Beruflich Dokumente
Kultur Dokumente
3, 301317
K . H O R I KO S H I a n d M . F. R A N D O L P H {
The majority of piled raft foundations have been La conception de la plupart des fondations a
designed ignoring any contribution from the raft dalle ottante sur pieux ne tient aucun compte
or pile cap, although it is well known that the de l'inuence de la dalle ottante ou de la
raft plays an important role in the overall per- structure de liaison, bien qu'on sache fort bien
formance. In such conventional designs, the que la dalle ottante joue un role important
overall settlement of the foundation is likely to dans la performance d'ensemble des fondations.
be very small, owing to the installation of more Dans de telles conceptions classiques, le tasse-
piles than are necessary. However, from an eco- ment general des fondations tend a etre tres
nomical point of view, it is preferable that the faible, car on utilise plus de pieux qu'il ne faut.
foundation is designed in such a way that the Toutefois, d'un point de vue economique, il est
average and differential settlements are limited preferable de concevoir les fondations de facon
to an acceptable level, but where the load-carry- a limiter a un niveau acceptable le tassement
ing capacity of the raft is taken into account. In moyen et le tassement differentiel, en tenant
this paper, a framework for a new design con- compte de la capacite portante de la dalle
cept, in which piles are installed only beneath ottante. L'article presente, au moyen des resul-
the central area of a relatively exible raft to tats d'une parametrisation extensive, le schema
minimize the differential settlement, is presented d'un nouveau type de conception, ou les pieux
through the results of an extensive parametric ne sont places que sous la partie centrale d'une
study. The so-called `hybrid' approach, devel- dalle ottante relativement souple an de mini-
oped by Clancy, was used in the analyses. The miser le tassement differentiel. Les analyses ont
study showed that piled rafts may be designed repose sur la methodologie hybride mise au
for negligible differential settlements by includ- point par Clancy. L'etude montre qu'on peut
ing a pile group over the central 1625% (by concevoir des dalles ottantes sur pieux qui
area) of the raft, with the pile group stiffness reduiront a un niveau negligeable le tassement
approximately equal to that of the raft alone. differentiel, en placant un groupe de pieux sous
The total pile capacity should be about 4070% la partie centrale de la dalle ottante (1625%
of the total applied load, depending on the pile de sa supercie), la rigidite du groupe de pieux
group area ratio and the Poisson's ratio of the etant approximativement egale a celle de la dalle
soil. The validity of the method is examined ottante seule. La capacite portante totale des
through the results of centrifuge model tests pieux devrait etre egale a 4070% de la charge
conducted by the authors. totale appliquee, en fonction du rapport de sec-
tion du groupe de pieux et du coefcient de
Poisson du sol. Les auteurs evaluent la validite
KEYWORDS: design; piles; rafts; settlement; soil/ de cette methode en examinant les resultats de
structure interaction. leurs essais sur modele centrifuge.
301
302 HORIKOSHI AND RANDOLPH
Rigid raft
Flexible raft
Contact pressure
Pile group
(a) (b)
Fig. 1. Principle of settlement-reducing piles: (a) rigid raft; (b) exible raft with
small central pile group
OPTIMUM DESIGN OF PILED RAFTS 303
therefore necessary to assume a representative soil validity of this approach for a wide variety of
modulus for piled rafts in non-homogeneous soils. conditions. In this paper, the equivalent-pier mod-
For piled rafts designed on the philosophy out- ulus, Eeq , and the equivalent-pier diameter, d eq , are
lined here, the average settlements may still be calculated using:
moderately large. Therefore, non-linear response of Atp
the piles may be an important aspect of the analy- Eeq Es (Ep Es ) (3)
sis. To cope with this, HyPR has the capability of Ag
q
taking non-linear pile behaviour into account. A
d eq 2 Ag = (4)
simple load cut-off is performed for the load-
transfer spring for the elements where pilesoil where Ep is the Young's modulus of the piles, Es
slip is computed to occur, using an incremental is the Young's modulus of the soil layer, Atp is the
elastic analysis. This mechanism is also implemen- total cross-sectional area of the piles in the group
ted for the spring at the pile base. and Ag is the gross cross-sectional area of the pile
group. The suitability of assuming an equivalent
pier can be categorized by the factor R, which was
NON-DIMENSIONAL PARAMETERS FOR proposed by Randolph & Clancy (1993):
PARAMETRIC STUDY r
It is of particular importance to establish a set ns
R (5)
of non-dimensional parameters to provide a frame- Lp
work for optimum design. In this paper, the follow- where n is the number of piles in a group, s is the
ing raftsoil stiffness ratio is introduced in order pile spacing and Lp is the pile length. It was
to express the relative exibility of rectangular suggested that the equivalent-pier approach was
rafts: appropriate for pile groups with R less than 4.
