Académique Documents
Professionnel Documents
Culture Documents
Arthur E. Lord Jr
Geosynthetic Research Institute, Drexel University, Philadelphia,
Pennsylvania 19104, USA
ABSTRACT
1 INTRODUCTION
r 1
counter wff.
or
I d t n f i c o l model
L.-J
(at resf)
Cat rest}
prototype. One possible remedy for this problem is the use of an artificial
gravitational field, which makes the model appear heavier. The centrifuge
provides a method for supplying an enhanced gravitational effect. 1 The
increased gravitational field is supplied by the centripetal (radial)
acceleration. (A common use of centrifuges is for separation of the
components in liquid suspensions, e.g. separating the cream in milk and
separating suspended particles in medical laboratory samples.)
Figure 1 is a schematic diagram of a centrifuge. The container, with the
soil model, hangs downward when the centrifuge is at rest. As the centrifuge
is rotated the model rotates upward and, at sufficiently high rotational
speed, the model rotates in a horizontal plane. In this final position the
centripetal acceleration is
V2
a, - - rJ (1)
r
2 PHYSICS O F T H E C E N T R I F U G E
F 2 - El = d F = d m a , (2)
where
F = force
dm = mass of the element
ar = -- = radial acceleration
r
Furthermore,
dm = tadV = odrA (3)
where
p = mass density
dV = volume of mass dm
dr = radial thickness of dm
A = area of drn
136 A r t h u r E. L o r d Jr
ro
/r-A
tO
f,, II A
Fig. 2. Schematic drawing of a soil element in the rotating body, showing the radial forces.
dF
do" - (4)
A
dF = pdrAar
dF
do" - - pa~dr (5)
A
The radial acceleration is given by eqn (1) yielding
d o " = proJ 2 d r
Integrating from the free inner surface of the model (r = r0) where the stress
is zero, we obtain, for a given to, the radial stress at the position ri
O'i =
P°J2
2
(4 - ro2) (6a)
Geosynthetic/soil studies using a centrifuge 137
tr = p g h (7)
1-2(105) = 2000(9.8)h
1.2(105 )
-h
2(0-98)(104 )
6.12m = h
6.12
n - -- - 61.2
0.1
w h i c h f o r o u r e x a m p l e c o m e s o u t to be
n(9.8) = (8702(1)
(8~r) 2
n - - - - 64.4
9.8
138 Arthur E. Lord Jr
The two values of n are slightly different due to the slight nonlinearity of eqn
(6a).
If the model's dimension is small with respect to the average radius of the
centrifuge, then the radial stress distribution is approximately linear with
depth from the free surface. This can be seen from eqn (6a).
2
(3"i _- - r0
p¢o
o'i = T [ ( r i + r0)(ri- r0)] (6b)
This gives the approximate linear dependence with depth. Equation (6a)
also demonstrates the scale factor concept. Equation (6c) can be written
which agrees with the standard stress/depth relation (eqn (7)) o-~ = p g h if
Ar = h
and
a, = geff = ng
The deviation of the linear eqn (6c) from the true stress distribution eqn
(6a) can be shown to be
O't . . . . 1 + - Ar
- (9)
O'linear 2r0
Table l shows this ratio as a function of Ar/2ro and also the percentage by
which o-~. . . . is below the true stress. It must be realized that if Ar represents
the m a x i m u m depth of the model, then the average stress is only about 50%
of this departure. The table shows that if the dimension of the model is 10%
or less of the radius of the centrifuge, then the maximum deviation from the
Geosynthetic/soil studies using a centrifuge 139
TABLE 1
Deviation of True Stress from Linear Approximation
3 H I S T O R Y A N D USES OF T H E G E O T E C H N I C A L C E N T R I F U G E
This section will borrow heavily from the excellent review of Chaney and
Fragaszy. ~Phillips 3 in France (in 1869) made the first suggestion of the use of
a centrifuge to study self-weight stresses in beams. Bucky 4 (USA) and
Pokrovsky 5 (USSR) and Davidenkov 6 (USSR) worked on centrifuges in t h e
140 Arthur E. Lord Jr
90
80
70
60 //z/ THEORY
"~ 50 , ~ .... EXPERIMENT
w
N
Z
40
3O
f
20
I0
0 I i i i i i i i
0 I 2 3 4
FREQUENCY (revo/./sec)
60 120 180 240
RPM
Fig. 3. Angle of model (with respect to the vertical) versus the frequencyof rotation.
