Mechatronics 33 (2016) 23–33

Contents lists available at ScienceDirect

Mechatronics
journal homepage: www.elsevier.com/locate/mechatronics

Multi positive feedback control method for active vibration suppression

in flexible structures
Ehsan Omidi, S. Nima Mahmoodi∗, W. Steve Shepard Jr.
Nonlinear Intelligent Structures Laboratory, Department of Mechanical Engineering, The University of Alabama, Tuscaloosa, AL 35487-0276, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents a novel controller for active vibration reduction in piezoelectric-actuated flexible struc-
Received 27 May 2015 tures. The new Multi Positive Feedback (MPF) control consists of two sections where each operates using a
Accepted 6 December 2015
set of actuators. One section uses position and the other uses velocity feedback to perform real-time vibra-
Available online 30 December 2015
tion control. The 90° phase difference between the feedbacks leads to non-zero resultant control output. This
Keywords: property results in a more effective operation of the available actuators. Sections of the MPF controller are
Active vibration control designed based on the concept of the Modified Positive Position Feedback (MPPF) control approach, where
Collocated system the damping of the control system is increased by addition of a first-order term parallel to a second-order
H2 and H∞ norms compensator. The controller is designed to simultaneously control single or multi resonant frequency vibra-
Multi-input multi-output system tions. Due to the high influence of gain values on the control performance, H2 and H∞ norms are used to
Piezoelectric actuator optimize these gains. For validation purposes, the controller is verified here numerically and experimentally
for vibration control of a clamped–clamped beam and a cantilever at resonance. According to the results, the
MPF controller has a superior performance compared to the MPPF controller, in addition to effective suppres-
sion on both vibration displacement and velocity. Using the MPF controller in multimode and by utilizing the
H∞ gain optimization method, vibration displacement amplitudes were reduced to 19.4% of the uncontrolled
state for the beam.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Active vibration control using piezoelectric actuators is one of the
most popular and practical approaches to reduce the amplitude of
unwanted vibrations in flexible structures. This popularity is due to
the light weight of the actuating elements, the lack of moving me-
chanical parts, the relatively easy fabrication process, and the low
cost and low power consumption. Piezoelectric materials have been
used as ceramics and PVDFs for actuation and sensing in studies and
for various purposes [1–4]. This broad usage of piezoelectric mate-
rials and their special application in vibration control at both micro
and macro scales requires effective and compatible controllers. As a
matter of fact, performance of an active vibration control setup could
be highly improved by modifications in the control system software;
similar to other systems that the controllers play very important roles
[5–7]. This improvement in efficiency provides the chance of using
smaller and fewer numbers of piezoelectric actuators and sensors,
which results in a lower implementation cost.
There are several reasons for applying active vibration to struc-
tures. One is to increase the precision of machines, such as in space-
∗
Corresponding author. Tel.: 1 205348 5056.
E-mail addresses: eomidi@crimson.ua.edu (E. Omidi), nmahmoodi@eng.ua.edu
(S.N. Mahmoodi), sshepard@eng.ua.edu (W.S. Shepard Jr.).
craft [8]. Another reason is to prevent the damaging effects of res-
onant vibrations in spatial structures [9,10] or to reduce the gener-
ated noise [11,12]. In flexible structures, large deformations or de-
flections case geometric nonlinearities due to nonlinear curvatures
[13]. In these cases, the vibration controller should also account
for the nonlinearities and methods such as multiple scales are re-
quired to obtain the modulation equation [14,15]. Even if the vi-
bration amplitude does not exceed the allowable level, transverse
resonant vibrations may eventually lead to fatigue failure. Suitable
controllers have been proposed and utilized for each of the pur-
poses mentioned above. Just a few examples of these controllers
include: positive position feedback (PPF) [16], filtered dynamic in-
version [17], Hybrid Positive Feedback (HPF) [18], state-switched
method [19], sliding mode control (SMC) [20], H∞ static-output
feedback control [21], and delayed position feedback [22]. Recently,
a set of network-based vibration control methods have been pro-
posed, where consensus PPF method is the leading example [23].
Although vibration amplitude reduces to a lower level using the
PPF controller, the resulting closed-loop system is more flexible
and consequently, the steady-state error is large [24]. It has been
shown in Ref. [25] that the performance of the PPF controller im-
proves when a first-order term is used parallel to the second-order
compensator. This increases the damping of the system and results
in lower steady-state error compared to the conventional method,

