MIT - 16.20
Fall, 2002
Unit 4
Equations of Elasticity
Readings
:
R
2.3, 2.6, 2.8
T & G
84, 85
B, M, P
5.1-5.5, 5.8, 5.9
7.1-7.9 6.1-6.3, 6.5-6.7
Jones
(as background on composites)
Paul A. Lagace, Ph.D.
Professor of Aeronautics & Astronautics
and Engineering Systems
Paul A. Lagace
? 2001
MIT - 16.20
Fall, 2002
Let
’s first review a bit
…
from Unified, saw that there are 3 basic considerations in elasticity:
1.
Equilibrium
2.
Strain - Displacement
3.
Stress - Strain Relations (
Constitutive
Relations)
Consider each:
1.
Equilibrium
(3)
?
Σ
F
i
= 0,
Σ
M
i
= 0
?
Free body diagrams
?
Applying these to an infinitesimal element yields
3
equilibrium equations
Figure 4.1
Representation of general infinitesimal
element
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
MIT - 16.20
Fall, 2002
?
σ
11
+
?
σ
21
+
?
σ
31
+
f
1
=
0
(4-1)
?
y
1
?
y
2
?
y
3
?
σ
12
+
?
σ
22
+
?
σ
32
+
f
2
=
0
(4-2)
?
y
1
?
y
2
?
y
3
?
σ
13
+
?
σ
23
+
?
σ
33
+
f
3
=
0
(4-3)
?
y
1
?
y
2
?
y
3
?
?
+
σ
mn
m
n
y
f
0
=
2.
Strain - Displacement
(6)
?
Based on geometric considerations
?
Linear considerations
(
I.e.,
small
strains only --
we will talk about
large strains later
)
(and
infinitesimal
displacements only)
Paul A. Lagace
? 2001
Unit 4 -
p
. 3
MIT - 16.20
Fall, 2002
ε
11
=
?
u
1
(4-4)
?
y
1
ε
22
=
?
u
2
(4-5)
?
y
2
ε
33
=
?
u
3
(4-6)
?
y
3
ε
21
=
ε
12
=
1
??
?
u
1
+
?
u
2
??
2
? ?
?
y
2
?
y
1
? ?
ε
31
=
ε
13
=
1
??
?
u
1
+
?
u
3
??
2
? ?
?
y
3
?
y
1
? ?
ε
32
=
ε
23
=
1
??
?
u
2
+
?
u
3
??
2
? ?
?
y
3
?
y
2
? ?
(4-7) (4-8) (4-9)
1
??
?
u
m
+
?
u
n
??
ε
mn
=
2
? ?
?
y
n
?
y
m
? ?
Paul A. Lagace
? 2001
Unit 4 -
p
. 4
MIT - 16.20
Fall, 2002
3.
Stress - Strain
(6)
σ
mn
=
E
mnpq
ε
pq
we’
ll come back to this
…
Let
’s review the
“
4th important concept
”:
Static
Determinance
There are there possibilities
(as noted in U.E.)
a.
A structure is not sufficiently restrained
(fewer reactions than d.o.f.)
degrees of freedom
?
DYNAMICS
b.
Structure is exactly (or
“
simply
”
)
restrained
(# of reactions = # of d.o.f.)
?
STATICS (statically
determinate)
Implication
:
can calculate stresses via
equilibrium (
as done in Unified
)
Paul A. Lagace
? 2001
Unit 4 -
p
. 5
MIT - 16.20
Fall, 2002
c.
Structure is overrestrained
(# reactions > # of d.o.f.)
?
STATICALLY
INDETERMINATE
…must solve for reactions simultaneously
with stresses, strains, etc.
in this case, you must employ the stress-strain equations
--> Overall, this yields for elasticity:
15 unknowns
and
15
equations
6 strains =
ε
mn
3 equilibrium (
σ
)
6 stresses =
σ
mn
6 strain-displacements (
ε
)
3 displacements = u
m
6 stress-strain (
σ
-
ε
)
IMPORTANT
POINT:
The first two sets of equations are
“universal
”
(independent of the
material) as they depend on geometry (strain-displacement) and equilibrium (equilibrium).
