Salgado, R. et al. (2015). Géotechnique 65, No. 3, 169–187 [http://dx.doi.org/10.1680/geot.13.P.

124]

The role of soil stiffness non-linearity in 1D pile driving simulations

R . S A L G A D O , D. L O U K I D I S † , G . A B O U - JAO U D E ‡ a n d Y. Z H A N G 

Drivability studies and determination of pile static capacity from driving resistance are an integral part
of present foundation engineering practice. The success of both pile driving simulations and back-
calculation of static resistance from pile driving resistance (signal matching analyses) depends on the
model used for the interaction of the shaft and base of the pile with the soil. In this paper, a set of
advanced base and shaft dynamic reaction models is formulated that accounts explicitly for soil
stiffness non-linearity. Comparisons between predictions of the proposed soil reaction models and of
various existing reaction models used in practice with measured pile driving data from two well-
instrumented field cases allow assessment both of the importance of stiffness non-linearity in accurate
pile driving simulations and of the usefulness of various levels of sophistication in the shaft and base
reaction models.

KEYWORDS: dynamics; numerical modelling; piles

INTRODUCTION
Driven piles made of steel or concrete are commonly used Hammer mass Ram v0
as foundation elements in engineering projects. Installation
of these piles requires selection of driving systems that will
be effective in driving the piles to required depths without
damaging them. This is done in what is usually referred to
as a drivability study. For design purposes, it is also im-
portant to correlate the static pile resistance to the resistance
to driving, which then allows decisions on when to stop mi
driving a pile, for example. These two needs have led to Lumped pile
masses
studies of theoretical modelling of the pile driving process. Soil reaction model
Smith (1960), in the first numerical study of pile driving, along pile shaft
discretised the pile into segments, each of which is modelled Interconnecting
pile springs
as a mass connected to the segments above and below
through linear elastic springs corresponding to the axial
stiffness of the pile (Fig. 1). In the Smith model, the soil
reaction acting on the perimeter of any pile segment and at
the pile base (bottom segment) results from a linear elastic,
perfectly plastic spring-slider and a dashpot connected in Soil reaction
model at pile
parallel. The hammer (ram) is modelled as one or more base
elements with suitable inertia and stiffness. A hammer blow
is applied through an initial velocity imposed on the ham-
mer, which then propagates down the pile one-dimensionally Fig. 1. Formulation of one-dimensional (1D) dynamic analysis:
as a mechanical wave. This type of analysis is often referred discretisation of pile into lumped masses with soil reactions
to as a wave equation analysis. In this paper, the same applied to each lumped mass
lumped-mass approach is used, but with different models,
described later, for the shaft and base reactions.
Over the years, a number of improved and less empirical detail on the soil behaviour under dynamic loading and more
soil reaction models have been proposed for use in vertical rigorous treatment of radiation and hysteretic damping, soil
pile response analysis (e.g. Holeyman, 1985; Randolph & stiffness non-linearity and rate effects on soil strength.
Simons, 1985; Lee et al., 1988; Nguyen et al., 1988; EI- In this paper, a set of shaft and base soil reaction models
Naggar & Novak, 1994; Holeyman et al., 1996; Michaelides is developed that takes soil non-linearity and hysteresis
et al., 1998; Vanden-Berghe & Holeyman, 2002). These explicitly into account. The input parameters for the soil
models have typically grown in sophistication, with better reaction models have physical meaning and are well an-
chored to standard soil properties. The response of two well-
instrumented piles driven at different sites is simulated using
Manuscript received 19 July 2013; revised manuscript accepted 9 the advanced shaft and base reaction models, as well as a
January 2015. Published online ahead of print 20 March 2015. number of simpler, existing reaction models. Comparison of
Discussion on this paper closes on 1 August 2015, for further details the simulation results with the pile driving data allows an
see p. ii. assessment of the ability of the proposed and existing
 School of Civil Engineering, Purdue University, West Lafayette, IN,
USA.
reaction models to accurately predict the observed pile
† Department of Civil and Environmental Engineering, University of response. The comparison also allows examination of the
Cyprus. importance of accurately modelling various features of soil
‡ Department of Civil Engineering, Lebanese American University, behaviour, such as soil stiffness non-linearity, in the pile
Lebanon. driving analysis.

169

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 170 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
SOIL REACTION MODELS small increments ˜t. Using the central finite-difference
Shaft reaction model approximation, the second derivative appearing in equation
Near-field response. The shaft reaction model proposed in (1) for node i is given by
this paper, which produces the unit shaft resistance applied
on each of the lumped masses into which the pile is divided, @ 2 w w(tþ˜t)
i  2w(t) (t˜t)
i þ wi
 (2)
follows the outlines of the model proposed by Holeyman @t2 ˜t2
(1985), Holeyman & Legrand (1994) and Vanden-Berghe &
Holeyman (2002). The proposed model considers soil non- Each pair of adjacent nodes in Fig. 2 defines a segment
linearity, radiation damping, hysteretic damping and rate where the strain and thus the stress are assumed to be
effect on strength, with the soil modelled as a continuum. constant. The central difference approximation of the first
Vanden-Berghe & Holeyman (2002) modelled the soil mass derivative of the shear stress for node i is expressed as
surrounding the pile as concentric solid cylinders instead of (t) (t)
springs, dashpots and sliders. These cylinders extend @ R,i  L,i
 (3)
vertically throughout the length of the pile, thus requiring @r ˜r
averaging of the soil properties along the pile shaft. This also
implies that all segments of the pile are assumed to move in where (t) (t)
L,i and R,i are the shear stress values within the
phase, in other words, that the pile is rigid. In order to be segments on the left (inner) and right (outer) sides of node i,
able to consider motion phase differences along the pile and respectively.
soil layers of various types with different soil properties, the The soil is assumed to follow a ‘hyperbolic’ stress–strain
present formulation uses thin soil discs instead of cylinders. law in rate form
This approach has been used to simulate the soil response Gmax
around the pile shaft under static conditions (e.g. Randolph & _ ¼  2 ª_ (4)
Wroth, 1978; Potts & Martins, 1982; Loukidis & Salgado, j  LI rev j
1 þ bf
2008). According to this approach, the soil around the pile (LI þ 1) 3 jsgn(ª)
_ f  j
shaft is divided in thin horizontal discs with an annulus
(occupied by the pile) of radius rpile (Fig. 2). The thin soil where Gmax is the maximum (small-strain) shear modulus, f
disc has an outer radius rf. The equation of vertical motion of is the shear strength of the soil in simple shear conditions,
a soil element at any radial position r and time t can be rev is the shear stress at the last stress reversal, LI is a
written as loading index parameter that takes the values 0 for virgin
@  @2w loading (loading starting from zero shear stress) and 1 for
þ ¼r 2 (1) subsequent unloading and reloading, and the variable sgn(ª) _
@r r @t is the sign of the strain rate ª:
_ The loading index LI follows
where w is the vertical soil displacement,  is the vertical from the second Masing rule (Kramer, 1996); it helps
shear stress, r is the soil density, while the term @z /@z, achieve closed and symmetrical stress–strain loops during
which is negligibly small, is omitted. The signs in equation cycling. The parameter bf controls the rate of degradation of
(1) are consistent with the positive directions (Fig. 2) the shear modulus. The well-validated experimental shear
adopted for stress, displacement and depth z. Equation (1) modulus degradation curves of Vucetic & Dobry (1991) can
can be solved analytically if it is assumed that the material be approximated adequately by equation (4) with bf calcu-
is linear elastic with constant damping ratio and the soil lated using the following equation (Loukidis et al., 2008)
motion is a steady-state harmonic oscillation (Novak et al.,
bf ¼ 5 exp [0.05(PI)] (5)
1978; Michaelides et al., 1998). Since the goal here is to
produce a formulation that can accommodate soil non- where PI is the plasticity index in percentage units. For PI
linearity, variable hysteretic damping ratio and any type of from 0 to 60%, bf ranges from 5 to 0 .25.
motion – especially that due to an impact-type pulse propa- The strength f (equation (4)) inside the continuum
gating down the pile – equation (1) must be solved numeri- depends on the strain rate according to the following loga-
cally. For this purpose, Ns points (soil nodes) are assumed rithmic law (Graham et al., 1983; Kulhawy & Mayne, 1990;
along the radial axis, separated by distance ˜r, from the pile Sheahan et al., 1996)
radius rpile to the outer disc radius rf (Fig. 2). The discretisa-   
ª_
tion of the soil disc must be fine enough to capture the f ¼ f ,stat 1 þ ºs log (6)
details of the propagating wave form. Time is divided into ª_ ref

rpile
vibration

r G, ρ
Pile

τ
Horizontally travelling ∂τ
vertical shear waves τ⫹
∂r

Radius of pile influence, rm

Extent of the analysis domain, rf

1 2 i⫺1 i⫹1 Ns

rpile Δr wi Soil nodes

ri

Fig. 2. Soil disc around pile shaft and finite-difference discretisation

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 171
where ª_ ref is a reference shear strain rate taken equal to Having established the shear stress values, the finite-
1%/h (Kulhawy & Mayne, 1990; Einav & Randolph, 2006). difference approximation of equation (1) can be written as
For clays, a value of 0 .1 is adopted for the parameter ºS
(Kulhawy & Mayne, 1990; Sheahan et al., 1996), while a (t) (t)
R,i  L,i (t) w(tþ˜t)  2w(t) (t˜t)
i þ wi
þ i ¼r i (13)
value of 0 .05 is assumed for non-plastic soils based on the ˜r ri ˜t 2

experimental observations of Omidvar et al. (2012).

