Professional Documents
Culture Documents
Negative Derivative Feedback For Vibration Control of Flexible Structures
Negative Derivative Feedback For Vibration Control of Flexible Structures
This content has been downloaded from IOPscience. Please scroll down to see the full text.
(http://iopscience.iop.org/0964-1726/21/7/075024)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 148.247.124.101
This content was downloaded on 08/08/2017 at 18:25
An H2 norm approach for the actuator and sensor placement in vibration control of a smart structure
P Ambrosio, F Resta and F Ripamonti
Reduction of spillover effects on independent modal space control through optimal placement of
sensors and actuators
S Cinquemani, D Ferrari and I Bayati
E-mail: gabriele.cazzulani@mail.polimi.it
1. Introduction IMSC is the so-called spillover effect [10], which can lead
the system to instability associated with significant modes
In recent decades, vibration control has assumed more and neglected in the control tuning.
more importance owing to the significant development of light To solve this problem several approaches have been
and highly flexible structures. The main problem for these proposed, such as those in [9, 11]. In this sense, one of the
systems is related to the amplification of vibrations near their best known control strategy families is the so-called ‘resonant
natural frequencies, mainly due to the low damping ratio control’. These control logics calculate the feedback action
associated with the first structure modes. A high vibration through a dynamic system, typically a first- or second-order
level can lead to a degradation of system performance in terms compensator. It returns a control force that, at the resonance
both of positioning error (consider, for example, a robotic frequencies under investigation, is opposite in phase to
manipulator) and system health and lifetime. the velocity and so produces a damping effect on the
In order to prevent this issue a possible approach is to corresponding vibration modes.
act directly on the input disturbance. In this sense many One of the first resonant control logics, developed
examples about the disturbance estimators/compensators [1, by Caughey and Fanson in the 1980’s [12, 13], is the
2] and the input shaping method [3, 4] are present in the positive position feedback (PPF) [14–17]. In this solution,
scientific literature. On the other side, vibration suppression the generalized displacement is fed back with a positive
can be applied, improving the system frequency response sign through a second-order dynamic system acting, in
function (FRF) through a suitable control action able to reduce other words, as a low-pass filter. As a consequence, the
the resonance peaks. PPF effectively reduces the spillover on the higher modes,
Since in many flexible structures only few vibration but it introduces a significant static error and in general
modes are significant, modal space control techniques are a worsening of system response at lower frequencies. A
very attractive for active vibration suppression. One of the dual solution is represented by the active modal tuned mass
most important control logics is the independent modal space damper (AMTMD) [18]. This control logic, developed in
control (IMSC) [5–7], developed in the 1980s, but still used modal approach as well, is based on an active realization
with success nowadays [8]. Under a number of assumptions, of the classic passive TMD theory and supplies a control
it allows one to modify the damping and stiffness of each force 90◦ shifted with respect to the displacement at the
controlled mode independently. The most important limit of system resonance frequencies. As shown below, it works as a
0964-1726/12/075024+10$33.00 1 c 2012 IOP Publishing Ltd Printed in the UK & the USA
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
2
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
3
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 4. Frequency response function between the AMTMD Figure 5. Frequency response function between the NDF control
control force and the modal velocity. force and the modal velocity.
and, in general, it performs less well for low-frequency modes auxiliary variable η
owing to its inertial nature.
Starting from the TMD formulation, it is possible to ui = −gi η̇
(12)
define a resonant controller, called active modal tuned mass η̈i + 2ξi ωi η̇i + ω2 ηi = ki (q̇i − η̇i ) ,
damper (AMTMD), which works as a PPF on the system
where ki and gi are two gains and ξi and ωi are the natural
modal coordinates and overcomes the limits of passive TMD.
