You are on page 1of 17

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

net/publication/260061662

Model for prediction of resilient modulus incorporating matric suction for


recycled unbound granular materials

Article  in  Canadian Geotechnical Journal · November 2013


DOI: 10.1139/cgj-2012-0406

CITATIONS READS

52 1,172

3 authors:

Abdelhalim Moawad Azam Donald Anthony Cameron


Mansoura University University of South Australia
33 PUBLICATIONS   282 CITATIONS    64 PUBLICATIONS   759 CITATIONS   

SEE PROFILE SEE PROFILE

Md Mizanur Rahman
University of South Australia
155 PUBLICATIONS   1,818 CITATIONS   

SEE PROFILE

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

Expansive clay soils, soil suction and ground movement prediction View project

Model for prediction of resilient modulus incorporating matric suction for recycled unbound granular materials View project

All content following this page was uploaded by Abdelhalim Moawad Azam on 23 November 2015.

The user has requested enhancement of the downloaded file.


1143

ARTICLE
Model for prediction of resilient modulus incorporating matric
suction for recycled unbound granular materials
A.M. Azam, D.A. Cameron, and M.M. Rahman
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Abstract: This paper presents experimental results on the effect of matric suction on the resilient modulus of four recycled
unbound granular materials. The recycled materials were prepared at moisture contents ranging between 70% and 90% of
optimum moisture content (OMC) and tested in a repeated load triaxial test (RLTT) apparatus under various stress regimes.
Soil-water characteristic curves (SWCC) were established for each material by preparing samples at various moisture contents
and measuring matric suction with filter papers. To obtain the wet end of the SWCC, further samples were conditioned on a
tension plate at suctions controlled by the hanging water column method. Some published models for prediction of resilient
modulus were applied to the experimental data, but the correlations were unsatisfactory generally, and so an improved model
was sought. Subsequently, a model with four terms and six constants was developed, which followed the general power law. A
single set of material constants was found for all recycled materials to provide satisfactory predictions of resilient modulus (R2 =
0.88), over a wide range of stresses and moisture states.

Key words: recycled aggregates, matric suction, soil-water characteristic curve, resilient modulus, predictive model.

Résumé : Cet article présente les résultats expérimentaux de l’effet de la succion matricielle sur le module de résilience de
quatre matériaux granulaires lâches recyclés. Les matériaux recyclés ont été préparés à des teneurs en eau variant de 70 à 90 %
de la teneur en eau optimale (TEO) et ont été testés dans un appareil d’essai triaxial à charge répétée (ETCR) sous différents
For personal use only.

régimes de contraintes. Les courbes de rétention d’eau (CRE) ont été établies pour chaque matériau en préparant des échantillons
à différentes teneurs en eau et en mesurant la succion matricielle avec des papiers filtres. Afin d’obtenir la portion humide de la
CRE, d’autres échantillons ont été conditionnés sur une plaque à tension à des succions contrôlées par la méthode de la colonne
d’eau suspendue. Certains modèles de prédiction du module de résilience publiés ont été appliqués aux résultats expérimentaux,
mais les corrélations étaient généralement insatisfaisante, alors un modèle amélioré a été recherché. Par la suite, un modèle
comportant quatre termes et six constantes a été développé, qui est conforme à la loi de puissance générale. Un seul ensemble
de constantes de matériaux a été déterminé pour tous les matériaux recyclés afin d’obtenir des prédictions satisfaisantes du
module de résilience (R2 = 0,88), et ce, pour une variété de contraintes et d’états d’humidité. [Traduit par la Rédaction]

Mots-clés : agrégats recyclés, succion matricielle, courbe de rétention d’eau, module de résilience, modèle prédictif.

Introduction affected by the compressive strength of the original concrete, the


Resilient modulus is an important engineering parameter for amount of softer material in the RCA, and the aggregate shape.
the characterization of unbound granular materials (UGM) as it is Many predictive models have been developed, such as the
used in mechanistic pavement thickness design, for example, in K–Theta (K–␪) model (Hicks and Monismith 1971), the modified
the 2002 Design guide of the American Association of State Highway and universal model (AASHTO 2002), and the models of May and
Transportation Officials (AASHTO 2002). The resilient modulus is an Witczak (1981), Gupta et al. (2007), Liang et al. (2008), and Cary and
indication of the elastic behaviour of unbound pavement materi- Zapata (2011). Those models were based on the behavior of sub-
als and soils under repeated traffic loads. grade soils and (or) virgin aggregate, not recycled products, with
The resilient modulus depends on various factors, such as their added complexity. A few of the models referred to include
deviator stress, confining stress, density, and moisture content. the effect of matric suction on resilient modulus and have been
George (2004) and Mohammad et al. (1999) reported that the most shown to be successful, but how practical they are to implement
influential soil index properties with respect to resilient modulus can be an issue. So the challenges are to develop a model for
were moisture content, degree of saturation, material passing
recycled materials, which preferably incorporates soil suction in a
#200 sieve, liquid limit, plasticity index, and density. Witczak
simple but effective way.
et al. (2000) studied the influence on the resilient modulus of
Thus, the paper sets out to review previously published models
cohesive soils of changes of relative moisture content either side
of optimum moisture content (OMC) for material compacted at for resilient modulus (Mr) and the effect of matric suction. Exper-
maximum dry density (MDD). It was observed that even low vari- imental data from repeated load triaxial test (RLTT) studies on
ations in moisture content resulted in significant changes in re- four unbound recycled granular materials are presented, and the
silient modulus. influence of initial matric suction on material behaviour is high-
Nataatmadja and Tan (2001) found that the resilient modulus of lighted. An improved and simplified predictive model for resil-
well-graded recycled (or crushed) concrete aggregate (RCA) can be ient modulus is presented for all these materials. A single set of

Received 6 November 2012. Accepted 2 September 2013.


A.M. Azam, D.A. Cameron, and M.M. Rahman. School of Natural and Built Environments, University of South Australia, Australia.
Corresponding author: D.A. Cameron (e-mail: donald.cameron@unisa.edu.au).

Can. Geotech. J. 50: 1143–1158 (2013) dx.doi.org/10.1139/cgj-2012-0406 Published at www.nrcresearchpress.com/cgj on 26 September 2013.
1144 Can. Geotech. J. Vol. 50, 2013

constants and exponents was found to be adequate for the drying a saturated sample or by wetting a dry sample; the former
materials tested. is generally preferred.
Walker (1997) reported that significant matric suctions can be de-
Literature review veloped in a UGM. Walker (1997) measured both matric and total
Models for prediction of resilient modulus suctions for three unbound pavement materials (crushed basalt
Several constitutive models for modeling the resilient modulus base, sandstone base, and clay soil as subgrade) taken from two ex-
of soils have been proposed in the past. Hicks and Monismith perimental pavements. The materials were compacted by both stan-
(1971) found from their studies that stress level is the most influ- dard and modified compactive effort and suctions were measured by
encing factor on the resilient modulus of granular pavement the filter paper (FP) method. Matric suction was found to range from
materials. The authors proposed the simple K–␪ model, which 0 to 45 kPa for crushed basalt, from 0 to 3500 kPa for sandstone, and
includes only bulk stress from 500 to 9500 kPa for clay material. The matric suctions were
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

recorded at OMC and were found to be 5, 400, and 650 kPa for
(1) Mr ⫽ k1␪k2 crushed basalt, sandstone, and clay subgrade, respectively. Therefore
it may be concluded that matric suction exists within an UGM, albeit
of a lower level than in clay, which will contribute to strength and
where Mr is resilient modulus (MPa), ␪ is bulk stress (= ␴1 + 2␴3 stiffness (Heath et al. 2004; Gupta et al. 2007).
(MPa)), ␴1 is major principal stress, ␴3 is confining pressure, and k1 Nokkaew et al. (2012) measured the matric suction of three
and k2 are the regression constant and exponent, respectively. recycled asphalt pavements (RAP) and three RCA base course prod-
May and Witczak (1981) proposed a power equation (eq. (2), ucts. The authors used a hanging water column test to determine
hereafter referred to as the May and Witczak model) that relied on
the SWCC by progressive drying. The measured matric suctions
normalized measures of the mean stress and the octahedral shear
ranged from 0.05 to 75 kPa for the recycled products.
stress as follows:
Several empirical equations have been proposed to fit suction –

