San Jose State University

SJSU ScholarWorks
Master's Theses Master's Theses and Graduate Research

Fall 2009

Active vibration control of a flexible beam.

Shawn Le
San Jose State University

Follow this and additional works at: http://scholarworks.sjsu.edu/etd_theses

Recommended Citation
Le, Shawn, "Active vibration control of a flexible beam." (2009). Master's Theses. 3983.
http://scholarworks.sjsu.edu/etd_theses/3983

This Thesis is brought to you for free and open access by the Master's Theses and Graduate Research at SJSU ScholarWorks. It has been accepted for
inclusion in Master's Theses by an authorized administrator of SJSU ScholarWorks. For more information, please contact [email protected].
 ACTIVE VIBRATION CONTROL OF A FLEXIBLE BEAM

A Thesis

Presented To

The Faculty of the Department of Mechanical and Aerospace Engineering

San Jose State University

In Partial Fulfillment

of the Requirement for the Degree

Master of Science

by

Shawn Le

December 2009
 UMI Number: 1484309

All rights reserved

INFORMATION TO ALL USERS

The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript
and there are missing pages, these will be noted. Also, if material had to be removed,
a note will indicate the deletion.

UMT
Dissertation Publishing

UMI 1484309
Copyright 2010 by ProQuest LLC.
All rights reserved. This edition of the work is protected against
unauthorized copying under Title 17, United States Code.

ProQuest LLC
789 East Eisenhower Parkway
P.O. Box 1346
Ann Arbor, Ml 48106-1346
 ©2009

Shawn Thanhson Le

ALL RIGHTS RESERVED

 SAN JOSE STATE UNIVERSITY

The Undersigned Thesis Committee Approves the Thesis Titled

ACTIVE VIBRATION CONTROL OF A FLEXIBLE BEAM

by
Shawn Thanhson Le

APPROVED FOR THE DEPARMENT OF

MECHANICAL AND AEROSPACE ENGINEERING

ta/oS /0*f
cruy OlS*

Departeft^itof Mechanical and Date

Aerospace Engineering

lojte/ol
Dr. Neyram Hemati, Department of Mechanical and Date
Aerospace Engineering

^ 1 /0/o8/o?
Dr. Winncy Du. Department of Mechanical and Date
Aerospace Engineering

APPROVED FOR THE UNIVERSITY

Associate Dean, Office of Graduate Studies and Research Date

 ABSTRACT

ACTIVE VIBRATION CONTROL OF A FLEXIBLE BEAM

by Shawn Le

There has been tremendous growth in the study of vibration suppression of smart

material structures with lead zironate titanate (PZT) material by the control engineering

community. This thesis considers a cantilever beam with bonded piezoceramic actuators

and a sensor for the study of vibration control. The flexible beam dynamic model is first

derived analytically according to the Euler Bernoulli Beam Theory. The first three mode

shapes and natural frequencies of the beam are constructed analytically and verified with

finite element analysis. The validity of the smart structure was experimentally verified.

The natural frequencies and damping parameters for each mode were experimentally

verified and adjusted. In this study, a transfer function consisting of the first three modes

is constructed to implement both classical derivative (D) and proportional and derivative

(PD) control. Then a state space model consisting of the first two modes of the beam is

constructed to design and implement the modern linear quadratic regulator (LQR) state

feedback control algorithm. A smart-structure beam station was built according to the

instruction of Steven Griffin [6]. The Griffin's analog circuit was modified to integrate

with the Matlab-Quanser real-time control unit. In the analytical and experimental study,

the D, PD, and LQR state-feedback controller provided significant vibration suppression.
 ACKNOWLEDGEMENTS

First and foremost I would like to thank my committee chair and advisor,

Professor Ji Wang, for his guidance and support for making this work possible. I would

like to thank Professor Winncy Du and Professor Neyram Hemati for taking the time and

interest in serving as my committee members. I would like to thank my two close

electrical engineering friends from San Diego, Khang Nguyen and Lam Tran. They have

been great in helping me understand the electrical circuit of this work. I would especially

like to thank my friend and classmate Howlit Ch'ng for keeping me company while

working on this thesis in the San Jose State University Control Lab. In addition, I would

like to thank him for helping me set up and use the Matlab-Quanser real-time control

system.

v
 TABLE OF CONTENTS

LIST OF FIGURES viii

LIST OF TABLES xi

1. Introduction 1

2. Modeling of the Beam 2

2.2 Piezoceramic Actuator Model 6

2.3 Piezoceramic Sensor Model 8

2.4 Derivation of the Transfer Function with Actuator 9

2.5 Derivation of the Transfer Function with Actuator-Sensor 10

2.6 Mode Shapes of the Beam 12

2.7 Impulse Response and Bode Plot of the Transfer Function 14

2.8 Parameters of the Piezoceramic Laminate Beam 17

3. Experimental Setup 18

3.1 Real-Time Experimental Setup with Simulink and Quanser 20

3.1.1 Real-Time Hardware Setup 21

3.2 Experimental Identification 24

4. Vibration Control Method 27

vi
 4.1 Method 1: Derivative Control 28

4.2 Method 2: Proportional and Derivative Control, PD Controller 29

4.2.1 PD Controller Simulation 30

4.2.2 PD Real-Time Control 33

4.3 Method 3: LQR State Feedback with Observer Design- LQR Controller 35

4.3.1 State Space Dynamic Model Derivation 35

4.3.2 Observability and Controllabilty 36

4.3.3 Observer Design 38

4.3.4 LQR State Feedback Gain 39

4.3.5 LQR State Feedback Controller Simulation 39

4.3.6 LQR Real-Time Control 47

5. Results and Discussion 52

6. Conclusion and Recommendations 54

BIBLIOGRAPHY 56

Appendix A Mathcad Analysis 57

Appendix B Matlab M Files 72

vn
 LIST OF FIGURES

Figure 1. Cantilever Beam with Bonded PZT Actuators and Sensor 3

Figure 2. Bending Moment of Actuator 6

Figure 3. Charge Amplifier and PZT Sensor 8

Figure 4. Mode Shape of Beam Derived Theoretically 12

Figure 5. Mode 1 and Mode 2 Respectively with Pro-Mechanica 13

Figure 6. Mode 3 of the Beam with Pro-Mechanica 13

Figure 7. Impulse Reponse of Beam Deflection at the Tip of the Beam Simulated in

Simulink 15

Figure 8. Bode Plot of the Transfer Function of Equation (20) 16

Figure 9. The Impulse Response of the Voltage Sense by the Piezoceramic Material. ... 16

Figure 10. Bode Plot of the Transfer Function Equation (27) 17

Figure 11. Beam Circuit Detail of LM324 Operational Amplifier 19

Figure 12. Beam Circuit Interfacing to Quanser between Griffin's Analog Circuit with

the Maltab-Quanser System 20

Figure 13. Experimental Beam Station Connected to Simulink-Quanser 21

Figure 14. View of Beam Station and Circuit 22

Figure 15. Close Up Top View of Beam Station Showing Piezoceramic Actuator and

Sensor 22

Figure 16. Close Up Bottom View of Beam Station Showing Piezoceramic Actuator ... 23

Figure 17. Quanser DAQ Board with Analog 23

viii
Figure 18. Real-Time Implementation of Open Loop Actuation at the Beam Natural

Frequency 25

Figure 19. Plot of Open Loop Actuation at Resonant Frequency of 97.5 rad/s 26

Figure 20. Open Loop Reponse of Beam Deflected at Approximately 1 inch 26

Figure 21. Plot to Calculate Damping Coefficient of First Mode Open Loop Response. 27