http://dx.doi.org/10.1016/j.mechatronics.2015.12.003
0957-4158/© 2015 Elsevier Ltd. All rights reserved.
24 E. Omidi et al. / Mechatronics 33 (2016) 23–33
Fig. 1. Clamped–clamped beam with the implemented actuators and sensors.
which is the basic of Modified Positive Position Feedback (MPPF)
control [26,27].
According to the concept of transverse vibrations in continuous
systems, all points on the structure cross their considered neutral
axes at the same time. If the positive feedback controller uses just one
type of feedback i.e. displacement, velocity or acceleration, the net
control output of all actuators would have infinite zero points while
the vibration is still in progress. Harmonic vibration displacement
phase lags velocity by 90°. Therefore, utilization of both displacement
and velocity feedbacks will result in a non-zero overall control out-
put. This is the main contribution of the current work in comparison
to aforementioned methods.
In this paper, the Multi Positive Feedback (MPF) controller is pre-
sented as a new method for active vibration suppression in smart
structures. In this method, the vibration control system is divided
into two sections where one uses the vibration position and the other
uses the vibration velocity for positive feedback control. This con-
current implementation of displacement and velocity feedback un-
der two separate modified second-order PPF compensators provides
a high-level suppression using the non-zero overall control output.
Another advantage of the MPF approach is in suppression of both vi-
bration displacement and velocity to lower levels in multimode con-
Fig. 2. Block diagram of closed-loop control system.
trol. Since in vibration control systems magnitude of the acquired
feedback is compared to the zero reference of the same type, that
specific vibration signal is suppressed to a lower level compared to
the other signals. This is due to the fact that each resonant mode has
a different contribution to the aggregate vibration displacement, ve-
locity or acceleration. For instance, typically first mode contributes
to the vibration displacement more that velocity, and second mode
to velocity more than displacement. Hence, when both displacement
and velocity feedbacks are implemented, resonant modes gain the re-
quired feedback amplitude and consequently control effort for both
vibration displacement and velocity. The MPF always senses the high
amplitude vibrations in the structure in either form of displacement
or velocity, and implements the control effort toward suppression of
both of them.
The system model is considered linear, and the MPF is naturally a
multi-input multi-output (MIMO) control system due to its require-
ment for more than one input/output. Additionally, since the MPF
controller is a collocated control system, it is resilient to the desta-
bilizing effect of spillover, which is the result of channeled control
energy to higher modes [28]. To achieve a high level of vibration sup-
pression for a wide range of disturbances, it is essential to perform
multimode control rather than single-mode control [27]. As it has
been shown in [27], vibration suppression under multimode condi-
tions has its own challenges. The performance of the controller in
suppression of each mode is not only sensitive to the assigned val-
ues in the gain vectors of that mode, but also to the gain values of
the other modes. This cross-sensitivity is due to the limited avail-
able control power by the actuators. In addition, there are too many
gain values to be found by trial and error for an effective perfor-
mance. The MPF controller has two gains for each mode on each actu-
ation/sensing patch. Hence, efficient gain optimization methods have
to be utilized to guarantee the optimal level of suppression in the
system. Here, H2 and H∞ norms of the closed-loop system have been
used for controller performance optimization using the gain vectors.
Stability criterion of the closed-loop system is also extracted and the
 E. Omidi et al. / Mechatronics 33 (2016) 23–33
Fig. 3. The clamped–clamped beam with the piezoelectric actuators.
Fig. 4. Implemented devices in the experiment setup.
resulting gains should be constrained. To evaluate the new approach,
a clamped–clamped aluminum beam is used as a test platform to
numerically and experimentally investigate the effectiveness of the
controller.
2. Problem definition
The MPF controller is designed based on the dynamics and char-
acteristics of the flexible structures. Thus, the dynamics of the system
has to be studied first. Fig. 1 shows the schematic top view of the con-
sidered clamped–clamped beam and the MPF controller components.
There are two sets of actuation/sensing patches and they are placed
at the same distance from each end.
One of the difficulties with active vibration control of flexible
structures is the large number of participating resonance modes.
The reason is that the flexible structures are distributed parameter
systems. There is a possibility that if the dynamic modeling of the
system does not include the higher modes, control energy is chan-
neled to these residual modes resulting in spillover [29]. On the other
hand, when there are more modes included in the controller de-
sign, effective performance of the controller on the considered modes
decreases. Therefore, a reasonable number of modes have to be in-
cluded in the controller design, while no dominant mode with rea-
sonable amplitude is neglected. Considering that requirement and
to make the closed-loop system analysis possible, the first N modes
of the flexible structure are considered and the resulting transfer
25
function of

N
σi σ  i
G (s ) = +d (1)
i=1
s2 + ηi s + ωi2
is extracted via the modal analysis technique. In Eq. (1), ηi = 2ςi ωi ,
where ς i and ωi are damping ratio and frequency of the ith mode,
respectively. σ i is a vector of size 2n × 1, where 2n is the number of
the actuators/sensors on the beam. The prime shows the transpose of
the vector. d is a constant diagonal matrix of size 2n × 2n consisting
of the feed-through terms to compensate for the neglected higher
modes [30].
3. MPF controller
In this section, first the MPF controller is explained and then, the
stability criterion for the proposed controller is extracted.
3.1. MPF controller design
As discussed earlier, the MPF controller consists of two sections
that use two types of feedback signals. The first section is the Dis-
placement Feedback Section (DFS) and the other is the Velocity Feed-
back Section (VFS). These sections are constructed based on the con-
cept of positive feedback control for resonant systems. A first-order
term is added to the second-order compensator for each mode of the
26 E. Omidi et al. / Mechatronics 33 (2016) 23–33

Fig. 5. Frequency response of the clamped–clamped beam.

