Dynamic behaviour of single piles under strong harmonic excitation

AND

MILOSNOVAK
Fnculry of Engineering Scietfce, The Utliversity o f Wesrertl Otltario, Lot~rlotl,Otlf., C C I I I N C ~ ( ~5B9
N6A
Received October 15, 1987
Accepted March 17, 1988

Dynamic experiments were conducted on large-scale model piles in sand subjected to strong horizontal and vertical
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13

excitation. The theoretical response curves are calculated using DYNA2 and PILAY2 computer codes. and using also for the
vertical direction, the theory of nonlinear vibration. The theoretical curves are compared with the experimental results. The
dynamic behaviour of the pile is presented as frequency response curves for displacement, stiffness, dashpot constants, and
damping ratios. The influence of excitation intensity, repeated loading, and cap contact with soil on the dynamic behaviour of
single piles is investigated.
Key ~vorcls:dynamics, vibration, piles, foundations, nonlinear vibration, testing, deep foundations.

L'on a rCalisC des expCriences dynamiques sur des pieux rnodkles 2 grande Cchelle soumis B de fortes sollicitations
horizontales et verticales dans le sable. Les courbes des rkponscs thCoriques ont CtC calculCes au moyen des programmes
d'ordinateur DYNA2 et PILAY2 et en utilisant en plus, pour la direction verticale, la thCorie de vibration non IinCaire, et elles
ont it6 comparCes avec les risultats expirirnentaux. Le comportement dynamique du pieu est prksentd au moyen de courbes de
riponses pour le dkplacement, la rigidit6 et les constantes et rapports d'arnortissenient. L'on itudie ['influence, sur le
comportement de pieux simples, de I'intensitC de la sollicitation, et de la rCpCtition du chagenlent et du contact de capuchon
avec le sol.
Mots elks : dynan~ique,vibration, pieux, fondations, vibration non IinCaire, essais, fondations profondes.
For personal use only.

[Traduit par la revue]

Can. Geotech. 1. 25, 523-534 (1988)

