Fleming, W. G. K. (1993). G&Xechnique 43, No.

4, 615-619

DISCUSSION

A new method for single pile settlement prediction and analysis

W. G. K. FLEMING (1992). GBotechnique 42, No. 3, 411-425

T. J. Poskitt, Queen Mary College, University of apparent transition from shaft dominance to base
London dominance. This appears to conflict with the type
The state of stress in the soil surrounding a pile of behaviour indicated by Fig. 1, and so I suggest
is complex. This is due to the process of install- that abrupt changes could be due to brittle
ation and subsequent reconsolidation, and also to behaviour in the soil.
the stresses which existed in the soil prior to The Paper gives persuasive arguments, mainly
installation, which may not be known. As a of a practical kind, for collecting together all the
result, when the pile is loaded the settlement is characteristics of shaft behaviour into a single
difficult to predict, and the level of sophistication hyperbolic relationship. I believe a better method
which it is sensible to use in theoretical methods is to represent shaft behaviour in terms of hyper-
must be matched accordingly. The Author recog- bolic load transfer functions. This has been done
nizes this. The complex problem of single pile set- in connection with the related problem of finding
tlement is globally characterized by a few the form of dynamic load transfer functions which
parameters related to the hyperbolic law, and should be used in pile driving studies. The practi-
these are then found from the load-settlement cal difficulties of taking dynamic measurements
curve. The numerous cases which the Author has initially led me to study static load-settlement
successfully analysed give confidence in the curves. Several factors arising from these studies
method. have a direct counterpart in the present Paper.
Perhaps the greatest objection to the hyper- The first concerns the law used for the basic
bolic law is the assumption that, irrespective of load transfer function. In pile driving this is gen-
soil type or pile make-up a load-settlement curve erally taken as bilinear (Fig. 17). The Author
when plotted on hyperbolic axes (namely regards this as an inferior law, but it is necessary
settlement/load against settlement) should give a to remember that its use is necessitated by the
straight line. To define the hyperbola requires practical need to develop simple numerical pro-
two parameters, and it is difficult to see how these cedures for the unloading and reversed loading
relate to the engineering parameters of the pile ranges. The bilinear law, like the hyperbola,
and soil. requires two parameters to define it, one of which
In the Paper this is partially overcome by the is the ultimate resistance. However, unlike the
use of hyperbolic laws for both the shaft and the hyperbola, this is reached after a finite displace-
base. The four parameters are related to basic soil ment Q, which is known as the quake. For the
constants and the method as presented is a sig- shaft, Q can be associated with M,D,; for the
nificant step forward in the understanding of pile base its counterpart is 0.6 U,/E, D, (see Table 1).
behaviour. However, a consequence of using two
hyperbolic relations is that the original assump-
tion that load and settlement conform to a hyper-
bolic law is now violated. This is readily seen in
the case of a rigid pile, given by equation (14),
where the graph of AT/PT against Ar is no longer
linear.
To overcome this problem, the Author suggests
that the first part of the curve be associated with
shaft parameters, while the second is associated
with base parameters. The difficulty with this is
knowing where shaft influence ends and base
influence takes over. This can be seen in the case
study of pile H at Wembley. Using the data in
Fig. 6, and a standard non-linear structural pro-
gramme, the load-settlement curve for this pile
was obtained. This is shown on a hyperbolic plot
in Fig. 16. Over the range of loading considered SetUemeniAoad~ mmlkN

by the Author this gives a gentle curve with no Fig. 16

615
Downloaded by [ UNIVERSIDAD NACIONAL DE COLOMBIA] on [17/11/19]. Copyright © ICE Publishing, all rights reserved.
 616 DISCUSSION

acceptable whether they are interpreted for load

testing or for the less usual application of quake
in pile-driving calculations.