Only
the stress-strain equations are
dependent on the material.
Paul A. Lagace
? 2001
Unit 4 -
p
. 6
??
??
MIT - 16.20
Fall, 2002
One
other
point
:
Are all these equations/unknowns independent?
N
O
Why?
-->
Relations between the strains and displacements (due to
geometrical considerations result in the
Strain
Compatibility
Equations
(as you saw in Unified)
General form is:
?
2
ε
nk
+
?
2
ε
m
l
?
?
2
ε
n
l
?
?
2
ε
mk
=
0
yy
l
??
y
k
yy
k
??
y
l
m
y
n
m
y
n
This results in
6
strain-compatibility (in 3-D).
What a mess!!!
What do these really tell us???
The strains must be
compatible
, they cannot be prescribed in
an arbitrary fashion.
Let
’s consider an
example:
Step 1
:
consider how shear strain (
ε
12
) is related to displacement:
1
?
?
u
1
?
u
2
?
?
+
?
ε
12
=
2
? ?
?
y
2
?
y
1
? ?
Paul A. Lagace
? 2001
Unit 4 -
p
. 7
??
??
??
MIT - 16.20
Fall, 2002
Note that deformations (u
m
) must be continuous
single-valued functions for continuity. (or it doesn
’
t
make
physical
sense
!)
Step 2
:
Now consider the case where there are gradients in the strain
field
?
ε
12
≠
0
,
?
ε
12
≠
0
?
y
1
?
y
2
This is the most general case and most likely in a general structure Take derivatives on both sides: ?
2
ε
12
1
?
?
3
u
1
?
3
u
2
?
?
=
2
+
2
yy
2
2
? ?
yy
2
yy
2
? ?
1
1
1
Step 3
:
rearrange slightly and recall other strain-displacement
equations
?
u
1
=
ε
1
,
?
u
2
ε
2
=
?
y
1
?
y
2
Paul A. Lagace
? 2001
Unit 4 -
p
. 8
??
MIT - 16.20
Fall, 2002
?
2
ε
12
1
?
?
2
ε
11
+
?
2
ε
22
?
?
=
2
yy
2
2
? ?
?
y
2
?
y
1
2
??
1
So, the gradients in strain are related in certain ways since they are all related to the 3 displacements.
Same for other 5 cases
…
Let
’s now go back and spend time with the
…
Stress-Strain Relations and the Elasticity
Tensor
In Unified, you saw particular examples of this, but we now want to generalize it to encompass all cases. The basic relation between force and displacement (recall 8.01) is Hooke
’s
Law:
F =
kx
spring constant (linear case)
Paul A. Lagace
? 2001
Unit 4 -
p
. 9
MIT - 16.20
Fall, 2002
If this is extended to the three-dimensional case and applied over infinitesimal areas and lengths, we get the relation between stress and strain known as:
Generalized
Hooke
’
s
law:
σ
mn
=
E
mnpq
ε
pq
where E
mnpq
is the
“elasticity
tensor
”
How many components does this appear to have?
m, n, p, q = 1, 2, 3 ?
3 x 3 x 3 x 3 = 81 components
But there are several symmetries:
1.
Since
σ
mn
=
σ
nm
(energy considerations)
?
E
mnpq
= E
nmpq
(symmetry in switching
first two
indices
)
2.
Since
ε
pq
=
ε
qp
(geometrical considerations)
?
E
mnpq
= E
mnqp
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
0
2212
MIT - 16.20
Fall, 2002
(symmetry in switching
last
two
indices
)
3.
From thermodynamic considerations
(1st law of thermo)
?
E
mnpq
= E
pqmn
(symmetry in switching
pairs of
indices
)
Also note that:
Since
σ
mn
=
σ
nm
are only 6!
, the apparent 9 equations for stress
With
these symmetrics, the resulting equations are:
?
σ
11
?
?
E
1111
E
1122
E
1133
2
E
1123
2
E
1113
2
E
1112
?
?
ε
11
?
?
?
?