Introduction
A number of solutions have been developed in the last 20
years for dynamic analysis of single piles subjected to har-
monic forces or moments. These solutions are based on dis-
crete models, continuum models, and finite element models.
However, they are usually formulated under the assumptions
that the soil behaviour is governed by the laws of linear elastic-
ity or viscoelsaticity and that the soil is perfectly bonded to the
piles.
In fact, nonlinearity in the area with high strain levels
and slippage, or even separation, in the contact area between
the soil and the piles often occur, especially for piles subjected
to strong excitation, as is the case for pile driving, offshore
rigs, seismic environments, and other applications. A rigorous
approach to slippage and nonlinearity is extremely difficult and
therefore approximate theories have to be used. Lumped mass
models with nonlinear discrete springs, dashpots, and friction
elements have been in use, particularly in the petroleum indus-
try (see, e.g., Matlock et al. 1978). Although such models can
reproduce strong nonlinear effects, it is difficult to relate the
properties of the discerte elements to standard geotechnical
parameters such as soil shear modulus and Poisson's ratio and
to account for pile group effects. Another approximate
approach, which includes a weak cylindrical zone around the
pile and is readily applicable to practice, was proposed by
Novak and Sheta (1980). In this approach, the stiffness -and
damping of a single pile are computed on the basis of a method
'Presently at Faculty of Engineering Science, The University of
Western Ontario, London, Ont., on leave from Institute of Engineer-
ing Mechanics, Harbin, China.
due to Novak (1974) and Novak ef nl. (1978, 1981, 1983)
whose application is facilitated by an efficient computer pro-
gramme, PILAY2 (Novak et 01. 198 1). The dynamic response
of the pile-supported foundation can be computed by a com-
puter programme, DYNA2 (Novak et al. 1983), which incor-
porates PILAY2.
Although there have been a large number of analytical
studies on the dynamic response, the published record of
experimental data is rather scarce. Novak and Grigg (1976)
and El-Sharnouby and Novak (1984) presented the results
obtained on small-scale pile groups. The dynamic response of
vertically loaded small-scale piles was measured by Roesset
et al. (1986). Full-scale pile foundation field test results were
provided by Gle (198 1). A large-amplitude vertical vibration
test of a full-scale group of piles was conducted by Blaney
et al. (1987). Simulated earthquake loading tests were con-
ducted by Finn and Gohl (1987) on a model pile foundation in
sand. A few other experimental investigations are described in
Nogami (1987). Most of these dynamic tests of piles are lim-
ited to small amplitudes of vibration or to small-scale piles.
In this paper, dynamic experiments on large-scale model
piles with strong horizontal and vertical vibration are
described. The objective is to study the dynamic behaviour of
single piles under strong excitation and to determine whether
the basically linear theory can reproduce the behaviour of piles
under strong excitation if suitable adjustments of the input
parameters are made. To this end, frequency response curves
and deflection curves of piles were experimentally established
in the field for different intensities of excitation and contact
conditions between the pile cap and the soil surface and were
compared with those calculated using the computer pro-
 524 CAN. GEOTECH. 1. VOL. 25, 1988
grammes DYNA2 and PILAY2. The influence of repeated
loading on dynamic behaviour of piles is investigated as well.
For vertical excitation, the nonlinear restoring force, effective
vibrating mass, and damping are also determined from the
measured response.
Theory
Before comparing the experimental results with the theory, it
may be useful to briefly expound on the assumptions used in
the theory and to introduce the soil and pile properties
employed in it. The main assumptions are that (1) the soil is
composed of independent horizontal layers that are linearly
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
viscoelastic with material damping of the hysteretic type and
(2) the pile is divided into beam elements along with the soil
layers and the soil below the pile is an elastic half-space.
Soil stiffness for the horizontal and rocking directions is
given by (Novak et al. 1978)
[ll Kus = G [ S " I ( ~ O , Y ,+
D )iSU2(ao,v,D)1
in which G is the shear modulus of soil, a. is dinlensionless
frequency (ao = Ro/Vs, where R is the pile radius, w is the cir-
cular frequency, and I/, is the soil shear wave velocity), D =
tan 6 = 2/3 is the material damping (/3 = the material damping
ratio), v is the soils's Poisson's ratio, and i = m; finally,
For personal use only.
S j l , Sj2 are the real and imaginary parts of the dimensionless
parameters, respectively. This approach to soil reactions is
obviously approximate but detailed con~parisonswith the finite
element solutions due to Kuhlemeyer (1976), Blaney et nl.
(1976), and others show that the approximate theery based on
[ I ] and [2] gives very good results. The advantage of this
approach is that nonhomogeneous soil profiles, that is, layer-
ing, can be included, as well as the cylindrical weakened zone
around the pile in which the soil properties differ from those of
the outer region. The properties of both the weak zone (inner
medium) and the outer region enter the soil stiffness expres-
sions. This makes it possible to account, in an approximate
way, for the effects of soil nonlinearity, slippage, and lack of
bond between the pile and soil that may lead to separation. The
role of all these factors is quite essential, as will be seen.
With harmonic motion, U(z,t) = U(z)efWr,the complete sol-
ution for the pile amplitude U(z) and pile impedances can be
obtained. From these, the stiffness and damping constants of
the pile are, for the horizontal direction,
for rotation,
and for the vertical direction,
In [3] - [5], V, is the shear wave velocity in the lowest layer of
soil at the pile tip; Ep is the Young's modulus of the pile, and I,
A are the pile's moment of inertia and cross-sectional area,
respectively; dl,?are dimensionless functions depending pri-
marily on the slenderness ratio e/R (where e is the pile length)
-.-
0-
1.5
0 I I I
0 5 10 15
E
FIG.1. Inverse value of sand Young's modulus vs. strain (1 psi =
6.895 kPa).
and the ratio Ep/G. The coupling impedances Ku4 and Cu4 are
similarly defined.
With the stiffness and damping constants of the pile estab-
lished, the response of a pile-supported footing to dynamic
loading can be predicted. This is done using the computer pro-
gram DYNA2 allowing for the coupling between horizontal
translation and rotation of the cap.
Experimental setup
The field tests of the single pile subjected to strong harmonic
excitation were conducted at the Institute of Engineering
Mechanics, Harbin, China, by Han et al. (1987).
The pile was a steel pipe with a diameter of 133 mm and a
length of 3.38 m. The pile properties are given in Table 1. The
pile was first place in a pit approximately 3.6 m deep and 1.5 m
in diameter. The pile cap was a concrete block 200 mm thick,
770 mm long, and 690 mm wide with a mass of 250 kg. The
exciter was fixed on the cap; its mass was 120 kg. The centre
of gravity of the cap-exciter system was 3 mm below the cap
surface.
The washed medium sand was then placed in the pit in
200 mm lifts after the pile had been positioned in the centre of
the pit. The sand was compacted in place with a mechanical
vibrator. Density and water content measurements were taken
for each lift. In this way good contact between the pile and soil
was secured.
Soil properties that are basic in dynamic analysis are the bulk
density and the shear wave velocity. Several types of static and
dynamic tests were run in the laboratory and in situ on the
washed medium sand and the undisturbed natural deposit
around the pit. The static tests included specific gravity, den-
sity, and water content, while the dynamic tests were triaxial
and seismic crosshole tests. The samples of sand taken from
the pit were tested under dynamic triaxial conditions in the
laboratory. The Young's modulus variation with strain and
effective stress i~~ are shown in Fig. 1. At a dynamic strain of
E = the shear modulus at the pile tip was G, =
14.69 MPa. With the mass density of sand p, = 1700 kg/m3,
the shear wave velocity was Vr = JG,/p, = 93 m/s.
The density and water content of the medium sand were uni-
form in every lift. The properties of the backfill soil are given
 HANANDNOVAK

F.. Pi*'
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13

FIG.2. Properties of pile and soil used in analysis (1 psi = 6.895 kPa).

TABLE
1. Pile properties (steel pipe) TABLE2. Properties of backfill soil
(medium sand)
Property Value
Property Value
For personal use only.