The static soil strength of the soil f,stat is set to be greater Provided that the displacements at all the nodes at times t
than the strength of the shear band so that the slider engages and t  ˜t are known, it is possible to calculate the
at a shear strain level at the first near-field segment in the displacement at time t + ˜t at any node i by solving
5–10% range. This is the shear strain range within which equation (13) for wi(t+˜t )
soils are expected to reach peak strength under simple shear " #
2 (t)  (t) (t)
conditions. In order to achieve this ‘failure strain’, in the ˜t R,i L,i 
wi(tþ˜t) ¼ þ i þ 2w(t)
i  wi
(t˜t)
(14)
case of clays and silts, f,stat is defined by the following r ˜r ri
equation
f ,stat ¼ RF qsL (7)
where qsL is the unit limit shaft resistance and the factor RF Far-field-consistent boundary conditions. The discretised soil
(.1) is set to be a function of the parameter bf (equation disc represents only part of the soil around the pile (the near
(5)) and, consequently, PI field). The presence of the outer soil mass (far field) needs to
be represented in the model formulation because it absorbs
1 the waves travelling radially in the soil disc away from the
RF ¼ . (8)
1  0.015b1f 5 pile. As in Vanden-Berghe & Holeyman (2002) and Charue
(2004), a radiation dashpot with constant cFF following Novak
et al. (1978) is placed at the outer boundary of the near field
In the case of sands (PI ¼ 0, bf ¼ 5), f,stat is defined by
(Fig. 3)
the following equation
cFF ¼ rV s (15)
f ,stat ¼ 1.2nF qsL (9)
where Vs is the soil shear wave velocity. The absorbing
where the factor nF is introduced to account for the effect of boundary formulations commonly employed in finite-
roughness of the shaft wall on the pile–soil interface difference or finite-element analysis of seismic problems
strength. It takes the values 1 .0 and 1 .22 for piles made of consist only of radiation dashpots. This approach is not
concrete and mild steel, respectively. suitable for pile driving analysis, because the static solution
The stress–strain relationship expressed by equation (4) should be recovered after all vibrations have dissipated; that
can be seen as a variation of the f–g model (Fahey & Carter, is, there must be a permanent soil displacement with magni-
1986, 1993; Foray et al., 1998; Dos Santos & Correia, 2001; tude decreasing with increasing radial distance from the pile
Lee et al., 2004). Using the hyperbolic stress–strain model, axis caused by the residual shear stress on the pile shaft.
the shear stresses L and R at time t are calculated based on This is achieved by the addition of a linear spring with
their values at time t  ˜t and on the values of the displace- stiffness kFF at the outer boundary of the soil disc (Fig. 3).
ment rates in the time increment from t  ˜t to t using the The static solution derived by Randolph & Wroth (1978)
following equations leads to a spring constant kFF given by
Gmax Gmax 1
(t)
L,i ¼ 2   32 k FF ¼ (16)
 (t˜t)  rf ln (rm =rf )
 L,i  LI L,i L,rev,i 
6  7
41 þ bf  (t˜t) 
5
(LIL,i þ 1) 3 sgn(ª)
_ f  L,i  (10) where rf is the radius of the outer boundary of the near-field
soil disc (Fig. 3). The parameter rm (Fig. 2) is the radius of
pile influence (the so-called ‘magical’ radius); it represents
(w(t) (t)
i  wi1 )  (wi
(t˜t)
 w(t˜t)
i1 ) (t˜t)
3 þ L,i the radial distance from the pile axis at which the displace-
˜r ment is zero (negligible, for practical purposes) and can be
Gmax estimated by the following equation (Randolph & Wroth,
(t)
R,i ¼ 2   32
 (t˜t)  1978)
 R,i  LI R,i R,rev,i 
6  7 rm ¼ 2.5L(1  ) (17)
41 þ bf  (t˜t) 
5
(LI R,i þ 1) 3 sgn(ª) _ f  R,i  (11)

(w(t) (t) (t˜t)

iþ1  wi )  (wiþ1  w(t˜t) ) (t˜t)
3 i
þ R,i
˜r Pile Shear band behaviour

The shear stress at the position of node i is calculated in Far field

a similar way CFF

Gmax
(t)
i ¼ 2   32 1 2 i⫺1 i i⫹1 Ns
 (t˜t) 
 i  LIi rev,i  KFF
6  7
41 þ bf  (t˜t) 
5 Near-field response
(LIi þ 1) 3 sgn(ª)_ f  i  (12)
Fig. 3. Proposed shaft reaction model consisting of three parts:
continuous near field, far-field consistent spring and radiation
(w(t)  w(t)
i1 ) (w(t˜t)  w(t˜t)
i1 )
3 iþ1 iþ1
þ (t˜t)
i
dashpot, and rheological shear band model at the soil–pile
2˜r interface

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 172 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
where L is the length of the pile and  is the Poisson ratio (t) (t)
R,1  sf (t) w(tþ˜t)  2w(t) (t˜t)
1 þ w1
of the soil. þ sf ¼ r 1 (20)
˜r=2 r1 ˜t2
If rf is set to 0 .102rm, kFF becomes identical to the
analytical solution of Novak et al. (1978) for steady-state and solved with respect to w1(t+˜t )
harmonic motion. On the other hand, if rf ¼ rm, equation !
2 (t)  (t) (t)
(16) yields an infinitely large spring stiffness, thus negating ˜t R,1 
w(tþ˜t) ¼ sf
þ sf þ 2w(t) (t˜t)
1  w1 (21)
the effect of the radiation dashpot. The near-field domain 1
r ˜r=2 r1
needs to extend only up to the point where the shear strains
become small enough for the soil non-linearity to cease to The denominator ˜r/2 in equation (2), instead of the ˜r
be significant. Based on trial analyses, it was established that appearing in equation (14), reflects the fact that node 1
this happens for values of the ratio rf /rm in the 0 .2–0 .5 where sf acts is at a distance ˜r/2 from the centre of the
range. In the analyses presented subsequently in this paper, a first segment where 1,R is operative. Slippage in the slider
value of 0 .35 is used. Given what is known today about stops and the soil disc rejoins the pile once the relative
shear strain localisation along the pile shaft, even this num- velocity between the pile segment and node 1 returns to zero
ber may be high. and, according to equation (19), sf becomes equal to qsL. In
Having established the boundary conditions at the outer other words, the effect of the slider vanishes when the
edge of the soil disc, equation (14) becomes mobilised shear stress at the interface drops below the static
2 ! limit.
w(t)  w (t˜t) The main output of the soil disc computations is the soil
6 N N
þ k FF wN s  (t)
(t) reaction at the pile shaft wall. Knowing the shear stress that
2 4 cFF
s s

˜t ˜t L,N s
acts on the pile shaft (given by either equation (12) or
w(tþ˜t) ¼
Ns
r ˜r=2 equation (19)), the shear force acting on each shaft segment
!3 can be computed as
w(t)  w(t˜t) n o
 cFF N s Ns
þ k FF w(t) 7 R(t)  (t)
 —pile ˜L
Ns 7 ¼ min , sf (22)
˜t 5 s 1
þ
rN s where —pile is the perimeter of the pile cross-section and ˜L
is the length of the pile segment. These shaft reaction forces
þ 2w(t)
Ns  w(t˜t)
Ns are then used in the calculation of the motion of the pile
(18) lumped masses described in the Appendix.

for the outmost (i ¼ Ns) point. Because the radius at the

outer boundary rN s is much larger than ˜r/2, the second Base reaction model
term in the brackets could, in fact, be omitted without loss General approach. The base failure mechanism has a size of
of accuracy, as observed in trial analyses. roughly one to two pile diameters (e.g. Salgado, 2012;
Arshad et al., 2014). Therefore, in contrast to the very thin
shaft shear band, the base failure mechanism encompasses a
substantial part of the near-field soil, in which waves of
Shear band at soil–pile interface. The rheological model significant amplitude propagate; the development of limit
proposed by Simons & Randolph (1985), consisting of a base resistance cannot be decoupled from the mechanics of
plastic slider and a viscous dashpot connected in parallel, is the near field. Hence, the base resistance model follows a
placed between the first node of the soil disc and the traditional macro-element approach of a spring and dashpot
corresponding pile segment (Fig. 3). Whenever there is no connected in parallel instead of a continuum approach.
slippage between soil and pile (i.e., the rheological model is The proposed base model consists of a non-linear spring
inactive), the soil reaction at the pile shaft wall, s, is taken in parallel with a radiation dashpot (Fig. 4) connected to the
as the shear stress in the first segment (segment between bottom pile lumped mass. The total base reaction is the
nodes 1 and 2) of the soil disc, 1,R. sum of the spring reaction Rb(S) and the radiation dashpot
The limit resistance of the plastic slider is equal to the reaction Rb(D)
static unit limit shaft resistance qsL. Sliding initiates once
1,R ¼ qsL, at which point the viscous dashpot is activated. Rb ¼ Rb(S) þ Rb(D) ¼ R(S)
b þ Cbw
_b (23)
The reaction of the viscous dashpot is a power function of
the relative velocity across the shear band (between the pile where w_ b is the velocity of the pile base and Cb is the
and the first node of the near field component). The radiation damping constant.
total (static + viscous) resistance of the rheological model is
given by
"   #
w_ pile  w_ 1 ns Pile
sf ¼ qsL 1 þ ms (19)
vref

Non-linear spring
where ms and ns are input parameters and vref is a reference with rate effect on strength
Radiation dashpot
relative velocity equal to 1 m/s. .
Rb w
Once slippage initiates, the displacement of the first node
of the soil disc component (node 1) and the pile displace-
ment become different. Hence, the displacement of node 1
ceases to be a known input and its value needs to be wb
calculated as part of the solution, while 1 is now equal to
sf and constitutes the known input. So, during slippage, Fig. 4. Proposed base reaction model consisting of a non-linear
equation (13) can be rewritten for node 1 as spring and a radiation dashpot

