Professional Documents
Culture Documents
DENSE SAND
By D. Bruno1 and M. F. Randolph2
ABSTRACT: Systematic studies of pile response during static and dynamic load tests are generally too expensive
to conduct in the field, and at model scale may be limited by scaling effects and the ability to obtain accurate
stress waves. This paper describes such a study, conducted at model scale in a geotechnical centrifuge, for piles
driven into dense sand. Adverse scaling effects were minimized by the use of extremely fine sand (silica flour),
and accurate stress waves were obtained using high-frequency data logging, together with a Hopkinson bar
arrangement for the measurement of pile-head velocity. The overall aim of the study was to compare dynamic
and static test data, for open- and closed-ended piles driven into dense sand, for a range of delivered hammer
energies. Open- and close-ended model piles were driven into dense sand and statically load tested at different
Downloaded from ascelibrary.org by Aalborg University on 01/23/24. Copyright ASCE. For personal use only; all rights reserved.
penetrations, without stopping the centrifuge. Stress-wave data were collected, during continuous driving and
from single blows immediately prior to the commencement of static load testing. An assessment of the accuracy
of the mobilized soil resistance estimated from dynamic testing, using different levels of analysis, has been made
by direct comparison with the static load test data. Particular emphasis has been placed on the performance of
the dynamic analyses in light of different driving conditions and delivered hammer energy. From measurements
of the delivered hammer energy, the efficiency of the centrifuge pile driving system was also assessed.
INTRODUCTION tests arises for open-ended piles, where the mode of penetra-
The assessment of the static bearing capacity of piles using tion under dynamic conditions, with soil moving up inside the
dynamic testing is of major interest because of the lower cost open pipe pile, differs from that during static loading, where
of dynamic testing, and also in situations such as offshore the pile will generally behave as fully plugged. It might be
operations, where static loading is generally not feasible. Com- expected that dynamic testing would significantly underesti-
parisons between dynamic and static estimates of pile capacity mate the fully plugged pile capacity under static loading, par-
have been reported widely in the literature (Broms 1981, 1985; ticularly under the relatively high-energy levels needed to
Fellenius 1988; Barends 1992; Goble and Likins 1996; Town- cause sufficient displacement of the pile under dynamic con-
send 1996). Open-ended piles are an area of particular diffi- ditions.
culty because of the different modes of penetration under This paper presents results from an experimental study of
dynamic and static loading conditions. Systematic studies are pile capacity under dynamic and static conditions, with the
needed in uniform soil conditions, where differences between main focus being the effect of delivered hammer energy on
dynamic and static loading can be evaluated for both open- the dynamic response (and estimated soil resistance), and the
and closed-ended piles, for a range of delivered hammer effect of the pile tip conditions (open- and closed-ended) on
energy. the dynamic and static responses. Both compression and ten-
The high cost of field testing necessitates that such studies sion load tests were performed, at varying pile embedment
are undertaken, at least initially, at model scale. An extensive ratios. However, the present paper is restricted to a comparison
centrifuge testing program has been performed with the aim of the responses in compression only.
of examining the drivability of open- and closed-ended piles
in dense sand, and comparing capacities measured under both CENTRIFUGE MODELING
dynamic and static loading conditions. Because it is difficult, Centrifuge modeling has a number of advantages over al-
at model scale, to obtain accurate measurements of the high ternative modeling techniques in geotechnical engineering, al-
accelerations associated with pile driving, the pile velocity lowing a wide range of soil and pile parameters to be varied
during a hammer blow has been determined using a Hopkin- independently while maintaining correct scaling of the self-
son bar arrangement (Zelikson 1991; Bruno and Randolph weight stress gradient in the soil inside and outside the model
1998). In this approach, the pile velocity is deduced directly pile. Correct reproduction of internal and external stresses ad-
from the stress wave traveling in a light plastic rod attached jacent to the pile wall allows an accurate assessment of the
near the pile head. Measurement of both force and velocity at mobilization and degradation of shaft and end-bearing resis-
the pile head during hammer blows allows dynamic analyses tance during pile installation and loading.
