Professional Documents
Culture Documents
Abstract: This paper presents a plastic limit analysis of the lateral load capacity of suction caissons in an anisotropic, purely cohesive
Downloaded from ascelibrary.org by YILDIZ TEKNIK UNIVERSITESI on 03/05/21. Copyright ASCE. For personal use only; all rights reserved.
soil assuming conditions of rotational symmetry about the vertical or gravity axis. The formulation utilizes a form of the Hill yield
criterion that is modified to allow for different soil strengths in triaxial compression and extension. Using this yield criterion, energy
dissipation relationships are formulated for continuous and discontinuous deformation fields. These dissipation relationships are then
applied to a postulated caisson failure mechanism comprising a wedge near the free soil surface 共mudline兲, a two-dimensional flow-around
failure at depth, and a hemispherical slip surface at the base of the rotating caisson. The plastic limit analysis predictions compared
favorably to predictions obtained from finite-element simulations employing a Hill yield criterion. For the range of anisotropic undrained
strength properties commonly reported for normally K 0 -consolidated clays, parametric studies indicate that suction caisson horizontal load
capacities predicted using a conventional approach 共a von Mises yield surface fitted to the soil simple shear strength兲 will differ from
anisotropic predictions by less than 10%.
DOI: 10.1061/共ASCE兲1532-3641共2003兲3:2共225兲
CE Database subject headings: Soil suction; Cohesive soils; Plastic analysis; Anchors; Offshore structures.
f ⫽J 1/2⫺k⫽0 (1a)
where
J⫽a 1 共 z ⫺ x 兲 2 ⫹a 2 共 y ⫺ z 兲 2 ⫹a 3 共 x ⫺ y 兲 2 ⫹a 4 2zx
冋冉 冊 册
2
1 S uSSV 2
a 3⫽ ⫺2a 1 (4b) Using a soil mechanics sign convention, positive and negative
4 S uPM
values of Ė TX correspond to triaxial compression and extension,
a 4 ⫽1 (4c) respectively. In cases of discontinuous deformations across a slip
In an isotropic von Mises material, S uSSV /S uTX⫽2/) and surface, the strain rates are actually infinite. However, it will be
S uSSV /S uPM⫽1. Substitution of these values into Eq. 共4兲 again later shown in this paper that the relative magnitudes of the strain
yields a 1 ⫽a 3 ⫽1/6, and a 4 ⫽1. rate components can be evaluated; hence, the sign of Ė TX in Eq.
共5兲 can always be determined.
Rather than the modified Hill approach adopted in this paper,
Strength Variations due to Sense of Loading
other authors 共e.g., Davis and Christian 1971兲, utilize a formula-
Undrained shear strength can also vary according to the sense of tion in which the center of an ellipsoidal yield surface 共Fig. 3兲 is
loading for a given mode of shearing. This effect can be important set equal to (S uTC⫹S uTE)/2 to model strength differences in tri-
when considering triaxial compression versus extension, for ex- axial compression and extension. A comparison of the yield sur-
ample. In K 0 -consolidated soils, the strength in compression faces corresponding to differing assumptions is shown in Fig. 3
(S uTC) along the consolidation axis can be more than twice that in for a typical normally consolidated clay in which S uTC⫽1.33
extension (S uTE). The Hill model expressed by Eqs. 共1兲 and 共2兲 S uSSV and S uTE⫽0.67 S uSSV . Differences between the modified
cannot directly model strength differences according to the sense Hill surface and the Davis and Christian yield surface are rela-
of loading for a given shearing mode. However, it is possible to tively small. The chief difference is that, for the modified Hill
specify different yield surfaces according to whether triaxial com- approach, the maximum resistance in simple shear, xy f , is mo-
pression or extension is expected within a given element of soil. bilized under conditions of zero triaxial stress, ( y ⫺ x )⫽0,
Such a procedure is quite straightforward in upper bound plastic- while the Davis and Christian surface has the maximum simple
ity calculations, where velocities and strain rates are completely shearing resistance occurring when ( y ⫺ x )⫽(S uTC⫹S uTE)/2.
