ANS&A
GEOTECHNICAL CENTRIFUGE,
‘TECHNOLOGY
With Compliments
ANDREW N SCHOFIELD AND ASSOCIATES LIMETED, 9 LITTLE ST MARYS LANE, CAMBRIDGE CB? IRR. TELEPHONE (0223) 460555
FACSIMILE (0223) 460777 REG. IN ENGLAND NO. 1866438 VAT NO.393 15 7827
( Centrifuge and Theoretical Modelling
of Conical Footings on Sand
by
Finian Seng Chin TAN
A dissertation submitted for the Degree of
Doctor of Philosophy at Cambridge University
Cambridge, England.
Christ's College, January, 1990To Mum and DadSUMMARY
CENTRIFUGE AND THEORETICAL MODELLING OF CONICAL FOOTINGS ON SAND
Finian Seng Chin TAN January, 1990
SUMMARY :
‘This dissertation is concerned with investigating the behaviour of conical and spud-can footings subjected
to vertical and combined static and cyclic loading on sand. The investigation consists of Physical
Centrifage modelling involving both the Beam and Drum Centrifuges provided by the Cambridge
University Engineering Department, and Theoretical (conceptual) modelling based an the Theory of
Plasticity.
Experimentally, the vertical and static combined loading of rough and semi-rough conical and spud-can
{footings on saturated sand were carried out in the Drum Centrifuge, simulating prototype footings of 1.6m
in diameter, and the cyclic combined loading of semi-rough spud-can footings on dry and oil saturated
sand, im the Beam Centrifuge, simulating prototype footings of 4 m in diameter. Theoretically, Perfect
Plastic, Work and Kinematic Hardening models were employed to described the experimeatal behaviour of
Conical footings when subjected to vertical, static, and cyclic combined loading respectively.
‘The vertical loading tests of conical and spud-can footings conducted in the Drum were unique in the sense
that up to 20 footing tests of 1.6 m prototype diameter, could be conducted on the same specimen, thereby
reducing errors due to variability in soil density between soil specimens. ‘These results were then used to
verify the theoretical perfect plastic constant @ analysis for axi-symmetric footings, based on the Method of
(Characteristics to take the weight of the material into consideration, assuming the Harr-Von Karman
hypothesis for the hoop stress, and using for the frst time, the Lundgren- Mortensen mathematically correct
siress field analysis to include footing roughness for surface conical footings on a Mohr-Coloumb material,
toarrive ata means of estimating the vertical penetration of conical footings in sand when subjected to pure
vertical load.
‘The static combined vertical and horizontal load tests of conical and spud-can footings also conducted in
the Drum Centrifuge provided a new insight into the exploration of the static yield locus, in that load paths
‘Which almost traced the entire yield locus were obtained, as opposed to the determination of several yield
‘points from which the yield locus could only be postulated, as has been done previously. A load state was
identified on the yield locus, at which continuous horizontal displacement could take place with no change
in vertical displacement or vertical and horizontal loads, separating the space within the yield locus into
‘work hardening and work softening regions. ‘This point was termed the ‘parallel point’ to describe the fact
thatthe incremental displacement vector is parallel to the horizontal load axis at this point. These results
‘were then used to verify the theoretical elasto-plastic work hardening model obeying a non-associated {low
rule, whose predictions were shown to be quite good both qualitatively and quantitatively.
‘The cyclic combined vertical, horizontal and moment load tests conducted in the Beam Centrifuge and
‘modelling prototype spud-can footing diameters of 4 m, showed that the displacement range over which the
response is truly elastic and non-dissipative was very small when compared with the displacement
necessary to reduce the stiffness 10 that corresponding to plastic yield. ‘These results were then used for
incorporation into a simple associated elasto-plastic work hardening model of an additional translating
inner true yield locus, in order to arrive ata kinematic hardening model. ‘The results, although qualitative,
‘were shown to exhibit most of the important aspects of cyclic behaviour.
