Sunteți pe pagina 1din 10

EXPERIMENTAL AND NUMERICAL ASSESSMENT OF OSTERBERG LOAD TESTS

ON LARGE BORED PILES IN SAND



Roberto Nova, Politecnico di Milano, ITALY
Bruno Becci, Ce.A.S. S.r.l., Milano, ITALY



The selection of the Osterberg Cell (O-Cell) technology as the standard testing
method for deep foundations of the Railways Po Viaduct in Northern Italy offered
a relevant number of field measures on large bored pile behaviour in alluvial
soils. The assessment of pile response during such non standard testing
procedure was performed through the comparison with standard top load tests
on companion piles as well as by numerical models of the tests including a
simplified pile-soil interaction scheme. As shown in this paper, the O-Cell
technology allowed a careful assessment of the non-linear pile behaviour even at
quite small loading levels as those required by a posteriori proof tests on
production piles. Simple numerical models proved to be very effective in the
simulation and the interpretation of pile behaviour during such unconventional
testing procedure.

1. INTRODUCTION

In the construction of the High Speed Railway
Viaduct crossing the Po river near Piacenza
(Italy), large diameter bored piles were routinely
adopted as foundation system.

High design loads and complex site conditions
suggested the use of the Osterberg Cell (O-Cell)
technology (Osterberg (1989)) as the first choice
testing method for these piles. At onshore
viaduct piers, usual kentledge load tests were
also performed, thus permitting worthwhile
comparison between different testing methods.

All the performed tests were also reproduced
through simple numerical models that offered a
useful and thorough assessment of the
measured pile behaviour during the O-Cell tests
and a confirmation of the post-processing
procedures employed in their interpretation.

In the following, a review of the pile design
criteria is reported; then a discussion of both
design and proof load tests is included;
numerical analyses of all the performed tests are
presented and relevant parameters that best fit
experimental observations are outlined. In the
light of all these observations, some general
remarks on the design of large diameter piles in
sand are proposed.



2. SITE AND VIADUCT DESCRIPTION

The Po Viaduct (Evangelista et al. (2003))
includes 23 bays, of which 20 approaching the
river at both sides. A cable-stayed bridge, whose
central bay is 192 m long, is placed at the
permanent riverbed crossing (fig. 1).


Figure 1: the Po Viaduct near Piacenza (Italy)

Two meter diameter bored piles were used at
the base of all the 24 piers, with pile lengths
ranging from 40 to 62.5 m to ensure allowable
loads between 10 to 18 MN. Bentonite slurry
was employed in borings. Pile number per pier
varies from 4 to 28.

Nova. R, Becci B., Experimental and Numerical Assessment of Osterberg Load Tests on Large Bored Piles in
Sand, 33
rd
Annual 11
th
International Deep Foundations Institute Conference Proceedings, New York, NY,
Oct 15-17 2008, pp 225-233.
Subsoil conditions at the site are represented by
very deep submerged alluvial deposits: currently
a shallow 15 to 20 m thick sand layer is
overlying a 7 to 15 m thick silty clay stratum,
which in turn is resting over a very deep silty
sand deposit.

Shallow and deep granular layers were mainly
investigated by means of SPT and CPT tests
that revealed almost normally consolidated
sands; limited to offshore piers, very deep layers
were sampled through a special CPT test, using
self-penetrating equipment into advancing hole.

In the granular deposits, SPT blow count N
could be approximated by the following
analytical expression:

N=(0.78 to 1.0) z (1)

in which z is the depth in [m]; typical relative
densities between 50% and 60% were estimated
and an almost constant peak friction angle equal
to 31 was evaluated, including dilatancy
reduction due to high pressure, according to
Bolton (1986).

The intermediate, slightly overconsolidated, silty
clay layer was analyzed through both in-situ
CPT and lab tests, showing characteristic
undrained shear strength values ranging from 50
to 100 kPa at depths between 15 and 30 m.

This intermediate cohesive layer plays an
important role in the hydraulic design of the
offshore piers, as it limits the expected scour
depth during extreme design flood conditions.