of varying degrees of sophistication to be undertaken to esti- The experimental work presented in this paper was carried
mate the static capacity of the model piles, in addition to as- out on the fixed beam geotechnical centrifuge facility at the
sessing impact energy and hammer efficiency. The deduced University of Western Australia. The facility houses an Acu-
pile capacities may then be compared with results from direct tronic 661 centrifuge with a capacity of 40 g-tonnes and a
static loading of the pile. platform radius of 1.8 m. A detailed account of the geotech-
A potential problem in comparing dynamic and static load nical centrifuge and associated equipment may be found in
1
PhD Student, Sp. Res. Ctr. for Offshore Found. Sys., Univ. of Western
Randolph et al. (1991). The tests conducted here were per-
Australia, Nedlands 6907, Perth, Australia. formed in a strongbox of width 390 mm, length 650 mm, and
2
Prof. of Civ. Engrg., Sp. Res. Ctr. for Offshore Found. Sys., Univ. of height 325 mm, containing between 260 and 275 mm depth
Western Australia, Nedlands 6907, Perth, Australia. of soil.
Note. Discussion open until April 1, 2000. To extend the closing date Centrifuge modeling laws have been discussed in general
one month, a written request must be filed with the ASCE Manager of terms by Schofield (1980), and in the context of modeling pile
Journals. The manuscript for this paper was submitted for review and
possible publication on October 9, 1998. This paper is part of the Journal
driving by de Nicola (1996). Relevant scaling relationships are
of Geotechnical and Geoenvironmental Engineering, Vol. 125, No. 11, summarized in Table 1. It may be seen that all accelerations
November, 1999. 䉷ASCE, ISSN 1090-0241/99/0011-0988–0998/$8.00 are scaled by the modeling ratio N. The tests described here
⫹ $.50 per page. Paper No. 19420. were undertaken at a scale of 1:100, so that acceleration levels
988 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / NOVEMBER 1999
Model Piles
A detailed description of the model piles has been given by
Bruno (1998), so only the key features are summarized here.
Fig. 3 shows a schematic diagram of the model piles used
throughout the centrifuge testing program. At 100g the model
piles have a prototype diameter of 0.95 m, with a wall thick-
ness of 0.05 m and a maximum embedment depth of 20 m.
Downloaded from ascelibrary.org by Aalborg University on 01/23/24. Copyright ASCE. For personal use only; all rights reserved.
into Sample E. Installation effects due to different hammer the distinct transition in the incremental filling ratio (IFR),
drop heights and pile toe conditions were examined in this after a depth of around 5 m, where the IFR is given by (Brucy
sample. At an embedment depth of around 5.5 m, the drop et al. 1991)
height for two of the open-ended piles was increased to 1.7
m, whereas 1 m was maintained for the third open-ended pile. change in plug height
IFR = (2)
change in pile penetration
This is believed to be due primarily to the change in drop
height, from 1 to 1.7 m, at which the ram mass is delivered
during driving. Brucy et al. (1991) also observed that the IFR
was strongly affected by the input energy, with increased en-
ergy leading to a higher IFR.
It is interesting to observe that, over most of the installation
with a 1.7-m drop height, the IFR is steady at about 0.9. Per-
haps the main factor that limits the IFR to 0.9 is the relatively
large pile wall thickness (or small internal pile diameter) com-
pared with the mean soil grain size. This may be responsible
for inhibiting the initial ingression of soil into the pile.
FIG. 9. Typical Stress-Wave Data from Open-Ended Pile, Depth 7.15 m (Test J ) Hammer Delivered at Drop Height of 1.8 m: (a) Force
and Zv Data; (b) Integrated Energy and Displacement
FIG. 10. Typical Stress-Wave Data from Closed-Ended Pile, Depth 7.15 m (Test J ) Hammer Delivered at Drop Height of 1.7 m: (a) Force
and Zv data; (b) Integrated Energy and Displacement
displacements up to the maximum displacements, which com- scatter, due to varying hammer efficiencies, but a trend of in-
prised both temporary compression and permanent displace- creasing efficiency of the driving system as the hammer drop
ment. The permanent pile sets were taken from the driving height increases. This trend was also observed in a study of
records immediately preceding the static pile test sites, en- dynamic tests conducted on field scale piles by Hussein et al.
abling the magnitude of temporary compression to be esti- (1992), and is perhaps due to certain energy losses that are
mated. This procedure gave values of temporary compression largely independent of drop height.
that were similar to those deduced from (4). An increase in input hammer energy results in an increase
in pile displacements (s and c), and hence, leads to higher
DYNAMIC STRESS-WAVE ANALYSES mobilized soil resistance. This is reflected in Fig. 12, which
shows a clear trend of increasing Rs with increasing drop
Energy Balance Analysis
height.
