Professional Documents
Culture Documents
A modiJed independent modal space control (MIMSC) method is developed for designing active vibration control systems for large
flexible structures. The method accounts for the interaction between the controlled and residual modes. It also incorporates optimal
placement procedures for selecting the optimal locations of the actuators in the structure in order to minimize the structural vibrations
as well as the actuation eflort.
The MIMSC method relies on an important feature which is based on 'time sharing' of a small number of actuators, in the modal space,
to control eflectively a large number of modes.
Numerical examples are presented to illustrate the application of the method to generic flexible systems. The obtained results suggest
the potential of the devised method in designing eficient active control systems for large flexible structures.
6, & = displacement and acceleration of the nodal much higher frequency bands than the controlled
points of the structure modes. If these two conditions are not satisified, then
F = vector of the external and control forces acting there will be considerable interaction between the con-
on the structure trolled and residual modes.
The MIMSC method considers such interaction by
Equation (1) is put in the modal space by using the calculating the optimal modal control forces fc using
following weighted modal transformation : the IMSC closed-form solution of the Riccati equation
s=4u (2) such that the control forcefi of the ith mode, as given
by reference (7), is
where
U = modal coordinates of the system
4 = system's modal shape matrix
With such transformation, equation (1) reduces to the where
following uncoupled form: R = factor that weighs the importance of mini-
U+nu=f (3) mizing the vibration with respect to the
control forces
where mi = resonant frequency at the ith normal mode
1 = eigenvalues diagonal matrix of the system ui ,u, = modal displacement and velocity respectively
f = modal force matrix given by gl, g2 = modal position and velocity feedback gains
f = given by reference (7) as
(4)
or gl= -u,R+ J{(O,R)~+U?R) (10)
+
g2 = ,/(2Roi[ - mi R + &(mi R)2 wz R ) ] + O! R )
(1 1)
Accordingly, the displacement ui and velocity u, of the
ith mode can be fed back and used along with equa-
tions (9), (10) and (11) to determine the modal control
forcef;,.
Once these forces are calculated, equation (7) is
solved to give the physically applied control forces F ,
as follows :
Fc = B,-,'fc
Then equation (8) is used to calculate the modal
where forcesf, that would excite the residual modes which are
f,, = modalforces on the controlled and residual generated by the spillover from the controlled modes.
modes respectively These forces, f R , are not equal to zero as originally
F,, = corresponding physical forces acting on the assumed in the IMSC method.
system Equations (3) can then be integrated with respect to
&ij) is the modal shape at mode i and location ij the time to determine the modal displacements ui and
The above equation can be rewritten as velocities ir, which, in turn, can be used to compute the
modal forcesf and so on.
From the modal displacements and velocities, the
physical state 6 of the flexible system can be determined
from equation (2). A relationship can therefore be estab-
If only C modes are controlled with equal number of lished between the physical state 6 of the system and the
control forces F,, then FR= 0 and equation (6) reduces physical control forces F, applied to it.
to
2.3 Optimum placement of actuators
fc = BCCFC
It is very important to point out that the magnitude of
and the modal forcesf, depends primarily on the magnitude
fR = BRC FC
of wi's of the controlled modes as well as on the modal
state variables u and u. On the other hand, the magni-
tude of the actual physical control forces F , depends
2.2 Effect of control spillover mainly, for a given controlled mode, on the application
In the IMSC method, it is assumed that the control point of these forces as defined by the matrix B;:.
forces F, will not excite the residual higher-order Therefore, minimizing fc does not necessarily mean
modes. Accordingly, it was assumed that there is no that F, will be minimum in spite of the fact that it is
control spillover from the controlled modes into the represented as a linear combination of fc . This is simply
uncontrolled modes. Mathematically, this means that because the coefficients of the linear combination, which
fR = 0. This of course can only be true if the number of are elements of the B&! matrix, depend on the place-
controlled modes is very large compared to the number ment strategy of the control forces F , . It would still be
of residual modes or when the residual modes are at possible to find an optimally placed set of physical
Part C : Journal of Mechanical Engineering Science @ IMechE 1989
INPUT
Geometry. elastic and inertial
displacements and control forces would assume a propertiesof structure
minimum value.
This optimum placement of the physical control
forces is an important feature of the MIMSC method
and will be demonstrated to be an essential part of the
active control system design.
It should also be stressed that if all the modelled
modes are controlled then the conditions for minimizing
fc will make Fc minimum as well. In very large struc-
tures this would be unlikely to happen as the number of
controlled modes is much smaller than the number of
modelled modes. Therefore, it is essential to augment
the IMSC method with an optimal placement algorithm
to guarantee efficient control system design.