3. PILE DESIGN CRITERIA AND PRELIM-
INARY EXPERIMENTAL FINDINGS

Preliminary design criteria are discussed in
Becci et al. (2007). In Table 1 (left column) a
summary of initial design assumptions is
included.
Unit shaft resistance in sands was assumed
linearly varying with depth through a constant
coefficient K set to 0.6 and a friction parameter
tan(): as for , the critical friction angle of soil,
set to 30, was selected. Such unit resistance
was assumed to develop at a differential pile-soil
local movement of 0.5% of pile diameter D. As
for toe resistance, the design value in Table 1
was considered to develop corresponding with a
toe displacement of 5% D.

Before actual production pile construction
stages, two design load tests were performed,
according to the ASTM D1143 Quick Load Test
Method, on sacrificial piles (50 and 55 m long),
near actual Viaduct pier on the left hand
riverside, i.e. at the north side of Po river. By
means of a pair of Osterberg cells, installed in
each shaft, as shown in fig. 2, ultimate loads
could be almost reached.
-21 m
O-cells
SAND
SILTY CLAY
SILTY SAND
-57 m
-52 m
-36 m
-55 m
-45 m
-50 m
-42 m
-2 m
0 m
D=2 m
PILE A PILE B

Figure 2: preliminary pile load test assembly
In both piles, the lower O-Cell was placed 2 m
above the toe, thus allowing a close depiction of
toe behaviour. Moreover, through the installation
of ten strain gauge levels along the shaft, shear
distribution at relevant test stages was obtained
Table 1
Preliminary Design assumptions Preliminary load tests results
Shaft
resistance
Sands q
s
= Ktan()
v
= 0.346
v
(2)

Clay q
s
0 (3)
Sands q
s
= (0.62 to 0.85)
v
(5)

Clay q
s
(0.20 to 0.25)
v
(6)
Toe
resistance
Sands q
b
N
q

v
10
v
(4) Sands q
b
(7 to 9)
v
(7)

v
=effective overburden stress = z = buoyancy unit weight= 10 kN/m
3

by the measurement of the pile axial strain
distribution.

Based on these measures (Becci et al. (2007)),
shaft and toe resistances listed in right column
of Table 1 were estimated.

It should be noted that a slightly lower toe
resistance at 5%D displacement was
measured, whilst a higher shaft strength could
be observed, the latter, however obtained
through unusually large shaft displacements that
could be imposed thanks to the particular testing
method offered by the O-Cell technology.
Increasing shaft resistance with depth was
confirmed.

The inspection of the equivalent top load curves
obtained by an a posteriori processing of
measures (solid gray curves in fig. 3) shows that
preliminary design assumptions (dashed lines)
had predicted a lower ultimate loading due to an
underestimate of shaft resistance.

0
10
20
30
40
50
60
70
0 50 100 150 200 250
PILE HEAD SETTLEMENT- [mm]
T
O
P

L
O
A
D

[
M
N
]








-
-
-
-
-
PRELIMINARY DESIGN CURVES
NUMERICAL BACKANALYSIS
O-CELL EQUIVALENT CURVES
(POSTPOCESSING FROM TEST
RESULTS)
PILE A
PILE B
PILE A
PILE B

Figure 3: Top Load vs Top Displacement
curve for preliminary test piles
On the other hand, it was noted that the
presumed pile stiffness at working load level had
been well matched (see fig. 3, again); moreover,
actual pile construction would have implied more
difficult working conditions than those occurred
during test pile installation; finally, long term
extreme scour conditions for offshore piers
would have represented an important issue that
could have been hardly investigated by similar
preliminary tests.

For all these reasons, in spite of the apparently
conservative preliminary design criteria, no pile
length was reduced in the final design. This
choice demonstrated to be wise, in the light of
actual construction observations.

4. NUMERICAL SIMULATION OF PREL-
IMINARY LOAD TESTS

In the practice, the simulation of the soil-pile
interaction by means of non-linear Winkler
springs (ONeill & Reese, (1999)) is still probably
the most popular method to analyse single pile
behaviour in inhomogeneous soil conditions.

Adopting this approach within a finite element
framework, various pile and loading conditions
can be easily modelled, including the simulation
of an Osterberg test as well.

Using this method, the dark solid curves in fig. 3
have been computed, including strength and
stiffness parameters summarized in fig. 4.

