Procedia Engineering

Volume 143, 2016, Pages 300–307

Advances in Transportation Geotechnics 3 . The 3rd

International Conference on Transportation Geotechnics
(ICTG 2016)

Characterisation of Permanent Deformation of Silty

Sand Subgrades from Multistage RLT Tests
Farhad Salour 1 and Sigurdur Erlingsson 2, 3
1
Swedish Transport Administration, Trafikverket, Gävle, Sweden
2
Swedish National Road and Transport Research Institute, VTI, Linköping, Sweden
3
Faculty of Civil and Environmental Engineering, University of Iceland, Reykjavik, Iceland
farhad.salour@trafikverket.se, sigurdur.erlingsson@vti.se

Abstract
In thin flexible pavements, unbound aggregate layers and subgrade, contribute largely to the rut
formation of the pavement system. Therefore, realistic prediction of surface rutting requires models that
can reliably capture the cumulative plastic deformation of pavement unbound materials under repeated
loads. Here four models were evaluated to capture the accumulation of permanent deformation of two
fine grained subgrade materials. The models were modified based on the time hardening concept
allowing them to be used in a multistage repeated load triaxial (RLT) test procedure, in which a single
specimen is exposed to a series of consecutive stress paths of varying magnitudes. The RLT tests were
conducted on two non-plastic silty sand subgrade soils with different fines content and at four different
moisture contents. The used triaxial testing system was capable of controlling the pore-air and pore-
water pressures of the specimen and therefore the permanent deformation tests were carried out with
controlled matric suction of the soil samples throughout the test. An effective stress approach was
therefore used in the modelling. Using the test data, the model parameters were determined using a
parameter optimisation algorithm and the qualities of fits achieved was analysed and compared through
goodness-of-fit statistics. In general, the four permanent deformation models were found to work
satisfactorily in capturing the accumulation of the permanent deformation of the tested subgrade
materials. It was concluded that the multistage RLT test procedure has the potentials to be used for
characterizing the permanent deformation behaviour of subgrade soils. This can considerably reduce the
effort and time required for permanent deformation characterization of subgrade materials.

Keywords: Permanent deformation, moisture content, subgrade, unsaturated soil, suction, modelling, triaxial test.

1 Introduction
Rutting is one of the main modes of deterioration in pavements which is often associated with
accumulation of plastic deformation in different pavement layers under repeated traffic loads. In thin
pavement structures, unbound layers and subgrade significantly contribute to the load bearing capacity
of the system and calculating surface rut requires calculation and summation of the permanent

300 Selection and peer-review under responsibility of the Scientific Programme Committee of ICTG 2016
c The Authors. Published by Elsevier B.V.

doi:10.1016/j.proeng.2016.06.038
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

deformation of each individual layer using an appropriate permanent deformation model and material
properties.
Pavement unbound materials generally exhibit an elastoplastic behaviour under repeated cyclic loads
which are usually small compared to the strength of the material (Lekarp, 1999). The deformation of the
material under these loading conditions consists of two parts: the recoverable (resilient) strain and the
plastic (permanent) strain. Resilient and permanent deformation characterization of pavement materials
are generally done through laboratory-based repeated load triaxial (RLT) studies in which loading and
environmental conditions that are often experienced by the material in the field are simulated. The data
obtained from these studies are then used to develop models that can predict the mechanical behaviour
of the materials. (Hornych and El Abd, 2004; Lekarp et al., 2000, Li and Selig, 1996).
From a micromechanical perspective, the development of permanent strains under repeated cyclic
loading can be described using the shakedown concept (Sharp and Brooker, 1984; Werkmeister et al.
2001) and can be classified into three different categories (Figure 1): the plastic shakedown (Range A),
the intermediate response or plastic creep (Range B) and the intermediate collapse (Range C). The
criteria for defining the shakedown concept ranges from RLT tests is described in Figure 1 (Werkmeister
et al., 2004).