?
?
??
E
1122
E
2222
E
2233
2
E
2223
2
E
2213
2
E
2212
?
?
ε
22
?
?
σ
22
?
?
σ
33
?
?
E
1133
E
2233
E
3333
2
E
3323
2
E
3313
2
E
3312
?
?
ε
33
?
?
?
=
?
?
?
?
?
σ
23
?
?
E
1123
E
2223
E
3323
2
E
2323
2
E
1323
2
E
1223
?
?
ε
23
?
?
σ
13
?
?
E
1113
E
2213
E
3313
2
E
1323
2
E
1313
2
E
1213
?
?
ε
13
?
?
?
?
?
?
?
?
σ
12
?
?
E
1112
E
2212
E
3312
2
E
1223
2
E
1213
2
E
1212
?
?
ε
12
?
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
1
MIT - 16.20
Fall, 2002
Results in
21
independent components of the elasticity tensor
?
Along diagonal (6)
?
Upper right half of matrix (15)
[don
’
t
worry about 2
’s]
Also
note
:
2
’s come out automatically
…
don
’
t
put them in
ε~
For example:
σ
12
=
… E
1212
ε
12
+ E
1221
ε
21
…
= …
2E
1212
ε
12
…
These E
mnpq
can be placed into
3 groups
:
?
Extensional strains to extensional stresses
E
1111
E
1122
E
2222
E
1133
E
3333
E
2233
e.g.,
σ
11
= … E
1122
ε
22
…
?
Shear strains to shear stresses
E
1212
E
1213
E
1313
E
1323
E
2323
E
2312
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
2
MIT - 16.20
Fall, 2002
e.g.,
σ
12
= … 2E
1223
ε
23
…
?
Coupling
term
:
extensional strains to shear stress
or
shear strains to extensional stresses
E
1112
E
2212
E
3312
E
1113
E
2213
E
3313
E
1123
E
2223
E
3323
e.g.,
σ
12
= …E
1211
ε
11
…
11
= …2E
1123
ε
23
…
σ
A material which behaves in this manner is
“fully”
anisotropic
However, there are
currently
no useful engineering materials which
have
21
different and independent components of
E
mnpq
The
“
type
”
of material (with regard to elastic behavior) dictates the number
of
independent
components of
E
mnpq
:
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
3
MIT - 16.20
Fall, 2002
2
Isotropic
3
Cubic
5
“Transversely Isotropic”*
6
Tetragonal
9
Orthotropic
13
Monoclinic
21
Anisotropic
# of Independent
Components of E
mnpq
Material Type
Useful
Engineering
Materials
Composite
Laminates
Basic
Composite
Ply
Metals
(on average)
Good Reference
:
BMP, Ch. 7
*not in BMP
For
orthotropic
material
s (which is as complicated as we usually get),
there are no coupling terms in the
principal
axes
of
the
material
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
4
MIT - 16.20
Fall, 2002
?
When you apply an extensional stress, no shear strains arise
e.g., E
1112
= 0
(total of 9 terms are now zero)
?
When you apply a shear stress, no extensional strains arise
(some terms become zero as for
previous condition)
?
Shear strains (stresses) in one plane do not cause shear strains (stresses) in another plane
( E
1223
, E
1213
, E
1323
=
0)
With these additional terms zero, we end up with
9
independent
components:
(21 - 9 - 3 = 9)
and the equations are:
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
5
???
13
MIT - 16.20
Fall, 2002
?
σ
11
?
?
E
1111
E
1122
E
1133
0
0
0
?
?
ε
11
?
?
?
??
E
1122
E
2222
E
2233
0
0
0
?
?
?
?
σ
22
?
?
?
ε
22
?
?
σ
33
?
?
E
1133
E
2233
E
3333
0
0
0
?
?
ε
33
?
?
?
=
?
?
?
?
σ
23
?
? ?
0
0
0
2
E
2323
0
0
?
?
ε
23
?
?
σ
13
?
?
0
0
0
0
2
E
1313
0
?
?
ε
13
?
?
?
?
?
?
?
σ
12
?