The toe reaction pattern has been included by
assigning the behaviour as directly measured by
O-Cells tests.

At 5%D toe displacement, the ratio N
q
=q
b
/
v

was found to be about 8.6 for Pile A and 6.7 for
Pile B.
S
I
L
T
Y

C
L
A
Y
S
A
N
D
10
20
30
40
50
60
10
20
30
40
50
60
200 200 400 400 q [kPa]
s
q [kPa]
s
-2 m
-52 m
-57 m
-21 m
-36 m
0 m
-2 m
403
300
60
40
130
170
60
40
180
400
=3
=1
=4 =2.5
=1.5
=1
0.4
0.8
1.2
1.6
0.
SHAFT s / (D)
0.5% 1% 1.5% 2%
q











s
q











/











CLAY
SANDS
5% 10% 15% 20%
2
0
4
6
8
B
a
s
e

r
e
a
c
t
i
o
n

q


[
M
P
a
]









b
PILE B
PILE A
BASE s / D
b
PILE A PILE B
measured
measured
S
I
L
T
Y

S
A
N
D
assumed in
numerical
analysis
assumed in
numerical
analysis
Figure 4: numerical back-analysis on preliminary
test piles
As for shaft reaction, trial shear strength profiles
(solid lines in fig. 4) along with a threshold side
displacement have been assigned. The latter is
included via a scale factor , which is actually
one of the free parameters in this back-analysis
process: note that, taking =1, the normalized
shaft displacement s/D as usually considered by
ONeill & Reese, (1999) is recovered.

Final set of data, giving a good reproduction of
target results, shows that is currently higher
than usual values; as for pile B, shaft reaction
was found to be weaker yet stiffer than in pile A,
in the deep sand layer. Improved numerical
simulations can be obtained provided a slightly
less smooth curve for shaft reaction in sand is
adopted: this modification is not relevant for
current discussion, however.

5. PROOF LOAD TESTS WITH OSTER-
BERG CELLS

In preliminary load tests, due to high expected
ultimate loadings, the O-Cell technology was
deemed almost mandatory; however, proof load
tests, up to 1.20 the maximum service pile load
N
max
, could have been performed by means of
traditional methods as well, unless quite
complex conditions would have to be dealt with
in offshore piers no. 7 and 8 (fig. 5).
Following some debate, three production piles in
offshore piers and two onshore piles (fig. 6)
were equipped with one O-Cell only, properly
designed to impose a sufficient equivalent load
for proof purpose. Near to the onshore O-Cell
piles, three traditional top load tests were also
requested.


Figure 5: site conditions for Piers 7 and 8
The O-cell was placed at about 20% of the pile
length above the toe, as deep as possible to
avoid significant lateral pile resistance reduction
and, in the same time, to allow desired
equivalent top load imposition without excessive
cell loadings or toe movements.
Figure 6: piling layout and maximum load levels for proof load tests
Dealing with one O-Cell only offered the
opportunity to perform a numerical analysis of
the test itself, with a straightforward modification
of the finite element model used to analyze the
behaviour of the pile subjected to vertical top
loads.

To model the single O-Cell test, the finite
element corresponding with cell position must be
annealed and two equal loads must be applied
in opposite directions to the nodes
corresponding with annealed element.

It should be noted that the simulation of a single
O-Cell test requires no complex constitutive law
for soil springs, since all such elements basically
undergo monotonic loading path only.

6. PROOF LOAD TESTS DISCUSSION

In the following, the results from each O-Cell test
are discussed. Also these tests have been
conducted according to the Quick Load Test
Method.

Limited to tests 1 to 3 (listed in fig. 6),
remarkable nonlinear aspects have been
highlighted during the tests; therefore some
considerations about ultimate loading can be
attempted as well.

In onshore tests 4 and 5, nonlinear behaviour is
less clear: on the other hand, a worthwhile
comparison with traditional load tests is
available, which can be considered a
contribution to experimental assessment of O-
Cell technology.

OFFSHORE PIERS

Fig. 7 outlines the results obtained for offshore
piles in piers 7 and 8, by plotting the O-Cell plate
absolute movements in upward (top plate) and
downward (bottom plate) direction.