Figure 22. Derivative Controller of Make Analog Circuit 29

Figure 23. Simulink Open Loop Response Simulation of Transfer Function for First

Three Modes 30

Figure 24. Plot of Simulink Open Loop Response Simulation 30

Figure 25. PD Controller Simulation of Beam Vibration Suppression 31

Figure 26. PD Vibration Suppression Simulation, P=40, D=1.5 32

Figure 27. PD Vibration Suppression Simulation, P=17, D=0.01 32

Figure 28. Real-Time PD Controller Implementation Block Diagram Simulink 33

Figure 29. Real-Time Plot of PD Vibration Suppression, P=17, D=0.01 34

Figure 30. Real-Time Plot of Control Voltage 34

Figure 31. State Space Open Loop Response with Initial Condition 40

Figure 32. Plot of State Space Open Loop Response with Initial Condition 40

Figure 33. State Space with Observer and LQR State Feedback 41

Figure 34. Voltage Sense of LQR Controller at a=\, p=\ 42

Figure 35. LQR Control Voltage at a=\, (3=\ 42

Figure 36. Voltage Sense of LQR Controller at «r=10, /?=1 43

Figure 37. LQR Control Voltage at or = 10, p-\ 43

ix
Figure 38. Voltage Sense of LQR Controller at «r=100, /?=1 44

Figure 39. LQR Control Voltage at «r = 100, ft = 1 44

Figure 40. Voltage Sense of LQR Controller at a=l, /? =0.001 45

Figure 41. LQR Control Voltage at a=l, f3 =0.001 45

Figure 42. Voltage Sense of LQR Controller at a=\, /3 =0.005 46

Figure 43. Real-Time Control of LQR Controller with Observer and State Feedback in

Simulink 48

Figure 44. Real-Time Voltage Sense Plot of LQR Control at a=\, fi=\ 49

Figure 45. Real-Time Voltage Sense Plot of LQR Control at a=10, /3=\ 49

Figure 46. Real-Time Control Voltage of LQR Controller at a=10, (3=\ 50

Figure 47. Real-Time Voltage Sense Plot LQR Control at or=100, fi=\ 50

Figure 48. Real-Time Control Voltage of LQR Controller at a=100, /?=1. This data

shows the control voltage before the ±36V limitation of the hardware 51

x
 LIST OF TABLES

Table 1. Constant Values for A. 5

Table 2. Comparison of the 3 Mode Shape Between Pro-Mechanica and Theoretical

Method 14

Table 3. Parameters of Aluminum 6064 Beam 17

Table 4. Parameters of PZT PSI-4A4E 18

Table 5. Comparison Controller Performance Based on Settling Time 53

XI
1. Introduction

The interest of this study is active vibration damping in a flexible structure

bonded with piezoelectric materials such as piezoelectric ceramic material (PZT).

Piezoceramic layers bonded to the surface of or into a manufactured flexible structure

member can act as either control actuator or sensor [6]. The piezoelectric effect consists

of the ability to strain when the crystalline material is exposed to voltage. Oppositely, it

produces electrical charge when strained [1]. A flexible structure with the piezoelectric

elements bonded on it becomes what is called a smart structure. Application of smart

structures range from K2 skis to space structures, where minimal vibration is highly

desirable [6]. This smart material technology may be applied to the construction of high-

rise buildings to counter the devastating effects of vibration from an earthquake [4].

In this study, a cantilever beam with the smart material (PZT) bonded on it was

modeled with the Euler Bernoulli Beam theory [6]. With the model derived, different

controllers could be designed and simulated in Simulink and implemented in real-time to

study the improvement of the dampening effect on the beam.

1
2. Modeling of the Beam

A flexible aluminum cantilever beam with a pair of PZT actuators and a single

PZT sensor was modeled with the Euler-Bernoulli Beam theory. There was a derivation

of the transfer function of the system relating the elastic deflection of the beam to a

voltage applied to the piezoceramic actuator [1]. There was also a derivation of the

transfer function of the relationship between the voltage applied to the actuator and the

voltage induced in the piezoceramic sensor. The transfer function derived was verified

by comparing the first three mode shapes and natural frequencies of the beam to the finite

element analysis result in Pro-Mechanica [1].

2.1 Derivation of the Flexible Beam Mode Shape

A piezoceramic laminate cantilevered beam is illustrated in Figure 1. The beam is

fixed at one end and free at the other end. Two piezoceramic actuators patches and one

piezoceramic sensor (PZT) are used as shown in Figure 1. The parameters in Figure 1

are given in Table 4 . The Euler Bernoulli Beam theory gives the partial differential

beam equation in Equation (1) [1, 8].

2
 -Xs2-

-Xsl-

FZT ACTUATOR PZTSfflSCR

PZT ACTUATOR
Aluninum Baam
I hb
ha 1~
'A
-Xal-
T
-Xa2-

Figure 1. Cantilever Beam with Bonded PZT Actuators and Sensor.

The partial differential equation describing the dynamic of the flexible beam without

damping force is:

d2w 9 2 / ? W
PubbAi 2 + E , r ^ =M (1)
bb 4 a
dt d x dx2

where:
p, = mass density of material of beam
A, = cross sectional area of beam
b
E, = Young's Modulus of beam
b
3
ba /z
h
/. = first moment of inertia, /, = of beam
b b 3
h, = thickness of beam

b = width of beam

w = beam transverse displacement

M = bending moment acting on beam.

3
where R is the generalized location function

R(x) = H(x-x ,)-H{x-x .) (2)

a\ al
and H is the Heaviside function, x , and x ^ are locations of the actuators [81.
al al
The transverse displacement is expressed in terms of infinite series

w(x,t)= S <f»Xx)riAt) (3)

i= l

where 0,(x) is the /' mode shape for the cantilever beam and rj.(t) is the corresponding

generalized displacement. The homogeneous solution to the partial differential Equation

(l)is

Fh ( JC) = A cos fix + B sin fix + C cosh fix + D sinh fix (4)

For a clamp-free cantilever beam, the boundary conditions are:

w(0,t) = 0 (4a)

Bw(0,t)
=0 (4b)
dx

d2w(l,t)
=0 (4c)
dx2

dw\l,t)
=0 (4d)
dx'

where / is the length of the beam. The boundary conditions applied to Equation (4) give

the following 4 equations:

4
 A+C =0 (5)

B+D = Q (5a)

- A cos pi - B sin J3l + C cosh pi + D sinh J3l = 0 (5b)

Acospi -fisin pi + Ccosh pi + Dsinh/?/ = 0 (5c)

The substitution of the first three equations into the last equation results in

(cos pi + cosh pi)2

1+- 2 =0 (6)
(sin yfl/ + sinh2y#/)

To satisfy the boundary conditions, A=0 is to a trivial solution and B = C = D = 0.

This result to

(cos pi + cosh pif

= -1 (7)
(sin2 pi + sinh2 pi)

This equation is reduced to

(cos pi cosh pl) = -\ (7a)

The equation is solved with an infinite number constant of Ptl 's. The first three values

are given in Table 1. Note that (/?/ = A.).

Table 1. Constant Values for X.

i x = P!
1 1.875104069
2 4.694091133
3 7.85475743
The m o d e shape, <f>. (JC) is

'/L^
COS r^i + cosh
(4*
f
-cosh U^ \ I / lx^
0.(*) = C, cos sin -sinhf^ (8)
l
v i y
l
V J sin I —— | + sinh V \ J
v I J

where the constant, Ci, can be determined from the orthogonality expression:

\</>.2(x)dx = l (9)

and

| $.</> .dx — 0 if i^ j (9a)

2.2 Piezoceramic Actuator Model

Two PZT patches are laminated to the top and the bottom of the beam structure

with epoxy glue as shown in Figure 2. The PZT patches have an actuating capability,

which is governed by the piezoelectric constant (d 3 1 ).

Ma=Ca-Va
w
PZT ACTUATOR

BEAM —X

PZT ACTUATOR

Figure 2. Bending Moment of Actuator.

6
The 3 in the d3i implies that the charge is collected on the polarized surfaces or along the

w-axis as shown in Figure 1, and the 1 implies that the force is generated along the

longitudinal x-axis. When a voltage (Va) is applied in the same direction as the

polarization of the piezoceramic electric material, the material is elongated along the x-

axis. The bending moment (Ma) is shown in Figure 2 [2]. When an opposite Va is

applied to the polarized direction, the material is contracted along the x-axis [2]. The

moment induced by the voltage is given in the form of

Ma{t) = CaVa(t) (10)

where the constant, Ca, is given as

C -Ead3lba(hh+hJ (10a)

where

Ea - Young's modulus of the piezoceramic actuator

d3i = electric charge constant (isotropic plane)

b - width of the actuator

hb = thickness of the beam

h, = thickness of the actuator

The total distributed load, qa(x,t), in Equation 1 is given in the form of

2
M JL*&. = qa(Xtt) (11)

7
2.3 Piezoceramic Sensor Model

A PZT sensor is laminated on the top surface of the beam as shown in Figure 3. It

also shows the PZT sensor connected to a charge amplifier.

