Vous êtes sur la page 1sur 17

Murff, J. D., Wagner, D. A. & Randolph, M. F. (1989). GCotechnique39, No.

2,213-229

Pipe penetration in cohesive soil


J. D. MURFF, *D. A. WAGNER and M. F. RANDOLPH

Penetration of a pipe into cohesive soil is an impor- La pknktration dun conduit dans un sol coherent
tant consideration in offshore pipeline engineering, reprbente un aspect important de la construction
especially as such penetration affects on-bottom des pipelines au large, surtout parce quune telle
stability of the pipeline. If the soil is described as a p&&ration affecte la stabiliti! du pipeline sur le
perfectly plastic cohesive material then the calcu- fond de la mer. Si le sol est dbcrit comme Ctant une
lation of the limit load at a given penetration mat&e coherente parfaitement plastique le calcul
reduces to a plane strain problem in plasticity du chargement limite pour une pktitration don&
theory. Upper and lower bound solutions to this se rkduit 1 un probkme de dkformation plane dans
penetration problem are presented. The maximum la th6orie de la plasticitk Larticle prbente des sol-
range between the hounds occurs at one radius pen- utions donnant des valeurs plus Clevbs et plus r&
etration. The difference between the upper and duites pour ce prohkme de plasticit& Lkcart
lower bounds varies from about ten per cent for the maximum entre ces limites correspond A un p&n&
rough pipe case to approximately 25% for the tration dun rayon. La diffkence entre les limites
smooth pipe case. Parametric studies demonstrate supkrieures et inf(trieures varie dentre environ 10%
the effect of embedment depth, pipesoil adhesion, pour un conduit a surface lisse et environ 25% pour
soil surface heave, and increasing soil strength on un conduit P surface rugueuse. Des etudes param&
the vertical limit load. The solutions presented are triques demontrent Ieffet exerck par la profondeur
shown to compare favorably with test data. denrobage, Iadhesion du conduit au sol, le soule-
vkment de la surface du sol et la r&&stance crois-
sante du sol sur le chargement limite vertical. Les
KEYWORDS: analysis; bearing capacity; clays; limit solutions p&e&es sont en trb bonne concordance
state design/analysis; offshore geotechnics; plasticity. avec les essais effect&s.

pipe-soiladhesion normal stress on pipe-soil interface


Acontact area of the pipe (in plan) mean normal stress
c soil shear strength shear stress on pipe-soil interface
ca average soil shear strength over depth rotation of major principal stress from
D x axis
CO strength at soil surface +I - A/2
Cl linear rate of strength increase with arcsin (1 - P/r,)
depth
D depth for averaging soil strength
F vertical collapse load The offshore pipeline designer attempts to mini-
P pipe penetration mize cost by specifying the least amount of pipe-
r radius of curvature of characteristics line weight coating that will yield a safe
r0 pipe radius on-bottom condition without burial. Pipe pen-
Vi tangential virtual velocity along char- etration is critical for lateral soil resistance which
acteristic in turn is critical to pipeline lateral stability
up virtual downward velocity of pipe (Lyons, 1973; Jones, 1976). Test data confirm that
W rate of plastic work the lateral resistance of pipes on cohesive soils
x3 Y Cartesian co-ordinates depends strongly on the pipe penetration (Lyons,
Yl r. - P, distance to soil surface from 1973; Karal, 1977; Gulhati, Venkatapparao &
origin (center of pipe) Varadarajan, 1978; Wantland, ONeill, Reese &
ale Kalajain, 1979; Wagner, Murff, Brennodden &
: arcsin (a/c) Sveggen, 1987). Thus an accurate prediction of
r angle of soil surface to the horizontal pipeline penetration in cohesive soils is important
e1,0,, 03 angles defining characteristics for efficient pipeline design.
Early attempts to predict the penetration of
Discussion on this Paper closes on 6 October 1989. pipelines into cohesive soil applied traditional
* Exxon Production Research Co., Houston. bearing capacity theory (Small, Tambuvello &

213
214 MURFF, WAGNER AND RANDOLPH

RIGID, CIRCULAR PIPE

Y
s RIGID, PERFECTLY PLASTIC SOIL 4

Fig. 1. Idealized pipe penetration problem

Piaseckyj, 1971; Ghazzaly & Lim, 1975; Audi- the upper bound approach, a virtual velocity field
bert, Lai & Bea, 1978) for flat indenters with that satisfies all of the kinematic constraints is
shallow embedment. These methods often used an postulated. Equating the virtual energy dissi-
equivalent bearing pressure on a footing width pation rate within the deforming soil (based on
equal to the chord length of the embedded pipe. the yield condition) to the virtual work rate done
Wantland, ONeill, Reese Kz Kalajain (1979) by external loads and body forces provides an
offered a slightly different approach by achieving equation that can be solved for the unknown load
equilibrium iteratively between an approximation factor. In this case the calculated load factor will
to the maximum bearing stress imposed by the be greater than or equal to the true value. In a
pipe and classical bearing capacity values. These case where the lower bound load factor equals
methods make severe simplifying assumptions on the upper bound factor, the computed value is
the pipe-soil geometry and the soil bearing necessarily the exact one.
capacity. The desired relationship between vertical force
Karal applied the upper bound method of and penetration is determined by associating it
perfect plasticity to develop a prediction of pipe with the relationship between collapse load and
penetration, idealizing the pipe as a rigid wedge pipeline embedment. For the most part the soil
indenter. The wedge indenter approximation is surface is assumed horizontal but some discussion
reasonable at negligible penetrations but the of the effect of a sloping sea-floor and of soil
error becomes significant with increasing embed- mounding that results from penetration is provid-
ment. ed. The relationship of vertical force against pipe
This Paper develops both upper and lower penetration offers important design information
bound solutions to the pipe penetration problem for pipeline lateral on-bottom stability consider-
in uniform strength cohesive soil. These solutions ations.
incorporate the circular pipe-soil interface
exactly. The soil is considered to obey rigid per-
fectly plastic yield behaviour. An extension of the THEORETICAL SOLUTIONS
solutions to non-homogeneous soil conditions is The problem of pipeline penetration shown in
also presented. Taken together these solutions Fig. 1 is idealized as a plane strain geometry in
provide an accurate prediction of pipe penetra- which a rigid, unyielding, cylindrical indenter is
tion in cohesive soil as demonstrated by the partially embedded (less than or equal to one
agreement with large scale tests and model data. radius) in a rigid, perfectly plastic half space. The
The upper and lower bound theorems of clas- half space material obeys the Tresca, or
sical plastivity provide a method for estimating maximum shear stress, yield criterion. It is desired
collapse loads (Chen, 1975). In the lower bound to determine the vertical force F, on the cylinder
approach, a stress distribution is postulated that necessary to cause collapse of the supporting
satisfies equilibrium, does not exceed the yield cri- material.
terion within the soil, and balances the applied There are several possible related approaches
load. This provides a means for estimating the that can be taken to solve the problem. For
unknown load or load scale factor which will be example, legitimate upper and lower bounds can
less than or equal to the true collapse value. In be developed by some enlightened guesswork of
PIPE PENETRATION IN COHESIVE SOIL 215

