Transportation Geotechnics 38 (2023) 100866

Contents lists available at ScienceDirect

Transportation Geotechnics
journal homepage: www.elsevier.com/locate/trgeo

Forward calculation of displacement fields with multilayered unsaturated

highway system induced by falling weight deflectometer using dynamic
response method
Jing Zhang a, b, Zheng Lu a, c, *, Chuxuan Tang a, b, Jie Liu d, Hailin Yao a
a
State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan 430071, China
b
University of Chinese Academy of Sciences, Beijing 100049, China
c
Hubei Key Laboratory of Geo-Environmental Engineering, Wuhan 430071, China
d
Xinjiang Transportation Planning Survey and Design Institute Co., Ltd., Urumqi 830006, China

A R T I C L E I N F O A B S T R A C T

Keywords: A new dynamic forward model of a multilayered highway structure, which considers the unsaturated charac­
FWD teristics of subgrade soil, is established in this paper to assess the dynamic response induced by the falling weight
Elastic modulus deflectometer (FWD) load. The Laplace-Hankel transform and dynamic stiffness matrix method are employed to
Saturation
solve the governing equations of the unsaturated soil layer. Combined with the global stiffness matrix and
Subgrade deterioration
Vertical displacement
boundary conditions, analytical solutions of the entire system can be derived in the transform domain. A nu­
merical integration method is utilized to perform the Laplace-Hankel inverse transform to obtain time-domain
solutions. An analysis of the effects of saturation on the stress, displacement, and pore water pressure is pre­
sented. It is revealed that the effect of saturation of unsaturated subgrade soil on vertical stress is negligible,
whereas vertical displacement and pore water pressure increase significantly with increasing saturation. Addi­
tionally, the vertical displacement response on the pavement surface is sensitive to subgrade deterioration and
variations in material properties. Moreover, subgrade deterioration can be reflected in the variation in vertical
displacement at approximately 1.2 m from the load center for the studied road structure. The proposed model,
which is more realistic and reasonable, can provide a theoretical basis for the back-calculation of the elastic
modulus of multilayered highway systems.

Introduction requires field sampling. In recent years, the falling weight deflectometer
(FWD), a non-destructive testing equipment with high accuracy and
As the subgrade forms the foundation of pavements, its poor per­ efficiency, has been widely used to assess pavement structure perfor­
formance, particularly the insufficient stiffness characteristics that mance and back-calculate the layer dynamic modulus [1,2]. The FWD
determine the deformation resistance of the entire road, significantly can apply a brief half-sine impact load on the pavement surface by
impacts the long-term stability and service life of highways. In addition, dropping a mass on a 0.3 m diameter circular plate. A load cell placed
many pavement damages have been attributed to the reduced stiffness of over the plate measures the impact load, and a series of geophones that
the subgrade. In pavement design, the resilient modulus is often used as are arranged along the radial direction on the pavement surface records
an indicator of the stiffness of the subgrade. Therefore, it is crucial to the vertical displacements at different positions [3–5]. The back-
accurately assess and predict the resilient modulus of the pavement and calculation of layer dynamic modulus by using FWD is accomplished
subgrade. The bearing plate method is one of the most commonly used by optimally matching the measured field displacement with the
test methods to determine the resilient modulus; however, the high cost calculated theoretical displacement. The forward model, which can
of time limits its application. The laboratory dynamic triaxial test of obtain the calculated theoretical displacement, becomes the theoretical
subgrade soil is a destructive and probabilistic testing method that basis and precondition of the back-calculation of the dynamic modulus

* Corresponding author at: State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of
Sciences, Wuhan 430071, China.
E-mail addresses: [email protected] (J. Zhang), [email protected] (Z. Lu), [email protected] (C. Tang), [email protected] (J. Liu),
[email protected] (H. Yao).

https://doi.org/10.1016/j.trgeo.2022.100866
Received 25 May 2022; Received in revised form 15 September 2022; Accepted 19 September 2022
Available online 24 September 2022
2214-3912/© 2022 Elsevier Ltd. All rights reserved.
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