C2=4.7nf

Figure 3. Charge Amplifier and PZT Sensor.

The structural deformation of the beam induces strain to the laminar sensor. The electric

charge of the piezoceramic sensor (Qs (t)) is equal to the integral of the electric charge

distribution over the entire length of the piezoceramic materials multiplied by the sensor

width (bs) [1]. The electric charge distribution (q(x,t)) is given as

q(x,t) = £c(x,t) (11a)

#31 ,

where /c31 is the coupling coefficient, g31 denotes the piezoelectric voltage coefficient,

and ec (x,t) is the strain in the sensor patch. The strain (£c (x,t)) is related to the

curvature of the beam in the form of

8
 ^^2 w
ec(x,t) = - + (12)
vt S dx 2

The total charge accumulated on the sensing layer can by found by integrating q(x,t)

over the entire area of the piezoelectric sensing element.

x
**2 s2
b
f
^ d2w(x,t)
Qs(t) = -bs \q(x,t)dx = -bs - +t
V^ J V^3iy dx2
%i *.5l

(13)

2.4 Derivation of the Transfer Function with Actuator

From [1], the substitution of the Equation (2) into (1) results to Equation (14).

z
1 =1 dx
(14)

Because \0.2(x)dx = 1, Equation (14) is integrated by w. (x)dx to yield

o
2
Pb\\^2Mdx ijt(t) + EbIh U (x)0"{x)dx W(t) = M (t) \<L*£l0(x)dx (15)
V o V 0 J 0 OX

where

*:{X)=M.{X) (15a)

Equation (15) becomes

(pbAb)fJi(t) + EJ^rJi(t) = Mjt)\^-^-^(x)dx

T
(15b)
o dx2 i

9
which simplifies Equation (15) to the second order Equation (1)

fj.{t) + (foiXt) = ka[4(xa2)-4i(xal)]Va(t) i = l,2,3...

where
6£!=ML#
PA
b b ix

-Ead3lba(hh+ha)
k= ± I
PaK

To include damping, £ , the Equation (16) becomes

fj. (0 + 2£<ym77. (0 + 0)2mTj. (t) = ka [#' (xa2)- ft (xal )]va

Take Laplace transformation of Equation (19) and substitute into Equation (3)

Va(s) ~h s2+2C,conis + co2m

2.5 Derivation of the Transfer Function with Actuator-Sensor

The total charge of the sensor in Equation (13)

x
s2 ( *s2
d2w(x,t)
' * 3 ^
Qs(t) = -bs \q(x,t)dx = -bs • + t„
V ^31 J dx2
%i *s\

evaluated at the position of the sensor, xs2 and xs] give

x
s2
d2w(x,t)
= I ^(Ot^)-^,,)]
dx K 1 =1
sl

10
results to

Qs(t) = K Z Tj.(t)W(xs2)-t(xsl)] (22)

1 =1

where
*31
*,="A \2 a
(23)
j §31,

The relationship between the voltage, Vs(t), and the total charge, Q(t), is given [1] as

G,(0 (24)
vs(t) = c
A(**2-*,i)
where

C is the capacitance per unit area of the piezoelectric sensor

bs (xs2 - xsl) is the surface area of the piezoelectric sensor

Substituting Qs(t) into Vs(t) yields:

Vs(t) = i' = l (25)

C
A(*,2-*5l)

From Equation (19)

(26)

Substitute Equation (26) into Equation (25) yield Equation (27), the transfer function

relating the input voltage of the actuator to the voltage induced by the piezoelectric

sensor.

11
 V,(S) = y KK [<Pi (Xs2 ) ~ <Pi (Xsl )] [#' (*„2 ) ~ J (XaS\
(27)
Va(s)~h Csbs{xs2-xsl)(s2+2Zl(oms + CDl)

2.6 Mode Shapes of the Beam

The first 3 bending mode shapes were plotted from Equation (8) as shown in

Figure 4 and verified with Pro-Mechanica finite element analysis as shown in Figure 5

and Figure 6. The analytical mode shapes analysis agrees well with the finite element

analysis. Table 1 shows a comparison of the 3 natural frequencies of the first 3 modes of

the beam. The analytical natural frequencies of the cantilever beam are obtained from

Equation (17). The natural frequencies of both methods are very close to one another.

Mode Shape 1,2,3

0.5

1 4>Kx) 0
| <l>2(x)
u
.,..""'"-' --r7^~~ __^
1 ^ 0 . 5
Q

-1

"1.5

Length of Beam 0-11.8in

Figure 4. Mode Shape of Beam Derived Theoretically.

12
Figure 5. Mode 1 and Mode 2 Respectively with Pro-Mechanica.

Figure 6. Mode 3 of the Beam with Pro-Mechanica.

13
Table 2. Comparison of the 3 Mode Shape Between Pro-Mechanica and Theoretical
Method.

Model (rad/s) Mode 2 (rad/s) Mode 3 (rad/s)

Pro-Mechanica 95.3618 597.5569 1714
Theoretical (Euler 92.677 580.79 1626
Bernoulli)

2.7 Impulse Response and Bode Plot of the Transfer Function

The impulse response for two cases was simulated in Simulink per Equation (20)

and (27). The impulse response of the first case is a tip deflection (x = I) of the beam

and is shown in Figure 7. The second case is the impulse response of the sensor voltage

and is shown in Figure 9. The Bode plots of the two cases are shown in Figure 8 and

Figure 10. Both Bode plots show the resonant peaks to be at the same location. The

damping coefficient is assumed to be £", 2 3 = 0.01.

14
 w(l,t)

Time (s)

Figure 7. Impulse Reponse of Beam Deflection at the Tip of the Beam Simulated in
Simulink.

15
 ^eUatV^i^pVHpf

••'5 ;

180

so

: i -180

Figure 8. Bode Plot of the Transfer Function of Equation (20).

Voltage

Time (s)

Figure 9. The Impulse Response of the Voltage Sense by the Piezoceramic Material.

16
 -jsof ' yy' ;'^r;:; ::;'r:—; ::: T v'ffnr ""; :i: T : :

RrNueftcy 09$f$es?ji

Figure 10. Bode Plot of the Transfer Function Equation (27).

2.8 Parameters of the Piezoceramic Laminate Beam

Table 3 shows the parameters of the aluminum cantilever beam. The properties

and locations of the PZT actuators and sensor are shown in Table 4.

Table 3. Parameters of Aluminum 6064 Beam.

Properties Units Beam

E (Young Modulus) lb/inA2 1.09E+07
p (density lb/inA3 0.0975
w (width) in 0.6
t (thickness) in 0.065
1 (Length) in 11.8

17
 Table 4. Parameters of PZT PSI-4A4E.

Properties Units Sensor Actuator

d31 (Charge constant) m/V -1.90E-10 -1.90E-10
g31 (Voltage constant Vm/N -1.16E-02 -1.16E-02
k31 (coupling coef.) 0.35 0.35
ba (width) in 0.4 0.4
t (thickness) in 0.0105 0.0105
L (Length) in 0.5 1
x1 (location on Beam) in 1.5 0
x2(location on Beam) in 2 1

3. Experimental Setup

The first beam station is constructed based on Griffin's station from Make [6].

Griffin's beam station suppressed the vibration of the beam without a microcontroller. A

LM324 quad amplifier chip is used for signal processing, derivative control, and as a

bridge amplifier. Figure 11 shows a detail circuit schematic of Griffin's beam station.

First the charge signal from the piezoceramic sensor is passed through a charge

amplifier in the first operational amplifier circuit. The second operational amplifier in the

LM324 serves as a low-pass filter that boosts the input voltage of the first vibration

mode. The potentiometer resistance, R2 in Figure 11, is adjusted to match the resonance

frequency of the beam. The last two sets of operational amplifiers power the two

actuators in tandem. The two bridge amplifiers are then connected to a double pole

double throw (DPDT) phase switch. The phase switch can be switched to the up position

to suppress the beam vibration or to the down position to excite the beam resonant

18
vibration. The principle of the up position of the phase switch is to have the actuator

function in a 180 degree phase shift to counteract the vibration of the beam [6].

^^iM'M'Hilir-ff1

Figure 11. Beam Circuit Detail of LM324 Operational Amplifier.

