Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

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

net/publication/290599390

Bearing capacity of driven piles in sand, the NGI approach

Article · August 2005


DOI: 10.1201/NOE0415390637.ch75

CITATIONS READS

78 2,409

3 authors, including:

Kjell Karlsrud
Norwegian Geotechnical Institute
31 PUBLICATIONS   534 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Kjell Karlsrud on 17 March 2017.

The user has requested enhancement of the downloaded file.


Bearing Capacity of Driven Piles in Sand, the NGI Approach
C.J.F. Clausen
P.M. Aas
K. Karlsrud
Norwegian Geotechnical Institute, Oslo, Norway

ABSTRACT: Based upon comparisons between calculated and measured axial capacity of driven piles in
sand, it is concluded that the present API RP2A recommendations should be revised to better reflect the meas-
ured capacities. The authors propose a new empirical calculation method called NGI-99. This method was
calibrated against well documented pile tests with results from CPTs in a database established by NGI. Based
upon this method a best-fit conversion between SPT and CPT was established. Results of comparisons be-
tween calculated and measured pile capacities for the NGI-99 and two other methods are presented.The pro-
posed new method gives a good agreement between calculated and measured capacities for the most relevant
tests in NGI's database.

1 INTRODUCTION pile tests were included in the database, Fugro


(2004) and Tvedt & Fredriksen (2003).
The calculation of axial bearing capacity for off- Figure 1 indicates the range of all the pile tests in
shore piles has traditionally been carried out as rec- NGI's database in terms of tip depth and the average
ommended by the American Petroleum Institute, relative density Dr along the pile shaft.
API RP2A, API (1993). After high quality test re-
sults from driven piles in dense sand became avail- Sand average relative density Dr (-)
able to the participants in the Euripides project, 0
0 0.2 0.4 0.6 0.8 1 1.2

Zuidberg & Vergobbi (1996), several authors have


pointed out the important difference between the ca-
Pile tip depth below ground surface (m)

pacities calculated by the API RP2A method and


those actually measured. Over the last 10 years, the 25
Norwegian Geotechnical Institute (NGI) has built a
database with well-documented results from tests on
driven piles in sand, and carried out comparative
calculations with different methods. 50

In the following, results of such comparisons are


presented, together with a new calculation method,
called NGI-99, that leads to an improved agreement
75
between predicted and observed capacities. An ac-
Steel, open
companying paper by Karlsrud et al (2005) deals Steel, closed
with the capacity of driven piles in clay. Concrete