Comparison of theoryand experiment.
early 1930s. A few projects were carried out subsequently in the U S A , 4"7'8
while in the USSR more than 50 centrifuges were built for model testing of
dams, foundations, earth fill embankments, etc., in the period 1932-1980. 9
In the late 1960s Rowe at University of Manchester Institute of Science
and Technology (UMIST) and Schofield at Cambridge University
developed excellent geotechnical centrifuge programs. In the early 1970s
Rowe built a large machine at Simon Engineering Laboratory at the
University of Manchester and a 3.7 m (12 ft) radius machine was developed
at Cambridge by Roscoe and James. The Cambridge project was taken over
by Schofield on his return to Cambridge from Manchester in 1974. England
is a most active center for geotechnical centrifuge studies and probably has
produced more PhDs and hosted more sabbatical visitors than any other
country in the world.
At present centrifuge activities are ongoing in many European countries
and Japan. The USA had a late start in the use of the centrifuge on
geotechnical problems. A small centrifuge was built at the University of
California at Davis in 1972 under Chaney. Scott at the California
Institute of Technology was active by 1975. Schmidt at Boeing in
Washington State worked on cratering starting in 1976 and Ko at the
University of Colorado began a program in 1978. An indication of the
present activity in the USA is given in Table 2 (from Chaney and Fragaszyt).
The very large centrifuges (8-10 m radius) at Moffet Field (previously
Geosynthetic/soil studies using a centrifuge 141
TABLE 2
List of U S Centrifuge Facilities
a 1 ft = 0.304 8 m.
b l Ib = 0"453 6 kg.
c I ton = 1.016 x 103 kg.
N A S A and run by the University of California at Davis) and Sandia are not
operational at present as far as the author is aware. (The Davis effort has
now been moved back to campus.) The Russian centrifuge at Baku is also
extremely large and its status at present is unknown to the author.
Centrifuges have been employed in research on a large number of
geotechnical-related problems:
• Beam-soil interaction. 10
• Bearing capacity. 11-13
• Earthquake effects. 14
• Excavations. 15
• Flow simulation. 16
• Foundations. iv
• Ice forces. 2
• Liquifaction. TM
• Phreatic surfaces observations.19
142 Arthur E. Lord Jr
• Sand drains. 23
• Slurry walls. 24
• Slope stability--many projects. 25'26
• Subsidence. 27
• Tunnels. 28
4 SCALING CONSIDERATIONS
where
u = pore water pressure
e = void ratio
k = coefficient of permeability
y = unit weight of water
a = coefficient of compressibility
z and t = the spatial and temporal variables
Zp = nZm (12)
where the subscripts p and m refer to the prototype and model, respectively.
The other quantities are scaled accordingly
Um= O[uU p (12a)
'Ym = ay~/p (12b)
tm = a t t p (12c)
Geosynthetic/soilstudies using a centrifuge 143
km = akkp (12d)
am = a.ap (12e)
Comparing eqn (11) with eqn (13a) shows that for similarity of the model
and prototype, the following relation must exist among the various scale
factors
aka'n-----~2- 1 (14)
OgaO( 3,
If the permeability, unit weight of the fluid and the compressibility of the
soil are the same between model and prototype, i.e., ak = a, -- % = 1,
then
1 (14a)
O/t - - ?/2
Thus
1
tm= -~tp (15)
which says that the same process (e.g. consolidation of the model) happens
nZ-times faster in the model than in the prototype. (Obviously with choices
of the o~s other than one, tm 4=1/n 2tp.) This is easy to understand in physical
terms. Consolidation is a diffusion-type process, in which the time for a
process to occur depends on the pertinent distance squared. The model is
only (1/n)-times as thick as the prototype, therefore (with the same k) the
process takes only (1/n)2-times as long. In exactly the same m a n n e r one can
easily show that a process such as Darcy flow which has a velocity associated
with it will take (1/n)-times as long in the model as the prototype. Due to the
144 A r t h u r E. L o r d Jr
TABLE 3
List of Scaling Relations
great time savings involved, it would appear that flow and consolidation
studies would be most popular in the centrifuge. In point-of-fact, they are
not as yet. Therefore there may be either significant difficulties with the
logistics of such measurements or with the scaling theory.