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 173
Non-linear base spring. The spring stiffness is formulated so 3.4R2 pffiffiffiffiffiffiffiffiffiffiffiffiffi 3.4R2
that the quasi-static response is consistent with the base load– C Lysm ¼ rGmax ¼ rV s,max (28)
1 1
settlement curve that is observed in static pile load tests. The
non-linear spring follows a ‘hyperbolic’ load–settlement In the proposed model, the radiation dashpot reaction R(D)
b
relationship in rate form (equation (23)) is given by the following equation
(S) K b,max R(D)
b ¼ (C Lysm cemb chys )w
_ b,el (29)
R_ b ¼  2 w_ b (24)
jRb  LIb Rb,rev j
1 þ bfb where w_ b,el is the elastic component of the pile base velocity,
(LIb þ 1) 3 jsgn(w_ b )RbL  Rb j
which, according to equation (24), can be calculated as
where Kb,max is the maximum (small-strain) base spring stiff- 1
ness, RbL is the limit base capacity, Rb,rev is the spring reaction w_ b,el ¼  2 w_ b (30)
jRb  LIb Rb,rev j
R(S)
b at the last displacement reversal, and LIb is a loading 1 þ bfb
index parameter that takes the values 0 for virgin loading and (LIb þ 1) 3 jsgn(w_ b ) RbL  Rb j
1 for subsequent unloading and reloading. The base reaction
reaches RbL asymptotically with increasing wb. The variable The velocity w_ b,el is used instead of w_ b in equation (30)
sgn(w_ b ) is the sign of the base velocity w_ b , and bfb is a model because the wave energy released by the pile base to the far-
parameter that controls the rate of degradation of the base field soil mass decreases as the plastic mechanism develops
spring stiffness. The model captures the energy loss inside the under the base. After the plastic zone around the base is
failure mechanism due to material hysteresis with the help of fully formed, there is only minimal transmission of energy
the variables LIb, sgn(w_ b ) and Rb,rev. At points of displace- through the lateral boundaries of the plastic mechanism.
ment reversals (onset of unloading or reloading), the spring Hence, the influence of radiation damping should decrease
stiffness is equal to the initial stiffness values Kb,max. Because with increasing deformation inside the plastic zone.
tensile stresses cannot be transmitted through the soil–pile The factor cemb captures the effect of embedment on
base interface, R(S)b is not allowed to take negative values.
radiation damping. The radiation dashpot coefficient takes
The curvature of the base reaction plotted against settle- also into account the influence of hysteretic damping inside
ment curve depends on the pile type in addition to the the far-field soil through the factor chys. Expressions for the
intrinsic and state variables of the soil. The settlement factors cemb and chys are proposed later based on analyses
required for displacement piles embedded in clay to reach using the software FLAC.
limit base resistance can be as low as 0 .1B to 0 .2B. In Based on equation (30), the radiation dashpot coefficient
contrast, very large displacements are required to achieve for the pile base in equation (23) becomes
the limit base resistance in non-displacement piles in sand C Lysm cemb chys
(e.g. Lee & Salgado, 1999, 2000). The parameter bfb needs Cb ¼  2 (31)
jRb  LIb Rb,rev j
to be in the 1 .0–2 .0 range for a displacement pile and the 1 þ bfb
10–20 range for non-displacement piles in order to produce (LIb þ 1) 3 jsgn(w_ b )RbL  Rb j
results consistent with experimental observations of base
resistance against base settlement. According to equations (29) and (31), the base radiation
The rate dependence of soil strength is included in the dashpot coefficient decreases gradually towards zero as the
model through RbL. The limit base resistance RbL is set to be base reaction Rb approaches the limit resistance RbL.
a function of the base velocity
"  nb #
w_ b Effects of base embedment on stiffness and damping coeffi-
RbL ¼ QbL 1 þ mb (25) cients. The embedment effect on Kb,max and Cb is accounted
vref
for in equations (26) and (31) through the factors Df,dyn and
cemb and chys, respectively. In order to develop expressions for
where mb and nb are input parameters. For zero base
these factors, the present authors performed static and
velocity, RbL is equal to QbL, which is the limit base
dynamic analyses of a circular rigid footing, representing
capacity under quasi-static loading.
the pile base, embedded to depth D in an elastic half-space
The maximum base spring stiffness is expressed as
(G ¼ 100 000 kPa, v ¼ 0 .15, r ¼ 1 .8 t/m3) using the program
K b,max ¼ K Lysm Df ,dyn (26) FLAC. Typical mesh and boundary conditions are shown in
Fig. 5. The footing had diameter B ¼ 0 .5 m and mass
The term KLysm corresponds to the spring stiffness for a M ¼ 1 .96 t. This combination of G and M leads to a natural
circular footing resting on an elastic half-space and is the period of oscillation of the system in the FLAC analyses that
spring stiffness used in the classical Lysmer model (Lysmer is roughly the same as the one observed in pile driving (e.g.
& Richart, 1966) around 0 .02 s in the two case histories presented later). The
4RGmax frequency of the pile–soil system is in a range for which the
K Lysm ¼ (27) dimensionless frequency factor øB/Vs is less than 1 .0. For
1 such value of øB/Vs, the effect of frequency on the spring
where R is the pile radius. stiffness and dashpot constants is very small (Gazetas, 1991).
The depth factor Df,dyn introduces the effect of depth of Results of static analyses in terms of depth factor static
embedment on the spring stiffness under dynamic loading stiffness Df (Fig. 6) can be fitted closely by the following
conditions. An expression for Df,dyn is proposed later based equation.
on analyses using the software FLAC (Itasca Consulting ( "  0.83 #)
Group, 1998). D
Df ¼ 1 þ (0.27  0.12 ln ) 1  exp 0.83
B
(32)
Radiation damping. Most existing base reaction models (e.g.,
Holeyman, 1985; Simons & Randolph, 1985) employ the Predictions made using equation (32) agree well with
radiation dashpot coefficient of the Lysmer analogue existing solutions for embedded flexible footings (Fox, 1948;

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 174 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
B/2 1·6

1·5

Embedment, D 1·4

Df 1·3

Initial velocity
FLAC: ν ⫽ 0·15
1·2
FLAC: ν ⫽ 0·33
Impact mass M FLAC: ν ⫽ 0·49
at base of a
cylindrical hole Equation (32): ν ⫽ 0·15
1·1
Absorbing Equation (32): ν ⫽ 0·33
boundary Equation (32): ν ⫽ 0·49
Elastic material
with hysterisis 1·0
G, ν, ρ, ξ 0 5 10 15 20 25
Absorbing D/B
boundary
Fig. 6. Variation of depth factor Df for static footing stiffness with
embedment ratio D/B
1·0 m 2·0 m 3·0 m 4·0 m
C
L
matching the observed response to that predicted by the
Fig. 5. Mesh and boundary conditions used in dynamic FLAC single degree-of-freedom mass–spring–dashpot analogue.
analysis of circular footing at the bottom of cylindrical well The values of Df,dyn ¼ K/KLysm can be fitted by
.
Df ,dyn ¼ (Df )1 6 (33)
Burland, 1970; Pells & Turner, 1978) and embedded rigid
footings in an unsupported well (Gazetas et al., 1985).
while the values of the ratio chys ¼ C/C ¼ 0 can be fitted by
In dynamic analyses, the footing is excited by an initial
the equation proposed by Dobry & Gazetas (1986) based on
vertical velocity, and the ensuing oscillations are allowed to
the principles of correspondence
dissipate by radiation and hysteretic damping. Stiffness-
proportional Rayleigh damping is assigned to the soil in  
2K b,max =(C Lysm cemb )
order to simulate hysteretic damping. It is common in the chys ¼ 1 þ  (34)
analysis of machine foundations to assume  between 2% ø
and 5% (Richart et al., 1970). Here, the upper value of 5%
was adopted since the amplitude of pile base motion oscilla- The fundamental period of the pile–soil system is con-
tions are in the upper range of what could be observed in trolled by the base spring stiffness, since it is much smaller
foundation vibration problems. than the longitudinal pile stiffness. Hence, a first approxima-
The dynamic simulation results are summarised in Table tion for the ø of the base motion can be calculated as
1. The resulting values of K and C are determined by follows

Table 1. Summary of results from Flac simulations of embedded oscillating footing

D/B Hyst. damp, : % Df K: kN/m C: kN s/m K/KLysm C/CLysm new C/Cemb. ¼ 0 C/C ¼ 0

0 0 .01 1 123 529 121 1 .05 1 .14 1 .00 1 .00

0 5 1 123 529 154 1 .05 1 .62 1 .00 1 .42
0 15 1 123 529 212 1 .05 2 .55 1 .00 2 .24
0 30 1 123 529 308 1 .05 4 .00 1 .00 3 .51
0 .5 0 .01 1 .19 171 765 207 1 .40 1 .95 1 .71 1 .00
1 0 .01 1 .28 194 118 244 1 .65 2 .30 2 .02 1 .00
1 5 1 .28 194 118 280 1 .65 2 .95 1 .82 1 .28
1 15 1 .28 194 118 353 1 .65 4 .30 1 .69 1 .87
1 30 1 .28 194 118 467 1 .65 6 .50 1 .63 2 .82
2 0 .01 1 .38 223 529 196 1 .90 1 .85 1 .62 1 .00
2 5 1 .38 223 529 244 1 .90 2 .65 1 .64 1 .43
5 0 .01 1 .48 223 529 133 1 .90 1 .25 1 .10 1 .00
5 5 1 .48 223 529 175 1 .90 1 .90 1 .17 1 .52
5 15 1 .48 223 529 255 1 .90 3 .40 1 .33 2 .72
5 30 1 .48 223 529 377 1 .90 5 .80 1 .45 4 .64
10 0 .01 1 .5 229 412 130 1 .95 1 .23 1 .08 1 .00
10 5 1 .5 229 412 173 1 .95 2 .00 1 .23 1 .63
10 15 1 .5 229 412 255 1 .95 3 .50 1 .37 2 .84
15 0 .01 1 .5 229 412 130 1 .95 1 .23 1 .08 1 .00
15 5 1 .5 229 412 174 1 .95 2 .00 1 .23 1 .63

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 175
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
0.5K b,max histories, Lagrange County and Jasper County, which are
ø (35) described in the two preceding sections. In addition to the
M pile
proposed shaft and base models, simulations are performed
with various existing shaft and base reaction models and
where Mpile is the mass of the pile. The stiffness reduction their combinations; the results are then compared in the
factor of 0 .5 is introduced to approximately account for the ‘Results of pile driving simulations’ section. The existing
non-linearity of the base spring. In case the frequency ø of reaction models considered in this comparison are shown in
the resulting pile base motion is substantially different from Fig. 7. Apart from the classical Smith (1960) model (re-
the value yielded by equation (35), the analysis could be ferred to henceforth as J-Q), there are two shaft reaction
repeated using the observed ø. The difference between the models – S1, proposed by Simons & Randolph (1985) and
ø observed in the simulations of the field cases presented Randolph (2003), and S2, which determines soil yielding
later in the paper and the corresponding values yielded by based solely on the spring force, like the model variants by
equation (35) does not exceed 25%. Lee et al. (1988) and Nguyen et al. (1988) – and two base
The ratio cemb ¼ C/CLysm for a given  increases rapidly reaction models (B1 and B2).
towards a peak value at D/B ¼ 1, and then decreases towards The base model B1, proposed by Deeks & Randolph
a plateau. This trend is captured adequately by the following (1995), includes an added mass m0 and an auxiliary mass
equation m1, following Wolf (1988). For v ¼ 0 .5 (fully saturated soil
at pile base, as is the case in both field cases investigated),
2(D=B)  0.15
cemb ¼ 1.15 þ (36) the mass m1 and the corresponding spring and dashpot
exp (D=B) vanish. The base model B2 (without m0) was proposed by
Simons (1985) and Simons & Randolph (1985). The Simons
& Randolph (1985) model, referred to as model B2
Determination of model parameters henceforth, assumes that the radiation dashpot disconnects
The input parameters directly used in the shaft and base once the slider gets activated. In the analyses shown in the
reaction models are: (a) soil density, r; (b) small-strain shear following section, the mass m0 was added to the bottom
modulus, Gmax; (c) Poisson ratio, ; (d ) plasticity index, PI; (base) pile lumped mass, irrespectively of base reaction
(e) pile diameter, B, and length, L; ( f ) static unit shaft model (including the proposed hyperbolic model).
resistance, qsL; (g) static unit base resistance, qbL; and (h) Two sets of analyses were performed using models S1,
soil viscosity parameters, ms, mb, ns, nb. The small-strain S2, B1 and B2: one with shear modulus G equal to the Gmax
shear modulus for sandy and clayey soils can be calculated calculated using the Hardin & Black (1968) equations, and
using either the Hardin & Black (1968) equations or cone the other with G equal to a secant shear modulus G1
penetration test (CPT)-based empirical correlations. Esti- following Lee et al. (1988): G1 ¼ 150su (su is the undrained
mated values of Gmax correspond to in-situ conditions, shear strength) for clayey soils and G1 ¼ 200 v9 for sandy
before pile installation, not reflecting changes induced to the soils. G1 calculated in this manner is roughly 10–50% of
soil density and stress state by the pile installation. Presently, Gmax.
reliable prediction of such changes would require advanced The simulations using the Smith (1960) reaction models
finite-element analyses. Nonetheless, trial analyses using the assumed the soil quake to be 2 .5 mm at the pile shaft and
proposed models have shown that the small differences in B/60 at the pile base. The corresponding damping values
Gmax values due to the use of different estimation methods were taken as Js ¼ 0 .16 s/m and 0 .65 s/m for coarse- and
for its estimation are not of critical importance. fine-grained soils, respectively, and Jb ¼ 0 .5 s/m.
The Poisson ratio used in the model is the small-strain
value, assumed equal to 0 .15 for sandy soils and 0 .22 for
clayey soils, unless the soil is saturated and loading un- First case history: Lagrange County, IN, USA
drained, in which case it is equal to 0 .5. The parameters ns The first case history involves the installation and testing
and nb fall in a relatively narrow range, with most values of an instrumented closed-ended steel pipe pile for a bridge
falling in the vicinity of 0 .2 (Coyle & Gibson, 1970; Dayal project in Lagrange County, Indiana (Paik et al., 2003). The
& Allen, 1975; Heerema, 1979; Litkouthi & Poskitt, 1980; soil profile at the site consisted of loose gravelly sand with
Lee et al., 1988). Hence, a single value for ns ¼ nb ¼ 0 .2 is DR ¼ 30% down to 3 m, followed by dense gravelly sand
selected independently of soil type. For sands, the assump- with DR ¼ 80% down to 13 to 14 m. The water table was at
tion ms ¼ mb ¼ 0 .3 is made, in accordance with Coyle & a depth of 3 m. The critical-state friction angle measured
Gibson (1970) and Randolph (2003). For clay, the param- from triaxial compression tests was 33 .38. CPT logs for this
eters ms and mb are expressed as linear decreasing functions site can be found in Paik et al. (2003).
of the undrained shear strength (su) according to Lee et al. The pile, which was 8 .24 m long with an outer diameter
(1988). of 356 mm and wall thickness of 12 .7 mm, was driven down
In practice, the limit base and shaft resistances can be to a depth of 6 .9 m. It was instrumented with two acceler-
evaluated based on measured soil properties (su, relative ometers and two strain transducers at the pile head. More-
density, DR) or CPT data using recently proposed static over, 18 strain gauges grouped in pairs were attached at nine
resistance equations (e.g. Randolph & Murphy, 1985; Jardine levels along the pile shaft in order to measure the strain in
et al., 1998, 2005; Lee & Salgado, 1999; Randolph, 2003; the pile during subsequent static load testing.
Lehane et al., 2005a; Salgado, 2008; Salgado et al., 2011). An ICE 42-S single-acting diesel hammer was used
(ICEUSA, 2011). The hammer had a ram weight of 18 .2 kN
and a maximum rated driving energy of 56 .8 kN m. Data
COMPARISON OF NUMERICAL PREDICTIONS WITH recorded during driving, using the Pile Driving Analyser1
PILE DRIVING DATA (PDA) (Pile Dynamics, 2001), showed that the energy
Models used in simulations transferred to the pile head ranged from 33% to 50% of the
To evaluate the performance of the proposed models, in maximum potential energy of the ram. An energy transfer
the section entitled ‘Results of pile driving simulations’, ratio equal to 0 .41 is used in the present calculations for this
model simulations are compared with the results of dynamic field case.
load tests performed in the context of two detailed case The simulations are done for the final penetration depth,