Fig. 6. Simulated results of single-mode vibration suppression using MPPF and MPF Fig. 7. FFT plots of vibration displacement amplitudes in the controlled and uncon-
controllers. trolled states, (a) MPPF approach, (b) MPF approach.

system to add to the damping of the closed-loop system. The con- system are symmetric in terms of amplitude. One objective of the
troller output targets both vibration displacement and velocity, which MPF controller in addition to its effective vibration suppression level
results in a more comprehensive vibration suppression. The transfer is to reduce both vibration displacement and velocity amplitudes. In
function of the DFS is order to secure this goal, it is essential that the DFS and VFS are placed

N
αi α  i α̃i α̃i symmetrically on the structure for the clamped–clamped case. As a
KDF S (s ) = + , (2) result, both sections can measure the same modes with similar am-
s2 + η̂i s + ω̂i2 s + ω̂i
i=1 plitudes, and implement the feedback for controller output genera-
and the transfer function for the VFS is tion and suppression. This objective cannot be achieved if the sec-
tions are placed asymmetrically. For instance, if the VFS is placed in

N
βi β  i s β̃i β̃i s
KV F S (s ) = + . (3) the middle of the structure, second mode is not observed and conse-
s2 + η̂i s + ω̂i
2 s + ω̂i quently, it is not suppressed by that section. Hence, the overall sup-
i=1
pression on this mode decreases. For other boundary conditions, it
In Eqs. (2) and (3), α i , α̃i , β i and β̃i are gain vectors with sizes of n
is essential to place the DFS and VFS patches such that each active
× 1, (.) stands for transpose of a vector or matrix, and η̂i and ω̂i are
mode is sensed and eventually controlled, in either form of vibration
related to the structural damping and frequency of the compensators.
displacement or velocity.
The value for η̂i is in direct relation with the effectiveness of the com-
pensator for the resonant disturbance. It has to be small enough, but
not too small, to diminish the effectiveness of the first-order term. Ac- 3.2. Stability analysis
cording to [25], the optimal values for ω̂i is equal to ωi , the frequency
at each mode. A block diagram representation of the closed-loop sys- In this section, the stability of the closed-loop system is studied.
tem for the MPF controller is given in Fig. 2. The disturbance, Ds , is Parameters of the closed-loop system are presented in matrix form
exerted to the system that excites the resonant modes. The closed- as
loop control system calculates the required input voltage to the actu-
H = diag(η1 , . . . ηN ),  = diag(ω1 , . . . , ωN ),
ators for the suppression process. The reference is considered to be
zero, which implies the objective of reducing the measured vibration Ĥ = diag(η̂1 , . . . , η̂N ),
ˆ = diag(ω̂1 , . . . , ω̂N ),
amplitudes to minimum. A = [α1 , . . . , αN ], Ã = [α̃1 , . . . , α̃N ],
The positions of the DFS and VFS actuators on the flexible struc-
B = [β1 , . . . , βN ], B̃ = [β̃1 , . . . , β̃N ],
ture are important parameters. Since the structure for vibration sup-
pression is a clamped–clamped flexible beam, mode shapes of the = [σ1 , . . . , σN ]. (4)
 E. Omidi et al. / Mechatronics 33 (2016) 23–33 27

Fig. 8. Experimental vibration amplitude in the MPPF-controlled vibration suppres-

sion, (a) displacement, (b) velocity.
Fig. 9. FFT plot of vibration amplitude in the MPPF-controlled and the uncontrolled
state, (a) displacement, (b) velocity.

Next, the equations of the system are rewritten in the state-space

form. The corresponding equations are
written in matrix form as

ẍ(t ) + H ẋ(t ) + 2 x(t ) =  u(t ), (5)

⎡ẍ(t )⎤ ⎡ H 0 0 0 0
⎤⎡
ẋ(t )
⎤
y(t ) = x(t ) + du(t ), (6) ⎢ v̈ (t )⎥ ⎢ 0 Ĥ 0 0 0 ⎥⎢v̇(t )⎥
⎢ 0 ⎥+⎢ ⎥⎢
0 ⎥⎢ ż(t ) ⎥
⎥
⎣¨ ⎦ ⎢ 0
⎣−B 0
0 I 0
⎦ ⎣
ˆ 2 v(t ) = A y(t ),
v̈(t ) + Ĥ v̇(t ) +  (7) v̂(t ) 0 Ĥ − B dB 
−B dB̃ v̂(t )⎦
˙
0 −B̃ 0 0 −B̃ dB 
I − B̃ dB̃ ẑ˙ (t )
ˆ z(t ) = Ã y(t ),
ż(t ) +  ⎡ 2 ⎤
−  B̃ ⎡x(t )⎤
(8)
 −  A −  Ã −  B
⎢ −A  
ˆ 2
− A dA −A dà 0 0 ⎥⎢v(t )⎥
ˆ 2 v̂˙ (t ) = B ẏ(t ),
v̂¨ (t ) + Ĥ v̂˙ (t ) +  (9) ⎢ ⎥
+⎢−Ã −Ã dA ˆ − Ã dÃ
 0 0 ⎥⎢ z(t ) ⎥ = 0
ˆ ẑ(t ) = B̃ ẏ(t ), ⎣ 0 
ˆ2 ⎦⎣v̂(t )⎦
ẑ˙ (t ) +  (10) 0 0 0
0 0 0 0 
ˆ ẑ(t )

and (12)