19
3.1 Real-Time Experimental Setup with Simulink and Quanser

Four 9V
Batteries

^^iltJilhJilHilb-fT

Figure 12. Beam Circuit Interfacing to Quanser between Griffin's Analog Circuit
with the Maltab-Quanser System.

Figure 12 shows the connection between Griffin's beam station and real-time with

the Matlab-Quanser system. Figure 12 shows the second amplifier with a gain of 0.1

which replaces the low pass filter, as shown in Figure 11. The 0.1 gain amplifier

attenuates the input voltage and its output voltage is sent to Simulink-Quanser data

acquisition board. The output signal from the Simulink control block unit is multiplied

by a gain of 10 from the amplifier.

20
3.1.1 Real-Time Hardware Setup

The real-time experimental setup of the beam station is shown in Figure 13.

Figure 14-16 show a close-up view of the cantilevered beam along with actual PZT

actuators and sensor. Figure 17 shows the analog input and output signal connection to

the Quanser DAQ board.

Figure 13. Experimental Beam Station Connected to Simulink-Quanser.

21
 i -'Wrrt'.
™I . T,. ^••p*-! • •
*»

Figure 14. View of Beam Station and Circuit.

IMP

m
ri * i
iiiii

' - ! / • '

• iHB
Ntyi

111 **»&.,. ^» #ff->. i3& 'i|§lll

Figure 15. Close Up Top View of Beam Station Showing Piezoceramic Actuator and
Sensor.

22
 ;
" - ; ^

m
1

Figure 16. Close Up Bottom View of Beam Station Showing Piezoceramic Actuator

Analog Output Analog Input

to Bridge from Charge
Amplifier Ch 1 Arnplifilr Chf

gum • * aims

•^'.••^1 Q i mil II vjy?.>r rvf • •• ' - ^ i •--

tv.
Figure 17. Quanser DAQ Board with Analog.

23
3.2 Experimental Identification

The first experimental study was to examine the open loop beam vibrations. The

experimental setup is shown in Figure 12. The beam was manually deflected by the

operator at the free end of the beam with a tip deflection of approximately 1 inch. With

the PZT sensor and the charge amplifier connected to the Quanser DAQ board, the open

loop response of the beam was examined. The damping parameter was identified to be

£, = 0.005 in Equation (27). The damping parameters for £2 3 was approximated to be

0.001 in this study.

It takes approximately 40s for the beam to settle without the lamination of sensor

and actuators on the beam. The lamination of the PZTs significantly decreases the

settling time of the beam to approximately 7.7s as shown in Figure 20. Due to the

inconsistency of tip deflection of the beam from using a finger, the vibration response

peak voltage is not the same in each measurement in real-time control implementation.

Therefore, a method is imposed to measure the settling time of each vibration response

case without bias. The settling time in this study is defined as

* settling ~ *V=0.\ ~ *V=\0 (27a)

where TV=Q, is the time where the vibration level will be less than 0. IV for t > Tv=0,.

Figure 20 shows a visual detail for Tsettli .

The natural frequency of the first mode matched well with the experimental and

analytical results. The sampling time of the Quanser DAQ board is limited to 100 Hz.

Therefore only the first two modes of the beam vibration could be evaluated for system

24
identification and vibration control. An experimental method to verify the first two

modal frequencies of the beam is to excite the structure and examine its resonance

response, as seen in Figure 19. For the first mode, the resonant frequency was observed

to be at 97.5 rad/s. Similarly, the second mode was observed to be 589.7 rad/s.

• •
signl sign
output
97.5 589.5

Qmnsex

Pv K-
Gain Saturation!
Q4 DAC

Sine Wave An.d.1 o g Output 1

Scope Quansex
1 1 + -K Q4 ADC

+
0.00169493+1
Filter Gainl Anal o g I n p u t Z

-.2

Constant

Figure 18. Real-Time Implementation of Open Loop Actuation at the Beam Natural
Frequency.

25
 T 1 1 r

I I I I I 1 _
0 2 4 6 8 10
Time (s)

Figure 19. Plot of Open Loop Actuation at Resonant Frequency of 97.5 rad/s.

T
settling

T
"1 IIIIII 1III i l l -

illl1 ill
iBiiiiiii it 1HI P1l n nUMppiw
Hnlll'llllllHll^tHWWWJWJUfYJWirtiunkiiLuMuu.
T

1 1
fl
1 IP
ri-
-

i'
0
IIIWIIIP:
"rl"" l_l
2
Ii
4
1i
6
iI
8 10
I I
12

Time (s)

Figure 20. Open Loop Reponse of Beam Deflected at Approximately 1 inch.

26
The damping parameter for the first mode is calculated in Equation (27.2) from [5].

Figure 21 is a close up view of Figure 20 and it provides the data points to calculate the

damping parameter in Equation (27.2).

£ = 2M0•In 'A"
1 1 3.946
-In = 0.0052 (27.2)
2M0 2.848

i ( 1 r
5 X: 4.015
Y: 3.946
4

% 0

M .1
+±
o

>
-2

-3

-4

-5
_I I L
3.8 4.2 4.4 4.6 4.8 5 5.2 5.4 5.1

Time (s)

Figure 21. Plot to Calculate Damping Coefficient of First Mode Open Loop
Response.

4. Vibration Control Method

Three control methods that have been successfully implemented to suppress the

vibration of the beam in this study. The first method of active control comes from

27
Griffin's analog circuit. The second method of active control is the Proportional and

Derivative (PD) control with the modified Griffin's circuit. Simulink in PD simulation

and real-time control was implemented. The third method of active control is the state

feedback with a full state observer. A linear quadratic regulator (LQR) optimal control

method was implemented for vibration suppression in simulation study and real-time

control.

4.1 Method 1: Derivative Control

The derivative control method uses the low pass filter as shown in Figure 11 to

filter the high frequency vibration of the beam, allowing only the first modal vibration of

15.5 Hz to pass through. In addition to low pass filtering, this filter also provides the

derivative of the input signal for derivative control action. This control method is in the

form of

dV
V=-K—'- (27.3)
dt

dV
where K is the control gain, and — L is the derivative of the voltage signal from the
dt

sensor.

Figure 22 shows the result of the active derivative control of the beam. From
Figure 22, the vibration observed (Vs(t)) is considerably reduced within the first 1.5s.

After 1.5s, there is still some minimal vibration that lasts for another 2s until the beam

completely settles to OV.

28
 20

15

10

& 0 Kf*»»>****<w*d V0Htt(Htm\ll<tft*mw >ii.m> > i •wmm

-10

-15 _L
6 10 12
Time (s)

Figure 22. Derivative Controller of Make Analog Circuit.

4.2 Method 2: Proportional and Derivative Control, PD Controller

An impulse open loop response is simulated as seen in Figure 23 and plotted in

Figure 24. Then the PD beam vibration control system is investigated. The PD

controller provides good damping in the beam vibration, resulting in Is settling time in

the simulation study with P=40 and D=1.5, and 2.25s settling time for real-time control

with P=17 and D=0.01.

29
4.2.1 PD Controller Simulation

Figure 23 shows the Simulink diagram for the open loop response. Figure 24

shows the simulated open loop response.

I I
Vs
Open Loop
St ep stari 1
numoverall 1
+ I I
n den overall .01s+1
p +
Scope4
Ju L
Impulse3
VsA/a Low Pass Filter

Figure 23. Simulink Open Loop Response Simulation of Transfer Function for First
Three Modes.

Figure 24. Plot of Simulink Open Loop Response Simulation.

30
The PD control Simulink simulation is shown in Figure 25. The low pass filter with

bandwidth of 100 rad/s passes only the first mode of beam vibration to the feedback loop.

Essentially, the PD controller is only damping the first mode of vibration. If there is no

low pass filter, the noise and the higher frequency modal vibration are amplified with the

derivative action, causing instability in the beam vibration. The best PD controller gains

were found to be at P=40, and D=1.5, as shown in Figure 26. Figure 27 shows the

simulation case with P=17 and D=0.01. The best settling time for PD control simulation

is Is.

n m I |

St ep sta t2 Cent olVol a . 1 Vs

after Low p
* • numoverall
1
PID

M
mi uls «4
P=2
D--0.1
denoverall
VWa2
01s*1
First Order Loirg Pass
Filter

Figure 25. PD Controller Simulation of Beam Vibration Suppression.