Figure 1: Theoretical behaviour of unbound materials under repeated cyclic load (left); different categories of
permanent deformation development (right).
In permanent deformation characterization of unbound pavement materials usually a single stress
path is used (single stage RLT test) where the specimen is subjected to a large number of load pulses
with constant-amplitude and under constant confinement pressure. On the other hand and in the field,
the material generally undergoes stress paths of different magnitudes. Therefore, characterization of
material under different stress conditions requires preparation and testing of several specimen that is
tedious and costly. To overcome this, so-called multistage RLT testing was introduced in which a single
specimen is exposed to a series of consecutive stress paths of varying magnitudes.
Most of the permanent deformation models are developed and calibrated based on single-stage RLT
tests and modelling the accumulated permanent deformation tests from multistage RLT tests has been a
challenge. Erlingsson and Rahman (2013) formulated a general approach based on the time-hardening
concept to mathematically estimate the accumulated permanent deformation of unbound granular
materials from multistage RLT tests.
The objective of this paper was to study the feasibility of using the multistage RLT tests for
unsaturated fine grained subgrades and modelling their permanent deformation accumulation using four
different models that were originally developed for single stage RLT test procedures. For this purpose
two silty sand subgrade soils were tested at different moisture conditions typically observed in field
(Salour et al., 2015). The effective stress approach was used (Equation 1) in which the effect of soil
suction was taken into account when calculating the material stress state (Nuth and Laloui, 2008).
V c V net  S \ m [1]

301
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

Where V c is the effective stress, V net is the total exterior stress, S is the saturation degree and \ m
is the matric suction. Furthermore, the effect of moisture content variation on the accumulated
permanent deformation of the tested materials and the model parameters was investigated.

2 Permanent Deformation Modelling from RLT Tests

2.1 Single-Stage RLT Tests
Studies on permanent deformation behaviour of pavement unbound materials are limited and most
of the predictive models are based on experimental data obtained from single-stage RLT tests. These
permanent deformation models can generally empirically describe the influence of number of load
applications or level of stress (or a combination of them) on the permanent development (Hornych and
El Abd, 2004; Lekarp et al., 2000).
In this paper, four models (described further in the paper) that combine the influence of the number
of the load applications with the effect of material stress level are chosen. These models were originally
developed for single stage RLT test procedures and can be presented as followed:
HˆpN f1N f2 p, q,Hr  [2]
where Hˆ p is the accumulated permanent strain, N is the total number of load cycles, p is the hydrostatic
stress, q is the deviator stress and H r is the resilient strain.
Tseng and Lytton model
The Tseng and Lytton model (1989), combines the effect of the material stress state with the number
of load applications and assumes a direct relationship between the permanent strain and the resilient
strain. For a single stage RLT test the model is expressed as:
U
( ) E [3]
Hˆ p  N H rH0e N

where H r is the resilient strain at Nth load cycle and H 0 , U and E are material parameters.
Gidel et al. model
The Gidel et al. (2001) model combines the stress state and the number of load application and is
expressed as:
§ B ·§ L u 1
0¨ § N · ¸¨ max · § s q ·
Hˆ p N  H 1 ¨ ¸ ¸ ¨m   max ¸ [4]
¨ © 100 ¹ ¸¨ p a ¸ ¨ p max p max ¸
© ¹© ¹ © ¹
where H 0 , B and u are material parameters. p max and q max are the maximum applied hydrostatic
stress and deviator stress, respectively and L max 2
p max 2
 q max . The parameters m and s are the slope
and the intercept of the Mohr-Coulomb failure line in the p  q space, respectively, and are obtained
from static triaxial tests. Finally pa is a reference stress (100 kPa).
Korkiala-Tanttu model
The Korkiala-Tanttu (2005) model describes the development of permanent deformation using a
power law function of the number of load repetitions. In this model, the effect of stress level of the
material on the accumulated permanent deformation is taken into account using the shear ratio. The
model is expressed as:
R
Hˆ p N  CN b [5]
A R

302
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

where C and b are material parameters; A is the maximum theoretical the shear stress ratio ( A = 1.05
according to the model author’s recommendation) and R is the shear stress ratio defined as follow:
q
R [6]
qf
q mp  s [7]
f

Erlingsson and Rahman model

The Erlingsson and Rahman (2013) model directly relates the development of permanent strains to
the stress level ( S f ) which is also a component of the number of load cycles ( N ) together with the
material parameter b . The model is expressed as follow:
bS f
Ĥ p N  aN Sf [8]
where a and b are regression parameters associated with the material and the term S f takes into account
the effect of stress state applied given as:
§ q ·
¨¨ ¸
© p a ¸¹
Sf [9]
D
§ p ·
¨ ¸
¨ p ¸
© a¹
where p a is the reference stress (100 kPa) and D is a parameter obtained from regression analysis.