defined in advance of stress and energy dissipation calculations. Also shown in Fig. 3 is the von Mises yield surface as it is
This ‘‘modified Hill’’ formulation is used subsequently through- commonly utilized, i.e., when the second invariant of the devia-
out this paper. toric stress tensor at yield is assumed equal to the square of
For relatively simple collapse mechanisms, zones where tri- simple shear strength of the soil, S uSSV . As is evident in Fig. 3,
axial compression and extension occur can be identified by in- the von Mises surface significantly underestimates the triaxial
spection. Fig. 2 illustrates these zones for a rigidly rotating cais- compression strength and overestimates the triaxial extension
son for the typical case of the load attachment point being above strength.
the center of rotation. In this case, triaxial extension occurs in the Randolph 共2000兲 adopted the Davis and Christian 共1971兲 yield
surface failure wedge ahead of the caisson and on the backside of surface described previously in his study of the effects of strength
the failure surface at the base of the caisson. Triaxial compression anisotropy on bearing capacity of shallow foundations, penetrom-
occurs in the surface failure wedge behind the caisson and on the eter resistance, and lateral load capacity of suction caissons. An-
Continuous Deformation
Murff 共1978兲 invokes the normality rule with the yield relation-
⫹ 共 v ⬘ n z ⫺ v z⬘ n 兲 2 兴 ⫹
1
共 v ⬘ n ⫺ v ⬘ n r 兲 2
2 共 a 1 ⫹2a 3 兲 r 冎 1/2
(8a)
ship in Eq. 共2兲 to obtain the rate of energy dissipation per unit The rate of dissipation per unit surface area of discontinuity is
volume in a continuously deforming material then found to be
Ḋ⫽k 再 a1
共 a 1 ⫹2a 3 兲 2 冋冉 a3
˙ ⫺˙ x
a1 z 冊 冉
2
⫹ ˙ y ⫺
a3
˙
a1 z 冊册
2
Ḋ A ⫽ lim 共 tḊ 兲
t→0
(8b)
⫹
2
˙ 2
共 a 1 ⫹2a 3 兲 xy 冎 1/2
(6a)
Laterally Loaded Caisson Failure Mechanism
z0
n r ⫽⫺ (12a)
冑 z 0 ⫹ 共 r 0 ⫺R 兲 2
2
r 0 ⫺R
n z ⫽⫺ (12b)
冑 z 0 ⫹ 共 r 0 ⫺R 兲 2
2
n ⫽0 (12c)
Substitution of these velocity and unit vector terms into Eq. 共7兲
provides the rate of energy dissipation per unit surface area along
the discontinuity. Integrating over the area yields
冑 冉
冉 cz
冊
冊 冕
R ␣ 1⫺
r 0 ⫺R 2 z0 z0
冉 冊
Ė 2 ⫽2k v 0 1⫹ C1 ␣⫺1 dz
z0 0 z 共 r 0 ⫺R 兲
r 0⫺
Fig. 5. 共a兲 Failure mechanism for laterally loaded caisson; 共b兲 energy z0
dissipations for laterally loaded caisson (13)
where
冉 冊冎
nates (r, , z), corresponding to a virtual radial velocity v 0 at the 2 1/2
top center of the leading edge of a caisson of radius R 共Murff and a3 1 z0
⫹ n 2⫹ n ⫺ n
Hamilton 1993兲 共 a 1 ⫹2a 3 兲 2 r a 4 r r 0 ⫺R z
v r⫽ v 0 冉 冊冉 冊
R
r
␣
1⫺
cz
z0
cos (9a) Wedge-Caisson Interface
冉 冊
v z ⫽g 共 r, 兲 z⫺
cz 2
2z 0
⫹h 共 r, 兲 (9b)
To permit total or partial adhesion at the pile-soil interface, Eq.