冉 冊冉 冊
moisture content data to an SWCC (e.g., van Genuchten (1980);
␴m k1 ␶oct k2
(2) Mr ⫽ K0 Fredlund and Xing (1994)). These equations rely firstly on the sat-
pa ␶ref urated water content of the soil. Houston et al. (2006) provided
SWCC fitting parameters for both plastic and nonplastic soils
where pa is atmospheric pressure, ␴m is mean normal stress (= (␴1 + based on the particle size distribution and the soil plasticity. The
2␴3)/3); ␶oct is octahedral shear stress (= 共兹2/3兲共␴1 ⫺ ␴3兲); ␶ref is refer- basic equation for the SWCC was an extension of that proposed by
For personal use only.

ence shear stress (= 兹2/3qf); qf is peak shear strength (= d + ␴d tan␤); Fredlund and Xing (1994).
d and ␤ are Drucker–Prager failure parameters; and K0 is a regression In 2012, Gould et al. (2012) introduced a new model for an SWCC
parameter. (hereafter referred to as the Gould et al. model) that was built on the
The mechanistic empirical pavement design guide proposed a characteristic features of the SWCC, as shown in Fig. 1, which in-
variation of eq. (3) as the modified universal model (AASHTO 2002) cluded the
for prediction of resilient modulus. The universal model is:
• air-entry value (uae);

冉p␪ 冊 冉 冊
• residual suction (ur);
k2 ␶oct k3
(3) Mr ⫽ K1pa ⫹1 • minimum suction (umin);
a pa • maximum suction (umax);
• saturated gravimetric water content (wsat);
where K1 and k3 are regression parameters or exponents. • slope of the initial portion of the curve (m);
It should be noted at this point that not one of these models • slope of the central portion (n); and
incorporated the moisture state of the UGM. The importance of • slope of the final portion of curve (o).
moisture state on the performance of UGMs is discussed in the
next section. In addition, two fitting parameters were required to consider
the curvature of the SWCC at uae and ur, (i.e., ␾ae and ␾r).
Soil suction in UGM and soil-water characteristic curve The Gould et al. model is
(SWCC) models
m⫺n ∗
In Australia, UGM of a pavement are placed generally well
above soil water tables, and tend to remain unsaturated while in
(4) w(u∗) ⫽ ⫺
␾ae␲
关f(u ) ⫺ f共umin

兲兴 ⫺ n␾⫺␲o 关g(u∗) ⫺ g共umin

兲兴
r
service. The change in total suction is usually represented by the m⫹n ∗ m⫹o ∗
change in matric suction as the solute suction does not seem to be ⫹ ␺ ⫺ umin ⫹ wsat
2 2
sensitive to the changes in the soil water content, according to
Fredlund and Rahardjo (1993) and Khoury et al. (2003). In the
longer term and with saline soils this may not be the case. where
Nonetheless, recent developments in the theory of unsaturated
soil behavior have concentrated on matric suction. Although (5) u∗ ⫽ log10(u), ∗
uae ⫽ log10(uae), and ur∗ ⫽ log10(ur)
Kodikara (2012) recognized the importance of matric suction on
the volume change of compacted soils, their concern about the (6) f(x) ⫽ ␾ae共x ⫺ uae

兲 tan⫺1关␾ae共x ⫺ uae∗ 兲兴
hysteresis developed between wetting and drying of soils led 1
them to propose an alternative parameter, moisture ratio, ew, ⫺ ln兵1 ⫹ 关␾ae共x ⫺ uae
∗ 2
兲兴 其
2
which is the product of specific gravity and gravimetric moisture
content. Validation of Kodikara’s proposed new model is needed. (7) g(x) ⫽ ␾r共x ⫺ ur∗兲 tan⫺1关␾r共x ⫺ ur∗兲兴
Many researchers (e.g., Fredlund et al. 1994; Vanapalli et al.
1
1996, 1998; Sawangsuriya et al. 2008) have expounded the impor- ⫺ ln兵1 ⫹ 关␾r共x ⫺ ur∗兲兴2其
tance of the relationship between matric suction and moisture 2
content (gravimetric or volumetric) on soil behavior. This rela-
tionship is presented by a SWCC. The SWCC can be established by where x is matric suction (= um).

Published by NRC Research Press


Azam et al. 1145

Fig. 1. Characteristic features of SWCC (adapted from Gould et al. 2012).


Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

In summary, an UGM can have significant suction and a SWCC where (ua – uw)b (referred to as uae previously) is the air-entry value
may be established for the UGM. The next section discusses the of matric suction, at which point when drying the soil, air starts to
influence of suction and how it has been incorporated into pre- enter the largest pores in the soil matrix.
For personal use only.

dictive models for resilient modulus. Gupta et al. (2007) studied the effect of soil suction on the resil-
ient modulus of subgrade soils. The authors found that the rela-
Previous resilient modulus models incorporating soil
tionship between resilient modulus and matric suction was a
suction
It has been shown that resilient modulus of subgrade soils and power function. A new model was developed (referred to hereaf-
UGMs increases with increase of matric suction (or decrease of ter as the Gupta et al. model) as follows:
moisture content) (Craciun and Lo (2010) and Azam and Cameron
(2013a), for UGMs; Cary and Zapata (2011) for UGMs and subgrades;
Brown (1996), Parreira and Goncalves (2000), Khoury and Zaman
(11) Mr ⫽ K1pa 冉 ␴b ⫺ 3K6
pa 冊冉
k2 ␶oct
pa
⫹ K7 冊 k3
⫹ ␣1(ua ⫺ uw)␤1
(2004), Yang et al. (2005), Gupta et al. (2007), and Liang et al. (2008),
for subgrades). Correlations have been developed by most of these
researchers between resilient modulus and matric suction. where ␴b is bulk stress (= ␪ (kPa)), ua – uw is matric suction (kPa), K6
Fredlund et al. (1977) attempted to relate the resilient modulus and K7 are regression parameters (K6 ≤ 0 and K7 ≥ 1), k1 ≥ 0, k2 ≥ 0,
of compacted subgrades with deviator stress, with the regression and ␣1 and ␤1 are the intercept and slope of the resilient modulus
constants being dependent upon matric suction. Experimental relationship with matric suction (power equation), respectively.
data indicated good correlation between the observed and esti- The constants ␣1 and ␤1 were derived from the resilient modu-
mated resilient moduli. Equation (8) presents this early Fredlund lus data for samples prepared over a range of suction for each
et al. model as follows: material and conducted with a bulk stress of 83 kPa and an octa-
hedral shear stress of 19 kPa as recommended by the National
(8) log Mr ⫽ cld ⫺ mld(␴1 ⫺ ␴3) Cooperative Highway Research Program (NCHRP 2003; hereafter
referred to as NCHRP 1-28A) for subgrades under typical pavement
where cld and mld are functions of matric suction. stress conditions. Correlations between these parameters and ba-
Yang et al. (2005) explored correlations with either total or ma- sic soil properties were offered, however, they were not statisti-
tric suction for subgrade soils and found that matric suction pro- cally strong.
vided the better correlation. The authors proposed a predictive Liang et al. (2008) also investigated cohesive subgrade soils and
model incorporating deviator stress and matric suction presented a predictive model for resilient modulus (eq. (12)),
which was similar to the universal soil model. The mean stress
(9) Mr ⫽ k1(␴d ⫹ ␹wum)k2 was expanded to almost an effective stress term incorporating
matric suction and ␹w (Khalili and Khabbaz 1998). The accuracy of
the proposed model was validated against experimental data for
where ␹w is Bishop’s effective stress parameter.
A-4 and A-6 soils (silty and clay soils, respectively).
␹w has the same numerical limit as the degree of saturation, as