2.2 Multi-Stage RLT Tests

In order to use the multistage RLT test data, the generalized model described in Equation 2 needs to
be modified (Equation 10) using the time hardening approach (Erlingsson and Rahman, 2013) in which
the accumulated permanent strain from the preceding loading sequences is used to calculate the number
of load cycles required so that the same amount of accumulated permanent strain in the current stress
path (stress path i ) is obtained (Figure 2). This is called the equivalent load cycle ( N ieq ). The N ieq is
then used to modify the total number of load cycles ( N ) from the beginning of the test to calculate the
effective number of load cycles ( N  N i 1  N ieq ) in which N i 1 is the total number of load cycles at
the end of the ( i 1) stress path. Thus, it is possible to model the whole range of the multistage RLT tests.

Figure 2: Schematic overview of the time-hardening concept.

303
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

By implementing the time-hardening concept into the generalized permanent deformation model
(Equation 1), the following generalized equation for multistage RLT tests is obtained:
Hˆ p i  N  f1 ( N  N i 1  N ieq ) f 2 ( pi , qi , H ri ) [10]
When the time-hardening concept is applied to the four models described earlier the following
equations to be used in multistage RLT tests are obtained.
Tseng and Lytton model
The Tseng and Lytton (1989) model can be rewritten as followed:
U
( )E
N  N i 1  N ieq [11]
Hˆ pi  N  H ri H 0e
where:
1 E
§ § Hˆ p · ·
Nieq U¨  ln¨ i1 ¸ ¸ [12]
¨ ¨ Hr Ho ¸¸
© © i ¹¹
Gidel et al. model
The Gidel et al. (2001) model can be rewritten as followed:
§ § B
eq · · u
§ ·
1
0¨ N  N  N ¸§ L · s q
H ¨1  ¨ ¸ ¨ max,
i max,i
Hˆ pi N  i 1 i ¸ ¨m   ¸ [13]
¨ 100 ¸ ¸¨ pa ¸ ¨ p p ¸
¨ © ¹ ¹ ¸© ¹ © max,
i max,
i ¹
©
where:
1
§ u 1 ·  B
¨ Hˆ p § Lmax, i · § q max, i · ¸
N ieq 100¨1  i01 ¨¨ ¸¸ ¨m  s  ¸ [14]
¨ ¸ ¸
¨ H © pa ¹ © p max, i pmax, i ¹ ¸
© ¹
Korkiala-Tanttu model
The Korkiala-Tanttu (2005) model can be rewritten as followed:
Ri
Hˆ p i  N  C ( N  N i 1  N ieq ) b [15]
A  Ri
where:
1
§ Hˆ pi 1 ( A  R i ) · b
N ieq ¨ ¸ [16]
¨ CR i ¸
© ¹
Erlingsson and Rahman model
The Erlingsson and Rahman (2013) model can be rewritten as followed:
bS
Hˆ pi N a( N  Ni 1  Nieq ) fi S f i
  [17]
where:
§ Hˆ p ·
N ieq ¨ i 1 ¸ [18]
¨ aS ¸
fi
© ¹
In all of the abovementioned equations, the subscript i refers to the i th stress path.

304
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

3 Test Program and Result

Two non-plastic silty sand (SM) subgrade soils (Table 1) with different fines content were tested at
four different moisture contents (Table 2) along with corresponding matric suction measurement that
was obtained from Soil-Water Characteristic Curve using Fredlund pressure plate cell (Salour and
Erlingsson, 2015). The applied multistage RLT tests consisted of three load sequences each comprised
of three stress paths with a constant confinement pressure and ascending deviator stresses (Table 3). The
selected external applied stresses are chosen to simulate the stress state ranges often experienced in the
subgrade material in pavements. In each stress path, 10,000 haversine load pulse, (1 െ cos ߠ )/2, at
loading duration 0.2 sec and a rest period of 0.4 sec were applied followed by the succeeding stress path
(90,000 load applications in total). The LVDTs that measured the vertical deformations of the cylindrical
4x8 inch specimen were mounted on the middle half of the specimen. These LVDTs were attached to
the studs that were buried into specimens during the compaction.
Passing Max dry unit weight Opt. moisture Specific gravity
Subgrade
No. 200* [%] Ȗdmax [kN/m³] content wopt [%] SG [-]
Luleå 42.2 19.6 10.1 2.68
Torpsbruk 27.4 20.3 7.6 2.67
* 75 microns sieve opening.
Table 1 Summary of the subgrade soils index properties.