3
: Er 1 2s B tr Furthermore, the following dimensionless values
K rs 5 57 (1) were found to be important:
Es 1 2r L L
where Er and Es are Young's modulus of the raft relative pile length
and the soil, r and s are the Poisson's ratio of the
raft and the soil, tr is the raft thickness, and L and Lp =a (or Lp =aeq for rectangular rafts) (6)
B are the length and the breadth of the raft pile groupraft area ratio
respectively. Although Horikoshi (1995) used
1:0 in his analyses of rectangular rafts, a agr Ag =Ar (7)
recent study by the authors has shown that 0:5 equivalent piersoil stiffness ratio
gives the same magnitude of differential settlement
normalized by the average settlement, regardless of Eeq =Es (8)
B=L. As such, it provides a more consistent deni- pile groupraft stiffness ratio
tion of raftsoil stiffness ratio, as is documented in
a forthcoming publication (Horikoshi & Randolph, K pr k p =k r (9)
1997). In the present paper, a value of 0:5 has where aeq is thep equivalent circular raft radius,
therefore been adopted. Equation (1) is essentially dened as aeq (BL=) for rectangular rafts; Ar
compatible with the raftsoil stiffness ratio for is the raft area; and k p and k r are the overall
circular rafts that was suggested by Clancy (1993): stiffnesses of the pile group (or the equivalent pier)
3
Er 1 2s tr and the unpiled raft respectively. The pile group
K rc (2) raft area ratio denes the proportion of the central
Es 1 2r a
area over which piles are installed.
where a is the circular raft radius. If the raft area A measure of compressibility for the equivalent
and the thickness are the same for the circular and pier in a homogeneous soil can be expressed using
square rafts, the corresponding stiffness ratios K rc the expression introduced by Randolph & Wroth
and Krs are also the same. (1978) and Randolph (1994):
It was also found that the concept of the so- s
called `equivalent pier' was very useful in design- 2Gs 2Lp
ing piled rafts. The equivalent-pier method, where- Lp
Eeq d eq
by a group of piles is replaced by an equivalent
single, stubby pier, was introduced by Poulos & where
Davis (1980). Randolph (1994) suggested that once
ln [5 5(1 s )Lp =d eq ] (10)
the piles had been replaced by a pier of the same
length, the approximate solution for a single pile Here is a measure of the radius of inuence of
could be used to estimate the stiffness of the the pier. The compressibility Lp is a combination
equivalent pier. Horikoshi (1995) has explored the of the aspect ratio Lp =d eq and the square root of
304 HORIKOSHI AND RANDOLPH
the stiffness ratio Eeq =Gs . The larger the value of The differential settlement for square rafts is
Lp , the more compressible is the pier. dened as the difference between the central and
Among the non-dimensional parameters, it was the mid-side settlements as discussed by Horikoshi
found that the pile groupraft stiffness ratio K pr & Randolph (1997). This differential settlement is
was critical in designing optimum piled rafts. The normalized by the average settlement of the raft
magnitude of the differential settlement is largely (alone) with the same raftsoil stiffness ratio. The
governed by this parameter, and optimum results normalized differential settlement is dened as
correspond to a ratio close to unity. In this paper, w in the gures presented below.
an equivalent pier was assumed in calculating the
stiffness of the pile group, k p , while independent
HyPR analyses were conducted to calculate the raft PARAMETRIC STUDY OF SQUARE PILED RAFTS
stiffness. (Note that it is also possible to estimate An extensive parametric study of piled rafts was
k r assuming a fully rigid or a fully exible raft, performed by Horikoshi (1995). In the present
e.g. Giroud (1968).) paper, some of the results for square piled rafts are
The overall stiffness of a piled raft, k pr , can be presented. In the study, a group of only nine piles
estimated through the interaction factor between was modelled, beneath a exible raft with K rs
the pier and the raft, rp , which was proposed by 0:01 to 10. For practical applications of the new
Randolph (1983) as design philosphy, it will generally be necessary to
ln (rr =rp ) model a larger number of piles. In that case, the
rp 1 (11) equivalent-pier concept may be useful, replacing
ln (rm =rp ) the full pile group by a set of equivalent piers
with the piled-raft stiffness then expressed as (Randolph & Clancy, 1994).