 
Av(t ) + Ãz(t ) where I is an identity matrix of size N × N. According to the stabil-
u(t ) = , (11)
Bv̂(t ) + B̃ẑ(t ) ity theorem for second-order systems [31], the system of Eq. (12) is
asymptotically stable if the matrix inequalities of

where x(t) ∈ RN × 1 represents the state-space variables of the struc-

⎡ ⎤
ture and y(t) ∈ R2n × 1 is the measured output, over-dot represents H 0 0 0 0
the derivative with respect to time. v(t) ∈ RN × 1 and z(t) ∈ RN × 1 are ⎢ 0 Ĥ 0 0 0 ⎥
second-order and first-order compensator variables of the DFS, and ⎢ 0 ⎥
⎢ 0 I 0 0 ⎥>0 (13)
v̂(t ) ∈ RN×1 and ẑ(t ) ∈ RN×1 are the corresponding variables for the ⎣−B 0 0 Ĥ − B dB −B dB̃ ⎦

VFS, respectively. The last term, u(t) ∈ R2n × 1 , denotes the input volt-
−B̃ 0 0 −B̃ dB I − B̃ dB̃
age vector to the piezoelectric actuators. The above equations are
28 E. Omidi et al. / Mechatronics 33 (2016) 23–33

Fig. 10. Vibration amplitude in the MPF-controlled vibration suppression, (a) displace-
ment, (b) velocity.
Fig. 11. FFT plot of vibration amplitude in the MPF-controlled and the uncontrolled
state, (a) displacement, (b) velocity.
and
⎡ ⎤
2 −  A −  Ã −  B −  B̃
where GC is the closed-loop transfer function of the system
⎢−A ˆ − A dA
 2
−A dà 0 0 ⎥
⎢  ⎥ and tr. is the trace function. The resulting H2 -optimized MPF
⎢−Ã −Ã dA ˆ − Ã dÃ
 0 0 ⎥>0 (14)
⎣ 0 0 0 
ˆ2 0
⎦ controller is automatically stable, since the existence of a fi-
nite value for the H2 norm is constrained to the closed-loop
0 0 0 0 
ˆ
stability.
are satisfied. The constraints of Eqs. (13) and (14) have to be checked H∞ norm minimization is the other approach considered here.
after the gains are selected using the gain optimization methods to H∞ norm of a transfer function is the distance from the origin to the
ensure the stability. farthest point on the Nyquist graph [33]. The H∞ norm is expressed
as [34]

3.3. Gain optimization using H2 and H∞ norms
As shown earlier, the MPF controller has two gains for each piezo-
electric actuator for each mode. For instance, in the simplest case of
the MPF controller and considering two active modes, the controller
has eight gains that must be specified. It is essential to create a proper
control energy distribution over the active modes, so that the maxi-
mum possible vibration attenuation is secured. To this end, an op-
timization task for selecting those multiple gains is required. First,
an H2 -based gain optimization will be discussed, followed by an H∞
norm minimization approach.
H2 -based gain optimization is one of the methods used to obtain a
suitable set of gains for the MPF controller. This approach consists of
selecting the gains such that the H2 norm of the closed-loop output
is minimized. Subsequently, the H2 norm formulation is [32]
 ∞
1
GC 22 = trGC ( jω ) ∗ GC ( jω )dω,
2π −∞
GC ∞ = supσ̄ {GC ( jω )}, (16)
ω
where ‘sup’ is the supremum over the frequency domain of the sys-
tem. For the MIMO system under multimode conditions, the H∞
norm is known as the maximum of the largest singular value of the
transfer function GC (jω) [35]. In order to obtain the optimal gain val-
ues using H2 or H∞ norm, a search algorithm is designed to find the
minimum norm in the range of considered possible gain values. This
search algorithm is also constrained to the stability conditions of Eq.
(13) and (14), and the acceptable gain set complies with these con-
ditions. Other optimization methods such as genetic algorithm also
may be used to obtain the optimal gain set under each norm, such
as the PPF-based method in [36]. The major outcome of this opti-
mization task for the experimental results is the normalized ratios of
the gains to one another in DFS and VFS, which has the role of split-
(15) ting the available control power between the resonant modes in each
section.
 E. Omidi et al. / Mechatronics 33 (2016) 23–33 29

Fig. 12. FFT plot of vibration amplitude in the multimode MPF-H2 controlled and the uncontrolled state, (a) displacement, (b) velocity.

4. Results and discussions Table 1

The clamped–clamped beam and the piezoelectric actuators sizes.