In spite of applied (equivalent) loads slightly
higher than maximum expected working loads,
quite evident non-linear behaviour is obtained.

Whereas such a behavior would, in general, be
undesirable in a proof test, it is currently
acceptable in such a procedure: in particular, a
numerical simulation of each of these tests can
sufficiently explain that the observed behaviour
is physiological.


For back-analyses, the pile-soil interaction
described in the previous section is used, with
the normalized curves shown in fig. 8, same to
those included in fig. 4, for shaft reaction.


In fig. 8 again, q
s
profile and relevant parameters
obtained through the back-analyses are
summarized; obtained responses are included in
fig. 7, superimposed to experimental data.

The back-calculated critical toe pressure q
bLIM
=4
MPa, corresponding with a toe settlement equal
to 0.05 D is a good assumption for all the
analyzed cases, and well agrees with eq. 7.

-50
-40
-30
-20
-10
0
10
20
30
0 5 10 15 20
O-Cell Gross Load (MN)
O
-
C
e
l
l

D
i
s
p
l
a
c
e
m
e
n
t

[
m
m
]

TEST1 - PIER 7 PILE 9
TEST 3 - PIER 8 PILE 20
TEST 3 - numerical
DOWN
UP
-50
-40
-30
-20
-10
0
10
20
30
0 5 10 15 20
O
-
C
e
l
l

D
i
s
p
l
a
c
e
m
e
n
t

[
m
m
]

TEST 2 - PIER 7 PILE 21
TEST2 - numerical
DOWN
UP
Figure 7 measured displacements and back
analysis results for offshore tests
PILE HEAD +37.0 m
SILTY
CLAY
z [m]
10
20
30
40
50
60
0 50 100 150 200
q [kPa]
s
SAND
SILTY
SAND
O-CELL POSITION
T
E
S
T

1

&

3
T
E
S
T

2

1 1
1 1
0.75 1
1.5 1.5
TEST 1
TEST 2
q [MPa]
b LIM
4 4
0.4
0.8
1.2
1.6
0.
SHAFT s / (D)
0.5% 1% 1.5% 2%
q











s
q











/











CLAY
SANDS
0.4
0.8
1.2
1.6
0.
5% 10% 15% 20%
q











b

L
I
M
q











b
/











BASE s / D
b
TEST 3


In TEST 1 or 3, the shaft resistances q
s
and their
mobilization levels quite well adhere to original
design assumptions (as included in Table 1, left
column), except that a non-zero shaft resistance
is highlighted in the clay layer as well.

As for TEST 2, the shaft behaviour above the O-
Cell is almost the same as in the others; in the
lower part, some reduced shaft resistances had
to be included, to match the observed
behaviour. At this stage, it is however hard to
state whether this diminished response is due to
locally weaker shaft resistance or to higher local
compliance. To reliably depict the actual
behaviour, the load would have had to be
increased to a much higher level.

The scaling parameter decreases with q
s
: this
numerical effect is necessary to reproduce
similar side skin friction stiffness for all cases,
independently from maximum strength value.
Anyhow, the back-figured parameters in fig. 8
currently fall within typical ranges. It should be
noted that the shear strength increase with
depth is confirmed: however, the mobilization
levels tend to increase with depth as well, as
also noted in the analysis of preliminary load
tests discussed previously.

ONSHORE PIERS

In TEST 4 (fig. 9), both soil conditions and pile
length are very similar to those considered for
offshore piles.

The measured behaviour was consistently very
similar to TEST 1 or TEST3: in particular the
TEST3 back-analysis can reasonably reproduce
the behaviour of this pile too.

Further modifications may be included, to better
match higher stiffness of the upper part as well
as a slightly lower stiffness of the lower
segment: however such changes would not
modify the overall description for this pile.

Finally, experimental results and numerical
back-analysis of TEST 5 on the shortest pile in
this campaign, is reported in fig. 10, whilst back-
figured parameters are included in fig. 11.

Back-figured shaft reaction displays a better
response than in offshore piles: in sand layers,
q
s
from eq. (1) can be increased by a factor
1.33, thus obtaining an average ratio
=q
s
/
v
=0.46.