unsaturated model to describe the physical state and mechanical

behavior of subgrade soil than to use the saturated model. Fang et al.
[24] established a track-multilayer ground model to study the dynamic
response of an unsaturated subgrade under a moving train load. Tang
et al. [25] studied the dynamic responses of a pavement-unsaturated
ground system subjected to moving traffic loads. Shi et al. [26] exam­
ined a dynamic response considering different boundary conditions of
unsaturated soils under a strip load. Ye and Ai [27] proposed an un­
saturated poroelastic model to study the dynamic responses of a multi­
layered unsaturated medium subjected to a vertical time-harmonic load.
Notably, most of the aforementioned studies related to unsaturated soils
focus on the dynamic response of unsaturated subgrades under moving
loads or time-harmonic loads, while the time-history curve of the FWD
load has not been considered. Conversely, the dynamic response models
Fig. 1. Dynamic response model of three-layer dynamic highway structure of multilayered pavements induced by FWD are mostly elastic or
under FWD load. viscoelastic models [9,10,12], which do not consider the unsaturated
characteristics of subgrade soil. Therefore, unsaturated soil theory needs
of the subgrade and pavement, thereby determining the inversion ac­ to be applied to study the FWD test mechanism.
curacy and reliability. Therefore, establishing a complete and realistic The objective of this study is to establish a more realistic multilay­
highway structure forward model to accurately analyze the dynamic ered dynamic forward model considering the unsaturated characteristics
response induced by the FWD load is essential. of subgrade soil to predict the dynamic response of the entire highway
In previous studies, for simplicity, soils were assumed to be elastic or structure under FWD load, especially the pavement surface displace­
viscoelastic media. Lamb [6] first solved the vibration propagation ment response. This study contributes to advancing the development of
problem on a semi-infinite isotropic elastic solid surface. Since then, back-calculation of the subgrade and pavement elastic modulus. The
many scholars have investigated elastodynamic problems. Jones et al. governing equations of each layer are solved in the Laplace-Hankel
[7] investigated the vibration response of a viscoelastic half-space under transform domain, and the global stiffness matrix is then synthesized
a rectangular vertical load. Yue et al. [8] presented an analytical solu­ by the continuity condition. The space–time domain solutions of the
tion to the stress and displacement of an elastic half-space under a entire pavement-subgrade system are obtained by using the inverse
rectangular load. Han et al. [9] proposed a spectral element method Laplace-Hankel transform; the calculation results of other scholars can
based on elastic theory to calculate the displacement response for a be used to verify the accuracy and reliability of this forward model. As
transversely isotropic layered highway structure subjected to an FWD an essential parameter of subgrade soil, the effects of saturation on the
load. Using the spectral element method, Grenier et al. [10] simulated stress, displacement, and pore water pressure are discussed. Further­
the dynamic response of flexible asphalt pavement under FWD loads. Ai more, the feedback of the pavement surface displacement due to sub­
and Zhang [11] also studied the effects of anisotropy, load depth, and grade deterioration and the change in elastic moduli and thicknesses of
load frequency on dynamic response using the spectral element method. different structural layers is examined. This forward model, which is
Cai et al. [12] studied the dynamic response of a viscoelastic multilay­ more realistic and reasonable, can provide a theoretical basis for the
ered pavement under excitation by an FWD load. Park and Kausel [13] back-calculation of the elastic moduli of pavement-subgrade systems.
elucidated the response functions for an elastic half-space model based
on wave theory. Solutions for an elastic medium under dynamic loads Solution for the pavement-subgrade system
were reported by Lin et al. [14], Liu and Pan [15], Picoux et al. [5], and
Ruta and Szydło [16]. Because the load period of the FWD is short and the stress level of the
Although the aforementioned single elastic models have concise FWD load is typically low, the multilayered road system can be assumed
mathematical formulas, they only consider the soil skeleton, not the to be linearly elastic. In addition, considering the unsaturated properties
pore fluid; this approach is flawed. Biot [17] proposed a theory for of the subgrade soil, a three-layer dynamic model is proposed to char­
solving the wave propagation problem in a saturated porous elastic acterize the highway structure, as shown in Fig. 1. The layered pavement
medium. Many scholars have applied saturated porous elastic models to structure consists of two layers: a surface course and a base course. The
study the dynamic properties of soils. Lu and Hanyga [18] developed a subgrade is modeled as an unsaturated half-space. It is assumed that the
transmission and reflection matrix method to calculate the displace­ layers are frictionless, and the subgrade surface is permeable.
ments, stresses, and pore pressures in a layered porous half-space. Ai and
Wang [19] studied the dynamic response of a saturated elastic soil layer Solutions for the elastic surface course and base course
and demonstrated the influence of the load type, load depth, soil
permeability, and stratification characteristics on the calculation results. The surface and base courses were simulated as linear isotropic
Jin et al. [20] examined the influence of diffusivity on the stresses and elastic layers with thicknesses h1 and h2, respectively. On the basis of the
excess pore fluid pressure in saturated poroelastic soil induced by a method by Yao et al. [28], which derives the dynamic stiffness matrices
concentrated line load. Based on Kirchhoff’s hypotheses, Fang et al. [21] in Cartesian coordinates, the dynamic stiffness matrices can be obtained
calculated the vertical displacement of the pore elastic half-space under in a cylindrical system.
a moving traffic load and studied the influence of load velocity and The relationship between stress and displacement of the surface
material properties. Ba et al. [22,23] proposed a method for evaluating course can be expressed as:
plane strain responses subjected to time-harmonic loads or moving point {∑} { } { }
forces on a multilayered transversely isotropic saturated half-space = [Sp ][Dp ]− 1 U
̃ = [K p ] U
̃ (1)
p p p
surface.
Most classical theories simplify soils as a single elastic or two-phase where [Sp ] is the stress matrix for the surface course, [Dp ] is the
medium containing soil grains and pore water. However, subgrade soil is displacement matrix for the surface course, [Kp ] is the dynamic stiffness
{ }
a typical porous unsaturated medium, owing to precipitation, infiltra­ ∑
matrix for the surface course, and { }p , U ̃ are the stress and
tion, and evaporation. Therefore, it is more reasonable to use the p
displacement components of the upper and lower interfaces of the sur

2
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

[ ]− m
face course in the Laplace-Hankel transform domain, respectively. By using the relationship Se = 1 + (αpc )d = S1−r − SSw0
w0
[29], the time
Similarly, for the base course, the relationship between stress and
displacement can be written as. derivative of saturation can be obtained.
{∑} [ ]d−d 1 ( )
[ ][ ]− 1 { } [ ]{ }
(8)
m+1 1
= Sb Db ̃ = Kb U
U ̃ (2) Ṡr = − αmd(1 − Sw0 )(Se ) m (Se )− m − 1 ṗa − ṗw
b b b