冉 冊冉 冊
it varies between 0 and 1 (completely dry and saturated, respec-
␪ ⫹ ␹wum k2 ␶oct k3
tively). Yang et al. (2005) adopted the empirical expression for ␹w (12) Mr ⫽ k1pa ⫹1
developed by Khalili and Khabbaz (1998) pa pa

(10) ␹w ⫽ 冋 (ua ⫺ uw)b


ua ⫺ uw 册
0.55
The model of Liang et al. (2008) is attractive as it takes into
account the effect of both stress state and moisture content.

Published by NRC Research Press


1146 Can. Geotech. J. Vol. 50, 2013

Cary and Zapata (2011), proposed a predictive model that incor- Suction measurement and SWCC
porated the effect of seasonal variations in matric suction on re- The FP method according to ASTM (2003a) standard D5298 was
silient modulus, as given in eq. (13). A silty sand subgrade and a used for indirect measurement of matric suction for the blends.
base course were tested to validate the model. The base material Samples of 100 mm diameter were prepared at different levels of
was tested under drained and undrained conditions; however, the moisture content (dry side of OMC) and at a target dry density
subgrade was tested under a drained condition. ratio of 98% MDD.
Two samples were prepared at the same chosen moisture con-

冉 冊冉 冊冉 冊
␪net ⫺ 3⌬uw-sat k2 ␶oct k3 um0 ⫺ ⌬um k4 tent. The sample height was reduced to 100 mm; compaction was
(13) Mr ⫽ k1 pa ⫹1 ⫹1 conducted in two layers, each 50 mm high after compaction. Ini-
pa Pa pa
tial trials were made by placing the filter paper on top of the first
compacted layer, followed by compaction of the second layer.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

where k1 , k2, k3 , and k4 are regression parameters (k1 > 0, k2 > 0, However, it was found that compaction of the second half of the
k3 < 0, and k4 > 0); ␪net is net bulk stress (= (␪ – 3ua), where ua is pore sample caused damage to the filter paper, and so subsequent sam-
air pressure); ⌬uw-sat is build-up of pore-water pressure under sat- ples were compacted without rodding between the two layers to
urated conditions, in such cases ⌬um = 0; umo is initial matric soil facilitate separation of the two halves of the sample and insertion
suction; and ⌬um is relative change of matric soil suction with of filter papers. An air dry filter paper, Whatman No. 1 (85 mm
respect to umo due to build-up of pore-water pressure under un- diameter), was protected by a 90 mm diameter filter paper on each
saturated conditions, in this case ⌬uw-sat = 0. face to form a stack of three filter papers. The stack of papers was
placed in the middle of the sample, and the complete sample was
All these suction-inclusive models have practical limitations. As
subsequently held securely in a tight fitting polyvinyl chloride
already stated, the Gupta et al. (2007) model requires resilient
(PVC) tube. More details of the FP method are available in Azam
modulus testing of samples of the material prepared at different
and Cameron (2013a).
matric suctions under a set stress regime. A limitation of the
The hanging water column technique was conducted according
model of Liang et al. (2008) is that an accurate determination of to ASTM (2003b) standard D6836 to verify the values of matric
matric suction is required for wet soil conditions (low suctions) to suction less than 10 kPa, which is the lower limit of suction for
obtain the air entry value. For the application of the model of Cary filter papers (ASTM (2003a) standard D5298). The hanging column
and Zapata (2011), the pore-air pressure is required to determine (HC) apparatus consisted of four main parts: testing cell, outflow
the net bulk stress and the likely change of matric suction must column, manometer, and the HC as shown in Fig. 2. The testing
also be known because of pore-water pressure build-up during cell contained a low air entry (35 kPa) ceramic plate of 300 mm
For personal use only.

repeated loading. diameter to condition the specimens to the required suction. Sam-
To simplify and possibly to improve these models, the authors ples of dimensions, 72.6 mm diameter and 69 mm height, were
devised a testing program to generate data for unbound recycled prepared in a stainless steel mould at OMC and at 98% MDD by
granular material and subsequently evaluated existing models static compaction in one layer. The samples that were kept in the
before developing a new model. mould were soaked in water for one week before testing to ensure
saturation of the sample.
Materials Thereafter the samples were placed on the ceramic plate of the
Two products from different South Australian suppliers were testing cell and then the upper chamber of the cell was closed. The
blended in this research to form the recycled product; 20 mm RCA reservoirs were filled with water and the plastic tube and the air
base material and 20 mm recycled clay masonry (RCM). The prod- bubbles were removed. Thereafter, the specimens were allowed to
uct suppliers were ResourceCo (RCO) and Adelaide Resource Re- equilibrate for at least 48 h until movement of the air–water in-
covery (ARR). The RCO product was blended in the University terface ceased. After equilibration, the desired suction increment
laboratory at 10%, 20%, and 30% RCM by total aggregate dry mass. was applied by adjusting the elevation of the water reservoirs
The ARR material was delivered as a premixed combination of until the manometer represented the desired suction. Equilib-
RCA and RCM (80% RCA + 20% RCM). rium was achieved when water ceased to flow from the specimen.
The specimen was weighed for determination of the moisture
Testing program and discussion content corresponding to the end suction, and was then returned
to the testing cell for applying the next suction level. At the end of
The experimental program included particle size distribution, the test, the specimen was dried to obtain the final mass.
Atterberg limits, modified Proctor compaction and measurement The suction data from both methods of measurement were
of initial matric suction. In addition, unconsolidated, undrained, combined to form each SWCC. The average of the FP data for each
triaxial shear tests and RLTTs with multiple stress stages per- moisture content preparation was adopted; however, only one
formed. sample for each blend was used to provide the HC data.
The results are presented in Figs. 3 and 4 for the RCO and the
Basic engineering properties
ARR blends, respectively. The legend in these two figures differ-
The particle size distribution of each blend of RCO was com- entiates suction determined by HC and suction measured by FP.
pared with the Department for Transport, Energy, and Infrastruc- The fitting of the suction data in these figures is according to
ture (DPTI 2011) specifications for sub-base material and the Gould et al. (2012; eqs. (4–7)). The regression parameters and the
grading curves were found to be satisfactory. The grading curve of correlation coefficient (R2) for each SWCC are provided in Table 2.
the ARR product was compared with DPTI specifications for base The regression parameters were determined by manual fitting,
material and was found to be within the specified limits. In Table 1 using the soil properties from the experimental data, and the
the basic engineering properties of the blends are compared with parameters were found to fall within the allowable ranges re-
the DPTI specifications. ported by Gould et al. (2012).
The plasticity of fines as described by liquid limit and plasticity The shapes of the SWCCs were as expected, being generally
index generally met DPTI specifications for base material, except the S-shaped. The fitting of the SWCC data for the 30% RCO blend was
liquid limit for the ARR blend, which was one percentage point incomplete as a residual suction was not evident from the data;
higher than the maximum required. The MDD decreased and the consequently the value of residual suction was set equal to the
OMC increased as the percent of RCM increased for the RCO mate- maximum suction (106 kPa) to remove the second bending point
rial, which agreed with the finding of Poon and Chan (2006). from the SWCC, as suggested by Gould et al. (2012).