A set of scaling relations is given in Table 3 which is from Scott and
Morgan. 3° (All entries in Table 3 are not obvious to the author, but this will
possibly be addressed at a later date.)
In the second approach for determining the scaling relations, no
differential eqn(s) is (are) available describing the process and therefore
dimensional analysis must be used. A very readable treatment of this
approach as applied to elastic problems is given in Hoek. 29 Buckingham's
theory is used in which the displacement and stress at a point are written as a
function of a pertinent number of dimensionless parameters. These
parameters for the elastic problem are x, y, z, t, the Young's modulus of the
material (Poisson's ratio is already dimensionless), density, g, applied
stresses and forces and applied displacement, internal stresses and possibly
other variables. For similarity between the model and the prototype as
applied to the properties of the materials, the most pertinent result is (there
may also be other relations to be satisfied 29)
Lp _ Ep Pm gm
a (16)
Lm Em pp gp
Geosynthetic/soil studies using a centrifuge 145
where
L = the length
E = modulus of elasticity (it could be another pertinent elastic constant but not
Poisson's ratio)
g = acceleration
ct -- the ratio between the resulting stress at an equivalent point in prototype
and model, i.e. O-p = Carm
Lp _ Ep Pm
(16b)
Lm Em pp
where
F0 = the sum of moments about point 0
m = the mass of the slipping (rotating) region
r = the resisting shear stress on the failure arc (this is shear force per length of
arc)
lab = arc length 'ab'
T = ultimate stress in the geotextile
146 Arthur E. Lord Jr
Fig. 4. Schematic drawing of model for calculating limit equilibrium ofa geotextile reinforced
slope.
Equation (17) is written for one unit thickness perpendicular to the plane
of the drawing. Thus for dimensional analysis the one unit thickness is
included.
For a purely frictional material (e.g. a granular soil) we can use
7i = Ni t a n 0 i (18)
ri = m i g tan 01
mgR1 =
(h ~ migtanOi
)
liRo+ TR2 (17a)
i 1
(17b)
The first n (unparenthesized) on the right hand side comes from the fact
that 7-is a force/length along the arc 'ab'. (Note that there would appear to be
scaling problems with a c - ~bsoil because the cohesion would not scale with
mass. However, centrifuge slope stability studies are performed on cohesive
soils with no mention of this problem, to the author's knowledge.) From eqn
(17b):
I
aT -- n2 (19)
i.e. scaling says that the tensile strength of the geotextile should be (l/n2) -
times as large in the model as compared to the prototype. Blivet et al. 22find
that T should scale as 1/n. Regardless of which is correct, on physical
Geosynthetic/soil studies using a centrifuge 147
time observations on the model with a closed circuit T V system. These are
seen in Figs 5 and 6. Reflecting balls or painted stripes placed in a regular
array in the model, will allow general m o v e m e n t (and strains) to be
m o n i t o r e d during rotation.
Fig. 9. Model of reinforced sand slope on soft clay foundation after spinning to an equivalent
height of 335.3 cm.
152 Arthur E. Lord Jr
8 CONCLUSION
This article has given a brief review of the geotechnical centrifuge together
with descriptions of the Drexel centrifuge and a preliminary experiment.
