MIT - 16.20
Fall, 2002
Unit 23
Vibration of Continuous Systems
Paul A. Lagace, Ph.D.
Professor of Aeronautics & Astronautics
and Engineering Systems
Paul A. Lagace
? 2001
MIT - 16.20
Fall, 2002
The logical extension of discrete mass systems is one of an infinite number of masses. In the limit, this is a
continuous
system
.
Take the generalized beam-column as a generic representation:
2
d
2
?
EI
dw
?
dx
2
??
dx
2
? ?
?
d
??
F
dw
? ?
=
p
z
(23-1)
dx
?
dx
?
Figure 23.1
Representation of generalized beam-column
dF
=
?
p
x
dx
This considers only static loads. Must add the inertial load(s). Since the concern is in the z-displacement (w):
˙˙
Inertial
load
unit
length
=
mw
(23-2)
where:
m(x) = mass/unit length
Paul A. Lagace
? 2001
Unit 23 -
2
w
MIT - 16.20
Fall, 2002
Use
per
unit
length
since entire equation is of this form.
Thus:
2
d
2
?
EI
dw
?
dx
2
??
dx
2
? ?
?
d
??
F
dw
? ?
=
p
z
?
m
˙˙
dx
?
dx
?
or:
2
d
2
?
EI
dw
?
˙˙
dx
2
??
dx
2
? ?
?
d
??
F
dw
? ?
+
mw
=
p
z
(23-3)
dx
?
dx
?
Beam Bending Equation
often, F
=
0 and this becomes:
2
d
2
2
??
EI
dw
?
+
mw
=
p
z
dx
?
dx
2
??
˙˙
-->
This is a
fourth
order differential equation in x
-->
Need
four
boundary conditions
-->
This is a
second
order differential equation in time
-->
Need
two
initial conditions
Paul A. Lagace
? 2001
Unit 23 -
3
w
it
MIT - 16.20
Fall, 2002
Notes
:
?
Could also get via simple beam equations. Change occurs in:
dS
=
p
z
?
m
˙˙
dx
?
If consider dynamics along x, must include
mu
˙˙
in
p
x
term: (
p
x
?
m
u
˙˙
)
Use the same approach as in the discrete spring-mass systems:
Free Vibration
Again assume harmonic motion. In a continuous system, there are an infinite
number of natural frequencies (
eigenvalues
) and associated
modes (eigenvectors)
so:
ω
wx
(,
t
)
=
w
(
x
)
e
separable solution spatially (x) and temporally (t)
Consider the homogeneous case (
p
z
= 0) and let there be no axial forces
(p
x
= 0
?
F = 0)
Paul A. Lagace
? 2001
Unit 23 -
4
it
it
MIT - 16.20
Fall, 2002
So:
2
d
2
?
EI
dw
?
+
mw
=
0
dx
2
??
dx
2
??
˙˙
Also assume that EI does not vary with x:
4
˙˙
EI
dw
+
mw
=
0
(23-5)
dx
4
Placing the assumed mode in the governing equation:
4
EI
d
w
e
ω
?
m
ω
2
w
e
ω
=
0
dx
4
This gives:
4
EI
d
w
?
m
ω
2
w
=
0
(23-6)
dx
4
which is now an equation solely in the spatial variable (successful separation of t and x dependencies)
_
Must now find a solution for w(x) which satisfies the differential equations and the boundary conditions.
Note
:
the shape and frequency are
intimately
linked
(through equation 23-6)
Paul A. Lagace
? 2001
Unit 23 -
5
MIT - 16.20
Fall, 2002
Can recast equation (23-6) to be:
4
d
w
m
ω
2
dx
4
?
w
=
0
(23-7)
EI
The solution to this homogeneous equation is of the form:
wx
()
=
e
px
Putting this into (23-7) yields
4
px
m
ω
2
pe
?
e
px
=
0
EI
4
m
ω
2
?
p
=
EI
So this is an
eigenvalue
problem (spatially). The four roots are:
p = +
λ
, -
λ
, +
i
λ
, -
i
λ
where:
14
?
m
ω
2
?
/
λ
=
?
?
?
EI
?
This yields:
Paul A. Lagace
? 2001
Unit 23 -
6
MIT - 16.20
Fall, 2002
λ
wx
()
=
A
e
λ
x
+
B
e
?
λ
x
+
C
e
i
λ
x
+
D
e
?
ix
or
:
wx
()
=
C
1
sinh
λ
x
+
C
2
cosh
λ
x
+
C
3
sin
λ
x
+
C
4
cos
λ
x
(23-8)
The constants are found by applying the boundary conditions
(4 constants
?
