SOILS AND FOUNDATIONS Vol. 51, No.

3, 483–496, June 2011

Japanese Geotechnical Society

UPLIFT CAPACITY OF BELLED AND MULTI-BELLED PILES IN DENSE SAND

TSUYOSHI HONDAi), YOSHIO HIRAIii) and EIJI SATOiii)

ABSTRACT
This paper evaluates the uplift capacity of belled and multi-belled piles in dense sand. A two-dimensional distinct
element (DE) analysis was applied in pullout tests on single piles to investigate the uplift resistance of the piles, the soil
behavior around the piles, and the interaction between the soil and the pile surface. It was observed from the DE analy-
sis that the soil mass adjacent to the projections of the belled and multi-belled piles moved vertically, and that the soil
movements leaned slightly with the occurrence of relative displacement between the soil and the pile surface. A theoret-
ical solution for predicting the uplift capacity of belled and multi-belled piles was derived from an upper bound limit
analysis based on the soil movements in the DE analysis. The solution was able to reproduce the ultimate uplift
resistance in the DE analysis using the friction angles in the aggregates and on the pile surface that were evaluated from
a simulation of direct and simple shear tests. In addition, a continuity equation that satisˆed the relationship between
the displacement vector of the soil mass adjacent to the projections and the change in volume around the soil mass was
proposed for predicting the uplift capacity of actual piles under axisymmetric conditions. The theoretical solution ob-
tained with the continuity equation was in good agreement with the pullout resistance of the belled and multi-belled
piles in centrifuge model tests and full-scale tests conducted in situ.

Key words: (belled pile), dense sand, distinct element method, (multi-belled pile), pulling test, (uplift capacity), (IGC:
E4/E13)

plied to the uplift capacity of multi-belled piles. In con-

INTRODUCTION trast, the uplift resistance of anchor plates has been theo-
During an earthquake, the uplift force acts on the piles retically studied by many researchers (Balla 1961;
under high-rise buildings. Belled and multi-belled piles Meyerhof and Adams, 1968, among others). Rowe and
have been used to increase the uplift and the bearing Davis (1982) investigated the eŠects of anchor embed-
capacity of foundations. They are cast-in-place concrete ment, the friction angle, dilatancy, initial stress state K0,
piles with one or a plural number of enlarged projections and anchor roughness for anchor plates using an elasto-
that are built in the bearing stratum and the middle hard plastic ˆnite element analysis, and proposed the factors
layer ( see Fig. 1). Figure 2 shows photographs of the spe- accounting for these eŠects. Murray and Geddes (1987)
cial buckets used to build the projections, namely, the showed that the theoretical solutions obtained from an e-
middle and the bottom projections, of a multi-belled pile quilibrium analysis, a lower bound limit analysis, and an
constructed to execute control tests. However, a design upper bound limit analysis could produce good agree-
method for the uplift resistance of belled and multi-belled ments with the experimental results of pullout tests on an-
piles has not yet been established, because it is unclear chor plates in dense sand. Kumar (2003) proposed a theo-
how the uplift resistance of soil is induced. Therefore, retical solution for the uplift capacity of anchor plates in
several axial tensile load tests have been conducted in two-layered sand by an upper bound limit analysis.
many studies to ascertain the design method. From these Meriˆeld and Sloan (2006) conducted a numerical limit
load tests, several methods to practically predict the up- analysis on the pullout capacity of horizontal and vertical
lift capacity of belled piles have been proposed by Tsutsui anchor plates, and compared them with numerous theo-
et al. (1995) and Chatani et al. (2008). Most of them sug- retical solutions. Dickin and Leung (1990) attempted to
gest a design to estimate the lowest value of skin friction apply the theoretical solution for anchor plates to the up-
in the belled piles that was measured in the axial tensile lift capacity of belled piers in a centrifuge model. They
load tests. However, the methods frequently underesti- showed that the uplift capacity of belled piers was lower
mate the uplift capacity of belled piles. In addition, little than that of anchor plates. Dickin and Leung (1992) con-
is known about whether or not the methods can be ap- ducted many pullout tests on belled piers using centrifuge
i)
Associate Chief Researcher, Research and Development Institute, Takenaka Corporation, Chiba, Japan (honda.tsuyoshi＠takenaka.co.jp).
ii)
Senior Chief Researcher, ditto.
iii)
Deputy General Manager, Soil Remediation Engineering Department, Takenaka Corporation.
The manuscript for this paper was received for review on April 30, 2010; approved on April 12, 2011.
Written discussions on this paper should be submitted before January 1, 2012 to the Japanese Geotechnical Society, 4-38-2, Sengoku,
Bunkyo-ku, Tokyo 112-0011, Japan. Upon request the closing date may be extended one month.

483
484 HONDA ET AL.

models, and proposed an empirical design method of pullout tests on single piles was conducted to inves-
wherein appropriate factors that account for the founda- tigate the uplift resistance of piles, the soil behavior
tion geometry of belled piers are applied to an anchor for- around the piles, and the interaction between the soil and
mula. Thus, the theoretical solution for anchor plates the pile surface.
cannot be directly applied to the uplift capacity of belled In the DE analysis, two types of disks were used, name-
and multi-belled piles, because anchor plates are horizon- ly, circular disks and cluster disks, formed from two cir-
tal and are used in shallow foundations, whereas belled cular particles. Before starting the simulation of the pull-
and multi-belled piles under high-rise buildings have slop- out tests, the mechanical properties of the aggregates
ing surfaces and are used in deep foundations. This study comprising the circular or the cluster disks were evaluated
evaluates the uplift capacity of belled and multi-belled from simulations of direct and simple shear tests to iden-
piles embedded in dense sand as a deep foundation. tify the friction and the dilatancy angles in the aggregates
Belled and multi-belled piles in loose sand are not exam- and on the pile surface. Four kinds of piles were tested in
ined because belled piles are rarely constructed in loose the pullout tests on single piles, a straight pile, a belled
sand. pile, and two multi-belled piles. In addition, a theoretical
The distinct element method (Cundall, 1971) is a solution to predict the uplift capacity was obtained from
powerful tool for simulating large and discontinuous soil an upper bound limit analysis based on soil movements in
deformations. A two-dimensional distinct element (DE) the DE analysis. The theoretical solution was compared
analysis can directly reproduce the interaction between with the results of the DE analysis, centrifuge model
the soil and the pile surface, which is often di‹cult to do tests, and full-scale tests conducted in situ.
with ˆnite element methods. In this study, a DE analysis