[ ] [ ]
where Sb , Db are the stress and displacement matrices for the base where pc = pa − pw denotes the matrix suction, α, m and d are fitting
[ ]
course, respectively; Kb is the dynamic stiffness matrix for the base parameters; and Sw0 is the irreducible saturation.
∑ { } The aim of this study is to investigate the dynamic response of
course; and { }b , U ̃ are the stress and displacement components of
different saturations; therefore, the shear modulus of unsaturated soil
b
the upper and lower interfaces of the base course in the Laplace-Hankel must be obtained. The shear modulus of unsaturated soil is affected by
transform domain, respectively. many factors, including humidity, strain, suction, and suction path
[30,31]. Considering the complexity of the factors affecting the shear
Solution for the unsaturated porous elastic subgrade modulus of unsaturated soil, only saturation and suction are considered
in the proposed model. Seed and Idriss [32] proposed an indirect
Subgrade soil was simulated as an unsaturated porous elastic half- method for calculating the shear modulus of unsaturated clay from shear
space. According to Biot [17], the fundamental governing equations strength. Further, Fredlund et al. [33] proposed that the shear strength
include the following: of unsaturated soil could be predicted from matric suction. Based on
The equilibrium equation after ignoring body force can be expressed these theories, the shear modulus in the unsaturated state was deter­
as. mined as follows [24,25]:
[√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ]
σ ij,j = ρüi + ρw Ẅ i + ρa V̈ i (3)
G = Gs + 2200ln (Se )− 2 − 1 + (Se )− 1 tanφ/α (9)
Here, σ ij is the total stress; the total density ρ = (1 − n)ρs + nSr ρw +
n(1 − Sr )ρa , where ρs , ρw and ρa are the mass densities of soil particles, where Gs is the shear modulus when the soil is saturated and φ is the
( )
water, and air, n is porosity; Sr is saturation; Wi = nSr uwi − ui and, Vi = internal friction angle.
( a )
n(1 − Sr ) ui − ui denote the displacement components relative to soil Applying the constitutive equation of soil particles
dσ sij
3Ks dt , and
dρs
=
particles in the i of pore water and air, respectively; ui , uwi and uai are the
ρs dt
substituting σ ij = (1 − n)σsij − nSr pw δij − n(1 − Sr )pa δij can yield Eq. (10).
displacements of particles, water, and air, respectively; and (˙) denotes
( )
the time derivative. aχ − nSr a(1 − χ ) − n(1 − Sr ) Kb
The relationship between the stress and strain of unsaturated soil is ṅ = ṗw + ṗa + 1 − n − ∇⋅u̇ (10)
Ks Ks Ks
given in Eq. (4).
where σsij is the stress component on the soil skeleton and Ks is the bulk
σ ij = 2Gεij + λεδij − apδij (4)
compression modulus of the soil particles.
Here, εij is the strain; ε denotes the volumetric strain; λ and G are the Substituting Eqs. (8) and (10) into Eq. (7), the mass-conservation
lame constants of soil; δij denotes the Kronecker delta; the coefficient equation can be obtained.
a = 1 − Kb /Ks , where, Kb , Ks are the compressive moduli of the soil
skeleton and soil particles, respectively; the average pore pressure p = A11 ṗw + A12 ṗa + A13 ∇⋅u̇ + A14 ∇⋅Ẇ = 0 (11a)
χ pw + (1 − χ )pa , where χ is an effective stress parameter; and pw , pa
denote the pore pressures of water and air. A21 ṗw + A22 ṗa + A23 ∇⋅u̇ + A24 ∇⋅V̇ = 0 (11b)
Combining Eqs. (3) and (4), the following equation is obtained:
where A11 = (aχ−KsnSr ) − nAs + nS Kw ; A12 =
r [a(1− χ )− n(1− Sr ) ]Sr
Ks + nAs ; A13 =
G∇2 u + (λ + G)∇ε − aχ ∇pw − a(1 − χ )∇pa = ρü + ρw Ẅ + ρa V̈ (5) ( )
1 − KKbs Sr ; A14 = 1; A21 = (1 − Sr ) (aχ−KsnSr ) + nAs ; A22 =
[ ] ( )
where u, W, and V are the displacement vectors of soil particles, water, (1 − Sr ) a(1− χ)−Ksn(1− Sr ) − Kna − nAs ; A23 = 1 − KKbs (1 − Sr ); A24 = 1; and
and air, respectively. [ ]d−d 1
The seepage motion equations for water and air are given as.
m+1 1
As = − αmd(1 − Sw0 )Se m (Se )− m − 1 .
ρw ρw g The m-order Laplace-Hankel transform and its inverse transform are
− pw,i = ρw üi + Ẅ i + Ẇ i (6a)
nSr kw employed to solve the governing equation, as shown in Eq. (12).
∫ +∞ ∫ +∞
ρa ρg
− pa,i = ρa üi + V̈ i + a V̇ i (6b) ̃f m (ξ, z, s) = f (r, z, t)Jm (ξr)re− st drdt (12a)
n(1 − Sr ) ka 0 0

∫ ∫
where kw = ρw gκkrw /ηw and ka = ρa gκkra /ηa denote the permeability +∞ +∞
f (r, z, t) = ̃f m (ξ, z, s)Jm (ξr)ξest dξds (12b)
coefficients of water and air, respectively; krw and kra are the relative 0 0
permeability coefficients of water and air, respectively; ηw and ηa are the
m
dynamic viscosity coefficients of water and air, respectively; and κ is the where f is a variable in the space–time domain, ̃f is the m-order
inherent permeability of the soil. Laplace-Hankel transform of f in the transformed domain, ξ and s are the
Upon utilizing the change rates of n, Sr and density in space are far Hankel and Laplace transform parameters, respectively, and Jm denotes
less than those with respect to time, and the continuity equation can be the m-order Bessel function.
derived using Eq. (7). By applying the Laplace transform to Eqs. (5), (6), and (11), Eq. (13)
ρ̇w can be obtained as follows:
nS˙r + Sr ṅ + Sr n + nSr ∇⋅u̇w = 0 (7a)
ρw G∇2 u
̃ + (λ + G)∇̃ε − aχ ∇̃ u + ρw s 2 ̃
pa = ρs2 ̃
pw − a(1 − χ )∇̃ ̃
W + ρa s2 V
ρ̇a (13a)
− nṠr + (1 − Sr )ṅ + (1 − Sr )n + (1 − Sr )n∇⋅u̇a = 0 (7b)
ρa

