore, axial plastic strain induced by shearing.
the first cycle is much more significant than each incremen-
tal plastic axial strain induced by any subsequent cycle.
During the last ten cycles, there is no obvious additional Suction effects on axial plastic strain accumulation
accumulation of axial plastic strain. This implies that the Figure 8 shows the relationship between åpa and qcyc : It
soil specimen has reached a stable resilient response. can be seen that åpa increases with an increase in qcyc at all
To quantitatively investigate soil deformation under cyclic suction and temperature conditions, but at different rates. At
CYCLIC BEHAVIOUR OF AN UNSATURATED SILT AT VARIOUS SUCTIONS AND TEMPERATURES 715
40
First ten cycles
30
Deviator stress, q: kPa
20
MR
10
ε pa
0
0 0·1 0·2 0·3 0·4
Axial strain, εa: %
(a)
40
First ten cycles
30
Deviator stress, q: kPa
20
10
0
0 0·05 0·10 0·15
Volumetric strain, εv: %
(b)
4·0 S0T20 S30T20 S60T20 zero suction, measured åpa increases exponentially with an
S0T40 S30T40 S60T40 increase in qcyc : At suctions of 30 and 60 kPa, the relation-
S0T60
ship between qcyc and åpa is almost linear. Furthermore, at a
3·0 given stress and temperature condition, åpa is consistently
lower at higher suction. The decrease in axial plastic strain
Axial plastic strain: %
specimen in the drained condition does not alter the pre- 1·3 S0T60
consolidation pressure and OCR. This is because effects of
thermal softening could be compensated by effects of strain 1·2
hardening. Cekerevac & Laloui (2010) measured contractive
volumetric strains of up to 1% when normally consolidated 1·1
kaolin clay was heated from 22 to 908C. Thermal loads
densified the soil specimen. At the subsequent stage of
cyclic loading–unloading, a smaller axial plastic strain was 1·0
induced on the heated specimen than on the unheated
specimen. 0·9
It is revealed in Fig. 8 that thermal effects on accumulated 0 20 40 60 80 100
axial plastic strain are not obvious at a suction of 60 kPa. Number of cycles, N
(b)
The OCR of the soil specimen is up to about 6 and axial
plastic strain induced by cyclic loads is very small. This is Fig. 9. Influence of number of cycles on resilient modulus: (a) at
probably because the preconsolidation pressure increases sig- zero suction; (b) at cyclic stress of 70 kPa
CYCLIC BEHAVIOUR OF AN UNSATURATED SILT AT VARIOUS SUCTIONS AND TEMPERATURES 717
continuously with N : With 100 cycles, the increase in may be due to the fact that temperature could affect M R
M NR =M 1R is up to 20%. On the other hand, soil response at through different mechanisms. On the one hand, the electri-
608C is qualitatively similar to that at 208C. However, the cal force between soil particles becomes relevant for fine-
increase in M NR =M 1R with N is larger at 608C than at 208C. At grained soil. Based on Derjaguin–Landau–Verwey–Overbeek
cyclic stresses of 30 and 40 kPa, M NR =M 1R is larger at 608C (DLVO) double layer theory (Israelachvili, 2011), electric
than at 208C by about 10%. At cyclic stresses of 55 and repulsive forces between soil particles increase as tempera-
70 kPa, M NR =M 1R is larger at 608C than at 208C by only about ture increases. This would reduce inter-particle forces and
5%. More discussion on the thermal effects will be given in hence cause a decrease in soil stiffness (Santamarina et al.,
the next section. 2001). On the other hand, as illustrated in Fig. 9(a), the
The influence of N on M NR =M 1R is at least partially attrib- increase in M NR =M 1R with the number of cycles is larger at
uted to the accumulation of contractive volumetric strain, as 608C than at 208C by about 5%. This suggests that effects of
shown in Fig. 7(b). Contractive behaviour induces soil soil densification on M R are more significant at higher
densification and an increase in M NR =M 1R : Furthermore, it is temperature. These two mechanisms appear to produce op-
illustrated in Fig. 8 that the threshold qcyc is between 40 and posite effects on M R :
55 kPa. When qcyc is below the threshold value (30 and Figure 9(a) reveals that thermal effects on M NR =M 1R at
40 kPa), accumulated axial plastic strain is low so that the equilibrium are dependent on cyclic stress. At cyclic stresses
increase in M NR =M 1R is small. When qcyc is above the thresh- of 30 and 40 kPa, M NR =M 1R is larger at 608C than at 208C by
old value (55 and 70 kPa), the increase in M NR =M 1R becomes about 10%. At cyclic stresses of 55 and 70 kPa, M NR =M 1R is
significant due to a much higher axial plastic strain. larger at 608C than at 208C by only about 5%. Therefore, at
Figure 9(b) shows the relationship between M NR =M 1R and N low cyclic stress (30 and 40 kPa), the second mechanism is
at cyclic stress of 70 kPa for different suction and temperature more significant. Measured M R increases with increasing
conditions. At 208C, M NR =M 1R increases with N by up to 20% at temperature. At high cyclic stress (55 and 70 kPa), the
zero suction, but it remains almost constant at suctions of 30 second mechanism is relatively less important. Soil behav-
and 60 kPa. This is because axial plastic deformation is much iour is likely dominated by the first mechanism and in this
smaller at suctions of 30 and 60 kPa than at zero suction, as case measured M R decreases as temperature increases.