Outer diameter 133 mm

Inner diameter 121 mm
Moment of inertia 4.84 x lo6 mm"
Area 2390 mm2
Length 3380 mm
Young's modulus 2.06 X lo5 MPa
Poisson's ratio 0.25
Damping rate 0.01
Specific weight 7.65 X N/mm3
in Table 2. It is assumed that the distribution of shear modulus
in sand is parabolic with depth.
The soil profile around the pit was established from ground
surface to a depth of 20 m. The soil is homogeneous sandy
clay, with yellow and brown colouring. The shear wave veloc-
ity of the undisturbed natural deposit was measured using seis-
mic crosshole tests. The properties of the natural deposit are
given in Table 3.
The shear wave velocity of the clay outside the pit is about
twice that of the sand backfill and therefore the effect of the
interface between the two soil media has to be assessed. To
approximate this effect the PILAY2 programme was used to
evaluate the pile dynamic deflections for two cases: a horizon-
tally homogeneous medium comprising only the sand and a
composite medium comprising the inner zone of sand in the pit
and the outer zone of clay. For low dimensionless frequencies
ao, typical of pile response, the differences in the deflections
were quite small. However, it is possible that the geometric
damping of the piles tested was smaller than it would have
been in a horizontally homogenous medium. This question is
discussed in more detail in the last part of this paper.
Displacement pickups, strain gauges, and compressive stress
transducers were fixed along the pile shaft. Displacement and
acceleration pickups were mounted on the pile cap. A diagram-
matic sketch of the pile and the arrangement of the instruments
are shown in Fig. 2a.
Bulk density 16.7 kN/m3
Dry density 15.7 kN/m3
Void ratio 0.67
Moisture content 7%
Shear wave velocity 93 m/s
(at the tip of pile)
An exciter with two counterrotating eccentric masses was
used to produce the harmonic excitation. The magnitude of the
exciting forces was changed by adjusting the angle of the
eccentric masses, 8. Several magnitudes of the exciting force
were selected in the experiment as shown in Table 4 (f in the
table is the frequency of the exciting force in hertz).
In the experiments, two types of contact conditions between
the pile cap and the soil surface were considered. In the first
case, the pile cap rested directly on the soil surface (connected
situation) and two different excitation intensities were applied,
8 = 5 and 8" (the symbols 8 = 5, and 8, are used to identify
this case later in the paper). In the second case, the pile cap
was situated 0.20 m above the soil surface (unconnected situa-
tion) and the excitation intensities were 8 = 5, 8, and 14".
Because of the nonlinearity of the soil-pile system, the
magnitude of exciting force and the sequence of loading would
affect the dynamic characteristics of the system. The steady-
state response of the pile to harmonic forces was measured
with increasing excitation intensities and different frequencies.
Pile response under horizontal excitation
Experimental horizontal response
Using the equipment and technique described, response to
horizontal excitation was measured.
Three frequency response curves obtained with 8 = 5, 8,
and 14" for the unconnected case are shown in Fig. 3 and
denoted by the angle 0 of the exciter. The horizontal displace-
 526 C A N . GEOTECH. J. VOL. 2 5 , 1988
TABLE
3. Properties of natural deposit (hon~ogeneoussandy clay)
Depth Unit weight Void rrltio Shear wave velocity
(m) (kN/m3) e (mls)
ments of the pile cap vary with frequency and indicate a sharp
resonance peak in all cases. The maximum displacement mea-
sured on top of the cap amounts to 2-3 mm, the maximum
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
acceleration amounts to 0.8-0.9 g. For steady-state vibration
this represents very intense motion.
In a linear system the resonance peaks occur at one fre-
quency, no matter what the excitation intensity is. However,
Fig. 3 shows that with increasing excitation intensity, the mea-
sured resonance peak shifts from about 7.6 to about 4.7 Hz.
This indicates a reduction of stiffness-proportional to f2-to
almost one-third of the highest value associated with the lowest
excitation intensity. Also, the vibration amplitudes are not pro-
portional to excitation intensity at all frequencies. These are
typical features of nonlinear vibrations. However, the inidvid-
ual response curves observed intersect each other and are not
centred around a joint backbone curve, Q, as is inherent to a
system whose properties are defined by one nonlinear charac-
For personal use only.
teristic of the restoring force and one nonlinear characteristic
of nonlinear damping. An example of such theoretical response
curve is shown in Fig. 4. This response curve was calculated
for one softening characteristic of the restoring force and non-
linear damping proportional to the nth power of vibration
velocity with n < 1, that is, with damping decreasing with
vibration amplitudes, using the technique described in Novak
(1971). Such nonlinear characteristics of the restoring force
and damping are, in general, indicated by Fig. 3. The ampli-
tude increments grow with excitation intensity, and are com-
parable to the amplitude increments of the theoretical system
(Fig. 4) at the frequency of the linear system, wo. The presence
of the response curve intersections in Fig. 3 and their absence
in Fig. 4 suggest that the restoring forces of the real system do
not follow one nonlinear characteristic. Rather, it seems that
with increasing excitation intensity the restoring force charac-
teristic stabilizes after many cycles of load application and to a
high degree linearizes within a range of peak amplitudes, A ;
with a further increase in excitation intensity the restoring
force again stabilizes but on a lower level of stiffness. These
notions are further supported by observations made with hori-
zontal repeated loading and by the analysis of the vertical
response later herein. The features of the response observed
suggest that a linear theory might reproduce the main nonlinear
features of the steady-state response if the input parameters are
suitably adjusted. This possibility is explored in the next para-
graphs.
Comparison of e+xperirnental horizorztal response with theo-
retical predictions
The theoretical response curves are calculated using the code
DYNA2. The weak zone around the pile and free pile length
available in this code give the analysis considerable versatility.
The shear modulus distribution of the sand backfill is
assumed to be parabolic along the pile, which is consistent
with the effect of confining pressure apparent in Fig. 1, and the
soil below the pile tip is assumed to be homogeneous. For dif-
w
0 5 10 15
Frequency ( Hz )
FIG.3. Theoretical and experimental horizontal response of pile cap
for different excitation intensities (pile cap separated from soil).
TABLE4. Magnitude of horizontal exciting force (frequency f
in Hz)
Angle of eccentric
masses, 0 (deg) 5 8 14 28
Exciting force (N) 2.39 f' 3.79f' 6.75f' 10.23f"
ferent excitation intensities, the soil properties in the weakened
zone and at the top of the sand backfill are adjusted so that the
theoretical response curves approach the measured data. The
characteristics of the pile and soil used in the calculations are
given in Fig. 2. The material damping of the weak zone is of
less importance than the thickness of the weak zone and its
shear modulus. The comparison of experimental results with
theoretical response curves is shown in Figs. 3 and 5 . The fit
achieved is quite good.
From Fig. 2, it can be seen that the soil properties in the
weakened zone and at the top of the backfill varied with the
intensity of excitation and the change of the contact condition
between the pile cap and the soil surface. As the excitation