Finally, recommendations were made for furure study. These recommendations were divided into direct
and indirect extensions to this theoretical and experimental program. Direct extensioas which included
‘more realistic jack up footing and leg models, subjected to more general static and cyclic load paths. And
indirect extensions, which included different soil conditions (e.g. layered soil) and experiments on whole
‘model rigs in order to identify the load paths and structure interaction problems.PREFACE
PREFACE
First of all, I would like to thank my supervisoe, Dr.R.G. James, the Assistant Director of Research, for his
‘most valuable judgment, advice and suggestions throughout all the stages of this research. Dr. James has
‘always been there and I much appreciate his dedication and constant willingness to help.
thas been a privilege to be a member of the Cambridge Soil Mechanics Group, which under the leadership
of Prof.Andrew Schofield, has provided a most intellectually stimulating and friendly environment. I wish
to thank all its past and present members, in particular Dr-Richard Dean, Mr-Malcolm Bolton, Dr-Ryan
Phillips and Prof.David Wood, for the many discussions and extremely beneficial exchange of information.
To the former I am also indebted for reading the draft copy of this dissertation, and making valuable
suggestions.
| would also like to thank the past and present staff members who have provided excellent technical help
and advice, in particular; Chris Collison, Clive Hooper, Neil Baker, Steve Chandler, Paul Ford, Wally
‘Gwizdala, John Chandler, Ralph Ward, Jim Doherty, Bill Balodis and Charlie Potter.
1 wish to express my gratitude to the Shell Group of Companies and the Cambridge Commonwealth Trust
for my Shell/Cambridge Scholarship and to my College, for giving me the honour of being a Christ
Scholar.
On a more personal level, I would not have completed this work without the practical and emotional
support of my wife, Pauline, and it would have been less of a joy had it not been for our son Colin's
‘constant distraction.
| cemtfy that, except where specific reference is made in the text to the work of others, the contents of this
dissertation are original and have not been submitted to any other University. This dissertation is the result
‘of my own work and includes nothing which is the outcome of work done in collaboration, It contains less
‘than 300 pages as authorized by the Board of Graduate Studies in their lewer dated 17 November 1989 (ref.
JCH/FIV/X/GS8604147).
FSCTan
January, 1990NOTATIONS
NOTATIONS
All symbols are defined at their first occurrence in the text. All stresses are effective stresses unless
otherwise stated. Compressive strains and compressive stresses are taken positive, following usual soil
‘mechanics practice. Work is taken as positive when it is done on a body. All angular dimensions are in
degrees. Real numbers are expressed in terms of 10 10 the power of an interger and deaoted by E followed
‘by that integer, where E means ‘times 10 to the power of, e.g. 0.004 is expressed as 4E-03, This table of
‘notations excludes symbols which are used in only one place in the text.
DEFINITIONS OF SYMBOLS USED IN SEVERAL PLACES IN THE TEXT
Roman (Lowercase)
c
e
ec
max
min
8
h
bn hardening modulus
hy additional hardening function when yield and bounding loci are notin contact
hy hardening function when yield and bounding loci are in contact
»P plastic horizontal displacement
i parameter to determine conditions of axial symmetry (j= 1) or plane strain (j = 0)
1 istortional displacement
r ‘elastic distortional displacement
Pp plastic distorional displacement
a ‘gravity level (all chapters) or number of cycles (Chapter 7)
> twiaxial mean stress [(6 + 263)/3]
Pu ‘ultimate bearing capacity (stress)
q triaxial deviator stress (6 - 63)
radial distance
t time
v veatical displacement
ve clastic vertical displacement
we plastic vertical displacement
x horizontal displacement‘NOTATIONS
vertical distance
Roman (Uppercase)
aon
Pee SOR PEE ro
LyDT
x
Ze
22z
3
aa teee
ee
determinant for Method of Characteristics
plan area of footing
effective plan area of footing
Plastic potential parameter
width of strip or diameter of circular footing
effective B
depth of overburden to base of fat footing or top of conical section (neglecting heave)
clastic modulus
‘component for superposition factor jt
specific gravity of soil grains
horizontal load acting on footing
horizontal load at centre of the yield locus
horizontal load atthe conjugate load point of the bounding locus
theoretical horizontal load capacity
‘peak horizontal load capacity in load space
relative density
ilataney index for triaxial and shear box tests
dilatancy index for footing tests
length of stress field measured from centre of footing (Chapter 3), magnitude of
distortional load (Chapters 5 & 8)
linear variable differential transformer
‘moment
sradient of line of parallel points in load space
‘gradient of line joining the peaks of the yield loci in load space
‘intercept of the projected elasto-plastic vertical load - vertical displacement line with the
vertical displacement axis
bearing capacity factor (cohesion)
bearing capacity factor (surcharge)
‘bearing capacity factor (self-weight)
Ny using axi-symmetrc angle of shearing resistance $y,
Ny using plane strain angle of shearing resistance
Component for superposition factor 1
Pore pressure transducer
‘normalized surcharge [2yD/(yB) = 2D/B]
‘normalized radial distance (21/(yB)] (Chapter 3), ratio of size of yield to bounding loci
(Chapter 8)
ratio of areas (A'/A)cat = 30°, 35°, 40°
é = 0°, 13°,
a = 0°, 13°, 25°, 45°, 60°
BAB+D)= 0, 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1.0
However, due tothe fact that this would require a total of 495 cases, only the Ny and Na cases (i.