The optimum placement of the actuators is imple-
mented through the use of the uni-variate search
method (13) which varies the location of one actuator at
a time in order to
minimize
s (ti2 + RFZ) dt
-0.60
-I.ooI
k
0.00
I 1
5.71
I I
11.43
I 1
17.14
I ,
22.86
I I
28.57
I I I
34.2940.00
1 0.00 5.71 11.43 17.14 22.86 28.57 34.29 40.00
Time
__ Time
2.40
-2.401
-4.00
0.00 5.71 11.43 17.14 22.86 28.57 34.29 40.00
I -2.40
-4.001
i?0.00
I I
5.71
I ,
11.43
I 1
17.14
I 1
22.86
r ,
28.57
I , .
34.29 40.00
J
Time Time
__
(b) S (b) S
Fig. 4 Time history of amplitude of vibration and control Fig. 5 Time history of amplitude of vibration and control
forces for the spring-mass system using the MIMSC forces for the spring-mass system using the MIMSC
method with sequential time sharing of two actuators method with modal energy time sharing of two actu-
(R = 1) ators (R = 1)
operate unless the vibrations of the system are com- of time, then the resulting time history of the three
pletely damped out. Accordingly, in the considered masses and the associated control forces are as shown
example, the two actuators are powered by signals to in Fig. 5a and b respectively. Figure 5a demonstrates
eliminate all the three modes and not only the first two the effectiveness of the time-sharing scheme based on
modes as in the IMSC. This is achieved by time sharing the modal energy ranking as compared to the sequential
the two actuators, among the three modes, either scheme or the IMSC method.
sequentially or based on the maximum modal energy A better quantitative comparison between the two
ranking. schemes can be established based on the displacement,
Figure 4a and b shows the time history of the vibra- control force and control energy indices U , , U , and UE
tion amplitudes and the associated control forces of the which are given by
two actuators respectively when the MIMSC utilizes a t* N
sequential time-sharing strategy. Figure 4a indicates Ud= C@At (14)
t=O i=l
that sharing the small number of actuators among a N
9 5 F: At
t*
larger number of modes has been effective in damping U, =
out the amplitudes of vibration of all the modes. Such a f=O i = l
process is done by making these actuators work as long t* N
as there is vibration to be damped out, as can be seen U, = 1 1d i F i At
r=O i=l
(16)
from Fig. 4b. This is unlike the same two actuators
which have been only partially utilized by the IMSC where
method as illustrated in Fig. 3b. N = number of d.0.f. of system
When the time-sharing strategy is based on dedicat- At = integration time increment
ing the two actuators to control the two modes that t* = maximum time limit of integration
have the highest modal energy (u: h:), at any instant + Table 2 summarizes the results of such a comparison.
@ IMechE 1989 Proc Instn Mech Engrs Vol 203
Table 2 Effect of the strategy of time sharing two actuators associated control forces is shown in Figs 6a and 6b
on displacement, control force and control energy respectively. The figures emphasize the same trends
indices for the spring-mass system with R = 1 observed for R = 1, however, and more importantly
Displacement Control forces Control energy show that the MIMSC method is still very effective in
Strategy index index index quickly damping out the vibrations, even with smaller
Sequential 2.44 32.36 7.14 control forces. For example, when R = 100 the control
Modal energy 2.15 13.99 3.04 forces required with sequential and modal energy time
sharing assume maximum values of 0.460 and 0.524
3.1.2 Control by two actuators with weighting factor respectively. These magnitudes are at least six times
R = 100 lower than those computed for R = 1, as can be seen in
The effect of increasing the weighting factor R to 100 on Figs 4b and 5b.
the time history of the vibration amplitudes and the Table 3 lists the effect of increasing R to 100 on the
IMSC
IMSC
I .OO
I I (Mass number) ' 0 I
6 . o
0.361
-0.36
1
i I
-0.601 l . I I I I I I . I
0.00 11.43 22.86 34.29 45.71 57.14 68.57 80.00 0.00 11.43 22.86 34.29 45.71 57.14 68.57 80.00
Time Time
__
~
S S
MIMSC sequential
I .OO
1 (Mass number)
o,36-
I I (Actuator number)
2 (Actuator number)
0.36
aJ
L
P
p 0.12
c
6
-0.12
-0.36
i2 -0.60
0.00 11.43 22.86 34.29 45.71 57.14 68.57 80.00 0.00 11.43 22.86 34.29 45.71 57.14 68.57 80.00
Time Time
S S
Fig. 6a Time history of the amplitudes of vibration of the Fig. 6b Time history of the control forces on the
spring-mass system with two actuators and R = 100 spring-mass system with two actuators and R = 100
Part C : Journal of Mechanical Engineering Science @ IMechE 1989
Table 3 Effect of the strategy of time sharing two actuators reduction in the control force and energy indices that
on displacement, control force and control energy are more dominant than the increase in the
indices for the spring-mass system with R = 100 displacement index.
