Principles

p
of centrifuge
g
modeling
TNA Workshop on Centrifuge Modelling 3/4 March 2011
Dr Gopal Madabhushi,
Madabhushi
Reader in Geotechnical Engineering,
University of Cambridge

Introduction
Modern day geotechnical engineer uses
the following
g methods in design:
g
For simple problems
Empiricall or S
Semi-empiricall methods
Semih d
Closed form solutions
For complex problems
Finite
Fi it Element
El
tA
Analysis
l i

When using the FE method

method, we need to
know the constitutive behaviour of soil
i.e. the stress strain behaviour of the
soil under the type of loading it
experiences
p
needs to be known.
If Yes, we can program this into the FE
codes and proceed with the design
If No, ???
Let us look at the stress strain behaviour
of soil

Stress Strain behaviour of soil

Let us consider the shear stress Vs shear
strain for loose and dense soils

Shear
Stres
s

Dense
Sand
Loose Sand

Shear box test

Shear Strain

Soil is highly NON-LINEAR and PLASTIC

Physical Modelling in Geotechnics

As soil is nonnon-linear we cannot simply model

our problem (say a caisson subjected to lateral
load) at small scale
Dense Sand

Shear
Stress

Model behaviour at
small stresses and
strains

Loose Sand

Prototype
behaviour at
large stresses

Prototype behaviour
at large stresses
and when large
strains are
mobilised

Shear Strain
We need to create prototype stresses and strains in our models !!!

Centrifuge Modelling Technique

Consider, for simplicity, we have a block structure of dimensions
L B H sitting on a horizontal soil bed.

Principle of Centrifuge Modelling

In a centrifuge
if a reduced
d d scale
l model
d l is
i subjected
bj d to centrifugal
if l acceleration
l i so that
h
correct prototype stresses and strains are created in the model.
Stress under the block is given by
Centrifuge
Model

Idealised Field
Structure
1g

Ng
3

H/N

M/N

B/N
L/N

B
L

Mg
=
LB

L
L

Ng
Mg
N3
=
L B
LB
N
N

L / N
L/ N

L
L

Centrifugal Acceleration:
The easiest way to create high gravity is by spinning our soil models
in a centrifuge.
When the centrifuge is rotating with an angular velocity of , the
centrifugal acceleration at any radius r is given by

centrifugal acceleration = r 2
We wish to match this centrifugal acceleration to be the same factor
as the one we used to scale down our prototype by ie N
(geometrical scaling factor).

N g = r 2

For example, the Sudarshan Centrifuge at IIT, Bombay will

be doing;
In a 100g Centrifuge Test:

100 9.81 = 4.25

..RPM 145

Various centrifuges around the world

T
Turner
B
Beam Centrifuge
C t if
10 m diameter
di
t and
d 150 g-ton
t
capacity
it

University of Cambridge

University of California, Davis

IFSTTAR / LCPC Centrifuge - Nantes

The beam Centrifuge at LCPC

Radius 5.50 m
2T in the basket at 100 g
11

US Army Waterways Experiment Station

Takenaka Construction Corporation

Sudarshan National Geotechnical Centrifuge Facility, Bombay

Scaling Laws
We need a set of rules which link the behaviour
we observe in the model to that of an
equivalent prototype in the field
These rules are called Scaling
Scaling Laws
Laws
We already know that the Scaling Laws for
a) stress and strain
b) dimensions
c) mass
d) acceleration due to gravity

Scaling Laws
Parameter

Model/Prototype

Stress

St i
Strain

Length
g

1/N
/

Area

1/N2

Volume

1/N3

Mass

1/N3

Acceleration

Scaling law for force

By definition Force = Mass X
Acceleration
Using
Us
gd
dimensional
e s o a analysis:
a a ys s

Fprototype = M 1 a
Fmod el

M
= 3 (N a )
N

Fmodel
1
= 2
Fprototype N

Suppose we have a situation in the field where we are expecting

a 200 ton load on a pile cap.
cap
Let us say we want to model this load in a 100g centrifuge test
Then:

Fprototype = 200t = 2 MN
2 10
=
=
200
N
2
100
6

Fmod el

Thus we only need to apply 200 N or 20 kg(f) in the model

Scaling law for energy

By definition energy = work done = Force
X distance
Using
Us
gd
dimensional
e s o a analysis:
a a ys s

E prototype = F d
E mod el

F d F d
= 2 =
N N
N3

Emodel
1
= 3
E prototype N

Let us suppose
pp
an explosion
p
from a g
given device is
anticipated to yield 1 TJ of energy
p
in a 100g
g centrifuge
g test,, then
If we model this explosion

E prototype = 1
1 10 J
12

1 10
=
= 1000.kJ
3
100
12

Emod el

Thus,, we would need a much smaller explosion

p
in the
centrifuge test, that can be produced by few grams of
TNT !!!
Effects of blast on structures can be studied in this way

Scaling law for consolidation

Consolidation of clayey soils plays an important
part in geotechnical design
Long term settlement of many structures is a
function of the consolidation settlements
We can write the 33-D consolidation equation as
f ll
follows:
2u 2u 2u
u
= C v 2 + 2 + 2
t
x y z
We can identify the coefficient of consolidation Cv as an
i
important
t t parameter
t

We know from Soil Mechanics that degree of consolidation

is linked to co-efficient of consolidation Cv, drainage path d
and time of consolidation t.
Time factor Tv indicates the degree of consolidation and is
linked to the above parameters as follows:

t
Tv = C v . 2
d
We wish to achieve the same degree of consolidation in the
centrifuge model as that has occurred in the field
i.e. the Tv should be the same in the model and prototype

In the prototype:

(Tv ) prototype = (Cv ) prototype .

t prototype

(d )
2

prototype
t t

In the model:

(Tv )mod el

t mod el
= (Cv )mod el . 2
(d )mod el

Dividing the above equations with one another and noting

that on the LHS we will get 1 as we are aiming for same
degree of consolidation, we get;

(T v )mod el
(T v ) prototype

=1=

(C v )mod el
(C v ) prototype

t modd ell
.

t prototype

(d )
(d )
2

mod el

prototype

If we use the same soil in the centrifuge model as that is

present in the field, then we will have the same co-efficient
Cv both in the model and the prototype, therefore:

(Cv )mod el
(Cv ) prototype

=1

Therefore, the time of consolidation in the model and the

prototype are related by:

t mod el
t prototype

d mod el
1
= 2
=
d

N
prototype

Let us suppose that

L
h a 10m
10 layer
l
off Marine
M i
clay
l
off
ff Mumbai
M
b i shore
h
takes about 10 years to reach 95% consolidation
In a 100g test the same cla
clay will
ill reach
each 95% consolidation in

t mod el

10 365 24 60 60
=
s
2
100

t mod el = 31536 s = 8 . 8 . hours

This is very attractive to us, as we can now start to model
l
long
term consolidation
ld
events b
by running the
h centrifuge
f
ffor
relatively short periods of time.

Thus Centrifuge Acts as a Time Machine !!!

Advantages of Centrifuge Modelling

We create CORRECT prototype stresses and strains in

the centrifuge model,
model thereby the soil will mobilise the
RIGHT stiffness and TRUE behaviour is captured
Using centrifuge modelling new materials with unknown
stress--strain relationships or known materials with new
stress
stress paths can be tested to reproduce prototype
behaviour
The big advantage of Centrifuge Modelling is that the
TRUE failure
f l
mechanism
h
will
ll b
be captured
d
The Centrifuge Modelling can model TRUE Three
Three-Dimensional Behaviour (say in Pile Groups), which can
be very expensive and time consuming using FE
analysis
y
even today
y

Soil-Structure Interaction is automatically accounted for

Soilin centrifuge models
Long term events such as consolidation settlements can
be modelled in relatively short durations in a centrifuge
test.
test This is also true for other diffusion processes
Forces required in the centrifuge model are small
compared to the prototype, so centrifuge model tests
are cheap compared to the field testing
Energy scaling law dictates that we can model events
such as blast loading using very small charges in
centrifuge
t if
models.
d l