Optimum moisture Moisture Degree of Corresponding matric

Subgrade suction \ m [kPa]
content wopt [%] content w [%] saturation S [%]
6.0 47.3 90.0
8.0 63.1 53.0
Luleå 10.1
9.3 73.3 44.0
10.1 79.6 35.0
4.0 37.0 140.0
5.5 50.8 49.0
Torpsbruk 7.6
6.8 62.8 28.0
7.4 68.4 25.0
Table 2 Subgrade moisture content, degree of saturation and measured matric suction for multistage RLT tests.

Loading Stress Confinement stress Cyclic deviator. Maximum deviator

sequence path ı 3 [kPa] stress ¨q [kPa] stress q0-max [kPa]
1 27.6 13.8 19.3
1 2 27.6 27.6 33.1
3 27.6 41.4 46.9
4 41.4 20.7 26.2
2 5 41.4 48.3 53.8
6 41.4 69.0 74.5
7 55.2 27.6 33.1
3 8 55.2 82.8 88.3
9 55.2 124.2 129.7
Table 3 Applied external stress levels used for the multistage RLT tests.

The data from the multistage RLT tests were used to calibrate the permanent deformation models.
The optimization of the model parameters were carried out so that the sum of squares for error (SSE)
from the test data and the model predictions is minimised (Erlingsson and Rahman, 2013). This required
some experience and engineering judgment in selecting a suitable set of parameters due to the fact that

305
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

different set of parameter combinations can provide a reasonable fit (Erlingsson and Rahman, 2013;
Korkiala-Tanttu, 2005; Rahman and Erlingsson, 2014). The optimization of the parameters was
performed for both of the subgrades and all of the tested moisture contents. Based on the discussion
above and for each of the models, a set of nonlinear parameters that would provide a reasonable fit for
all the moisture contents were chosen (Table 4). Figure 3 shows an example of the measured permanent
deformation development from the multistage RLT test and predictions for different calibrated models
for the Luleå subgrade at moisture content 9.3%.

Figure 3: Example of permanent deformation development from multistage RLT tests and calibrated models
for Luleå subgrade at moisture content 9.3% (A, B and C are test data shakedown ranges).
Tseng and Korkiala-
Gidel et al.2 Erlingsson and Rahman 4
Lytton1 Tanttu3
İ0×10-3 2 ݺ ×10 -3 2 C ×10 -3 2 a ×10-3 2
Subgrade w [%] Radj Radj Radj b [-] Radj
[-] [-] [-] [-]
6.0 6.2 0.99 1.19 0.92 0.39 0.97 0.110 0.300 0.91
8.0 9.2 0.99 1.03 0.79 0.77 0.82 0.127 0.300 0.95
Luleå
9.3 18.8 0.99 4.10 0.82 3.07 0.96 0.508 0.300 0.87
10.1 32.0 0.97 6.70 0.83 4.09 0.96 0.637 0.300 0.97
4.0 6.0 0.97 0.93 0.80 0.57 0.94 0.135 0.307 0.88
5.5 10.0 0.99 1.36 0.87 0.58 0.88 0.092 0.307 0.93
Torpsbruk
6.8 45.2 0.97 6.0 0.86 2.35 0.72 0.312 0.307 0.99
7.4 54.0 0.94 6.9 0.85 2.67 0.68 0.339 0.307 0.99
1
ȡ [-] = 250,000 and ȕ [-] = 0.1, 2 B [-] = 0.2 and u [-] = 0.65, 3 b [-] = 0.09122, 4 Į [-] = 1.0
Table 4 Permanent deformation calibrated model parameters and quality of model fit.

From visual examination of the calibrated models and the measured data as well as the quality of the
fit of the models it can be inferred that the modified models based on the time-hardening concept
performed satisfactorily for the selected materials. Generally, reasonable quality of fit was observed
with some differences between the models.
The Tseng and Lytton model showed better predictions for the tested subgrades for all the moisture
condition and stress levels compared to the Korkiala-Tanttu, Gidel et al. and Erlingsson and Rahman
models. The advantage of the Erlingsson and Rahman model on the other hand was that it contains
relatively fewer numbers of material parameters, and unlike the Korkiala-Tanttu, Gidel et al., it does not
require the shear strength parameters and directly utilizes applied stress level as a predicting variable,
simplifying its application.