The raft meshes used for the analyses are shown
k p k r (1 2rp ) in Fig. 3 for a quarter raft. The raft consisted of
k pr (12)
1 (k r =k p )rp 2 49 plate-bending elements, and each pile was
where rr denotes the radius of the raft, rp is the divided into 15 rod elements. Uniform vertical
pile radius and rm is the maximum radius of inu-
ence of an individual pile. The accuracy of the
estimated overall stiffness was checked with the
HyPR results.
In the design process, the critical decisions in
regard to the pile support are (a) pile geometry
(plan area of the pile group, length of the piles in
relation to the raft geometry) and (b) pile capacity
in relation to the local and total loading applied to (a) (b)
Pile
s
Table 1. Parameters used for basic parametric study of square piled rafts: basic parameters
Pile Square raft Soil
Pile number, n 9 Length, L 133 m Depth, h Deep
Diameter, d p 03 m Young's modulus, Er 35 GPa Young's modulus, Es 35 MPa
Young's modulus, Ep 35 GPa Poisson's ratio, r 016 Poisson's ratio, s 05
306 HORIKOSHI AND RANDOLPH
0.3
Raft
0.2 0.8
0.6
0.1
w * 0.5
0.4
0
s /b 5 0.3
20.1
20.2
0 0.2 0.4 0.6 0.8 1
1/P *t
(a)
0.3
Raft
0.2
0.8
0.1 0.6
w *
0.5
0 0.4
20.1 s /b 5 0.3
20.2
0 0.2 0.4 0.6 0.8 1
1/P *t
(b)
0.3
Raft 0.8
0.2
0.1
0.6
w *
0 0.5
0.4
20.1
s /b 5 0.3
20.2
0 0.2 0.4 0.6 0.8 1
1/P *t
(c)
0.2
0.1
Lp/aeq 5 3
0
0 0.2 0.4 0.6 0.8 1
1/P *t
Fig. 5. Effects of pile spacing on pile load (Krs 01) ( pgt proportion
of load carried by piles)
0.8 s /b
0.3 Lp/aeq 5 3
0.4
0.6
w/(w at m 5 1)
0.5
0.6
0.4
0.2
20.2
0 0.2 0.4 0.6 0.8 1
m
(a)
s /b
0.8 0.3 Lp/aeq 5 3
0.4
0.5
0.6 0.6
P g*
0.4
0.2
0
0 0.2 0 .4 0.6 0.8 1
m
(b)
in Fig. 6(b). Considering that the combination optimized by adjusting the pile radius from 005 m
of Lp =aeq 3, s=b 0:4 (i.e. agr 0:16) and to 048 m but keeping the same equivalent pier
K rs 0:1 resulted in very small differential settle- raft stiffness ratio Kpr as for the basic case
ment after some mobilization of pile capacity (Fig. (Ep =Es 1000). Note that in this situation, the
4(b)), then Pg 0:4 and thus Pt 2:5 (equation equivalent pier modulus Eeq and diameter d eq are
(14)) appears to form a set of optimum conditions kept constant, as are the number and length of the
for m 0:8. When Pg 0:4 and Pt 2:5, the piles (n 9, Lp =aeq 3). The total pile capacities
corresponding pgt (proportion of total load to be were no longer the same, since the pile radius was
carried by piles) is calculated as 032 using equa- varied. The value of Gs =s was kept constant
tion (15). ( 500) in all the cases.
On the basis of the above results, the case of The effect on the differential settlement of vary-
K rs 0:1, s=b 0:4 and Lp =aeq 3 was chosen ing Ep =Es is shown in Fig. 8. The gure shows
as a base, from which the effect of other para- that the calculated behaviour is very similar regard-
meters on piled-raft performance could be exam- less of the different Ep =Es values, and that all the
ined. The raftpier stiffness ratio was calculated as cases give very small differential settlement when
K pr 1:02 for these conditions, meaning that the 1=Pt is in the range 03 to 04.
pile group and the unpiled raft have similar stiff-
ness.