Having described the control system and design of the MPF con- Clamped–clamped beam Piezoelectric actuator
troller, numerical and experimental evaluation of the system and Description Variable Value Variable Value
controller is described in this section. First, the experimental setup is
described and then, different sections of numerical and experimental Length Lcc 535.90 mm Lpzt 31.75 mm
Width wcc 12.70 mm wpzt 12.70 mm
evaluations are presented and discussed.
Thickness tcc 1.58 mm tpzt 0.50 mm

4.1. Experiment setup

The experiment consists of an aluminum beam with Young’s mod-
ulus of 70 GPa and density of 2710 kg/m3 , which is clamped at both
ends. Two identical piezoelectric actuators are attached to the beam
at 8.7 cm from each end. Those locations are not at nodes of any of the
dominant modes, hence, the vibration amplitudes of those modes are
sensible and subsequently, controllable. Fig. 3 shows the beam with
the attached piezoelectric actuators. PZT No. 1 and 2 are the assigned
piezoelectric actuators for the DFS and VFS of the MPF controller, re-
spectively. The aluminum beam and piezoelectric actuator sizes are
listed in Table 1.
One end of the beam is clamped to a shaker made by Labworks,
Inc.® to produce the required disturbance within the beam. The input
voltage for the piezoelectric actuators is amplified by an amplifier of
model PZD350A, made by Trek, Inc.® . The amplification gain for the
shaker is kept constant at 3.5. A laser displacement sensor by MTI In-
struments, Inc.® is used to measure displacement for the DFS of the
MPF controller at PZT No. 1. Velocity feedback for the VFS is measured
by a Polytec VibraScan Laser Vibrometer® system at PZT No. 2. The
Vibrometer consists of an OFV056 Scanning Head and an OFV 3001S
Vibrometer Controller. Signal processing and controller implementa-
tion is through a dSPACE® board that uses ControlDesk® and MATLAB
Simulink® software. Fig. 4 shows the devices and layout used in the
experiment investigation. Note that the laser sensors are on a plat-
form that is separate from that supporting the excitation source and
beam.
4.2. Frequency response of the beam
The first step in the experimental evaluation is to investigate
the characteristics of the resonant system in the frequency domain.
30 E. Omidi et al. / Mechatronics 33 (2016) 23–33

A chirp input is applied to the clamped–clamped beam using the

shaker, and the vibration displacement is measured at the location
of PZT No. 1. Fig. 4 shows the frequency response of the system to the
chirp signal. This graph is normalized by the maximum amplitude. In
the range of the applied signal, the first three modes exist with fre-
quencies of 25.8, 68.0 and 129.4 Hz, respectively. According to these
results, the amplitude of vibration for the first and the second modes
are significantly greater that the third mode amplitude, and conse-
quently the rest of the modes. Damping ratios of the first and second
resonant modes are selected as 0.032 and 0.013 respectively, to have
the agreement between the numerical and experimental models.

4.3. Comparison of the MPPF and MPF controllers

In order to draw a comparison between the performance of the

newly proposed controller and current approaches, the former MPPF
[25] and the new MPF controllers are each used to suppress the vibra-
tions at the second mode of the clamped–clamped beam. The beam
is excited at its second resonant frequency using a pure tone sig-
nal, and then, the MPPF and MPF controllers are applied in differ-
ent evaluations. Numerical evaluation is conducted first, where the
MPPF and MPF controllers are activated after 5 and 4 s, respectively.
Fig. 6 shows the vibration displacements, normalized to the uncon-
trolled vibration amplitudes. According to the obtained results, vi-
bration displacement amplitudes using MPPF and MPF methods have
been reduced to approximately 29.5% and 10.5% of the uncontrolled
values, respectively. Since vibration velocities also had similar pro-
files, they have not been added here. Fig. 7 compares the Fast Fourier
Transform (FFT) of the illustrated displacement amplitudes of Fig. 6,
for the controlled and uncontrolled states. As it is depicted, ampli-
tude reductions for MPPF and MPF approaches are 23 and 48 dB,
respectively.
Next, similar suppression tasks are performed experimentally.
Fig. 8 shows the corresponding results of the MPPF controlled sys-
tem, for vibration displacement and velocity amplitudes. Vibration Fig. 13. FFT plot of vibration amplitude in the multimode MPF-H∞ controlled and the
displacement is reduced to 35% of the uncontrolled value and the ve- uncontrolled state, (a) displacement, (b) velocity.

locity down to 38% of the uncontrolled value. According to the FFT

comparisons depicted in Fig. 9, vibration reductions are 21 and 17 dB
for displacement and velocity, respectively.
Next, results of the MPF controlled system are illustrated in
Figs. 10 and 11, for vibration amplitudes and FFT results, respectively.
These results parallel the MPPF controller results depicted in Figs. 8
and 9. From Fig. 10, it can be seen that the vibration displacement
amplitude is reduced to 12.5% of the uncontrolled state and the vi-
bration velocity is reduced to 15% of the uncontrolled value. Accord-
ing to the FFT plots in Fig. 11, vibration displacement and velocity re-
ductions are 44 and 37 dB, respectively. Comparing the results of the
controllers shows a remarkable improvement in vibration suppres-
sion performance for the MPF controller over the MPPF controller of
23 dB for displacement and 20 dB for velocity. Now it is useful to eval-
uate the vibration suppression capabilities of the MPF controller in a
multimode case.
Here at the end of this section, a comparison between the
numerical and experimental results is drawn, in order to evaluate the Fig. 14. DFS and VFS placements on the cantilever beam in two cases.
modeling accuracy and agreement of the numerical and experimental
models. Comparing the numerical vibration displacement ampli-
tudes depicted in Fig. 6, with experimental results of Figs. 8 and 10, 4.4. Multimode vibration suppression using the MPF controller
and the FFT results of Fig. 7 with Figs. 9 and 11, it can be seen that the
numerical and experimental results are in close agreement with each In this section, performance of the MPF controller is tested in mul-
other. It is also traceable that the numerical results are slightly better timode state. According to Fig. 5, which is the frequency response of
that the parallel experimental results, as also the numerical results the beam, the first and the second modes of the vibration have rela-
reach to the steady-state condition faster. This is due to the fact that tively large amplitudes when compared to the third mode. Based on
not all real-world parameters and imperfections of the experimental this fact, the first two modes have the most contribution to the un-
devices are considered. However in all results, the better performance wanted vibrations of the beam. Hence, the first and second modes
of the MPF controller over the MPPF approach is obvious. of the clamped–clamped beam are excited simultaneously. The
 E. Omidi et al. / Mechatronics 33 (2016) 23–33 31