MECHANICAL PROPERTIES OF AGGREGATES

Two types of disks were used in the DE analysis to in-
vestigate the in‰uence of the mechanical properties of soil
on the uplift capacity of single piles. Figure 3 illustrates
the shapes of the circular and the cluster disks used in the
DE analysis. The cluster disks were formed from two cir-
cular particles with diameter ``b'' and size ``a'' in the
long axis. The ratio in length of the short axis to the long
axis was 0.7 for all disks. The circular and the cluster
disks have the same grain size distribution curve, which is
plotted in Fig. 4. Table 1 shows the numerical
parameters used in the DE analysis.
Before starting the simulation of the pullout tests on

Fig. 1. Schematic diagrams of a belled pile and a multi-belled pile

Fig. 3. Shapes of the circular and the cluster disks used in the DE
analysis

Fig. 2. Photographs of the special bucket to construct the parts of Fig. 4. The grain size distribution curve of the aggregates comprising
middle and bottom projections the circular or the cluster disks
 UPLIFT CAPACITY OF PILES 485

Table 1. Numerical parameters used in the DE analysis

Average of diameter 2 mm
Density of particles 1.8E＋03 kg/m3
Normal stiŠness (kn) 1.5E＋08 N/m
Shear stiŠness (ks) 3.8E＋07 N/m
Normal damping (hn) 3.8E＋01 Ns/m
Shear damping (hs) 9.6E＋00 Ns/m
Cohesion (c?) 0.0 N
Friction angle (q?) 40.09

Fig. 5. Schematic diagrams of the models in direct and simple shear Fig. 6. Results in the simulation of direct shear tests using the circular
tests to evaluate mechanical properties of the aggregates disks

the single piles, direct and simple shear tests under con-
stant pressures were reproduced to evaluate the mechani-
cal properties of the aggregates comprising the circular or
the cluster disks, as shown in Fig. 3. The direct shear tests
directly give the friction and the dilatancy angles of the
aggregates on failure lines. The mechanical properties ob-
tained from the simple shear tests are useful for evaluat-
ing the interaction between the soil and the pile surface
for the upper bound limit analysis described in a subse-
quent chapter. Schematic diagrams of the models in the
direct and simple shear tests are presented in Fig. 5. The
models were 100 mm in both width and height. The top
and bottom boundaries of the models had a friction angle
of 409 . The side walls in the direct shear tests had no fric-
tion. Repeated boundaries were used for the right and left
boundaries of the simple shear tests. In the packing of the
disks, the friction angle for the disk-to-disk was tem-
porally decreased from 409to 109to achieve a dense con-
dition. The disks were randomly arranged and com-
pressed at vertical pressures of 100, 200, 300, and 400
kPa. The models were sheared after the friction angle was
reset to 409. Fig. 7 Results in the simulation of simple shear tests using the circular
Figures 6 and 7 show the mobilized friction angles, disks
qmob＝tan－1 (t/sn), and the vertical displacements against
the shear displacements in the direct and simple shear
tests using the circular disks. The diŠerence in vertical did not change, because the soil attained the critical state.
pressure hardly aŠected the mobilized friction angle or From the initial gradient of the vertical displacements
the vertical displacement. In the direct shear tests, strain against the shear displacement, a dilatancy angle of about
softening was remarkably observed after the mobilized 279was obtained. In the simple shear tests, the mobilized
friction angle reached about 289at the peak strength, and friction angles and the vertical displacements were
then it decreased to about 229 . The vertical displacement remarkably smaller than those in the direct shear tests,
due to dilatancy gradually increased until the strength and the shear displacement at the peak strength was larg-
dropped to the residual strength. Once the strength er than that in the direct shear tests. The friction angles at
reached the residual strength, the vertical displacement the peak strength in the simple shear tests were about 199 ,
486 HONDA ET AL.

Fig. 10. Deformations in the direct and simple shear tests using the
circular disks

Fig. 8. Results in the simulation of direct shear tests using the cluster
disks
Fig. 11. Deformations in the direct and simple shear tests using the
cluster disks