with the characteristic net. These solutions may


coincide, to yield the exact solution, or they may
differ, to provide only bounds. In the latter case
the bounds may often be close enough to be
useful for engineering purposes.

Lower bound solution


The equations of equilibrium and the yield
condition form a set of hyperbolic equations that
can be solved by the method of characteristics.
For the case of non-homogeneous soil strength,
these governing equations have been derived by
Davis & Booker (1973). Solution strategies to
these equations using finite difference methods
are well known (Houlsby & Wroth, 1982). Exact
solutions, however, must also satisfy kinematic
constraints. Solutions satisfying the equations of
equilibrium and the yield criterion are not neces-
sarily unique and can only be asserted to be
lower bounds.
Figure 2 shows the characteristic net developed
Fig. 2. Characteristic net for smooth pipe
by integrating the governing equations for a soil
of uniform shear strength c and a smooth pipe,
i.e. zero adhesion between the soil and pipe
respective velocity and stress fields. For the surface. It may be shown from the equations that
purpose of this Paper, however, a more formal the mean stress at the pipe is equal to the mean
approach is taken similar to that described by stress at the free boundary (u, = c) plus 2cA4
Randolph & Houlsby (1984). First a lower bound where A4 is the rotation of the major principal
is sought through the method of characteristics stress along a characteristic from the boundary to
solution of the governing equations (yield condi- the pipe. Assuming that the soil is at yield, the
tion and equilibrium). Then an upper bound is entire stress state at any point on the pipe bound-
sought with the use of a velocity field consistent ary may be determined.

Fig. 3. Characteristic net for rough pipe


216 MURFF, WAGNER AND RANDOLPH

Following Randolph and Houlsby, for a rough where the symbols are defined in Fig. 4.
pipe the limiting shear stress or adhesion a at the For shallow penetrations such that n/2 -
pipe soil interface is a = ~(0 < a < 1). The incli- A/2 < w the collapse load F, is simply
nation of a characteristic at the pipe surface is
then x/4 - A/2 where sin A = CL.The character-
istic solution for a rough pipe is shown in Fig. 3. Note that for level ground surface, the collapse
This solution incorporates the wedge proposed pressure F,/A (where A = 2r, cos o) approaches
by Randolph and Houlsby at the line of sym- the well known value 2 + A, as the embedment
metry. The wedge consists of straight character- becomes very small. For embedments of a full
istics which intersect the line of symmetry at 45. radius (w = 0) with a level ground surface the
The shear stress mobilized at the pipe-soil inter- smooth and rough solutions are respectively
face then varies from zero at the centre line to the
limiting value a at point F, shown on Fig. 3. smooth: F, = 8r,c (2)
For these fields the stress state at the pipe rough: F, = 2[1 + 2/J(2) + n/2]r,c
boundary depends only on the boundary condi-
tions at the soils free surface and the orientation = 10.80r,c (3)
of the characteristics at the pipe boundary. Addi-
These solutions are respectively 78 and 105% of
tionally, the orientation is dictated by the adhe-
the classical result for a strip footing.
sion. Thus a closed form solution for the
In order to establish the solution as a lower
consistent lower bound collapse load can be
bound, the stress field must be extended beyond
determined. This solution is developed in
the slip line field into the rigid region. To accom-
Appendix 1.
plish this extension, Randolph & Houlsby have
A general geometry of the pipe and surround-
adopted a scheme suggested by Shield (1954).
ing soil is shown in Fig. 4. The soil surface adjac-
Owing to the similarities between their stress
ent to the pipe lies at an angle w below the
fields and the one presented in this Paper, this
horizontal pipe diameter and the surface itself lies
solution can be extended in an almost identical
at an angle I to the horizontal. For penetrations
way. The only slight modification is the soil
such that n/2 - A/2 > w, the vertical collapse
wedge at the free surface which differs only
load F, is
modestly from Randolph and Houlsbys circular
fan. Both fields are amenable to direct extension
F, = 2r, c[cos (A + w) + 2 sin (A/2)
from the slip line field, however, and therefore the
+(l+A+w-22r)cosw solution is established as a rigorous lower bound.
For a material that is non-homogeneous the
+ 2 cos A/2 - sin o] (1) solutions are not so obvious and the finite differ-