306
Characterisation of Permanent Deformation of Silty Sand Subgrades ... Salour and Erlingsson

4 Conclusions
A series of multistage RLT tests was conducted on two silty sand subgrades at four different moisture
contents and under controlled matric suction. The test data were then used to calibrate four permanent
deformation models using the time-hardening concept. Generally it was observed that the modified
models performed quite satisfactorily in predicting the relationship between the total number of load
applications, stress level and the accumulation of permanent deformation. Among the models and
considering the goodness of fit of the data Tseng and Lytton model exhibited the best performance in
most of the cases. The Tseng and Lytton and the Erlingsson and Rahman models do not require
determination of the static parameters, which considerably reduces the number of tests that needs to be
performed for calibration of the prediction model, making them more practice-ready. A drawback of the
current permanent deformation prediction models is that they include several parameters, making the
calibration troublesome and increasing the uncertainties in the model predictions. Therefore developing
models that contain less parameters and does not need the shear strength parameter of the material would
be very beneficial.

References
Erlingsson, S., and Rahman, M. S. (2013). Evaluation of permanent deformation characteristics of unbound
granular materials from multi-stage repeated load triaxial test. Transport Research Records: Journal of the
Transportation Research Board. No. 2369, 11-19.
Gidel, G., Hornych, P., Chauvin, J., Breysse, D., and Denis, A. (2001). A new approach for investigating the
permanent deformation behavior of unbound granular material using the Repeated Load Triaxial Apparatus.
Bulletin des laboratoires des Ponts et Chaussées, 5-21.
Hornych, P., and El Abd, A. (2004). Selection and evaluation of models for prediction of permanent
deformations of unbound granular materials in road pavements. Sustainable and Advanced Materials for Road
Infrastructure, SAMARIS EU project, Work Package 5, Performance based specifications.
Korkiala-Tanttu, L. (2005). A new material model for permanent deformations in pavements. Proceedings of
the 7th international conference on bearing capacity of roads and airfields, BCRRA ’05. Trondheim, Norway.
Lekarp, F. (1999). Resilient and permanent deformation behavior of unbound aggregates under repeated
loading. PhD Thesis. The Royal Institute of Technology, KTH, Stockholm, Sweden.
Lekarp, F., Isacsson, U., and Dawson, A. (2000). State of the art. II: Permanent strain response of unbound
aggregates. Journal of Transportation Engineering, 126(1), 76-83.
Li, D., and Selig, E. T. (1996). Cumulative plastic deformation for fine-grained subgrade soils. Journal of
geotechnical engineering, 122(12), 1006-1013.
Lytton, R. L., Uzan, J., Fernando, E. G., Roque, R., Hiltumen, D. and Stoffels, S. M. (1993) Development and
validation of performance prediction models and specifications for asphalt binders and paving mixes. Report
SHRPA-357. TRB, National Research Council, Washington, D.C., USA.
Nuth, M. and Laloui, L. (2008). Effective stress concept in unsaturated soils: clarification and validation of a
unified framework. International Journal for Numerical and Analytical Methods in Geomechanics, 32(7), 771-801.
Rahman, M. S., and Erlingsson, S. (2014). Predicting permanent deformation behaviour of unbound granular
materials. International Journal of Pavement Engineering, 16(7), 587-601.
Salour, F., and Erlingsson, S. (2015). Permanent deformation characteristics of silty sand subgrades from
multistage RLT tests. International Journal of Pavement Engineering, DOI:10.1080/10298436.2015.1065991.
Salour, F., Erlingsson, S., and Zapata, C. E. (2015). Model for seasonal variation of resilient modulus in silty
sand subgrade soil: evaluation with falling weight deflectometer. Transportation Research Record: Journal of the
Transportation Research Board, 2510, 65-73.
Tseng, K. H., and Lytton, R. L. (1989). Prediction of permanent deformation in flexible pavement materials.
Implication of Aggregates in the Design, Construction, and Performance of Flexible Pavements, ASTM STP, 1016,
154-172.
Werkmeister, S., Dawson, A. R., and Wellner, F. (2001). Permanent deformation behavior of granular materials
and the shakedown concept. Transportation Research Record: Journal of the Transportation Research Board,
1757(1), 75-81.

307