Professional Documents
Culture Documents
Vibration Absorbers As Controllers
Vibration Absorbers As Controllers
#
(6)
where W(j) is a weighting function describing the frequencies around which F is concentrated, and is scaled so that |F(t)|
#. Equation (6) leads to a difficult optimization problem, especially for multi-degree of freedom systems. The approach
taken in this paper starts from defining a function (s) approximating S(s) and then obtains the absorber parameters by
solving a simpler optimization problem. In addition to its simplicity, this approach gives valuable insights as to how to
select the absorber locations for multi-DOF systems, as will be explained later.
Defining the desired sensitivity function as (s), which behaves qualitatively as #/W(s), we can write:
(s) =
+
S s
G s Z s
( )
( ) ( )
#
#
(7)
form which, Z(s) can be solved as:
Z s
G s
( )
( )
=
# # (s)
(s)
(8)
A more physically motivated definition for (s) might be to define it as a ratio of the vibration amplitude of the main mass
with the absorber X
WA
(s) to the one without the absorber X
NA
(s). That is:
(s) =
+
=
#
# G s Z s
X s
X s
WA
NA
( ) ( )
( )
( )
(9)
Typically (s) will have the shape of a notch filter. A possible choice for (s) for single sinusoidal forcing with frequency
0
is:
( )
( ) ( )
( ( ) )
s
s s
s
=
+ +
+
0
2
0
0
2
2 #
#
(#0)
A passive absorber such as the one in equation (2) cannot satisfy (8) for all frequencies, so it is reasonable to define a
criterion J to be minimized in order to obtain absorber parameters:
J Z j Z j
k k
k
n
=
=
( )
( )
#
2
(##)
where
k
are a discrete set of frequencies around which a good fit is required and
0
(#2)
where f is the internal force transferred to the system by the impedance of the absorber:
f Z X
A
= (#3)
If a model system is defined by the following relationship:
X s s G s F s
F I I
( ) ( ) ( ) ( ) = (#4)
using equations (#2) and (#3), we can solve for Z(s) as:
Z
G
G G G G
F
A AA F
=
+
( )
( )
#
#
I
FA I I
(#5)
For the special case of X
I
= X
A
= X
F
= X, the above equation (#5) reduces to equation (8), as in the case of single DOF
systems. Given the structure of the absorber, parameters of the absorber
>
( ) Z s can be obtained by minimizing (##).
3. Selecting the Absorber Locations
In a multi degree of freedom system, there is more than one choice for the location of the absorber (or the location of the
collocated sensor and actuator for the active control case). At first sight, for a multi DOF system, location of the 'best' place
for the absorber is not obvious. The best absorber should minimize the vibration amplitude at the selected degree of
freedom with the least control effort possible. If a qualitative measure
i
for an absorber located at i'th degree of freedom
is defined, it might be possible to evaluate the effectiveness of the alternative locations.
i
can be defined as:
i
I
i
I i
X
X
= =
f F
f
F
#
(#6)
Evaluating the above expressions (#2) to (#5) we can obtain the simple expression for
i
:
( )
i
i
=
#
G
AI
(#7)
Note that
i
is independent of the controller used and it can be used as a screening tool amongst the possible absorber
locations. The absorber location i with the highest |
i
| at the frequency of interest would be the best location for placing
the absorber, in the sense that it would decrease X
I
to the desired value with the least control effort. Since the factor /( -
#) is common for all absorber locations,
i
is proportional to (G
AI
)
i
at the frequency of interest. This result is reasonable
because the absorber location with the highest G
AI
would have the most effect on X
I
. The interesting point is that the |
i
|,
being a function of frequency, can differ at different frequencies of interest, therefore, the 'best' absorber location for a fixed
X
I
can change with changing forcing frequency
0
. This will be illustrated by an example later .
4. Design for Multiple Forcing
In the case of multiple sinusoidal forces with different frequencies acting on a structure, a single absorber will not be
sufficient. Instead, multiple absorbers located at different locations and 'tuned' to the frequency of excitation might be used,
as in [Igusa, Xu, #994]. The approach developed in the sections above might also be used to design multiple absorbers for
multiple forces, using the matrix equivalents of the above equations. When X
F
and X
A
are vectors of equal length n, the
absorber impedance Z becomes a full nn matrix. In addition to the problems to be encountered in parameter estimation, an
impedance Z with cross-coupling terms might be hard to implement on a structure. Therefore, a solution with diagonal Z
has to be sought. An approach similar to multivariable sequential loop closing [Maciejowski, #989], used in decentralized
control, can be taken to design multiple absorbers. First, an absorber is designed for one forcing, ignoring all others. Other
absorbers are designed sequentially adding the already designed absorbers to the structure. It is necessary to perform a few
iterations until the absorber parameters converge.