Fig. 4. Geometry for lower bound solution


PIPE PENETRATION IN COHESIVE SOIL 217

ence approach is used to assess the stress state at involutes ABC, FGH, etc. have the same
the pipe interface. Extension of the stress field shape as the associated stress characteristics
into the rigid region is more difficult and depends defined in the lower bound solution.
on the details of the non-homogeneity. Where (4 FGHQK is a deforming shear region, The
this extension is not accomplished the character- virtual velocities along the curved character-
istic solution can be viewed as merely an approx- istics such as FGH are defined similarly to the
imate solution, not a rigorous lower bound. velocities in the fan region ABCHGF but with
Results for non-homogeneous soil conditions are a smaller radius of curvature and a tangential
presented in the Paper. virtual velocity gradient along lines such as
GK due to the decreasing radial distance
Upper bound solutions from the evolute. The strain rate changes sign
The mechanisms to be used for upper bound within this region which complicates the cal-
solutions are associated with characteristic stress culations.
fields. As will be demonstrated, however, the (4 CDIH is a circular fan shear zone centred at
virtual velocity field implied by the stress charac- Q, the pipe-soil-surface interface.
teristic net already described does not always give (4 HIQ is a deforming shear region of circular
the least upper bound. A similar field based on arc virtual velocity characteristics centred at
that suggested by Randolph & Houlsby for the Q. The tangential velocities along the circular
lateral capacity of a pile yields slightly better arc characteristics satisfy normal velocity
results for pipe penetrations approaching one compatibility across HQ; the virtual tangen-
radius. tial velocity at Q is not zero.
In discussion of the postulated velocity field (f) DEJI is a rigid sliding region whose virtual
consistent with the lower bound stress character- velocity satisfies normal velocity compatibility
istic net described, taking advantage of the sym- across DI.
metry, only one half of the deformation pattern (g) IJQ is a deforming shear zone with straight
need be considered, as depicted on Fig. 5. Seven virtual velocity characteristics whose velo-
separate regions define the virtual velocity field. cities along characteristics such as IJ satisfy
normal velocity compatibility across IQ.
(a) AFF is a rigid zone attached to the pipe
moving downward with the same virtual The convenient angles 19i, f12, O3 define the
velocity as the pipe. virtual velocity characteristic net (Fig. 5). The
(b) ABCHGF is similar to a fan shear zone with angle O1 locates a specific curved characteristic
the fan centre moving anticlockwise along the that is an involute in the region ABCHQKF.
evolute defined within the pipe by the These characteristics are extended as circular arcs
normals to the involute ABC; these curved in region CDIQH and as straight lines in region

w = ARCSIN(1 p/ro)
A = ARCSIN (ADHESION/SHEAR STRENGTH)
v,, = VIRTUAL VELOCITY OF PIPE

Fig. 5. Velocity field consistent with lower bound


218 MURFF, WAGNER AND RANDOLPH

DEJQI. The angles e2 and 8, locate specific determination of the upper bound collapse load.
straight velocity characteristics. For the postulated failure mechanism, consistent
The full extent of the virtual velocity field is with the lower bound stress characteristic solu-
defined by the angle w which depends on the pipe tion, the upper bound solution is
radius and penetration, the angle A which
depends on the ratio of pipe-soil adhesion to soil F
2 = sin [3 + n - A - 20 - sin A tan $1
cohesive strength, along with the three angles 20,
mentioned above (Fig. 5).
The calculation of the internal energy dissi-
pation requires determination of the velocities +coS(i)[(t-i-,)(,a,*+2
along the characteristics in all of the deforming
regions. These velocities all stem from the down-
ward virtual velocity of the rigid pipe u,,. The +i(g-i-w))+sinA]
velocity normal to the region boundary must be
consistent between regions. Tangential slip + sin A[tan $ cos w - sin w]
(tangential velocity discontinuities) along bound-
aries between regions is admissible. However, + 2 cos w[A + 20 - l]
normal velocity discontinuities between regions n/2-A/2 8,
would imply gapping within the soil and are + lsin e1 -(e, - O2
inadmissible. s lo I0
The tangential velocity along the initially + tan $) cos @,I de2 de, (5)
curved characteristics inside region ABCDE-
JIHGF is D, = u,/,/(2). However, the tangential The integral in the last term can be evaluated in
velocity along the initially curved characteristics closed form for certain special cases but for other
within the region FGHIJQK is v, = general cases must be evaluated numerically. This
v,, sin e,/cos $ where I(/ = n/4 - Af2. Therefore a difficulty is a result of the absolute value taken of
tangential velocity discontinuity exists across the integrand.
FGHIJ equal to AU, = o,[cos (A/2) set $ - 1/ This result, however, yields an upper bound
J(2)]. The relative velocity between the pipe and collapse load greater than the lower bound com-
the soil along FKQ is u,(cos 6i + sin 0i tan $). puted earlier. Hence this exercise does not estab-
The tangential velocity along the straight char- lish either solution to be the true solution. For
acteristics is zero since they terminate in the rigid the smooth pipe case (zero pipe-soil adhesion)
region. Therefore no relative slip exists along the with one radius pipe embedment, the upper
characteristics such as KGB, QHC and QID. bound collapse load is ll.llcr, compared with
The integrations of the internal energy dissi- 8cr, for the lower bound. For the case of the
pation require the radii of curvature of the curved rough pipe with one radius penetration, the upper
characteristics. Within the region FGHQK the bound is 11.84crc compared with 1080cr, for the
curved characteristics, which are involutes, have a lower bound.
radius of curvature of A second mechanism, suggested by Randolph
& Houlsby (1984), improves the agreement
r = r,[sin Ic, + (0, - e,) cos I)] (4) between the upper and lower bounds for smooth
noting that o < e2 < 0i. The radius of curvature pipes at deep penetration (near one radius). This
outside this region may be obtained by setting 0i virtual velocity field is similar to the field already
equal to its maximum value of (n/2 - A/2), and considered with the exception of the regions
adding on the radial distance beyond FGH. along the soil surface. Fig. 6 shows that the
The internal energy dissipation terms fall into wedge shaped deforming region IJQ and the rigid
two classes: work rate due to tangential slip region DEJI from the previous virtual velocity
along an interface and work rate due to shear field (Fig. 5) have been replaced by extensions of
deformations within a region. The work rate fan region CDIH and deforming region HIQ. For
along an interface is calculated by the integration convenience, the points I and D have been
along the interface of the product of the relative retained on Fig. 6 although the characteristic
tangential velocity across the interface and the QID no longer forms a boundary between dis-
shear strength. Similarly, the product of the shear tinct regions.
strength and the shear strain rate integrated over The similarities between the velocity fields
a deforming region represents the work rate. The defined in Figs 5 and 6 allow the velocities and
specific work rates for this mechanism are derived geometry for the Randolph-Houlsby mechanism
in Appendix 2. to be determined readily from the previous
Equating the internal energy dissipation to the mechanism. The expressions for the tangential
external work rate of the collapse load allows velocity along the characteristics, the tangential
PIPE PENETRATION IN COHESIVE SOIL 219