4 boundary conditions)
Example
:
Simply-supported beam
Figure 23.2
Representation of simply-supported beam
E
I
, m = constant with x
4
2
EI
dw
+
m
dw
=
p
z
dx
4
dt
2
Paul A. Lagace
? 2001
Unit 23 -
7
MIT - 16.20
Fall, 2002
Boundary conditions:
@ x = 0
w = 0
2
@ x =
l
M
=
E
I
dw
=
0
dx
2
with:
wx
()
=
C
1
sinh
λ
x
+
C
2
cosh
λ
x
+
C
3
sin
λ
x
+
C
4
cos
λ
x
Put the resulting four equations in matrix form
w
0
2 ()
=
0
?
0
1
0
1
?
?
C
1
?
?
0
?
dw
dx
2
()
=
0
??
0
1
0
?
1
?
?
?
?
?
0
?
C
2
?
?
0
?
?
?
?
=
?
?
()
=
0
?
sinh
λ
l
cosh
λ
l
sin
λ
l
cos
λ
l
?
?
C
3
?
?
0
?
wl
dw
? ?
sinh
λ
l
cosh
λ
l
?
sin
λ
l
?
cos
λ
l
? ?
?
C
4
?
?
0
?
2
?
?
?
?
l
dx
2
()
=
0
Solution of determinant matrix generally yields values of
λ
which then
yield frequencies and associated modes (as was done for multiple mass systems in a
somewhat
similar
fashion)
Paul A. Lagace
? 2001
Unit 23 -
8
MIT - 16.20
Fall, 2002
In this case, the determinant of the matrix yields:
C
sin
λ
l
=
0
3
Note
:
Equations (1 & 2) give C
2
= C
4
= 0
Equations (3 & 4) give
2
C
3
sin
λ
l
=
0
?
nontrivial:
λ
l
= n
π
The nontrivial solution is:
λ
l
= n
π
(eigenvalue problem!)
Recalling that:
14
?
m
ω
2
?
/
λ
=
?
?
?
EI
?
4
m
ω
2
n
π
4
(change n to r to be consistent with
?
=
EI
l
4
previous notation)
22
?
ω
r
=
r
π
EI
ml
4
<-- natural frequency
Paul A. Lagace
? 2001
Unit 23 -
9
rx
rx
MIT - 16.20
Fall, 2002
As before, find associated mode (eigenvector), by putting this back in the governing matrix equation. Here (setting C
3
= 1
…..one
“
arbitrary
”
magnitude):
π
wx
()
=
φ
r
=
sin
<-- mode shape (normal mode)
l
for:
r = 1, 2, 3,……
∞
Note
:
A continuous system has an
infinite
number of
modes
So total solution is:
π
?
22
EI
ml
4
?
wx
(,
t
)
=
φ
r
sin
ω
r
t
=
sin
sin
?
r
π
t
?
l
?
?
--> Vibration modes and frequencies are:
Paul A. Lagace
? 2001
Unit 23 -
1
0
MIT - 16.20
Fall, 2002
Figure 23.3
Representation of vibration modes of simply-supported beam
1st mode
ω
=
π
2
1
4
EI
ml
2nd mode
ω
=
4
π
2
2
4
EI
ml
3rd mode
ω
=
9
π
2
3
4
EI
ml
etc.
Same for other cases
Paul A. Lagace
? 2001
Unit 23 -
1
1
MIT - 16.20
Fall, 2002
Continue to see the similarity in results between continuous and multi-mass (degree-of-freedom) systems. Multi-mass systems have predetermined modes since
discretization
constrains system to deform
accordingly. The extension is also valid for
…
Orthogonality
Relations
They take basically the same form except now have
continuous
functions
integrated
spatially over the regime of interest
rather
than
vectors
:
l
∫
0
mx
x
x
()
φ
r
()
φ
s
()
dx
=
M
r
δ
r
s
(23-9)
= 1
for
r = s
where:
δ
rs
= kronecker delta
= 0
for
r
≠
s
l
()
φ
r
2
()
d
x
M
r
=
∫
0
m
x
x
generalized mass of the
rth mode
Paul A. Lagace
? 2001
Unit 23 -
1
2
MIT - 16.20
Fall, 2002
So:
l
∫
0
m
φφ
s
d
x
=
0
r
≠
s
r
l
∫
0
m
φφ
r
d
x
=
M
r
r
Also can show (similar to multi degree-of-freedom case):
2
l
d
2
? ?