Table 2. Mechanical properties of the aggregates comprising the circu-

lar or the cluster disks

Unit weight of the aggregates 15.5 kN/m 3

(a) Circular disks
Soil in shear zone
Friction angle 289
Dilatancy angle 279
Interaction between soil and the pile surface
Friction angle 199
Dilatancy angle 8.49
(b) Cluster disks
Soil in shear zone
Friction angle 379
Dilatancy angle 339
Interaction between soil and the pile surface
Friction angle 279
Dilatancy angle 129

lar disks. In the direct shear tests, the friction and the
dilatancy angles at the peak strength were 379and 339,
Fig. 9. Results in the simulation of simple shear tests using the cluster respectively. In the simple shear tests, the friction and the
disks dilatancy angles were 279and about 129 , respectively.
The ˆnal deformations in the direct and simple shear
tests under the vertical stress of 400 kPa are illustrated in
and a dilatancy angle of about 8.49was obtained. This Figs. 10 and 11. The shear deformation in the direct shear
must be because the disks easily slipped along the top and tests was concentrated in the gap between the top and
bottom boundaries in the simple shear tests. bottom boxes. The shear deformation in the simple shear
Figures 8 and 9 show the results of the direct and sim- tests was predominant around the top and bottom
ple shear tests using the cluster disks. The curves of the boundaries, since the disks easily slipped. Accordingly,
mobilized friction angles and the vertical displacements this is why the friction and the dilatancy angles in the sim-
against the shear displacements in both tests were very ple shear tests were smaller than those in the direct shear
similar to those using the circular disks. However, their tests. In the simulation of the pullout tests on single piles,
values were larger than those in the cases using the circu- the pile surface was also reproduced by the same bound-
 UPLIFT CAPACITY OF PILES 487

aries as the top and bottom boundaries in the simple

shear tests. Therefore, it might be assumed that the disks
on the pile surface also slip easily, and that the mechani-
cal properties of the disks on the pile surface are similar
to those in the simple shear tests.
To sum up the results of the direct and simple shear
simulation, the mechanical properties of the aggregates
with the circular or the cluster disks were almost the same
in quality, although they were quantitatively diŠerent.
The friction and the dilatancy angles on the boundaries,
such as the pile surface, are assumed to be signiˆcantly
lower than those in the soil. Table 2 summarizes the
mechanical properties obtained from the simulation of
the direct and simple shear tests. Fig. 12. Schematic diagrams of the piles used in the DE analysis (the
sizes of piles are expressed in prototype scale)

PULLOUT TESTS ON SINGLE PILES

Two series of pullout tests on single piles were conduct-
ed in the DE analysis. One case used the circular disks,
while the other used the cluster disks, as shown in Fig. 3.
Four kinds of single piles were tested in both series.
Figure 12 presents the schematic diagrams of the single
piles in prototype scale. Case 1 was a straight pile with a
diameter of 1.4 m; Case 2 was a belled pile with diameters
of 1.4 m in the shaft and 2.5 m at the base; Cases 3 and 4
were multi-belled piles with diameters of 1.4 m in the
shaft and 2.5 m in the projections-they were diŠerent in
terms of the distance between the upper and lower projec-
tions. The total length of the piles was 23.5 m. The incli-
nations at the upper and lower surfaces in the projections
were 129and 459, respectively. The pile surface had the
same friction angle of 409as the top and bottom bounda-
ries used in the simulation of the simple shear tests. As an
example, the numerical model that has half a cross sec-
tion in Case 4 is shown in Fig. 13. The piles were repro-
duced as a rigid body with a self-weight of about 425
kN/m in unit depth. The DE analysis of the pullout tests Fig. 13. Numerical model of the multi-belled pile in Case 4
on the single piles was conducted in a 50-times gravity
ˆeld (50 G) using miniature models with a scale of 1/50.
The reasons for this are to use the same circular and The uplift resistance of the piles was calculated by in-
cluster disks as those used in the simulation of the direct tegrating the increment of downward forces induced by
and simple shear tests and to reduce the number of disks the contact forces between the disks and the pile surface.
for minimizing the calculation time. The similarity rule in Figures 14 and 15 show the uplift resistances against the
the DE analysis is the same as that in centrifuge model uplift displacements of the piles in the series using the cir-
tests. The simulation was conducted in the following cular or the cluster disks. In these ˆgures, a sudden reduc-
manner: tion in uplift resistance repeatedly occurred. The reason
(i) All the disks with a disk-to-disk friction angle of for this is that the skeleton of the disks' structure col-
109were randomly arranged in the analyzed area. lapsed locally and the uplift resistance suddenly
(ii) An acceleration of 50 G was applied to all the decreased. This is a particular phenomenon in two-
disks, and then the model ground was compressed dimensional DE analyses. The straight pile in both series
by the self-weight of the disks to yield a dense con- showed yielding at the lowest value of uplift resistance.
dition. The uplift resistance of the straight pile did not increase
(iii) After the friction angle in all the disks was reset to just after yielding, but maintained the resistance of about
409 , the disks that overlapped with the area of a 1200 kN/m. In contrast, the belled and multi-belled piles
single pile were removed; then the single pile with showed a gradual increase in uplift resistance after the
self-weight was inserted. yielding occurred at the uplift resistance of about 1200
(iv) The pile was pulled out at a constant rate of 5 kN/m. The stiŠness of the belled and multi-belled piles in
mm/sec up to the uplift displacement of 500 mm the initial gradient of the uplift resistance was slightly
in the prototype scale. larger than that in the straight piles. It can be said from
488 HONDA ET AL.

this that the uplift resistance is ˆrstly induced by the skin uplift displacement. The multi-belled piles showed the
friction, and that the eŠects of the projections in the highest stiŠness and uplift resistance. The multi-belled
belled and multi-belled piles then develop after a certain piles of Case 4, with the longer distance between two
projections, had the highest uplift resistance of all the
cases. Comparing the series using the circular and the
cluster disks, the uplift resistance of the straight piles was
almost the same for both series. The uplift resistance of
the belled pile in the cluster disks was larger than that in
the circular disks. Thus, it was found that the eŠects of
the projections in belled and multi-belled piles largely de-
pend on the mechanical properties of the soil.
Figures 16(a), (b), (c), and (d) show the ˆnal ground
deformation at the vertical displacement of 500 mm in
the series using the cluster disks. It is found from these
ˆgures that the ground deformation was very small in all

Fig. 14. Uplift resistance of the single piles against uplift displace-
ments in the simulation using the circular disks

Fig. 15. Uplift resistance of the single piles against uplift displace- Fig. 16. Final ground deformations in the simulation using the cluster
ments in the simulation using the cluster disks disks

Fig. 17. Loci of soil movements until the uplift displacement of 250 mm in the simulation using the circular disks
 UPLIFT CAPACITY OF PILES 489

Fig. 18. Loci of soil movements until the uplift displacement of 250 mm in the simulation using the cluster disks

cases, and that no remarkable failure line was observed.