Published by NRC Research Press


Azam et al. 1147

Table 1. Engineering properties for blends.


Cohesion (kPa) Friction angle (°)
RCM Fines Liquid Plasticity MDD OMC 70% 80% 90% 70% 80% 90%
Blend (%) (%) limit (%) index (%) (t/m3) (%) OMC OMC OMC OMC OMC OMC
DPTI-SA for base — — 25 (max) 6 (max) — — 150 (max) 150 (max) 150 (max) 45 (min) 45 (min) 45 (min)
DPTI-SA for sub-base — — 28 (max) 8 (max) — — 250 (max) 250 (max) 250 (max) 40 (min) 40 (min) 40 (min)
RCO 10 7.0 23.5 2.5 1.88 12.1 81 6 25 41 53 51
20 7.0 23.0 2.5 1.86 12.5 40 46 30 49 44 46
30 7.0 23.0 2.0 1.84 14.2 72 0 32 37 53 45
ARR 20 6.0 26.0 0.0 1.92 11.6 76 54 70 56 56 51
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 2. Hanging water column apparatus.

Upper Reservoir
Outflow
Column

Manometer
For personal use only.

Lower Reservoir

Testing Cell

For samples of the two 20% crushed masonry blends prepared at horizontal line, with the numerical value above or near the point
moulding moisture contents between 60% and 80% relative mois- being the measured suction for that moisture state and dry den-
ture content (RMC), matric suction ranged from 35 to 450 kPa for sity. It should be noted that the zero air void line intersected the
the ARR material, but was less for the RCO blend (22 to 275 kPa). suction data at 2.6 and 4.6 kPa for the 20% RCO and the ARR
The greater plasticity of fines in the ARR product is the probable blends, respectively. The zero air void line (degree of saturation =
reason for this difference in water retention. 1) should intersect the suction data at 1 kPa, which theoretically is
It was expected that with the addition of more crushed masonry considered the minimum. The small difference in measured suc-
tion and theoretical suction values at this wet soil end may be
in the blend, the matric suction would increase because of the
attributed to the inaccuracy of both measuring low suction and
greater plasticity of the RCM fines. This indeed was observed at
determination of specific gravity of mixed soil.
matric suctions greater than 10 kPa for the RCO blends (Fig. 3).
Kodikara (2012) constructed suction contours alongside com-
Unexpectedly however, the 30% RCO soil showed less water retain- paction curves for different compaction energies to develop a
ing capacity at low suction than did the 20% RCO material. constitutive model for compacted soil. Suction contours could be
constructed for each material in the current study (refer to Figs. 5
Suction and compaction
and 6), provided further dry densities were explored.
The matric suction data from the previous section are shown
against the compaction curve of each of the two materials with Static triaxial test
20% RCM in Figs. 5 and 6. All materials were prepared to the same As undrained triaxial testing is recommended by DPTI (2011) for
dry density (98% of MDD) and so the data appear along a single appraisal of recycled products for pavements, unconsolidated–

Published by NRC Research Press


1148 Can. Geotech. J. Vol. 50, 2013

Fig. 3. SWCC for RCO blends.


Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 4. SWCC for ARR blend.


For personal use only.

Table 2. Regression parameters for each blend using the Gould et al. (2012) model.
Blend uae ur umin m n o ␾ae ␾r wsat(%) R2
10% RCO 7 35 1 −0.0001 −0.093 −0.026 3.5 80 14.5 0.996
20% RCO 5 30 1 −0.0085 −0.099 −0.008 15.0 6.5 16.0 0.999
30% RCO 6 106 1 −0.001 −0.105 −0.005 0.8 0.22 15.6 0.992
ARR 5 15 1 −0.0045 −0.099 −0.016 60.0 70.0 14.5 0.997

undrained triaxial shear tests were performed according to deviator stress and initial matric suction. Specimens were sub-
Australian standard AS1289.6.4.1 (Standards Australia 1998) to jected to three confining pressures; 25, 50, and 75 kPa. Owing to
determine total shear strength parameters (apparent cohesion, c; the high shear strength of these granular materials, greater con-
and angle of friction, ␾). Samples 100 mm in diameter and 200 mm fining pressures could not be considered with the available equip-
high were prepared for the triaxial cell at 70%, 80%, and 90% OMC, ment.
and at 98% of MDD. Samples were prepared by static compaction Figure 7 provides stress–strain curves for a confining pressure
in two layers. The 20% blends were tested at two additional mois- of 50 kPa for three different RMC values for the 10% RCO blend. It
ture levels (60% and 100% OMC) to verify the relationship between is evident from this figure that the moulding moisture content of

Published by NRC Research Press


Azam et al. 1149

Fig. 5. Suction data against the compaction curve for 20% RCO blend.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13
For personal use only.

Fig. 6. Suction data against the compaction curve for ARR blend.

Published by NRC Research Press


1150 Can. Geotech. J. Vol. 50, 2013

Fig. 7. Examples of stress–strain curves at different relative moisture contents for 10% RCM (RCO) at 50 kPa confining pressure.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 8. Variations of deviator stress and static Young’s modulus with initial matric suction for 20% blends at 75 kPa confining pressure
For personal use only.

each sample had a significant effect on the peak deviator stress; same units. The correlation coefficient for this power function
increased moisture content decreased the peak stress. was 0.86. This function may be implemented in finite element
The static (monotonic) Young’s modulus was determined from programs, such as AFENA (Carter and Balaam 1995) and will be
the stress–strain graph after making any necessary zero strain used in the future to predict the stresses and strains under static
adjustments. The secant modulus was determined to half of the wheel loading for the UGM.
maximum deviator stress. The variations of the peak deviator
stress and Young’s modulus with initial matric suction are plotted (14) E ⫽ 184␴m
0.85
in Fig. 8; power functions have been fitted to the data. In general,
the stiffness and the strength of the blended materials increased
as the initial matric suction increased. Gupta et al. (2007) also The values of apparent cohesion and friction angle for blends at
found that power law relationships could be successfully applied different levels of moisture content are presented in Table 1. The
to deviator stress and suction data for subgrade soils. maximum allowable cohesion of 150 kPa for a base was met by all
The impact of mean normal stress on Young’s modulus is de- blends. Seventy-five percent of the specimens had friction angles
picted in Fig. 9 for all blends at different confining pressures and meeting the DPTI specification for base products, which was a
moisture contents. As expected, the static Young’s modulus in- minimum value of 45°.
creased as the mean normal stress increased. Again, the relation- The apparent cohesion and friction angle of each blend did not
ship for all the data in this plot can be represented by a power trend consistently with moisture content. However, the shear
function as provided in eq. (14). Modulus and mean stress have the strength for a given normal stress (e.g., 100 kPa) did decrease as

Published by NRC Research Press


Azam et al. 1151

Fig. 9. Young’s modulus against mean normal stress.


Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 10. Apparent cohesion against percentage of optimum moisture content.