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 176 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS

.
τsf ⫽ qsL[1 ⫹ ms(wpile)ns]

Cs ⫽ jsqs,lim

τsf ⫽ qsL G
cs,rad ⫽
G Vs
ks ⫽ 1·375
πR
qs,lim
Ks ⫽
Qs

J-Q model Model S2

(Smith, 1960)

τvisc ⫽ qsLms(wpile ⫺ wsoil)ns

τsf,stat ⫽ qsL

G
G cs,rad ⫽
ks ⫽ 1·375 Vs
πR

Model S1
(a)

.
Rbf ⫽ QbL[1 ⫹ mb(wpile)nb]

Cb ⫽ JbRbL
Rbf ⫽ QbL 3·4R 2
Cb ⫽ 冪ρG
4GR 1⫺ν
Kb ⫽
1⫺ν
RbL
Kb ⫽
Qb

J-Q model Model B2

(Smith, 1960)

m0 ⫽ Kbα0(R /Vs )2
. .
Rvisc ⫽ QbLmb(wpile ⫹ wsoil)nb K1 ⫽ γ1Kb

C1 ⫽ Kb β1(R/Vs )
Rbf ⫽ QbL
m1 ⫽ Kbα1(R /Vs )2

3·4R 2
4GR Cb ⫽ 冪ρ G
Kb ⫽ 1⫺ν
1⫺ν

Model B1
(b)

Fig. 7. Existing reaction models used in the analyses of field pile tests: (a) for pile shaft;
(b) for pile base

at end of driving (EOD), since it is of most importance in Table 2 shows the soil parameters used in the simulations.
practice, with a measured pile set (permanent pile head The limit unit shaft resistances were obtained from the strain
displacement after the hammer blow) wset ¼ 10 mm. Data gauge data of the static load test, while Gmax is estimated
from the final hammer blows during pile installation are using the Hardin & Black (1968) equation. The static load
often used in the estimation of the static pile capacity. test was performed within 72 h from the pile installation.

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 177
Table 2. Model parameters for pile driving case study in Lagrange County

Depth: m Soil type qsL: kPa Gmax: kPa G1: kPa r: t/m3 ms ns

0 .5 Loose sand 26 .4 (12 .1) 24 257 1633 1 .67 0 .3 0 .2

1 .5 Loose sand 45 .1 (11 .4) 42 074 4914 1 .67 0 .3 0 .2
2 .5 Loose sand 17 .1 (27 .9) 54 317 8190 1 .67 0 .3 0 .2
3 .5 Dense sand 54 .4 (83 .7) 78 941 10 428 2 .14 0 .3 0 .2
4 .5 Dense sand 161 .7 (109 .2) 87 001 12 666 2 .14 0 .3 0 .2
5 .5 Dense sand 99 .8 (156 .7) 94 375 14 905 2 .14 0 .3 0 .2
6 .5 Dense sand 131 .6 (140 .7) 100 882 17 031 2 .14 0 .3 0 .2

Base depth: m Soil type Measured qbL: kPa Gmax: kPa G1: kPa r: t/m3 mb nb

6 .9 Dense sand 12 166 .1 103 822 18 038 2 .14 0 .3 0 .2

Several studies have shown that pile set-up effects on pile two strain transducers at the pile head to obtain the accelera-
base resistance in sand are either non-existent or minimal, tion– and force–time histories during driving. Thirty-four
irrespectively of time frame (Chow et al., 1998; Axelsson, strain gauges grouped in pairs were attached at 17 levels
2000; Lehane et al., 2005b). According to field test data by along the pile shaft in order to measure the strain in the pile
Bullock et al. (2005) and Jardine et al. (2006), the shaft during two static load tests, performed 50 days and 90 days
resistance gains in 72 h could be of the order of 10–20%. after driving. Data recorded during driving (PDA), showed
On the other hand, laboratory chamber test data by Lim & that the energy transmitted from the diesel hammer to the
Lehane (2014) suggest that pile set-up in sand is negligible pile head in the last stages of driving was 36% of the
for the first 7 days after installation. To balance any doubt maximum potential energy of the ram. Hence, the ef in
regarding the applicability of the qsL obtained from the static equation (44) for the calculation of the initial ram velocity
load test to EOD conditions, analyses were also performed was set equal to 0 .36, thus accounting for the total energy
using qsL estimates obtained from the CPT-based, semi- loses inside the hammer and the entire assembly between
empirical ICP-05 formula (Jardine et al., 1998, 2005), which ram and pile head.
captures very well the observed shaft resistance values just a The input soil parameters that were used in the EOD
few hours after pile installation at the Labenne and Dunkirk simulation of the Jasper County case appear in Table 3. The
test sites (Lehane et al., 2005b; Basu et al., 2011). These qsL values outside parentheses in Table 3 are those measured
qsL estimates are shown in parentheses in Table 2. in the static load test 50 days after installation. This time
The hammer impact in the present analyses is represented interval is substantial and, in this field case, set-up effects on
by the instantaneous application of an initial velocity to the the shaft resistance values from the static load test cannot be
topmost mass (the ram mass) of the modelled lumped mass ignored. Hence, the analyses presented in the next section
system (see Appendix). In reality, the ram of the diesel were done using the qsL values in parentheses in Table 3.
hammer never comes in contact with the anvil and the For soil layers that exhibited negligible to no plasticity, these
hammer action on the pile head is in essence a pressure are unit limit shaft resistance estimates established using the
associated with detonation of the diesel fuel mix, which ICP-05 formula, which produces estimates consistent with
creates a pulse of short duration. Moreover, pile helmet and time intervals of only few hours after EOD, as described in
cushion are not explicitly considered in the simulations of the previous section. In fine-grained, plastic soil layers
the present study. The energy losses in the assembly placed (labelled in Table 3 with an asterisk), the qsL operative at
between the ram and the pile head (along with energy loses the time of pile installation is estimated based on the recent
in the hammer) are indirectly taken into account by using an work by Basu et al. (2014), by multiplying the qsL observed
efficiency factor ef in equation (44) that is equal to the in the static load test by the ratio ÆST /ÆLT of short- to long-
energy transfer ratio of the one recorded by the instrumenta- term Æ values (where Æ is the coefficient by which the
tion of the pile head (equal to 0 .41). In this way, the energy undrained shear strength must be multiplied to produce the
applied to the pile head in the simulations is the same as the limit unit shaft resistance, qsL ¼ Æsu) provided by Basu et al.
energy measured in the field test. (2014). This approach assumes that the time interval of 50
days is enough for consolidation around the pile shaft to be
practically complete. This assumption is supported by the
Second case history: Jasper County, IN, USA fact that qsL values from the first (t ¼ 50 days) and second
The second case history relates to piling work done for a pile load tests (t ¼ 90 days) are practically the same (Kim et
bridge in Jasper County, Indiana (Abou-Jaoude, 2006; Kim al., 2009), as well as by the findings of Basu et al. (2014).
et al., 2009). An instrumented closed-ended steel pipe pile,
with length 19 .9 m, outer diameter of 356 mm and wall
thickness of 12 .7 mm, was driven to a depth of 17 .4 m. The Results of pile driving simulations
soil profile at the site consisted of soil layers of clayey silt Figure 8 through Fig. 11 compare the displacement–time
(plastic), silty clay and clayey silty sand down to a depth of histories at the pile head predicted using combinations of
17 m. These layers were underlain by a layer of very dense the proposed shaft and base models (hereafter referred to as
non-plastic silt (bearing layer), underlain by an 8-m-thick the ‘disc’ and ‘hyperbolic’ models, respectively) and the
soft to stiff clay layer. The results of an extensive laboratory simpler models shown in Fig. 7. The resulting values of pile
testing programme performed to characterise this soil profile set are summarised in Table 4 and Table 5.
and obtain the soil properties for each layer, along with CPT Figure 8 and Fig. 9 show results for the Lagrange County
logs at the location of the test pile, can be found in Kim et pile using qSL based on static load test data. The correspond-
al. (2009). The water table was at a depth equal to 1 m. ing pile set and relative error values are the numbers outside
The pile was instrumented with two accelerometers and parentheses in Table 4. The values inside the parentheses

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 178 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
Table 3. Model parameters for pile driving case study in Jasper County