共8b兲 may be modified as follows:
h 共 r, 兲 ⫽ v 0 z 0 cos 冉 冊 冉 冊再
R
r
␣
r 0 ⫺r
r 0 ⫺R
⫺
1
r 0 ⫺r
⫹ 冉 冊c
r 0 ⫺R
The relative velocity components across the soil-caisson inter-
face 共i.e., where r⫽R) are as follows:
⫹
1⫺␣
冋 冉 冊册 冎
c r 0 ⫺r v r⬘ ⫽0 (15a)
再 冉 冊 冋 冉 冊册
1⫺
r 2 r 0 ⫺r 共 ␣⫺1 兲 cz 2 c⫺1 1⫺␣ c
v z⬘ ⫽ v 0 cos z⫺ ⫹z 0 ⫹ 1⫺
R 2z 0 r 0 ⫺R R 2
冉 冊 冎
Internal Energy Dissipation in Wedge
cz R
The dissipation rate per unit volume within the wedge is com- ⫹ 1⫺ (15b)
z 0 z 0 /c⫺z
puted by differentiation of Eq. 共9兲 to obtain strain rates and sub-
stitution of these strain rate terms into Eq. 共6兲. The total dissipa-
tion rate is obtained by integrating over the volume of the wedge v ⬘ ⫽ v 0 L 1⫺ 冉 cz
z0 冊
sin (15c)
Ė 1 ⫽2 冕 冕
r⫽r 0
r⫽R
z⫽z 0 [(r 0 ⫺r)/(r 0 ⫺R)]
z⫽0
冕⫽0
⫽/2
Ḋrddzdr (10) Noting that the unit normal vectors along this slip surface are
n r ⫽1 and n z ⫽n ⫽0, substitution of Eq. 共15兲 into Eq. 共8兲 yields
0
/2 1 2
a4
vz ⫹
1
册
v 2 ddz
2 共 a 1 ⫹2a 3 兲
(16)
Flow-Around Zone
For the case of rotational strength anisotropy, the energy dissipa-
tion for flow-around (E 4 and E 5 ) is computed on exactly the
same lines as for the isotropic strength case, provided one recog-
nizes that the appropriate strength parameter is the pressuremeter
shear strength or simple shear strength in a horizontal plane,
S uPM⫽S uSSH . The dissipation rate along a unit length 共measured
Downloaded from ascelibrary.org by YILDIZ TEKNIK UNIVERSITESI on 03/05/21. Copyright ASCE. For personal use only; all rights reserved.
v ⫽˙ 冉 冊 冉
z0
c
⫺z ⫽ v 0 1⫺
cz
z0 冊 (17b)
Ė 4 ⫽ 冕z0
z 0 /c
2R N p S uPMv 0 1⫺ 冉 冊 cz
z0
dz (18)
Ė 5 ⫽ 冕 lf
z 0 /c
2R N p S uPMv 0 冉 冊 cz
z0
⫺1 dz (19)
Ė 4 ⫽ 冕 lf
z0
2R N p S uPMv 0 1⫺ 冉 cz
z0
dz冊 (20)
Ė 5 ⫽0 (21)
Tip Resistance
Dissipation over the caisson tip is idealized as a spherical seg-
ment fixed to the pile tip with a center at the point of rotation.
Using a spherical coordinate system 共, , 兲 with origin at the
intersection of the center of rotation of the mechanism and the Fig. 6. 共a兲 Failure mechanism at tip of caisson; 共b兲 spherical coordi-
centerline of the pile 共Fig. 6兲, the dissipation is integrated over the nate system
spherical surface, limited by its intersection with the pile wall.
The velocity components are transformed to Cartesian coordi-
nates for convenient implementation in Eqs. 共5兲 and 共6兲
v z⬘ ⫽⫺R 3
v0
cos (22c)
Ė 6 ⫽ 冕 冕⫽2
⫽0 ⫽0
冑
⫽sin⫺1 (R/ R 2 ⫹R 1 )
2 kR 32 v 0
共 z 0 /c 兲
A 1 C 2 sin dd
共 z 0 /c 兲 (24)
where ⫽tan⫺1关cos /(sin cos )兴; R3 where
⫽R 2 冑cos ⫹(sin sin )2; R 2 ⫽ 冑R 2 ⫹R 21 ; R 1 ⫽L f
再 冋冉 冊 冉 冊册
2
and 2 2
⫺ z 0 /c. a1 a3 a3 a3
C 2⫽ A 2⫹ A ⫹ A ⫹ A2
Unit normal vectors are similarly transformed 共 a 1 ⫹2a 3 兲 2 a1 3 a1 3 共 a 1 ⫹2a 3 兲 2 2
n x ⫽sin cos
n z ⫽cos
(23a)
(23b)
⫹
1
a4
关共 A 4 ⫺A 5 兲 2 ⫹A 26 兴 ⫹
1
冎 A2
2 共 a 1 ⫹2a 3 兲 7
1/2
冉
F v 0 1⫺
Li
⫽冊 兺i Ė i ⫺Ẇ g (25)
Downloaded from ascelibrary.org by YILDIZ TEKNIK UNIVERSITESI on 03/05/21. Copyright ASCE. For personal use only; all rights reserved.