frequency and damping ratio of the compensator, equal to
Considering again the single degree of freedom, the modal
those of the system. Equation (12) allows us to calculate the
force and the compensator can be now defined as
transfer function between the control force and the modal
ui = mci (ωc2 (q − η) + 2ξc ωc (q̇ − η̇)) velocity as
(11)
η̈i + 2ξc ωc η̇ + ωc2 η = 2ξc ωc q̇ + ωc2 q, Ui (s) ki gi s
=− 2 , (13)
Q̇i (s) s + (2ξi ωi ) s + ω2
where ηi is again the compensator variable, ξc and ωc
represent respectively the damping ratio and natural frequency where s is the Laplace variable. Equation (13) represents the
of TMD, while mc,i represents the inertial term of the mass transfer function of a band-pass filter. For this reason, the
damper, typically set between 5% and 10% of the modal control force rolls off both for frequencies higher and lower
mass. The values of these parameters can be tuned using than the controlled one (figure 5), reducing both the spillover
optimization rules typical of passive TMD [23]. Considering effects typical of IMSC and AMTMD and the low-frequency
the single degree of freedom equation this formulation, thanks amplification occurring with PPF.
to its passive nature, is stable for every parameter value. From the stability point of view, the NDF compensator
Owing to its formulation, the relationship between the control equation (12) must be evaluated together with the system
force and the modal velocity is represented by a high-pass modal equation (5)
filter (figure 4). Similarly also NPF, being a particular active ( ) " #( )
realization of the mechanical dynamic absorber in which in q̈i 2ξi ωi gi q̇i
+
the equivalent mechanical system the spring and damper are η̈i −ki 2ξi ωi + ki η̇i
in series instead of parallel, acts as a high-pass filter. " #( ) ( )
ωi2 0 qi fd,i
+ = . (14)
0 ωi 2
ηi 0
4. Negative derivative feedback
It can be seen that, while the inertia and stiffness matrices
As described in section 3.3, all the resonant controllers
are always symmetrical and defined positive, the damping
presented are designed to reduce the interaction between
matrix is not symmetrical. It means that the stability of the
control action and higher or lower uncontrolled modes.
controlled system depends on the values of the two gains gi
However, each control logic introduces a side effect that can
and ki and on the damping of the uncontrolled system.
reduce the stability margin on the uncontrolled modes.
To overcome these problems, a control logic able to 5. Numerical and experimental analysis
reduce the effects both for high and low frequencies is
proposed. This logic has been called negative derivative This section describes the numerical and experimental tests
feedback (NDF) because of the formulation of the control carried out to validate the proposed control logic and compare
force, including a negative feedback of the velocity of the it with the solutions already known. In order to consider
4
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 7. Position of the accelerometers with respect to the modal shape of mode 1 (a) and 2 (c) and position of the actuators with respect
to the modal curvature of mode 1 (b) and 2 (d).
5
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 8. Numerical/experimental comparison of first (a) and second (b) modal shapes.
6
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 10. Frequency response function between disturbance force and tip acceleration with IMSC control: numerical amplitude (a) and
phase (b) and experimental amplitude (c) and phase (d).
Figure 11. Frequency response function between disturbance force and tip acceleration with AMTMD control: numerical amplitude (a)
and phase (b) and experimental amplitude (c) and phase (d).
parameters considered for this application compared with the Table 4. Parameters of PPF for the first and second mode.
optimal ones of classical passive TMD.
Mode 1 Mode 2
PPF, compared with the control logics presented so far,
has low sensitivity to high-frequency modes owing to the gi (—) 0.1 0.1
low-pass behaviour of its compensator. As a consequence, ξc,i (%) 20 60
the increase of the damping imposed on the controlled modes
causes a negligible effect on the higher modes if compared, for modes (typical of IMSC and AMTMD) and low-frequency
example, with IMSC. For each controlled mode, setting the amplification (typical of PPF). Setting the cut-off frequency
cut-off frequency of the compensator equal to the considered and damping of the compensator equal to those associated
eigenfrequency, the behaviour of PPF control is governed by with the controlled mode, the only control parameter is the
two parameters: the damping of the compensator and the gain gain. For this application, the optimal control gains are 0.045
(see section 3.2). These two values were optimized in order to and 0.002 for the first and second modes respectively. The
obtain the maximum damping effect on the controlled modes, optimal tuning of NDF allows us to obtain very good results
but limiting the ‘double peak’ effect on the frequency response in terms of peak reduction of frequency response function
function amplitude. Table 4 shows the optimized parameter amplitude, avoiding low-frequency worsening and spillover.