3
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Table 1 Table 2
Parameters of saturated half-space soil [19]. Parameters of elastic pavement and unsaturated subgrade [24,25,34].
Parameter (Symbol) Unit Value Soil phase Parameter Unit Surface Base Subgrade
(Symbol) course course
Shear modulus (G) MPa 30
Lame constant (λ) MPa 30 Framework Thickness (h) m 0.18 0.35 ∞
Density of solid (ρs) kg/m3 2000 Elastic modulus MPa 2300 910 80
Density of water (ρw) kg/m3 1000 (E)
Permeability coefficient (kw) m/s 1 × 10-6 Poisson ratio (ν) – 0.25 0.25 0.35
Porosity (n) – 0.4 Density (ρ) kg/ 2100 2000 –
m3
Porosity (n) – – – 0.4
Saturation (Sr) – – – 0 ~ 1.0
Effective stress – – – Sr
parameter (χ )
Irreducible – – – 0.05
saturation (Sw0)
Inherent m2 – – 1 × 10-9
permeability (κ)
Internal friction ◦
– – 27
angle (φ)
Fitting parameter – – – 4 × 10-5
(α)
Fitting parameter – – – 0.5
(m)
Fitting parameter – – – 2
(d)
Grain Density (ρs) kg/ – – 1861
m3
Bulk modulus (Ks) GPa – – 2.96 × 109
Water Density (ρw) kg/ – – 1000
m3
Bulk modulus (Kw) GPa – – 0.155
Viscosity (ηw) N⋅s/ – – 1.005 ×
m2 10-3
Air Density (ρa) kg/ – – 1.29
m3
Fig. 2. Comparison of the vertical displacement along the r-direction with Bulk modulus (Ka) kPa – – 145
Ai [19]. Viscosity (ηa) N⋅s/ – – 1.5075 ×
m2 10-5
( )
ρw s2 ρw gs ̃
pw =
− ∇̃ + W + ρw s2 u
̃ (13b)
nSr kw
0
d2 ̃
pw
[
ρa s 2
ρa gs ̃
] p0w − b12 ̃
− (ξ + b11 )̃ ε0 = 0
p0a − b13̃ (16a)
pa =
− ∇̃ + V + ρa s2 ̃
u (13c) dz2
n(1 − Sr ) ka
0
d2 ̃
pa
p0w − (ξ + b22 )̃
− b21 ̃ ε0 = 0
p0a − b23̃ (16b)
pw + A12 ̃
A11 ̃ u + A14 ∇⋅̃
pa + A13 ∇⋅̃ W=0 (13d) dz2

(13e)
0
pw + A22 ̃
A21 ̃ pa + A23 ∇⋅̃ ̃=0
u + A24 ∇⋅V d2̃ε
p0a − (ξ + b33 )̃ε0 = 0
p0w − b32 ̃
− b31 ̃ (16c)
dz2
By diversifying Eqs. (13b) and (13c) and substituting them into Eqs.
(13d) and (13e), the following equations are obtained. The solutions to the ordinary differential equations can be written as
follows:
2
pw − b11 ̃
∇̃ pa − b13 ε = 0
pw − b12 ̃ (14a)
∑
3
[ ]
p0w =
̃ fwn Cn eλn (z− h)
+ Dn e− λn z
(17a)
∇2 ̃
pa − b21 ̃ pa − b23 ε = 0
pw − b22 ̃ (14b) n=1

where b11 = ρw bw A11 , b12 = ρw bw A12 , b13 = ρw bw A13 − ρw s2 ; b21 = ∑

3
[ ]
ρa ba A21 , b22 = ρa ba A22 , b23 = ρa ba A23 − ρa s2 ; bw = s2 gs p0a = fan Cn eλn (z− h)
+ Dn e− λn z
(17b)
kw , ba =
+ ̃
nSr n=1
s2 gs
n(1− Sr ) + ka .
After substituting Eqs. (13b) and (13c) into Eq. (13a) and performing ∑
3
[ ]
ε0 =
̃ Cn eλn (z− h)
+ Dn e− λn z
(17b)
the divergence calculation, Eq. (15) can be obtained as follows: n=1

∇2 ̃
ε − b31 ̃pw − b32 ̃pa − b33̃ε = 0 (15) b13 dn − b13 b22 +b12 b23 b23 dn − b11 b23 +b13 b21
where fwn = d 2 ,f
− (b11 +b12 )dn +b11 b22 − b12 b21 an
=d 2
− (b11 +b12 )dn +b11 b22 − b12 b21
, n=
n n

where b31 = − b1 b11 +b2 b21

λ+2G , b32 = − b1 b12 +b2 b22
λ+2G ,and b31 = − b1 b13 +b
λ+2G
2 b23 +b3
; 1, 2, 3, Cn and Dn are the integration constant, h is the thickness, λn =
4 4
and b3 = − ρs2 + bww + ρba as .
s
2 2
further,b1 = − aχ + bsw , b2 = − a(1 − χ ) + bs a , ρ √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
dn 2 + ξ2 is the characteristic root of the equation, and dn is obtained
Applying the zero-order Hankel transform to Eq. (14) and (15), the
second-order ternary linear ordinary differential equations can be ob­ using Eq. (18).
tained. dn 3 − I1 dn 2 + I2 dn − I3 = 0 (18)

where I1 = b11 + b12 + b13 , I2 = b11 b22 + b22 b33 +

4
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 3. Curves of stress and displacement with depth at r = 0 m under different saturation: (a) vertical stress; (b) vertical displacement.

and Dz are the integration constant.

0
1 u
ε0 = ξ̃
Substituting Eqs. (17) and (20) into the relationship ̃
d̃
ur + dzz ,
the solution of the displacement in the r-direction can be obtained.

λ0 [ λ0 (z− ] 1∑ 3
[ ]
u1r = −
̃ Cz e h)
+ Dz e− λ0 z
+ (1 − fun λn ) Cn eλn (z− h)
+ Dn e− λn z
ξ ξ n=1
(21)
Applying the zero and first-order Hankel transforms to the normal
stress and shear stress components in Eq. (4), respectively, and
substituting based on Eqs. (17), (20), and (21), the mathematical for­
mulas for stress can be obtained as follows:

[ ] ∑
3
[ ]
σ 0z = 2Gλ0 Cz eλ0 (z−
̃ h)
+ Dz e− λ0 z
+ fzn Cn eλn (z− h)
+ Dn e− λn z
(22a)
n=1

( ) ∑
3 [ ]
b3 [ λ0 (z− ]
̃τ0zr = − G 2ξ − Cz e h)
− Dz e− λ0 z
+ frn Cn eλn (z− h)
− Dn e− λn z
Gξ n=1