EI
d
dx
φ
2
r
?? ?
φ
s
dx
=
δ
rs
M
r
ω
r
2
(23-10)
∫
0
dx
2
?
This again, leads to the ability to transform the equation based on the normal modes to get the
…
Normal Equations of Motion
Let:
∞
wx
(,
t
)
=
∑
φ
r
(
x
)
ξ
r
()
t
(23-11)
r
=
1
normal mode
normal coordinates
Paul A. Lagace
? 2001
Unit 23 -
1
3
px
px
MIT - 16.20
Fall, 2002
Place into governing equation:
2
2
d
2
2
??
EI
dw
?
+
m
dw
=
z
()
dx
?
dx
2
??
dt
2
l
multiply by
φ
s
and integrate
∫
0
dx
to get:
∞
˙˙
∞
d
2
?
2
∑
ξ
r
∫
0
l
m
φ
r
φ
s
d
x
+
∑
ξ
r
∫
0
l
φ
s
dx
2
? ?
EI
d
dx
φ
2
r
???
dx
=
∫
0
l
φ
s
f
d
x
r
=
1
r
=
1
Using
orthogonality
conditions, this takes on the same forms as
before:
˙˙
M
ξ
+
M
ω
2
ξ
=
Ξ
(23-12)
rr
r
r
r
r
r
= 1, 2, 3,……
∞
l
with:
M
r
=
∫
0
m
φ
r
2
d
x
- Generalized mass of
rth mode
l
Ξ
r
=
∫
0
φ
r
z
(,
t
)
d
x
- Generalized force of
rth mode
ξ
r
(t)
=
normal coordinates
Paul A. Lagace
? 2001
Unit 23 -
1
4
MIT - 16.20
Fall, 2002
Once
again
?
each equation can be solved independently
?
allows continuous system to be treated as a series of
“simple
”
one degree-of-freedom systems
?
superpose solutions to get total response (
Superposition
of
Normal
Modes
)
?
often only lowest modes are important
?
difference from multi degree-of-freedom system:
n -->
∞
-->
To find Initial Conditions in normalized coordinates
…
same
as
before:
(,
0
)
=
∑
φ
r
(
x
)
ξ
r
()
wx
0
r
etc.
Thus:
1
l
0
ξ
r
()
=
M
∫
0
m
φ
r
w
0
(
x
)
d
x
r
(23-13)
˙
r
()
=
1
∫
0
l
m
φ
r
w
˙
0
()
d
x
ξ
0
x
M
r
Paul A. Lagace
? 2001
Unit 23 -
1
5
MIT - 16.20
Fall, 2002
Finally, can add the case of
…
Forced Vibration
Again
… response is made up of the natural modes
?
Break up force into series of spatial impulses
? Use Duhamel
’
s
(convolution) integral to get response for each
normalized mode
ξ
ω
τ
τ
τ
r
rr
r
t
M
t
t
(
)
=
(
)
?
(
)
∫
1
0
Ξ
sin
ω
r
d
(23-14)
?
Add up responses (equation 23-11) for all normalized modes
(Linear
?
Superposition)
What about the special case of
…
-->
Sinusoidal
Force
at
point
x
A
Paul A. Lagace
? 2001
Unit 23 -
1
6
xF
MIT - 16.20
Fall, 2002
Figure 23.4
Representation of force at point
x
A
on simply-supported
beam
Ft
()
=
F
o
sin
?
t
As for single degree-of-freedom system, for each normal mode get:
t
A
ξ
r
()
=
φ
r
()
o
2
sin
?
t
M
r
ω
r
2
? ?
1
?
?
2
? ?
?
ω
?
r
for steady state response (Again, initial transient of sin
ω
r
t dies out
due to damping)
Add
up all responses
…
Paul A. Lagace
? 2001
Unit 23 -
1
7
MIT - 16.20
Fall, 2002
Note
:
?
Resonance can occur at any
ω
r
? DMF (Dynamic Magnification Factor) associated with each normal
mode
--> Can apply technique to any system.
?
Get governing equation
including
inertial
terms
? Determine Free Vibration Modes and frequencies ? Transform equation to uncoupled single degree-of-freedom system
(normal equations)
? Solve each normal equation separately ? Total response equal to sum of individual responses
Modal superposition is a very powerful technique!
Paul A. Lagace
? 2001
Unit 23 -
1
8