冕 冕 冕
r⫽r 0 z⫽z 0 [(r 0 ⫺r)/(r 0 ⫺R)] ⫽/2 agreement between PLA and FEM solutions for shorter aspect
Ẇ g ⫽2 v z ␥ ⬘ rddzdr ratios 共Table 1, L f /D⫽10, 6, and 4兲 suggest that the latter effect
r⫽R z⫽0 ⫽0 is increasingly significant and the tendency for the errors to can-
(26) cel increases with decreasing caisson aspect ratio. However, for
where v z is given by Eq. 共9b兲. very short caissons, L f /D⫽2, the trend reverses and the FEM
The virtual velocity terms are cancelled and the lateral load F simulation overestimates the PLA capacity by nearly 20%. This
latter case, L f /D⫽2, suggests that the postulated failure mecha-
is minimized with respect to the four optimization parameters
nism used in the PLA analysis is probably less accurate for very
described earlier.
short caissons. More detailed comparisons of FEM and PLA so-
lutions for suction caisson load capacity in an isotropic soil are
presented by Aubeny et al. 共2001兲.
Comparison to Finite-Element Method Solutions For the anisotropic simulations, the following conditions were
selected for the test case: load application point at the mudline,
To evaluate the analytical procedure described previously, a series full adhesion between the soil and the caisson, and full develop-
of plastic limit analysis 共PLA兲 predictions of lateral load capacity ment of suction behind the caisson. A series of suction caissons of
was compared to finite-element method 共FEM兲 simulations. The varying aspect ratios (L f /D⫽2, 4, 6, and 10兲 were considered.
finite-element simulations were performed using the anisotropic Undrained strength properties typical of normally consolidated
yield option available in the computer code ABAQUS 共HKS clay 共to be discussed subsequently兲 were selected for the
2000兲. The material model features linearly elastic behavior be- analyses—that is, with the axis of symmetry aligned with vertical.
neath the yield surface, a Hill yield criterion 关Eq. 共1兲兴, an associ- All analyses assigned a pressuremeter shear strength equal to the
ated flow rule, and perfect plasticity following yield. The FEM simple shear strength, S uPM /S uSSV⫽1. Three cases were consid-
model used three-dimensional, eight-node, linear brick elements. ered for characterizing the triaxial shear strength:
Prior to presenting the FEM/PLA comparisons, the limitations • A triaxial shear strength typical of triaxial compression,
of both approaches should be recognized. Accurate estimates of S uTX /S uSSV⫽1.57, was assigned to all soil elements.
limit load capacity from FEM analyses are difficult to achieve, • A triaxial shear strength typical of triaxial extension,
particularly for a rotating body. The PLA formulation utilized in S uTX /S uSSV⫽0.73, was assigned to all of the soil.
this paper is actually an assemblage of three independent upper • A strength S uTC /S uSSV⫽1.57 was assigned to the zones of tri-
bound solutions and hence not a rigorous upper bound solution in axial compression, and S uTE /S uSSV⫽0.73 was assigned to
zones of triaxial extension depicted in Fig. 2.
itself. Therefore, neither solution can be considered as a bench-
It should be noted that the first two cases are not realistic
mark or ‘‘true’’ solution. Nevertheless, comparisons between the
conditions of anisotropy and that these runs were performed pri-
methods can provide useful insights into the limitations of each
marily for the purpose of evaluating the FEM and PLA numerical
method.
models.