Displacement Control forces Control energy In the case of the IMSC method, changing R did not
Strategy index index index influence at all the amplitudes of vibration at steady
Sequential 8.40 1.960 5.05 state but it did, however, prolong the duration of the
Modal energy 5.03 1.462 4.03 transition time needed to eliminated the first two
modes.
Accordingly, with optimally selected weighting factor
displacement, control force and control energy indices R the MIMSC method can effectively suppress the
when the time sharing is based on sequential and modal vibration without the need for excessively large control
energy strategies. The table indicates a considerable forces.
-1.00( I
0.00 11.43
1 , ,
22.86
' I' "
34.29 45.71
I ' * 57.14 ' 68.57 80.001
. I
I
!
S S
MIMSC sequential
I .OO
MIMSC sequential
1.00 , I
1 (Mass number)
,
S
S
-1.00 I
0.00
I
1 I
11.43
. ,
22.86
[ I
34.29 45.71
I .
57.14
I . , ,
68.57 80.00 0.00 11.43 22.86 34.29 45.71 57.14 68.57 80.00
Time Time
__
S
S
Fig. 7a Time history of the amplitudes of vibration of the Fig. 7b Time history of the control forces on the
spring-mass system with one actuator and R = 100 spring-mass system with one actuator and R = 100
0 IMechE 1989 Proc Instn Mech Engrs Vol 203
Table 4 Effect of the strategy of time sharing one actuator maximum amplitude of oscillation of the beam is
on displacement, control force and control energy observed to be, in this case, about 10.7 per cent lower
indices for the spring-mass system with R = 100 than that obtained with the IMSC method.
Displacement Control forces Control energy
Strategy index index index IMSC
1.753
30.00
Sequential 14.129 2,752
Modal energy 11.833 2.960 1.289 A 4 (Node number)
3.1.3 Control by one actuator with weighting factor
R = 100
To demonstrate more dramatically the effectiveness of
the MIMSC in controlling the vibration of a large
number of modes with a small number of actuators a
single actuator, placed at mass 1, is used to actively
control the three spring-mass system.
Figures 7a and 7b show the time history of the
amplitudes of vibrations and the associated control -30.001 . , . I . , . I , . , .
forces respectively as obtained by the IMSC and 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14
Time
MIMSC methods. Again the figures emphasize the
s
potential of the MIMSC, particularly with its modal MIMSC seauential
energy time-sharing strategy, as a viable active control.
Table 4 summarizes the results obtained from the 30'00 I 4 (Node number)
analysis of these figures.
-30.00!
1
0.00
1
0.02
. 1
0.04
.
0.06
1
0.08
. I - I
0.10
. I .
0.12 0.14
I
higher modes remain totally undamped. On the Time
contrary, the MIMSC with the modal energy time S
sharing exhibits complete control over all the modes Fig. 9 Time history of the amplitudes of transverse vibration
and effective damping is demonstrated. Also the of the cantilever beam with one actuator and R = 100
Part C : Journal of Mechanical Engineering Science @ IMechE 1989
Table 5 Effect of the strategy of time sharing one actuator 3.2.4 Beam subjected to sinusoidal excitation
on displacement, control force and control energy
indices for the cantilever beam with R = 100 Unlike the previous examples where the flexible systems
were subjected to an initial disturbance, the MIMSC
Displacement Control forces Control energy
Strategy index ( x 10’) index ( x 10’) index ( x lo6) method is used to control the cantilever beam under
sinusoidal excitations. The excitation force is 1.0 N
Sequential 5.919 0.9081 3.205 applied at node 4 at a frequency of 10 Hz.
Modal energy 1.243 2.9000 1.596
Figure 10a shows the time response of the
uncontrolled beam. Figures 10b and c displays the
Considering, however, the MIMSC method with
sequential time sharing it can be seen that this strategy Uncontrolled
- 10.00
is not as effective as the modal energy strategy or the
IMSC method. The reasons are obvious and it is 4 (Node number)
important to cite them. Firstly, as the beam is a
six-mode system controlled by one actuator, it is
essential to utilize this actuator in the best possible way.
Dedicating it to the first mode, as in the IMSC, is found
adequate to damp out large amplitude oscillations but
inadequate to take care of the high-frequency jitters.
Using this actuator to control any mode in any
sequence without due consideration to its contribution
to the system’s vibration or energy is definitely a reason
behind the ineffectiveness of the sequential time-sharing
strategy. However, once the actuator is set to control
the highest modal energy mode then the time sharing is 0.00 0.04 0.08 0.12 0.16 0.20 0.24 0.28
efficient and fast. The second reason for the inadequacy Time
__
(a) S
of the sequential time sharing is due to the fact that as
the actuator is sequenced to control the high-frequency MIMSC modal energy (R = 100)
modes, which may have a low energy level, it leaves the
high energy modes uncontrolled. Therefore, the more
the number of modes between which an actuator is
shared, the longer it will take to go through them all
and come back to control the low-frequency modes that
may still have the high energy. The delay period in
controlling the high-energy modes increases as the
number of modes is increased and accordingly this will
prolong the time needed to actively bring the structure
under control.