100

2 DATABASE Figure 1. Range of the 85 pile tests in NGI's database for piles
in sand
The NGI database for pile tests in sand includes 85
tests from 35 different locations, NGI (2001). These 3 SOIL PARAMETERS
tests are all in the public domain. Results from CPTs
(see below) are available for 56 of these tests. The To calculate the capacity of a driven pile in sand, the
database includes the detailed soil layering and key following soil parameters are needed for each of the
soil parameters at each of the pile test locations. Re- sand layers (the term "sand" is used herein for any
cently the Euripides and the Drammen E18 full scale non-cohesive and non-carbonate soil type):
- Unit weight testing is not known for the majority of the pile tests
- Porewater pressures in the database.
- Relative density
Estimates of the relative density must be based
upon results from some type of in situ test, for ex- 5 THE NGI-99 CALCULATION METHOD
ample SPT (standard penetration test), CPT (cone
penetrometer test), PMT (pressuremeter test) or ob- Several authors have previously pointed out that the
served pile driving resistance. For most offshore API RP2A method does not correctly predict the ca-
projects, CPTs are routinely carried out as a part of a pacity of driven piles in sand, e.g. Toolan et al
normal site investigation program. (1990), Randolph et al (1994) and Jardine & Chow
For some of the older pile tests in NGI's database (1996). Studies by Clausen & Aas (2001) confirmed
the site investigations only include SPT results. these earlier findings. NGI therefore opted to de-
Since the calculation method proposed in the follow- velop a simple CPT-based method for the calcula-
ing is linked to CPT results, one needs a conversion tion of axial capacity of piles in sand. This method is
between SPTs and CPTs. The approach used to ob- referred to as NGI-99 in the following.
tain this conversion was to: For the majority of the pile tests loaded in com-
• Calibrate the NGI-99 method against the pression, only the total capacity, and not the separate
high quality pile tests with CPT results. contributions from skin friction and tip resistance,
• Find the SPT to CPT conversion that gives are known. To develop the NGI-99 method, the piles
the best agreement between the observed ca- loaded in tension were therefore considered first :
pacity, and the capacity calculated by the • Find an expression for the average skin fric-
NGI-99 method for the pile tests with SPTs tion that acts upon piles loaded in tension.
only. • Assume that the skin friction in compression
This resulted in the following simple expression : is a constant times the skin friction in ten-
sion.
NSPTcorr = NSPTmeas·0.77·log10(1915kPa/σ'vo) (1) • Assume that the skin friction in a homogene-
qc = 2.8 · NSPT corr · σ atm (2) ous deposit has a triangular distribution with
depth.
where NSPTcorr is the corrected SPT N-value, Peck et • Determine the tip resistance in compression
al (1974), NSPTmeas the measured SPT N-value, σ'vo as some function of the cone tip resistance.
the vertical effective stress, σ atm the atmospheric Based upon calculations with the NGI sand data-
reference pressure (100 kPa), and qc the calculated base, and guided by the detailed results from instru-
cone tip resistance. mented pile tests, the following expressions were
Section 6 includes a comparison between the reached.
scatter in the ratio of calculated to measured capac- The local unit skin friction on a driven pile in
ity for SPT- and CPT-based pile tests. sand, τ skin, is given by:
The NGI-99 calculation method (Section 5) uses τ skin (z) = z/z tip·σ atm·FDr·Fsig·Ftip·Fload·Fmat (4)
the following relationship between CPT tip resis- τ skin (z) > 0.1 · σ'vo
tance qc and the sand relative density Dr:
z = Depth below the ground surface
Dr = 0.4 · ln{qc / [22 · (σ'vo · σ atm) 0.5]} (3) z tip = Pile tip depth
σ atm = Atmospheric reference pressure = 100 kPa
This expression is a compromise between the two
diagrams shown on Figure 5.47 of Lunne et al FDr = 2.1 · (Dr - 0.1) 1.7 (5)
(1997). It should be noted that for very dense sands F sig = (σ'vo / σ atm) 0.25 (6)
at shallow depth, equation (3) may result in Dr > 1.0. F tip = 1.0 for a pile driven open-ended,
Such a result is not considered as unrealistic, and 1.6 for a close-ended pile
should be used in the calculations. F load = 1.0 for tension, 1.3 for compression
F mat = 1.0 for steel, 1.2 for concrete.
4 EFFECT OF TIME BETWEEN PILE DRIVING The relative density Dr to be used for FDr is the
AND TESTING value calculated from equation (3) at the depth z. It
should be noted that the ratio z/z tip leads to a "fric-
Chow et al (1998) and Fugro (2004) present results tion fatigue" effect, i.e. as the pile is driven deeper
that indicate an important effect of time upon the the local unit skin friction at depth z goes down.
pile skin friction. However, the authors decided not The tip resistance acting against a pile driven
to include a time correction on the measured pile ca- close-ended is given by:
pacities, since the time between driving and load
σ tip = 0.8 · qc / (1+Dr2) (7)
- API RP2A from 1993, API-93
The tip resistance acting against a pile driven open- - Jardine & Chow (1996), MTD-96
ended is taken as the smallest of the coring and the - NGI's method as presented above, NGI-99
plugged tip resistance. The coring tip resistance is - Fugro (2004), Fugro-04
calculated assuming a stress against the pile wall of For each of these methods the average value of the
qc, and an internal pile/plug unit skin friction taken ratio between calculated and measured capacity
as 3 times the external skin friction. This higher in- (called C/M below) was found, together with the co-
side friction is caused by arching near the pile tip. efficient of variation (CoV), i.e. the standard devia-
The plugged tip resistance of an open-ended pile is tion divided by C/M. The comparative calculations
calculated as: were carried out for tests on tubular steel piles, with
the highest quality rating and with CPT data, 28 tests
σ tip = 0.7 · qc / (1+3·Dr2) (8) met these criteria.

From the above it follows that the skin friction is Table 1. Comparison between calculated and measured
mainly governed by the relative density, while the capacities for tests on steel piles at sites with CPT data and the
highest quality rating.
vertical effective stress has only a modest influence.
This is supported by the observed skin friction on Calculated/measured pile capacity
Method
piles subjected to tension loading, Figure 2, where Average C/M CoV
Piles loaded in tension, 8 tests
the average vertical effective stress along the pile is API-93 0.57 0.33
given for each data point. MTD-96 0.96 0.15
NGI-99 0.95 0.20
250
Fugro-04 0.79 0.43
Tension Piles Piles loaded in compression, 20 tests
Measured average skin friction in tension (kPa)

Steel, open API-93 0.67 0.39


Steel, closed 274
200 MTD-96 0.85 0.27
Concrete 316 NGI-99 0.95 0.23
Numbers give Fugro-04 0.94 0.39
average σ'vo in kPa
314
150
68 As one would expect, the NGI-99 method has lit-
tle bias and a relatively small CoV in this compari-
69
son, since that method was calibrated against many
100 of the same pile tests as those used for the Table 1
API-1993 55 comparison. On Figure 3 the C/M ratios calculated
Limiting
values
95 by the NGI-99 method for the 28 pile tests in Table
50
1 are plotted against pile tip depth
85
44 Calculated NGI-99 pile capacity / Measured capacity
91
129 52 0 0.5 1 1.5 2
0 0
0 0.2 0.4 0.6 0.8 1 1.2 40