w = ARCSIN (1 - p/r,,)
A = ARCSIN (ADHESION/SHEAR STRENGTH)
v. = VIRTUAL VELOCITY OF PIPE

Fig. 6. Velocity field consistent with Randolph-Ho&by (1984) solution

velocity discontinuities along FGHIJ and at the failure load of 9.97cr, for the smooth pipe case
pipe wall as well as the radii of curvature of the with one radius embedment compared with
characteristics remain the same. The only differ- ll~llcr, for the previous mechanism and 8.0cr,
ence between the dissipation rate calculations for for the lower bound. The comparison of the
the two mechanisms is the limits of integration various solutions shown in Fig. 7 highlights the
associated with extending the fan region CDIH difference between the lower bound and the two
and the deforming region HIQ to the soil surface. upper bound solutions.
Using the Randolph-Houlsby (1984) mecha-
nism described in Fig. 6 the following upper
bound solution for the collapse load is deter- Observations on diference between upper and
mined. lower bounds
F At first sight, it is surprising that the collapse
-=sini[l+n-A-22w-tan$sinA] load estimated from the upper bound solution
2cr,
based on the consistent velocity field shown in
Fig. 5, does not agree with the result from the
+cos(i){(i-$--c0)[tan* lower bound solution. However, the discrepancy
may be attributed to the fact that the velocity
field is consistent only in overall appearance, and
+i(i--$--m)]+sinA} not in terms of the sign of the implied shear strain
rates.
+ sin A[cos w tan $ - sin w] Therefore, for the region of soil considered in
Fig. 5, the CLstress characteristics (positive, or
anticlockwise shear stress) in the corresponding
+coso[;+A+20]
lower bound solution are the straight lines
n/Z-A/Z 8,
normal to the bounding surface ABCDE. The
+ 1 sin 0i curved characteristics (straight only in DEJQI)
5 s are j3 characteristics, where the shear stress is
negative, implying clockwise shear.
- (ii - e2 Ptan II/) cos 0,l df?, dfJ, (6) In the upper bound solution, the velocity along
The integral in the expression cannot be evalu- the fi characteristics is uniform [at v, = q,,/(2)]
ated in general because of the absolute value in the region ABCDEJIHGF. In the part of that
operator in the integrand. This upper bound solu- region where the characteristics are curved, the
tion represents a slight improvement over the constant velocity implies clockwise (or negative)
previous upper bound solution for deep embed- shear strain rates, which is consistent with the
ments and a smooth pipe. This solution gives a lower bound solution. However, in region IJQ,
220 MURFF. WAGNER AND RANDOLPH

5.92

5.56
5.40

4.99

= 0

.2 .4 .6 .a 1 .o
P/r,

Fig. 7. Comparison of lower and upper bound solutions for smooth


and rough pipe

the velocity along the b characteristics is u, = a,, Two additional parameters should be considered
sin O&OS $, which increases with increasing when comparing the idealized solution with the
values of 0, (going from Q to IJ). The shear strain field conditions of actual pipelines: soil heave
rate is thus anticlockwise, or positive, which con- next to the pipe and varying soil strength with
flicts with the lower bound solution. In region depth. It is a straightforward matter to include
FGHIQK, the shear strain rate changes sign soil buoyancy effects and this will not be dis-
except in the case of a fully rough pipe, where the cussed.
shear strain rate along the curved characteristics The influence of embedment depth and adhe-
is negative throughout. For cases where o! < 1, sion can be examined directly in the form present-
there will therefore be conflict with the lower ed. The method of characteristics lower bound
bound solution in part of the region. solution provides a convenient vehicle for con-
The change in sign of the shear strain rates sidering the influence of soil heave and non-
within region FGHIQK is a matter that was homogeneity, although those features can also be
overlooked in the solution developed by Ran- incorporated into the upper bound. Soil heave
dolph & Houlsby (1984) for the limiting pressure can be accounted for approximately by using the
on a circular pile. If the upper bound solution is lower bound solution for a sloping soil surface as
modified to ensure positive dissipative work at all developed in Appendix 1. The numerical recur-
points within the region, the agreement between sion formulae of the method of characteristics can
the lower and upper bounds is lost except for the be used to include the effects of increasing soil
case of a rough pile. The maximum discrepancy strength with depth.
arises for the case of a smooth pile, where the
revised upper bound result is 9.1% greater than
the lower bound result. Embedment depth
The non-dimensional plots of failure load
against pipe penetration presented in Fig. 7
PARAMETRIC STUDIES WITH LOWER demonstrate the marked influence embedment
BOUND depth has on the vertical collapse load. For all of
For the solutions presented, two parameters the solutions, the collapse load is extremely sensi-
affect the nondimensional collapse load; the pipe tive to embedment at small penetrations and
embedment depth and the pipe-soil adhesion. gradually becomes less sensitive as the embed-
PIPE PENETRATION IN COHESIVE SOIL 221