Depth: m Soil type qsL: kPa Gmax: kPa G1: kPa r: t/m3 ms ns

0 .5 Organic clay 29 (25 .3) 522 4350 1 .37 1 .43 0 .2

1 .5 Organic clay 29 (24 .4) 786 4350 1 .37 1 .43 0 .2
2 .5 Organic clay 29 (24 .1) 875 4350 1 .37 1 .43 0 .2
3 .5 Silty sand 29 (12 .1) 65 734 5356 2 .24 0 .3 0 .2
4 .5 Clay sandy silt 21 (17 .9) 103 019 11 700 2 .2 1 .07 0 .2
5 .5 Clay sandy silt 21 (17 .7) 117 631 11 700 2 .2 1 .07 0 .2
6 .5 Silt clayey sand 21 (16 .2) 99 836 12 498 2 .24 0 .3 0 .2
7 .5 Sandy silty clay 7 (19 .8) 120 136 14 833 2 .14 0 .3 0 .2
8 .5 Silt clayey sand 88 (53 .5) 117 009 17 168 2 .24 0 .3 0 .2
9 .5 Silty clay 0 (0) 108 128 33 000 2 .05 0 .0 0 .2
10 .5 Clayey silt 47 (39 .9) 106 347 48 000 2 .1 0 .0 0 .2
11 .5 Clayey silt 47 (39 .7) 111 551 48 000 2 .1 0 .0 0 .2
12 .5 Silty clay 47 (38 .0) 190 550 15 450 2 .23 0 .88 0 .2
13 .5 Silty clay 47 (37 .8) 199 208 15 450 2 .23 0 .88 0 .2
14 .5 Clayey silt 109 (90 .6) 180 755 43 800 2 .2 0 .0 0 .2
15 .5 Clayey silt 109 (90 .2) 187 544 43 800 2 .2 0 .0 0 .2
16 .5 Clayey silt 250 (206 .3) 194 096 43 800 2 .2 0 .0 0 .2
17 .2 Very dense non-plastic silt 160 (188 .2) 148 548 38 357 2 .14 0 .3 0 .2

Base depth: m Soil type Measured qbL: kPa Gmax: kPa G1: kPa r: t/m3 mb nb

17 .4 Very dense non-plastic silt 6350 148 548 38 357 2 .14 0 .3 0 .2

 Soil treated as plastic.

30
S1 and B1
Lagrange County
G ⫽ Gmax
25
Pile head displacement: mm

20

15

S1 and hyperbolic
Observed pile set ⫽ 10 mm
10 Disc and hyperbolic

S1 and B2*
5 S1 and B2
S2 and B1
J-Q
S2 and hyperbolic
S2 and B2/S2 and B2*
0
0 0·02 0·04 0·06 0·08 0·10
Time: s

Fig. 8. Predicted and measured pile head response for Lagrange County pile using G Gmax

come from the analyses with qSL values obtained using the models S1, S2, B1 and B2 perform significantly better when
ICP-05 formula. Comparing values inside and outside par- setting G equal to G1 (based on Lee et al. (1988)) rather
entheses in Table 4, it is possible to conclude that any pile than Gmax, indicating thus the importance of soil stiffness
set-up that may have occurred in the 72 h between EOD and non-linearity in pile driving simulations. The use of secant
static load test has negligible effect on the pile driving values for soil moduli has long been a way to indirectly take
response. into account non-linearity (the challenge usually being what
In Fig. 8, it can be seen that the analysis using the disc fraction of Gmax to take for realistic results). Large over-
shaft model and the hyperbolic base model (disc–hyperbolic) estimation of the pile set occurred for the combination of
produces a pile set wset of 10 .1 mm, which approximates the models S1 and B1. This suggests that it is better to place
observed pile set very well. All other model combinations the slider of the base reaction model in series with the
produce relative errors in the range of 13–184%, with the spring only, as in model B2.
lower value corresponding to the analysis using shaft model Figure 10 and Fig. 11 present displacement–time histories
S1 (Simons & Randolph, 1985) with G equal to G1 and the for the Jasper County pile for a hammer blow near EOD
hyperbolic base model of this paper. Results show that with measured pile set 7 .8 mm. The analysis using the disc–

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 179
30
S1 and B1
Lagrange County
G ⫽ G1 for S1, S2, B1, B2
25

Pile head displacement: mm

20

15
S1 and B2/S1 and B2*

S1 and hyperbolic
10 Disc and hyperbolic
Observed pile set ⫽ 10 mm
S2 and B1

S2 and B2/S2 and B2*

5
S2 and hyperbolic
J-Q

0
0 0·02 0·04 0·06 0·08 0·10
Time: s

Fig. 9. Predicted and measured pile head response for Lagrange County pile using G G1 for
models S1, S2, B1, B2

16
S1 and B1
Jasper County
G ⫽ Gmax

12
Pile head displacement: mm

S1 and hyperbolic
Disc and hyperbolic
Measured response S1 and B2*
8
S1 and B2

4
J-Q

S2 and hyperbolic
S2 and B1
S2 and B2/S2 and B2*
0
0 0·04 0·08 0·12 0·16 0·20
Time: s

Fig. 10. Predicted and measured pile head response for Jasper County pile using G Gmax

hyperbolic model combination yields wset ¼ 8 .75 mm, which as long as the slider is active (B2 ) appears to improve only
is 12 .4% greater than the observed set. Using the S1–B2 very slightly the set predictions, and even then only if G is
model combination with G ¼ Gmax and G ¼ G1 produces set equal to Gmax. Analyses were also performed with
slightly better wset predictions (relative error 5 .3% and models B1 and B2, in which the radiation damping includes
14 .6%, respectively). The rest of the combinations produce the effect of far-field hysteresis (based on the principle of
relative errors in the range of 22–99%, with the larger correspondence), and analyses using the hyperbolic base
values corresponding again to analyses using the S1–B1 model but with zero hysteretic damping ratio in the far field.
combination. In these analyses, the predicted displacement Inclusion of the far-field hysteretic damping in the radiation
response appears also unrealistic, since it lacks the pile dashpot has negligible impact (generating differences less
rebound of roughly 6 mm observed in the measured response than 5%) on the results.
(PDA record). Taking into account the absolute relative errors from the
The option in B2 of disconnecting the radiation dashpot analysis of both field cases, the best performing combination in

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 180 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
16
S1 and B1
Jasper County
G ⫽ G1 for S1, S2, B1, B2

12
Pile head displacement: mm

S1 and hyperbolic
S1 and B2/S1 and B2*
Disc and hyperbolic
8
Measured response

4 J-Q
S2 and hyperbolic
S2 and B1
S2 and B2

0
0 0·04 0·08 0·12 0·16 0·20
Time: s

Fig. 11. Predicted and measured pile head response for Jasper County pile using G G1 for
models S1, S2, B1, B2

Table 4. Pile set results from simulation of Lagrange County field test

Model G ¼ Gmax G ¼ G1

Shaft Base Pile set: mm Rel. error: % Pile set: mm Rel. error: %

Disc Hyperbolic 10 .07 (9 .91) 0 .7 (0 .9)  

S1 Hyperbolic 12 .24 (12 .07) 22 .4 (20 .7) 11 .31 (11 .47) 13 .1 (14 .7)
S2 Hyperbolic 2 .53 (2 .44) 74 .7 (75 .6) 4 .26 (4 .17) 57 .4 (58 .3)
S1 B1 28 .37 (28 .06) 183 .7 (180 .6) 28 .36 (28 .0) 183 .6 (180 .0)
S1 B2 4 .90 (4 .79) 51 .0 (52 .1) 13 .81 (13 .58) 38 .1 (35 .8)
S1 B2 5 .24 (4 .94) 47 .6 (50 .6) 13 .81 (13 .58) 38 .1 (35 .8)
S2 B1 4 .07 (4 .06) 59 .3 (59 .4) 8 .34 (8 .34) 16 .6 (16 .6)
S2 B2 1 .40 (1 .33) 86 .0 (86 .7) 6 .23 (6 .11) 37 .7 (38 .9)
S2 B2 1 .40 (1 .33) 86 .0 (86 .7) 6 .23 (6 .11) 37 .7 (38 .9)
J-Q J-Q   3 .18 (3 .26) 59 .18 (58 .15)

Table 5. Pile set results from simulation of Jasper County field of the Jasper County field case (in which the pile is discretised
test into 19 segments) obtained using the disc–hyperbolic model
combination is a calculation time approximately eight times
Model G ¼ Gmax G ¼ G1 greater than the corresponding time using the S1–B2 model
combination. The difference in computational cost is due to the
Shaft Base Pile set: Rel. error: Pile set: Rel. error: much larger number of degrees of freedom in the soil disc
mm % mm % model. However, since both analyses are completed in a short
amount of time in absolute terms, this additional cost is not
Disc Hyperbolic 8 .75 12 .4  
S1 Hyperbolic 9 .55 22 .7 9 .51 22 .09 considered significant even in routine design. Analyses using
S2 Hyperbolic 0 .96 87 .6 2 .30 70 .49 the proposed model are performed very fast on both a laptop
S1 B1 15 .16 94 .6 15 .48 98 .73 (Intel1 CoreTM i5 M520 @ 2 .40 GHz) and workstation (In-
S1 B2 7 .37 5 .3 8 .93 14 .63 tel1 Xeon(R) CPU E5-1650 @ 3 .20 GHz). For the case of
S1 B2 8 .25 5 .9 8 .93 14 .68 Jasper County pile, simulation of a 0 .2-s-long response to a
S2 B1 1 .03 86 .8 2 .82 63 .84 hammer blow takes 27 .39 s of CPU process time on the laptop
S2 B2 0 .96 87 .7 2 .56 67 .15 and 6 .02 s on the workstation. For the case of the Lagrange
S2 B2 0 .96 87 .7 2 .56 67 .15 County pile, the same 0 .2-s-long response requires 8 .94 s on
J-Q J-Q   4 .88 37 .36 the laptop and 1 .89 s on the workstation.
Figure 12 and Fig. 13 show the velocity– and force–time
histories at the pile head for the Jasper County case for
terms of prediction of the pile set is the disc–hyperbolic model these three model combinations. It can be seen that all three
proposed in the present paper, followed by S1–hyperbolic and analyses, discussed above, simulate reasonably closely the
S1–B2, with overall relative errors 7%, 18% and 26%, respec- measured velocity and force responses. The numerical simu-
tively. The cost of the better prediction for the simulated 0 .2 s lations underpredict by 15% the peak velocity, while they

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 181
6·0 0
Measured (PDA)
Jasper County
Disc and hyperbolic
Field test data
Disc and hyperbolic S1 and B2 (G ⫽ Gmax)
S1 and hyperbolic S1 and B2 (G ⫽ G1)
4·0
S1 and B2 4
J-Q
Pile head velocity: m/s

Disk and hyperbolic (50% qsL)

2·0

Depth of pile base: m

8

12

⫺2·0
0 0·01 0·02 0·03 0·04 0·05
Time: s

Fig. 12. Predicted and measured velocity at pile head for Jasper 16
County pile

3000
Jasper County
20
Field test data 0 20 40 60
Disc and hyperbolic Pile set: mm
2000 S1 and hyperbolic
S1 and B2 Fig. 14. Predicted and measured pile set for Jasper County pile at
various pile embedment depths
Force at pile head: kN