‘BAB+D) values of approximately 0 and 1) were investigated forall the different values of , 8 and o, and
the effect of varying values of . with B/(B+D) studied only for the cases 6 = 30° and 35°, 8 = 13° and @ and
= 0°, 13°, 25°, 45° and 60° resulting in 90 + 180 = 270 cases.
3.4.1 Verification of Computer Program
To verify the results of the computer program, 6 additional cases were studied in order to enable
‘comparisons to be made with published results of numerical and exact solutions, and with observed
‘mechanisms of collapse, bringing the total number of cases to 276.
1. Plane strain case of a perfectly smooth footing on weightless soil with = 30°, so as to compare
with the exact Prandil solution (Terzaghi & Peck, 1967) and Hill-type stress field (Lau, 1988).
2. Plane strain case ofa perfectly rough footing on weightless soil with @ = 30°, so as to compare with
345‘THEORETICAL BEARING CAPACITY
the exact Prandl solution (Terzaghi & Peck, 1967) and Prandi type stress field (Lau, 1988).
3. Plane strain case of a perfectly rough footing on soil with weight and @ = 30°, so as to compare
‘with the numerical solution of Davis and Booker (1971).
4. __Axi-symmetric case of a perfectly smooth footing on weightless soil with @ = 35°, so as to compare
‘with the numerical solution of Lau (1988).
5. __Axi-symmetric case of a perfectly smooth footing on soil with weight and = 35°, so as to
Compare withthe numerical solution of Lau (1988).
6 Plane stain case ofa semi-rough (6 = 20°) wedge-shaped footing (a= 13°) on soil with weight and
@ = 45°, soas to compare with actual observed mechanisms of collapse.
Figures 3.17a & 3.17b show the derived stress fields for cases 1 and 2, and are almos identical to the Hill
and Prandil stess fields (Lau, 1988). The normalized length of the failure zone 11(0.5B) = 5.29 and
1i(0-5B)=9.58 measured from the cenre ofthe footing and the solution for Nq= 18.40 obtained numerically
for both the cases are exactly equal to those derived analytically using the following equations for
‘weightless soil.
L1OSB)smooth = exPICK /2nang leos( 45° - 6/2)leos(45° + 9/2)+1 = 5.29
LOSB rough = 2LM(0.5B) smooth ~ 1+ 1 = 958 3.25)
Ng = (C1 + sinoV/(1- sing)lexp(r tan) = 1840
For case 3, the plane strain fully rough solution of Davis and Booker (1971) obtained Ny = 14.96, which
‘when compared with the derived N, = 14.90 is notas close but is considered sufficiently similar considering
the differences in mesh spacing, radial divisions, target accuracies of iteration error, length of failure zone
[Land the deviation atthe centreline of the final angle W of the major principal stress from zero.
For axi-symmetric cases, asin cases 4 and 5, Lau’s simulation of both the weightless soil and soil with self-
weight gives Ny = 18.2 and N, = 61.0, which is almost identical vo the derived Ny = 18.16 and Ng = 60:97.
‘This verifies the accuracy of the program for the flat footing in plane strain for both rough and smooth
‘cases and with axial symmetry for the smooth case. The extension of the method to include both smooth
‘and rough conical and wedge-shaped footings has been described and may now be used to generate the
results.