6 EQUIVALENT
SOLUTION
5.40
5.14
4.84

4.0

Fv = 2cr, (2+8) J2

EQUIVALENT
SOLUTION
I I I I I I I I I
.l .2 .3 .4 .5 .6 .7 .8 .9 0
P/r,

Fig. 8. Effect of p&soil adhesion on nondimensional collapse


load against embedment, lower bound solution

ment approaches one pipe radius. At very shallow In the consistent upper bound solution, some of
penetrations the solution approaches the stan- the higher intermediate adhesion values were
dard bearing capacity solution (2 + a) where the found to provide results that are apparently
area is replaced by the chord length of the pipe at worse (higher) than the full adhesion solution.
the soil surface. This solution is shown in Fig. 8 This is a characteristic of this particular solution
and is a reasonable solution for a rough pipe up and is not seen, for example, in the Randolph-
to full penetration. The lower bound solutions at Houlsby solution (1984). Where this occurs, it is
full radius embedment are considerably less sensi- possible to use the mechanism for the full adhe-
tive to slight variations in penetration than the sion case but only account for dissipation at the
upper bounds. interface that results from partial adhesion. This
remains a valid upper bound and provides a
more realistic estimate of the adhesion effects.
Pipe-soil adhesion The effects of adhesion can be traced to the
For the design problem of immediate pipe pen- ensuing enlargement of the deforming zone. The
etration in clay under the pipes self weight, the nose cone effect at the pipe apex described in the
zero adhesion case for undrained conditions lower bound discussion, area AFF in Fig. 5,
seems most reasonable. Observations made exists only for non-zero adhesions. As the adhe-
during experiments show that typically several sion increases, the deforming zone enlarges and
hours are required for significant pipe-soil adhe- the interface ABCDE pushes deeper into the soil
sion to develop. However, for some applications as depicted in Fig. 9. This resultant shearing of
it may be reasonable to consider partial or full additional soil increases the collapse load.
adhesion between the pipe and the soil, for
example, additional penetration after consoli-
dation which might occur on filling the pipe with Soil surface heave
water before use. Significant increases in collapse It is clear that the volume of soil displaced by
load result from adhesion as shown in Fig. 7. The the penetrating pipe must heave up on either side
relative increase in the non-dimensional collapse of the pipe. The upper and lower bound theorems
load against embedment with increasing adhesion apply only for small displacement incremental
is depicted in Fig. 8 for the lower bound solution. collapse problems; therefore, for the purposes of
222 MURFF, WAGNER AND RANDOLPH

- SOIL
SURFACE

0 0.5 1 .o 1.5 2.0 2.5 : 3

Fig. 9. Deforming zone for full pipe-soil adhesion for embedment of one-half radius

this Paper, it is assumed that the displaced soil is valid only up to nominal penetrations of about
forms a wedge adjoining the pipe. The size of the 0.4r, since beyond this point the soil pipe contact
wedge and the slope angle r from the horizontal is above the midpoint of the pipe. An alternative
can be determined by trial and error by matching upper bound solution including the soil heave
(a) the extent of the deforming zone along the region as a deforming zone is expected to yield
original soil surface, (b) the soil volume displaced similar results. The compensating effects of the
by the embedded pipe, and (c) the slope angle I- sloping soil surface against increased embedment
that determines the volume of soil heave. This reduce the solution sensitivity to heave. However,
matching is illustrated in Fig. 10 for a smooth it should be realized that heave can have an effect
pipe with the total penetration and the nominal of lGlS% on the predicted collapse load.
penetration noted.
By applying this matching approximation, the
lower bound solution is used with the appropri- Increasing soil strength with depth
ate value of r to assess the effect of heave. Fig. 11 The recursion formulae for the method of char-
shows the plots of non-dimensional load against acteristics solution allow for inclusion of the effect
embedment when heave is either included or of variable soil strength on penetration resistance
neglected. This solution to account for soil heave as discussed by Davis & Booker (1973). However,

EXTENT OF VOLUME I = 2 X VOLUME II


DEFORMING ZONE
*
t-

DEFORMING ZONE LIMIT---/

Fig. 10. Approximation for effect of soil heave


PIPE PENETRATION IN COHESIVE SOIL 223

As the soil strength increases the collapse load


also increases. Fig. 12 is a plot of the non-
dimensional collapse load against rate of strength
increase for a smooth pipe at various embed-
ments. The change on the abscissa scale from
Giro/Co to C,/C,r, allows the plotting of the
entire range of strength parameters.
FV The deforming region for the increasing
2cr,
strength case differs considerably from that
observed in the homogeneous strength case. As
the strength increases with depth, the deforming
region becomes more localized and in general the
shear fan sector becomes curved, i.e. the charac-
teristics such as QHC in Fig. 5 are curved rather
than straight. Fig. 13 shows the effect that the
non-dimensional strength increase term of two
and infinity (zero strength at the surface) have on
Fig. 11. Effect of soil heave OII aon-dimeasional lower the deforming region for the smooth pipe case.
bouad collapse load against embedmeot for smooth pipe Being restricted to numerical approaches for
case
the increasing strength case adds a certain incon-
venience to the estimation of the collapse load
the solution has not been extended in general and
and requires the designer to use either a tabular
is therefore considered an approximate solution
or a graphical solution. It is clear that for a par-
rather than a rigorous lower bound. Here the
ticular pipe embedment and strength rate increase
case of a linearly increasing strength with depth is
there is some depth D over which the averaged
considered but othr continuous strength profiles
strength when used in the homogeneous solution
are readily included.
yields the same collapse load as for the more rig-
The strength profile for a linear function of
orous non-homogeneous characteristics calcu-
depth below the original soil surface is taken to
lation. The following expression for D was found
be
to yield a surprisingly robust estimate over a con-
c = co + C,(Y- Yd (7) siderable range of penetrations and strength rate
increases.
where co is the strength at the soil surface, ci is
the rate of strength increase with depth, y, = r. D/r, = P/r, + 0.075 (8)
- P and y is the depth beneath the centre of the
pipe. A linear increase such as this is often used to Using the nominal soil strength determined by
model the strength of many offshore clay de- averaging over the depth D produces reliable esti-
posits. mates of the collapse load estimated from the