1000 It can be seen that analyses with the proposed model

yields pile set predictions that are closer to those observed
(based on PDA measurements) than the best performing
combination of existing models, namely S1–B2, and the
traditional J-Q model. Moreover, the S1–B2 combination
0 yields better prediction when using G ¼ G1 than G ¼ Gmax,
However, it can also be seen that the predictions increasingly
deviate as the depth of pile penetration decreases, irrespec-
tive of the reaction model used. The deviations could be
⫺1 000 attributed to the following factors
0 0·01 0·02 0·03 0·04 0·05
Time: s (a) the inherent uncertainties in estimation of base and
especially shaft resistances based on empirical correla-
Fig. 13. Predicted and measured force at pile head for Jasper tions, particularly in layers of ‘borderline’ soils (e.g.
County pile sandy clayey silt)
(b) the fact that ICP-05, although performing very well for
pile penetrations close to the final installation depth, may
overpredict the peak pile head force by roughly 10%. Since
overestimate qsL at shallower penetration depths (Basu et
this is seen in all analyses irrespective of soil model
al., 2011)
combinations, the observed discrepancies can be attributed
(c) the possibility that values predicted by ICP-05 may not be
to: (a) differences between the actual values of the various
representative of (be greater than) the resistance operative
soil properties operative during driving and the estimates
at the time of driving (i.e., as argued by Jardine et al.
used in the analyses, and (b) the approximations made in
(2006))
modelling the diesel hammer impact.
(d ) the use of values of qbL in the simulations that are too
Figure 14 shows the pile set resulting from simulations at
high. If the shaft resistance at the time of driving is less
depths shallower than the EOD scenario considered so far
than that measured at the time of the load test, the
for the Jasper County pile. The shaft resistance in these
corresponding sets would increase for shallow and
cases was estimated using the same approach as in the near-
intermediate depths, reducing the difference between
EOD simulation. In the absence of static load testing data at
predicted and observed sets. Likewise, if excessively
embedments other than 17 .4 m (the final installation depth),
large values of qbL were to be used, this would also be
the static limit base resistance qbL at shallower depths was
underpredicting the set.
assumed to be equal to the CPT cone resistance qc averaged
over a depth range from 1 3 B above to 1 .5 3 B below the Figure 15 shows measured and predicted pile sets at
pile base (Salgado, 2008). This assumption can be consid- various intermediate depths for the Lagrange County pile.
ered acceptable if a sufficiently large pile set results in the The shaft resistance in these cases was estimated using the
simulated blow; otherwise, the operative qbL for the pile same approach as in the near-EOD simulation. The method
would be less than qc. used to estimate the limit base resistance at shallower depths

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 182 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
0 Measured (PDA) the strain gauges, and the velocity (plotted multiplied by the
pile impedance Z ¼ EA/c, where E is the Young’s modulus, A
Disc and hyperbolic
is the cross-sectional area of the pile and c is the wave speed
S1 and B2 (G ⫽ Gmax) in the pile) at the pile top produced by signal matching at the
end of driving for the Lagrange and Jasper Counties case
S1 and B2 (G ⫽ G1) histories using the models proposed in the present paper. Fig.
J-Q 16(c) and Fig. 17(c) show the comparison between the meas-
2 ured and predicted unit shaft resistances at the end of driving
for the two case histories. For the pile in Lagrange County,
the limit base resistance predicted by the signal-matching
algorithm is 14 .9 MPa, and the predicted static unit base
resistance at the mobilised peak base displacement of
Depth of pile base: m

15 .5 mm during the blow is 10 .2 MPa. This compares with a

unit base resistance value of 7 .3 MPa for the Lagrange
4 County pile as obtained from the static load test for the same
displacement. For the pile in the Jasper County case history,
the predicted limit base resistance is 10 .4 MPa, and the static
unit base resistance mobilised at the peak displacement of
15 .2 mm is 9 .0 MPa. For this value of displacement, the base
unit load from the load test was only 3 .4 MPa. In general,
there appears to be underestimation of shaft and overestima-
6
tion of base resistance, which could be partly a result of
residual loads being different at the time of driving and at the
time of the static load test.

SUMMARY AND CONCLUSIONS

Pile driving simulations have become an integral part of
design (through wave equation analysis) and quality control
8
0 20 40 60 (through signal matching analyses). This paper presented
Pile set: mm new shaft and base soil reaction models that may be used to
obtain the response of a pile (in terms of pile set and
Fig. 15. Predicted and measured pile set for Lagrange County
pile at various pile embedment depths displacement–time history) to hammer impact. The new
models account explicitly for soil stiffness non-linearity.
Predictions using the new models as well as other models in
was the same as for the Jasper County pile. The proposed existence were compared with measurements from two field
model yields pile set predictions that are close to the ob- cases. Based on the simulation results, the proposed hyper-
served set based on PDA measurements even at shallower bolic base model and disc shaft reaction model yield the
depths; the proposed model again performs better than closest predictions to the observed pile set. Nonetheless,
combinations of existing models, with the exception of the sufficiently accurate predictions can also be obtained by
S1–B2 combination with G ¼ G1, which again performs using the Simons & Randolph (1985) shaft model combined
well. The uniform sand layers (DR ¼ 30% from ground level with the hyperbolic base model, even when Gmax is used.
to 3 m and DR ¼ 80% down to 8 m) in the Lagrange site From this, it can be concluded that the effect of pre-failure
appear conducive to good accuracy of pile set prediction soil non-linearity is not as important in shaft reaction model-
using the proposed model, possibly owing to the more ling as in base reaction modelling.
accurate estimation of static resistances that it enables. The combination of shaft and base reaction models pro-
posed by Simons & Randolph (1985) performed adequately.
Use of secant shear modulus values substantially less than
Signal matching Gmax in the existing shaft and base reaction models, which
So far, results have been presented for use of the analysis assume linear pre-failure behaviour, lead to accurate simula-
to simulate the response of the pile–soil system to hammer tions of pile driving, as pointed out by Lee et al. (1988).
blows given estimated static resistance. It is also possible to In signal matching, measured response at the pile head is
use the analysis to do the reverse: estimate the static matched by varying the pile resistances. The proposed
resistance from the force and velocity measured during a models were also used to match force and velocity signals
hammer blow. This approach to static resistance calculation measured during impact; this matching produced estimates
is referred to as signal matching (Rausche et al., 1985, of static resistance at the end of driving that were reasonably
2000, 2010; Likins et al., 2012; Ng & Sritharan, 2013) and close to values measured in the load tests. Observed discre-
may be viewed as a more refined tool for doing what the pancies may be due to any one of the factors present in this
Case method (Rausche et al., 1985) was intended to do: complex simulation problem, including residual loads in-
estimate static resistance from force and velocity measured duced by driving that may evolve during the time between
during a hammer blow. Signals may, in theory, be matched driving and loading. The present paper shows that both wave
by more than one distribution of static resistance along the equation analysis and signal-matching applications could
pile (in other words, signal matching, as many inverse meth- benefit from more advanced base and shaft resistance mod-
ods, is subject to the potential limitation of non-uniqueness). els, as those presented in the paper.
The signal-matching algorithm must minimise differences
between predicted and observed signals while producing
static resistances that are physically acceptable. ACKNOWLEDGEMENT
Figures 16(a) and 16(b) and Figs 17(a) and 17(b) show the This research was supported in part by the FHWA-
predicted time histories of displacement, force measured with INDOT–Purdue Joint Transportation Research Program.

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 183
0·020 0·020

Lagrange County Jasper County

Field data Field data
0·016 Signal matching 0·016 Signal matching
Pile top displacement: m

Pile top displacement: m

0·012 0·012

0·008 0·008

0·004 0·004

0 0
0 0·02 0·04 0·06 0·08 0 0·02 0·04 0·06 0·08 0·10
Time: s Time: s
(a) (a)
3000 3000
Lagrange County Jasper County
Measured velocity Measured velocity
Calculated velocity
Force F and transformed velocity Zv: kN
Force F and transformed velocity Zv: kN

Calculated velocity
Measured force Measured force
2000 2000

1000 1000

0 0

⫺1 000 ⫺1 000
0 0·02 0·04 0·06 0·08 0 0·02 0·04 0·06 0·08 0·10
Time: s Time: s
(b) (b)

qsL: kPa qsL: kPa

0 40 80 120 160 200 0 50 100 150 200 250 300
0 0

Lagrange County Jasper County

Static load test
Static load test
4 Basu et al. (2014)
ICP-05
2 Signal matching
Signal matching

8
Depth: m
Depth: m

12

6
16

20
8 (c) (c)

Fig. 16. Signal matching of a pile blow at 6 .9 m embedment depth Fig. 17. Signal matching of a pile blow at 17 .4 m embedment
for Lagrange County: (a) displacement at pile top; (b) force and depth for Jasper County: (a) displacement at pile top; (b) force
velocity measured at pile top; (c) measured and predicted unit and velocity measured at pile top; (c) measured and predicted
shaft resistance unit shaft resistance

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 184 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
APPENDIX. PILE DRIVING ANALYSIS SOLUTION M1
ALGORITHM Ram v0
The 1D pile dynamic analysis is based on discretisation of the pile K2
into N  1 lumped masses (Mi), each subject to a soil reaction Rs,i M2
(Fig. 18). The ram (hammer) is also represented in the system as an K3
additional mass (M1). The lumped masses are connected to each M3
Rs,3
other through linear springs (Ki). The base reaction Rb acts on the
last mass in addition to the corresponding shaft reaction Rs,N. K4
Each lumped pile mass Mi is given by Ki⫺1

