Analytical Modeling of Hammer Impact For Pile Driving

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/229485100
Analytical modeling of hammer impact for pile driving
Article in International Journal for Numerical and Analytical Methods in Geomechanics · May 1993
DOI: 10.1002/nag.1610170502
CITATIONS READS
89 3,474
2 authors:
Andrew J Deeks Mark Randolph

Murdoch University University of Western Australia
118 PUBLICATIONS 4,079 CITATIONS 398 PUBLICATIONS 19,099 CITATIONS
SEE PROFILE SEE PROFILE
All content following this page was uploaded by Andrew J Deeks on 05 December 2017.
The user has requested enhancement of the downloaded file.

ANALYTICAL MODELLING OF HAMMER IMPACT FOR PILE DRIVING
A.J. Deeks and M.F. Randolph
Department of Civil and Environmental Engineering,
University of Western Australia,
Nedlands, Western Australia, 6009.
Phone: (+61-9-) 380 3186
Facsimile: (+61-9-) 380 1044

Analytical modelling of hammer impact for pile driving A.J Deeks & M.F. Randolph
Summary
Estimation of the drivability of piles requires modelling of the hammer impact in order
to provide the input force wave at the pile head. Historically, this has been achieved
through a numerical model of the hammer components (ram, cushion and anvil), which
is then linked with that of the pile in order to effect the drivability analysis. This paper
presents an analytical model of hammer impact, based on lumped ram and anvil masses
separated by a cushion with internal damping, and connected to the pile which is
modelled as a dashpot. Force-time responses derived from the analytical model are
compared with actual field data, and also with results from commercially available
numerical models of hammer impact. The analytical model is then used to explore the
characteristics of hammer impact, with particular attention to combinations of
parameters that lead to hammer bounce, and hence significant loss of energy transferred
to the pile.
Introduction
Analysis of the dynamic response of a pile during driving requires, as input, the force-
time signal generated by the hammer impact. In dynamic testing of piles, this signal
may be measured directly by means of strain sensors attached near the head of the pile;
the measured force response is then used in a numerical analysis of the pile-soil system,
adjusting the soil parameters in an iterative fashion in order to achieve consistency
between the results of the analysis and other field measurements (for example, velocity
near the pile head, force measured at other positions down the pile, temporary
compression and permanent set). Such data are not available at the design stage,
however, where it is necessary to assess an appropriate size and type of hammer in order
to drive the piles to the required penetration, within a reasonable blow count and
without generating excessive compressive or tensile stresses in the pile. This aspect of
design is known as a ‘drivability study’.
1
Historically, the force response at the head of a pile for a drivability study is computed
through a numerical model of the hammer. The main components of a hammer consist
of the ram, a cushion, and an anvil (or helmet) which sits on the pile. Smith1 proposed
modelling these as lumped masses, separated by a spring representing the cushion. Any
cushion between the anvil and the pile (for example, for concrete piles) may be
incorporated into the first element of the pile. More recently, progressively more
complex models of hammer impact have been developed, particularly for modelling
diesel hammers (for example, in the WEAP family of programs, Goble and Rausche2,
Rausche et al3). In most cases, however, relatively simple hammer models provide a
sufficiently accurate representation of the force-time response. This has been
demonstrated by Middendorp and van Weele4 who compared results from different
numerical models of an MRBS 8000 hammer with field data, and showed that precise
modelling of the hammer geometry is not necessary in order to achieve an adequate
match to the field data.
For simple hammer models, it is possible to derive analytical solutions for the force-
time response at the pile head. Such solutions provide a useful alternative to numerical
modelling, since they may be readily incorporated into spreadsheet or other programs to
analyse the driving response of piles. A particular advantage of the analytical solutions
is the ease with which parametric studies may be undertaken, to explore different
combinations of hammer and pile properties. Randolph5 has presented an analytical