Fig. 15. FFT plots of vibration amplitudes in the multimode MPF control of the cantilever, (a) case 1, displacement, (b) case 1, velocity, (c) case 2, displacement, (d) case 2, velocity.

Table 2 bration reduction in multimode is expected to be less than the reduc-

Coefficients of the applied disturbance signal to the
tion in single-mode control. The reason for this expected reduction
shaker.
in performance is that the control power is distributed between all
Amplitude (V) Frequency (Hz) of the participating modes as opposed to being concentrated on con-
First mode 3 25.8
trolling a single mode. The optimal vibration suppression level is to
Second mode 4.5 67.0 reduce the vibration amplitude to the vibration level at the shaker-
clamped base. In this experiment, the vibration displacement ampli-
tude at the base which is caused by the disturbance is 15% of the vi-
Table 3
bration magnitude in the uncontrolled mode at the location of the
Optimized gain values for the MPF controller in multimode suppression. piezoelectric actuator. Taking this into the account, the resonant vi-
bration displacement amplitude with the MPF controller is reduced
α1 α̃1 α2 α̃2 β1 β̃1 β2 β̃2
to 24.1% of the uncontrolled amplitude. Vibration velocity in this pro-
H2 optimized 0.7 0.1 0.3 0.1 0.1 0.1 0.6 0.5 cess is also reduced to 26.0% of the uncontrolled amplitude. Accord-
H∞ optimized 0.2 0.4 0.7 0.5 0.7 0.6 0.2 0.1 ing to Fig. 12 (a), the second mode of vibration experiences a greater
level of reduction than the first mode using the H2 -optimized MPF
controller. In contrast to this, the vibration velocity of the first mode
is reduced by 27 dB, which is marginally more than the level of sup-
pure-tone voltage amplitudes of the applied excitation signal for the pression for the second mode at 25 dB.
shaker in order to produce the multimode disturbance are listed Next, the H∞ optimization approach is used for the gain selection
in Table 2. This signal is amplified before being applied to the task in a similar fashion to the H2 norm optimization, and the gain
shaker. values are listed in Table 3. Here also the stability constraints are con-
First, the H2 -optimized MPF controller is used to attenuate the sidered and the minimum eigenvalue of the matrices of Eqs. (13) and
resonant vibrations. In this approach, the gain vectors of A, Ã, B (14) are obtained as 0.072 and 141.521, respectively. Using the same
and B̃ are used as parameters to create the minimum H2 norm in disturbance input, vibration suppression is performed and the result-
the closed-loop control system in a search algorithm to find the lo- ing frequency spectrums for the response are shown in Fig. 13. Us-
cal minimum. Gain values are listed in Table 3. The stability con- ing this gain-selection approach, vibration displacement amplitude
straints of Eqs. (13) and (14) are also considered in this algorithm, is reduced to 19.4% of the uncontrolled amplitude, and vibration ve-
and the minimum eigenvalue of the matrices are obtained as 0.081 locity is reduced to 23.5%. Although the results of the MPF controller
and 162.106, respectively. The fact that these numbers are positive using both optimization approaches are close and both are satisfac-
implies the stability of the closed-loop system. Fig. 12 shows the fre- tory, the result of the H∞ -optimized MPF controller is slightly better.
quency spectrums of the resulting vibration response when using the The effect of the optimization approaches is traceable in the results.
H2 -optimized MPF controller compared to the uncontrolled state. Vi- Fig. 13 (a) has a lower peak value compared to Fig. 12 (a). This is the
32 E. Omidi et al. / Mechatronics 33 (2016) 23–33

Fig. 16. Time histories of the system response to unknown disturbances, (a) initial disturbance, open-loop, (b) initial disturbance, closed-loop, (c) repeating pulse, open-loop, (d)
repeating pulse, closed-loop.

Table 4
Selected gain values for multimode suppression on cantilever in both
cases.