(22b)
[ ( )]
Fig. 4. Curves of pore water pressure with subgrade depth at r = 0 m under where fzn = 2Gfun λn + λ − aχ fwn − a(1 − χ )fan , frn = Gξ λn − fun λn 2 + ξ2 .
different saturation. Applying the zero-order Hankel transform to the z-direction com­
ponents of the displacement in Eqs. (13b) and (13c) and then
b33 b11 − (b12 b21 +b23 b32 +b31 b13 ), I3 = b11 b22 b33 + b12 b23 b31 + substituting based on Eqs. (17) and (20), the vertical displacement of
b21 b32 b13 − (b11 b23 b32 +b22 b31 b13 +b33 b12 b21 ). water and air can be obtained.
A second-order ordinary differential equation can be obtained by
applying the zero-order Hankel transform to Eq. (13a) and substituting
2 [
̃ 0z = − s Cz eλ0 (z− h) λ0 z
] ∑3
fwn λn + ρw s2 fun [ λn (z− h) λn z
]
W − Dz e− − Cn e − Dn e−
based on Eqs. (17). bw n=1
ρw bw
0 (23a)
d2 ̃
uz ∑3
G u0z = −
− (Gξ − b3 )̃ λn [(λ
fan λn + ρa s2 fun [ λn (z− ]
dz 2 2 [ ] ∑3
n=1
[ ] ̃ 0 = − s Cz eλ0 (z−
V h)
− Dz e− λ0 z
− Cn e h)
− Dn e− λn z
(19)
z
+ G) + b1 fwn + b2 fan ] Cn eλn (z− h)
− Dn e− λn z ba n=1
ρa b a
(23b)
The solution for the vertical displacement in the transformed domain
can be obtained by solving Eq. (19). The relationship between the stress and displacement of the unsat­
urated subgrade can be expressed as a matrix formulation:
∑
3
[ ]
u0z = Cz eλ0 (z−
̃ h)
− Dz e− λ0 z
+ fun Cn eλn (z− h)
− Dn e− λn z
(20) {∑} { } { }
n=1 = [Ss ][Ds ]− 1 U
̃ = [K s ] U
̃ (24)
s s s
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅ [ ]
where λ0 = ξ2 − bG3 , fun = − λn
(λ + G) +b1 fwn +b2 fan , and Cz where [Ss ], [Ds ] are the stress and displacement matrices for the unsat­
Gλn 2 − Gξ2 +b3
urated subgrade, respectively; [Ks ] is the dynamic stiffness matrix for the

5
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 5. Maximum vertical displacement of the surface along the r-direction under different saturation: (a) pavement; (b) subgrade.

stress and displacement components of the different interfaces of the

multilayered system in the Laplace-Hankel transform domain,
respectively.
The pavement surface was only subjected to an FWD vertical load, as
shown in Fig. 1. The stress boundary conditions can then be written into
a mathematical expression as.
{
σz (r, z = 0, t) = P(r, t)
(26)
τzr (r, z = 0, t) = 0
By substituting the stress boundary conditions in Eqs. (26) into Eq.
(25), the displacements in the transformed domain at the three in­
terfaces can be obtained. Subsequently, eight integral constants can be
calculated. Furthermore, the dynamic response at any position inside
the multilayered system can be solved in the transformed domain by
substituting the integral constant into the corresponding mathematical
expression. Finally, the solutions in the space–time domain can be ob­
tained using a numerical integration method that achieves both the
Hankel inverse transform and Laplace inverse transform using Eq. (12b).

Model comparison and verification

Fig. 6. Schematic of the FWD system with a weak layer.

{ } The model was retrogressed into a saturated half-space state to verify

∑
unsaturated subgrade; and { }s and Ũ are the stress and its reliability and accuracy and compared with the calculation results of
Ai [19]. A circular harmonic load Peiωt was applied on the surface of the
s
displacement components of the upper interface of the subgrade in the
saturated half-space, where P = 10 kPa, ω = 20 rad/s, and the radius of
Laplace-Hankel transform domain, respectively.
the load R = 1 m. The relevant physical and mechanical parameters of
the saturated half-space soil are listed in Table 1.
Solution for the multilayered pavement-subgrade system
As shown in Fig. 2, the vertical displacement along the r-direction
was calculated and compared with the results obtained by Ai [19]. The
Assuming that the surface course, base course, and subgrade are
results show good consistency between the two methods, thus validating
entirely in contact, the stress and displacement for the different layers
the accuracy of the proposed model.
are continuous at the same interface. Then, based on Eqs. (1), (2), and
(24), and considering the total surface permeability of the unsaturated
Numerical results and analysis
subgrade, the relationship between the stress and displacement of the
multilayered system can be expressed by the following matrix formu­
A typical three-layer highway model was selected to study the dy­
lation:
namic characteristics of FWD loads under different conditions, as shown
{ } {∑} {∑}
̃ = [K]− 1
U = [B] (25) in Fig. 1. The entire highway structure is only subjected to the vertical
load of the FWD at the pavement surface without considering damping
and temperature. It is assumed that the layers are frictionless, and the
where [K] is the global stiffness matrix of the multilayered system with a
∑ { } subgrade surface is permeable. The related parameters of FWD with a
size of 8 × 8; [B] is the inverse matrix of [K]; and { } and U ̃ are the
duration of 30 ms and a peak load of 707 kPa (FWD standard load) and
the loading radius r0 = 0.15 m were selected from Ref. [34], where P(r,t)

6
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 7. Time history curve of vertical displacement of pavement surface for different elastic modulus of the weak layer: (a) r = 0 m; (b) r = 0.6 m; (c) r = 1.2 m; (d) r
= 1.8 m.