solution for the simple case of a (lumped) ram mass impacting a cushion seated directly
on the pile, and discusses extension of the solution to include an anvil. The latter
solution made use of Laplace transforms, which were inverted numerically, and did not
allow for separation of the hammer from the pile.
This paper presents new analytical solutions for the hammer model shown in Figure
1(d). The model uses lumped masses for the ram and anvil, and a combined spring and
dashpot for the cushion (modelling energy dissipation in the cushion). The pile is
modelled as a dashpot, of impedance Z = EA/c where EA is the cross-sectional rigidity
2
of the pile and c is the (one-dimensional) wave speed in the pile. The derived force-
time responses at the pile head are therefore valid up to the time when reflections arrive
back at the pile head. The solutions allow for separation of the ram mass, by restricting
the cushion force to be non-negative.
The solutions are presented in terms of non-dimensional quantities where, for the
idealised case of a lumped ram mass impacting directly onto a pile, the non-dimensional
peak force and the time constant for the exponential decay of the force would both be
unity. Appropriate non-dimensional ratios are introduced to allow for finite cushion
stiffness, anvil mass and cushion damping. For completeness, analytical solutions for
the simpler cases (for example, zero anvil mass) are presented in addition to that for the
full model.
Results from the lumped hammer model are compared with field data and it is shown
that, despite the obvious idealisation of the model, typical force-time responses
measured in the field may be simulated with remarkable accuracy. The analytical
solutions provide an efficient and valuable alternative to numerical modelling of the
hammer.
In the final part of the paper, results of a detailed parametric study are presented,
showing how the force-time response is affected by different combinations of hammer
and pile properties. Particular attention is paid to conditions which lead to separation of
the hammer from the pile, and which result in significant loss in energy transfer.
Curves of hammer efficiency are presented in terms of the non-dimensional parameters.
Ram/pile model
The simplest model of a pile hammer is that of a falling mass impacting directly on the
top of a pile. The response of the pile to the impacting mass can be modelled by
replacing the pile with a dashpot of equal impedance to the pile. This impedance is
given by
3
Ep Ap
Z = cp (1)
where Ep is the Young's modulus of the pile, Ap is the cross-section area of the pile, and
cp is the axial wave velocity in the pile.
The force exerted by the hammer on the pile can be obtained by analysing the system
shown in Figure 1(a), where mr is the mass of the ram, ur is the displacement of the ram
after it strikes the top of the pile, and vo is the velocity of the ram when it strikes the
pile. Dot notation will be used to represent differentiation with respect to time. The
equation of motion for the ram is
mr ür + Z u̇ r = 0 (2).
Defining a dimensionless time as
Z
t* = m t (3)
r
and a dimensionless velocity as
du* u̇
u̇ * = * = v (4),
dt o
the dimensionless displacement and acceleration are
Z u
u* = m v (5)
r o
and
d2u* mr ü
ü* = *2 = Z v (6).
dt o
Substituting these equalities into equation (2), a dimensionless equation of motion can
be formed.
ür* + u̇ * = 0 (7)
4
At a dimensionless time of zero (the instant the hammer strikes the pile), the
dimensionless velocity of the ram is unity, and the displacement is zero. Applying these
initial conditions, the solution to equation (7) can be found.
-t*
u̇ r* = e (8)
The force exerted on the pile is given by multiplying the pile impedance by the ram
velocity.
fp = Z u̇ r
Since the appropriate non-dimensionalisation of force is
f
f* = Z v (9),
o
the dimensionless force exerted on the pile is simply
u̇r -t*
fp * = v = u̇ r* = e (10).
o
The dimensionless impulse function exerted on the pile by this pile hammer model is
shown in Figure 2, denoted by an infinite cushion stiffness. At time zero the
dimensionless force rises instantaneously to unity, and then decays exponentially with a
non-dimensional time constant of unity.
Ram/cushion/pile model
The accuracy of the model described above can be increased by using a linear spring to
model the effect of a cushion between the ram and the pile. The system then has two
degrees of freedom, since the motion of the pile head can be different from that of the
ram. The displacement of the pile head will be denoted by ua, and the stiffness of the
cushion spring by kc. The model is illustrated in Figure 1(b).
The two equations of motion for this model are
5
mr ür + kc (ur - ua) = 0 (11)
and
Z u̇ a + kc (ua - ur) = 0 (12).
Using the dimensionless variables given in equations (3) to (6), these equations can be
re-written as
ür* + kc* (ur* - ua*) = 0 (13)
u̇ a* + kc* (ua* - ur*) = 0 (14)
where
kc mr
kc* = (15).
Z2
Applying the Laplace transformation to equations (13) and (14) and solving
simultaneously, the Laplace transform of the velocity at the pile head can be found.
kc*
*
L u̇ a = 2 (16)
s + kc* s + kc*
Factorising the quadratic denominator
kc*
L u̇ a* = (17)
kc* kc*
(s + 2 + µ) (s + 2 - µ)
where
kc*2
µ = 4 - kc
* (18).
The inverse transform of equation (17) depends on the value of kc*. If kc* is greater
than 4, equation (17) becomes
6
kc*  1 1 
L u̇ a* =  -  (19),
2µ kc * kc*
s + - µ s + + µ
 2 2 
and performing the inverse transform yields
kc*
- t* kc*
f* = u̇ a* = e 2 sinh(µ t*) (20).
µ
If kc* is equal to 4,
kc*
L u̇ a* = (21)
 kc*2
s + 
 2
and
kc*
- t*
fp* = u̇ a* = t* e 2 (22).
Finally, if kc* is less than 4, µ is imaginary. Let
kc*2
µ' = µ i = kc* - 4 (23).
Equation 17 becomes
kc*  1 1 
*
2µ' i  kc* 
L u̇ a = - (24)
kc*
s + 
 2 - µ' i s + 2 + µ' i
and
kc* k *
* c
fp = u̇ a = e 2 t
-
* * sin(µ' t*) (25).
µ'
The force in the cushion spring is equal to the force exerted on the pile. If kc* is less
than 4, equation 25 indicates that the force will become negative after time ts*, where
π
ts* = (26).
µ'
7
Since the cushion is not attached to the pile or the ram, negative stresses cannot develop
in the cushion spring, so gaps will form and the ram will become separated from the
pile. The force exerted on the pile after time ts*will be zero. Equation (25) is only valid
for times less than or equal to ts*.
The dimensionless response of the ram/cushion/pile model is dependent solely on the