intensity increases, the shear modulus ratio, G,,/G, is reduced,
whereas the thickness ratio, t,/R, and the material damping are
increased (G, and t,, are the shear modulus and thickness of
the weak zone, respectively). For the connected situation (6 =
5, and 8,), the soil shear modulus at the top of the backfill is
reduced with the increase in the excitation intensity. When the
connected situation changes to the unconnected situation (from
6 = 8, to 6 = 8), G,/G decreases, t,/R increases, and material
damping increases.
Using the ratio G,,/G = 0 in the topmost layer, the separa-
tion between the pile and the soil is accounted for. For the
unconnected situation, the separation was caused by strong
excitation. The depth of the anticipated separation, eo, ranges
from 6.8d for 6 = 8" to 9.8d for 6 = 14", where d is pile
diameter.
From Figs. 3 and 5 it can be seen that as the excitation inten-
 HANANDNOVAK
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
FIG. 4. Theoretical response curves of a system with nonlinear
restoring force and nonlinear damping (11 < 1) for increasing excita-
tion intensity ( p = excitation moment/mass = m,e/n2).
For personal use only.
Unconnected Connected
Measured Data
Frequency ( Hz 1
FIG. 5. Theoretical and experimental response curves with different
excitation intensity and contact between cap and soil (horizontal exci-
tation).
sity increases, from 8 = 5 " to 8 = 14", the resonant frequency
of the pile decreases and the resonant displacement increases.
Meanwhile, for the same magnitude of exciting forces, the
resonant frequency of the unconnected situation is markedly
lower than that of the connected situation and its resonant dis-
placement is larger than that of the connected situation. Appar-
ently, for the single pile tested, the cap contributed to the
stiffness and damping of the pile, as was also found by Novak
and Grigg (1976); however, in practice this observation should
0 0
0 5 10 15 20 25
Frequency ( H z )
FIG. 6. Pile horizontal stiffness and damping coefficient.
be applied with caution because the sand under the cap may
settle more than the piles, particularly with end bearing piles.
For embedded caps the soil reactions acting on the vertical
sides can generate a significant amount of damping and be
quite beneficial.
Pile stifjizess arzd damping
The theoretical stiffness and damping coefficients of the pile
for the horizontal and rocking directions are shown in Figs. 6
and 7 , respectively. The subscript u indicates horizontal
translation while C$ stands for pile head rotation (rocking). The
stiffness is almost frequency independent, as would be
expected, because of low dimensionless frequencies the pile
dynamic stiffness is known to be close to the static one.
Figures 6 and 7 also show the pile damping coefficients. These
rapidly increase as frequency decreases. This is primarily due
to the conversion of frequency-independent soil material
damping to equivalent viscous damping coefficient c = 2610-1.
To provide a dimensionless measure of pile damping, the
damping ratios can be evaluated from the stiffness and damp-
ing coefficients as cwl(2k). These damping ratios are shown in
Figs. 8 and 9. The damping ratios increase with frequency but
the rate of increase is lowest for the strongest excitation. It can
also be seen from Figs. 6-9 that the damping coefficients and
damping ratio mostly decrease with excitation intensity and
thus with vibration amplitudes for frequencies higher than the
lowest resonance frequency observed (4.7 Hz). This is further
illustrated by Table 5. In soils, an increase in damping with
strain is usually assumed. This apparent discrepancy suggests a
growth in pile separation with vibration amplitudes and the
weakening of the soil around the pile, as reflected in the
necessary reduction in the shear modulus in the weak zone.
Pile deflections
The measured deflections of the pile varied with frequency
 528 CAN. GEOTECH. J . VOL. 25. 1988
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
"
0 5 10 15 20 25"
Frequency ( Hz
FIG.7. Pile rocking stiffness and damping coefficient.
TABLE
5. Damping parameters of the pile
Excitation test
Parameter B=5, B=8, B=8
Resonance frequency (Hz) 12.2 8.5 5.6
Damping ratio
Horizontal, D,, 0.163 0.1 I0 0.090
For personal use only.
Rocking, D,> 0.080 0.070 0.063
and excitation intensity. A comparison of the measured data
with the theoretical curves calculated by the PILAY2 code is
shown in Figs. 10 and 1 1 for I9 = 8 and 14", respectively. The
theoretical characteristics of the pile and soil used in the
PILAY2 code are the same as those used in the response analy-
sis. It can be seen that the measured data agree well with the
theoretical deflection curves.
Comparing the dynamic deflection with the static deflection
of the pile, it appears that the effective pile length is about
13.5d under strong horizontal excitation, and about 11.0d for
static loading. That is, the effective pile length for dynamic
loading is slightly longer than that for static loading, although
the horizontal displacements along the pile length caused by
dynamic loading may be larger than those due to static loading.
Effect of repeated loading
The soil around the pile softens and separates under repeated
loading. Accordingly, the stiffness, damping, and dynamic
response of the pile varies with repeated loading. To obtain
data on these effects, dynamic experiments on a single pile
under repeated loading were also conducted. Another pile was
embedded in another sand pit, with the properties of the pile
and soil, as well as the test conditions, being the same as
before. The contact condition between the pile cap and the soil
surface was of the two types, that is, connected and uncon-
nected. For each case, the pile was excited three times, from
low to high frequency, with the excitation intensity kept the
same in all runs. The comparison of measured data with theo-
retical response curves for the connected and unconnected
cases is shown in Figs. 12 and 13, respectively. The measured
resonance peaks shift to lower frequencies and become sharper
with loading repetition, indicating reduction in both stiffness
and damping. With judicious choice of input data, the theoret-
ical response curves agree with the experimental results quite
I I I I
0 5 10 15 20 25
Frequency ( H z 1
FIG.8. Pile horizontal damping ratio.
0=14
4.7
0.090
0.060
I I I I
0 5 10 15 20 25
FREQUENCY ( Hz 1
FIG.9. Pile rocking damping ratio
well. The reduction in theoretical stiffness and damping with
load repetition can be seen from Table 6. Because of the
loosening of soil around the pile, soil nonlinearity, and pos-
sibly slippage under the pile cap, the pile stiffness is reduced in
the connected condition also.
The difficulty with the inclusion of the weak zone in the
analysis is that its parameters are difficult to assess before-
hand. Thus the question arises of whether a correction for pile
separation in the form of a pile free length would not be suffi-
cient to account for the reduction in pile stiffness and damping
observed in the experiments. The need for a free length was
already observed by other investigators, for example, Novak
and Aboul-Ella (1978a), Novak and El-Sharnouby (1984), and
Roesset et al. (1986).
For the unconnected cap, the theoretical separation between
the pile and soil is eo = 3.8d for the first excitation, eo = 5.3d
for the second, and eo = 6.8d for the third. The separation
lengths required in the analysis to provide a good fit with the
experimental data for the different resonant amplitudes mea-
sured are plotted in Fig. 14. Analyses both with and without
the weak zone were employed and gave satisfactory results,
although the weak zone allows finer tuning. The separation
lengths needed increase with displacement amplitude and are
 HANANDNOVAK