In the cases of belled and multi-belled piles, the ground
deformation was slightly induced in the limited area adja-
cent to the projections. The ground deformation in the
series using the circular disks was also small and similar
to that in the series using the cluster disks. The 10-times
displacement vectors of the soil movements at the uplift
displacement of 250 mm are illustrated in Figs. 17 and
18. The circles in Figs. 17 and 18 indicate the initial posi-
tions of some disks, while the curves show the displace-
ment vectors of the soil movements. It was found that the
soil movements in the straight pile were very small and Fig. 19. Distributions of horizontal stress at the uplift displacement of
were limited in the closed area around the pile. The soil 250 mm in the simulation using the cluster disks
movements in the cases of the belled and multi-belled
piles were larger than that in the case of the straight pile,
and the areas where large movements were induced were
wider. In addition, the soil movements in the series using
the cluster disks with the large friction angle became larg-
er than that in the other series. With respect to the behav-
ior of the soil movements in the cases of the belled and
multi-belled piles, it was observed that the soil move-
ments around the projections were larger than those
around the shaft in all the cases. The soil adjacent to the
projections ˆrstly moved in an upward direction, and
then gradually inclined in a horizontal direction with the
increase in uplift displacement. It was also observed that
the soil mass with the shape of a reverse triangle moved as Fig. 20. Distributions of vertical stress at the uplift displacement of
a body in the area adjacent to the projections, and that 250 mm in the simulation using the cluster disks
the soil movements in that area were not fully transmitted
to the upper layer.
Figures 19 and 20 show the distributions of vertical bearing eŠect, was seen in the limited area adjacent to the
and horizontal stress at the vertical displacement of 250 projections. In addition, compressive and shear deforma-
mm in the series using the cluster disks. The concentra- tion, due to the load-bearing eŠect, was intensively in-
tion of vertical and horizontal stress, due to the load- duced in the area where vertical and horizontal stress was
490 HONDA ET AL.

concentrated. No remarkable failure line was detected in line BC developed (Kanatani et al., 1973). However, it
the distribution of the maximum shear strain. was derived from a comparison of several failure
mechanisms that the uplift capacity became the lowest
value at the relative displacement of dwba＝0.
UPLIFT CAPACITY OF BELLED AND The external loads in Fig. 21(a) represent the gravity
MULTI-BELLED PILES UNDER acting on blocks ABD and BCD and the overburden pres-
TWO-DIMENSIONAL CONDITIONS sure on lines AD and DC. They are calculated by Eqs. (2)
In this chapter, the ultimate uplift resistance of belled and (3).
and multi-belled piles under two-dimensional conditions 1 1
is evaluated by the upper bound limit analysis based on Wa＝ gH2 tan u, W b＝ gH2/tan (a－q) (2)
the soil movements in the DE analysis. The uplift capaci- 2 2
PA＝paH tan u, PB＝pbH/tan (a－q) (3)
ty in actual piles under axisymmetric conditions will be
discussed in the next chapter. Figure 21 shows the where g is the unit weight of the soil, Wa and Wb are the
mechanism of the plastic collapse around the projections weight in blocks ABD and BCD, and Pa and Pb are the
of the belled and multi-belled piles. The plastic collapse overburden forces that are calculated by integrating the
can reproduce the phenomenon whereby soil adjacent to overburden pressure on lines AD and DC. The increment
the projections moves vertically as a body and then leans of work done by the external loads is expressed as Eq.
slightly, and that the soil movements are not fully trans- (4).
mitted to the upper layer in Figs. 17 and 18. In Fig. 21, P
DW＝P･dv0－( Pa＋Wa)･dwa sin a－( Pb＋Wb)･dwb sin a
is the ultimate uplift resistance, pa and pb are the overbur-
(4)
den pressures on lines AD and DC, D1 and D2 are the di-
ameters of the shaft and the projection, H and u are the The work done by the internal stress is calculated on
height and the inclination of the upper slope in the lines AB and BC. Now, the dilatancy angle is assumed to
projection, a is the direction of the soil movements in equal the friction angle. The increment in work done by
blocks ABD and BCD, c and q are the cohesion and the the internal stress is due to cohesion, and it is expressed as
friction angle of the soil, c is the dilatancy angle of the follows:
soil that is assumed to equal friction angle q, d is the fric-
DE＝c･H/cos u･dwa0 cos d＋c･H/sin (a－q)･dwb cos q
tion angle on the slope in the projection, which depends
(5)
on the roughness of the pile surface and loosening during
the excavation, dv0, dwa, and dwb are the displacement From the theory of the upper bound method, the incre-
vectors of the pile, block ABD, and block BCD, and dwa0 ment in work done by the external load and the internal
is the relative displacement vector on line AB. stress are equal.
The displacement vectors in Fig. 21(b) must satisfy the
D W ＝ DE (6)
following equations:
The ultimate uplift resistance is calculated by substituting
dv0＝dwa sin a＋dwa cos a/tan (u＋d) (1a)
Eqs. (1a)–(1d) into Eq. (6). It is expressed as Eq. (7a),
d w b ＝ dw a (1b)
namely,
dwba＝0 (1c)
dwa0＝dwa cos a/sin (u＋d) (1d) 1
P＝ s( Pa＋Wa)･sin a
sin a＋cos a/tan (u＋d)
where dwba is the relative displacement on line BD, which
＋( Pb＋Wb)･sin a
was observed in the tests on an anchor plate after failure
＋c･H/cos u･cos a/sin (u＋d)･cos d
＋c･H/sin (a－q)･cos qt (7a)
The uplift capacity is obtained by searching for the an-
gle of a to minimize the uplift resistance. Using the fric-
tion angle of 289in the aggregates for the circular disks,
the friction angle of 8.49between the disks and the pile
surface and the unit weight of 15.5 kN/m3 in Table 2, the
uplift capacity of a belled pile is about 1508 kN/m ( see
Fig. 14), which includes the piles' self-weight of about
425 kN/m. For the cluster disks, the uplift capacity of
1931 kN/m was obtained from the friction angle of 379in
the aggregates, the friction angle of 129on the pile sur-
face, and the unit weight of 15.5 kN/m3. The angles of a,
applied to minimize the uplift resistance in the models us-
ing the circular and the cluster disks, are 80.49and 84.19,
respectively.
Fig. 21. A mechanism of plastic collapse of the soil adjacent to the The uplift capacity of the belled piles is actually com-
projections in belled and multi-belled piles posed of the skin friction on the shaft and the bearing
 UPLIFT CAPACITY OF PILES 491