dimensionless stiffness of the cushion, kc*. The effect of this stiffness on the response
is shown in Figure 2. Finite stiffness of the cushion causes the force on the pile to have
a finite rise time. A soft cushion causes a large rise time. As the cushion stiffness
becomes large, the rise time decreases, and the solution approaches that for the ram/pile
model.
Ram/cushion/anvil model
The hammer model can be further improved by introducing a finite anvil mass (Figure
1(c)). This does not change the number of degrees of freedom, but increases the
complexity of the solution.
The two equations of motion governing the system are as follows, where ma is the anvil
mass.
mr ür + kc (ur - ua) = 0 (27)
ma üa + Z u̇ a + kc (ua - ur) = 0 (28)
Again using the dimensionless variables given in equations (3) to (6), these equations
can be re-written as
ür* + kc* (ur* - ua*) = 0 (29)
1 k*
üa* + u̇ * + c (u * - u *) = 0 (30)
ma* a ma* a r
where kc* is given by equation (15), and the dimensionless anvil mass ma* is given by
8
ma
ma* = m (31).
r
simultaneously, the Laplace transform of the velocity of the anvil and the pile head can
be found.
kc*
ma*
L u̇ a* = (32)
kc*  1  1
* + kc* * + 1 s + * s2 + s3
ma ma  ma
The anvil velocity and the force on the pile head can be found by factorising the cubic
denominator and performing the inverse Laplace transform, as detailed in Appendix A.

An expression of the following form is obtained. Expressions for the parameters Fp, c1,
c2, ω, and φ are presented in the appendix.
-c1t* -c2t* cos (ωt* - φ)

fp* = u̇ a* = Fp e 1 - e  (33).
 cos φ 
The dimensionless force in the spring is given by
fs* = ma* üa* + u̇ a* (34).
Differentiating equation (33), substituting into equation (34), and simplifying, the spring
force can be expressed as
 c2t* cos (ωt* - θ)

fs* = Fs e -  (35),
 cos θ 
where Fs and θ are also specified in Appendix A.
The ram will separate from the top of the cushion when the spring force is zero, which
occurs at time ts*, where
c2ts* cos (ωts* - θ)

e = (36).
cos θ
9
An analytical solution for ts* cannot be found. For a certain range of anvil mass and
cushion stiffness, ram separation does not occur. (This range is discussed in a
subsequent section). For those cases where separation does occur, equation (36) must
be solved numerically to identify ts.
For most combinations of dimensionless anvil mass and cushion stiffness, the cubic
denominator in equation (32) has one real root and two imaginary roots, and the
solution presented above applies. For a small range of combinations, the cubic
denominator has three real roots, and different solutions must be used. In this case,
equation (32) can be written in the following way.
a0
L u̇ a* = (s + b ) (s + b ) (s + b ) (37)
1 2 3
When b1, b2, and b3 are distinct, the solution is given by
fp* = u̇ a* = Fp e-b1t - Ap e-b2t - Bp e-b3t 

* * *
(38).
The force in the spring is
fs* = Fs e-b1t - As e-b2t - Bs e-b3t 

* * *
(39).
These solutions are developed in full in Appendix A, together with expressions for all
the parameters.
However, if two of the real roots are the same, the solution is different again. When b3
is equal to b2, the solution becomes
fp* = u̇ a* = Fp e-c1t 1 - e-c2t (1 + b2 t* )

* *
(40)
fs* = Fs e-c1t 1 - e-c2t (1 + b2 As t *)

* *
(41),
as detailed in Appendix A.
10
For cases where b1 is equal to b3 or b2, the above solution can be adjusted by
interchanging subscripts. However, in a practical computer implementation of these
analytical solutions, equations (40) and (41) are probably not needed, since a very small
change in ma* or kc* will allow one of the solutions discussed previously to be used.
Equations (40) and (41) are presented here for completeness.
When ram separation occurs, the equation of motion of the anvil is
ma üa + Z u̇ a = 0 (42)
or
ma* üa* + u̇ a* = 0 (43).
The dimensionless velocity of the anvil at the time the ram separates will be denoted by
vs*. The velocity of the anvil after separation is then given by
t*-ts*
- *
fp* = u̇ a* = vs* e ma (44).
The effect of the anvil mass on the force exerted on the pile head is shown in Figure 3
for a dimensionless cushion stiffness of 4. As the anvil mass increases, the time taken
for the force to peak increases. The peak force also increases for a range of ma* values,
but decreases after ma* exceeds a certain value. The anvil mass which leads to
maximum peak force on the pile head is discussed in a later section.
Damped cushion model
The model above assumes that the cushion behaves as an elastic spring. However, in
practice the cushion deforms non-linearly, and absorbs energy. The simplest analytical
method of accounting for this non-linearity and energy loss is to model the cushion as a
linear spring in parallel with a dashpot (Figure 1(d)). Only one extra parameter is added
to the model, and this method has been found to adequately model the observed
11
behaviour of real pile hammers. An example of the ability of this model to represent
real hammer performance is presented later in the paper.
Adding the dashpot again increases the complexity of the solution, but does not
introduce any extra degrees of freedom. Using cc to represent the damping in the
cushion, the equations of motion of the system are
mr ür + cc (u̇ r - u̇ a) + kc (ur - ua) = 0 (45)
ma üa + Z u̇ a + cc (u̇ a - u̇ r) + kc (ua - ur) = 0 (46)
Again using the dimensionless variables specified in equations (3) to (6), these
equations can be re-written as
ür* + cc* (u̇ r* - u̇ a*) + kc* (ur* - ua*) = 0 (47)
1 cc* kc*
*
üa + u̇ * + * *
(u̇ - u̇r ) + (u * - ur*) = 0 (48)
ma* a ma* a ma* a
where the dimensionless cushion damping cc* is given by
cc
cc* = Z (49).
simultaneously, the Laplace transform of the anvil velocity can be found.
cc*  kc*
s + 
ma*  cc*
L u̇ a* = (50)
kc*   1  cc 
*
  1  1 
+ kc* *+1+ * s + cc* *+1+ * s2 + s3
ma*  ma  ma   ma  ma 
Performing the inverse Laplace transform, the pile force and anvil velocity can be
expressed in the same form as equation (33).
-c1t* -c2t* cos (ωt* - φ)