0 ' 9'-' - I
/Or
/ -
0 '0
/ /
- o/
/
0 - /O
a3 C - /
M
M
I - - - Calculation I

C-
o Measurement
w -
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13

FIG. 10. Comparison of theoretical curves and measured deflections along pile shaft (excitation 0 = 14).

0
0 0.5 1.0 1.5mm 0 0.25 0.5

;
/

' ' f7
For personal use only.

- - - Calculation
0 Measurement

f = 6Hz
(Fx = 136 IN
1
f = 8Hz
(Fx = 2 4 2 N )
FIG. 11. Comparison of theoretical and measured deflections along pile shaft (excitation 6 = 8).

greater if no weak zone is considered. 'The maximum separa-
tion is about 10 pile diameters. The separation is given in terms
of diameter but: in general, the absolute depth must be a basic
parameter because it controls the confining pressure. With the
pile diameter of 133 mm the maximum separation length was
thus 1.33 m.
Pile separation was indicated in full-scale offshore towers
where it was observed that piles loosen up during a storm,
lowering the tower fundamental frequency for a period of time,
but recover the stiffness during a subsequent period of calm.
Pile response under vertical excitation
Experimental vertical response
After the testing of the pile under horizontal excitation, ver-
tical excitation was applied, with all the other experimental
conditions remaining the same.
The excitation intensity was increased stepwise and three
frequency response curves, shown in Fig. 15, were obtained.
Sharp intersecting resonance peaks are indicated like the ones
shown in Fig. 5. The maximum displacement measured on the
top of the cap is about 1.4 mm and the maximum acceleration
reaches about 1.3 g. Markedly nonlinear features are again
manifested particularly by the drop in the resonant frequency
with the increase in excitation intensity (O), but this drop is less
than it was under horizontal excitation.
In the case of vertical excitation, the nonlinear features of
the measured response curves can be studied more easily
because this response can be treated in one degree of freedom
(S. D.O.F.) and is, therefore, quite amenable to nonlinear
analysis. Such analysis is employed in the next section.
Evaluation of the vertical response
Nonlinear S. D. 0.F. analysis
To identify the parameters of the soil -pile system from the
measured response displaying nonlinear features, the inverse
problem of nonlinear vibration has to be approached. This
means that the system mass, stiffness, and damping have to be
established from the experimental response. To this end, the
methodology formulated in Novak (1971) is employed. This
procedure assumes that system stiffness is independent of fre-
quency; this is also usually assumed for slender piles in
medium soil in practical applications.
Assuming that the restoring force is nonlinear but the damp-
ing force is linear (viscous), the backbone curve, which
defines the undamped natural frequencies a,can be deter-
mined to a given response curve using the simple relation
(Novak 1971)
 530 CAN. GEOTECH. J . VOL. 25, 1988
Measured Data
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
0 5 10 15
Frequency ( Hz
FIG. 12. Theoretical and experimental response curves for repeated
loading (0 = 5 , cap separated).
Fx e i a t
0.7 -1-
For personal use only.
/ I Measured Data
Frequency ( H z )
FIG. 13. Theoretical and experimental response curves for repeated
loading (0 = 5,, cap in contact with soil).
In [6], ol and w2 are the frequencies corresponding to the
points of intersection between the response curve and a line
passing through the origin of coordinates (Fig. 15). Intersect-
ing the response curve by a trace of such lines, the backbone
curve a(A) can be constructed to each response curve. These
backbone curves are plotted in Fig. 15. It can be seen that each
response curve has its own backbone curve, unlike the theoret-
ical response curves shown in Fig. 4. This indicates that the
stiffness characteristic of the system varies with the overall
level of vibration intensity rather than being unique. This is
confirmed by establishing the stiffness characteristics corre-
sponding to the individual backbone curves.
The stiffness characteristic can be assumed to be of two
Max. Amplitude,
Pile Diameter x
FIG. 14. Theoretical pile separation vs. dimensionless anlplitude.
types: A truly nonlinear characteristic that is followed for all
amplitudes A and a characteristic, F(A), that is nonlinear but
for every steady-state amplitude A linearizes to give the equiv-
alent linear stiffness depending on amplitude A , that is,
The latter assumption appears more suitable and is adopted
here. Then, for a restoring force expressed by a power series
[8] F ( A ) = k l + k 3 A 3 + k 5 A 5 + . . . +k,,A"
where ki are constants, the amplitude-dependent natural fre-
quency is given by
in which m is the system effective mass.
The effective mass and damping can be calculated exploiting
the geometric properties of the response curves (Novak 1971).
The mass and damping found are given in Table 7. The value
of the effective mass, m , is much greater than the total mass of
the pile, cap, and exciter, M. The apparent additional mass can
be expressed in terms of the coefficient
ill - M
[ lo] { = ------
M
which is also given in Table 7. The additional mass is very
substantial and decreases with increasing excitation intensity,
suggesting that partial separation of the pile from soil might
have occurred with higher excitation intensity. It seems that
the apparent mass increase results from the assumption that
pile stiffness is constant, that is, frequency independent. In the
continuum approaches, such as those of Novak (1974) and
Novak et al. (1981), no additional mass is needed because soil
mass is automatically accounted for, making the pile stiffness
frequency dependent. This effect can be quite strong, particu-
larly for stiff piles vibrating vertically and will be demon-
strated in the last part of this paper.
With the mass pertinent to each response curve the restoring
force characteristic follows from [7] and [9] as
 HANANDNOVAK
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13

0 10 20 30
Frequency ( Hz
For personal use only.

FIG. 15. Vertical pile response measured and back-calculated for three levels of excitation intensity (w = 2af).

For the three response curves shown in Fig. 15 the correspond-

ing restoring force characteristics, calculated by [ l l ] , are
plotted in Fig. 16. A few observations emerge: each response
curve has its own characteristic, the overall stiffness markedly
decreases with increasing excitation intensity (8) but the non-
linearity of individual characteristics is much less pronounced
than the overall nonlinearity of the whole set of the response
curves. The latter observation is consistent with those made
with regard to the horizontal response and supports the
assumption of linearization implied in [7].
The damping ratio derived from the response curves is also
given in Table 7. The damping found is much smaller than
what would be expected for a lightly loaded fully embedded
pile and increases with excitation intensity. Under horizontal
excitation, the damping decreased with increasing amplitudes.
This indicates that the separation effect is somewhat less
severe in the vertical vibration than it is in the horizontal vibra-
tion and that slippage generating friction may contribute to the
increase in the vertical damping. 0
With the soil-pile system mass, damping, and restoring 0 0.5 1 .O 1.5
force characteristic derived, the nonlinear response curves Vertical Displacement ( rnrn 1
were back-calculated using the nonlinear theory (Novak 1971) FIG. 16. Pile restoring force characteristics vs. pile head displace-
and are plotted as solid lines in Fig. 15. The theoretical non- ments corresponding to response curves from Fig. 15.
linear response curves agree with the measured ones very well.
Thus, a very good theoretical prediction can be made assuming
a nonlinear restoring force characteristic and a linear, viscous of soil-pile interaction analysis and the properties of soil as
damping whose constant depends on the peak amplitude. well as the pile. This analysis was conducted using the code
DYNA2 (Novak et al. 1983), which incorporates the code
Linear soil -pile itzteractiotz analysis PILAY2. This interaction analysis confirms that extensive pile
The theoretical response curves shown in Fig. 15 were back- separation must have occurred. This can be seen from Fig. 17
calculated using the S.D.O.F. data derived from the experi- in which the large amplitude experimental response is com-
ments. Another task is to predict the pile response on the basis pared with the theoretical one calculated using DYNA2. When
 532 CAN. GEOTECH. J . VOL. 25. 1988

TABLE
6. Dynamic parameters of pile under repeated loading

Connection
Parameter Connected Unconnected
Order of excitation 1st 2nd 3rd I st 2nd 3rd
Resonance frequency (Hz) 13.8 12.6 12.2 7.6 6.5 5.7
Stiffness
Horizontal, K, (kN/mm) 7.43 6.50 5.89 3.04 2.15 1.78
Rocking, Kd (kN.mm) 2.75 x 1 0 v . 6 4 x loh 2.59 x 1 0 N . 2 9 x loh 2.08 x lo6 1.96 x lo6
Damping ratio
Horizontal, D,,
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13

Rocking, D,

TABLE
7. Nonlinear vibrational parameters of single pile in vertical direction

Effective mass
Stiffness
Mass coefficient Damping ratio k
Exciting force Mass 171 (kg) < P (kN/mm)
0 = 8" 1730 3.01 0.042 23.9
6 = 14" 1350 2.05 0.060 15.7
0 = 28" 993 1.24 0.075 7.39
For personal use only.