Figure 8: back analysis proof load test
assumptions and results for offshore piles
-30
-20
-10
0
10
20
30
0 5 10 15 20
O-Cell Gross Load (MN)
O
-
C
e
l
l

D
i
s
p
l
a
c
e
m
e
n
t

[
m
m
]

TEST 4 - PIER9 - PILE 4
TEST3 - numerical
DOWN
UP
Figure 9 measured displacements and
back analysis results for onshore TEST 4
At adjacent pier no. 13 location (about 50 m far
from pier 12, in S-E direction), corresponding
with same piling layout and very similar soil
properties, a traditional top load test was
performed, using a steel kentledge.

These results could be effectively compared with
both equivalent top load curve obtained through
the standard O-Cell post-processing procedure
and numerical prediction based on back-figured
parameters from TEST 5, in fig. 11.

The comparison included in fig. 12 shows an
excellent agreement among experimental data
from traditional test (dots), equivalent O-Cell
curve (solid gray line) and numerical prediction
(solid black line).

The top load test could not investigate non-linear
pile behaviour. Therefore, unless a top loading
could be revised as a more natural way to test
actual pile behaviour, the information offered by
the traditional procedure is poor indeed as
compared with an O-Cell test.
The remaining two top load tests were
conducted at piers 11 and 18, on 46 m long
piles. In both tests, an almost linear behaviour
could be obtained, with top settlements of about
3 mm for both piles, at a proof load of 12 MN.

7. DISCUSSION AND CONCLUSIONS

The selection of the Osterberg cell technology
allowed the conduction of load tests to very high
load levels that would have been hardly imposed
due to complex environmental conditions.

The O-Cell method in proof load tests could
provide engineers with more useful and precise
information than those currently available by
traditional methods. In particular, some non-
linear behaviour of part of the pile could be
activated even at quite low loads.

In addition to the routine post-processing of
sampled data as a part of the standard O-Cell
procedure, the authors performed simple
numerical simulations of the tests using the
Winkler method, and found that observed
FREE FIELD +45.5 m
PILE HEAD +41.0 m
SILTY
CLAY
10
20
30
40
50
0 50 100 150 200
q [kPa]
s
SAND
SILTY
SAND
O-CELL POSITION
T
E
S
T

5

1 0.5
1
1.5
TEST 5
q [MPa]
b LIM
4
z [m]
250
0.5
Figure 11: back analysis proof load test
assumptions and results for TEST5
-20
-10
0
10
20
0 5 10 15 20
O-Cell Gross Load (MN)
O
-
C
e
l
l

D
i
s
p
l
a
c
e
m
e
n
t

[
m
m
]

-
-
-
-
-
TEST 5 - PIER12 - PILE 1
TEST 5 - numerical
DOWN
UP
Figure 10 measured displacements and
back analysis results for onshore TEST 5
PROOF LOAD
11.91MN
0
2
4
6
8
10
12
14
16
0 3 6 9 12 15
PILE HEAD SETTLEMENT [mm]
T
O
P

L
O
A
D

[
M
N
]


-
PRELIMINARYDESIGN ASSUMPTIONS
TOP LOAD TEST- PIERNo. 13, PILE 1
TEST 5 - EQUIV. TOP LOAD CURVE BY
O-CELL PROCEDURE
NUMERICAL ANALYSIS (TEST5)

Figure 12 comparison between traditional
top load test, O-Cell test and numerical
prediction for onshore pile

behaviour during these unconventional tests
could be reasonably reproduced by means of
interaction curves similar to those that would
have been used in a traditional pile model.

It is the authors opinion that these conclusions
may contribute to increase the confidence by
practicing engineers in the selection of modern
testing techniques like the one discussed in this
paper.

It is important to realize that new testing
methods currently allow the conduction of pile
load tests, in almost all the real conditions.
Complex site conditions and/or very high loads
can, therefore, hardly be used as an excuse to
limit or even omit load tests at all.

On the other hand, a careful assessment of the
obtained results is always recommended. In
particular, heavy modifications to initial design,
based on reasonable and well-established
assumptions, should be considered with great
care.

As for the observed behaviour of these large
diameter shafts, drilled under bentonite in sands,
the authors found almost uniform toe behaviour,
in good agreement with most frequently used
correlations in the practice.