For personal use only.

expected (Azam and Cameron 2013b). The reasons for this appar- Repeated loading triaxial test
ent contradiction may be the steepness of the failure envelope of The Australian test method (AustRoads) for resilient modulus eval-
these materials and the small range of applied confining stresses, uation was conducted for the blends (AustRoads 2007). In general,
which could have led to errors in interpretation of the peak the AustRoads procedure is similar to the approach specified by
strength parameters. AASHTO (2007) standard T307-99 in the multistage and the precon-
For material compacted to the same dry density but at different ditioning stage. However, the shape of the cyclic load pulse differs
initial moisture contents, strength will vary largely because of soil between the two approaches. The shape is trapezoidal for AustRoads,
suction differences and not soil structure, so in keeping with and ramping, loading, and unloading occur over a 3 s period. In con-
unsaturated soil theory regarding soil strength (e.g., Khalili and trast, the AASHTO protocol adopted a haversine shaped pulse over a 1 s
Khabbaz 1998), the angle of friction was assumed constant for period. Furthermore, the number of load repetitions applied in the
each blend. Subsequently the average friction angle for each AustRoads method in each stress stage is 200, but AASHTO stipulates
blend was enforced and the cohesion intercept was adjusted ac- just 100 cycles. Lekarp et al. (2000), quoting Hicks (1970), stated that the
cordingly. The resultant values of apparent cohesion are depicted loading period made no difference to the resilient modulus, al-
in Fig. 10. The cohesion is observed to decrease as the moulding though they suggested that saturated samples having low permea-
moisture content increased, generally speaking. bility may be affected if drainage becomes ineffective.
It is evident from Figs. 8, 9, and 10 that both initial moisture Combinations of applied repeated vertical and static confin-
content (or matric suction) and mean normal stress influence the ing stresses were applied over 66 stress stages for each stage to
strength and stiffness of recycled UGM. characterize the vertical resilient strain response. The required

Published by NRC Research Press


1152 Can. Geotech. J. Vol. 50, 2013

Fig. 11. Stress levels for resilient modulus (base material) (Austroads 2007).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 12. Resilient modulus testing for the ARR blend at different levels of RMC.
For personal use only.

values of repeated deviator stress and static confining pressure for Replicate samples were prepared at target moisture contents of 70%,
the 66 stress stages for base material are shown in Fig. 11. 80%, and 90% OMC and a target dry density of 98% of MDD (modified
Based on the basic engineering properties reported earlier in compaction effort) for each blend. Sample deformations were mea-
this paper, the RCO blends were expected to be suitable for sub- sured with two pairs of inductance coils (“Emu” coils, Dawson 1994;
base applications, while the ARR blend might be suitable as a base Janoo et al. 1999; Gabr et al. 2013) mounted on the sample.
course (in the top 150 mm below the pavement surface). Sub-base The resilient modulus was calculated from the average over the
materials are divided into upper and lower sub-base layers in the last six cycles of each stress stage. An example of the observed
approach taken by AustRoads (2007). The upper sub-base layer is variations of resilient modulus with RMC is shown in Fig. 12 for
150 to 250 mm below the pavement surface, whereas the lower the ARR blend. It is evident from Fig. 12 that the resilient modulus
sub-base is more than 250 mm below the pavement surface. decreased as RMC increased within all stress stages. The resilient

Published by NRC Research Press


Azam et al. 1153

modulus of samples in the last loading stage prepared at 67% proved by taking the effect of moisture content into consider-
RMC was 120 MPa greater than for samples of the same material ation, as discussed previously.
prepared at 86% RMC. Therefore it is clear that it is important to The measured data were applied to the model of Yang et al. (2005),
take the effect of moisture content on resilient modulus into which considered moisture state in the form of matric suction, as
consideration, when considering a predictive model for resil- given by eq. (9). A comparison of predicted and measured resilient
ient modulus. modulus data are provided in Fig. 14. Predictions were only fair (R2 =
According to the recommendations of Vuong and Arnold 0.50 and %RMSD = 20.7), which is supported by the plot.
(2006), the number of stress stages may be reduced for relatively Thereafter predictions were sought from the model proposed
lower strength materials, which are more suitable for upper and by Gupta et al. (2007), which also incorporated matric suction. The
lower sub-bases, to avoid failure of the material. Accordingly, initial matric suction of each specimen was applied in this model,
some samples failed before reaching the last stages of resilient according to the SWCC for the blend. The model parameters, ␣1
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

modulus testing where failure is increasingly more likely, as the and ␤1, were derived for a specific bulk stress level of 83 kPa and
ratio of deviator stress to confining pressure is higher. an octahedral shear stress of 19 kPa, for subgrades, as recom-
In summary, the importance of moisture content has been shown mended by NCHRP 1-28A. As the tested material was UGM, more
on the static Young’s modulus, undrained strength, and resilient appropriate stress levels were sought.
modulus of these recycled materials. Thus the moisture state should The specific recommendation by NCHRP 1-28A for the stress levels
be incorporated in predictive models for resilient modulus. suitable for base materials is a bulk stress of 208 kPa and an octahe-
dral shear stress of 49 kPa. Stresses close to the recommended values
Evaluations of models against new data of NCHRP are applied in the first stage of the AustRoads test protocol,
The resilient modulus testing of 24 samples generated 1406 data (bulk stress of 250 kPa and an octahedral shear of 47 kPa), and so the
points. The goodness of fit was accomplished by the statistical data from this stage were used to find the parameters, ␣1 and ␤1, for
software PASW 18, which was verified with the data analysis func- application of the Gupta et al. model. These stresses form the lowest
tions in MS EXCEL. Statistical measurements are the coefficient of stress combination applied to base materials through the AustRoads
correlation (R2), adjusted coefficient of correlation (Radj2
) and rela- protocol. Admittedly, these stresses are much higher than those ap-
tive accuracy (standard error/standard deviation or Se/Sy). Subjec- plied on subgrades, however, the Gupta et al. model cannot be eval-
tive criteria for goodness of fit were reported by Witczak et al. uated without the parameters of ␣1 and ␤1. It was hoped that the
(2002). An excellent fit can be defined as Se/Sy ≤ 0.35 and R2 ≥ 0.90. general form of the model could be appraised for granular bases
Values of Se/Sy of 0.36 to 0.55 and an R2 of 0.70 to 0.89, represent a using this modification.
good fit. A fair fit is defined by Se/Sy of 0.56 to 0.75 and an R2 of 0.40 It was soon realized that the model was material dependent, as
For personal use only.