0 Measured
Real Mass = 433 kg
Separation != 1.8rn
Weak Zone G m / G = 0.1
+
c
1.0- \ Dm=0.25tm/R=2.0
EaJ
v
-
0 0
n 0
.-In
n
- O 0
0.5 - o
.-
+
0
L

8
0 Calculated i-
No Separation
Gm/G = 0.5
0-5 10 15 -,!La-
2 25
0
Frequency ( Hz
FIG. 17. Experimental pile response in vertical vibration vs. theoretical prediction calculated using DYNA2 (0 = 28).

no separation is assumed (curve A) the theoretical response is
heavily damped and the stiffness is very high, bringing the
resonance peak beyond the experimental frequency range. A
large separation of 1.80 m, a weakened zone around the pile,
and a true mass of the system yielded curve B and a reasonable
agreement with the measured data. The discrepancy in the
descendent branches of the response curves is due to the linear
theory employed in this case, which cannot give a perfect
agreement with the actual nonlinear response over the whole
frequency range. Additional data on the vertical response
analysis can be found in Han and Novak (1988).
Effect of limited size of test pit
Finally, it is of interest to examine the effect of the limited
extent of the pit in which the tests were conducted. The pit was
filled with sand whose properties differed from those of the
surrounding soil. Consequently, an interface occurs between

Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
FIG. 18. Notations and geometry for pile in test pit.
For personal use only.
-I
0 100 200
FREQUENCY 0 = 2 7T f
FIG. 19.Pile stiffness parameter f,, for different test pit dimensions
(Gn,/G=0.25,Dm = D = O.lO,pm = p, v = 0.4).
the two media from which waves may reflect back into the pit,
complicating the wave pattern within the pit and creating what
is known in analysis as the "box effect." To examine this
effect, the vertical pile stiffness and damping were reanalyzed
using the theory of Novak and Sheta (1980) in which an annu-
lar zone is allowed for around the pile with soil properties dif-
ferent from those of the outer soil. Because of the large extent
of the pit the mass of the inner zone (sand) was accounted for,
whereas it is neglected in PILAY2. (A similar analysis was
presented by Lakshmanan and Minai (1981) and Veletsos and
Dotson (1986).) The geometry and notations for this analysis
HANANDNOVAK
0
0 100 200 300
FREQUENCY O = 2 T f
FIG.20. Pile damping parameter fw2for different test pit dimensions
(Gm/G = 0.25, Dm = D = 0.10, p, = p , v = 0.4).
are shown in Fig. 18. The subscript m denotes the soil in
the pit.
The dimensionless stiffness and damping parametersfwl and
fw2 defined by [5] are plotted in Figs. 19 and 20. The results
are shown for three pit dimensions characterized by the ratio of
the zone thickness, t,, to pile radius, R, and the actual soil
properties of the site. Shown as solid lines for comparison are
the pile parameters corresponding to homogeneous soil
medium, that is, t, - co. The site t,/R ratio is close to 10.
The presence of the interface can be seen to have only a small
effect on pile stiffness (Fig. 19) but has a profound effect on
radiation damping. At low frequencies the damping is slightly
increased but at higher frequencies it is dramatically decreased
compared with that of the horizontally homogeneous medium.
This effect of the interface on pile damping is marked but is
much less significant than pile separation, as the response
curves plotted in Fig. 17 demonstrate.
The strong variation in the stiffness parameter fwl with fre-
quency (Fig. 19) is also of interest. It is close to parabolic and
can be interpreted as k,, - where k,, is the pile static stiff-
300 ness and p is the effective mass. Because the pile stiffness, like
k,,, is assumed to be frequency independent in the S.D.O.F.
nonlinear analysis, the frequency variation of the actual stiff-
ness calls for the inertia term, ~ w " and hence for additional
mass as found in Table 7.
Conclusions
Dynamic experiments with large-scale model piles subjected
to strong harmonic excitation in the field were conducted. The
measured response curves are compared with the theoretical
curves calculated using DYNA2 and PILAY2 codes. The com-
parison suggests the following conclusions:
1. Pile response at large amplitudes exhibits typical nonlinear
features.
2. Consideration of a weakened zone around the pile and pile
separation is necessary for piles subjected to strong vibration.
3. Good agreement can be achieved between the experimental
 534 CAN. GEOTECH. J . VOL. 25. 1988
and theoretical response curves when the characteristics of soil
are suitably chosen, without resort to truly nonlinear analysis.
The excitation intensity and contact conditions between the
pile and soil and between the cap and soil change the stiffness
and damping of piles.
4. Under repeated loading, the stiffness and damping of single
piles decrease, so that the resonant frequency decreases and
resonant displacement increases.
5. The weak zone around the pile considered in the computer
codes yields lower damping of the pile-soil system than that
evaluated without a weak zone. T h e damping values calculated
in this way are closer to the experimental results.
6. Under dynamic horizontal deflections the effective pile
Can. Geotech. J. Downloaded from www.nrcresearchpress.com by Duke University on 07/18/13
length is slightly longer than it is under static loading.
7. Dynamic experiments conducted in a test pit may be affected
by wave reflections from the interface between the soil in the
pit and the soil outside it. Damping of the test body can be
more affected than its stiffness.
8. Further research is needed to develop criteria for pile separa-
tion prediction and pile nonlinear dynamic analysis.
Acknowledgements