= 707sin(100πt/3)H(t)H(0.03-t), (0 m ≤ r ≤ 0.15 m), and H(t) is the step increasing depth. The vertical displacement increases with the increase
function. We selected the vertical surface displacement monitoring point in saturation, especially when saturation changes from 0 to 0.2 and 0.8
at a radial distance of 0, 0.2, 0.3, 0.45, 0.6, 0.9, 1.2, 1.5, and 1.8 m from to 0.999, as shown in Fig. 3(b). Because the deformation characteristics
the load center. The necessary physical and mechanical parameters of of dry, unsaturated, and saturated soils are different, soil saturation must
the elastic pavement and unsaturated subgrade were selected from the be considered in the assessment of pavement structure properties and
literature [24,25,34], as shown in Table 2. When the effect of a layer moduli back-calculation. The vertical displacement response can
parameter is examined, the other parameters are considered to remain be an effective index for reflecting the variation in the subgrade hu­
constant. midity state.
The pore water pressure can influence the stiffness and deformation
behavior of unsaturated subgrades. However, it is difficult to accurately
Influence of soil saturation
measure dynamic pore water pressure in field or model tests. The nu­
merical models developed in this study can be applied to investigate the
To fully comprehend the influence of saturation on the dynamic
pore water pressure inside a subgrade when subjected to traffic loads.
response of the road structure, the variations in stress, displacement, and
The distributions of the pore water pressure amplitude along the depth
pore water pressure with depth below the center of the FWD load under
with different soil saturations are shown in Fig. 4. It was found that the
different saturations were studied, as shown in Figs. 3 and 4.
peak value of the pore water pressure decreases as the soil saturation
As shown in Fig. 3, the influence of saturation on the vertical stress is
decreases, and the reduced amplitude is significant when soil saturation
negligible, whereas saturation has a significant effect on the vertical
drops from 0.999 to 0.8. This may be attributed to the much higher
displacement. The vertical stress decreases rapidly with the increase of
compressibility of pore air compared with pore water. Moreover, the
depth and mainly exists within 0.5 m, for all saturation levels. Unlike the
pore water pressure remains low level when the saturation is less than
vertical stress, the vertical displacement decreases more evenly with

7
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 8. Time history curve of vertical displacement of pavement surface for different thickness of the weak layer: (a) r = 0 m; (b) r = 0.6 m; (c) r = 1.2 m; (d) r =
1.8 m.

0.6. As regards the distribution of pore water pressure with the depth of when Sr varies from 0.8 to 0.999. The increase rates of the subgrade
the subgrade under a certain saturation, it can be observed that the surface are 12.9 % and 24.9 %, when Sr varies from 0.6 to 0.8 and 0.8 to
dynamic pore water pressure gradually decays after reaching the peak 0.999, respectively. The results imply that the subgrade performance
value near the surface of the subgrade. This is because the subgrade deteriorates with an increase in subgrade soil saturation. This adverse
surface was assumed to be fully permeable resulting in a pore water effect of saturation can be attributed to a decrease in the dynamic shear
pressure of 0 on the subgrade surface. modulus, as shown in Eq. (9). As aforedescribed, saturation has a sig­
When the FWD is used for pavement performance evaluation and nificant effect on the vertical displacement and pore water pressure.
layer modulus back-calculation, the deflection values of the pavement Thus, the unsaturated characteristics of subgrade soil must be consid­
surface are typically selected as inputs. As shown in Fig. 5, the distri­ ered when establishing a forward model to assess pavement structure
butions of the amplitude of the dynamic vertical displacement at the performance and back-calculation of the layer modulus.
pavement surface and subgrade surface along the r-direction with
different saturations were calculated. Evidently, the variations in the
Influence of subgrade deterioration
vertical displacement amplitude on the surface of the pavement and
subgrade are similar when the saturation changes and the impact of
Part of the service subgrade stiffness may be reduced owing to
saturation on the subgrade surface is more significant than that on the
insufficient construction quality, extreme climatic conditions, and
pavement surface. This may be because pore water directly affects the
continuous traffic overload. The reduction in the subgrade elastic
subgrade, whereas the effect on the pavement is indirect. The increase
modulus is the most direct characteristic of subgrade deterioration. A
rate of vertical displacement amplitude at r = 0 m of pavement surface is
weak layer on the upper part of the subgrade is established to study the
7.77 % as Sr increases from 0.6 to 0.8, while the value reaches 16.9 %
pavement surface displacement response after subgrade deterioration,

8
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 9. The increasing rate of vertical displacement of pavement surface along the r-direction for different parameters of the weak layer:(a) elastic modulus; (b)
thicknesses.