to 0.69. a very poor statistical fit was achieved for all the data (R2 of 0.15
Additional statistical checks on each predictive model were per- and %RMSD = 28). Nonetheless, the Gupta et al. model provided
formed in terms of root mean square deviation (RMSD). The RMSD reasonable values of R2 for individual blends as shown in Table 3.
is a measure of the differences between values predicted by a As this model could not achieve the objective of this paper, which
model and the values actually measured. RMSD is used to compare was to predict resilient modulus over a wide range of materials, it
forecasting errors of different models for a particular variable, as was not considered any further.
it is scale-dependent (Hyndman and Koehler 2006). In the follow- It should be noted that the Gupta et al. model tended to over-
ing discussion, percentage differences within the data were ap- estimate modulus (refer Fig. 15 for 20% RCO) and that the regres-
plied to obtain %RMSD to eliminate dependency on scale. This sion parameters seemed inconsistent between blends (refer
modified RMSD, or %RMSD, is defined by Table 3). It may also be observed that the trends for 10% and 30%
RCO, were better and had correlation coefficients of 0.85 and 0.89
and %RMSD of 10.7 and 9.7, respectively.
兺 [(x
冪 ⫺ xmeas)/xmeas]2
pred
Finally the measured resilient moduli and specimen data were
(15) %RMSD ⫽ 100 trialled against the model of Liang et al. (2008), which is given in
n eq. (12). Figure 16 presents the fitting of the resilient modulus data.
The regression constants and exponents and the correlation coef-
where xpred is predicted value, xmeas is measured value, and n is ficient, R2, and % RMSD are presented in this figure. It was found
number of datasets. that the Liang et al. model provided reasonably good predictions
Predictions of modulus were made using the various models, for the resilient modulus, but the R2 value of 0.74 fell in between
and constants and exponents for each equation were found. The that for the models of May and Witczak (1981) and K–␪. A practical
correlations were assessed as previously discussed for goodness of limitation of this model is that an SWCC is needed to determine
fit and %RMSD. The simple K–␪ model (eq. (1)) was the first model the parameter ␹w.
investigated. The values of the constant and the exponent of the The model of Cary and Zapata (2011) was unable to be applied to
K–␪ model, were 457 and 0.65, respectively. The correlation coef- the RLTT data as the change in suction with repeated loading was
ficient was 0.72 and the relative accuracy was 0.53, which implied unknown. The build up of pore-water pressure and soil pore-air
a good correlation. The %RMSD was 16.3. pressure during RLTT were not measured, primarily as it was
Subsequently, the May and Witczak predictive model (eq. (2)) considered to be difficult and so was seen as an inappropriate test
was applied and evaluated. Figure 13 presents the resilient modu- procedure for commercial laboratories to conduct.
lus data using the May and Witczak model. Information concern- In summary, all of the investigated resilient modulus models
ing the constants and exponents, the total number of data points were unable to provide adequately reliable predictions for the
(n), and statistical values are presented in a legend to the figure. It range of UGMs in this study. The K–␪ and May and Witczak models
is apparent from Fig. 13 that the May and Witczak model provided provided good predictions for resilient modulus, as R2 was higher
improved predictions for the resilient modulus for all blends, as than 0.70, despite the exclusion of the effect of moisture state. The
R2 was 0.75 compared with 0.72 for the K–␪ model. Similarly, three models that do incorporate moisture state, did not perform
%RMSD was less at 14.9. any better, when appraised against the data from the materials in
The modified universal model (eq. (3)) provided an R2 of 0.72 and this study. Indeed the model of Gupta et al. (2007) appeared not to
a value of Se/Sy of 0.53, and % RMSD = 16, which was as good a be suitable for UGMs. The Liang et al. model (2008) performed
statistical fit as the K–␪ model. So the May and Witczak model better than the Yang et al. (2005) model. A practical limitation
provided better predictions. Nonetheless, this model may be im- common to these suction-based models is the difficulty in the

Published by NRC Research Press


1154 Can. Geotech. J. Vol. 50, 2013

Fig. 13. Predicted resilient modulus against measured resilient modulus for all blends, May and Witczak (1981) model.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13
For personal use only.

Fig. 14. Predicted resilient modulus against measured resilient modulus for all blends by Yang et al. (2005) model.

Table 3. Regression parameters for each blend using the Gupta et al. determination of the parameter, ␹w, which generally requires ac-
(2007) model. curate determination of the SWCC.
Blend K1 k2 k3 K6 K7 ␣1 ␤1 R2 To improve the prediction of resilient modulus and to improve
practicality, a new model was developed for recycled UGM, which
10% RCO 0.30 1.08 0.0 1.05 0.43 47.0 0.33 0.84 is presented in the following section.
20% RCO 0.20 1.20 −0.10 1.00 1.20 105 0.18 0.77
30% RCO 0.25 1.14 0.0 1.02 1.20 31.3 0.32 0.89 Proposed simplified model incorporating soil
ARR 0.002 0.78 1.33 3.87 83.1 9.58 0.41 0.84
suction
The model for prediction of modulus could be improved by con-
sidering particle size grading, and aggregate properties, however,

Published by NRC Research Press


Azam et al. 1155

Fig. 15. Predicted resilient modulus against measured resilient modulus for 20% RCM (RCO) blend by Gupta et al. (2007) model.
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 16. Predicted resilient modulus against measured resilient modulus for all blends by Liang et al. (2008) model.
For personal use only.

particle size gradings were generally similar for the materials in this and Witczak model. RMC was incorporated as another multiplying
study. Moreover, as discussed throughout this paper, the major defi- power term. The correlation coefficient was improved to 0.80 and
ciency of current models is consideration of moisture state through %RMSD was reduced to 13.1.
degree of saturation or soil suction. Although some models already Further improvement was sought by adopting matric suction
exist, they have been based on correlations with subgrades and have into the model, which has been shown to be a much better
been found wanting when applied to UGM. predictor of engineering behaviour than moisture content
The multiplication of power terms is common to most models, (Khoury et al. 2003; Cary and Zapata 2011). Because Gupta et al.
although the model of Gupta et al. (2007), as expressed by eq. (11), (2007) and Azam and Cameron (2013a) found that the relation-
multiplied the stress terms and added the suction term. ship between the resilient modulus and matric suction of a
A trial was conducted to introduce RMC, which represents the material could be represented by a power function, the initial
compacted moisture content normalized by the OMC, in the May sample matric suction was adopted in the model in the form of

Published by NRC Research Press


1156 Can. Geotech. J. Vol. 50, 2013

Fig. 17. Predicted resilient modulus against measured resilient modulus for all blends by eq. (16).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

Fig. 18. Predicted resilient modulus against measured resilient modulus for all blends by eq. (17).
For personal use only.

a power law, as provided in eq. (16). The initial matric suction moduli (Fig. 17) may be compared directly with Fig. 13 (May and
after compaction was chosen to be normalized by the atmo- Witczak model). It is evident that the suction term has brought in
spheric pressure. most of the outlying data points. Indeed, the addition of the ma-
tric suction term in the May and Witczak model increased the
(16) Mr ⫽ k0 冉 冊冉 冊冉 冊
␴m
pa
k1 ␶oct
␶ref
k2 um
pa
k3 correlation coefficient by 10 percentage points.
Further trials to produce even better correlations were con-
ducted. Soil density has been shown by numerous researchers
to affect resilient modulus (Mohammad et al. 1999; George
where um is matric suction, after compaction and before load- 2004; Hopkins et al. 2007; Ba et al. 2011). Therefore the dry
ing (kPa). density ratio was considered for incorporation into the model.
The effect of adding matric suction to the model was strong; R2 In terms of the recycled material in this current study, the blend
improved to 0.85. The plot of measured to predicted resilient composition (RCM percent) was found to have a significant effect on

Published by NRC Research Press


Azam et al. 1157

Table 4. Correlation coefficients for each blend by eq. (17). ulus reliably for all blends using a single set of regression con-
Blend R2 2
Radj Se/Sy %RMSD Model rating stants, over ranges of stress and moisture states (R2 of 0.88).
The final resilient modulus model requires crushed masonry
10% RCO 0.86 0.86 0.37 10.6 Good content (can be zero), dry density, shear strength, initial matric
20% RCO 0.82 0.81 0.43 11.1 Good suction, and stress state. The proposed model does not require the
30% RCO 0.93 0.93 0.27 8.7 Excellent determination of air entry value from SWCC to evaluate the effec-
ARR 0.89 0.89 0.33 9.2 Good
tive stress factor, ␹w. It does require knowledge of the initial ma-
Note: RMSD, root mean square deviation. tric suction prior to loading, which may be measured by FP or by
other testing methods, or by using available empirical equations
Table 5. Comparisons between predictive models for the RLTT data. for the SWCC, combined with the moisture content of the UGM.
Furthermore, the model does not require specialized testing
Model Equation R2 %RMSD
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

equipment for testing of unsaturated soils.