Here itis interesting to compare a stress field obtained for a wedge-shaped footing with actual observed
‘mechanisms of collapse as in case 6. It is realized that real soil may not obey normality and thus
Quantitative comparisons should not be made. However, it may be still worthwhile to examine whether
false bases do indeed form beneath a rough or semi-rough wedge-shaped footing as hypothesized,
Consider the stress field (fig.3.18) derived from a plane strain analysis of a semi-rough (8 = 20°) wedge-
shaped footing («.= 13°) on a soil (p = 45°) with self-weight and compare it qualitatively with the actual
Observed mechanism of collapse as obtained by Tan (1987) from x-ray photographs of a 1g test as shown in
3:16THEORETICAL BEARING CAPACITY
Plates 3.1 and 3.2. As can be seen, false bases not unlike that derived numerically, do indeed form beneath
«8 semi-rough wedge-shaped footing on real soil, with the false base not extending to the comers of the
footing when it is of intermediate roughness.
342 Numerical Details and Accuracies
‘The following lists the details and accuracies of the input to the program:
1, The iteration at each point was continued until Ay/y and AE py/E,y was less than or equal to
‘0.0001
2 The mesh spacing was kept to a minimum by specifying a constant number of 51 nodal points
along the free or surcharged surface.
3. ‘The number of radial divisions was fixed at 50,
4, Since a singularity point exists at the edge of the footing, the mesh spacings of the first 11 nodal
points from and including the singularity point were divided by 4.
5. The "N," case was investigated with a minimal surcharge of 0.001(yB/2) in an attempt to simulate
a surface footing.
6. The "Ng" case was achieved by an input of surcharge of 1000(yB/2) in an attempt to simulate
‘weightless soil.
‘The two cases involving bounded cones beneath a rough footing and the way the bounded cones and the
cntire stress field change in size as the depth of surcharge D/B, angle of roughness 8 and cone angle &
change are clearly illustrated in the following figures. fig.3.19 to fig.3.22. The results of the calculated
bearing capacity factors Ny and Nand their dependence on ¢, 6 and ot for the 90 cases are presented in
tabular and graphical form in Table 3.1 and fig.3.23 tofig.3.25. respectively, with the remaining 180 cases
oA =
‘Thus, although simplistic, the stress field calculations have shown to be very useful in the understanding of
the inclined load behaviour of wedge shaped footings on weightless perfect plastic material, in particular
the effects of various parameters. Experimental results will prove how good they are for conical footings
‘on real sand,
5.3 COMPLETE MATHEMATICAL MODEL
‘The theories and equations described so far have only dealt with the ultimate state, and are, in continuum
‘mechanics terms, perfect plastic models which only give details of the point of failure. However if the
footing behaves more or less elastically within the interaction locus, and the locus grows larger as the
Footing settles, then it is believed that a work hardening elasto-plastic model might well be more useful to
‘compietely describe the qualitative behaviour of footings when subjected to combined loading, in particular
With regard tothe prediction of displacements,
53.1 The Interaction Diagram as a Yield Locus
Consider Shi's empirical equation 1 (egn.5.5). For the case when there is no moment, Eqn.5.12 may be
replaced by
Hg = My VV Gq) - (VV?) (5.20)
where My =tané
‘As canbe seen, for one particular value of 8, My, isa constant and the locus becomes an ellipse of identical
shape with the yield locus of Modified Camclay (Britto and Gunn, 1987) when the gradient ofthe critical
state line M is replaced by Myy. Fig.5.16 shows an illustration of the model when compared with the
lemental model of Modified Camclay when M and My, are equal to 1. This similarity in shape is quite
"useful since the equations of the elemental model have already been derived. However when the results of
the stress field calculations of the previous chapter are viewed again (fig.5.11), it can be seen that they
59THEORETICAL STATIC COMBINED LOADING
Predict peak points that are below V/V, values of 0.5, and with the gradient at V/V n= 1, not at infinity,
lunlike the proposed model. ‘Thus although simple, the elliptical interaction locus does not seem the best
solution to the problem.