2.45

I I 0.14
I
0 1.0 2.0
0.5 0.25 0.0

C1ro CO
CO Cl'0

Fig. 12. Effect of hear strength increase on lower bound collapse


load for smooth pipe case
224 MURFF, WAGNER AND RANDOLPH

0.5 IL
ACE

y/r0

1 .o

1.5
0 0.5 1 .o 1.5 2.0 2.5

Go

Fig. 13. Effect of increasing strength with depth on deforming zone


for embedment of one-half radius

characteristics solution. Table 1 compares the col- in Fig. 8, where the footing width is taken as the
lapse load given by the numerical solution that chord width of the pipe at the soil surface. The
uses the increasing strength with the lower bound total capacity is then normalized in the same
solution that uses the nominal strength in the manner as the other solutions in Table 1. These
homogeneous solution for the smooth pipe case. values fall both above and below the theoretical
The errors are small for a wide range of condi- predictions, a result of competing effects. A pipe
tions. Care should be exercised when applying in contact with stronger soil below the soil
this average strength method for very small pen- surface tends to make the Davis and Booker
etrations (P/r, -c 0.1) and very high rates of equivalent solution smaller, but a smooth pipe-
strength increase. soil interface has the opposite effect.
As well as the solutions described above, a
third column based on Davis & Bookers (1973) COMPARISONS WITH TEST DATA
solution for a surface footing is included. This uti- The upper and lower bound solutions present-
lizes the equivalent solution idealization as shown ed compare favorably with test data. Some of the

Table 1. Prediction error using equivalent homogeneous solution for increasing strength
with depth in lower bound solution

Embedment Linear Predicted Predicted Predicted force


Plr0 strength rate force force (Davis and
increase (empirical) (theoretical) Booker (12))
clrOIcO F F F
2r&, + clrO) 2r& + cIrO) 2r&, + clrO)
0.1 0.5 1.42 1.43 1.59
0.1 2.0 0.80 0.80 0.94
0.1 5.0 0.49 0.48 0.60
0.1 10.0 0.35 0.33 0.41
0.1 0.18 0.14 0.10
0.25 :5 2.11 2.11 244
0.25 2.0 1.30 1.31 1.48
0.25 5.0 0.89 0.88 0.99
0.25 10.0 0.70 0.68 0.71
0.25 0.48 0.43 0.22
0.5 :5 2.77 2.78 4.45
0.5 2.0 1.91 1.93 2.01
0.5 5.0 1.48 1.49 140
0.5 10.0 1.28 1.29 1.04
0.5 co 1.05 1.05 0.38
1.0 0.5 3.38 3.48 3.75
1.0 2.0 2.77 2.96 2.37
1.0 5.0 2.46 2.70 1.69
1.0 10.0 2.32 2.58 1.29
1.0 co 2.15 2.45 0.50
PIPE PENETRATION IN COHESIVE SOIL 225

CONSISTENT UPPER BOUND c_c- 5.56

5- . .
* .

UPPER BOUND

0 LYONS [ll
a KARAL 131
+ WANTLAND ETAL [51
. REF. 161

0 I I I I I I I I I I I I I I I I I I I

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
PENETRATION/RADIUS

Fig. 14. Nondimensional collapse load against pipe embedment

available test data on pipe penetration in clay are CONCLUSIONS


plotted with the non-dimensional collapse load The penetration of a pipe into cohesive soil on
against penetration curves for the upper and the ocean floor is an important design consider-
lower bounds on Fig. 14. The penetration report- ation for offshore pipelines. The problem in this
ed in the data results from the pipe self weight. It Paper is idealized as a rigid cylinder, partially
was allowed to take place relatively rapidly. embedded in an ideal, rigid plastic material. The
Therefore undrained conditions prevail and thus relationship between vertical force and penetra-
the use of undrained shear strength and zero tion is associated with the relationship between
pipe-soil adhesion is appropriate. the vertical collapse load and embedment.
The bulk of the data reported in Fig. 14 comes The method of characteristics is used to deter-
from a study on pipe-soil interaction described mine a solution to the idealized problem for
by Wagner, Murff, Brennodden & Sveggen (1987). homogeneous soil. The solution may be extended
The tests focused on large radii pipes, 0.25 m and into the rigid region and is therefore a rigorous
0.50 m, and on remoulded soft clay with strengths lower bound. Upper bound solutions, associated
in the range 0.7 kPa-1.6 kPa. Vane shear mea- with the derived characteristic net, and an exten-
surements, fall cone results and miniature cone sion of the solution by Randolph & Houlsby
penetration tests provided the information on soil (1984) have been developed. The associated veloc-
strength and variation with depth. The strength ity fields are found to be inconsistent with the
was found to increase modestly with depth and lower bound stress fields over the range of
the reported results include the average strength embedments and roughnesses, and therefore the
approximation. The results have also been adjust- solutions are not exact. The bounds, however, are
ed for the effects of soil buoyancy. reasonably close and should be useful for engin-
The resolution of shear strength measurements, eering purposes. The solutions have been
especially at the relatively low values measured, extended in an approximate way to include the
probably contributes most of the scatter in the effects of soil heave and nonhomogeneous
data. Much of the scatter evident in Fig. 14 could strength.
be reduced if minor but reasonable adjustments The force penetration relationships have been
are made in the strength used in the prediction. compared with available data and found to
No such adjustments have been included here. provide a favorable prediction. Although scatter
Considering the scatter in strength measure- exists, the magnitude and trends of the data are
ments and the nonunique nature of undrained correctly predicted.
shear strength itself, these results are believed to
be quite good. The comparison between the
upper and lower bound solutions in Fig. 14 APPENDIX 1. LOWER BOUND SOLUTION
includes a wide range of pipe sizes and soil The lower bound solution is obtained by considering
strengths which lend credibility to the validity two regions along the pipe-soil interface as shown in
and completeness of the non-dimensional solu- Fig. 4. This solution is valid for penetrations such that
tion. (x/2) - (A/2) > w.
226 MURFF, WAGNER AND RANDOLPH