Sand average relative density 36 3693 58


93
99 9999 99
63 42
Tip depth below ground surface (m)

Figure 2. Measured average skin friction on piles in sand 10 65 24


46

subjected to tension loading 27 66 66


40

The limiting skin friction values recommended by 20


the API RP2A for non-silty sands are included on
Figure 2 for comparison. For long piles in loose sand 30 Steel piles with CPTs
Numbers give D r (%)
the API-93 limiting skin frictions values are too op- 79
30 Open/Compression
timistic, whereas they can be too pessimistic in very Open/Tension
dense sands. 41 Closed/Compression
94 94 Closed/Tension

40 C/M average = 0.95


CoV = 0.22
6 COMPARISON BETWEEN CALCULATED
AND MEASURED CAPACITY 100 100
97 97

50
The following presents a comparison between calcu-
lated and measured capacities for the pile tests in
NGI's sand database. Four calculation methods were Figure 3. Comparison between calculated and measured pile
included: capacity for the pile tests in Table 1 and the NGI-99 method
The high CoV values found for Fugro-04 in Table Local unit skin friction (kPa)
1 are puzzling since this method has also been cali- 0 200 400 600 800 1000 1200
brated against essentially the same tests, in particular 20

the Euripides tests. Calculated values


Table 2 compares calculated and measured ca- NGI-99
25
pacities for these high quality tests and the three MTD-96

Depth below ground surface (m)


CPT-based calculation methods. Fugro 2004

30
Table 2. Measured and calculated capacities for six of the Eu-
ripides piles Measured
location 2
35
Pile Measured Calculated/measured capacity
test (MN) MTD-96 NGI-99 Fugro-04
Tension loading
40
L1 - 38.7 m 8.4 0.85 0.85 0.83
Measured
L1 - 47.2 m 12.5 0.85 0.87 0.73 location 1
L2 - 46.9 m 9.7 1.13 1.15 0.97
45
Average C/M 0.94 0.96 0.84
Compression loading
L1 - 38.7 m 12.5 1.08 1.04 1.19
50
L1 - 47.0 m 19.1 1.00 0.96 0.94
L2 - 46.7 m 18.8 1.06 1.00 0.99
Average C/M 1.05 1.00 1.04 Figure 5. Comparison between measured and calculated local
skin friction for the Euripides tension piles with 47 m tip depth
For these tests the main difference between the
It is observed that the Fugro-04 method matches
methods is that Fugro-04 gives somewhat lower ten-
the low skin friction measured between 22 m and 37
sile capacity. The CPT profiles used by the authors
m quite well. However, in order to obtain the meas-
for this comparison are shown on Figure 4.
ured total capacity as well, the Fugro method needs
CPT tip resistance, qc (MPa) to introduce very high unit skin friction values near
the pile tip, higher than 1000 kPa.
0 25 50 75 100
20 This means that the properties of the soil layers
Location 1 Location 2 close to the pile tip will have a dominating influence
upon the calculated capacity. The authors believe
25 that this could explain the high CoV values found
for the Fugro-04 method in Table 1.
Depth below ground surface (m)