31
Figure 26o PO VnlbraMoim Suppression! Snmplatnoim, P=4(D), D=l..

200

Time (s)

Fngunr® 27o PD Vnbraftnonn Smppressioim § , P=17 9 D=(

32
4.2.2 PD Real-Time Control

A simulation implementation of the real-time digital PD control in Simulink is

shown in Figure 28. The sensor voltage is connected to the Quanser analog input box.

The signal goes through a series of signal conditioning units: first a low pass filter, an

offset constant unit and then an amplifier with the gain of 10. Then the signal is fed back

for PD control. Figure 29 and Figure 30 show the real-time PD control beam vibration

and the control voltage with P=17 and D=0.01 with a settling time of 2.25s. The

maximum sensor voltage in Figure 29 is much less than that of the PD simulation in

Figure 26 due to real-time hardware limitation.

10

P Gain
output

Quansez
z-1 Q4 DAC
I Gain5 Discrete-Time
Integratorl Saturation! Analog Outpiitti
Setpoint
0
K- tiu/dt
Derivative
D
Gain

Quansez
Q4 ADC
0.01063+1
Gain2
Low Pass Analog Input 1
10
Filter

.02
input
Offset
Constant

Figure 28. Real-Time PD Controller Implementation Block Diagram Simulink.

33
 Proportion^ Gain =17
Derivative Gain = 0.01

10 12
Time (s)

Figure 29. Real-Time Plot of PD Vibration Suppression, P=17, D=0.01.

Control Voltage - P D Controller

Time (s)

Figure 30. Real-Time Plot of Control Voltage.

34
4.3 Method 3: LQR State Feedback with Observer Design- LQR Controller

4.3.1 State Space Dynamic Model Derivation

From Equation (27), the transfer function is transformed to a state space vector

dynamic equation for state feedback control system design. Since the first two modes are

dominant, and due to the limitation of sampling rate of the Quanser DAQ board of 100

Hz, only the first two modes of the transfer function in Equation (27) will be considered

for state space base optimal control. A second order transfer function for each mode

requires two state variables. There are two modes and one input, so a 4 by 4 system A

matrix and a 4 by 1 system input matrix is needed. The output y matrix is the sensor

voltage, and is a combination of the state from the first mode and second mode of the

beam bending vibration. The state variables are in the form

(28)
X,=T]2(t)

*4=#2(0 = * 5

where the state space matrix dynamic model is in the form

x = Ax + BVa
(28a)
y = Cx

The details of the A, B, and C matrices are in Equation (28a) and shown as

35
 *, 0 1 0 0 \
-aj -2£>Bl 0 0 0
+ k„ K(0
0 0 0 1 ^(•*fl2)-&(*fli)
0 0 -co.nl 0
(28b)

y=Cb [ti(xs2)-tiM ° ^(^2)-^2(^,1) °]

s s{Xs2-Xsl)

4.3.2 Observability and Controllabilty

Observability and controllability of the state space dynamic model are examined

to prove whether the system is state controllable and state observable. The following

relationships give the controllability matrix Co

Co = (B,AB,AzB,...,An-lB) (29)

where A and B are the state space matrices of the system. The matrix Co must be full

rank to be state controllable. Controllability is calculated in Matlab with the obsv and

rank command.

Co = obsv(A,B)
(30)
Controllability = rank{Co)

The observability matrix is given by

0 = (CT,ATCT,...,(AT)"~lCT) (31)

where C is the output state space matrix of the system. Observability is calculated in

Matlab with the following command

36
 ob = obsv(A, C)
(31a)
observability = rank(ob)

The duality between the observer design and the state feedback regulator design

allows for an observer design with the transpose of the A and the C matrix. In this study,

the observability and controllability matrix are full rank. Full rank is the maximum

number of linearly independent columns of the matrix A. The observer design was based

on the pole placement method. The observer gain is calculated with the Matlab

command in Equation (32) and Equation (32a).

Ke=place(A',C',po) (32)

where,

po = the desired poles location (32a)

37
4.3.3 Observer Design

An observer design is required since all 4 states cannot be individually measured.

The separation principle allows the design of the observer to be independent from the

design of the state feedback regulator. The full state observer is in the form of Equation

(33)

x = Ax + Bu + Ke (y - y) (33)
where

y = Cx
y = Cx (34)
u=V„

x = (A-Kec)x + Bu + Key

x = Aobx + [B Ke] (35)

x = Aohx + Bob

where X is the estimated state, and y is the estimated output. For this flexible beam

system, the eigenvalues of the observer matrix are assigned as

-5.0000 + 2.0000i, -5.0000 - 2.0000i, -1.0000 + l.OOOOi, -1.0000 - l.OOOOi

with observer gain (K e ) of

[0.0007, -0.6349, -0.0120, 405.723 if

38
4.3.4 LQR State Feedback Gain

Similar to the observer design, the separation and the duality principle applies to

the LQR design. The state space system in Equation (28a) is controllable. Therefore,

there is a linear state feedback gain (k) that can be found such that the quadratic cost

function (J) is minimized.

J = $(xTQx + v'RVa)dt (36)

where
' l 0 0 0^
0 1 0 0
Q=a (37)
0 0 1 0
0 0 0 1

R=P (38)

where a and /? are scalar value.

The Matlab command in Equation (39) is used to compute the LQR gain matrix.

[k,S,E) = lqr(A,B,Q,R) (39)

The control voltage (Va ) is generated in the form of

V, = -kx

4.3.5 LQR State Feedback Controller Simulation

The open loop response (Figure 32) is simulated in Simulink (Figure 31) to verify

that the model is close to the experimental open loop response. With the observer

39
designed separately, the LQR feedback controller gain matrix was calculated via Matlab

command with Equation (39). Different combinations of a and /? in the Q and R

weighting matrices were evaluated to find the response with the best settling time. Figure

33 shows the complete LQR base control system in Simulink form.

x1 = A>ffBu
y = C>c(-Du
#JOu •
Setpoint = 0 Vs
State-Space
Voltage Sense
Open Loop Response

Figure 31. State Space Open Loop Response with Initial Condition.

9 10

Figure 32. Plot of State Space Open Loop Response with Initial Condition

[.01 0 .01 0].

40
 Control Voltage

* = Ax<-Bu
• y = Cx+Du
Setpoint C1 Vs
atO State-Space 1 Voltage Sense

x1 = Axt-Bu
y = C>ft-Du
^
Observed States
State-Space Observer

-<J?ul
Full State feedback Gain

Figure 33. State Space with Observer and LQR State Feedback

Simulink Simulation.

Some of the Simulink simulation results of the LQR control system are shown in Figures

34-42.

41
 -i r-

Q = 1[1 00 0
0 100
0 010
0 00 1]

R =1
^VH^Ift^TMMftWiftv^wft^'W***''* » ' « « » ' » » ^ n n ^ « ^ ^ w * ^ ^ *

_l I I 1_
4 5 6
Time (s)

Figure 34. Voltage Sense of LQR Controller at a =1, J3=l.

n 1 1 1 r~

Q = 1[1 0 0 0
0 100
0 0 10
0 0 0 1]

R =1

j(t(l0f)>HlimtMttimmMiimim*mm m

0 1 2 3 4 5 6 7 9 10
Time (s)

Figure 35. LQR Control Voltage at a=l, J3=l.

42
 Q = 10*[1 0 0 0;
1.5
0 100;
0 0 10;
0 0 01]

0.5

Oh

-0.5
>

-1.5

4 5 6 10
Time (s)

Figure 36. Voltage Sense of LQR Controller at a =10, j3=\.

0.8 10*1 1 0 0 0;
0 1 0 0;
0.6 0 0 10;
0 0 0 1]

0.4 R=1

0.2

0 IJ^Iiyyiwvwfiftftit^w^v-^*^
0.2
>
0.4

-0.6

-0.8

-1
0 4 5 9 10
Time (s)

Figure 37. LQR Control Voltage at or =10, J3=l.

43
 Time (s)

Figure 38. Voltage Sense of LQR Controller at a =100, p =1.

Voltage Sense, LQR Controller

30

20

10

0 llllni™ L
ppvw--""
-10 Q = 100*[ 1 0 0 0;
<D
0 100;
O)
0 0 10;
B -20
o 0 0 01]
-
>
- R=1
-30

-40
-
-50

-60
I l l
4 5 6 10
-70 Time (sec)

Figure 39. LQR Control Voltage at er =100, p=\.