fp* = ua* = Fp e 1 - e  (51)
 cos φ 
12
The intermediate steps and expressions for the parameters are presented in Appendix B.
The spring force can be found by applying equation (34), and is of the form
-(c1+c2)t*  c2t* cos (ωt*-θ)

fs* = Fs e e - As  (52),
 cos θ 
where the coefficients are also detailed in Appendix B.
The ram will separate from the top of the cushion when the spring force is zero, which
occurs at time ts*, where
c2ts* cos (ωts* - θ)

e = As (53).
cos θ
Again, this equation cannot be solved analytically. For those cases in which hammer
separation occurs, the time for separation must be established numerically. The
subsequent force response is then given by equation (44).
The above solution applies when the denominator has one real and two imaginary roots.
If it has three distinct real roots, Appendix B shows that the solution can be expressed
as
fp* = u̇ a* = Fp e-b1t - Ap e-b2t + Bp e-b3t 

* * *
(54).
fs* = Fs e-b1t - As e-b2t + Bs e-b3t 

* * *
(55).
When two of the roots are identical, the solution below applies (Appendix B).
* *
(
fp* = u̇ a* = Fp e-c1t 1 - e-c2t (1 + γ t *) ) (56)
fs* = Fs e-c1t 1 - e-c2t (As + γ Bs t *)

* *
(57)
If either of these solutions applies, hammer separation will not occur.
13
Figure 4 shows the effect of the dimensionless cushion damping on the solution for the
case when kc* is 4 and ma* is 0.2. As expected, increased damping in the cushion
decreases any oscillation of the force. When the damping is small, the solution
approaches that obtained for the model with a linear cushion.
Figure 5 shows which of the above solutions must be used for different combinations of
kc*, ma* and cc*. The range of possible combinations of kc* and ma* is broken into two
regions, depending on the roots of the polynomial denominator in the Laplace

transform. Each value of cc* changes the extent of the two regions. When there is no
cushion damping (cc* = 0), equation (33) must be used in the region with two imaginary
roots. In the region with three distinct real roots, equation (38) applies, and along the
boundary between the two regions, equation (40) can be used. For non-zero values of
cc*, equation (51) applies in the region with two imaginary roots, equation (54) applies
when there are three distinct real roots, and equation (56) should be used along the
dividing boundary.
Matching field data
The ability of the analytical model to represent real pile driving hammers will be
demonstrated by comparing two sets of field data to the analytical solutions.
In the first case, a BSP 357 hammer was used to drive a 762 mm diameter by 18.5 mm
wall thickness pile with an impedance of 1750 kNs/m. The cushion was steel with a
stiffness of 1.6x106 kN/m. The recorded variation of the pile head force with time is
shown in Figure 6.
The specified ram and anvil masses for this hammer are 6860 kg and 850 kg
respectively, yielding a dimensionless anvil mass of 0.124 and a dimensionless cushion
stiffness of 3.653. Using these values in equation (33), which applies for zero cushion
damping, yields the solution also plotted in Figure 6. The agreement between the
analytical solution and the field data is excellent, suggesting that there is little damping
14
in the cushion. Equation (51), with a dimensionless cushion damping of 0.05 or less,
was found to give similar results.
The second data set was obtained from a BSP HA40 hammer, driving a 762 m diameter
by 44 mm wall thickness pile with an impedance of 4000 kNs/m. In this case the
cushion was Bongassi with a stiffness of 7.5x106 kN/m. The force measured at the pile
head is shown in Figure 7.
The specified cushion stiffness, together with a ram mass of 39 300 kg and an anvil
mass of 6000 kg, give a dimensionless cushion stiffness of 18.8, and a dimensionless
anvil mass of 0.153. Equation (51) was found to best match the field data when a
dimensionless cushion damping of 0.6 was used with a kc* of 25. The resulting solution
is also plotted in Figure 7. The agreement between the analytical model and the field
measurements is very good.
An appropriate cushion damping must be selected in order to reproduce the observed
response. However, a damping constant determined from one blow has been found to
adequately represent subsequent blows of the same hammer.
The two examples above show that the analytical model described in this paper can be
used to represent a hammer with reasonable accuracy, or to construct a hammer model
from field results. Any computer spreadsheet with graphing facilities can be used to
match the model to field results, or to rapidly evaluate the effect of altering hammer
parameters.
Range of dimensionless parameters
The analytical solutions presented in this paper show that the behaviour of the hammer
model is governed by three dimensionless parameters: the cushion stiffness, the anvil
mass, and the cushion damping.
15
The dimensionless cushion stiffness is dependent on the actual cushion stiffness, the
ram mass, and the pile impedance. The range of dimensionless cushion stiffness for a
number of commercial pile driving hammers has been calculated, based on piles with
impedances ranging from 1400 to 18 000 kNs/m. The dimensionless anvil mass for
each was also calculated. The results are presented in Figure 8.
The range of practical interest for kc* is 1 to 1000, and for ma* is 0.1 to 0.7. Since the
use of cushion damping is a technique to account for dissipation of energy in the

