Professional Documents
Culture Documents
is of interest.
While improved data
acquisition and parameter
extraction
techniques can improve results for some
structures, the more fundamental approach
taken here is to question whether the
assumptions made in the development of
modal theory are valid when considering the
response of some large structures or
structures with high damping.
The intent of this paper is to present
a dynamic model of the response of a finite
Timoshenko beam supported by fixed-fixed
boundary conditions and subject to an
impulse excitation.
The assumptions
regarding the separation of the temporal and
spatial response made in a modal approach
need not be made in the development of this
model. In this manner following an impulse
excitation the response can be observed as
a propagating waves and transitions to a
modal response.
In this paper we will first have a
discussion of the motivation for constructing
this model. Secondly, the development of
the model will be explained. Last is a
description of how the model was validated
and confirmed to adhere to known behavior
of vibrating structures.
Background
This
section
discusses
the
motivation to develop the model presented
in this paper, as well as a comparison of
how real normal modes are realized in
modal analysis and wave propagation.
Justification is also given for the use of the
Timoshenko beam model.
Motivation
The primary purpose of the model
presented in this paper is to explore if the
assumptions made in the development of
modal analysis are valid for very large
structures.
Experimental modal tests
performed on very large structures such as
spacecraft, buildings, and bridges have
proved to be a challenge.
Traditional
excitation techniques may not be able to
adequately excite the structure resulting in
poor coherence in the measurements.
Isolating large structures with either very stiff
or very soft interfaces, as is often done with
small test articles, is often impossible. As a
result, accurate representation of the
boundary conditions can be difficult.
Channel counts are often rather high on
tests of large structures making it
cumbersome to manage all the data. In the
case of these problems the difficulty is in the
test setup.
Some have made efforts to
improve techniques with the aim of
improving the results in these situations.
However, the goal of this effort is to consider
a more fundamental obstacle to modal
testing of large structures, the ability of the
structure to physically realize real normal
modes. If true, modal analysis would not be
an appropriate method of characterizing the
vibration of these structures and wave
propagation methods may be more suitable.
Real Normal Modes of Vibration
In order to consider if modal
analysis is appropriate for very large
structures, an understanding about real
normal modes must first be developed
relative to wave propagation theory and
modal analysis theory.
Modal analysis
theory will be considered first.
In this
discussion the modes of a beam in bending
are considered for simplicity, but the
principles extend to more complex
structures.
Interestingly enough, even though
the dynamic responses of structures consist
of the summation of many propagating
waves, modal theory can bypass all those
details. The modal solution is based on the
roots of the equations of motion. Using an
eigenproblem approach, as is commonly
done, the natural frequencies are found by
the square root of the eigenvalues, and the
mode shapes are scaled versions of the
eigenfunctions. These natural frequencies
and mode shapes are the basis of forming
the system response when using the modal
superposition method. The response is
composed of a weighted summation of the
mode shapes. The weighting functions are
a function of time. This weighted summation
is where the separation of variables is
employed as a part of the assumed solution
= q ( x, t )
2
x
t 2
x
2 ( x, t )
2 (x, t )
y ( x, t )
GA
= I
( x, t ) + EI
2
x
t 2
where
x=
t=
y(x,t) =
(x,t) =
shear
G=
A=
E=
I=
=
position
time
transverse displacement
slope of the centroidal axis due to
shear modulus
cross-sectional area
elastic modulus
second moment of area
Timoshenko shear coefficient
=
q(x,t) =
t 2
( x, t ) 2 y ( x, t )
x 2
x
2
2 ( x, t ) GA y ( x, t )
E ( x, t )
=
( x, t ) +
2
2
I x
t
x
z1 = y
z3 =
y
t
z4 =
t
z2 =
A
x
t
x 2
z3 ( x, t )
= z 4 ( x, t )
t
2
z4 ( x, t ) GA z1 ( x, t )
E z 3 ( x, t )
=
z 3 ( x, t ) +
2
I x
t
x
Applying
the
discrete
spatial
variable and substituting the derivative
2h
h2
A
t
z3 (xi , t )
= z 4 ( xi , t )
t
z 4 (xi , t ) GA z1 (xi +1 , t ) z1 (xi 1 , t )
E z3 ( xi+1 , t ) 2 z3 ( xi , t ) + z3 (xi 1 , t )
=
z3 (xi , t ) +
h2
t
I
2h
Addition of Damping
Viscous damping can be added to
this set of equations by adding a term
proportional to the velocity of the deflection.