AA
 Voltage Sense, LQR Controller

Q = 1*[ 1 0 0 0;
0 100;
0 0 10;
0 0 01]
<D "4
O)

•o R=0.001
> -6

-14
10
Time (sec)

Figure 40. Voltage Sense of LQR Controller at a =1, ft =0.001.

50 -| 1 -
-I 1 1 T" -I 1-

Q = 100*[1 0 0 0
-50 0 100
0 0 10
0 00 1]
R=0.001
>-100

-150

-200 _i i_
2 3 4 5 6 10
Time (s)

Figure 41. LQR Control Voltage at a =1, /? =0.001.

45
 0.3 -1 1 1-

Q = [10 0 0;
0.2 0 1 0 0;
0 0 1 0;
0 0 0 1]
0.1
R==0.005

J-0.1
o
>
-0.2

-0.3 -

-0.4
4 5 6 10
Time (s)

Figure 42. Voltage Sense of LQR Controller at a =1, j5 =0.005.

The increase of a in the Q matrix from a = 10 in Figure 36 to a = 100 in Figure 38

significantly dampens the beam vibration from 4.5s to 1.5s settling time with low control

voltage of 8V. The decrease of ft from 1 to 0.001 also significantly dampens the beam

vibration from 6s settling time to 0.7s. However, there is an increase in the maximum

control voltage from 5V volt to 175V. In real-time implementation 175V is not feasible.

The maximum voltage could be applied to the PZT actuator is ±90V. The actual

hardware configuration shown in Figure 12 has a limited control output voltage (Va) of

±36V.

46
4.3.6 LQR Real-Time Control

The LQR controller implementation setup in real-time is shown in Figure 43. It is

important to note that the sensor voltage signal is signal conditioned by a first order low

pass filter with the time constant of T = 0.0109s. The voltage signal is offset to 0V and

multiplied by a gain of 10 because there is an operational amplified signal with a gain of

0.1 before the signal goes to the DAQ board. Similar to the simulation in Figure 33, the

sensor voltage and the actuator control voltage input to the state observer. The observer

estimated states vector is multiplied by the computed LQR full state feedback gain (k) for

control action.

47
 volt sense

0.010949S+1 Quansez
Q4 ADC
Gain 10 Low Pass
Filter
Analog Inputs
-.02

Offset
Constant

£H AX+BJ
y = Cx+l>i
Observer

K*u
CI •
Estimated
State State
Feedback Gain
all states Control Output
Quansei
Q4 DAC

Analog Oufcpirbl

Step

Figure 43. Real-Time Control of LQR Controller with Observer and State Feedback

in Simulink.

The plots in Figure 44-48 show the vibration response with real-time control

implementation of the LQR controller. Note that the settling time in real-time control is

calculated using Equation (27a). For the real-time LQR control, the best control

performance is shown in Figure 47 with a settling time of 1.8s, where a = 100 and (3 = \.

The control voltage calculated by Simulink is about 100V, but the feasible maximum

control voltage is limited to ±36V. The control voltage in Figures 46 and 48 is the

calculated voltage before the ±36V cutoff of the hardware.

48
 15

= 1*[ 1 0 0 0;
0 100;
10 0 0 10;
0 0 01

R=1

IM|HVflliriiMyw"A<ww»>»w».» »

-101

-15
6 10 12
Time (s)

Figure 44. Real-Time Voltage Sense Plot of LQR Control at a =1, f3 =1.

15

Q = 10*[ 1 0 0 0;
10 0 100;
0 0 10;
0 0 0 1]

R=1

? 0 f|***»«uW«~

> -5

-10

-15
6 10 12
Time (s)

Figure 45. Real-Time Voltage Sense Plot of LQR Control at a =10, J3 =1.

49
 Q = 10*[ 1 0 0 0;
0 100;
0 0 10;
0 0 01]

R=1

tfiy^WMW^^w^l^**'^***^!" »^«». iMrtwu^^n

4 6
Time (s)

Figure 46. Real-Time Control Voltage of LQR Controller at a =10, J3=\.

Q = 100"[1 0 0 0;
0 100;
0 0 10;
10 0 0 01]

R=1

f * f t » ^ » » ^ » p * « > » - » ^ f t ^ T V • • » > * w « f * V * f * » • • « * > « • ' » ' * » - • » * * • »w»< *»»•*•*• frV^" W*»»*»

>

-10

10 12 14 16
Time (s)

Figure 47. Real-Time Voltage Sense Plot LQR Control at a =100, fi =1.

50
 Q = 100*[1 0 0 0;
0 1 0 0;
0 0 10;
0 0 01]

R=1

/**tn^0»*+'if*tm*'&**i*mm>m*#W*<ii0*v »** \<l ' mf'i n • n "l " IM »yi i'*«* vtl 'wtf* •

Time (s)

Figure 48. Real-Time Control Voltage of LQR Controller at or =100, J3=l. This

data shows the control voltage before the ±36V limitation of the hardware.

51
5. Results and Discussion

Table 5 summarizes the performance results of the open loop and close loop response

with different controllers. It indicates that the LQR controller with a = 100 and (5 = 1

provided the best vibration suppression with a settling time of 0.5s. As mentioned in

Section 4.1, the classical control method such as the derivative control from [6] provided

good damping in the first 1.5s. However, there was a small sustaining vibration that was

not quickly suppressed after 1.5s. This resulted in a longer settling time of 2.5s. PD real-

time control also provided a fast settling time of 1.75s. In this study, both classical

control and modern control theory were successfully applied for vibration suppression of

the smart structure.

The PD simulation controller 4 and real-time PD controller 5 in Table 5 have the

same P and D gain, but the real-time PD gain provides better settling time performance.

Thus a more aggressive controller gain, such as P = 40 and D = 1.5, is needed for faster

settling time response.

52
Table 5. Comparison Controller Performance Based on Settling Time.

Note: Real-time settling time calculated using Equation

(27a).
Settling
Time
Control Control Response Control
Method Parameters (s) Voltage (V)
Open Loop
1 (Real-Time) 8.75
Open Loop
2 (Simulation) 10

3 D (Make) 2.5
PD P = 40
4 (Simulation) D=1.5 1
PD P = 17
5 (Simulation) D=0.01 <6
PD (Real- P = 17
6 Time) D=0.01 2.25
LQR a = 10
7 (Simulation) 3.5
LQR a=1
8 (Simulation) P = 0.005 1
LQR a = 100
9 (Simulation) 1.2
LQR a =1
10 (Simulation) f3 =0.001 0.75
LQR (Real- a=1
11 Time) 7.5
LQR (Real- a = 10
12 Time) /?=1 2.5 25 max
LQR (Real- a = 100
13 Time) 0.5 12 max

53
6. Conclusion and Recommendations

First a dynamic model of a flexible beam was investigated in this thesis. The

derived model obtained was based on the mode shapes, natural frequencies of the beam

vibration, and the experimental damping parameter of the flexible beam. The response of

the analytical beam model did not exactly fit with that of the experimental model of the

beam. However, the model parameters were adjusted to match closely to the

experimental data. The adjusted model provided a good understanding of the vibration of

the beam. The damping coefficient for the 2nd and 3 rd mode vibration was best estimated

from the open loop response data. Because of the limitation of the hardware sampling

rate, the experimental frequency response of the system was not obtained. Therefore the

damping coefficient parameters also could not be obtained.

In future work, a frequency response experiment should be conducted to obtain a

closer fit of the analytical model to the experimental data. Having a closer fit model, a

more accurate and realistic controllers can be designed.

Other control approaches such as the H controller [8] and the sliding mode

method [10] could also be implemented. The sliding mode has distinct advantages over

the conventional PD approach. The sliding mode method provides robustness, and

improves transient response and control accuracy [10]. This sliding method is worth

studying because of its applicability in wind and seismic structure control.

54
 In this study, the LQR state feedback method provided the best vibration

suppression compared to the derivative control and PD control. Vibration suppression

could be better improved by changing the Q and R weighting matrices. For future work,

the maximum output voltage of the operational amplifiers powering the PZT actuators

can be upgraded from ±36V to ±90V. Doing so will increase the vibration suppression

effectiveness.

55
 BIBLIOGRAPHY

[1] G. L. C. Abreu, and J. Ribeiro, "Spatial Hoo Control of a Flexible Beam Containing
Piezoelectric Sensors and Actuator," University of Uberlandia, Brazil: 2004.