Shaft reaction models

M i ¼ rpile ˜Li A (37) Rs,i⫺1 Mi⫺1
Ki
where rpile is the density of the pile material, A is the pile cross-
sectional area and ˜Li is the length of pile segment i. Rs,i Mi
The stiffness Ki of the pile spring between pile segments i and Ki⫹1
i  1 is Rs,i⫹1 Mi⫹1
Epile A Ki⫹2
Ki ¼ . (38)
0 5(˜Li þ ˜Li1 )
KN⫺1
where Epile is the Young’s modulus of the pile material. The spring Rs,N⫺1 MN⫺1
connecting the pile head to the pile hammer combines the stiffness
KN
of the first pile element and the stiffness of the ram Base reaction
   model Rs,N MN
Epile A Eram Aram
0.5˜L2 Lram Rb
K 2 ¼ K com ¼ (39)
Epile A Eram Aram
. þ
0 5˜L2 Lram Fig. 18. Formulation of 1D dynamic analysis: discretisation of pile
into lumped masses with soil reactions applied to each lumped
where Eram, Aram and Lram are the Young’s modulus, the cross-
mass
sectional area and the length of the ram.
The system of differential equations (Phan et al., 2012, 2013)
describing the problem is .
where VE (¼ (Epile /rpile)0 5) is the wave velocity inside the pile. Trial
€ pile g þ [C][w_ pile ] þ [K]fwpile g þ fRg ¼ 0
[M]fw (40) simulations show that results are practically no longer influenced by
the time step size for ˜t values less than 0 .1˜tmin.
where [M], [C] and [K] are the global mass, damping and stiffness The analysis starts by imposing an initial velocity to the ram mass
matrices of the lumped mass system; fw € pile g, fw_ pile g and fwpile g are (M1) given by
the acceleration, velocity and displacement vectors of the pile rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Ei pffiffiffiffiffiffiffiffiffiffiffiffiffi
segments, and fRg is the vector of corresponding soil reactions Rs,i w_ (t¼0) ¼ 2 ef ¼ 2gHef (44)
1
and Rb calculated after each time increment using equations (22) and M ram
(23).
The global stiffness matrix of the system is tridiagonal
2 3
K com K com 0  0 0 0
6
6 K com K com þ K 3 K 3  0 0 0 7
7
6 0 K 3 K3 þ K4  0 0 0 7
6 7
6 . .. .. .. .. .. .. 7
[K] ¼ 6 .. . . . . . . 7 (41)
6 7
6 0 0 0    K N 2 þ K N 1 K N 1 0 7
6 7
4 0 0 0  K N 1 K N 1 þ K N K N 5
0 0 0  0 K N KN
The global mass matrix of any system comprising lumped masses
is a diagonal matrix
2 3 where g is the acceleration of gravity, H is the hammer drop height
M ram 0 0  0 0 0 and ef is the driving system efficiency expressing the energy losses
6
6 0 M2 0  0 0 7
70 of the driving system. Ei is the initial potential energy of the hammer
6 0 0 M3  0 0 70 (Mram 3 g 3 H; also called the ‘theoretical’ energy).
6 7
6 . .. .. .. .. .. 7..
[M] ¼ 6 .. . . . . . 7 . (42)
6 7
6 0 0 0  M N 2 0 0 7
6 7
4 0 0 0  0 M N 1 0 5 NOTATION
0 0 0  0 0 MN A pile cross-sectional area
In the present 1D analysis formulation, all elements in the B pile diameter
damping matrix [C] are equal to zero except the last diagonal bf parameter that controls rate of degradation of
element (corresponding to the pile base), which is equal to Cb shear modulus
(equation (31)). This is because any shaft reaction damping bfb parameter that controls rate of degradation of base
component is already included in the force Rs,i. spring stiffness
The system of equations of motion (equation (40)) is solved using C damping coefficient
Newmark’s algorithm. Cb radiation damping coefficient for pile base
To obtain an accurate solution, the time increment size ˜t must be CLysm radiation dashpot coefficient of Lysmer model
small enough that information does not propagate beyond the c wave speed in pile
distance between adjacent nodes during a single time step. This cemb depth factor for radiation damping
means the time increment needs to be smaller than the threshold cFF radiation dashpot constant
value chys factor representing influence of hysteretic damping
inside far field soil
min (˜Li ) D pile embedment depth
˜tmin ¼ (43)
VE Df depth factor

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 185
Df,dyn depth factor for dynamic loading —pile perimeter of pile cross-section
DR relative density r soil density
E Young’s modulus  v9 vertical effective stress
Ei initial potential energy of hammer  shear stress
ef hammer efficiency _ rate of shear stress
G shear modulus f shear strength of soil in simple shear condition
Gmax maximum (small-strain) shear modulus f,stat static soil strength
G1 secant shear modulus (t)
i shear stress at position of node i at time t
g acceleration of gravity (t) ,  (t)
shear stress values within the segments on the left
L,i R,i
H hammer drop height (inner) and right (outer) sides of node i
Jb Smith damping coefficient for pile base rev shear stress at the last stress reversal
Js Smith damping coefficient for pile shaft S soil reaction at pile shaft wall
K spring stiffness at pile base sf total (static + viscous) resistance of the rheological
Kb,max maximum (small-strain) spring stiffness at pile model on the pile shaft
base 1,R shear stress in the first segment (segment between
KLysm spring stiffness used in Lysmer model nodes 1 and 2) of the soil disc
kFF linear spring stiffness at outer boundary of soil ø fundamental period of the pile–soil system
disc
L pile length
LI loading index parameter for pile shaft
LIb loading index parameter for pile base REFERENCES
[M], [C], [K] global mass, damping and stiffness matrices of the Abou-Jaoude, G. G. (2006). Assessment of static pile design meth-
lumped mass system ods and non-linear analysis of pile driving. PhD thesis, Purdue
Mpile mass of pile University, West Lafayette, IN, USA.
m0 added mass in base model Arshad, M., Tehrani, F., Prezzi, M. & Salgado, R. (2014). Experi-
m1 auxiliary mass in base model mental study of cone penetration in silica sand using digital
mb, nb soil viscosity parameters for pile base image correlation. Géotechnique 64, No. 7, 551–569, http://
ms, ns soil viscosity parameters for pile shaft dx.doi.org/10.1680/geot.13.P.179.
Ns number of points (soil nodes) along radial axis Axelsson, G. (2000). Long-term set-up of driven piles in sand. PhD
nF effect of roughness of pile shaft on pile–soil thesis, Royal Institute of Technology, Stockholm, Sweden.
interface strength Basu, P., Loukidis, D., Prezzi, M. & Salgado, R. (2011). Analysis
PI plasticity index of soil in percentage units of shaft resistance of jacked piles in sands. Int. J. Numer. Analyt.
QbL limit pile base capacity under quasi-static loading Methods Geomech. 35, No. 15, 1605–1635.
qbL static unit base resistance Basu, P., Prezzi, M., Salgado, R. & Chakraborty, T. (2014). Shaft
qc cone resistance resistance and setup factors for piles jacked in clay. J. Geotech.
qsL static unit shaft resistance Geoenviron. Engng, ASCE 140, No. 3, 04013026.
Rb total base reaction Bullock, P. J., Schmertmann, J. H., McVay, M. C. & Townsend, F.
RbL limit base resistance C. (2005). Side shear setup. II: results from Florida test piles. J.
Rb,rev spring reaction R(S)
b at last displacement reversal Geotech. Geoenviron. Engng, ASCE 131, No. 3, 301–310.
R(D)
b dashpot reaction for pile base Burland, J. B. (1970). Discussion on Session A. Proceedings of the
R(S)
b spring reaction for pile base conference on in situ investigation in soils and rocks, pp. 61–62.
R(t)
s total shaft reaction at time t London, UK: British Geotechnical Society.
r radial coordinate of soil node Charue, N. (2004). Loading rate effects on pile load–displacement
rf radius of outer boundary of near field soil disc behaviour derived from back-analysis of two load testing
rm radius of pile influence procedures. PhD thesis, Université catholique de Louvain,
rpile radius of pile Belgium.
su undrained shear strength of soil Chow, F. C., Jardine, R. J., Brucy, F. & Nauroy, J. F. (1998). Effects
sgn(x) sign of variable x of time on capacity of pipe piles in dense marine sand.
t time J. Geotech. Geoenviron. Engng, ASCE 124, No. 3, 254–264.
VE wave velocity inside pile Coyle, H. M. & Gibson, G. C. (1970). Empirical damping constants
Vs shear wave velocity of soil for sands and clays. J. Soil Mech. Found. Div. 96, No. 3, 949–
vref reference relative velocity 965.
w vertical displacement Dayal, U. & Allen, J. H. (1975). The effect of penetration rate on
w_ b vertical velocity of pile base the strength of remolded clay and sand samples. Can. Geotech.
w_ b,el elastic component of pile base velocity J. 12, No. 3, 336–348.
w_ pile vertical velocity of pile Deeks, A. J. & Randolph, M. F. (1995). A simple model for
wset pile set for a hammer blow inelastic footing response to transient loading. Int. J. Numer.
w_ 1 vertical velocity of first soil node of the soil disc Analyt. Methods Geomech. 19, No. 5, 307–329.
f€
wpile g, fw_ pile g, acceleration, velocity and displacement vectors of Dobry, R. & Gazetas, G. (1986). Dynamic response of arbitrarily
fwpile g pile segments shaped foundations. J. Geotech. Engng, ASCE 112, No. 2, 109–
Z pile impedance 135.
z depth Dos Santos, J. A. & Correia, A. G. (2001). Reference threshold
Æ coefficient by which the undrained shear strength shear strain of soil. Its application to obtain an unique strain-
must be multiplied to produce the limit unit shaft dependent shear modulus curve for soil. Proceedings of the 15th
resistance international conference on soil mechanics and geotechnical
ÆLT long-term value of Æ engineering, Istanbul, Turkey. Rotterdam, the Netherlands: AA
ÆST short-term value of Æ Balkema.
ª_ shear strain rate EI-Naggar, M. & Novak, M. (1994). Nonlinear model for dynamic
ª_ ref reference shear strain rate axial pile response. J. Geotech. Engng 120, No. 2, 308–329.
˜Li length of pile segment i Einav, I. & Randolph, M. F. (2006). Effect of strain rate on
˜r radial distance between soil nodes mobilised strength and thickness of curved shear bands. Géo-
˜t time increment technique 56, No. 7, 501–504, http://dx.doi.org/10.1680/
ºS rate factor geot.2006.56.7.501.
v Poisson ratio of soil Fahey, M. & Carter, J. P. (1986). Some effects of rate of loading
 hysteretic damping ratio and drainage on pressuremeter tests in clay. Proceedings of the