force on the slope at the bell. It is di‹cult to exactly esti-

mate the skin friction. The skin friction causes a reduc-
tion in the overburden forces on lines AD and DC in Fig.
21. Therefore, evaluating the share of uplift resistance in
the shaft and the bell is a complicated problem. However,
the total uplift resistance of belled piles can be calculated
simply from Eq. (7a) and the overburden forces that are
obtained by computing the self-weight of the soil directly,
since the skin friction on the shaft is in balance with the Fig. 22. Three mechanisms of plastic collapse in the multi-belled piles
reduction in uplift resistance due to the decrease in over- constructed in two-layer grounds
burden forces. Similarly, Dickin and Leung (1992) meas-
ured the uplift resistance in the shaft and the bell
separately, and showed that the total uplift resistance
correlated with the theoretical solutions for the anchor
plates.
The uplift capacity of the multi-belled piles is obtained
as the sum of the uplift resistance in each projection. As-
suming that soil does not possess any cohesion and that
the friction angle on the pile surface is adequately rough,
for example, it is equivalent to that of soil, the direction
of the soil movements, a in Fig. 21, is always 909 . Conse-
quently, Eq. (7a) is also expressed in the following:
P ＝ P a＋ W a＋ P b ＋ W b (7b) Fig. 23. Relationship between the uplift resistance and the roughness
on the pile surface
The uplift capacity of each projection is determined by
only the weight of the soil mass, which is similar to that
of a horizontal anchor plate proposed by Kumar (2003).
Equation (7b) also suggests that the soil masses used to
induce the uplift resistance are clearly separated in the up-
per and lower projections, as seen in Fig. 22, because the
soil mass above the upper projection is in balance with
the uplift force in the upper projection and does not aŠect
the ultimate uplift resistance in the lower projection.
Figure 22 shows three failure mechanisms of the multi-
belled piles in a two-layer ground. The hatched areas in
Fig. 22 indicate the soil mass resisting the uplift force.
The shape of the soil mass in the hatched areas depends
Fig. 24. Relationship between the direction of soil movements and the
on the geometry of the piles and the friction angles in the
roughness on the pile surface
soil and on the pile surface. When multi-belled piles with
the same dimensions in the upper and lower projections
are constructed in a homogeneous layer, as seen in Fig.
22(a), the uplift capacity almost equals that of a belled that the soil movements in Figs. 17 and 18 were induced
pile without the upper projection. In other words, multi- in the outer area of block ABC in Fig. 21 with the large
belled piles do not always increase the ultimate uplift uplift displacement.
resistance in comparison to belled piles. However, multi- Equation (7a) implies that the uplift capacity of belled
belled piles can make the stiŠness in uplift resistance larg- and multi-belled piles is considerably in‰uenced by the
er than belled piles, for the uplift capacity in each projec- friction angle on the pile surface. Figures 23 and 24 show
tion is reduced by the share in the upper and lower projec- the in‰uence of the friction angle on the pile surface in
tions, and the uplift displacement to reach the ultimate the uplift capacity and the direction of soil movements.
uplift resistance is also decreased, as seen in Figs. 14 and With a decrease in the friction angle on the pile surface,
15. Now, when the model ground in the DE analysis is re- the uplift capacity is decreased and the direction of the
garded as a homogeneous one, the uplift capacities of the soil movements is inclined.
belled and the multi-belled piles are equal, as shown in
Figs. 14 and 15. It can be seen that the proposed method
by Eq. (7a) is in good agreement with the results of the UPLIFT CAPACITY OF BELLED AND
DE analysis. However, the belled pile of Case 2 and the MULTI-BELLED PILES UNDER AXISYMMETRIC
multi-belled piles with the longer distance between two CONDITIONS
projections in Case 4 exceeded the upper limit value of The theoretical solution for the uplift capacity in actual
Eq. (7a) in the large uplift displacement. The reason is piles under axisymmetric conditions is derived by in-
492 HONDA ET AL.