cushion, determination of cc* is only possible from available field data. However,
examination of the effect of cc* on the pile force has revealed that the range from 0 to 1
is of practical interest.
The analytical solutions allow various parametric studies to be performed quickly and
easily. Three such studies will be discussed here.
Ram separation
For most values of dimensionless cushion stiffness and anvil mass, the ram will
separate from the cushion as it rebounds. The time at which this separation occurs is
given by equation (36) when there is no cushion damping, and equation (53) when
cushion damping is present. However, for a certain range of kc* and ma*, ram
separation will not occur. This region is dependent on the cushion damping. Figure 8
shows that the size of the region of no separation increases as the cushion damping
increases.
When separation does occur, the ram usually has sufficient upward velocity after
separation from the cushion to prevent it from restriking the pile within the time domain
of interest. This means that the kinetic energy contained in the separating hammer is
effectively lost, since when the hammer does restrike the pile, the magnitude of the
resulting stress wave will usually be insufficient to cause any further inelastic pile
penetration.
16
It is interesting to observe that the majority of hammer data fall into the region where
separation occurs, leading to loss of transmitted energy.
Transmitted energy
The amount of energy transmitted from the ram to the pile can be found by integrating
the pile head force multiplied by the pile head velocity from zero to infinity. Any
energy contained in the ram after separation from the cushion is considered lost. A
transmitted energy factor can be formed by dividing the transmitted energy by the
kinetic energy in the ram at the instant it strikes the pile.
The transmitted energy factor is dependent on the dimensionless anvil mass, cushion
stiffness, and cushion damping. Figure 9 shows the variation of transmitted energy
factor with cushion stiffness for various anvil masses, when there is no cushion
damping. Since there are no energy absorbing components in the hammer system, the
only energy which is not transmitted to the pile is the kinetic energy in the ram after
separation. Consequently, when the ram does not separate, the transmitted energy factor
is unity. At low values of cushion stiffness, the velocity of the ram during rebound is
small, so the energy loss is small. However, at higher stiffnesses, the ram velocity
during rebound is larger, and the energy loss is correspondingly larger. The sudden
drops in the curves show transitions from no ram separation to ram separation. The
energy loss at high cushion stiffness is greatly effected by the anvil mass. A small anvil
mass combined with a large cushion stiffness can be seen to result in an inefficient
hammer.
Comparing Figures 8 and 9, it may be seen that the majority of hammers considered
have anvil mass ratios in excess of 0.4. While separation will generally occur for these
hammers, the consequential loss in transmitted energy is relatively small. Thus, even
for very high dimensionless cushion stiffnesses, the minimum transmitted energy for an
ma* of 0.4 is about 83%.
17
The effect of varying cushion damping was examined by analysing a system with a
dimensionless anvil mass of 0.3. The plots of transmitted energy factor against cushion
stiffness for different damping ratios are shown in Figure 10. Since the damping
represents energy loss in the cushion, increasing the damping generally decreases the
efficiency of the hammer system, but there are some small regions in which the
efficiency is actually increased by the cushion damping. These regions occur where the
addition of damping has prevented ram separation.
Holding the cushion damping at 0.4 and varying the anvil mass, the curves presented in
Figure 11 result. For large cushion stiffnesses there is little change from the case with
no damping. However, the efficiency of larger anvil masses with smaller cushion
stiffnesses is reduced quite markedly. At small cushion stiffnesses, smaller anvils are
seen to be more efficient than larger ones (Figure 9).
Maximum transmitted force
Figure 3 shows that, for a given spring stiffness, as the anvil mass is increased from
zero, the maximum force exerted on the pile increases, up to a certain point. After this
point, increasing the anvil mass further decreases the maximum force on the pile. For
each combination of cushion stiffness and damping, there is a particular anvil mass
which causes the maximum instantaneous force to be exerted on the pile. This
instantaneous force can be greater than the pile impedance multiplied by the initial ram
velocity.
The value of anvil mass which leads to maximum pile force for various values of
cushion damping is shown in Figure 12. For a dimensionless cushion stiffness of less
than 100, the mass is within the range of real pile hammers. The effect of cushion
damping is small for cushion stiffnesses greater than 10, increasing the mass slightly.
At small cushion stiffnesses, cushion damping is much more significant, decreasing the
required anvil mass considerably.
18
Figure 13 shows the maximum instantaneous pile force which will be obtained. If kc* is
100, an increase of more than 40% in the peak pile force can result. The effect of
cushion damping is to slightly decrease the maximum force.
Limitations
In this paper, a simple mechanical model has been used to represent the pile hammer
system. This model can adequately represent the performance of many real drop
hammers, but does have limitations.
The lumped mass model of the hammer assembly may be criticised on the grounds that
the rams of real hammers are several metres long, which will influence the force
response at the pile head. To illustrate this, Figure 14 shows a comparison of the force
response calculated using (a) the lumped mass model described in this paper; and (b) a
full numerical analysis of the hammer impact using the impedance approach
(Middendorp and van Weele4). The hammer, modelled approximately on an MRBC

7000, has a steel ram of 5 m long, 1.5 m diameter, with a Young's modulus of 2.1x108
kPa and a density of 7.85 t/m3. The pile was 1.83 m in diameter with a wall thickness
of 65 mm (giving a pile impedance of 1.46x104 kNs/m). As may be seen from Figure