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 186 SOIL STIFFNESS NON-LINEARITY IN 1D PILE DRIVING SIMULATIONS
specialty geomechanics symposium: Interpretation of field test- Lee, S. L., Chow, Y. K., Karunaratne, G. P. & Wong, K. Y. (1988).
ing for design parameters, Adelaide, South Australia, pp. 50–55. Rational wave equation model for pile-driving analysis. J. Geo-
Barton, ACT, Australia: Institution of Engineers, Australia. tech. Engng, ASCE 114, No. 3, 306–325.
Fahey, M. & Carter, J. P. (1993). A finite element study of the Lehane, B. M., Schneider, J. A. & Xu, X. (2005a). The UWA-05
pressuremeter test in sand using a nonlinear elastic plastic method for prediction of axial capacity of driven piles in sand.
model. Can. Geotech. J. 30, No. 2, 348–362. In Proceedings of the international symposium on frontiers in
Foray, P., Balachowski, L. & Rault, G. (1998). Scale effect in shaft offshore geotechnics (IS-FOG 2005)(eds C. Mark and G. Susan),
friction due to the localisation of deformations. In Proceedings pp. 683–689. Perth, Australia: Taylor and Francis.
of the international conference centrifuge 98 (eds Y. Kimura, O. Lehane, B. M., Schneider, J. A. & Xu, X. (2005b). A review of
Kusakabe and J. Takemura), pp. 211–216. Amsterdam, the design methods for offshore driven piles in siliceous sand, UWA
Netherlands: Balkema. Report Geo 05358. Perth, Australia: University of Western
Fox, E. N. (1948). The mean elastic settlement of a uniformly Australia.
loaded area at a depth below the ground surfaces. Proceedings Likins, G. E., Liang, L. & Hyatt, T. (2012). Development of
of the 2nd international conference on soil mechanics and automatic signal matching procedure-iCAP. Proceedings of test-
foundation engineering, Rotterdam, the Netherlands, pp. 129– ing and design methods for deep foundations, pp. 97–104.
132. Kanazawa, Japan: IS-Kanazawa.
Gazetas, G. (1991). Foundation vibrations. In Foundation engineer- Lim, J. K. & Lehane, B. M. (2014). Set-up of pile shaft friction in
ing handbook (ed. H.-Y. Fang), pp. 553–593. Boston, MA, USA: laboratory chamber tests. Int. J. Phys. Modelling Geotech. 14,
Springer US. No. 2, 21–30.
Gazetas, G., Dobry, R. & Tassoulas, J. L. (1985). Vertical response Litkouthi, S. & Poskitt, T. J. (1980). Damping constants for pile
of arbitrarily shaped embedded foundations. J. Geotech. Engng, driveability calculations. Géotechnique 30, No. 1, 77–86, http://
ASCE 111, No. 6, 750–771. dx.doi.org/10.1680/geot.1980.30.1.77.
Graham, J., Crooks, J. H. A. & Bell, A. L. (1983). Time effects on Loukidis, D. & Salgado, R. (2008). Analysis of the shaft resistance
the stress–strain behaviour of natural soft clays. Géotechnique of non-displacement piles in sand. Géotechnique 58, No. 4,
33, No. 3, 327–340, http://dx.doi.org/10.1680/geot.1983.33.3.327. 283–296, http://dx.doi.org/10.1680/geot.2008.58.4.283.
Hardin, B. O. & Black, W. L. (1968). Vibration modulus of Loukidis, D., Salgado, R. & Abou-Jaoude, G. (2008). Assessment of
normally consolidated clay. J. Soil Mech. Found. Div. 94, No. 2, axially-loaded pile dynamic design methods and review of
353–370. INDOT axially-loaded pile design procedure, JTRP Technical
Heerema, E. P. (1979). Relationships between wall friction, Reports, pp. 172. West Lafayette, IN, USA: Purdue University.
displacement, velocity and horizontal stress in clay and in sand Lysmer, J. & Richart, F. E. J. (1966). Dynamic response of footings
for pile drivability analysis. Ground Engng 12, No. 1, 55–61. to vertical loading. J. Soil Mech. Found. Div. 92, No. 1, 65–91.
Holeyman, A. E. (1985). Dynamic non-linear skin friction of piles. Michaelides, O., Bouckovalas, G. & Gazetas, G. (1998). Non-linear
Proceedings of the international symposium on penetrability and soil properties and impedances for axially vibrating pile ele-
drivability of piles, San Francisco, California, pp. 173–176. ments. Soils Found. 38, No. 3, 129–142.
Tokyo, Japan: Japanese Society of Soil Mechanics and Founda- Ng, K. W. & Sritharan, S. (2013). Improving dynamic soil param-
tion Engineering. eters and advancing the pile signal matching technique. Comput.
Holeyman, A. & Legrand, C. (1994). Soil modeling for pile Geotech. 54, 166–174.
vibratory driving. Proceedings of the US FHWA international Nguyen, T. T., Berggren, B. & Hansbo, S. (1988). A new soil
conference on design and construction of deep foundations, pp. model for pile driving and drivability analysis. Proceedings of
1165–1178. Orlando, Florida: FHWA. 3rd international conference on application of stress-wave
Holeyman, A. E., Legrand, C. & Van Rompaey, D. (1996). A theory to piles, Vancouver, Canada. pp. 353–367. Vancouver,
method to predict the drivability of vibratory driven piles. Canada: Bi Tech.
Proceedings of 5th international conference on the application Novak, M., Aboul-Ella, F. & Nogami, T. (1978). Dynamic soil
of stress-wave theory to piles, Orlando, Florida, pp. 1101–1112. reactions for plane strain case. J. Engng Mech. Div. 104, No. 4,
Orlando, FL, USA: University of Florida. 953–959.
ICEUSA (2011). ICE model 42S fuel-injected diesel pile hammers. Omidvar, M., Iskander, M. & Bless, S. (2012). Stress–strain behav-
See http://www.iceusa.com/files/uploads/75EE1F17-9106-4590- ior of sand at high strain rates. Int. J. Impact Engng 49, 192–
8021-AD86B608D08E.pdf (accessed 18/07/2013). 213.
Itasca Consulting Group (1998). Fast Lagrangian analysis of con- Paik, K., Salgado, R., Lee, J. & Kim, B. (2003). Behavior of open-
tinua (FLAC). Minnesota, USA: Itasca Consulting Group. and closed-ended piles driven into sands. J. Geotech. Geoenvir-
Jardine, R. J., Overy, R. F. & Chow, F. C. (1998). Axial capacity of on. Engng 129, No. 4, 296–306.
offshore piles in dense north sea sands. J. Geotech. Geoenviron. Pells, P. J. N. & Turner, R. M. (1978). Janbu, Bjerrum and
Engng, ASCE 124, No. 2, 171–178. Kjaernsli’s chart reinterpreted: discussion. Can. Geotech. J. 15,
Jardine, R. J., Chow, F., Overy, R. & Standing, J. (2005). ICP No. 436, 437–437.
design methods for driven piles in sand and clays. London, UK: Phan, T. L., Matsumoto, T. & Kobayashi, S. (2012). A matrix
Thomas Telford. method of wave propagation analysis in an open-ended pipe pile
Jardine, R. J., Standing, J. R. & Chow, F. C. (2006). Some based on equation of motion. Proceedings of the 9th interna-
observations of the effects of time on the capacity of piles tional conference on testing and design methods for deep
driven in sand. Géotechnique 56, No. 4, 227–244, http:// foundations – IS-Kanazawa 2012 (eds K. Kaneda and S.
dx.doi.org/10.1680/geot.2006.56.4.227. Kobayashi), pp. 105–112. Tokyo, Japan: Japanese Geotechnical
Kim, D., Bica, A. V. D., Salgado, R., Prezzi, M. & Lee, W. (2009). Society.
Load testing of a closed-ended pipe pile driven in multilayered Phan, T. L., Matsumoto, T. & Nguyen, H. H. (2013). Comparison of
soil. J. Geotech. Geoenviron. Engng 135, No. 4, 463–473. static and dynamic pile load tests at Thi Vai International Port
Kramer, S. L. (1996). Geotechnical earthquake engineering. Upper in Viet Nam. Int. J. Geoengng Case Histories 3, No. 1, 36–66.
Saddle River, NJ, USA: Prentice Hall. Pile Dynamics (2001). Pile driving analyzer (PDA). See http://
Kulhawy, F. H. & Mayne, P. W. (1990). Manual on estimating soil www.pile.com/pdi/products/pda/ (accessed 18/07/2013).
properties for foundation design, 1493-6(EL-6800), 308. Palo Potts, D. M. & Martins, J. P. (1982). The shaft resistance of axially
Alto, CA, USA: Electric Power Research Institutite. loaded piles in clay. Géotechnique 32, No. 4, 369–386, http://
Lee, J. H. & Salgado, R. (1999). Determination of pile base resistance dx.doi.org/10.1680/geot.1982.32.4.369.
in sands. J. Geotech. Geoenviron. Engng 125, No. 8, 673–683. Randolph, M. F. (2003). Science and empiricism in pile foundation
Lee, J. & Salgado, R. (2000). Analysis of calibration chamber plate design. Géotechnique 53, No. 10, 847–875, http://dx.doi.org/
load tests Junhwan Lee and Rodrigo Salgado. Can. Geotech. J. 10.1680/geot.2003.53.10.847.
37, No. 1, 14–25. Randolph, M. F. & Murphy, B. S. (1985). Shaft capacity of driven
Lee, J., Salgado, R. & Carraro, J. A. H. (2004). Stiffness degrada- piles in clay. Proceedings of the 17th annual offshore technology
tion and shear strength of silty sands. Can. Geotech. J. 41, No. conference, Houston, TX, USA, pp. 371–378.
5, 831–843. Randolph, M. F. & Simons, H. A. (1985). An improved soil model

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.
 SALGADO, LOUKIDIS, ABOU-JAOUDE AND ZHANG 187
for one dimensional pile driving analysis. Proceedings of the Salgado, R., Prezzi, M. & Tehrani, F. S. (2011). Soil property-based
3rd international conference of numerical methods in offshore methods for design of nondisplacement piles. Pan-Am CGS
piling, Nantes, France, pp. 3–17. Cambridge, UK: Cambridge geotechnical conference, Toronto, Canada. Toronto, Canada:
University Engineering Department. Canadian Geotechnical Society.
Randolph, M. F. & Wroth, C. P. (1978). Analysis of deformation of Sheahan, T. C., Ladd, C. C. & Germaine, J. T. (1996). Rate-
vertically loaded piles. J. Geotech. Engng Div. 104, No. 12, dependent undrained shear behavior of saturated clay. J. Geo-
1465–1488. tech. Engng 122, No. 2, 99–108.
Rausche, F., Goble, G. G. & Likins, G. E. (1985). Dynamic Simons, H. A. (1985). A theoretical study of pile driving. PhD
determination of pile capacity. J. Geotech. Engng 111, No. 3, thesis, University of Cambridge, Cambridge, UK.
367–383. Simons, H. A. & Randolph, M. F. (1985). A new approach to one
Rausche, F., Robinson, B. & Liang, L. (2000). Automatic signal dimensional pile driving analysis. Proceedings of the 5th inter-
matching with CAPWAP. Application of stress-wave theory to national conference on numerical methods in geomechanics (eds
piles, pp. 53–58. Rotterdam, the Netherlands: Balkema. T. Kawamoto and Y. Ichikawa), pp. 1457–1464. Nagoya, Japan:
Rausche, F., Likins, G. & Liang, L. (2010). Static and dynamic Balkema.
models for capwap signal matching. In Art of foundation Smith, E. A. L. (1960). Pile-driving analysis by the wave equation.
engineering practice, pp. 534–553. West Palm Beach, FL, USA: J. Soil Mech. Found. Div. 86, No. 4, 35–64.
American Society of Civil Engineers. Vanden-Berghe, J.-F. & Holeyman, A. (2002). Application of a
Richart, F. E., Hall, J. R. & Woods, R. D. (1970). Vibrations of hypoplastic constitutive law into a vibratory pile driving model.
soils and foundations. Upper Saddle River, NJ, USA: Prentice- Proceedings of the international conference on vibratory pile
Hall. driving and deep soil compaction (eds A. Holeyman, N. Charue
Salgado, R. (2008). The engineering of foundations. New York, NY, and J. F. Vanden Berghe), pp. 61–68. Louvain-la-Neuve,
USA: McGraw Hill. Belgium: Taylor and Francis.
Salgado, R. (2012). The mechanics of cone penetration: Contribu- Vucetic, M. & Dobry, R. (1991). Effect of soil plasticity on cyclic
tions from experimental and theoretical studies. Geotechnical response. J. Geotech. Engng 117, No. 1, 89–107.
and geophysical site characterization 4, ISC 4, pp. 131–153. Wolf, J. P. (1988). Soil–structure-interaction analysis in time do-
Porto de Galinhas, Brazil: CRC Press. main. Englewood Cliffs, NJ, USA: Prentice-Hall.

Downloaded by [ Zhejiang University] on [23/09/21]. Copyright © ICE Publishing, all rights reserved.