? ?
0
0
0
0
0
2
E
1212
?
?
ε
12
?
For other cases,
no
more
terms
become
zero
, but the terms are not
Independent.
For example, for isotropic materials:
?
E
1111
= E
2222
= E
3333
?
E
1122
= E
1133
= E
2233
?
E
2323
= E
1313
= E
1212
?
And there is one other equation relating E
1111
, E
1122
and E
2323
?
2 independent components of
E
mnpq
(we’ll see this more when we do engineering constants)
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
6
MIT - 16.20
Fall, 2002
Why, then, do we bother with anisotropy?
Two
reasons
:
1.
Someday, we may have useful fully
anisotropic
materials
(certain crystals now behave that way) Also, 40-50 years ago, people only worried about isotropy
2.
It may not always be convenient to describe a structure (i.e., write the governing equations) along the principal material axes.
How else?
Loading
axes
Examples
Figure 4-2
wing
rocket case
fuselage
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
7
MIT - 16.20
Fall, 2002
In these other axis systems, the material may have
“more
”
elastic
components.
But it really does
n
’
t
.
(you can’
t “
create
”
elastic components just by describing a material in
a different axis system, the inherent properties of the material
stay
the
same
).
Figure 4-3
Example
:
Unidirectional composite (transversely isotropic)
No shear / extension coupling
Shears
with
regard
to
loading
axis
but still no
inherent
shear/extension coupling
In order to describe full behavior, need to do …TRANSFORMATIONS
(we’ll review this/expand on it later)
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
8
MIT - 16.20
Fall, 2002
--> It is often useful to consider the relationship between stress and strain (opposite way). For this we use
COMPLIANCE
ε
mn
= S
mnpq
σ
pq
where:
S
mnpq
=
compliance
tensor
Paul A. Lagace
? 2001
Unit 4 -
p
. 1
9
MIT - 16.20
Fall, 2002
Using matrix notation:
σ
= E
ε
~~
~
and
E
-1
σ
~
~
=
ε
~
inverse
with
ε
= S
σ
~
~~
this means
E
-1
= S
~
~
?
E S =
I
~~
~
?
The compliance matrix is the
inverse of the elasticity matrix
Note
:
the same symmetries apply to
S
mnpq
as
to
E
mnpq
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
0
MIT - 16.20
Fall, 2002
Meaning of each:
?
Elasticity term
E
mnpq
:
amount of stress (
σ
mn
) related to the
deformation/strain (
ε
pq
)
?
Compliance term
S
mnpq
:
amount of strain (
ε
mn
) the stress (
σ
pq
)
causes
These are useful in defining/ determining the
“engineering
constants
”
All of this presentation on elasticity
(and what you had in
Unified
) is based on assumptions which limit their
applicability:
which we will review / introduce / expand on in the next lecture.
CAUTION
?
Small strain
?
Small displacement / infinitesimal (linear) strain
Fortunately, most engineering structures are such that these assumptions cause negligible error.
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
1
MIT - 16.20
Fall, 2002
However, there are cases where this is not true:
?
Manufacturing (important to be able to convince)
?
Compliant materials
?
Structural examples:
dirigibles,
…
So let
’s explore:
Large strain and the formal definition of strain
What we defined before are the physical manifestation of strain /
deformation
?
Relative elongation
?
Angular rotation
Strain is formally defined by considering the diagonal length of a cube:
Figure 4-4
undeformed
x
3
(small letters)
x
2
Paul A. Lagace
? 2001
x
1
Unit 4 -
p
. 2
2
MIT - 16.20
Fall, 2002
and looking at the change in length under general (and possibly large) deformation:
Figure 4-5
deformed
(capital letters)
x
3
x
2
x
1
The formal definition of the strain tensor is:
2
2
2
γ
mn
dx
m
dx
n
=
(
dS
)
-
(
ds
)
?
2
γ
11
dx
1
dx
1
+
2
γ
2
2
dx
2
dx
2
+
2
d
γ
3
3
d
x
3
d
x
3
+
2
(
γ
12
+
γ
21
)
dx
1
dx
2
+
2
(
γ
13
+
γ
31
)
dx
1
dx
3
2
2
+
2
(
γ
23
+
γ
32
)
dx
2
dx
3
=
(
dS
)
?