which is realized by changing the elastic modulus (Ed) and thickness (hd) parameters of the weak layer. The increasing rate of vertical displace­
of the layer, as shown in Fig. 6. In addition, four observations at r = 0, ment gradually increases to the peak value and then decreases as r in­
0.6, 1.2 and 1.8 m were chosen to study the sensitivity of the dynamic creases from 0 m to 1.8 m when the subgrade deteriorates. As can be
vertical displacement at different locations when the mechanical prop­ seen, this trend is more obvious as the elastic modulus of the weak layer
erties of the weak layer changed. decreases and the thickness of the weak layer increases. The increasing
Fig. 7 shows the time history curve of the dynamic vertical rate of vertical displacement reaches the peak values at r = 1.2 m with
displacement at different surface positions as the elastic modulus of the different Ed. Moreover, the increasing rate of vertical displacement
weak layer decreases. Four elastic moduli of the weak layer (Ed = 10, 30, reaches the peak values at r = 1.2 m when hd = 1.0 m and 0.7 m, while
50, and 70 MPa) were considered. The thickness sets as hd = 1.0 m, while the values of the increasing rate at r = 1.2 m and 0.6 m are very close
other parameters of surface course, base course, and no deterioration when hd = 0.4 m and 0.1 m. Therefore, the vertical displacement vari­
part of the subgrade are the same as those in Table 2 with Sr = 0.8. It can ation at r = 1.2 m is the most reasonable position to reflect subgrade
be observed that the vertical displacement at different positions deterioration.
increased as Ed decreased. The increasing amplitude of the vertical
displacement is only 4.07 % at r = 0.6 m when Ed decreases from 80 MPa
to 70 MPa. This implies that when the reduction in Ed is small, it is Influence of highway structure
challenging to identify the subgrade deterioration by the vertical
displacement of the pavement surface. Comparing the changes in the A model considered highway structures with different elastic moduli
peak values of the vertical displacement at different positions, it is found and thicknesses was developed to analyze the vertical displacement
that the vertical displacement response at r = 1.2 m is more sensitive to response of the pavement surface under the FWD load, and the results
the variation in Ed. The vertical displacement amplitude at r = 0 m are presented in Figs. 10 and 11, respectively. Fig. 10 shows the am­
increased 30.64 % as Ed decreased from 80 to 10 MPa. However, the plitudes of the vertical displacement of the pavement surface in the r-
increase of vertical displacement amplitude can reach 61.04 % at r = 1.2 direction as the elastic modulus of the surface course (E1), base course
m. Therefore, when FWD is used to identify subgrade deterioration, the (E2), and subgrade (E3). The magnitudes of the vertical displacement of
observation point for vertical displacement response is recommended to the pavement surface show a decreasing trend with an increase in the
be set at approximately 1.2 m away from the load center. In addition, it elastic modulus of each layer, and this effect tends to weaken with the
was discovered that the time to reach the peak displacement increased addition of the elastic modulus. The influence of E1 on the vertical
as r increased, which was attributed to wave propagation on the road. displacement of pavement surface is mainly concentrated within 0.6 m
Fig. 8 shows the vertical displacement at different positions as the of the load center, while the influence range of E2 is relatively larger,
thickness of the weak layer increased, and the elastic modulus was which extends to approximately 0.9 m. Further, the vertical displace­
selected as 10 MPa. The vertical displacement of pavement surface ment of the pavement surface of the whole monitoring area is affected by
shows an increasing trend when hd increases from 0.1 m to 1.0 m. The E3. This implies that the elastic modulus of each layer has a different
increase in the amplitude of the vertical displacement at all positions did influence on the vertical displacement of the pavement surface along the
not exceed 10 % when hd = 0.1 m. In addition, although the maximum r-direction; this finding can provide a theoretical basis for the back-
vertical displacement value appeared at r = 0 m, the increase in the calculation of the multilayered elastic modulus of the road. The verti­
amplitude of the vertical displacement was not the largest. By cal displacement of the pavement surface mainly reflects the properties
comparing the changes in the peak values of the vertical displacement at of the subgrade when r > 1.2 m.
different positions, the variation in the vertical displacement amplitude The variations in the vertical displacement amplitude of the pave­
at r = 1.2 m can also be applied to reflect the variation in the weak layer ment surface along the r-direction with different surface-course thick­
thickness. nesses (h1) and base-course thicknesses (h2) are shown in Fig. 11.
Fig. 9 shows the increasing rate of the vertical displacement of the Evidently, the vertical displacement of the pavement surface decreases
pavement surface as the radial distance increases for different with an increase in h1 and h2, and the influence range of both h1 and h2
on the vertical displacement response of the pavement surface is

9
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 10. Influence of the elastic modulus variation of each layer on the vertical displacement of pavement surface: (a) variation of E1; (b) variation of E2; (c) variation
of E3; (d) influence on r = 1.2 m.

approximately 1.2 m which is more extensive than that of the elastic thickness, and subgrade deterioration on the dynamic response. The
modulus. The aforedescribed analysis indicates that the elastic modulus following conclusions can be drawn based on the results of the present
and thickness of the highway structural layer significantly influence the study:
vertical displacement response of the pavement surface.
(1) The dynamic forward model can accurately predict the stress
Conclusions state and deformation characteristics at any position on the
highway and can thus provide a theoretical basis for the back-
In this study, a dynamic forward model of a multilayered pavement- calculation of the elastic modulus.
unsaturated subgrade system under an FWD load was developed to (2) Saturation has a negligible influence on the vertical stress dis­
calculate the dynamic response of the entire system. The pavement was tribution; however, it significantly affects the vertical displace­
simulated as two elastic layers with finite thickness. The subgrade was ment and pore water pressure. Therefore the unsaturated
modeled as a poroelastic half-space. The Laplace-Hankel transform and characteristics of the subgrade soil must be considered in the
dynamic stiffness matrix methods were employed to solve the governing back-calculation of the elastic modulus.
equations of the unsaturated soil layer. The analytical solutions of the (3) The vertical displacement of pavement surface induced by FWD
entire system were derived in the transform domain combined with the load can indicate subgrade deterioration, and the vertical
global stiffness matrix and boundary conditions. Finally, the space–time displacement at r = 1.2 m is most sensitive to subgrade
domain solutions are obtained using the Laplace-Hankel inverse trans­ deterioration.
form coupled with the numerical integration method. Several numerical (4) The elastic modulus of each layer has a significant impact on the
examples have revealed the impact of saturation, elastic modulus, dynamic vertical displacement of the road surface. The variation

10
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

Fig. 11. Influence of the thickness variation of each layer on the vertical displacement of pavement surface: (a) variation of h1; (b) variation of h2.