K–␪ (Hicks and Monismith 1971) 1 0.72 16.3 A further achievement of this model is the successful applica-
May and Witczak (1981) 2 0.75 14.9 tion to the resilient modulus of recycled C&D aggregate products,
Modified universal (AASHTO 2002) 3 0.72 16.0 which has not been achieved previously. Crushed masonry con-
Yang et al. (2005) 9 0.50 20.7 tents up to 30% in blends with RCA have been considered.
Gupta et al. (2007) 11 0.15 28.0
Liang et al. (2008) 12 0.74 15.4 Further research
Proposed model 17 0.88 9.8
Although the proposed empirical model appears to be a valu-
able contribution to the prediction of resilient modulus of recy-
cled UGM and perhaps other material, further validation is
the compaction characteristics of the materials (MDD and OMC) as needed across a wide range of materials, moisture contents, and
shown in Table 1. Therefore a further term was introduced that in- densities. Future developments in the theory of unsaturated soils
cluded dry density ratio moderated by the influence of RCM content. may lead to new insights, for example, the new framework for the
The final model proposed is behavior of compacted soils proposed recently by Kodikara (2012).
An attempt was made to incorporate aspects of this theory, how-
(17) Mr ⫽ k0 冉 冊冉 冊冉 冊冋
␴m
pa
k1 ␶oct
␶ref
k2 um
pa
k3 DDR(1 ⫺ k4RCM/100)
100
册 k5
ever, further testing would be required to acquire other recom-
mended parameters.
For personal use only.

References
where DDR is dry density ratio (%), RCM is percent of recycled clay AASHTO. 2002. Guide for the Design of New and Rehabilitated Pavement Struc-
masonry (%), and k4 and k5 are regression parameters (factors and tures. National Cooperative Highway Research Program, AASHTO NCHRP
exponents). Project 1-37A. American Association of State Highway and Transportation
Officials (AASHTO), Washington, D.C.
Validation of proposed model AASHTO. 2007. Standard method of test for determining the resilient modulus
of soils and aggregate materials. AASHTO standard T307-99. American Asso-
The application of the final proposed model (eq. (17)) improved
ciation of State Highway and Transportation Officials (AASHTO), Washing-
the correlation coefficient by a further three percentage points ton, D.C.
(R2 = 0.88); the comparison of measured and predicted moduli is ASTM. 2003a. Standard test methods for measurement of soil potential (suction)
provided in Fig. 18. This improvement is largely because of the using filter paper. ASTM standard D5298. ASTM International.
addition of the last term, which included dry density ratio and the ASTM. 2003b. Standard test methods for determination of the soil water charac-
teristic curve for desorption using hanging column, pressure extractor,
percent of RCM in the blend. The value of R2 of 0.88 is considered chilled mirror hygrometer, or centrifuge. ASTM standard D6836. ASTM In-
a good fit, but the value of Se/Sy of 0.35 lies on the boundary ternational.
between good and excellent. The %RMDS for the proposed model AustRoads. 2007. AustRoads repeated load triaxial test method: determination
was 9.8. of permanent deformation and resilient modulus characteristics of unbound
granular materials under drained conditions. Standard, AG-PT/T053. Aus-
The fitting of data of each blend by eq. (17) using the single set of tRoads, Sydney, Australia. (Reprint of APRG 00/33 (MA) June by Vuong. B.T.,
regression parameters in Fig. 18 provided a good prediction for and Brimble.
10% RCM, 20% RCM (RCO), and the ARR blend, and an excellent Azam, A.M., and Cameron, D.A. 2013a. Geotechnical properties of blends of
prediction for 30% RCM (refer to Table 4). recycled clay masonry and recycled concrete aggregates in unbound pave-
ment construction. Journal of Materials in Civil Engineering, 25(6): 788–798.
A comparison of the published models reviewed in this paper doi:10.1061/(ASCE)MT.1943-5533.0000634.
and the proposed model (eq. (17)) is given in Table 5, based on the Azam, A.M., and Cameron, D.A. 2013b. Laboratory evaluation of recycled con-
total dataset generated by RLTT testing of recycled UGM in this crete aggregate and recycled clay masonry blends in pavement applications.
project. It is evident from this comparison that the proposed Advances in Civil Engineering Materials, 2(1): 328–346. doi:10.1520/
model was superior in prediction of resilient modulus and had the ACEM20120016.
Ba, M., Fall, M., Samb, F., Sarr, D., and Ndiaye, M. 2011. Resilient modulus of
added advantage over two of the other models that only the initial unbound aggregate base courses from Senegal (West Africa). Open Journal of
matric suction is required. Civil Engineering, 1: 1–6. http://www.SciRP.org/journal/ojce. doi:10.4236/ojce.
2011.11001.
Conclusions Brown, S.F. 1996. Soil mechanics in pavement engineering. Géotechnique, 46(3):
383–426. doi:10.1680/geot.1996.46.3.383.
Matric suction has been shown to impact on the strength of Carter, J.P., and Balaam, N.P. 1995. A general finite element program for geotech-
unsaturated soil. The initial matric suction has an important ef- nical engineering. School of Civil and Mining Engineering, University of
fect on the behavior of UGM, such as the shear strength, the static Sydney, Sydney, Australia.
Young’s modulus, and resilient modulus. This research has dem- Cary, C., and Zapata, C.E. 2011. Resilient modulus for unsaturated unbound
materials. Road Materials and Pavement Design, 12(3): 615–638. doi:10.1080/
onstrated that matric suction should be incorporated as a state 14680629.2011.9695263.
variable in the prediction of resilient modulus. Craciun, O., and Lo, S.-C.R. 2010. Matric suction measurement in stress path
Resilient moduli from RLTT under the AustRoads protocol were cyclic triaxial testing of unbound granular base materials. Geotechnical Test-
determined for four blended products. After evaluation of exist- ing Journal, 33(1): 1–12. doi:10.1520/GTJ102390.
Dawson, A. 1994. The E-mu system, users manual. 2nd ed. University of Notting-
ing models for prediction of the resilient modulus of the recycled ham, UK.
UGM in this investigation, an improved and practical model was DPTI. 2011. Part 215 Pavement Materials Appendix 1, Master specification, Divi-
developed. The new model was found capable of predicting mod- sion 2, Road Works. Department for Transport, Energy and Infrastructure

Published by NRC Research Press


1158 Can. Geotech. J. Vol. 50, 2013

(DPTI, formerly DETI-SA). Available from http://www.dpti.sa.gov.au/ sponse of unbound aggregates. Journal of Transportation Engineering,
contractor_documents/specifications_-_division_2_roadworks. 126(1): 66–75. doi:10.1061/(ASCE)0733-947X(2000)126:1(66).
Fredlund, D.G., and Rahardjo, H. 1993. Soil mechanics for unsaturated soils. John Liang, R.Y., Rabab'ah, S., and Khasawneh, M. 2008. Predicting moisture-
Wiley & Sons, Inc. dependent resilient modulus of cohesive soils using soil suction concept.
Fredlund, D.G., and Xing, A. 1994. Equations for the soil-water characteristic Journal of Transportation Engineering, 134(1): 34–40. doi:10.1061/(ASCE)0733-
curve. Canadian Geotechnical Journal, 31(4): 521–532. doi:10.1139/t94-061. 947X(2008)134:1(34).
Fredlund, D.G., Bergan, A.T., and Wong, P.K. 1977. Relation between resilient May, R.W., and Witczak, M.W. 1981. Effective granular modulus to model pave-
modulus and stress conditions for cohesive subgrade soils. Transportation ment responses. Transportation Research Record, 810: 1–9.
Research Record, 642: 73–81. Mohammad, L.N., Huang, B., Puppala, A.J., and Allen, A. 1999. Regression model
for resilient modulus of subgrade soils. Transportation Research Record,
Fredlund, D.G., Xing, A., and Huang, S. 1994. Prediction the permeability func-
1687: 47–54. doi:10.3141/1687-06.
tion for unsaturated soils using the soil-water characteristic curve. Canadian
Nataatmadja, A., and Tan, Y.L. 2001. Resilient response of recycled concrete road
Geotechnical Journal, 31(4): 533–546. doi:10.1139/t94-062.
aggregates. Journal of Transportation Engineering, 127(5): 451–453. doi:10.
Gabr, A.R., Mills, K.G., and Cameron, D.A. 2013. Repeated load triaxial testing of 1061/(ASCE)0733-947X(2001)127:5(450).
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by University of New South Wales on 10/30/13