14, the distributed mass model leads to a series of waves propagating within the
hammer system, which give rise to steps in the pile head force response. The lumped
mass model gives a much smoother response, but in other respects the two curves are
identical. As such, it may be concluded that the lumped mass model gives an adequate
representation of real hammers with ram lengths of several metres.
The hammer assembly is separated from the pile model, and so the analytical solutions
are only valid until the arrival of significant reflected waves, usually from the soil at the
base of the pile. At subsequent times, the reflected waves interact with the hammer
assembly. However, in most cases this interaction does not significantly affect the pile
drivability.
19
The model presented here only represents drop hammers, not diesel hammers, and does
not currently permit a cushion to be used between the anvil and the pile.
Conclusions
This paper has presented an analytical model of the pile hammer system. The results
obtained from the model have been shown to closely represent data recorded during
field tests.
In contrast to numerical models of the pile hammer system, the analytical model allows
parametric studies of hammer performance to be carried out quickly and easily. The
parametric studies reported in the last part of the paper show the potential of the model
in this area.
The model is expected to find particular application in computer programs which
analyse pile drivability through characteristic methods, where the hammer must be
modelled separately.
References
1. E.A.L. Smith, 'Pile driving analysis by the wave equation', J. Soil Mech. and Found.,
ASCE, 86, 35-61, (1960).
2. G.G. Goble and F. Rausche, WEAP86 program documentation in 4 Vols, Federal
Highway Administration, Office of Implementation, Washington D.C., (1986).
3. F. Rausche, G.G. Goble and G.E. Likins, 'Recent WEAP developments', Proc. 3rd
Int. Conf. on Application of Stress-Wave Theory to Piles, Ottawa, 164-173, (1988).
4. P. Middendorp and A.F. van Weele, 'Application of characteristic stress wave method
in offshore practice', Proc. 3rd Int. Conf. on Num. Methods in Offshore Piling, Nantes,
Supplement, 6-18, (1986).
20
5. M.F. Randolph, 'Analysis of the dynamics of pile driving', in Developments in Soil
Mechanics IV: Advanced Geotechnical Analyses, eds P.K. Banerjee and R. Butterfield,
Elsevier Applied Science, also University of Western Australia Research Report No.
G1001, (1991).
Appendix A -- Solution of ram/cushion/anvil model
The Laplace transform for the anvil velocity in the ram/cushion/anvil model is
kc*
ma*
L u̇ a* = (A1).
kc*  1  1
* + kc* * + 1 s + * s2 + s3
ma ma  ma
Naming the coefficients of the cubic denominator
kc*  1  1
a0 = * , a1 = kc* * + 1 , a2 = * (A2),
ma ma  ma
equation (32) can be written as
a0
L u̇ a* = (A3).
a0 + a1 s + a2 s2 + s3
Letting
a1 a2 a0 a23
α = 6 - 2 - 27 (A4)
and
a13 a12 a22 a0 a1 a2 a02 a0 a23

β = 27 - 108 - 6 + 4 + 27 (A5),
if β 2 is greater than zero, the cubic denominator of equation (A1) has one real root and
two imaginary roots. The inverse transform can be found readily if the following
substitutions are made.
a2 1 1
b1 = 3 - (α + β) /3 - (α - β) /3 (A6)
21
a2 1 1 1
(
b2 = 3 + 2 (α + β) /3 + (α - β) /3 ) (A7)
3
ω = 2 ((α + β)1/3 - (α - β)1/3) (A8)
This allows the cubic denominator to be factorised, and equation (A1) becomes
a0
L u̇ a* = (A9),
(s + b1) ((s + b2)2 + ω2)
which can be expanded to
a0  1  s + b2 b2 - b1 
ω2+(b2-b1)2 s +b1 (s+b2)2+ω2 (s+b2)2+ω2
 
L u̇ a* = - + (A10).
Performing the inverse Laplace transform, the anvil velocity can be found.
a0  -b1t* -b2t*  b2-b1 

u̇ a* = e -e cos ωt* + sin ωt* (A11)
ω +(b2-b1) 
2 2  ω 
This solution can be written in a simpler, more convenient form by making
c2 a0
c1 = b1 , c2 = b2 - b1 , φ = atan , and Fp = 2 2 (A12).
ω ω +c2
The anvil velocity and the force on the pile head are then
-c1t* -c2t* cos (ωt* - φ)

fp* = u̇ a* = Fp e 1 - e  (A13).
 cos φ 
The spring force can be found by using the equilibrium equation
fs* = ma* üa* + u̇ a* (A14).
Differentiating equation (A13), substituting into equation (A14), and simplifying, the
spring force can be expressed as
 c2t* cos (ωt* - θ)

fs* = Fs e -  (A15),
 cos θ 
where
22
c2 (a2 - c1 -- c2) - ω2 a0 (a2-c1)

θ = atan   and Fs = (A16).
 ω (a2 - c1)  ω2+c22
For most combinations of ma* and kc*, β 2 is greater than zero, and the solution
presented above applies. For a small range of combinations, β*2 is less than or equal to
zero, and a different solution must be used. When this is the case, the cubic
denominator has three real roots, and equation (A1) can be written in the following way.
a0
L u̇ a* = (s + b ) (s + b ) (s + b ) (A17)
1 2 3
a2 θ
b1 = 3 + 2 Q cos 3 (A18)
a2 θ+2π
b2 = 3 + 2 Q cos 3 (A19)
a2 θ+4π
b3 = 3 + 2 Q cos 3 (A20)
a22 - 3 a1
Q = 9 (A21)
2 a23 - 9 a1 a2 + 27 a0
R = 54 (A22)
 R 
θ = acos   (A23)
 Q3
If b1, b2, and b3 are all distinct, equation (A18) becomes
 1 1 1 
L u̇ a* = a0 (s+b )(b -b )(b -b )+(s+b )(b -b )(b -b )+(s+b )(b -b )(b -b ) (A24)
 1 2 1 3 1 2 1 2 3 2 3 1 3 2 3 
and the solution is given by