Here itis worth noting again that although the plane strain stress field calculations were based on
weightless perfect plastic material, they have provided useful information about the effect of various
parameters, and thus itis appropriate to base the interaction locus shape on those obtained theoretically so
that the effects of these parameters (equations 5.17 and 5.19) may be included in any interaction locus
derivation. However the stress field equations are cumbersome to use directly in the derivation of the
interaction locus equation. It would be desirable if one single equation could describe the general shape of
the interaction locus.
‘A simple interaction locus, which predicts peak points that are below V/V, values of 0.5, and with the
‘gradients at V/V, = 1, not at infinity and also one which is well known, is an equation similar to the yield
ocus of the elemental model, Camclay (Schofield and Wroth, 1968). Fig.5.17 shows a comparison of such
an interaction locus and expressed as
HV 9g) = VM (VV g)10V (VII) (21)
‘with those derived from the plane strain stress field calculations. As can be seen, the comparison is much
‘better, and simplifies the yield locus equation greatly. Furthermore, if used in conjunction with equations
5.17 and 5.19, then Hy and hence My; may be estimated theoretically for different wedge angles with the
effect of the initial depth of overburden also taken into consideration.
‘So far, the equations of two yield loci have been proposed, and these were shown to be similar in shape
with the yield loci of well known elemental models. These comparisons of the interaction loci of the
overall behaviour of footings when subjected to combined vertical and horizontal loading with yield loci of
¢lemental elasto-plastic models for soil can be taken further, ifthe behaviour within the interaction locus is
Clastic. ILis well known, that a footing when loaded vertically in sand, exhibits a very stiff behaviour
uring vertical unloading and very nearly retraces its path when reloaded vertically. During horizontal
loading though, the hysteretic behaviour is more pronounced, bat if this is ignored for monotonic analyses,
then the interaction diagram of the footing may in fact be regarded as a yield locus which contains a space
of different tangent stiffness and with parameters derived elastically. Thus the shape of the yield locus and.
the behaviour within it may be defined. However, in order to obiain a complete elasto-plastic work
hardening model, the displacements predicted beyond the yield locus when plastic deformation is taking
place must be considered.
53.2 Flow Rule
Experiments on cones and spuds conducted at Cambridge by Shi (1985, 1988), Graham (1986), Brod
(1988), Wooley (1989) and Gardiner (1989) have shown that incremental displacement vectors are almost
5-10‘THEORETICAL STATIC COMBINED LOADING
normal to a symmetric yield locus for V/V q values greater than the point at which the peak of the yield
Jocus occurs, which in this case is 0.5, but at lower V/V, values, the footing sil settles when horizontally
loaded as shown in fig 5.18, This seems to indicate that perhaps, the footing behaviour does not obey an
associated flow rule, The shape of a non-associated plastic potential though, has not been defined. It has
‘been thought that perhaps it might be unlikely for the footing to heave even at very low V/Vjq and hence a
Plastic potential which was identical tothe yield locus above V/Vjq = 0.5, and then remains at a constant
sradient as shown in fig.5.19, might suffice. However, although it has the advantage of simplicity, it does
‘Rot account for the observed vertical settlement at V/V, values below the peak of the yield locus as was
seen infig.5.18. As a consequence, not much has been said about the behaviour at V/V, values below this
point, where the assumption of normality is thought to be no longer valid.