The simplest approach to estimating the static capacity of The dynamic tests were grouped into four main categories,
the pile (or more correctly, the mobilized soil resistance) is depending on the drop height at which the dynamic tests were
from an energy balance between the input energy from the performed. The static pile capacity estimates from each cate-
hammer and the work done by the pile movement. There is a gory were then compared with the pile capacity measured dur-
variety of pile driving formulas in common use, many of ing the static tests (at a pile-head displacement of 0.1D), with
which have been reviewed by Whitaker (1975). Rather than the results summarized in Table 5. On average, all estimated
describe all of the different approaches here, the basic for- mobilized soil resistances underestimate the measured static
mulation will be outlined and some of the main variables dis- capacity, with the best estimates obtained from dynamic tests
cussed. conducted at 1.5-m drop heights or greater. At low hammer
The basis of the fundamental pile driving formula is rep- drop heights (and hence, low impact energy), the driving for-
resented by mula approach tends to be conservative by around 35% and
improves to within 15% of the static pile capacity at higher
/Wh impact energies. Fig. 13 shows how the individual dynamic
Rs = (3)
(s ⫹ c/2) tests, conducted at drop heights of 1.7 m or greater, compare
with the measured static pile capacities. Generally, consider-
where Rs = mobilized soil resistance; = hammer efficiency; able scatter is present in the calculated soil resistance Rs, with
W = weight of the hammer; h = drop height; s = permanent most of the estimates underpredicting the measured static ca-
set; and c = elastic movement of the pile. The average per- pacities Qs, particularly for values in excess of 12 MN.
manent set was taken from the driving records, whereas the
temporary set was estimated from (de Nicola 1996)
Fp L
c= (4)
(EA)p
where Fp = measured peak impact force; L = length of the
pile; and (EA)p = cross-sectional rigidity of the pile.
The efficiency of a hammer blow can be determined from
the following equation:
冕 Fv dt
= (5)
mgh
where the energy measured during the impact is given by in-
tegrating the product of the pile-head force F and velocity v
over time. The efficiency of the hammer is therefore the ratio
of the energy measured during the impact, to the maximum
possible energy delivered by the hammer of mass m at a drop
height h.
The miniature pile driving hammer performed relatively
well with most hammer blows delivered with average effi-
ciencies of around 60%, which is in a similar range to field FIG. 11. Delivered Hammer Energy as Function of Hammer
scale steam or hydraulic hammers and it is slightly higher than Drop Height
sitive to the value adopted for the damping parameter jc. How-
ever, the above expression can provide useful guidance on the
static pile capacity, where it is possible to calibrate the param-
eter jc for a particular site, either by performing static tests or
by full numerical dynamic analyses of the stress wave data as
described later. Where no static load tests are carried out,
guidelines for jc, as given in Table 6, may be adopted (Rausche
FIG. 12. Estimated Static Pile Capacity Rs for Driving Formula
et al. 1985).