M. Maugeri, F. Castelli and E. Motta, University

of Catania, Italy
To evaluate non-linear single pile settlement,
the Author uses the hyperbolic load-transfer func-
t tion proposed by Chin (1970) which we used to
0 ND, 0 z
present a computer code based on a pile finite
Fig. 17 element discretization which takes into account
the non-linearity of the soil-pile interaction
(Castelli, Maugeri & Motta, 1992). We also pro-
posed a simplified procedure in a closed form
The Author’s comment that Ms does not
similar to that proposed by the Author.
appear to be sensitive to soil type is significant.
The computer code was used to back-analyse
This is also found to be the case with quake Q,
12 loading tests of bored piles. Numerical analysis
which engineers often assume to be 2.5 mm.
was carried out on instrumental full-scale pile
The position with regard to the factor
tests, collected from existing literature. The piles
0.6 U$E,D, is less well defined. Table 1 shows
were bored in clayey, silty, sandy and pyroclastic
this to be more sensitive than M, D, to soil type. I
soils, ranged between 14 m and 42 m long and
think that this is a consequence of the end resist-
had diameters ranging between 0.42 m and 2 m.
ance continuing to increase with displacement
Assuming a hyperbolic load-transfer function, the
and therefore showing no tendency to approach a
back-analysis was performed with the aim of
horizontal asymptote. This has implications for
deducing the most appropriate values of the main
the values of end quake currently assumed in pile-
parameters. The shaft flexibility factor can be
driving analyses.
evaluated using Ms = O.OOlLO.002 when the unit
I should like to comment on the influence of
ultimate skin friction is greater than 50 kPa and
shaft flexibility. The method devised by Poskitt &
M, = 0.002ZO.005 when the unit ultimate skin
Ward (1988) has some similarities with that pro-
friction is less than 50 kPa. These values are very
posed by the Author, but proved to be ill-
close to those given by the Author when applying
conditioned. As a result, the parameters obtained
the Randolph and Wroth theory (1978).
from back-analysis of field data were found to be
The Author suggests evaluating the contribu-
sensitive to small changes in the data. From this I
tion of the settlement due to the elastic shortening
concluded that it was necessary to treat the pile
by three stages AL, AZ and A3. An alternative
and soil as a properly formulated non-linear
procedure could be to assume the soil to be
structural system and solve it accordingly. By
homogeneous and to evaluate the elastic short-
comparison, the Author’s formulation appears to
ening in one stage, considering two laws of soil-
be well conditioned, and it is not immediately
pile interaction-along the pile shaft and at the
apparent why this should be so. Nevertheless, the
base-with different stiffnesses. This gives the
efficacy of the method seems beyond doubt and
second order differential equation
the parameters obtained in the examples are
y" - $y =o
(22)

where y is the settlement at depth z and @* is

Table 1. Load test parameters given by
ix= = 4K,/E, nD2
Site Ms&: mm 0.6 UJE,D,: mm
where K, is the stiffness of the lateral load-
Wembley H 1.32 23.4
transfer function and D is the pile diameter. With
Wembley N 0.94 21.4
Wembley P 1.13 23.7 the boundary conditions
Kallo 5 0.57 6.4
Kallo 2 0.32 3.2
Norwich 1 0.92 23.4
Shrewsbury TPl 0.90 13.6
Shrewsbury TP2 0.71 9.1
Bristol 0.42 35.5 E, T Y’(L) = -(PT - Ps) = - K, y(L) (25)

Downloaded by [ UNIVERSIDAD NACIONAL DE COLOMBIA] on [17/11/19]. Copyright © ICE Publishing, all rights reserved.
 DISCUSSION 617

where K, is the stiffness of the load-transfer func- Load. t

tion at the base and L is the total length of the

pile, the following equation can then be derived
for the elastic shortening

A, = C(4P&E, nD2) (26)

where

c = eaL(l + 8) + emaL(l- PI - 2

I
(27)
eaL(l + p) - emmL(l - B) x Whltaker & Cooke (1966)
0 Author
16
fi = 4K,/aE, nD2 (28) - Maugerl et al.
x

To consider the non-linearity of the load-

settlement curve, the value of K, must be chosen Fig. 19. Evaluation of total settlement for pile N
depending on the load level 4