5. An Example
Consider the 3 degree of freedom system shown in Figure 2, with M= #, K=4, C=0.#. A harmonic force of F(t)= F
0
Sin
0
t
is applied on the first mass. The aim is to minimize the effect of this force on the third mass, using a vibration absorber.
Figure 2. Three degree of freedom system
Both active and passive absorbers will be considered. For passive absorbers, the absorber impedance is :
( )
( )
Z s
M s C s K
M s C s K
a a a
a a a
=
+
+ +
2
2
where M
a
is specified as # and K
a
and C
a
is to be determined. For active absorbers, the controller transfer function is
defined as:
<
( ) Z s
a s a s a
s a s a
=
+ +
+ +
#
2
2 3
2
4 5
where a
#
a
5
is to be determined, with the constraint a
4
>0 and a
5
> 0, so that the controller is stable.
The original system has natural frequencies around 0.9 and 2.5 if its damping is neglected. The frequency of external
forcing
0
will be assumed to take these values. The desired sensitivity function (s) is selected as in equation (#0) with =
0.0#.
The frequency response of the transfer function G
AI
which determines the absorber effectiveness, is shown in Figure 3 for
three different absorber locations. It can be seen that as |G
AI
|
i
varies with frequency, the 'best' absorber location changes.
Table # gives the absorber effectiveness |
i
(j
0
)| for the forcing frequencies
0
= 0.9 and 2.5, for different locations of the
absorber. It can be seen that for
0
= 0.9, placing the absorber at the third mass is the best choice, whereas for
0
= 2.5, it is
best to place the absorber at the first mass. This result is not surprising if when one considers the mode shapes of the
original system. From the mode shapes of the undamped system given in Table 2, it can be seen that at the first natural
frequency
0
= 0.9, the third mass attains the maximum amplitude and at the second natural frequency
0
=2.5, amplitude of
the first mass is the greatest. In this example, the absorber location could have been selected observing the mode shapes,
however for general non-proportionally damped structures, mode shapes do not have a physical meaning, but the approach
proposed here still remains valid.
Figure 3. |G
AI
(j)| for different absorber locations.
Location of
absorber
|
i
(0.9j)| |
i
(2.5j)|
# 0.#39 0.0##
2 0.249 0.005
3 0.3## 0.009
Table #. Absorber effectiveness |
i
(j
0
)| for different locations.
Mass Mode # Mode 2 Mode 3
# 0.328 0.737 0.59#
2 0.59# 0.328 -0.737
3 0.737 -0.59# 0.328
Table 2. Normalized mode shapes of the undamped system.
The absorber impedances
( ) Z s for the active and passive case are obtained by solving the optimization problem defined by
equation (##). Optimization problem is solved by selecting #0 frequency points in the vicinity of
0
. Results are given for
the best absorber locations, dictated by highest |
i
(j
0
)|. For
0
= 0.9, the absorber is placed on the third mass and for
0
=
2.5, it is located on the first mass. The frequency response of the system with these absorbers, along with the frequency
response of the original system and the model closed loop system defined by G
FI
are given in figures 4 and 5, for the cases
of
0
=0.9 and
0
= 2.5.
For comparison, the obtained passive absorber parameters K
a
and C
a
for different absorber locations are given in Table 3.
It can be seen that the damping term C
a
increases as the absorber is placed in positions which have higher absorber
efficiency |
i
(j
0
)|. Also, K
a
approaches to undamped natural frequency squared of the original system as the absorber is
placed in less efficient locations. Therefore, the absorber creates an anti-resonance in less efficient locations, giving a
solution similar to the undamped absorber, whereas in more efficient locations it finds out a compromise between anti-
resonance and adding damping to the system. Obviously, the latter choice transmits less absorber force f to the system.
Figure 4. Frequency response of the original system, system with absorbers and the model system for
0
=0.9.
Figure 5. Frequency response of the original system, system with absorbers and the model system for
0
= 2.5.
Frequency
of forcing
0
=0.9
0
=2.5
Absorber
Location
K
a
C
a
K
a
C
a
# 0.830 0.080 6.#40 0.4#8
2 0.765 0.#80 6.263 0.060
3 0.723 0.245 6.247 0.259
Table 3. Passive absorber parameters for different locations.