* * *
(A25)
with
a0 b3-b2 b2-b1
Fp = (b -b )(b -b ) , Ap = b -b , and Bp = b -b (A26).
2 1 3 1 3 1 2 3
23
 (a -b )e-b1t* (a2-b2)e-b2t
*
(a2-b3)e-b3t 
*
 2 1
fs* = a0 (b -b )(b -b ) + (b -b )(b -b ) + (b -b )(b -b ) (A27)
 2 1 3 1 1 2 3 2 1 3 2 3 
or
fs* = Fs e 1 - As e 2 - Bs e 3 
-b t* -b t* -b t* (A28)
with
a0 (a2-b1) a2-b2 b3-b2 a2-b3 b2-b1

Fs = (b -b )(b -b ) , As = a -b b -b , and Bs = a -b b -b (A29).
2 1 3 1 2 1 3 1 2 1 2 3
However, if two of the roots are the same, the solution is different again. When β is
zero and b3 is equal to b2, and the solution becomes
fp* = u̇ a* = Fp e-c1t 1 - e-c2t (1 + b2 t* )

* *
(A30)
fs* = Fs e 1 1 - e 2 (1 + b2 As t *)

-c t* -c t* (A31)
a0 a0
Fp = , Fs = , As = 1 - ma* b2 (A32).
c22 ma*c22
Appendix B -- Solution of damped cushion model
The Laplace transform of the anvil velocity for the damped cushion model is
cc*  kc*
s + 
ma*  cc*
L u̇ a* = (B1)
kc*   1  cc 
*
  1  1 
+ kc* *+1+ * s + cc* *+1+ * s2 + s3
ma*  ma  ma   ma  ma 
Naming the coefficients of the cubic denominator as before
kc* *
*  cc  1 
1 1
a0 = * , a 1 = kc * + 1  + * , a2 = cc* * + 1 + * (B2)
ma ma  ma ma  ma
and introducing
24
cc* kc*
c0 = , b =
4 c * (B3),
ma* c
equation (69) can be written as
c0 (s + b4)
L u̇ a* = (B4).
a0 + a1 s + a2 s2 + s3
Introducing the substitutions specified in equations (A2) to (A8), the denominator may
be factorised as follows, providing β 2 is greater than zero.
c0 (s + b4)
L u̇ a* = (B5)
(s + b1) ((s + b2)2 + ω2)
This can be expanded to
c0(b4-b1)  1 s + b2 b22+ω2+b1b4-b4b2-b2b1
ω2+(b2-b1)2 s+b1 (s+b2)2+ω2 (b4-b1)((s+b2)2+ω2) 
L u̇ a* = - - (B6).
Performing the inverse Laplace transform, the anvil velocity can be found.
c0(b4-b1)  -b1t* -b2t* b b +b b -b b -b 2-ω2 

ua * = 2 e - e cos ωt *+ 2 1 4 2 1 4 2 sin ωt* (B7)
ω +(b2-b1) 
2
 ω (b4-b1) 
This solution can be written in a simpler form by making
c2(c4-c2)-ω2 c0 c4
c1 = b1 , c2 = b2 - b1 , c4 = b4 - b1 , φ = atan , and Fp = 2 2 (B8).
ω c4 ω +c2
The equation for the pile head force and velocity is then in the same form as equation
(A15).
-c1t* -c2t* cos (ωt* - φ)

fp* = ua* = Fp e 1 - e  (B9).
 cos φ 
The spring force can be found by applying equation (A14).
-(c1+c2)t*  c2t* cos (ωt*-θ)

fs* = Fs e e - As  (B10)
 cos θ 
where
25
 1 
c0 c4  * - c1
1
* -b2+γω  mγ * - ω - γb2 
Fs =
 a
m  ma
, As = 1
 a
, θ = atan 1
 (B11)
ω + c2
2 2
 * - b2 + γω
ma* 1
-b
ma 
and
(c4-c2)c2 - ω2
γ = tan φ = (B12).
ω c4
When β 2 is less than or equal to zero, the denominator again has three real roots.
Equations (A18) to (A24) define parameters b1, b2, and b3.
If b1, b2, and b3 are distinct, equation (B1) becomes
 (b4-b1) (b4-b2) (b4-b3) 

L u̇ a* = c0 (s+b )(b -b )(b -b ) + (s+b )(b -b )(b -b ) + (s+b )(b -b )(b -b ) (B13)
 1 2 1 3 1 2 1 2 3 2 3 1 3 2 3 
and the solution is given by

* * *
(B14)
c0 (b4-b1) b4-b2 b3-b1 b4-b3 b2-b1

Fp = (b -b )(b -b ) , Ap = b -b b -b , and Bp = b -b b -b (B15).
2 1 3 1 4 1 3 2 4 1 2 3
fs* = Fs e-b1t + As e-b2t - Bs e-b3t 