illustrated in Fig. 8. On the other hand, as temperature in-
creases from 20 to 608C, the change in M NR =M 1R is less than
5%. Compared to the suction effect, thermal effects on Influence of temperature on resilient modulus at different
M NR =M 1R are not so obvious. The relationship between M NR =M 1R suctions
and N is almost independent of temperature at suction of Figure 11 shows influence of temperature on M R at differ-
60 kPa. This observation is consistent with findings from Fig. ent suctions. Two different responses can be identified. At
8. At suction of 60 kPa, temperature seems to have no effect on zero suction, the difference in M R induced by temperature is
plastic strain accumulation and soil densification induced by less than 5 MPa. At suctions of 30 and 60 kPa, M R measured
cyclic loads. The variation of resilient modulus with number of at 408C is smaller than that measured at 208C by up to
cycles, which mainly results from soil densification, would be 10 MPa. This illustrates that thermal effects on M R are more
therefore very similar at various temperatures. significant at suctions of 30 and 60 kPa than at zero suction.
Figure 9 illustrates that M NR =M 1R generally achieves Through a series of suction- and temperature-controlled triax-
a constant value within 100 cycles of loading–unloading. ial tests, Uchaipichat & Khalili (2009) observed that, for
This is consistent with the observations from Fig. 7 that soil overconsolidated soil, the yield surface shrinks with an in-
response is almost reversible during the last ten cycles. crease in temperature. Moreover, the size of the yield surface
is more sensitive to temperature at higher matric suction. The
observations from their study and the current one suggest that
Influence of temperature on resilient modulus at zero suction thermal effects on mechanical behaviour are more significant
Following AASHTO (2003) standard for resilient modulus at unsaturated state than at saturated state. This may be
tests, resilient moduli M R during the last five cycles were attributed to the temperature dependency of surface tension.
determined for each stress level. Fig. 10 shows influence of The coefficient of surface tensions at temperatures of 20, 40
temperature on M R at zero suction. As temperature increases and 608C are 72 .8, 69 .6 and 66 .2 mN/m, respectively (Ng &
from 20 to 608C, M R increases by about 5% at cyclic Menzies, 2007). A decrease in surface tension affects unsatu-
stresses of 30 and 40 kPa, but decreases by about 5% at rated soil behaviour in at least two ways. First, as temperature
cyclic stresses of 55 and 70 kPa. The two different responses increases, the water retention ability of unsaturated soil
35 100
S0T20
S0T40
Resilient modulus, MR: MPa
80 S0T60
Resilient modulus, MR: MPa
30
S30T20
60 S30T40
25 S60T20
S60T40
40
qcyc ⫽ 30 kPa
20 qcyc ⫽ 40 kPa
20
qcyc ⫽ 55 kPa
qcyc ⫽ 70 kPa
15 0
0 20 40 60 80 0 20 40 60 80 100
Temperature, T: °C Cyclic stress, qcyc: kPa
Fig. 10. Influence of temperature on resilient modulus at zero Fig. 11. Influence of temperature on resilient modulus at different
suction suctions
718 NG AND ZHOU
100
slightly decreases and hence the degree of saturation at a
given suction becomes smaller (Romero et al., 2001). A