A series of isotropic analyses were performed initially to Because the ABAQUS Hill yield model cannot differentiate
evaluate the FEM model. For an infinitely long translating caisson between triaxial compression and extension strengths, different
with full adhesion on the boundary, the FEM analysis predicts a material parameters were assigned beforehand to elements in an-
normalized load capacity F/S u LD⫽13.12, which exceeds the ticipated zones of compression and extension. In front of the cais-
Randolph and Houlsby 共1984兲 solution of 11.94 by 10%. This son, an extension zone was assumed to occur above the rotation
overestimate is due somewhat to computer storage capacity re- point with compression occurring below the rotation point. The
strictions on the degree of mesh refinement for a three- opposite was assumed to occur behind the caisson: compression
dimensional model. For a long finite length caisson, L f /D⫽10, above and extension below the rotation point. An estimate of the
with a load attachment point at the mudline, the FEM simulation depth to the rotation point was made based on PLA solutions.
also overestimates the PLA by 10%. However, for shorter cais- Comparisons between FEM and PLA solutions for anisotropic
sons, L f /D⫽4 and 6, the FEM and PLA solutions are in much soils are summarized in Table 2. The horizontal load capacity F is
closer agreement. This improved agreement is attributed to com- normalized to a dimensionless ratio using the simple shear
pensating errors associated with the FEM solution. On one hand, strength S uSSV as a reference:
simple shear, S uSSH . In such an analysis, the strength of the soil in made based on a conventional isotropic analysis with the von
a triaxial shearing mode 共compression or extension兲 will be Mises yield surface matched to the simple shear strength of the
S uTX (isotropic)⫽0.866 S uSSV . Expressing actual triaxial soil S uSSV . The anisotropic strength cases A through D are shown
strengths in terms of the triaxial strengths derived from an isotro- in dashed lines in Fig. 7; while the conventional isotropic analy-
pic von Mises yield criterion 共Table 4兲 shows that actual triaxial ses are indicated by solid lines.
compression strengths will typically exceed von Mises triaxial The predictions in Fig. 7 indicate the following:
strengths by about 20–50%. Actual extension strengths can range • Predictions of horizontal load capacity using an anisotropic
from about 40% less to 10% greater than von Mises triaxial strength model differ from conventional isotropic predictions
strengths. Hence, the von Mises criterion will tend to be quite by less than 10%.
conservative for shearing modes involving triaxial compression. • When no gap occurs behind the caisson, the effects of strength
It will sometimes, but not always, be unconservative for shearing anisotropy become quite small for caissons with aspect ratios
in triaxial extension. L f /D⬎6. This result was to be expected, because for longer
Much less data are available regarding the pressuremeter caissons the relative influence of the surface failure wedges
strength S uPM , particularly in view of the unreliability of strength diminishes; hence, the load capacity is increasingly dominated
estimates from the pressuremeter test due to pressuremeter instal- by the simple shear strength in the flow-around failure zone.
lation disturbance and finite membrane length effects described • When no gap occurs behind the caisson, the conventional iso-
by Aubeny et al. 共2000兲. However, a review of shear strength data tropic analysis tends to give somewhat conservative results,
for resedimented K 0 -consolidated Boston Blue Clay 共BBC兲 by except for the condition of a low triaxial extension strength
Whittle and Aubeny 共1993兲 included strength comparisons for the 共Case D兲, in which case the isotropic analysis is unconserva-
triaxial compression and extension, direct simple shear, and pres- tive by a very small amount.
suremeter shear modes for normally to moderately overconsoli-
dated conditions. The pressuremeter strengths were derived from
true triaxial test data from Wood 共1981兲 on normally consolidated
BBC, and from directional shear cell test data from O’Neill
共1985兲 on moderately overconsolidated (OCR⫽4 – 5) BBC. In
both instances the pressuremeter shear strength S uPM was compa-
rable to the direct simple shear strength S uSSV . Based on this
somewhat limited data set, the pressuremeter strength was taken
equal to the direct simple shear strength for all cases considered
in this parametric study.
wedge is dominated by the relatively low 共for Cases B and D兲 axial extension strength S uTE should be borne in mind. Hence, the
triaxial extension strength, the optimization routine tends to effects of anisotropy on suction caisson load capacity should con-
maximize the depth of the surface failure wedge until it ex- tinue to be considered, particularly until sufficient experimental
tends to nearly the full depth of the caisson. Hence, when a data exist to permit the assumption of equality between S uPM and
gap develops behind the caisson, the influence of the surface S uSSV to be made with greater confidence.
failure wedge does not decline with increasing aspect ratio as
much as it does for the no-gap case.