From a physical point of view. it does not seem impossible for the footing to heave, as dilation takes place
‘when the soil grains ride over each other to accommodate the motion of the footing. If s, then the footing
‘might obey a non-associated flow rule with a stable plastic potential which passes through the origin, and
‘with the point at which the incremental displacement vector is parallel to the horizontal load axis, occurring
ata V/Vjq value less than that at which the peak of the yield locus occurs. Experiments will then have to
be very carefully conducted and at very low values of V/V yin order to prove that heave does indeed occur
in this region,
Mathematically, a mode! with an elliptical yield locus (eqn.5.20) can obey a non-associated flow rule if a
plastic potential is chosen such that its peak occurs at a V/V, value below 0.5. A good example is a
plastic potential equation similar to that of Camclay (eqn.5.21), but with the peak distonional magnitude of
the plastic potential modified so that it is equal to the peak magnitude of the elliptical yield locus, and
expressed as
CNV) = VM pg VV )10(V q/ VII) (522)
where My) = (Myexp(1)I/2
Differentiation of this equation produces
~ BHIBV = Mp - HV (8.23)
and thus the flow rule is
‘BvP/OHP = - BHIBV = Mp - HV (5.24)
where vP = plastic vertical displacement
hy = plastic horizontal displacement
This equation however, predicts horizontal displacement even when H is zero and V
the model is mirrored in the negative distortional space, then the negative and positive horizontal
I: OF course if
SL‘THEORETICAL STATIC COMBINED LOADING
displacements should cancel each other out. Nevertheless, a very small horizontal oad will result in an
incremental displacement vector inclined at Mpp, and thus it may not be considered an ideal plastic
potential for the behaviour of footings under combined load. By modifying the equation slightly, a solution
results which rectifies this problem and is given by
Hg, =Mpp\ (VexpC)V IV q/V)) 625)
‘This plastic potential together with an elliptical yield locus similar to that of equation 5.20, but assuming
My is any constant, is shown in fig.5.20 (for My, = 1), and wil be hereafter called Mode! 1a (Appendix
B1). As can be seen, the plastic potential meets the vertical load axis at right angles and thus automatically
Predicts no horizontal displacement with no horizontal load. Also the point at which the peak of the plastic
Potential intersects the yicld locus does not coincide with the peak of the yield locas where V/V, = 0.5,
‘bat occurs at V/V, = 0.35. This implies that for V/V,, values greater than 0.35, the footing settles when
Plastic deformation is taking place, and for smaller values of V/V ,,, the footing heaves.
‘The model now seems to satisfy some of the observed behaviour, and is one method of obtaining a non-
associated flow rule, however, the difference in shapes of the yield locus and the plastic potential might
‘make the model mathematically cumbersome to use. Efforts were made to arrive at a simpler plastic
potential which could still express the observed behaviour. Naturally, the simplest plastic potential would
bbe one of the same shape as the yield locus, and it has been found thatthe yield locus equation, bet with a
different distortional magnitude will be able to express the desired effect. This is achieved by changing the
‘equation ofthe yield locus very slightly as shown below.
Hy = Myp (VV (WMI (626)
where Mp) =A™My
AY = plastic potential parameter
‘Asan be seen, the form of the equation is the same as the form of eqn5.20, but the parameter My is now
replaced by the parameter Mp, which is made a function of My). If A” < 1, then the peak of the plastic
potential intersects the yield locus at a V/V p value above that at which the peak of the yield locus occurs
and the flow rule is non-associated. If A” = 1, then the plastic potential and the yield locus coincide and
the flow rule is associated. If A" > 1, then the intersection occurs ata V/V value below that at which the
peak of the yield locus occurs and the flow is mon-associated again. Fig.5.21 shows the incremental
displacement vectors and plastic potentials if A” = 2, and together with the elliptical yield locus of eqn.5.20
‘(Myj= constant, will be hereafter referred to as Model 1b (Appendix B2). As can be seen, the model now
predicts vertical settlement of the footing at points above V/V , = 0.2, and heave below it, when plastic
deformation is taking place, and also has the added advantage of simplicity. This method of simply
‘expressing the parameter Mp, as a function of My, may therefore be used whenever a non-associated flow
rule for a convex yield locus is desired.