* * *
(B16)
with
 1 
c0  * - b1(b4-b1)
 a
m 
Fs = (b -b )(b -b ) ,
2 1 3 1
1 1
- b2 -b
ma * b4-b2 b3-b1 ma* 3 b4-b3 b2-b1
Ap = 1 b4-b1 b3-b2 , and Bp = 1 b4-b1 b2-b3 (B17).
* - b 1 * - b1
ma ma
The solution for the case when β is zero and b3 is equal to b2 follows.
26
* *
(
fp* = u̇ a* = Fp e-c1t 1 - e-c2t (1 + γ t *) ) (B18)
fs* = Fs e-c1t 1 - e-c2 (As + γ Bs t *)

* t* (B19)
1 1
- b2 + γ -b
c0c4 c0c4 ma * ma* 2 (c4-c2)c2
Fp = 2 , Fs = 2 , As = 1 , Bs = 1 , γ = c4 (B20)
c2 c2
* - b1 * - b1
ma ma
27
Figures
Fig 1. Analytical pile hammer models.

Fig 2. Variation of pile head force/time response with dimensionless cushion
stiffness.
Fig 3. Variation of pile head force/time response with dimensionless anvil mass for a
cushion stiffness of 4.
Fig 4. Variation of pile head force/time response with dimensionless cushion

damping for kc* = 4 and ma* = 0.2.
Fig 5. Regions in which the cubic denominator has real roots.

Fig 6. Comparison of field data and analytical solution for BSP 357 hammer.
Fig 7. Comparison of field data and analytical solution for BSP HA40 hammer.
Fig 8. Range of dimensionless parameters for a selection of hammers, and regions in
which ram separation occurs.
Fig 9. Variation of transmitted energy factor with anvil mass for no cushion damping.
Fig 10. Variation of transmitted energy factor with cushion damping for ma* = 0.3.
Fig 11. Variation of transmitted energy factor with anvil mass for ca* = 0.4.
Fig 12. Anvil mass required to maximise pile force.
Fig 13. Maximum pile force obtained by varying anvil mass.

<This figure has been mislaid. Please refer to the published paper.>
Fig 14. The effect of finite ram length on the pile head force.
View publication stats

Analytical Modeling of Hammer Impact For Pile Driving

Uploaded by

Copyright:

Available Formats

Analytical Modeling of Hammer Impact For Pile Driving

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analytical Modeling of Hammer Impact For Pile Driving

Uploaded by

Copyright:

Available Formats

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

Analytical modeling of hammer impact for pile driving

Andrew J Deeks Mark Randolph

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

A.J. Deeks and M.F. Randolph

Department of Civil and Environmental Engineering,

University of Western Australia,

Nedlands, Western Australia, 6009.

Phone: (+61-9-) 380 3186

Facsimile: (+61-9-) 380 1044

characteristics of hammer impact, with particular attention to combinations of

adjusting the soil parameters in an iterative fashion in order to achieve consistency

design is known as a ‘drivability study’.

sufficiently accurate representation of the force-time response. This has been

modelling of the hammer geometry is not necessary in order to achieve an adequate

match to the field data.

combinations of hammer and pile properties. Randolph5 has presented an analytical

allow for separation of the hammer from the pile.

modelled as a dashpot, of impedance Z = EA/c where EA is the cross-sectional rigidity

the cushion force to be non-negative.

solutions provide an efficient and valuable alternative to numerical modelling of the

showing how the force-time response is affected by different combinations of hammer

Curves of hammer efficiency are presented in terms of the non-dimensional parameters.

cp is the axial wave velocity in the pile.

equation of motion for the ram is

Defining a dimensionless time as

and a dimensionless velocity as

the dimensionless displacement and acceleration are

initial conditions, the solution to equation (7) can be found.

Since the appropriate non-dimensionalisation of force is

the dimensionless force exerted on the pile is simply

shown in Figure 2, denoted by an infinite cushion stiffness. At time zero the

non-dimensional time constant of unity.

cushion spring by kc. The model is illustrated in Figure 1(b).

The two equations of motion for this model are

mr ür + kc (ur - ua) = 0 (11)

Z u̇ a + kc (ua - ur) = 0 (12).

ür* + kc* (ur* - ua*) = 0 (13)

u̇ a* + kc* (ua* - ur*) = 0 (14)

Factorising the quadratic denominator

than 4, equation (17) becomes

and performing the inverse transform yields

Finally, if kc* is less than 4, µ is imaginary. Let

The dimensionless response of the ram/cushion/pile model is dependent solely on the

complexity of the solution.

mr ür + kc (ur - ua) = 0 (27)

ma üa + Z u̇ a + kc (ua - ur) = 0 (28)

ür* + kc* (ur* - ua*) = 0 (29)

denominator and performing the inverse Laplace transform, as detailed in Appendix A.

c2, ω, and φ are presented in the appendix.

-c1t* -c2t* cos (ωt* - φ)

The dimensionless force in the spring is given by

fs* = ma* üa* + u̇ a* (34).

force can be expressed as

 c2t* cos (ωt* - θ)

where Fs and θ are also specified in Appendix A.

c2ts* cos (ωts* - θ)

equation (32) can be written in the following way.

-c1t -c2t cos (ωt* - φ)

ür* + cc* (u̇ r* - u̇ a) + kc (ur* - ua*) = 0 (47)

-c1t -c2t cos (ωt* - φ)