The parametric studies summarized herein apply to uniform Appendix. Anisotropic Material Parameters
soil strength conditions. Actual strength conditions often are more
appropriately characterized by a strength profile that increases If triaxial shear conditions ( x ⫽ y , xy ⫽ yz ⫽ zx ⫽0) are ap-
linearly with depth. Under these conditions, the effect of strength plied to Eq. 共2兲
anisotropy can normally be expected to be less than that corre-
sponding to uniform conditions. The reason for this expectation is k 2 ⫽2a 1 共 x ⫺ z 兲 2
that the effects of soil strength anisotropy are most significant If k is taken as S uSSV and, noting that ( x ⫺ z ) f ⫽2 S uTX , it
when resistance mobilized in the surface failure wedge comprises follows that
冉 冊
a significant portion of the total load capacity. Under conditions 2
where the soil strength profile increases with depth, this is not 1 S uSSV
a 1⫽ (28)
likely to be the case. Because the soil strength increases with 8 S uTX
depth, most of the caisson load capacity tends to be mobilized in Applying the conditions for undrained cylindrical cavity expan-
the flow-around zone beneath the surface failure wedge where, sion (2 z ⫽ x ⫹ y , xy ⫽ yz ⫽ zx ⫽0) to Eq. 共2兲 leads to
provided that the shearing resistance for loading in a horizontal
plane approximates that in a vertical plane, anisotropic effects are a1
k 2⫽ 共 x ⫺ y 兲 2f ⫹a 3 共 x ⫺ y 兲 2f
small. 2
Again, taking k as S uSSV and noting that ( x ⫺ y ) f ⫽2S uPM leads
to
冋冉 冊 册
Summary and Conclusions
1 S uSSV 2
a 3⫽ ⫺2a 1 (29)
A plastic limit formulation for horizontal load capacity for suction 4 S uPM
caissons in an anisotropic soil is presented. The formulation uti- Alternatively, one can consider the case of simple shear in a ver-
lizes a Hill yield criterion to characterize anisotropic strength be- tical plane. In this case, Eq. 共2兲 yields
havior, and the present formulation is restricted to conditions of
rotational symmetry about the vertical axis. Different yield sur- 2
S uSSV⫽2 共 a 1 ⫹2a 3 兲 S uSSH
2
faces are specified for the soil according to whether a soil element
Imposing a condition of rotational symmetry, S uSSV⫽S uPM , also
experiences triaxial compression or extension to realistically
leads to Eq. 共29兲.
simulate strength behavior for these loading conditions. Material
parameters characterizing the Hill yield surface are readily de-
rived from conventional geotechnical tests. The anisotropic yield
surface is implemented using dissipation formulations applicable Acknowledgments
to continuously deforming and discontinous failure mechanisms,
both of which are present in the failure mode postulated for a The writers would like to acknowledge the support of the Depart-
horizontally loaded caisson. The least upper bound is obtained by ment of the Interior Minerals Management Service 共Cooperative
optimizing four parameters using the procedure presented by Agreement No. 1435-01-99-CA-31003兲, the Offshore Technology
Murff and Hamilton 共1993兲. Research Center, and their colleagues at Texas A&M University
For the range of anisotropic undrained strength properties and the University of Texas.
commonly reported for normally K 0 -consolidated clays, paramet-
ric studies indicate that suction caisson horizontal load capacity
predicted using a conventional approach 共a von Mises yield sur- Notation
face fitted to the soil simple shear strength兲 will differ from an-
isotropic predictions by less than 10%. The isotropic analyses The following symbols are used in this paper:
tend to be conservative in situations where full suction is mobi- D ⫽ caisson diameter;
lized behind the caisson 共no gap兲. If a gap is assumed to occur Ḋ i ⫽ rate of energy dissipation per unit volume;