Quantitatively, the comparison between the two
time-sharing strategies is given in Table 5.
corresponding response when the beam is controlled by digitally the MIMSC to control the vibration of a
the MIMSC method with R = 100 and R = 1 cantilevered beam using a single piezoelectric actuator
respectively. It can be seen that the MIMSC damped and a single non-contacting sensor. The effect of
out all the high-frequency oscillations and attenuated sampling time, state estimation, modelling inaccuracy
the magnitude by about half (when R = 1) that of the and control strategy will be addressed.
uncontrolled case.
ACKNOWLEDGEMENTS
4 CONCLUSIONS
Special thanks are due to the Space Science and
This paper has presented a modified algorithm of the Technology Division at NASA Goddard Space Flight
independent modal space control method where control Center for providing the funds necessary to conduct this
spillover as well as optimum placement and time study under Grants NAG 5-520 and NAG 5-749.
sharing of actuators have been considered.
The considered numerical examples indicate the
importance of including such modifications to the REFERENCES
IMSC. 1 Simon, J. D. and Mitter, K. A theory of modal control. In$ and
It is shown that time-sharing a small number of Control, 1968, 13, 316353.
actuators between a large number of modes can be 2 Porter, B. and Crossley, R. Modal control, theory and applications,
effective in suppressing the vibration if the actuators are 1972 (Taylor and Francis, London).
dedicated to control the modes that have the highest 3 Balas, M. J. Feedback control of flexible systems. IEEE Trans. on
Automatic Control, August 1978, AC-23(4), 673-679.
modal energy at any particular instant. 4 Sesak, J. R., Likins, P. W. and Coradetti, T. Flexible spacecraft
The presented algorithm has the potential of being a control by model error sensitivity suppression. Proceedings of the
viable means for controlling large flexible structures. Second VPI and SU/AIAA Symposium on Dynamics and control
Furthermore, it is important to note that although the of largeflexible spacecraft, Blacksburg, Va., June 1979.
5 Coradetti, T. Orthogonal subspace reduction of optimal regulator
emphasis in this paper has been placed on undamped order. AIAA paper 79-1742, Presented at the AIAA Conference on
systems, the developed procedures can be equally Guidance and control, Boulder, Colo., 6 8 August 1979.
applied to damped and gyroscopic systems. The system 6 Sesak, J. T. Suppressed mode damping for model error sensitivity
reduction approach described by Meirovitch and Baruh suppression of flexible spacecraft controllers. AIAA paper 80-1710,
(11) can be used to compute numerically the optimal Presented at the AIAA Conference on Guidance and control,
Danvers, Ma., 11-13 August 1980.
gains of the mode to be controlled and the time sharing 7 Meirovitch, L., Van Landingham, H. F. and Oz, H. Control of
can be utilized accordingly to control these damped spinning flexible spacecraft by modal synthesis. Acta. Astronaut.,
modes one at a time. 1977,4(9-lo), 985-1010.
Also, it should be mentioned that the time-sharing 8 Meirovitch, L. and 05 H Computational aspects of the control of
large space structures. Proceedings of the eighteenth IEEE
concept should be considered carefully in conjunction Conference on Decision and Control, Fort Lauderdale, Flo.,
with the dynamic characteristics of the actuators. If one December 1979, pp. 220-229.
actuator is used to control several modes of vibrations 9 Meirovitch, L. and 0% H. Modal space control of distributed
then its frequency band should be wide enough to cover gyroscopic systems. J. Guidance and Control, 1980,3(3), 14G150.
the desired controlled modes. Practically speaking, it 10 0 2 , H. and Meirovitch, L. Optimal modal space control of flexible
gyroscopic systems. J. Guidance and Control, 1980,3(3), 218-226.
would be difficult to find a single actuator that covers 11 Meirovitch, L. and Bsruh, H. Optimal control of damped flexible
the entire frequency spectrum of a structure. Therefore, gyroscopic systems. J . Guidance and Control, 1981,4(2), 157-163.
one actuator can be dedicated to low-frequency modes 12 Meirovitch, L. and Baruh, H. Control of self-adjoint distributed-
and another to medium-frequency modes. The high- parameter systems. Presented as AIAA paper 8Gl707-CP at the
AIAA Conference on Guidance and Control, Danvers, Ma.,
frequency modes could presumably be damped quite August 1980.
easily and more effectively with passive dampers. 13 Fox, R. L. Optimization methods for engineering design, 1971, pp.
Future work will be carried out to implement 4 6 5 1 (Addison-Wesley Publishing Company, Reading, Mass.).