values, while figure 12 shows PPF performance with respect Figure 13 show the numerical and experimental comparison
to the uncontrolled system. Thanks to the optimization of the
between the controlled (with NDF) and uncontrolled system.
control parameters, only a small ‘double peak’ effect is visible
on the first mode. On the other hand, an amplification of the 5.3. Comparison between the control logics
system response is introduced, as expected, below the first
natural frequency. The experimental analysis performed with the different
Finally, the performance of the proposed negative control logics and shown in section 5.2 provided results
derivative feedback is analysed and compared with the state- consistent with the numerical analysis and with the results
of-the-art solutions. As shown in section 4, the compensator expected from the theory. IMSC allows us to directly set the
associated with this control logic works as a band-pass filter, damping associated with the controlled modes, but since these
limiting both the incidence of spillover instability on higher values are imposed starting from the reduced modal space,
7
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 12. Frequency response function between disturbance force and tip acceleration with PPF control: numerical amplitude (a) and
phase (b) and experimental amplitude (c) and phase (d).
Figure 13. Frequency response function between disturbance force and tip acceleration with NDF control: numerical amplitude (a) and
phase (b) and experimental amplitude (c) and phase (d).
unwanted spillover effects can arise, compromising control Table 5. Damping ratios (%) of the controlled modes with different
performance and leading the system to instability. AMTMD resonant control techniques compared with those of the
uncontrolled system.
calculates the control force starting from the formulation of
classical passive TMD. Thanks to the correspondence with Mode 1 Mode 2
a passive system, each compensator parameter has a precise No control 0.42 0.16
physical meaning and it can be tuned starting from the IMSC 6.50 0.51
knowledge of passive TMD optimized solutions. Since the AMTMD 1.90 0.90
PPF 4.20 1.25
tuning is done on the reduced model, spillover effects can
arise owing to the high-frequency sensitivity of this control NDF 14.12 1.96
algorithm. In this case, the reduction of the virtual mass of
the compensator allows us to reduce control forces and, as the other resonant controllers, achieving better performance in
a consequence, high-frequency problems, but at the same terms of vibration reduction and damping increase.
time it reduces also control performance on the considered A synthetic comparison between the resonant control
modes if compared with IMSC. PPF allows us to overcome logics is shown in table 5, where the damping ratios
the problems shown by IMSC and AMTMD, through the achieved on the controlled modes with the different solutions
introduction of a low-pass compensator tuned to the natural compared with the uncontrolled system damping are reported.
frequency considered. Unfortunately, as a side effect, PPF is Figures 14 and 15 show the decay tests respectively on the first
more sensitive to low-frequency disturbances. and second mode with and without control. In both cases the
Finally, the proposed negative derivative feedback allows damping increase due to NDF is higher than that due to the
us to obtain better performance, since it limits both the other controls.
problems of IMSC, AMTMD and PPF. The band-pass 6. Conclusions
formulation of its compensator allows NDF to be effective
in filtering out higher and lower frequency disturbances. For In the present work a resonant control technique for
this reason, the control gains can be increased with respect to damping increase and vibration suppression, called negative
8
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
Figure 14. Controlled and uncontrolled decay on the first mode using IMSC (a), AMTMD (b), PPF (c) and NDF (d).
Figure 15. Controlled and uncontrolled decay on the second mode using IMSC (a), AMTMD (b), PPF (c) and NDF (d).
derivative feedback (NDF), was proposed. This control logic, motion is numerically integrated on the control board. The
formulated considering a modal space approach, is based control action is provided through a negative feedback of the
on a second-order dynamic compensator, whose equation of first derivative of the compensator variable. NDF has been
9
Smart Mater. Struct. 21 (2012) 075024 G Cazzulani et al
designed to behave as a band-pass filter on the control force, [11] Fang J Q, Li Q S and Jeary A P 2003 Modified independent
in order to overcome the limitations typical of state-of-the-art modal space control of m.d.o.f. systems J. Sound Vib.