α1 α̃1 α2 α̃2 β1 β̃1 β2 β̃2

Gain values 0.7 0.5 0.4 0.2 0.4 0.2 0.7 0.5

result of H∞ norm minimization that minimizes the maximum peak

in the frequency domain, while the H2 norm considers the magnitude
of the overall vibration.
As it was discussed in Section 3, placement of the collocated DFS
and VFS patches has a very important effect on suppression perfor-
mance of the MPF controller. In order to evaluate this role in more
detail, a numerical simulation study has been conducted here, in ad-
dition to investigating the system response to unknown disturbances.
A flexible cantilever is considered as the sample structure with first
and second resonant modes at 7.1 and 45.6 Hz, and damping ratios of Fig. 17. Output voltage by each suppression section in case 1 of resonant vibration
0.035 and 0.006, respectively. First, simultaneous excitation of both control.

active resonant modes is studied under two cases of patch position-

ing. In order to provide a better illustration of the patch placements in
each case and their relation with the mode shapes, Fig. 14 is depicted
which shows the mode shapes and control patch in one picture. For
case 1, the DFS and VFS are positioned at 0.3 and 0.45 m from the
clamped end, and in case 2, these values are 0.15 and 0.78 m, respec-
tively. In order to only focus on the role of DFS/VFS positioning, sim-
ilar gain values have been considered for variables of both modes in
both cases, listed in Table 4. The same excitation amplitude has also
been applied for both cases.
Numerical simulations have been performed for the simultane-
ous excitation of the resonant modes and the FFTs of the steady-
state residual vibration amplitudes in comparison to the uncontrolled
states are depicted in Fig. 15. Comparing the results for the two cases
shows that the placement of the control patches under case 1 pro-
vides a more successful suppression performance, nearly twice the
suppression level obtained in case 2. Additionally, it is seen that sup-
pression on vibration velocity is affected more negatively than dis-
placement between cases 1 and 2. The reason is the position of the
VFS in case 2, which is on the node of the second vibration mode.
Next, performance of the controller in response to unknown (not
resonant) input disturbances is investigated. To this end, the sys-
tem of the cantilever beam with settings of case 1 is considered and
open- and closed-loop system responses to an initial disturbance and
a repeating pulse disturbance are studied. According to the results
illustrated in Fig. 16, peak response amplitude to the initial distur-
bance reduces by 80% in the MPF-controlled system. For the case
of repeating pulse input with amplitude of 3.5 and period of 4 s,
 E. Omidi et al. / Mechatronics 33 (2016) 23–33
the oscillations due to sudden changes are completely suppressed, in
addition to reduction in the displacement amplitude of the repeating
sequence.
As it has been mentioned, the control output provided by DFS and
VFS provide a non-zero output due to the phase shift which results
in higher level of suppression using MPF method. To this end, Fig. 17
illustrates the output voltage by each section in case 1 setting of the
cantilever, in vibration suppression regarding to Fig. 15. The outputs
in the figure are for each section in controlled steady-state response.
As it is shown, the outputs have few common zero points. Please note
that it is not physically possible to avoid all zero points when suppres-
sion is in multimode. However, there are not common zero points for
the same modes.
5. Conclusion
Multi Positive Feedback (MPF) control has been introduced in
this paper as a new method for active vibration reduction in flexible
smart structures. MPF uses two types of vibration feedback, both po-
sition and velocity, for two different sets of collocated actuator/sensor
patches. The phase lag between the feedbacks makes the overall out-
put of the control system to be non-zero during the control process.
This leads to a more effective suppression of vibrations, where both
vibration displacement and velocity are suppressed using their di-
rect feedbacks. The controller has been developed for both single-
and multi-mode vibration suppression. A stability analysis was per-
formed and stability criterion for the controller was extracted. Two
gain-selection approaches based on H2 and H∞ norms minimization
of the closed-loop control system were implemented for gain opti-
mization. An aluminum beam that was fixed at both ends was used
as the test platform. In order to evaluate the effectiveness of the pro-
posed controller, first, a comparison between the conventional MPPF
and the new MPF controller was made, both numerically and experi-
mentally. According to the results, the final vibration amplitude after
reduction was approximately 2.8 times lower than the results when
using the MPPF controller in terms of the displacement. Vibration
control in multimode was also evaluated. Vibration displacement us-
ing the H∞ -optimized MPF controller was reduced down to 19.4%
of the uncontrolled state and the reduction for velocity was 23.5%.
In conclusion, the presented MPF controller has a promising perfor-
mance in vibration suppression of flexible smart structures. Addition-
ally, the controller can be used for vibration controls of micro systems
with some minor modifications.
References