The vertical force, after some simplification, is then


r/2
F,, = ~(2 + n - 2r) sin Or, dtJ (15)
s n,2 -%,z
which after further simplifying gives

F,,=r,c(2+n-2T)sinq (16)

Combining the solutions and accounting for symmetry


gives

A
F, = 2r, cos (A + o) + 2 sin 2
f
+ (1 + A + 20 - 2l-J cos w

1
Fig. 15. MOWS circle for interface stresses A
+ 2 cos T - 2 sin 0 (17)
Region 1, (42) - (A/2) z 9 z o
In region 1 the normal stress on the interface oi, as For shallow penetrations where o z (n/2) - (A/2)
shown in Fig. 15 and Fig. 4, is region 1 vanishes and the vertical force is

oi = o, + c cos A n/2
F, = 2 c(2 + II - 2r) sin e rO d0 (18)
= c + 2cA$ + c cos A I 03
= ~(1 + 28 + A - 2r) + c cos A (9) F, = 2r,c(2 + PI- 2r) cos w (19)

The shear stress on the interface ~~is

~~= c sin A (10)


The differential vertical force acting on the pipe bound- APPENDIX 2. UPPER BOUND CALCULATION FOR
VIRTUAL VELOCITY FIELD CONSISTENT WITH
ary is then
LOWER BOUND
dF, = (TVcos 0 + oi sin g)r, dtI (11) The upper bound calculation requires the rate of
virtual work of the external collapse load and the rate
and consequently of virtual internal energy dissipation. For a plane strain
n,Z-A,2 geometry, the rate of energy dissipation may be calcu-
F,, = [c sin A cos 0 lated by integrating the product of the maximum shear
s0 strain rate and the shear strength over the deforming
+ ~(1 + 20 + A - 2r + cos A) sin e]r, de region. At a velocity discontinuity along an interface
between two regions, this calculation reduces to an inte-
which can be integrated to give gration of the product of the virtual relative velocity Av,
and the shear strength c, over the interface length. The
F,,=r,c cos(A+w)-(n-2r)sin: energy dissipation rates for the virtual velocity field
[ associated with the lower bound solution are given here.
The interface and region designations refer to Fig. 5.
+ (1 + A + 20 - 2r) cos o

(12)
For interface AF

1
A
+ 2 cos T- 2 sin 0
A@=c$r,J2sin($)=u,cr,sin(~) (20)

Region 2, 0 > (n/2) - (A/2) For interface ABC


In region 2 the major principal stress is vertical as
shown by Randolph
stresses are therefore
& Houlsby (1984). The interface

oi = c + 2cAB + c cos 5-e>I


Aw=c$

where r is the radius of curvature


s n,z -A,2
o r dtJ,

of ABC. Note that


(21)

[2( ABC is an involute. Its evolute is a circle of radius


rO cos + which is concentric with the pipe.
=c+2 f - r - c C~S 28
( >
= ~(1 + n - 2r) - c cos 28 (13)
=i = c sin 28 (14)
PIPE PENETRATION IN COHESIVE SOIL 221

ThUS

(321
Aw=,cr,$[(;-$o)
Thus

x(~Zsin(~)+sin$)

+;(;-;-+os$] (23)

For interface CD
For interface IJ, length DE is the same as radius QI due
n,%+~,2+o to the (n/4) intersection angle of IJ at the free surface
Am = c $ n,4 r d0, (241 and the velocity discontinuity is the same as across
s FGH and HI.
where I is the radius of curvature of circular arc CD
centred at Q. AI+=~~cr{cos(;)-y](;-q-m) (34)

r=r{J2sin(~)+(5-$-~)costi] (251 For interface FKQ along the pipe

Thus AB = c sin A cn/2 -*,2


Av,ra d0, (351

A~=r,cr,$[,/2sin(~) where r,, is the pipe radius and the tangential velocity
discontinuity is
Au, = u,(cos 6, + sin 6, tan $) (361
+(;-;-cc)cos+](;+c~) (26)
Thus
For interface DE, note that length DE is the same as
radius QE due to the n/4 intersection angle at the free Al@ = crOvO sin A cos 4 - sin w
surface. 1 0
A
- sin z tan $ + cos 0 tan i/j
0
Within the deforming regions ABCHGF, CDIH and
(271 DEJI, the tangential velocity along the curved charac-
For interface FGH, noting that Au, = o,[cos(A/2) teristics is constant, u,,/J2, so the shear strain rate is
x set JI - l/J21

Al@=c c x,2 -*,2


Au, r dt?, (281
where r is the radius of curvature in each region.
where r is the radius of curvature of FGH. FGH is an Within region ABCHGF

sI
involute similar to ABC and the radius of curvature of x,2 -A,/2
FGH is the radius of curvature of ABC less the distance I C
Aw= 0 T dr dlI, (381
AF. CD ., rJ2
where the radial integration limits rl and r2 denote the
r=r,[sin++(i-$-@,)cos$] (291 radius of curvature along FGH and ABC, respectively.
Noting that rz - ri = r,,J2 sin A/2, the same as dis-
Thus tance AF.