tegrating Eq. (7a) around the center axis of the piles in

Dn＝SAC･dwa･sin a－SAB･dwa0･sin d－SBC･dwb･sin q＝0
the following manner:
(15)
1
P＝ s( Pa＋Wa)･sin a where SAC is the surface area formed by rotating line AC
sin a＋cos a/tan (u＋d)
around the center axis of the pile. It was calculated from
＋( Pb＋Wb)･sin a＋c･SAB･cos a/sin (u＋d)･cos d
the following equation:
＋c･SBC･cos qt (8)
p
The weight of soil masses Wa and Wb are calculated by SAC＝ [sD2＋H/tan (a－q)t2－D21] (16)
Eqs. (9) and (10). 4
Equation (15) is also expressed as Eq. (17) by using Eqs.
{ }
p H 3 tan2 u
W a＝ g ( D22－D21)H－D1H 2 tan u－ (9) (1a)–(1d).
4 3
Dn＝SAC sin a－SAB cos a/sin (u＋d)･sin d－SBC sin q＝0

{ }
p D2H 2 H3 (17)
Wb＝ g ＋ (10)
4 tan (a－q) 3 tan2 (a－q)
Overburden forces Pa and Pb are obtained by Eqs. (11)
and (12). CASE STUDIES ON MODEL TESTS AND
FULL-SCALE TESTS IN SITU
p
( D22－D21)
P a＝ p a･ (11) Case studies on the uplift capacity of belled and multi-
4
p belled piles were conducted using centrifuge model tests
2
Pb＝pb･ [s D2＋H/tan (a－q)t －D22] (12) and in situ axial tensile load tests. This chapter discusses
4
the validity of the theoretical solution with the continuity
SAB and SBC are the surface areas formed by rotating lines equation that consists of Eqs. (8) and (17).
AB and BC around the center axis of the piles in Fig. 21.
They are expressed as follows: Case 1: Centrifuge model tests on a belled pile
2 Hirai et al. (2009) conducted centrifuge model tests on
SAB＝p( D1H＋H tan u) (13)
SBC＝psD2H＋H 2/tan (a－q)t (14) pullout tests of a belled pile. The model pile was made of
aluminum and had a length of 290 mm and diameters of
Note that the direction of soil movements, a, to 12.5 mm and 17.5 mm for the shaft and the base. The up-
minimize Eq. (8) does not always satisfy the continuity per slope of the bell had an inclination of 129. The sur-
condition: The vertical displacement on surface AC (Fig. face of the pile was smooth rather than rough, and its
21) is determined by the change in volume due to dilatan- boundaries were similar to those used in the DE analysis.
cy on failure surfaces SAB and SBC. Therefore, a new con- Toyoura sand with a relative density of 90z and a dry
tinuity equation was added to the theoretical solution for unit weight of 15.8 kN/m3 was used for the model
resolving the direction of the soil movements as Eq. (15), ground. Pullout tests were conducted in the centrifuge
namely, ˆeld of 80 G. The pile in prototype scale was 23.2 m in

Fig. 25. Comparison between the theoretical solution and pullout tests of a belled pile in centrifuge model tests conducted by Hirai et al. (2009)
 UPLIFT CAPACITY OF PILES 493

length and 1.0 m and 1.4 m in diameter for the shaft and ing the upper steel pipe and the lower cast-in-place con-
the base, respectively. Figure 25(a) presents a schematic crete pile. The self-weight of the pile was assumed to be
diagram of the model pile. The self-weight of the model 94 kN from the pile geometry. An uplift capacity of 4899
pile was 753 kN in prototype scale and a maximum uplift kN was measured from the axial tensile load tests. The
resistance of 3610 kN was measured in the tests. In the friction angle of the ˆne sand was estimated from the N-
computation of the theoretical uplift capacity, a friction value using Eq. (18).
angle of 409was given as the mechanical property of Toy-
q＝ 20×N＋15 (18)
oura sand. However, the friction angle on the pile sur-
face, d, was unknown. Therefore, theoretical uplift The uplift capacities and the directions of the soil move-
capacities were computed by changing the friction angles ments are calculated against several friction angles on the
on the pile surface. The results are illustrated in Fig. pile surface with the friction angle of 499in soil and a wet
25(b) with the measured uplift resistance. The directions unit weight of 17 kN/m3. The results are plotted in Fig.
of the soil movements are also shown in Fig. 25(b). It is 26(b). Using the relationship d＝q/2, the uplift capacity
noted that the relationship between the direction of the and the direction of the soil movements are 5181 kN and
soil movements and the friction angle on the pile is diŠer- 72.29 , respectively, which are illustrated as black symbols
ent from that of the piles under two-dimensional condi- in Fig. 26(b). Incidentally, the submerged unit weight
tions ( see Fig. 24). In Fig. 25(b), the black symbols show should generally be used instead of the wet unit weight
the uplift capacity and the direction of the soil move- for soil under the water surface, because the buoyancy
ments obtained using half a friction angle of soil as the force must be considered in calculating the work done by
friction angle on the pile surface, d＝q/2. Using the external loads in Eq. (4). However, the theoretical solu-
relationship d＝q/2, the uplift capacity and the direction tion using the submerged unit weight of the soil consider-
of the soil movements are 4072 kN and 65.89 , respec- ably underestimated the uplift capacity. The reason for
tively. this might be that the hydrostatic conditions in soil are
not always maintained during conventional axial tensile
Case 2: In situ axial tensile load tests on a belled pile load tests due to stress control. The uplift displacement
Kamimura et al. (2004) conducted in situ axial tensile of piles actually goes on to increase under a constant load
load tests on a belled pile with a length of 21.5 m and di- when the tensile load comes up to the uplift capacity. It
ameters of 1.0 m and 1.4 m for the shaft and the base, re- cannot be said that drained conditions are fully satisˆed
spectively. The upper slope of the bell had an inclination in the ground. However, there is insu‹cient data to prove
of 129. Figure 26(a) shows a schematic diagram of the this. A study on this topic is expected to be done in the fu-
belled pile and the N-value of the standard penetration ture.
tests in the ground. The water level was G.L. －5 m. The
base of the pile was embedded in dense ˆne sand with an Case 3: In situ axial tensile load tests on a belled pile
average N-value of 58. Skin friction of the pile's shaft, Axial tensile load tests on a belled pile with an inclina-
from G.L. －0 m to －16.5 m, was cut oŠ by disconnect- tion of 9.29 at the upper slope were conducted by

Fig. 26. Comparison between the theoretical solution and in-situ axial tensile load test of a belled pile conducted out by Kamimura et al. (2004)
494 HONDA ET AL.