5-12THEORETICAL STATIC COMBINED LOADING
‘This leaves us with the derivation of a plastic potential for the yield locus which is similar in shape to
‘Camelay (eqn.5.21), and which has been shown to describe the results ofthe stress field calculations quite
‘well. Now, because the plastic potential of eqn.5.25 predicts peak point which is at the same position as
2qn.5.21 when Mpy = Myj, it would not make the model able to predict settlement at V/V , values below
that at which the peak of the yield locus occurs as was seen in fig.5.18, Thus, this leaves us with the other
option of simply utilizing a plastic potential which is similar in equation with that of the yield locus, but
replacing the parameter Myy with Mp where Mp, isa function of My. However as said before, for this
articular yield locus equation, it makes the plastic potential rather unstable since the gradient of the
incremental displacement vectors change quite drastically from Mpg to 0 whea V/V, tends to 1. A
solution is to combine the wo methods. A slight modification was made to Mpp in eqn.5.25 to arrive at
BN, =Mypl(VferrOV Vg) 62
where Mpy = A'My,
‘and another complete yield locus and plastic potential is obtained, and will be hereafter referred to as
Model 2 (Appendix B3). Fig.5.22 shows an illustration of the model. Experiments will decide which
‘model best fits the experimental results,
Here it should be noted that models obeying a non-associated flow rule have been proposed in this chapter
in order to betier describe observed results. ‘There are however strong arguments which support the use of
the normality principle for conceptual models, even if the physical models appear to exhibit a non-
associated flow rule. Firstly, the problem is simplified when the yield locus and plastic potential coincide
since the number of functions that have to be generated in order to describe the plastic response is reduced
by one. Secondly, the solution of the equations that emerge in numerical analyses is faster and the
accuracy of the numerical computations can be more easily guaranteed for an associated flow model
(Wood, 1986). Calladine (1985) uses the idea of maximum plastic work to allow appreciation of the
generality of the normality principle.
However, itis important not to confuse the postulate of normality with Drucker’s (1964) postulate of
stability. Drucker developed a concept of stability related to the sign of the product 80 Be P of the stress
increment vector 89 with the plastic strain increment vector 8¢ P. If the product is positive and incremental
‘energy is dissipated then the increment is stable. If the product is zero and incremental energy is neither
dissipated nor released then the increment is neutral, And if the product is negative and incremental energy
is released then the increment is unstable. If all possible products of the stress increment vector with the
Plastic strain increment vector are positive, then Drucker considers the material or system a stable one.
But Drucker's stability postulate is derived mainly from energy considerations. Stability is also often used
to describe stress-strain or load-isplacement behaviour, where a work hardening material or system may
be described as stable, and a work softening material or system, described as unstable, A Drucker material
‘or system is necessarily associated and rules out work softening and hence is stable from considerations of
513THEORETICAL STATIC COMBINED LOADING
nergy and stress-strain or load-displacement behaviour. The converse however may not be true, A non-
Drucker material may work harden if the flow is non-associated. According to Drucker (1964), postulating
the stability of a material does not mean that all real materials ae stable. Drucker’s stability postulate is a
‘means of classification, and not a law of nature, Inthe original Camclay model (Schofield & Wroth, 1968),
the Normality Rule was adopted. This, together with other aspects of the model, meant that the product
50 Se P would always be zer0 or positive for stress increments at stress ratios a/p less than the critical
state stress ratio M where the model predicts work hardening behaviour, and would always be zero or
negative at stress ratios p greater than M where the model predicts work softening behaviour. Hence for
the model, the material at low stress ratio q < Mp was always stable according to Dracker’s criterion of
incremental energy dissipation, and also predicted stable behaviour, in the sense that work hardening takes
place. And the model was always unstable at high stress ratio q > Mp according to Dracker's release of,
incremental energy criterion, and also predicted unstable behaviour in the sease that the material work
softens. Thus Camclay truly exhibits stable and unstable behaviour both from Drucker’s point of view, and
‘when the stress-strain behaviour is considered,
In a practical context, loads are not known exactly, and it may therefore be wise to design structures which
are stable in the senses that small errors in estimating load conditions produce only small deviations from
the calculated displacements and that small changes in load or displacement conditions do not lead to
‘uncontrolled failure, By ensuring that Drucker’s stability postulate is not violated, then ican be envisaged
that instability will not result. In contrast, the purpose of this chapter is to obtain initial models to describe
the actual behaviour of the footings, and in this context itis appropriate to attempt to model the data rather
than attempt to produce a design method which would have this kind of stability or error-insensitivity.
Furthermore, Wood (1986) states that where normality might be regarded as a convenient modelling
assumption, stability can hardly be proposed as a necessary one. For these reasons, to better describe the
‘data, models with non-associated plastic flow have been developed here. For these models, it is possible,
for certain load increments, for the product ofthe load increment with the displacemeat increment to be
negative at low load ratios below My,, and thus incremental energy would be released during plastic
deformation. However at these low load ratios, growth of the yield locus occurs as willbe seen in section
5.3.5, and this hardening allows the load-displacement behaviour to be stable. It should be noted that some
Additional considerations might be necessary before this model might be used in the context of practical
‘design or analysis where applied loads cannot be estimated accurately.