Effect of Poisson's ratio of the soil
The effect of the Poisson's ratio of the soil was
Effect of soil depth examined, maintaining a constant equivalent pier
The effect of soil depth on differential settlement raft stiffness ratio K pr 1:02, by adjusting the pile
was examined by varying the soil depth from in- radius between limits of 007 m and 015 m. The
nitely deep to a nite depth of h=Lp 2, 4, 7 and raft thickness was also changed in the range
10, where h is the thickness of the soil layer. The 031 m to 038 m to keep the raftsoil stiffness
results are compared in Fig. 7. Horikoshi (1995) ratio K rs constant at 01. The pile length was
also carried out simple elastic analyses for the unchanged from the basic case, and the rigidity
centrifuge model described below, by using the code index Gs =s was kept constant at 500.
HyPR. In the centrifuge model, the soil depth was The calculated results are shown in Fig. 9. The
25 m at prototype scale which compared with the gure shows that the variations of differential
raft radius of 7 m. The results of the simple elastic settlement are essentially the same, irrespective of
analyses showed that the differential settlement of the Poisson's ratio of the soil, although the differ-
the piled raft on a soil layer with a thickness of ential settlements show a tendency to increase as
25 m was essentially the same as that for an in- Poisson's ratio reduces. The non-dimensional para-
nitely deep soil layer, even though the average meters Pg and 1=Pt are compared in Fig. 10.
settlement of the raft was smaller by about 40%. When the Poisson's ratio of the soil is small, the
corresponding Pg value for m 0:8 is slightly
lower (between 03 and 04) than the value found
Effect of pile compressibility for s 0:5. Therefore the appropriate Pt value
The pile stiffness ratio Ep =Es was varied from for smaller Poisson's ratio should be marginally
100 to 10 000. In the analyses, pile support was higher (between 20 and 25).
0.2
0.15
h/Lp
Lp/aeq 5 3 2
0.1 4
s /b 5 0.4 7
10
w *
0.05 `
20.05
20.1
0 0.2 0.4 0 .6 0.8 1
1/P *t
Ep/Es
0.15
Lp/aeq 5 3
100
500
0.1 1000
s /b 5 0.4
3000
w *
0.05 10 000
20.05
20.1
0 0 .2 0.4 0.6 0.8 1
1/P *t
0.2
0.15
s
Lp/aeq 5 3
0.1
0.1 s /b 5 0.4 0.3
0.5
w *
0.05
20.05
20.1
0 0.2 0.4 0 .6 0.8 1
1/P *t
0.6
0.5 s
0.1
0.4 0.3
0.5
0.3
P g*
0.2
0.1
0
0 0.2 0.4 0.6 0 .8 1
m
(a)
0.6
0.5 s
0.1
0.4 0.3
0.5
1/P t*
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
m
(b)
0.00
This is because the corresponding K pr values were
derived from fully elastic analyses. Slightly higher
20.10 Kpr values (by approximately 5%) than those
shown in Fig. 14(b) will give minimum differential
20.20
settlement after some mobilization of pile capacity.
1.5 2 2.5 3 3.5 Fig. 15(b) shows the maximum positive and nega-
Lp/aeq tive bending moments during the incremental cal-
culations, showing that a pile spacing of s=b 0:4
Fig. 11. Effects of pile length on differential settlement to 05 gives the minimum bending moments in the
(fully elastic solutions) raft in absolute terms. Here M denotes the bend-
OPTIMUM DESIGN OF PILED RAFTS 311
0.2
0.15 Lp/aeq
1.6
0.1 3.0
w *
0.05
20.05
20.1
0 0.2 0.4 0 .6 0.8 1
1/P *t
Lp/aeq
0.8
1.6
3.0
0.6
P g*
0.4
0.2
0
0 0.2 0.4 0.6 0.8 1
m
0.3 2.0
0.1
w *
1.0
Kpr
Lp/aeq 5 3
0.0
0.5
20.1
Fully elastic analyses
20.2 0.0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0 .1 0 .2 0 .3 0.4
s /b Area ratio agr
(a) (b)
Fig. 14. Effects of pile spacing on differential settlement (K rs 01): (a) normalized differential settlement; (b)
optimum Kpr from fully elastic analyses
ing moment per unit length normalized by the total tially independent of the pile spacing, compared
applied load Pt . with the variation of 1=Pt . When m 0:8 is
The relationships between the non-dimensional assumed, the corresponding Pg value is about 05
load ratios, Pg and Pt , and m are shown in Fig. 16. for s=b 0:4 to 06, which is slightly higher than
The variation of Pg with m appears to be essen- the value found in the previous parametric study
312 HORIKOSHI AND RANDOLPH
0.3
Raft s /b
0.2
0.3
Lp/aeq 5 3 0.4
0.5
w *
0.1
0.6
0.0
20.1
0.0 0.2 0.4 0.6 0.8 1.0
1/P *t
(a)
0.006
Raft
0.004
Max.