(
ds
)
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
3
MIT - 16.20
Fall, 2002
where
γ
mn
= formal strain tensor.
This is a
definition
. The physical interpretation is related to
this but not directly in the general case.
One can show (see BMP 5.1 - 5.4) that the formal strain tensor is related to relative elongation (the familiar
?
l
) via:
l
relative elongation in m-direction:
E
m
=
m
1
γ
m
+
2
?
1
(no summation on m)
and is related to angular change via:
2
γ
mn
sin
φ
=
mn
(
1
+
E
m
)
(
1
+
E
n
)
Thus, it also involves the relative elongations!
Most structural cases deal with relatively small strain. If the relative elongation is small (<<100%)
?
E
m
<<1
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
4
MIT - 16.20
Fall, 2002
look at:
E
m
=
1
+
2
γ
m
m
?
1
2
?
(
E
m
+
1
)
=
1
+
2
γ
m
m
E
2
+
2
E
m
=
2
γ
m
m
m
but if
E
m
<< 1,
then E
2
m
≈
0
?
E
m
=
γ
mm
Relative elongation = strain
?
l
=
ε
small strain approximation!
l
Can assess this effect by comparing 2E
m
and
E
m
(2 + E
m
)
relative elongation =
E
m
2E
m
E
m
(2 + E
m
)
% error
0.01
0.02
0.0201
0.5%
0.02
0.04
0.0404
1.0%
0.05
0.10
0.1025
2.4%
0.10
0.20
0.2100
4.8%
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
5
MIT - 16.20
Fall, 2002
Similarly, consider the general expression for rotation:
2
γ
mn
sin
φ
=
mn
(
1
+
E
m
)
(
1
+
E
n
)
for small elongations (
E
m
<< 1, E
n
<< 1)
?
sin
φ
mn
=
2
γ
mn
and
, if the rotation is small:
sin
φ
mn
≈
φ
mn
?
φ
mn
=
2
γ
mn
=
2
ε
mn
small strain approximation!
(as before)
Note
:
factor of 2 !
Even for a balloon, the small strain approximation may be good enough
So
:
from now on, small strain assumed, but
?
understand limitations
?
be prepared to deal with large strain
?
know difference between formal definition and the engineering
approximation which relates
dire
ctly
to physical reality.
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
6
MIT - 16.20
Fall, 2002
What is the other limitation? It deals with displacement, so consider
Large Displacement and Non-Infinitesimal (Non-linear) Strain
See BMP 5.8 and 5.9
The general strain-displacement relation is:
1
?
?
u
?
u
?
u
?
u
?
γ
mn
=
?
m
+
n
+
r
s
δ
rs
?
2
?
?
x
n
?
x
m
?
x
m
?
x
n
?
Where:
δ
rs
=
Kronecker
delta
The latter terms are important for larger displacements but are higher order for small displacements and can then be ignored to arrive back at:
1
??
?
u
m
+
?
u
n
??
ε
mn
=
2
? ?
?
y
n
?
y
m
? ?
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
7
MIT - 16.20
Fall, 2002
How to assess?
Look at
?
u
m
?
u
r
?
u
s
vs.
δ
rs
?
x
n
?
x
m
?
x
n
and compare magnitudes
Small
v
s
.
large
and
linear vs. nonlinear will depend on:
?
material(s)
?
structural configuration
?
mode of behavior
?
the loading
Examples
?
Rubber in inflated structures
?
Large strain (
Note
:
generally means larger displacement)
?
Diving board of plastic or wood
?
Small strain
but
possibly large displacement (will look at this
more when we deal with beams)
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
8
MIT - 16.20
Fall, 2002
?
Floor beam of steel
?
Small strain and linear strain (
Note
: linear strain must also be
?
small)
Next…back to constitutive constants
…now their physical
reality
Paul A. Lagace
? 2001
Unit 4 -
p
. 2
9