[2] P. Akella," Modeling, Analysis, and Control of Flexible and Smart Structure,"
Michigan: UMI, 1997.

[3] T. Bailey, and E. J. Hubbard, "Distributed Piezoelectric Polymer Active Vibration

Control of a Cantilever Beam," Journal of Guidance, Control, and Dynamics, vol. 8,
no. 5, pp. 605-611, 1985.

[4] S. Chow, Application of Piezo Film for Active Dampening of a Cantilever Beam,
Master's Thesis ed. , San Jose State University: 1993.

[5] J. Fei, "Active Vibration Control of Flexible Steel Cantilever Beam Using
Piezoelectric Actuators," Proceedings of the Thirty-Seventh Southeastern, Tuskegee
University, Tuskegee, Alabama: 2004.

[6] S. Griffin, "Smart Structure," Make: Technology in Your Time, vol. 13, pp. 135-
141,2008.

[7] K. Ogata, Modern Control Engineering , 4th ed. , Upper Saddle River, NJ: Prentice
Hall, 2002.

[8] H. R. Pota, S. O. R. Moheimani and M. Smith, "Resonant Controller for Smart

Structure," Smart Material Structure, vol. 11, pp. 1, 2002.

[9] A. Preumont, Vibration Control of Active Structure, 2nd ed., New Jersey: Kluwer
Academic Publishers, 2002.

[10] G. Song, and H. Gu," Active Vibration Suppression of a Smart Flexible Beam
Using a Sliding Mode Base Controller," Journal of Vibration and Control, vol. 13,
no. 8, pp. 1095-1107,2007.

56
Appendix A Mathcad Analysis

57
Active Vibration Control of a Flexible Beam

1. Beam Dimension and Properties for Aluminum 6061

1:= ll.Sin

Length of the beam

t := .065m

Thickness

w := .6in

Width

P-.W75™
. 3
in

Density

E:= 1.0878107 —
. 2
in

Young Modulus

h:=l
2

wz := .0285iti

Lz := 0.0765m

pz := 7650 - ^
3
m

a := t- w

a =0.039 in2

Cross sectional Area

58
Az := wz-Ls

w 3
I:=t
3

-3 4
I = 4.68 x 10 in

Moment of Inertia

tz := .0005m

f 3 2 ^
tz tz 2 tz
Iz := 2- — + t +t —
U 2 4/

2. Composite Material constant

d31:=-19010-12^-
volt

-9 in
d31 = -7.4S x 10
volt
Electric Charge Constant

ha := .0105in

Length of actuator

la := lin

Length of sensor

Is := .5in

Thickness of PZT actuator

Ea:=6.610 10 —
2
m

Ea = 9.572 x 10 psi

Young Modulus

59
ba:= .4in

width of actuator

-9 in
d31 = -7.48x 10
volt
_ Ead31ba(t+ha)

4
Ca=-1.222 x 10 C

Ea
Cacheck := b a ( t + ha)
2

Cacheck = 1.445 x 105 lbf

Ca=-1.081 x 10 3lbf- —
volt

C is lbf*in/volt

Cacheckl := Cacheck-7.43-10" 9 —
volt

1.445 x 105 lbf—7.4S10 9

— = -1.221 x 10 4

volt

Va := 100V

Max:=2Ca-Va

Max =-0.018 lbfft

k31 := 0.35

Electromagnetic coupling constant

2
g31 :=-11.6-10 •—
C

s31 = -2.031 volt —

lbf
 . 2
g31 = -17.98 —
C

Mode Shape of Beam

Az := wz-Lz

XI := 1.875104069

Eb:=75.1010ii
2
m

12 := 4.694091133

X3 := 7.85475743

cl := .292

c2 := .292

c3 := .292

Mode Shape 1

t = 0.065 in

x\ COS(M) + cosh(j.l)
<|>l(x) := cl cos| XI — cosh X\- sin XI-
1J sin(A-l) + sinh(xi) y i

4
h = 8.255 x 10 m

|1Q = -0.584

1
(j)l(x) dx = 1.006 in
0

Mode Shape 2

cos(&2) + cosh(&2) ( . ( xN
<j]2(x) := c2 cos| X2- cosh £2
sin(X2) + sinh(32) I, I. \j

61
(j»2(r) = 0.584

1
>(i2(x) dx = 1.006 in
0

Mode Shape 3

cosfcj) + coriifc) ( . { x\ . { x^
f3(x) := c3- cos| X3-- - cosh J 3 - - ; ; ; T~ \ Sin X5 - \ - Sintl A 3 -
U I 1 sin(X3) + sinh(A3) V V U V 1>

46(11.Sin) =-0.584

1
(|>3(x) dx= 1.006 in
J
0

Natural Frequency of the first 3 modes

"-T
61 = 1.907 -
ft

•1

Ql := *l(x)- " *l(x) dx

dx
J
0

4
Ql = 6.415 x 10 —
. 3
in

P:=T

S2 = 4.774 -
ft

A3
63:=

62
63 = 7.988 -
ft

Q2
r A
<(i2(x)- ® <(i2(x) dx
dx

Q2 = 0.025 —
. 3
in

r1 A
Q3:= ^(x)- " *3(x)dx
dx
0

Q3 = 0.198 —
. 3
in

H^f
wl = 9 2.677rad

?
I = 2.257 x 10 ft4

4
EI-61
wcheck :=
pa

wcheck= 1.024 x 10 -
ft

_1_

, EIQ2"2
w2
pa

w2 = 580.799rad

63
 „ (EIQ3
w3 :=
^ pa

w3 = 1.626 x 103rad

i ' ' i"

w(x,s) i;k a Kx)i ! Kxa2) - K x a l )

'<• r -2 + 2&
1 ^ Til *• Til'
2

4
Ca=-1.222 x 10 C

Ca
ka:=
pa

N01rE: A is ampere

3
•5 s A
ka = -8.321 x 10"
lb

lb
p a == 1.468
2
s

Vs(s) [ i V V ^ ' ' ^ x s 9 -K x s l ) ! 'K xa2 ) -K x a l )

Va(s) w 2 _ , T,T .,. 2

„ 2 .
-4s in
ka=-9.985 x 10 — C—•
ft lb

Location of Actuator

Position of Actuator

xal := Oin

Position of Actuator

xa2 := xal + la

64
xa2 = 1 in

Find the numerator of the Laplace transform of the system describing the beam tip

deflection

input.

kd:= 1

numlb := kd(j)l(I)

numlb = -0.584

num2b := kdiJSQ

num2b = 0.584

num3b := kdiJGQ

num2b = 0.584

numlb + num2b + num3b = -0.584

3
ka = -8.321 x 10~ 5 ^ - ^
lb

Find the numerator of the Laplace transform of the system describing the elastic

deflection

of the flexible beam due to a voltage applied by actuating the piezoelectric.

The Laplace transform of Vs(s)/Va(s). The is the relation between

the voltage applied to the actuator and the voltage induced in the piezoelectric sensor.

numl := -—<fil(xa2) I - ( -—$l(xal) •ka-^lQ

dxa2 J \ dxal

bs := ba

Width of the sensor is equal to width of the actuator

65
 -6 s A
numl = -8.098 x 10
lb-ft

hs := ha

Thickness of the sensor is equal to the width of the actuator.