Fig. 27. Comparison between the theoretical solution and in-situ axial tensile load test of a belled pile conducted by Takagaki et al. (1995)

Fig. 28. Comparison between the theoretical solution and in-situ axial tensile load test of a multi-belled pile conducted by Hirai et al. (2008)

Takagaki et al. (1995). The pile had a length of 18.8 m ˆne sand were used to calculate the uplift capacities and
and diameters of 1.0 m and 1.5 m in the shaft and at the the directions of the soil movements. Figure 27(b) shows
base. Figure 27(a) shows the schematic diagram of the the calculated uplift capacities and the directions of the
belled pile and the N-value in the ground. There was no soil movements against several friction angles on the pile
information about the water level. The base of the pile surface. The uplift capacity and the direction of the soil
was embedded in dense ˆne sand with an average N-value movements using the relationship d＝q/2 were 7387 kN
of 31. The self-weight of the pile was estimated to be 354 and 62.59 , respectively, which are illustrated as black
kN from the pile geometry. An uplift capacity of 8300 kN symbols in Fig. 27(b).
was measured from the axial tensile load tests. A wet unit
weight of 17.5 kN/m3 and a friction angle of 409in the
 UPLIFT CAPACITY OF PILES 495

the lower projection.

Finally, the theoretical uplift capacities are compared
with those in the centrifuge model tests and the full-scale
tests conducted in situ in Fig. 29, where the friction an-
gles in the soil and on the pile surface were calculated
from Eq. (18) and the relationship d＝q/2. It is obvious
from Fig. 29 that the theoretical solution with the con-
tinuity equation can reproduce the uplift capacity of both
the belled and the multi-belled piles.

CONCLUSIONS
Two-dimensional distinct element (DE) analyses on
direct shear tests, simple shear tests, and pullout tests on
single piles were conducted to investigate the uplift capac-
ity of belled and multi-belled piles in dense sand. In addi-
tion, a theoretical solution to predict the uplift capacity
Fig. 29. Summarized results in the comparison between the theoretical
of belled and multi-belled piles was derived from an up-
solution and the test results conducted in a centrifuge model and
situ per bound limit analysis based on the soil movements in
the DE analysis. The continuity equation was proposed
for application to actual piles under axisymmetric condi-
Case 4: In situ axial tensile load tests on a multi-belled tions. The theoretical solution with the continuity equa-
pile tion was compared with the results of centrifuge model
Hirai et al. (2008) conducted axial tensile load tests on tests and full-scale tests conducted in situ. The conclu-
a multi-belled pile with an inclination of 129at the upper sions obtained from this study are summarized as fol-
slope in the two projections. The pile had a length of 23.2 lows:
m and diameters of 1.0 m in the shaft and 1.7 m in the 1. It was found from the DE analysis on direct and sim-
projections. Figure 28(a) shows the schematic diagram of ple shear tests that the friction and the dilatancy an-
the multi-belled pile and the N-value in the ground. The gles on the pile surface were signiˆcantly smaller than
water level was about G.L. －4 m. The upper and lower those in soil.
projections were constructed in ˆne sand with average N- 2. The DE analysis of the pullout tests on single piles
values of 30 and 55, respectively. Skin friction of the demonstrated that the uplift resistance and the stiŠ-
pile's shaft, from G.L. －0 m to －15.8 m, was cut oŠ by ness in the early stage of uplift displacement were larg-
disconnecting the upper steel pipe and the lower cast-in- er in the order of straight piles, belled piles, and multi-
place concrete pile. The self-weight of the pile was esti- belled piles in the series using the cluster disks.
mated to be about 150 kN from the pile geometry. A total However, there was no remarkable diŠerence in the
uplift capacity of 12,500 kN was measured from the axial uplift capacities of the belled and multi-belled piles.
tensile load tests. The shared uplift resistance in the upper 3. Ground deformation induced by the uplift displace-
and lower projections was 7700 kN and 4800 kN, respec- ment of the piles was limited to the area adjacent to
tively. The theoretical uplift capacity of the multi-belled the projections of the piles. The soil movements
pile is calculated using the concept shown in Fig. 22(b). A around the projections were slightly inclined from the
wet unit weight of 17.5 kN/m3 in the soil was used. The vertical direction to the horizontal direction.
friction angles of the soil around the upper and lower 4. A failure mechanism for the soil adjacent to the
projections were estimated to be 409and 489 , respec- projection was proposed, and a theoretical solution
tively, from Eq. (18). The area of the soil mass to resist was derived for estimating the uplift capacity of belled
uplift force in the lower projection was wider than that in piles using the soil conditions and the shape of the
the upper projection, because the ground where the lower piles. A method for predicting the uplift capacity of
projection was embedded had a higher N-value than that multi-belled piles and the shared resistances at the
around the upper projection. The total uplift capacity of projections were also shown.
the multi-belled pile could be calculated by ignoring the 5. The fact that the uplift capacities calculated from the
upper projection. The calculated uplift capacities and theoretical solution corresponded well with the results
directions of the soil movements are illustrated in Fig. in the DE analysis suggests that the mechanical prop-
28(b). The uplift capacity and the direction of the soil erties of the aggregates were properly evaluated by the
movements using the relationship d＝q/2 are 12,643 kN simulation of direct and simple shear tests.
and 70.19, respectively, which are plotted as black sym- 6. A continuity equation was adopted for predicting the
bols in Fig. 28(b). The theoretical uplift capacity in the uplift capacity of actual piles. The theoretical solution
upper projection was obtained as 7753 kN by using fric- obtained with the continuity equation was compared
tion angles of 409in the soil and 209on the pile surface. with the results from centrifuge model tests and full-
As a result, the uplift resistance of 4890 kN was shared at scale tests conducted in situ, and it was in good agree-
496 HONDA ET AL.