K, = Ki,(l - q) (29)
from equations (17)-(21). Although our procedure
where Ki, is the stiffness at the origin and predicts a non-linear elastic shortening and the
Author’s predict a linear one, the results are in
rl = P&J, + U,) (30)
good agreement. The main difficulty in applying
Thus this procedure, which takes into account these methods is in the correct determination of
both the shaft and base interactions as well as the the function parameters. If it is possible therefore
load level, also shows the non-linearity of the for the load-transfer function to be characterized
elastic shortening. accurately by using a simplified procedure, ana-
Values of Ki, may be derived as Ki, =
Us/M, D, and the following approximate relation-
ship, deduced from parametric back-analysis, can
be used to calculate K, Load. I
800 1200
I
(q,/K, D) = 0.03 (31)
D,=D,=l-5m
where q, is the point resistance deduced from U, = 12530 kN
static penetration testing. U, = 2480 kN
Figures 18 and 19 compare the head settle- Lo= 10m
L, = 32 m
ments evaluated using the Author’s procedure M, = 0 0017
with ours for piles H and N (Whitaker & Cooke, Eb = 36460 kN/m*
E, = 26 x IO6 kN/m’
1966). Both procedures were also applied in one K, = o-45
of the 12 loading tests (Viggiani & Vinale, 1983), x
\
which was used for the back-analysis, as shown in x Vlgglanl 8 Vwlale (1983)
Fig. 20. All the results show good agreement o Author
- Maugw et al
between measured and computed settlements,
irrespective of the procedure used.
Fig. 20. Evaluation of total settlement for Viggiani &
For the Viggiani & Vinale (1983) pile, Fig. 21 Vinale (1983) pile
shows the elastic shortening A, derived from
equations (26)-(31) compared with that derived

Load: t

6 x Author
- Maugerl et al.

o.sL
Fig. 18. Evaluation of total settlement for pile H Fig. 21. Evaluation of elastic shortening A.

Downloaded by [ UNIVERSIDAD NACIONAL DE COLOMBIA] on [17/11/19]. Copyright © ICE Publishing, all rights reserved.
 618 DISCUSSION

lytical results will show good agreement with full- how locked in stresses are produced after
scale pile tests. unloading in a pile. Other developments associ-
ated with the method have also been published
Author’s reply (Fleming, 1992, 1993; England, 1992).
With regard to the violation of the single
Since the Paper was written the routine use of hyperbolic relationship when two hyperbolic
the method for the analysis and prediction of functions are added for a given deformation, this
single pile settlements and further research on it is indeed inevitable. It is easy to demonstrate
has been proceeding and the database has been mathematically and largely explains the frequent
widened to more than 500 case studies without ambivalence of engineers towards the plotting
any significant problems having arisen. Research method. Tables 2-4 illustrate the problem. In
has also been carried out on unloading recovery Table 2 a hyperbolic shaft function characteristic
and it now appears that the maximum value of for a rigid pile is shown. A similar characteristic
this can be predicted by a further extension of the for a pile base alone is shown in Table 3, and in
same mathematical functions. This also shows Table 4 these are mathematically added together

Table 2. Rigid pile with D, = 1 m, Us = 2000 kN, Ms = 0902

Applied load: kN Settlement: mm Settlement/load Interval slope:

X 103 kN
181.81 0.2000 1.100
2000
400 0~5000 1.250
2000
620.69 0.9000 1.450
2000
800 I.3333 1.6667
2000
1000 2~0000 2.000
2000
1200 3.0000 2.500
2000
1428.58 5~0000 3.500
2000
1600 8.0000 5.000
2000
1800 18~0000 10~000

Table 3. Rigid pile with D, = 1 m, U, = 1000 kN, Es = 50000 kN/m* (E,,

base soil mo&&)

Applied load: kN Settlement: mm Settlement/load Interval slope:

x 103 kN

16.39 0~2000 12.20

1000
40 0~5000 12.50
1000
69.77 0.9000 12.90
1000
100 1.3333 13.33
1000
142.86 2~0000 14.00
1000
200 3~0000 15GO
1000
294.12 5.0000 17.00
1000
400 8.0000 2oxKl
1000
600 18.0000 30.00

Downloaded by [ UNIVERSIDAD NACIONAL DE COLOMBIA] on [17/11/19]. Copyright © ICE Publishing, all rights reserved.
 DISCUSSION 619