recycled concrete aggregate for pavement base construction. Geotechnical NCHRP. 2003. Harmonized test method for laboratory determination of resilient
and Geological Engineering, 31(1): 119–132. doi:10.1007/s10706-012-9572-8. modulus for flexible pavement design. Final Report 1-28A, National Cooper-
George, K.P. 2004. Prediction of resilient modulus from soil index properties. ative Highway Research Program (NCHRPA), Transportation Research Board,
Final report. Department of Civil Engineering, University of Mississippi. National Research Council, Washington, D.C.
Gould, S., Rajeev, P., Kodikara, J., Zhao, X.-L., Burn, S., and Marlow, D. 2012. A Nokkaew, K., Tinjum, J.M., and Benson, C.H. 2012. Hydraulic properties of recy-
new method for developing equations applied to the water retention curve. cled asphalt pavement and recycled concrete aggregate. Geocongress 2012
Soil Science Society of America Journal, 6(3): 806–814. doi:10.2136/sssaj2011. Conference, Geotechnical Special Publication 225. American Society of Civil
0260. Engineers. pp. 1476–1485. doi:10.1061/9780784412121.152.
Gupta, S., Ranaivoson, A., Edil, T., Benson, C., and Sawangsuriya, A. 2007. Pave- Parreira, A.B., and Goncalves, R.F. 2000. The influence of moisture content and
ment design using unsaturated soil. Final Report. Minnesota Department of soil suction on the resilient modulus of a lateritic subgrade soil. Presented at
Transportation, Research Service Section. GeoEng - An International Conference on Geotechnical & Geological Engi-
Heath, A.C., Pestana, J.M., Harvey, J.T., and Bejarano, M.O. 2004. Normalizing neering, Melbourne, Australia.
behavior of unsaturated granular pavement materials. Journal of Geotechni- Poon, C.S., and Chan, D. 2006. Feasible use of recycled concrete aggregates and
cal and Geoenvironmental Engineering, 130(9): 896–904. doi:10.1061/(ASCE) crushed clay brick as unbound road sub-base. Construction and Building
1090-0241(2004)130:9(896). Materials, 20(8): 578–585. doi:10.1016/j.conbuildmat.2005.01.045.
Hicks, R.G. 1970. Factors influencing the resilient properties of granular mate- Sawangsuriya, A., Edil, T.B., and Bosscher, P.J. 2008. Modulus-suction-moisture
rials. Ph.D. thesis, University of California, Berkeley, Calif. relationship for compacted soils. Canadian Geotechnical Journal, 45(7): 973–
Hicks, R.G., and Monismith, C.L. 1971. Factors influencing the resilient response 983. doi:10.1139/T08-033.
of granular materials. Transportation Research Record, 345: 14–31. Standards Australia. 1998. Soil strength and consolidation tests -Determination
Hopkins, T.C., Beckham, T.L., and Sun, C. 2007. Resilient modulus of compacted of compressive strength of a soil-Compressive strength of a specimen tested
crushed stone aggregate bases. Kentucky Transportation Centre, University in undrained triaxial compression without measurement of pore water pres-
of Kentucky, USA, Research report No. KTC-05-27/SPR-229-01-1F, prepared in sure. AS1289.6.4.1. Standards Australia, SAI Global Limited, Sydney, NSW,
Australia.
For personal use only.

cooperation with the Kentucky Transportation Cabinet and the FHWA.


Houston, W.N., Dye, H.B., Zapata, C.E., Perera, Y.Y., and Harraz, A. 2006. Deter- van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic
mination of SWCC using one point suction measurement and standard conductivity of unsaturated soils. Soil Science Society of America Journal,
curves. Geotechnical Special publication. American Society of Civil Engi- 44(5): 892–898. doi:10.2136/sssaj1980.03615995004400050002x.
neers, 147: 1482–1493. doi:10.1061/40802(189)123. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E., and Clifton, A.W. 1996. Model for
Hyndman, R.J., and Koehler, A.B. 2006. Another look at measures of forecast the prediction of shear strength with respect to soil suction. Canadian
Geotechnical Journal, 33(3): 379–392. doi:10.1139/t96-060.
accuracy. International Journal of Forecasting, 22(4): 679–688. doi:10.1016/j.
ijforecast.2006.03.001. Vanapalli, S.K., Sillers, W.S., and Fredlund, D.G. 1998. The meaning and rele-
vance of residual state to unsaturated soils. In Proceedings of the 51st Cana-
Janoo, V., Irwin, L., Knuth, K., Dawson, A., and Eaton, R. 1999. Use of inductive
dian Geotechnical Conference, Edmonton, Alta.
coils to measure dynamic and permanent pavement strains. In Proceedings
Vuong, B.T., and Arnold, G. 2006. Predicting in-service performance of alterna-
of the Accelerated Pavement Testing International Conference, Nev. tive pavement materials for New Zealand conditions. Land Transport New
Khalili, N., and Khabbaz, M.H. 1998. A unique relationship for X for the deter- Zealand Research Report 304, Wellington, New Zealand. ISBN 0-478-28716-X,
mination of the shear strength of unsaturated soils. Géotechnique, 48(5): ISSN 1177-0600.
681–687. doi:10.1680/geot.1998.48.5.681. Walker, P.J. 1997. Measurement of total suction and matric suction in pavement
Khoury, N.N., and Zaman, M.M. 2004. Correlation between resilient modulus, materials at Dandenong ALF site. Road & Transport Research, 6(4): 48–58.
moisture variation, and soil suction for subgrade soils. Transportation Re- Witczak, M.W., Andrei, D., and Houston, W.N. 2000. Resilient modulus as func-
search Record, 1874: 99–107. doi:10.3141/1874-11. tion of soil moisture – summary of predictive models. Development of the
Khoury, N.N., Zaman, M., Nevels, J.B., and Manny, J. 2003. Effect of soil suction 2002 guide for the development of new and rehabilitated pavement struc-
on resilient modulus of subgrade soil using the filter paper Technique. In tures, NCHRP 1-37 A, Inter team technical report (Seasonal 1).
Proceedings of the 82nd Annual Meeting of Transportation Research Record. Witczak, M.W., Pellinen, T., and El-Basyouny, M. 2002. Pursuit of the simple
[CD-ROM]. performance test for asphalt concrete fracture/cracking. Journal of the Asso-
Kodikara, J. 2012. New framework for volumetric constitutive behaviour of com- ciation of Asphalt Paving Technologists, 71: 767–778.
pacted unsaturated soils. Canadian Geotechnical Journal, 49(11): 1227–1243. Yang, S.-R., Huang, W.-H., and Tai, Y.-T. 2005. Variation of resilient modulus with
doi:10.1139/t2012-084. soil suction for compacted subgrade soils. Transportation Research Record,
Lekarp, F., Isacsson, U., and Dawson, A. 2000. State of the art. I: Resilient re- 1913: 99–106. doi:10.3141/1913-10.

Published by NRC Research Press

View publication stats

You might also like