The Euripides tests were carried out in very dense


30
sands. Four pile load tests with open-ended tubular
steel piles of OD = 813 mm in loose to medium
35
dense sand were recently carried out for the E18 mo-
torway bridge in Drammen, Norway, Tvedt &
Fredriksen (2003). These tests are not part of the
40 data used for calibration of the MTD-96 and Fugro-
04 methods. It is therefore of interest to check how
the different CPT-based methods predict the Dram-
45 men tests. Results of such comparisons are shown in
Table 3.
50
Table 3. Measured and calculated capacities
for the Drammen test piles, loaded in compression
Figure 4. Idealisation of the CPT profiles used in NGI's calcu-
lations for the two Euripides test locations
Pile position Measured Calculated/measured capacity
and tip depth (MN) MTD-96 NGI-99 Fugro-04
In an attempt to explain the high CoV values Pier 25 - 15 m 1.89 0.44 0.79 1.29
found for the Fugro-04 method, measured and calcu- Pier 25 - 25 m 3.45 0.43 0.81 0.77
lated unit skin friction for the two Euripides tensile Pier 16 - 11 m 1.30 0.61 1.54 1.78
tests at 47 m tip depth were plotted against depth on Pier 16 - 17m * 1.30 0.84 0.67 0.53
Figure 5. Average C/M 0.58 0.95 1.09
* This pile has its tip in clay
MTD-96 gives too low capacity for these four report was sponsored by the API Subcommittee on
piles, probably as a result of a too low calculated tip Offshore Structures.
resistance. The two other methods give a good aver-
age C/M, but there is a wide scatter between the in- 9 REFERENCES
dividual results. Part of the difference between the
calculation methods for these tests is due to the as- American Petroleum Institute 1993. "Recommended Practice
sumed tip behaviour, plugged or coring. For all four for Planning, Designing and Constructing Fixed Offshore
Platforms - Working Stress Design". API RP 2A-WSD,
tests it was reported that the soil plug followed the 20th Edition, Washington, 1 July 1993.
pile as the load was applied. Chow F.C., R.J. Jardine, F. Brucy & J.F. Nauroy 1998.
The pile tests in the database with SPT data (29 "Effects of Time on Capacity of Pipe Piles in Dense Ma-
tests) were analysed by the NGI-99 method only and rine Sand." ASCE, JGGE, Vol. 124, No. 3, March 1998.
the proposed SPT to CPT conversion, equations (1- Clausen C.J.F. & P.M. Aas 2001."Capacity of Driven Piles in
2). This resulted in a C/M of 0.97 and a CoV of Clays and Sands on the Basis of Pile Load Tests." Pro-
ceedings of the 11th (2001) International Offshore and Po-
0.32. The same numbers from Table 1 for the piles lar Engineering Conference, ISOPE, Volume II, Stavanger
with CPT data are 0.95 and 0.22. This difference in June 2001. p. 581-586,
CoV most likely reflects the improved quality of the Fugro Engineers B.V. 2004. "Axial Pile Capacity Design
soil data obtained by the CPT. Method for Offshore Driven Piles in Sand". Report P1003
presented to the American Petroleum Institute, issue 3, 5
August 2004.
Jardine R.J. & F.C. Chow 1996. "New Design Methods for
7 CONCLUSIONS Offshore Piles." Marine Technology Directorate Ltd., Pub-
lication MTD 96/103, London 1996. ISBN 1 870553 31 4.
NGI established a comprehensive database with re- Karlsrud K., C.J.F. Clausen & P.M. Aas (2005)
sults from load tests on driven piles in sand. A new "Bearing Capacity of Driven Piles in Clay, the NGI
empirical calculation method, referred to as NGI-99, Approach". Proc., International Symposium on Frontiers in
was developed and calibrated against the results in Offshore Geotechnics, Perth Sept. 2005, A.A. Balkema
Publishers.
this database. Trial calculations with different meth- Lunne T., P.K. Robertson & J.J.M. Powell 1997. Cone Pene-
ods and the NGI database lead to the following con- tration Testing in Geotechnical Practice. Blackie Aca-
clusions: demic & Professional, ISBN 0 751 40393 8, London 1997.
Norwegian Geotechnical Institute 2001. "Bearing Capacity of
1 The API-93 method should be replaced by other Driven Piles, Piles in Sand." Internal report no. 525211-2,
21 January 2001.
methods that better reflect the results from recent Peck R.B., W.E. Hansen & T.H. Thornburn 1974. Foundation
high quality pile tests, e.g. the Euripides and the Engineering. John Wiley & Sons, New York 1974.
Drammen tests. Randolph M.F., J. Dolwin & R. Beck 1994. "Design of Driven
2 The MTD-96 method seems to work well for Piles in Sand." Geotechnique, Vol. 44, No. 3,September
piles in dense sand. Full scale tests on piles in 1994, p.427.
loose to medium sand in Drammen, Norway, do Toolan F.E., M.L. Lings & U.A. Mirza 1990. "An Appraisal of
API RP2A Recommendations for Determining Skin Fric-
however show capacities significantly higher than tion of Piles in Sand." OTC Paper 6422,Houston, May
those predicted by MTD-96. 1990.
3 The new method proposed by Fugro (2004) has Tvedt G. & F. Fredriksen 2003. "E18 Ny motorvegbru i
been calibrated against the Euripides tests and Drammen. Prøvebelastning av peler." Proceedings from
therefore "predicts" these tests quite well. How- the Conference on Rock Blasting and Geotechnics, Oslo,
November 2003.
ever, when this method is used for other well
Zuidberg H.M. & P. Vergobbi 1996. "EURIPIDES, Load Tests
documented tests in NGI's database, the scatter in on Large Driven Piles in Dense Silica Sands." OTC paper
the calculated results becomes higher than for the 7977, Houston, May 1996.
MTD-96 and the NGI-99 methods. This effect
could be caused by the very high local unit skin
friction calculated near the pile tip by the new
Fugro method.
4 The NGI-99 method presented herein results in a
good agreement between calculated and measured
capacities for the most relevant pile tests in NGI's
database.

8 ACKNOWLEDGEMENTS

The work presented herein was supported by Norsk


Hydro, Statoil and NGI. The authors gratefully ac-
knowledge their generous support. The Fugro (2004)

View publication stats

You might also like