As for shaft resistance, however, relevant
discrepancies among design correlations,
preliminary load tests and final proof load tests
findings have been highlighted and discussed.

These findings should, in general, suggest a
particular care in the selection of shaft
resistance parameters for bored piles in sand, all
the more because similar observations have
been reported by others, regarding bored shafts
or barrettes in different soil conditions (e.g.
Randolph (2003), Fellenius et al. (1999)).

Such discrepancies are primarily related to
construction issues, which can be hardly
incorporated in preliminary design models (see
Cernak (1976), Fleming & Sliwinski (1977), Ng &
Lei (2003)).

Even in the light of these limited observations, it
can be argued that the suggested partial safety
factors used in the growing LRFD method also
in geotechnical engineering, may need some
further discussion before being used in the
practice.
In particular, reference is made to Eurocode 7
(CEN (2003)), that recommends a partial safety
factor
B
=1.60 for toe resistance, higher than the
shaft resistance factor
S
=1.30, for bored piles.

The aforementioned values were likely tuned to
implicitly account for different settlements
necessary to activate each of the two
contributions. However the authors argue
whether such values may or may not conflict
with some actual findings like those reported in
this paper, as well as other frequent field
observations in practical pile constructions.

ACKNOWLEDGEMENT

The technical advice of Grandi Lavori Fincosit
staff, leaded by dr. Augusto Ba and dr. Raji
Haykal, is particularly acknowledged. Piling
Contractors TREVI S.p.A. and VIPP S.p.A.,
General Contractor SNAMPROGETTI and the
Client Italferr S.p.A. are also acknowledged, as
well as Loadtest Inc. engineers who provided
and supported the Osterberg cell technology.

REFERENCES

BECCI, B., NOVA, R., BA, A., and HAYKAL,
R., 2007 Prove di carico su pali di grande
diametro mediante limpiego di celle Osterberg,
Rivista Italiana di Geotecnica (RIG), Anno 41,
no. 4, pp 9-28, in Italian

BOLTON, M. D., 1986 The strength and
dilatancy of sands, Gotechnique, vol. 36, No. 1,
pp 65-78.

CEN, 2004 - EN 1997-1:2004: Eurocode 7:
Geotechnical design Part 1: General rules,
Brussels.

CERNAK, B., 1976, The Time Effect Suspension
on the Behavior of Piers, Institute of Civil
Engineering, Bratislava, CSSR, Proc. 6
th

European Conference on Soil Mechanics and
Foundation Engineering, Vienna, Vol. 1, pp.
111-114.

EVANGELISTA L., PETRANGELI M. P., TRAINI
G., 2003 The cable-stayed bridge over the PO
river, IABSE Symposium on structures for high-
speed railway transportation, Antwerp, August,
pp 138-139.

FELLENIUS, B. H., ALTAEE, A., KULESZA, R,
and HAYES, J., 1999 O-Cell Testing and FE
analysis of a 28 m Deep Barrette in Manila,
Philippines, ASCE Journal of Geotechnical and
Geoenvironmental Engineering, Vol. 125, No. 7.,
pp 566-575.

FLEMING, W. K. and SLIWINSKI Z. J., 1977
The Use and influence of bentonite in bored pile
construction, DOE / CIRIA Piling Development
Group Report PG 3.

NG, C. W. W. and LEI, G. H., 2003
Performance of Long Rectangular Barrettes in
Granitic Saprolites, ASCE J. Geotech. and
Geoenvir. Engrg., Volume 129, No. 8, pp. 685-
696

ONEILL, M. W. and REESE, L. C., 1999
Drilled Shafts: Construction Procedures and
Design Methods, report no. FHWA-IF-99-05,
U.S. Department of Transportation, Federal
Highway Administration.

OSTERBERG, J. O., 1989 New Device for
Load Testing Driven Piles and Drilled Shafts
Separates Friction and End Bearing, Proc.
International Conference on Piling and Deep
Foundations, London, A.A. Balkema, pp 421
427.

RANDOLPH, M. F., 2003 Science and
empiricism in foundation pile design.
Gotechnique, vol. 53, No. 10, pp 847875.

S-ar putea să vă placă și