533 3-Dimensional Model for Combined Loading
‘As mentioned previously in section 5.1.3, the inclusion of moment into the expressions is simplified greatly
if the Horizontal Load - Momen/Breadth relationship for a particular value of V/V y, isthat of a circle and
expressed as
(HN, + (MIBV,,)I? = constant 6.28)
‘This equation will enable the 3 models to also theoretically include the ability to consider moment loading,
4‘THEORETICAL STATIC COMBINED LOADING
1s well. For example, eqn.5.20 of Model 1b (Myj = constant) becomes
(CAIN gg)? + OMCBV pg)? = Myy1V/V py = (V/V gg) (5.29)
Here, if a distortional force Lis defined such that
(WV )?= HV)? + IMIBV I? (530)
then Vy =MyiVVVqq = (WV QQ") 631)
and the yield locus of Model 1b (My, = constant to be determined) becomes a complete 3 dimensional
yield locus defined in Vertical Load, Horizontal Load and Moment/Breadth space. Similarly, the plastic
potential of Model 1b (eqn.5.26) is changed for the combined load model to
UV = Mpp VV yq = Wg) 6.32)
which implies that the plastic potential too is circular in Horizontal Load - Moment/Breadth space.
Having, described the distortional term L, an associated incremental distortional displacement term &l must
exist such that no work is lost or gained. From consideration of the work equation, LBI must be equal to
Hh + MBO, and thus the incremental distortional displacement 81 becomes
8 = (Han + MBO) UH + (/B)) 6.33)
If itis further assumed that moment is produced by the action of the horizontal load at a height EB, then
M-= HEB. In this case, eqn.5.33 can be integrated to give
1 = (h+ OEB)N(I + £) (5:34)
Here, itis important to note that although the existence of a plastic potential of a different shape with the
‘yield locus indicates that the model is non-associated, the fact thatthe same circular equation in Horizontal
Load - Moment/Breadth space (eqn.5.28) is used for both the plastic potential and the yield locus implies
that the incremental displacement vectors appear normal in this space as shown in fig 5.23.
5.34 The Line of Parallel Points
‘Coming back to Vertical Load - Distortional Load space, itis worth noting, that the assumption of plastic.
Potentials as in Models 1a, 1b and 2 which predict settlement at high ratios of V/V,q and heave at low
ratios must also imply an existence of a transition point at which no vertical displacement takes place
during plastic deformation. This should occur at the point where the incremental displacement vector is
Parallel tothe distortional load and incremental distortional displacement axes. In thiscase, and in the case
‘of any convex plastic potential, the point also coincides with the peak point of the plastic potential and
S15‘THEORETICAL STATIC COMBINED LOADING
allows continuous distortion to take place with no change in vertical displacement. Therefore, itis
hereafter called the Parallel Point and the line joining the parallel points of different yield loc, the Line of
Parallel Points (Ipp) expressed as
HN = MypVVn (535)
535 Hardening Rule
‘The flow rule gives information about the direction of the plastic incremental displacement vector (i.e the
relative proportions ofthe various components); the magnitude of the displacement depends on the amount
cof work hardening or softening, and is controlled by the hardening rule, Considering the Vertical Load -
Vertical Displacement space, assuming constant bearing capacity factors Ny and Nq and thatthe depth of
overburden D is equivalent to the penetration v for a flat footing or for initially buried cones and wedges,
the classical Tercaghi (1943) bearing capacity relationship (eqn.3.1) for a cohesionless material may be
expressed as
vo EN#AV (536)
where Vij = pyr B/4
N= -05BNYNg
A =4nyB?Nge)
‘This relationship describes the combined plastic and elastic settlement behaviour in Vertical Load -
Vertical Displacement space during pure vertical loading. Differentiating this expression, we get
By ABV 630
For the elastic behaviour, it can be shown (e.g. Poulos and Davis, 1974) that the general elastic relationship
‘may be expressed by
Bve = KBV, (5.38)
where x = (1-v°)(EB)
v= Poisson's ratio
E = Young's Modulus
Combining equations 5.37 & 5.38, and considering that 5V is equal to V;q during elasto-plastic