Returning to the first of the equations of
motion and adding the damping term yields,
(x, t ) 2 y ( x, t )
2 y ( x, t )
y ( x, t )
+ A
GA
+C
= q ( x, t )
2
x
t 2
t
x
z 2 ( xi , t )
2h
t
h2
A
A
y ( x0 , t ) = 0
y (x0 , t )
=0
x
y ( x0 , t ) y (x1 , t ) y ( x0 , t )
FDA
=0
=
h
x
y ( x1 , t ) = y ( x0 , t ) = 0
Similarly at the end where x = L, the
backwards difference approximation is used
because of the restriction of the rightward
terms
y (x L , t ) y ( xL1 , t ) y ( x, t )
=0
BDA
=
h
x
can be rewritten as
y ( x L 1 , t ) = y ( x, t ) = 0
The same procedure can be used for the
slope due to shear and curvature due to
shear to find
1 < c
rect (c ) =
0 > c
will ensure frequency content to be below
c. The inverse Fourier transform of the
rectangular function is the sinc function,
sinc(t ) =
sin (ct )
c t
(a)
(b)
(c)
Displacement
(d)
(e)
(f)
(g)
beam length
frequency increases. The peaks in the autopower spectra indicate possible resonant
modes of vibration. The relative phase
information found in the FRF is required to
confirm that these peaks are indeed
resonances.
(a)
(a)
(b)
Figure 4: The deflection at (a) x = 0.75m,
and (b) x = 0.5m from time 0 to 100 ms.
Auto-Power Spectra and Frequency
Response Functions
To further validate both the impulse
and responses, the frequency domain is
also considered. The auto-power spectra
and FRFs were calculated. Figure 5(a)
shows the calculated auto-power spectra of
the excitation impulse. The plot shows the
result of the combined sinc function and
Gaussian distribution used to specify the
input to the beam.
As the frequency
increases the energy rolls off dramatically
with the -3dB point being lower than 500 Hz.
The auto-power spectra of the
response at x = 0.5 and x = 0.75 are shown
in figure 5(b). Much like the auto-power
spectrum of the impulse, the auto-power
spectra of the responses rolls off as the
(b)
Figure 5: The auto-power spectrum of (a)
the excitation impulse and (b) the beam
deflection at x = 0.5 and x = 0.75.
The FRFs plotted in figure 6 show
the telltale phase shifting associated with
resonant frequencies and confirms that the
model does indeed demonstrate the
resonant behavior that was sought. Also of
note are the differences between the FRF at
x = 0.5 and x = 0.75. Because the node of
the second mode exists at the midpoint of
the beam, the second peak is suppressed in
the FRF of the response at x = 0.5.
(a)
(b)
Figure 7: The smoothed pseudo WignerVille distribution of the deflection response
at (a) x = 0.75 and (b) x = 0.5
The smoothed pseudo Wiger-Ville
distribution of the response at x = 0.5 is
shown in figure 7(b). The observations
made of the response at x = 0.75 are again
made with the exception of the second
mode. As noted in the FRF and auto-power
spectrum of the response at x = 0.5, the
second mode is not evident at this location
because it lies on the node of the mode.
Conclusions
In this paper a model for the
response of a finite Timoshenko beam
supported
by
fixed-fixed
boundary
conditions and subject to an impulsive
excitation has been presented. The model
has been verified to behave like vibrating
beams are known to behave through time,
frequency, and time-frequency domain
observations. This model will be scaled to