The experiments were done in China and the theoretical
analysis was done at T h e University of Western Ontario,
London, Canada. The first author would like to express his
appreciation to Canadian International Development Agency -
For personal use only.
WUSC China Programme for the support that made his stay at
The University of Western Ontario possible.
BLANEY, G. W., KAUSEL, E., and ROESSET, J. M. 1976. Dynamic
stiffness of piles. 2nd International Conference on Numerical
Methods in Geomechanics, ASCE, New York, pp. 1001 - 1009.
BLANEY, G. W., MUSTER, G. L., and O'NEILL,M. W. 1987. Verti-
cal vibration test of a full-scale pile group. It1 Dynamic response of
pile foundations. ASCE Geotechnical Special Publication 1I , pp.
149- 165.
EL-SHARNOUBY, B., and NOVAK,M. 1984. Dynamic experiments
with a group of piles. ASCE Journal of Geotechnical Engineering,
110: 719-737.
FINN,W. D., and GOHL,B. 1987. Centrifuge model studies of piles
under simulated earthquake later loading. In Dynamic response of
pile foundations. ASCE Geotechnical Special Publication 11, pp.
21 -38.
GLE, D. R. 1981. The dynamic lateral response of deep foundation.
Ph.D. thesis, University of Michigan, Ann Arbor, MI.
HAN,Y. C., and NOVAK,M. 1988. Nonlinear vibration of single pile
under vertical harmonic excitation. Proceedings of the 3rd Interna-
tional Conference on the Application of Stress-Wave 'Theory to
Piles, Ottawa, Vol. 1.
HAN,Y. C., SHOU,P. X., and LIU,D. 1987. Research in vibration
tests of single piles. Journal of the Mechanics of Soil and Rock.
(To be published in Chinese.)
KUHLEMEYER, R. L. 1976. Static and dynamic laterally loaded piles.
Department of Civil Engineering, The University of Calgary,
Calgary, Alta., Research Report CE76-9, p. 48.
LAKSHMANAN, N., and MINAI,R. 1981. Dynamic soil reactions in
radially nonhomogeneous soil media. Disaster Prevention Research
Institute, Kyoto University, Bulletin 31, pp. 79- 114.
MATLOCK, H., FOO,H. C., and BRYANT, L. M. 1978. Simulation of
lateral pile behavior under earthquake motion. Proceedings, ASCE
Specialty Conference on Earthquake Engineering and Soil
Dynamics, Pasadena, CA, Vol. 11, pp. 600-619.
NOGAMI, T., editor. 1987. Dynamic response of pile foundations-
experiment, analysis and observation. In Dynamic response
of pile foundations. ASCE Geotechnical Special Publication 11,
p. 186.
NOVAK,M. 1971. Data reduction from nonlinear response curves.
ASCE Journal of the Engineering Mechanics Division, 97(EM4):
1187-1204.
1974. Dynamic stiffness and damping of piles. Canadian
Geotechnical Journal, 11: 574 -598.
NOVAK,M., and ABOUL-ELLA, F. 1978n. Impedance functions for
piles embedded in layered media. ASCE Journal of Engineering
Mechanics, 104(EM3): 643 -661.
1978b. Stiffness and damping of piles in layered media. Pro-
ceedings of the ASCE Specialty Conference on Earthquake Engi-
neering and Soil Dynamics, Pasadena, CA, Vol. 11, pp. 704-719.
NOVAK,M. and EL-SHAKNOUBY, B. 1984. Evaluation of dynamic
experiments on pile group. ASCE Journal of Geotechnical Engi-
neering, 110: 738-750.
NOVAK,M., and GRIGG,R. F. 1976. Dynamic experiments with
small pile foundations. Canadian Geotechnical Journal, 13:
372 - 385.
NOVAK, M., and SHETA,M. 1980. Approximate approach to contact
effects of piles. Proceedings, Dynamic Response of Pile Founda-
tions: Analytical Aspects, ASCE, Florida, pp. 53-79.
NOVAK, M., NOGAMI, T., and ABOUL-ELLA, F. 1978. Dynamic soil
reactions for plane strain case. ASCE Journal of Engineering
Mechanics, 104(EM4): 953 -959.
NOVAK,M., ABOUL-ELLA, F., and SHETA,M. 1981. PILAY2, a
computer program for calculation of stiffness and damping of piles
in layered media. Systems Analysis Control and Design Activity,
Faculty of Engineering Science, University of Western Ontario,
London, Ont.
NOVAK, M., SHETA,M . , EL-SHARNOUBY, B., and EL-HIFNAWY, L.
1983. DYNA, a computer program for calculation of response of
rigid foundation to dynamic loads. Systems Analysis Control and
Design Activity, Faculty of Engineering Science, University of
Western Ontario, London, Ont.
ROESSET, J. M., STOKOE, K. H., BAKA,J. E., and KWOK,S. T. 1986.
Dynamic response of vertically loaded small-scale piles in sand.
Proceedings of the 8th European Conference on Earthquake Engi-
neering, Lisbon, Vol. 2, pp. 5.6165-5.6172,
VELETSOS, A. S., and DOTSON,K. W. 1986. Impedance of soil layer
with disturbed boundary zone. ASCE Journal of Geotechnical
Engineering, 112: 363-368.