Table 4. Rigid pile with D, = 1 m, D, = 1 m, Us = 2000 kN, Ms = OM2,

lJB = 1000 kN, Es = 50 000 kN/m*

Applied load : kN Settlement: mm Settlement/load Interval slope:

X lo3 kN
198.2 0.2000 1.00908
2356.9
440 0~5000 1.13636
2393.5
690.46 0.9000 1.303479
2434.2
900 1.3333 1.4815
2482.9
1142.86 2.0000 1~7500
2545.2
1400 3.0000 2.1429
2633.3
1722.1 5.0000 2.9024
2733.2
2000 8.0000 4.0000
2857.1
2400 18~0000 75000

fcr given settlements. The result, considering necessary, as Figs 18-21 imply. The model used
slope over the selected settlement intervals, for elastic shortening in the Paper actually pro-
clearly shows that the relationship AT/PI. against duces a bilinear model which changes slope at the
Ar is no longer linear. It is undoubtedly true that point where all shaft friction has been mobilized.
in order to represent pile behaviour adequately, The values for Ms shown in the Paper are now
the functions representing the shaft and base, borne out, at least for the stiffer ranges of soil, by
swhich are individually hyperbolic, have to be a wealth of practical experience.
dealt with separately and subsequently combined
to represent the whole pile. The method advo-
cated by Chin (1970), for example, works well for REFERENCES
piles which have nearly all their load carried Castelli, F., Maugeri, M. & Mona, E. (1992). Analisi
either by shaft friction or end bearing, but is dis- nonlineare de1 cedimento di un palo singolo. Riu.
Ital. Geotecnica, 26, 115-135.
appointing when these components act together
Chin, F. K. (1970). Estimation of the ultimate load of
and are nearly equal.
piles from tests not carried to failure. Proc. 2nd S.E.
On the subject of quake in pile driving, where Asian Conj: Soil Engng, Singapore, 81-92.
elastic shortening is restricted to that within the Fleming, W. G. K. (1992). Limit states and partial
wave front and where volumetric strain along the factors in foundation design. Proc. Instn Ciu. Engrs
shaft length is slight, it seems highly probable Cit. Engng 92, 185-192.
that quake Q is directly related to Ms D, Studies Fleming, W. G. K. (1993). The improvement of pile per-
on base behaviour under impact seem to show formance by base grouting. Proc. Instn C’iu.Enyrs
that the stiffness E, approaches a limiting value Ciu. Enyny 97, May, 88-93.
England, M. (1992). Pile settlement behaviour: an accu-
of the stiffness of water (E, = 2 x lo6 kN/m’),
rate model. Application of stress waue theory to piles,
which might not be entirely unexpected in fully
pp. 91-96. Rotterdam: Balkema.
saturated soils. Poskitt, T. J. & Ward, G. (1988). The evaluation of pile
With regard to the question of the elastic short- load test data in calcareous soils for use in pile
ening model, at an early stage other forms of driving calculations. Engineering for calcareous sedi-
analysis using the same basic functions were con- ments, vol. 2, pp. 449-460. Rotterdam: Balkema.
sidered but it was decided to use the method in Randolph, M. F. & Wroth, C. P. (1978). Analysis of
the Paper because it is straightforward, easily deformation of vertically loaded piles. J. Geotech.
Engng Div. Am. Sot. Ciu. Engrs 104, GT12, 1465.
understood, and may be used as an everyday
Whitaker, T. & Cooke, R. W. (1966). An investigation of
design and analysis tool. The elastic shortening
the shaft and base resistances of large bored piles in
model has since been refined for analysis pur- London Clay. Large bored piles, pp. 7-49. London:
poses and the further suggestion of Messrs Institution of Civil Engineers.
Maugeri, Castelli and Motta is welcomed. For Viggiani, C. & Vinale, F. (1983). Comportamento di pali
design purposes, however, the more refined tech- trivellati di grande diametro in terreni piroclastici.
niques make slight differences and are scarcely Riu. Ital. Geotecnica, 17, 59-83.

Downloaded by [ UNIVERSIDAD NACIONAL DE COLOMBIA] on [17/11/19]. Copyright © ICE Publishing, all rights reserved.