resonant controllers, such as high-frequency spillover and 261 421–41
[12] Goh C J and Caughey T K 1985 On the stability problem
low-frequency amplification of system response. caused by finite actuator dynamics in the collocated control
The proposed NDF was tested on an experimental test of large space structures Int. J. Control 41 787–802
rig and the corresponding numerical model. It shows better [13] Fanson J L and Caughey T K 1990 Positive position feedback
performance with respect to all the other techniques analysed, control for large space structures AIAA J. 28 717–24
both in terms of achieved damping and robustness to low- and [14] Baz A, Poh S and Fedor J 1992 Independent modal space
control with positive position feedback J. Dyn. Syst. -
high-frequency problems.
Trans. ASME 114 96–112
[15] Friswell M I and Inman D J 1999 Relationship between
References positive position feedback and output feedback controllers
Smart Mater. Struct. 8 285–91
[1] Wang D A and Huang Y M 2002 Modal space vibration [16] Song G, Schmidt S P and Agrawal B N 2002 Experimental
control of a beam by using the feedforward and feedback robustness study of positive position feedback control for
control loops Int. J. Mech. Sci. 44 1–19 active vibration suppression J. Guid. Control Dyn.
[2] Bagordo G, Cazzulani G, Resta F and Ripamonti F 2011 A 25 179–82
modal disturbance estimator for vibration suppression in [17] Inman D J 2001 Active modal control for smart structures
nonlinear flexible structures J. Sound Vib. 330 6061–9 Phil. Trans. R. Soc. A 359 205–19
[3] Singhose W E and Singer N C 1996 Effects of input shaping [18] Cazzulani G, Resta F and Ripamonti F 2011 Active modal
on two-dimensional trajectory following IEEE Trans. tuned mass damper for smart structures Eng. Lett.
Robot. Autom. 12 881–7 19 297–303
[4] Shan J, Sun D and Liu D 2004 Design for robust component [19] Kim S M, Wang S and Brennan M J 2011 Comparison of
synthesis vibration suppression of flexible structures with negative and positive position feedback control of a flexible
on-off actuators IEEE Trans. Robot. Autom. 20 512–25 structure Smart Mater. Struct. 20 015011
[5] Meirovitch L and Baruh H 1983 Robustness of the [20] Bhikkaji B and Moheimani S O R 2008 Integral resonant
independent modal-space control method J. Guid. Control control of a piezoelectric tube actuator for fast nanoscale
Dyn. 6 20–5 positioning IEEE/ASME Trans. Mechatronics. 13 530–7
[6] Meirovitch L and Baruh H 1985 Implementation of modal [21] Bhikkaji B, Moheimani S O R and Petersen I R 2008
filters for control of structures J Guid. Control Dyn. Multivariable integral control of resonant structures Proc.
8 707–16 IEEE Conf. Decis. Control pp 3743–8
[7] Khulief Y A 2001 Vibration suppression in rotating beams [22] Pereira E, Aphale S S, Feliu V and Moheimani S O R 2011
using active modal control J. Sound Vib. 242 681–99 Integral resonant control for vibration damping and precise
[8] Resta F, Ripamonti F, Cazzulani G and Ferrari M 2010 tip-positioning of a single-link flexible manipulator
Independent modal control for nonlinear flexible structures: IEEE/ASME Trans. Mechatronics 16 232–40
An experimental test rig J. Sound Vib. 329 961–72 [23] Warburton G B and Ayorinde E O 1980 Optimum absorber
[9] Kim M H and Inman D J 2001 Reduction of observation parameters for simple systems Earthq. Eng. Struct. Dyn.
spillover in vibration suppression using a sliding mode 8 197–217
observer J. Vib. Control 7 1087–105 [24] Sun D, Shan J, Su Y X, Liu H and Lam C 2005 Hybrid control
[10] Cazzulani G, Resta F and Ripamonti F 2011 A feedback and of a rotational flexible beam using enhanced PD feedback
feedforward vibration control for a concrete placing boom with nonlinear differentiator and PZT actuators Smart
J. Vib. Acoust. 133 051002 Mater. Struct. 14 69–78
10