Approach
TABLE 5. Average Ratios and Standard Deviations of Cal- Modified Case Analysis
culated to Measured Static Pile Capacity
Eq. (7) must be modified to take into account the large
Hammer drop heights Driving change in impedance that is present in the model piles tested
used in dynamic Formula Case* IMPACT
in the centrifuge (Fig. 14). Consider a downward traveling
pile tests
wave of velocity vd arriving at a point in the pile where the
(m)
impedance changes from Z1 to Z2. The incidence wave will
(1) (2) (3) (4) (5) (6) (7)
give rise to a reflected wave with a velocity vu in Region 1,
<1.0 0.66 0.18 0.84 0.19 — — and a transmitted wave with a velocity vd in Region 2. Assum-
1.2 0.71 0.18 0.87 0.23 — —
1.5 0.86 0.20 0.96 0.18 — —
ing that there is no incidence upward traveling wave from
>1.7 0.84 0.18 0.97 0.18 0.98 0.05 Region 2 (which may be treated in an analogous way), the
particle velocity and force at the boundary of the two regions
just after arrival of the downward wave are given by (Ran-
dolph 1990):
FIG. 13. Driving Formula Approach, All Test Data for Drop
Heights of >1.7 m
Case Analysis
The governing equation that describes the motion of the
stress wave through an elastic rod in the axial direction is
given by
⭸2w 2 ⭸ w
2
2 = c (6)
⭸t ⭸z2 FIG. 14. Schematic of Change in Pile Impedance
v= 冉 2Z1
Z1 ⫹ Z2 冊 vincidence (10)
Fd2 = 冉 2Z2
Z1 ⫹ Z2 冊 Fd1 (11)
Fu1 = 冉 2Z1
Z1 ⫹ Z2
冊 Fu2 (12)
Downloaded from ascelibrary.org by Aalborg University on 01/23/24. Copyright ASCE. For personal use only; all rights reserved.
Eq. (11) and (12) can now be substituted into the standard
Case (7) to yield a modified formulation of the estimated static
pile capacity Rs, which will be referred to as the Case* method,
given by
CONCLUSIONS
Dynamic stress-wave data measured from centrifuge pile
tests have been used as input in three common methods of
dynamic analysis. Estimations of the static pile capacity (or
mobilized soil resistance) were derived from the dynamic anal-
yses and compared with the measured pile capacities from
static tests. It was evident that the dynamic tests performed
with the highest input energy provided the most accurate es-
timations of the static pile capacity because greater pile-head
displacements (and hence soil mobilization) were achieved.
The simplest approach was based on a fundamental energy
balance and it provided reasonably good approximations, typ-
ically underpredicting the static pile capacity by 15%. Pile
capacity estimates from the energy balance method were
highly dependent on input energy and provided less accurate
FIG. 18. IMPACT Analysis, All Test Data estimations as the measured static pile capacity increased.
A modified Case method, which incorporated large changes
analysis provides an excellent estimation to the static capacity in pile impedance, was also applied to the dynamic data. A
with the ratio of Rs /Qs = 0.98. With a standard deviation of value of 0.25 for the damping parameter jc provided the best
0.05, there is considerably less scatter in the estimates com- approximation to the measured static capacity and also com-
pared to both the Driving Formula and Case* analyses, which pared well with the suggested value from Rausche et al.
both had standard deviations of 0.18. As discussed earlier, a jc (1985). This value of jc was determined from the dynamic
value in the range of 0.15–0.25 provided the best fit, with the tests, conducted at the highest delivered energy, and calibrated
capacities deduced either from the IMPACT analyses or the with the static pile test data.
static load tests. Of the three dynamic analyses, the numerical analysis using
IMPACT provided, consistently, the best estimates for static
FAILURE MODES DURING STATIC AND pile capacity. This was expected because considerable effort is
DYNAMIC LOADING needed to determine a suitable profile of soil resistance along
It is interesting to consider the different failure mechanisms the pile shaft and base to match the measured and computed
that occur in pipe piles during dynamic and static loading. As stress waves. The deduced total soil resistance therefore tends
mentioned earlier, it is generally accepted that a pile remains to give an accurate prediction of the static pile capacity.
substantially fully plugged in the static loading state, but be- The resulting good agreement between pile capacity de-
comes unplugged under the dynamic impact loading condition. duced from dynamic and static load testing must still be
For a fully plugged pipe pile, sufficient inside shaft resistance viewed in the light of significant differences between the two
can be mobilized between the pile and the soil plug, and hence, types of test, particularly for open-ended piles. In the dynamic
an equivalent end-bearing capacity over the full cross-sectional test, the pile movement is much less than the static displace-
area is considered. For an unplugged pile, the integrated inside ment at which the capacity is quoted. Additionally, failure for
shaft resistance must be less than the sum of the end-bearing open-ended piles occurs in a (largely) unplugged mode, com-
resistance of the plug and the inertial resistance of the plug pared with a fully plugged mode during static testing. The
(Stevens 1988). good agreement therefore implies that, during continuous driv-
JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING / NOVEMBER 1999 / 997