If the two resonances at
0
= 0.9 and
0
= 2.5 are to be simultaneously supressed, then two absorbers are necessary. Using
the iterative procedure outlined in section 4, these absorbers can be designed. Forcing is assumed to be on the first mass and
the vibrations of the third mass are to be supressed. As in the previous cases, absorbers with mass M
a
= #, are placed on the
third and first masses. The design procedure used in single absorber design is used iteratively. The frequency response of
the resulting design is given in Figure 6, and the resulting absorber parameters are K
a#
= 0.867, C
a#
= 0.079, K
a2
= 5.895,
C
a2
= 0.248.
Figure 6. Frequency response of the original system (dashed) and the modified system with two absorbers (solid), which are
tuned to
0
=0.9 and
0
= 2.5.
In the above example, absorbers act relatively independently, since they are tuned to the resonances of the original system.
When the frequencies to be suppressed do not coincide with the resonances, the iterative solution tends to create
antiresonant absorbers with little damping and with
( / )
/
K M
a a
# 2
close to the frequency of forcing. An example is given
with
0#
=# and
02
= #.5, for the same structure. The parameters of the designed absorbers are : M
a#
= M
a2
= #, K
a#
= #.043,
C
a#
= 0.02#, K
a2
= 2.3#3, C
a2
= 0.0#2.
Figure 7. Frequency response of the original system (dashed) and the modified system with two absorbers (solid), which are
tuned to
0
=# and
0
= #.5.
6. Conclusions
In this paper, vibration absorber design is considered from a control point of view. The problem can be considered as
matching a sensitivity function over a specified frequency range where the disturbances are effective, with a fixed order
active or passive controller. An approach is proposed where the desired sensitivity function is specified and then the
absorber parameters are obtained solving a nonlinear optimization problem. In contrast to many control problems, the
selection of the manipulated variables (the location of the absorbers) is not an obvious choice in absorber design for multi
DOF systems. It is shown that the 'best' absorber location is at the point where the transfer function between the absorber
force and the displacement of the mass whose motion is to be minimized, is at its maximum, at the frequency range of
interest. The approach is extended to the design of multiple absorbers, using an iterative technique that is similar to
sequential loop closing used in decentralized control.
Acknowledgement
This research has been supported by Bo"azii University Research Fund, Project No : 95A0640.
References
Den Hartog J.P. (#956) Mechanical Vibrations 4'th Edition, McGraw Hill.
Doyle J. C., Francis B.A., Tannenbaum A.R. (#99#) Feedback Control Theory, Macmillan.
Herzog R. (#994) Active versus Passive Vibration Absorbers, ASME J. Dynamic Systems Measurement and Control V.##6
pp. 367-37#.
Hovd M., Skogestad S. (#992) Simple Frequency Dependent Tools for Control System Analysis, Structure Selection and
Design , Automatica, 28, 989-996
Igusa T. Xu K. (#994) Vibration Control using Multiple Tuned Mass Dampers, J. of Sound and Vibration V.#75 pp. 49#-
503.
Jacquot R.G. (#978) Optimal Dynamic Vibration Absorbers for General Beam Systems. J. of Sound and Vibration V.60 pp.
535-542.
Juang J.N., Phan M. (#992) Robust Controller Designs for Second Order Dynamic System : A Virtual Passive Approach.
AIAA J. of Guidance Control Dynamics, V. #5 pp. ##92-##98.
Juang J.N. (#984) Optimal Design of a Passive Vibration Absorber for a Truss Beam AIAA J. of Guidance Control
Dynamics V. 7 pp. 733-738.
Kitis L., Wang B.P., Pilkey W.D. (#983) Vibration Reduction over a Frequency Range. J. of Sound and Vibration V.89 pp.
559-569.
Maciejowski J.M. (#989) Multivariable Feedback Design, Addison Wesley.
Skelton R.E., Hanks B.R., Smith M., (#992) Structure Redesign for Improved Dynamic Response. AIAA J. of Guidance
Control Dynamics V. #5 pp. #272-#278.
Stech D.J. (#994) H
2
Approach for Optimally Tuning Passive Vibration Absorbers to Flexible Structures. AIAA J. of
Guidance Control Dynamics V. #7 pp. 636-638.
Wang B.P., Kitis L., Pilkey D., Palazzolo A. , (#985) Synthesis of Dynamic Vibration Absorbers, ASME J. of Vibration,
Acoustics, Stress and Reliability in Design V.#07 pp. #6#-#66.