A.=q,cr,,sin(t)(t-p-cc) (39)

Within region CDIH

AIb~~+A*+U~;$rdrdOa (40)
For interface HI, noting that the velocity discontinuity
across HI is identical to FGH where the radial integration limits ri and r2 denote the
++V2+o radii of curvature along HI and CD, respectively. Again
Aw=c AU,I dt?, (311 r2 - I, = r,, J2 sin A/2, so
I n/4
where r is the radius of curvature of circular arc HI AB=r,cr,sin(~)(q+w) (41)
centred at Q.
228 MURFF, WAGNER AND RANDOLPH

Within region DEJI, the radius of curvature is infinite; Thus


the velocity characteristics are straight lines so the n,* -A,,?
strain rate becomes zero and hence the region slides as A& = u,, crc (e, - w) cos 8, de,
a rigid body sw

Aw=O (42) =r,c,,[,in(t)+(54-0)


Within the deforming regions FGHQK, HIQ and IJQ,
the velocity along the straight characteristics emanating
from the pipe is zero. The tangential velocity along the Xcos(~)-cos(u]
initially curved characteristics is
The external virtual work rate of the collapse load is
0, = u0 sin ~&OS + (43)
AI@=F,u, (51)
The virtual shear strain rate in FGHQK, recalling that
the radius of curvature of the curved characteristics is Equating the internal dissipation rates and external
r = r,[sin + + (6, - 0,) cos $1, becomes work rates and multiplying by two for the symmetry
yields the following equation for the collapse load
fJg cos er u0 sin e1
jJ= - ~_- (44)
( r. co? I/J rcos * > 2 = sin 4 [3 + I[ - A - 2w - sin A tan $1
0
The unit differential area with FGHQK is
+cos(i)[(t-$w)(tan$+2
dA = r,, de, cos JI r de, (45)

Thus
+i(i-t-w))+sinA]

AB=
ss0
n/Z-A,2

0
0,
c)rr, cos JI de, dt?,
+ sin A [tan $ cos o - sin o]

s ssin
= 0 erg
e1 0
n,2-A/Z

w
0,

- (e, - e2 + tan +) cos e,l de, de,


(46) + 2 cos o[A + 20 - l]

+
ss
w
n,*-.4,2 8,

0
1sin e1 - (e, - e2 + tan $)

Since the integrand is not always non-negative within x cos ell de, de, (52)
the integration region, the absolute value must be used
where o = arcsin [l -(P/r,)], A = arcsin (a/c) and
in the integration. This is required to ensure that only
$ = (n/4) - (A/2). The integral in equation (52) can only
positive virtual work rate is allowed.
be integrated in closed form for special cases. Two par-
By similarity with the above calculation, within
ticular examples of interest are given here.
region HIQ, recalling that r = r,,(ei - w) cos $
Case A, zero adhesion A = 0, o 3 x/4, P < 0,2929r,
n,* -a,2 n,4+A,*+o
Al@= cjv de, r0 cos * de,
s0 I r/4 2 = 2 + A - 20 + 20 cos 0 - 2 sin w (53)
n,z -%,2 n/4+,%,2+ 0
1sin 8, (47)
= v0 crO Case B, full adhesion A = n/2, w < n/4, P 3 0,2929r,
s0 s n/4
- (e, - W) cos ell de, de, i 5= J2(3-w+;)
The integrand is always positive over the integration
region, thus
-3sinw+(5+2w-I)cosw (54)
A
2 cos 0 - 2 sin
0 ?
REFERENCES
Audibert, J. M. E., Lai, N. W. & Bea, R. G. (1978).
(48)
Design of pipelines-sea bottom loads and
restraints. Pipelines In Adverse Environments, Am.
Within the deforming region IJQ, a new set of axes with Sot. Civ. Engrs, 187-203.
origin at Q and y running along QID eases the calcu- Chen, W. F. (1975). Limit analysis and soil plasticity.
lations. The shear strain rate in IJQ stems from the Rotterdam: Elsevier.
tangential virtual velocity 0,. = v0 sin B,/cos $ and the Davis, E. H. & Booker, J. R. (1973). The effect of
relation y = rO(e, - 0) cos $. increasing strength with depth on the bearing capa-
Then city of clays. Gtotechnique 23, No. 4, 551-563.
Ghazzaly, 0. I. & Lim, S. T. (1975). Experimental inves-
tigation of pipeline stability in very soft clays. O$
(49)
shore Technology Con& Paper OTC 2277,315-326.
PIPE PENETRATION IN COHESIVE SOIL 229
Gulhati, S. K., Venkatapparao, G. & Varadarajan, A. Randolph, M. F. & Houlsby, G. T. (1984). The limiting
(1978). Positional stability of submarine pipeline. pressure on a circular pile loaded laterally in cohe-
Proceedings of Geocon India, Conf. on Geotechnical sive soil. Gtotechnique G, No. 4, 613-623.
enaineerina. New Delhi, 43&434. Shield. R. T. (1954). Plastic notential theorv and Prandtl
Ho&by, G. T: & Wroth, C. P. (1982). Direct solutions bearing capacity solution. J. Applidd Mechanics,
of plasticity problems in soils by the method of char- ASME, June, 193-194.
acteristics. Proc. Fourth Int. Conf. on Numerical Small, S. W., Tambuvello, R. D. & Piaseckyj, P. J.
Methods in Geomechanics 3, 1059-1072. (1971). Submarine pipeline support by marine sedi-
Jones, W. T. (1976). On-bottom pipeline stability in ments. Ofihore Technology ConjI, paper OTC 1357,
steady water currents. Offshore Technology_. Conf, 309-3 18.
paper OTC 2598,763777. Wagner, D. A., Murff, J. D., Brennodden, H. & Sveggen,
Karal. K. (1977). Lateral stabilitv of submarine oipe- 0. (1987). P&-soil interaction model. Ofihore
lines. O&ho;e Technology Co& paper OTC 2967, Technolo$y Cin$, paper OTC 5504, 181-190. -
71-78. Wantland, G. M., ONeill, M. W., Reese, U. & Kalajain,
Lyons, C. G. (1973). Soil resistance to lateral sliding of E. H. (1979). Lateral stability of pipelines in clay.
marine pipelines. Ofihore technology Con&, paper Ofihore Technology Con&, paper OTC 3477, 1025
OTC 1876.479484. 1034.

Vous aimerez peut-être aussi