ment with them. 3) Cundall, P. A. (1971): A computer model for simulating progres-
7. It was quantitatively revealed from the theoretical so- sive large scale movements in blocky rock system, Proc. of Sympo-
sium , ISRM, Nancy, France, 2, 129–136.
lution that the uplift capacity of belled and multi-
4) Dickin, E. A. and Leung, C. F. (1990): Performance of piles with
belled piles was intensively in‰uenced by the friction enlarged bases subject to uplift forces, Canadian Geotechnical
angle on the slope of the projections. The uplift capac- Journal, 27, 546–556.
ities in the in situ axial tensile load tests are appropri- 5) Dickin, E. A. and Leung, C. F. (1992): The in‰uence of foundation
ately estimated using half the frictional angle of soil as geometry on the uplift behavior of piles with enlarged bases, Cana-
dian Geotechnical Journal, 29, 498–505.
the friction angle on the pile surface.
6) Hirai, Y., Wakai, S. and Aoki, M. (2008): In-situ pull-out test for
The authors showed the validity of the theoretical solu- multi-belled cast-in-place concrete pile: –Part 3– In the Case of
tion by comparing it with the uplift capacity of belled and Diluvial Sand Ground, Annual Meeting of Architectural Institute
multi-belled piles in several axial tensile load tests. of Japan, 577–578 (in Japanese).
However, the authors caution that the use of the solution 7) Hirai, Y., Wakai, S. and Aoki, M. (2009): Centrifuge model tests
on uplift resistance of belled pile in sand, Journal of Structural and
for a design must be done with care, because the friction
Construction Engineering, Transactions of Architectural Institute
angle on the pile surface is intensively in‰uenced by the of Japan, 74(643), 1613–1619 (in Japanese).
construction process, namely, the type of buckets, the 8) Kamimura, K., Hirai, Y., Aoki, M. and Maruoka, M. (2004): In-
drilling slurry, and the looseness in the ground during the situ pull-out test for cast-in-place belled pile: –Part 3– Short Pile
excavation, which all sensitively aŠect the uplift capacity Embedded in Diluvial Laminated Ground, Annual Meeting of Ar-
chitectural Institute of Japan, 553–554 (in Japanese).
of belled and multi-belled piles. Therefore, it is recom-
9) Kanatani, Y., Sato, H. and Akino, Y. (1973): Experimental study
mended that the friction angle on the pile surface be care- on pullout tests of model foundations, Proceedings of the 8th An-
fully evaluated from axial tensile load tests of the piles nual Meeting of Japanese Geotechnical Society, 439–442 (in
that are constructed by exact execution controls. Japanese).
10) Kumar, J. (2003): Uplift resistance of strip and circular anchors in a
two layered sand, Soils and Foundations, 43(1), 101–107.
ACKNOWLEDGMENTS 11) Meriˆeld, R. S. and Sloan, S. W. (2006): The ultimate pullout
capacity of anchors in frictional soils, Canadian Geotechnical Jour-
The authors are grateful to Mr. Tamai, Mr. Wakai, nal, 43, 852–868.
and Dr. Koumura (Research and Development Institute, 12) Meyerhof, G. C. and Adams, J. L. (1968): The ultimate uplift
Takenaka Corporation) for their valuable suggestions re- capacity of foundations, Canadian Geotechnical Journal, 5(4),
225–245.
garding the mechanism of uplift resistance and soil be-
13) Murray, E. J. and Geddes, J. D. (1987): Uplift of anchor plates in
havior around belled and multi-belled piles. sand, Journal of Geotechnical Engineering, ASCE, 113(3),
202–215.
14) Rowe, R. K. and Davis, E. H. (1982): The behaviour of anchor
REFERENCES plates in sand, Geotechnique, 32(1), 25–41.
1) Balla, A. (1961): The resistance of breaking-out of mushroom 15) Takagagi, T., Tsutsui, M., Katoh, K. and Shimatani, Y. (1995): Ex-
foundations for pylons, Proceedings of the 5th International Con- perimental study on pulling resistance of cast-in-place belled pile,
ference on Soil Mechanics and Foundation Engineering, 1, –Part 1–, Annual Meeting of Architectural Institute of Japan,
569–576. 931–932 (in Japanese).
2) Chatani, F., Suzuki, N., Nishimura K. and Wada Y. (2008): Static 16) Tsutsui, M., Takagagi, T., Katoh, K. and Shimatani, Y. (1995): Ex-
axial reciprocal and tensile load tests of single cast-in-place concrete perimental study on pulling resistance of cast-in-place belled pile,
nodular piles, –Part 5–, Annual Meeting of Architectural Institute –Part 2–, Annual Meeting of Architectural Institute of Japan,
of Japan, 575–576 (in Japanese). 933–934 (in Japanese).