[1] Amin-Shahidi D, Trumper DL. Design and control of a piezoelectric driven reticle
assist device for prevention of reticle slip in lithography systems. Mechatronics
2014;24(6):562–71.
[2] El Khoury Moussa R, Grossard M, Boukallel M. Modeling and control of a piezo-
electric microactuator with proprioceptive sensing capabilities. Mechatronics
2014;24(6):590–604.
[3] Seigler T, Ghasemi A. Specified motion of piezoelectrically actuated structures. J
Vib Acoust 2012;134(2):021002.
[4] Seigler T, Ghasemi AH, Salehian A. Distributed actuation requirements of
piezoelectric structures under servoconstraints. J Intell Mater Syst Struct
2011;22(11):1227–38.
[5] Meng, F., Zhang, H., Cao, D., 2015, “System modeling and pressure control of a
clutch actuator for heavy-duty automatic transmission systems,” (doi: 10.1109/
TVT.2015.2404857).
33
[6] Zhang H, Liu X, Wang J. Robust H∞ sliding mode control with pole placement for
a fluid power electrohydraulic actuator (EHA) system. Int J Adv Manuf Technol
2014;73(5-8):1095–104.
[7] Zhu X, Zhang H, Fang Z. Speed synchronization control for integrated automotive
motor–transmission powertrain system with random delays. Mech Syst Signal
Process 2015;64–65:46–57.
[8] Hu Qinglei, Ma Guangfu. Adaptive variable structure controller for spacecraft vi-
bration reduction. IEEE Trans Aerosp Electron Syst 2008;44(3):861–76.
[9] Chomette B, Chesné S, Rémond D. Damage reduction of on-board structures us-
ing piezoelectric components and active modal control—Application to a printed
circuit board. Mech Syst Signal Process 2010;24(2):352–64.
[10] Omidi E, Mahmoodi SN. Multiple mode spatial vibration reduction in flexible
beams using H2 - and H∞ -modified positive position feedback. J Vib Acoust
2015;137(1):011004 (7 pages).
[11] Strassberger M, Waller H. Active noise reduction by structural control using
piezo-electric actuators. Mechatronics 2000;10(8):851–68.
[12] Zolfagharian A, Noshadi A, Khosravani MR. Unwanted noise and vibration con-
trol using finite element analysis and artificial intelligence. Appl Math Model
2014;38(9):2435–53.
[13] Malatkar P. Nonlinear vibrations of cantilever beams and plates. Virginia-Tech;
2003.
[14] Omidi E, Mahmoodi SN. Sensitivity analysis of the nonlinear integral positive
position feedback and integral resonant controllers on vibration suppression of
nonlinear oscillatory systems. Commun Nonlinear Sci Numer Simul 2015;22(1–
3):149–66.
[15] Omidi E, Mahmoodi SN. Nonlinear vibration suppression of flexible structures
using nonlinear modified positive position feedback approach. Nonlinear Dyn
2015;79(2):835–49.
[16] Fanson JL, Caughey TK. Positive position feedback control for large space struc-
tures. AIAA J 1990;28(4):717–24.
[17] Seigler TM, Hoagg JB. Filtered dynamic inversion for vibration control of struc-
tures with uncertainty. J Dyn Syst Meas Control 2013;135(4):041017 (16 pages).
[18] Omidi E, Mahmoodi SN. Hybrid positive feedback control for active vibration at-
tenuation of flexible structures. IEEE/ASME Trans Mech 2014;20(4):1790–7.
[19] Clark WW. Vibration control with state-switched piezoelectric materials. J Intell
Mater Syst Struct 2000;11(4):263–71.
[20] Li L, Song G, Ou J. Adaptive fuzzy sliding mode based active vibration control of a
smart beam with mass uncertainty. Struct Control Health Monit 2011;18(1):40–
52.
[21] Zhang H, Wang R, Wang J. Robust finite frequency H∞ static-output-feedback
control with application to vibration active control of structural systems. Mecha-
tronics 2014;24(4):354–66.
[22] Jnifene A. Active vibration control of flexible structures using delayed position
feedback. Syst Control Lett 2007;56(3):215–22.
[23] Omidi E, Mahmoodi SN. Consensus positive position feedback control for vibra-
tion attenuation of smart structures. Smart Mater Struct 2015;24(4):045016 (11
pages).
[24] Friswell MI, Inman DJ, Rietz RW. Active damping of thermally induced vibrations.
J Intell Mater Syst Struct 1997;8(8):678–85.
[25] Mahmoodi SN, Ahmadian M. Active vibration control with modified positive po-
sition feedback. J Dyn Syst Meas Control 2009;131(4):041002 (8 pages).
[26] Omidi E, Mahmoodi SN. Vibration control of collocated smart structures using
H∞ modified positive position and velocity feedback. J Vib Control 2014. doi:10.
1177/1077546314548471.
[27] Omidi E, Mahmoodi SN. Multimode modified positive position feedback to con-
trol a collocated structure. J Dyn Syst Meas Control 2015;137(5):051003 (7 pages).
[28] Friswell MI, Inman DJ. The relationship between positive position feedback and
output feedback controllers. Smart Mater Struct 1999;8(3):285.
[29] Balas MJ. Active control of flexible systems. J Optim Theory Appl 1978;25(3):415–
36.
[30] Clark RL. Accounting for out-of-bandwidth modes in the assumed modes ap-
proach: Implications on collocated output feedback control. J Dyn Syst Meas Con-
trol 1997;119(3):390–5.
[31] Goh CJ, Caughey TK. On the stability problem caused by finite actuator dynamics
in the collocated control of large space structures. Int J Control 1985;41(3):787–
802.
[32] Colaneri P, Geromel JC, Locatelli A. Control theory and design: An RH2 and RH
viewpoint. Academic Press; 1997.
[33] Voicu M. Advances in automatic control. Springer; 2003.
[34] Kwakernaak H. Robust control and H∞-optimization—Tutorial paper. Automatica
1993;29(2):255–73.
[35] Toscano R. Structured controllers for uncertain systems. Springer; 2013.