t = 0.065 in

ks := —bs I hs + -
2J g31

4
ks= 1.172 x 10 C

(Gustavo)

num2 := -—«|*2(xa2) I - ( -—(PCxal) ka-^CI)

dxa2 I \ dxal

•5 s3-A
num2 = 4.299 x 10"
lb-ft

xsl := xa2 + 0.5in

xsl = 1.5 in

Location of the sensor base

xs2 := xsl + Is

xs2 = 2 in

Location of sensor end

num3 := -—<(G(xa2) I - I -—<P(xal) ka-^Q

dxa2 } \ dxal

-4 s A
num3 = 1.009 x 10
lb-ft

4
ks = 1.172 x 10 C

66
Find the denominator of the transfer function

numsl -—<|.l(xs2) I - [ -—$l(xsl) ks

dxs2 } kdxsl

Numerator of TF for sensor 1

Q. := .007

Damping Coefficient for mode 1

Damping Coefficient for mode 2

- 6 s-A
numsl = -8.255 x 10
ft

-7 C
numsl = -6.879 x 10 —
in
Q := .007

Damping Coefficient for mode 3

Q := .007

nums2 := -—<|i2(xs2) I - [ -—42(xsl) ks

dxs2 J Idxsl

2-Q-wl = 1.297

w 1 is the natural frequency for mode 1

wl = 92.677

wl 2 = 8.589 x 103

• 5 s-A
nums2 =-1.947 x 10
ft

•6C
nums2 = -1.623 x 10

2-^2-^2 = 8.131

67
w2 is the natural frequency for mode 2

w22 = 3.373 x 105

nums3 -—<|G(xs2) 1 - f -—(f3(xsl) ks

dxs2 / Idxsl

w2 = 580.799

2 £ w 3 = 22.768

•5 £^A
nums3 = 1.876 x 10
ft

-6 C
nums3= 1.564 x 10 —
in
w3 is the natural frequency for mode 3

w3 2 = 2.645 x 106

For calculation of state space use for m file

w(x,s) i k a Kx)i- Kxa2) - K x a l )

2
V< S ) ^ S + 2 ^ W -s+iW.V
1
^ ni nr

numslss := -—<t>l(xs2) I - I -—<|>l(xsl) ks-(|>10

dxs2 } Idxsl

nums2ss := -—<|»2(xs2) | - (-—(|>2(xsl) ks$\Q)

dxs2 J \6xs\

numl num2 num3

V,s2 + 2^1wl + wl 2 s 2 + 2-g2w2+w2 2 s2 + 2Q-w3 + w32 J

(.
nums3ss -—<|»3(xs2) -(PCxsl) ks-4>i(rj
^dxs2 dxsl

• 6 s-A
numslss = 4.821 x 10
ft

68
 - 5 s_A
nums2ss = 1.137 x 10
ft

The Transfer function for mode 1,2, and 3 respectively calculated from Matlab

•5 s-A
nums3ss = -1.096 x 10
ft

for State Space B matrix

phil -—$l(xa2) I - f -—«t)l(xal)

dxa2 } \ dxal

phil = -0.014 —
in

phi2 := 'U (J2(xa2) - -—«f2(xal)

\dxa2 dxal /

phi2 = -0.074—
in

phi3:= -—(|3(xa2) J - (-—(f3(xal)

dxa2 J \ dxal

phi3 =-0.173 —
in

for State space C matrix

philxs : -—(fil(xs2) - f -—if.l(xsl)

dxs2 I Idxsl /J

3
philxs = -5.87 x 10 —
in

Combine the numerator for the sensor and actutator to find the overall numerator of the

transfer

function.

69
phi2xs := -—<t»2(xs2) j - f-—<j*2(xsl)
dxs2 ) Idxsl

phi2xs = -0.014 —
in

numlnumsl
Numl overall :=
m
Numl overall = -2.716 x 10 F

phi3xs : -<fG(xs2) P^xsl)

dxs2 dxsl

phi3xs = 0.013 —
in

num2nums2
Num2 overall:

Num2overall = -3.402 x 10 F

philpxs

num3nums3
Nurti3 overall :=

Numl overall = -2.716 x 10 F

phi2pxs

2 2
-19 s 7 A
Numl overall = -9.539 x 10 — s •-
ft lbin

philxa: -—<(>l(xa2) 1 - -—(fil(xal)

\dxa2 J ^dxal

1
philxa = -0.014-

phi2xa := | (Jx2(xa2) - <f(2(xal)

,dxa2 ) \dxz\

70
phi2xa = -0.074 —
in

phCxa := I —<J3(xaZ) - —<f<3(xal)

.dxa2 } Idxal

phi3xa = -0.173 —
in

71
Appendix B Matlab M Files

72
%Properties of Aluminum 6061

1= 11.8; %i n length of beam

t=0.05; %in thickness of beam

w=0.6; %in width of beam

ro=0.0975; %lb/inA3

E = 1.0878E7; %lb/inA2

a=t*w; %inA2

I=t*wA3/3 %Morrient of Inertia

% Properties of PZT

d31=-7.48E-9 %m/voit

ha=.0105 %in height of actuator

hs=ha; %in height of sensor

la=l ' %in, length of actuator

ls=.5 %in, length of sensor

Ea=9.572E6 %lb/inA2

ba=.4 %in, width of actuator

bs=ba %in, width of sensor

Cs=.008E-6; %capacitance per unit area

xsl=3.8; %location of sensor

xs2=4.3; %location of sensor

Ca=Ea*d31*ba*(t+ha)/2; %lb*in/volt Geometry coefficient

73
Ca=2*Ca %two collocated actuator

k31=.35; %coupling coefficient

g31=(-11.6E-3)*(39.368)A2 %inA2/C

omegal=6.415E-4; % I/inA3

omega2=0.025; %]/inA3

omega3=0.198; % l/inA3

wl=(E*I*omegal/(ro*a))A.5; %1st natural freq rad

w2=(E*I*omega2/(ro*a))A.5; %2nd natural freq rad

w3=(E*I*omega3/(ro*a))A.5; %3rd natural freq rad

wla=97.5;

w2a=589.5;

z 1=0.0052; %damping coefficient

z2=0.001;

z3=0.001;

%actuator and sensor constant.

ka=Ca/(ro*a) % inA2/volt

ks=-bs*(hs+t/2)*(k31A2/g31) %Coulomb or can be in*lb/volt

philxa=-0.014; %derivative of mode shape 1 of actuator at location 2 - location 1

phi2xa=-.074; ^derivative of mode shape 2 of actuator at location 2 - location 1

74
phi3xa=-.173; %derivadve of mode shape 3 of actuator at location 2 - location I

philxs=-5.87E-3; ^derivative of mode shape 1 of sensor at location 2 - location 1

phi2xs=-0.014; %derivative of mode shape 2 of sensor at location 2 - location 1

phi3xs=.013; %derivative of mode shape 3 of sensor at location 2 - location 1

% Transfer function of Vs/Va

numsl=ks*ka*philxs*philxa/(Cs*bs*(xs2-xsl)); Enumerator of transfer funciton. first

mode

nums2=ks*ka*phi2xs*phi2xa/(Cs*bs*(xs2-xsl));

nums3=ks*ka*phi3xs*phi3xa/(Cs*bs*(xs2-xsl));

denl=[l 2*zl*wla wla A 2]; %denominator of transfer function, first mode

den2=[l 2*z2*w2a w2aA2]; %denominator of transfer function, 2nd mode

den3=[l 2*z3*w3 w3A2]; %denominator of transfer function, 2nd mode

tf_model=tf(numsl,denl); ^transfer function of first mode.

tf_mode2=tf(nums2,den2);

tf_mode3=tf(nums3,den3);

t=0:.01:5;

Tf_mode=tf_model+tf_mode2+0; % add the transfer for first 3 mode

[numoverall,denoverall] = TFDATA(Tf_mode,V)

damp(conv(conv(den 1 ,den2),den3));

75
^calculate state space

Al=[0 10 0;

-wl A 2-(wl)*2*zlOO;

0001;

0 0-w2 A 2-(w2)*2*z2];

Bl=ka*[philxa 0 phi2xa 0]';

Cl=(ks/(Cs*bs*(xs2-xsl)))*[philxs 0 phi2xs 0];

D1=[0];

D2=[0 0 0 0]';

%step(Al,Bl,Cl,Dl)

p=[-100+j*100 -100-j*100 -500+j*2000 -500-j*2000]

pc=. 1 *p

kl=place(Al,Bl,pc)

%Bode( A1 ,B 1 ,C 1 ,D 1 );grtd;

AC=A1-Bl*kl;

%step(Al,BUCl,Dl)

%step(AC.Bl,Cl,Dl)

C2=[10 0 0;

0 100;

0 0 10;

0001];

76
po=[-l+li-l-li-5+2i-5-2i]';

pob=l*po

ke=place(Al',Cl',po);

ke=ke'

Aob=Al-ke*Cl;

Bob=[Bl ke]

%Contxollabitity and Observability

co=ctrb(Al,Bl)

ob=obsv(Al,Cl)

observability=rank(co)

controllability=rank(ob)

%Iqr

Q=l*[10 0 0 0;

0 100;

0 0 10 0;

0001];

R=.001; %R=.00I works

[ke2,S,E]=lqr(Al,Bl,Q,R);

77