0.002
0.000
M*
20.002
20.004 Min.
20.006
0.3 0.4 0.5 0.6
s /b
(b)
Fig. 15. Effects of pile spacing on piled rafts (K rs 01) (Kpr taken
from Fig. 14(b)): (a) normalized differential settlement; (b)
maximum bending moment of raft
described earlier (cf. Pg 0:4). It is considered Horikoshi & Randolph (1996). A concrete circular
that a slightly higher K pr value will give Pg closer raft with a diameter of 14 m at prototype scale was
to 04. When Pg is assumed to be 04, the corre- modelled. Note that the prototype scale is used in
sponding Pt value ranges from 15 to 25 for agr the following gures and calculations. The soil
varying from 016 to 025, from equation (14). prole in the centrifuge model was approximated
These values appear to be the optimum design by su 33 1:2z kPa after Horikoshi (1995),
conditions for piled raft foundations, implying a where z is the depth below the foundation in
total pile capacity which is 4060% of the total raft metres. By taking the soil strength at a depth of
load. The pile group extends over the central 16 one radius in the soil (i.e. 414 kPa), the ultimate
25% of the raft area, and, over that area, the total bearing capacity of the raft alone can be estimated
pile capacity is some 25 times the applied loading. as about 250 kPa, which corresponds to an ultimate
During the parametric study, it was also found load on the raft of 38 MN. A design load of
that as the area of the raft increases, and hence the 12 MN was then assumed by taking a safety factor
raftsoil stiffness ratio (K rs , see equation (2)) de- of 32 against bearing failure of the raft alone. This
creases, it is necessary to increase the area ratio in load corresponds to a uniform pressure of 78 kPa
order to avoid excessive settlement just beyond the on the raft.
pile-supported area. In estimating the response of the raft alone, a
uniform soil modulus of Gs 6 MPa has been
adopted, corresponding to Gs 145su and taking
CASE STUDY the modulus at depth of one radius. The Poisson's
The validity of the method presented here was ratio of the soil was assumed to be s 0:4 after
examined through the centrifuge tests reported by Horikoshi (1995). The overall stiffness of the raft
OPTIMUM DESIGN OF PILED RAFTS 313
1.0
0.8
s /b
0.3
0.6 0.4
0.5
P g* 0.4 0.6
0.2
0.0
0 0.2 0.4 0.6 0.8 1
m
(a)
1.0
0.8
0.6
1/P *t
s /b
0.3
0.4
0.4
0.5
0.2 0.6
0.0
0 0.2 0.4 0.6 0.8 1
m
(b)
method of Randolph & Wroth (1978) was then The computed non-dimensional parameters are
used with equation (10), allowing for the nite compared with the experimental results in Fig.
depth of soil by adjusting the base shear modulus 18(a) and (b) for pgt (proportion of total load
Gb (Randolph, 1989): carried by piles) and m (fractional mobilization of
Gl pile capacity) respectively. Note that the centrifuge
Gb (17) results are taken from `Test 3' described by
(h=Lp 1)
Horikoshi & Randolph (1996), and are plotted after
where h is the depth of the soil layer (valid only the end of the reconsolidation in ight with an
for 1 , h=Lp < 2). The equivalent-pier analysis initial load of 63 MN. In the test, loading and
gives the overall stiffness of the pile group as unloading cycles were applied as shown in the
K p 395 MN=m. The equivalent pierraft stiff- gures. The calculated results give non-dimen-
ness ratio K pr K p =K r is therefore calculated as sional parameters of pgt 0:23 (piles carrying
115. The overall stiffness of the piled raft is 23% of total applied load) and m 0:77 (piles
estimated as K pr 470 MN=m from equations (11) loaded to 77% of their ultimate capacity) under the
and (12), with the interaction factor rp 0:62. design load of 12 MN, indicating that pile capacity
The average settlement of the piled raft under the was effectively mobilized.