The centrifuge appears to be a viable modeling procedure to study soil/
geosynthetic problems at the large scale. It probably is the only full scale
modeling approach available. The design aspect of geosynthetics, which is in
its infancy, could very well use data from centrifuge related studies.
The author feels that there are fundamental aspects of centrifuge
modeling still to be researched. For example, lateral stress effects are not
well understood. Scaling has not been adequately addressed in many
problems (e.g. clay soils, geosynthetic and geosynthetic/soil interaction).
Placing instruments in small soil models could lead to unacceptable
disturbance effects.
Some possible future work on the Drexel centrifuge includes:
• soft soil base/sand berms, with geotextile reinforcing elements.
• spider netting soil-reinforcement.
• forces developed in geomembrane liners on slope walls of landfills.
• coastal erosion barriers.
• fundamental centrifuge experiments (not project oriented).
ACKNOWLEDGEMENTS
REFERENCES
Displacements in Centrifugal Models, Proc. 8th Int. Conf. on Soil Mech. and
Found. Engin., Moscow (1973), 1 (part 1), pp. 65-70.
25. Schofield, A. N., Cambridge Geotechnical Centrifuge Operations,
Geotechnique, 30(2), (1980), pp. 227-68. This was the 1980 Rankine Lecture.
26. Kim, M. M. & Ko, H. Y., Centrifugal testing of Soil Slope Models, Transp.
Res. Record, 872 (1982), pp. 7-15.
27. Sterling, H. J. & Ronayne, M., Simulating Landfill Cover Subsidence, Proc.
Conf. on Land Disposal of Hazardous Waste, U. S. E.P.A., Cincinnati, Ohio
(April, 1985), pp. 236--44.
28. Atkinson, J. H., Potts, D. M. & Schofield, A. N., Centrifugal Model tests on
Shallow Tunnels in Sand, Tunnels and Tunnelling, 9 (1977), pp. 59--64.
29. Hoek, E., The Design of a Centrifuge for the Simulation of Gravitational
Force Fields in Mine Models, J. South African Inst. Mining and Metall., 65
(1965), pp. 455-87.
30. Scott, R. F. & Morgan, N. R., Feasibility and Desirability of Constructing a
Very Large Centrifuge for Geotechnical Studies, Report 760-170 National
Science Foundation, Washington, DC, 1977.
31. Langhaar, H. L., Dimensional Analysis and the Theory of Models, Wiley, New
York (1951).
32. Personal communications with Mike Ronayne and Darryl Greer, previously
graduate students in Civil Engineering at the University of Kentucky.
33. Choudhury, S. P., A Study of Reinforced Earth Retaining Walls and Sand
Backfill by Centrifugal Modelling, PhD Thesis, University of Manchester
(1977).
34. Pang, P. L. R., Centrifugal Model Tests of Reinforced Earth Walls, PhD
Thesis, University of Manchester (1979).
35. Koerner, R. M., A new method of soil slope stabilization using geosynthetics,
Geotechnical Fabrics Report (Jan./Feb. 1984) pp. 18-25.
APPENDIX
Calculation of the angle of the model (with respect to vertical) versus the
rotational speed.
where
E F r = the sum of the forces in the radial direction = T (the tensile force in the
connecting rod)
m = total mass of the model
ar = radial acceleration of center-of-mass (c.m.) of model
Geosynthetic/soil studies using a centrifuge 155
and
Using
r = R+lsinO (A2)
I
N
.,~Ua 7" •
- R
mg
Fig. AI. Schematic diagram of the rotating model used to calculate the angle of model versus
speed of rotation.
mtoE(R+/sin0) = T (A3)
or
N= mg
156 Arthur E. Lord Jr
tan0 - T (A5)
N
i.e. the vector sum of T and N passes through the c.m. (the weight m g
already goes through the c.m.).
Combining eqns (A3) and (A5)
NtanO = T
mg tan 0 = mtoZ(R+/sin O)
gtanO = toZ(R + lsinO) (lo)