of elastic modulus of surface course mainly influences the vertical References

displacement of points within 0.6 m away from the load center.
The vertical displacement affected by the elastic modulus of the [1] Asli C, Feng ZQ, Porcher G, Rincent JJ. Back-calculation of elastic modulus of soil
and subgrade from portable falling weight deflectometer measurements. Eng Struct
base course is shown in a broader range (approximately 0.9 m 2012;34:1–7.
from the load center), while the subgrade effect is observed across [2] Ahmed MU, Tarefder RA. Incorporation of GPR and FWD into pavement
the entire monitoring area. This conclusion provides a precon­ mechanistic-empirical design. Constr Build Mater 2017;154:1272–82.
[3] Burak Goktepe A, Agar E, Hilmi Lav A. Advances in backcalculating the mechanical
dition for the back-calculation of the multilayered elastic properties of flexible pavements. Adv Eng Softw 2006;37:421–31.
modulus of highways. [4] Ruta P, Krawczyk B, Szydło A. Identification of pavement elastic moduli by means
of impact test. Eng Struct 2015;100:201–11.
[5] Picoux B, El Ayadi A, Petit C. Dynamic response of a flexible pavement submitted
In this study, the FWD load was simplified to a half-sine load, which by impulsive loading. Soil Dyn Earthquake Eng 2009;29:845–54.
can further simulate the FWD mechanical process comprehensively, [6] Lamb H. On the propagation of tremors over the surface of an elastic solid. Philos
considering the interaction between the falling weight, load plate, and Trans Royal Soc London. Ser A, Containing Papers of a Math Phys Character 1997;
203:1–42.
road. Second, the viscosity of the pavement was neglected in the
[7] Jones DV, Laghrouche, Petyt M. Ground vibration in the vicinity of a rectangular
calculation, and the effect of the viscosity of pavement materials such as load acting on a viscoelastic layer over a rigid foundation. J Sound Vibration, 203
asphalt on the FWD test results can be further investigated. (1997) 307-319.
[8] Yue ZQ, Xiao HT, Tham LG, Lee CF, Yin JH. Stresses and displacements of a
transversely isotropic elastic halfspace due to rectangular loadings. Eng Anal
CRediT authorship contribution statement Boundary Elem 2005;29:647–71.
[9] Han Z, Yang L, Fang H, Zhang J. Dynamic simulation of falling weight
Jing Zhang: Methodology, Software, Formal analysis, Writing – deflectometer tests on flexible transversely isotropic layered pavements. Soil Dyn
Earthquake Eng 2020;139:106353.
original draft. Zheng Lu: Supervision, Conceptualization, Funding [10] Grenier S, Konrad J-M, LeBœuf D. Dynamic simulation of falling weight
acquisition, Writing – review & editing. Chuxuan Tang: Data curation, deflectometer tests on flexible pavements using the spectral element method:
Writing – review & editing. Jie Liu: Visualization, Writing – review & forward calculations. Can J Civ Eng 2009;36(6):944–56.
[11] Ai ZY, Zhang YF. Plane strain dynamic response of a transversely isotropic
editing. Hailin Yao: Writing – review & editing. multilayered half-plane. Soil Dyn Earthquake Eng 2015;75:211–9.
[12] Cai Y, Hu B, Liu H. A method for dynamic response of a multilayered pavement
Declaration of Competing Interest loaded by FWD. ZAMM - J Appl Math Mechanics / Zeitschrift für Angewandte
Mathematik und Mechanik 2021;101(9).
[13] Park J, Kausel E. Impulse response of elastic half-space in the wave number-time
The authors declare that they have no known competing financial domain. J Eng Mech 2004;130:1211–22.
interests or personal relationships that could have appeared to influence [14] Lin G, Han Z, Li J. An efficient approach for dynamic impedance of surface footing
on layered half-space. Soil Dyn Earthquake Eng 2013;49:39–51.
the work reported in this paper.
[15] Liu H, Pan E. Time-harmonic loading over transversely isotropic and layered elastic
half-spaces with imperfect interfaces. Soil Dyn Earthquake Eng 2018;107:35–47.
Data availability [16] Ruta P, Szydło A. Drop-weight test based identification of elastic half-space model
parameters. J Sound Vib 2005;282:411–27.
[17] Biot MA. Mechanics of deformation and acoustic propagation in porous media.
Data will be made available on request. J Appl Phys 1962;33:1482–98.
[18] Lu J-F, Hanyga A. Fundamental solution for a layered porous half space subject to a
Acknowledgments vertical point force or a point fluid source. Comput Mech 2004;35:376–91.
[19] Ai Z, Wang L. Layer-element analysis of multilayered saturated soils subject to
axisymmetric vertical time-harmonic excitation. Appl Math Mech 2017;38:
This work was supported by the National Natural Science Foundation 1295–312.
of China (42077262 and 42077261). [20] Jin B, Yue ZQ, Tham LG. Stresses and excess pore pressure induced in saturated
poroelastic halfspace by moving line load. Soil Dyn Earthquake Eng 2004;24:
25–33.
[21] Fang X-Q, Yang S-P, Liu J-X, Zhang X-S. Dynamic response of road pavement
resting on a layered poroelastic half-space to a moving traffic load. Int J Numer
Anal Meth Geomech 2014;38:189–201.

11
J. Zhang et al. Transportation Geotechnics 38 (2023) 100866

[22] Ba Z, Kang Z, Lee VW. Plane strain dynamic responses of a multi-layered [28] Yao H, Hu Z, Lu Z, Wang H. Analytical model to predict dynamic responses of
transversely isotropic saturated half-space. Int J Eng Sci 2017;119:55–77. railway subgrade due to high-speed trains considering wheel-track interaction. Int
[23] Ba Z, Liang J. Fundamental solutions of a multi-layered transversely isotropic J Geomech 2016;16(2).
saturated half-space subjected to moving point forces and pore pressure. Eng Anal [29] van Genuchten MT. A closed-form equation for predicting the hydraulic
Boundary Elem 2017;76:40–58. conductivity of unsaturated soils. Soil Sci Soc Am J 1980;44(5):892–8.
[24] Fang R, Lu Z, Yao H, Luo X, Yang M. Study on dynamic responses of unsaturated [30] Ng CWW, Zhou C, Chiu CF. Constitutive modelling of state-dependent behaviour of
railway subgrade subjected to moving train load. Soil Dyn Earthquake Eng 2018; unsaturated soils: an overview. Acta Geotech 2020;15(10):2705–25.
115:319–23. [31] Chui CF, Ng CWW. A state-dependent elasto-plastic model for saturated and
[25] Tang C, Lu Z, Duan Y, Yao H. Dynamic responses of the pavement-unsaturated unsaturated soils. Géotechnique 2003;53:809–29.
poroelastic ground system to a moving traffic load. Transp Geotech 2020;25: [32] Seed HB, Idriss IM. Soil moduli and damping factors for dynamic response
100404. analyses; 1970.
[26] Shi L-W, Ma Q, Ma Y. Dynamic responses of unsaturated half-space soils to a strip [33] Fredlund DG, Vanapalli SK, Xing A, Pufahl DE. Predicting the shear strength
load at different boundary conditions. Arabian J Geosci 2021;14:1–11. function for unsaturated soils using the soil-water characteristic curve. in: 1st
[27] Ye Z, Ai ZY. Dynamic analysis of multilayered unsaturated poroelastic media International Conference on Unsaturated Soils. 1995.
subjected to a vertical time-harmonic load. Appl Math Model 2021;90:394–412. [34] Zhang J, Fan H, Zhang S, Liu J, Zheng J. Time-domain elasto-dynamic model of a
transversely isotropic, layered road structure system with rigid substratum under a
FWD load. Road Materials and Pavement Design 2021.

12