design load is therefore 26 mm. Figure 19 shows the measured average and
The ultimate shaft friction along the pile length differential settlements, compared with the predic-
was assumed uniform to maintain consistency with tions made using HyPR. Note that the average
the assumed homogeneous soil condition. The total settlement in this gure denotes the average of the
shaft friction and the base capacity were assumed central and edge settlements, to maintain consis-
to be 350 kN and 36 kN respectively for each pile, tency with the measured centrifuge results. The
after Horikoshi & Randolph (1996). Thus the total initial points of the experimental loadsettlement
single-pile capacity is estimated as qp 386 kN. curves have been constrained to fall on the curves
The non-dimensional parameters under the design predicted using HyPR. (This is necessary since the
conditions are then calculated as Pg pAg = initial displacements cannot be separated out from
nqp 0:56 and 1=Pt 0:29. The higher value of background reconsolidation of the soil, as dis-
Pg and the lower value of 1=Pt , compared with cussed by Horikoshi & Randolph (1996).) The
the optimum values found in the previous section, gure shows that there is good agreement between
suggest that the design load may be slightly higher the trends of the measured and computed results,
than optimum, and that a lower design load in the although the non-linearity of the experimental total
region of 9 MN might result in zero differential settlement curve is more pronounced than that of
settlement for the assumed conditions. the HyPR prediction. This may be attributed to the
Non-linear analysis of the piled raft was con- assumed elastic response of the raft in HyPR,
ducted by using HyPR with the soil depth of 25 m which only models non-linearity due to pilesoil
taken into account. Homogeneous soil was assumed slip. The calculated and experimental differential
in the HyPR analysis by taking the modulus at a settlements under the design load of 12 MN were
depth of two-thirds of the pile length in the soil. 4 mm and 3 mm, which may be considered negli-
The analytical parameters are summarized in gible from a practical point of view. The non-linear
Tables 3 and 4. HyPR analysis gives a true average settlement of
Table 3. Parameters used for analysis of centrifuge model: soil and pile properties
Soil properties Pile properties
Soil modulus, Gs : MPa 65 Pile length, Lp : m 150
Poisson's ratio of soil, s 04 Young's modulus, Ep : GPa 40
Soil depth, h: m 25 Total shaft capacity per pile: kN 350
Pile diameter, d p : m 0315 Base capacity per pile, qd : kN 36
Table 4. Parameters used for analysis of centrifuge model: properties of raft and
pile cap
Raft properties Properties of small pile cap
Radius, a: m 70
Length of square cap, Lc : m 089
Thickness, tr : m 0045 Thickness, tc : m 014
Young's modulus, Er : GPa 40 Young's modulus, Ec : GPa 40
Poisson's ratio, r 016 Poisson's ratio, c 016
A square-shaped pile cap with the same area as the circular cap model was assumed.
OPTIMUM DESIGN OF PILED RAFTS 315
0.5
pgt
0.3 Design load
0.2
0.1
0
0 5 10 15 20 25
Total applied load: MN
(a)
0.6
m
0.4
0.2
Design load
0
0 5 10 15 20 25
Total applied load: MN
(b)
290 mm when taken over the full area of the raft while that of the centrally piled raft was less than
(rather than taking the mean of edge and central 10 mm. The differential settlement of the piled raft
displacements); this is slightly greater than the is thus much smaller than that of the unpiled raft,
estimate of 26 mm obtained from the simple ap- in spite of the fact that the average settlement was
proach discussed earlier. The difference may be essentially similar.
attributed primarily to the effects of the non-linear
pile behaviour.
This example shows that central pile support is CONCLUSIONS
more effective in minimizing different settlements, An extensive parametric study of piled-raft be-
even though the total pile capacity is only about haviour has been performed to establish a frame-
10% of that of the raft. Differential settlements for work for the optimum design in terms of
the unpiled rafts modelled by Horikoshi & Ran- differential settlement. Although the cases consid-
dolph (1996) were about 40 mm after the end of ered in this paper are still limited, the following
the all loading steps (95 kPa applied loading), guidelines for optimum design are proposed.
316 HORIKOSHI AND RANDOLPH
Total applied load: MN
0 5 10 15 20 25
210
Design load
0
10
20
Settlement: mm
30
40
50
60 Ave., HyPR
70 Dif., HyPR
Ave., Centrifuge
80
Dif., Centrifuge
90
Fig. 19. Average and differential settlements of centrifuge model pile raft