You are on page 1of 6

18th IEEE International Conference on Control Applications

Part of 2009 IEEE Multi-conference on Systems and Control


Saint Petersburg, Russia, July 8-10, 2009

Adaptive Control of Structures with MR Damper


Şaban Çetin, Erkan Zergeroğlu, Selim Sivrioğlu and İsmail Yüksek

Abstract—In this study, an adaptive control scheme is the phenomenological model proposed by Spencer et al.
designed for vibration isolation of SDOF structural system based on a Bouc–Wen hysteresis model [4]. Although
using a semi-active MR (magnetorheological) damper. The Bouc-Wen model can accurately predict MR damper
main goal of using a MR damper system is for structural dynamics, in most cases they are complex and difficult to
protection against earthquake excitation and wind loading. use in control design and parameter adaptation schemes.
A structural system excited by an earthquake and An alternative modeling is based on LuGre friction
controlled by using MR damper is modeled as a semi-active
model [11,19,29-31].
controlled, seismically excited and a nonlinear system with
SDOF. A nonlinear observer is used to estimate of the Nonlinearities and hysteresis dynamics in structures
unmeasurable internal state variable. Unknown parameters incorporating semiactive devices have considerable
of MR damper are obtained by adaptation rules. The effects on the control performance. Various control
necessary input voltage to MR damper so as to generate algorithms used with the semiactive devices can be found
desired damping force is derived by the adaptive control. in the literature such as Lyapunov method [6,13,14],
Finally, response of semi-actively controlled building is decentralized bang bang control [6,14,15], clipped
obtained by simulation. optimal control [3], modulated homogeneous friction
method [16], maximum energy dissipation [15], optimal
I. INTRODUCTION control [9,17], stochastic optimal control [18], H∞ control
Significant progress has been made over the last three [19], backstepping control [20], adaptive control [11, 21],
decades on structural protection against earthquake and sliding mode control [22], robust control [23], fuzzy logic
strong wind loading. After Yao’s case study about [24,25], genetic algorithms [25]. Results of these studies
structural control [1], many control algorithms and show that the performance of the controlled system is
devices such as MR dampers have been investigated to highly dependent on the choice of control algorithm.
earthquake hazard mitigation [2-12]. In this paper a new nonlinear adaptive controller is
MR dampers are nonlinear semi-active control devices proposed for vibration attenuation of SDOF structural
that have significant potential to mitigate vibration and system using a MR damper. A structural system excited
shocks. MR dampers exhibit hysteretic behaviors when by an earthquake and controlled by using MR damper is
subjected to load. Due to their mechanical simplicity, modeled as a semi-active controlled, seismically excited
high dynamic range, low power requirements, low cost, and a nonlinear system with SDOF. The rest of the paper
large force capacity, and robustness, these devices are is organized as follows. In Section 2, mathematical model
suitable for many applications [9-10]. Although they can of SDOF semiactive structural system model is
only remove the energy from system, recent studies have represented. MR damper model is given in Section 3.
shown that the MR dampers can achieve the majority of Section 4 shows the general structure of control scheme.
the performance of fully active systems [3,6,7].
Section 5 discusses the simulation results. Finally Section
In order to describe the dynamic behaviour of MR
6 provides conclusions.
dampers, several models have been proposed. Several
quasistatic models have been researched [7,26-28].
Although these models are useful for MR damper design, II. MATHEMATICAL MODEL OF STRUCTURAL MODEL
they are not sufficient to describe the MR damper Consider a single degree of freedom semi-active
nonlinear behaviour under dynamic loading, especially structure model depicted in Figure 1. The MR damper
the force–velocity behaviour [10]. To overcome this, installed between the ground and the floor via a rigid
different dynamic models have been developed such as brace. Because of the stiffness of brace is usually strong,
the stiffness of brace can be neglected. The dynamical
differential equation of motion of this semi-actively
This work was supported in part by the İstanbul Metropolitan
Municipality under Grant Project İstanbul.
controlled SDOF structural system can be given as:
1
Ş. Çetin and İ. Yüksek are with the Faculty of Mecahnical
Engineering, Yildiz Technical University, İstanbul, 34349, Türkiye Mx + Bx + Kx = − f − Mxg (1)
(e-mail: {scetin , yukseki }@yildiz.edu.tr ).
2
E. Zergeroğlu and S. Sivrioğlu are with the Faculty of
Engineering, Gebze Institute of Thecnology, Gebze, 41400, Türkiye where f represent control force produced by MR damper;
(e-mail: ezerger@bilmuh.gyte.edu.tr and s.selim@gyte.edu.tr) M, B and K are the mass, damping and stiffness

978-1-4244-4602-5/09/$25.00 ©2009 IEEE 60


coefficients of the structure, respectively; xg is the in previous semi-active control studies [11,19]. Here, f is
the control force produced by MR damper; z denotes an
earthquake acceleration excitation; x is the horizontal
internal state variable related to the MR fluid deformation
displacement of the structure. In this paper, only
and v is the applied voltage that acts as the control input.
horizontal motion is considered, torsional and vertical
The internal state variable z cannot be measured directly
motions are not considered.
because MR fluid is actually enclosed within the damper
cylinder. The other parameters are defined in Table I.

TABLE I
PARAMETERS OF THE MR DAMPER
Symb
Quantity Unit Value
ol
σ0 stiffness of z(t) influenced by
N/(m.V) 28815
v(t)
σ1 damping coefficient of z(t) N.s/m 0.131
Fig. 1 SDOF structure model with MR damper
σ2 viscous damping coefficient N.s/m 29.6
σa stiffness of z(t) N/m 30542
III. MR DAMPER MODEL σb viscous damping coefficient N.s/(m.V
influenced by v(t) ) 16.3
The MR damper typically consist of a hydraulic
cylinder which houses the piston, the magnetic circuit, an a0 constant value V/N 3198
accumulator and MR fluid containing micron-sized
magnetically polarizable ferrous particles. Figure 2 is
depicted the schematic drawing of MR damper. The 60
0V
damper piston contains damper coil and annular flow 40 2V
channels. 20
4V
Force [N]

-20

-40

-60
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time [s]

(a)
60
0V
40 2V
4V
20
Force [N]

-20
Fig. 2 Schematic drawing of MR damper
-40

-60
-0.015 -0.01 -0.005 0 0.005 0.01 0.015
Displacement [m]

(b)
60
In order to describe the dynamic behaviors of MR 0V
40
dampers, several models have been proposed. These 2V
4V
20
include the LuGre friction model [11, 19, 29-31]. In the
Force [N]

problem considered here, the governing equation of 0

produced force by the MR damper is expressed as -20

-40

f = σ a z + σ 0 zv + σ 1 z + σ 2 x + σ b xv (2) -60


-0.1 -0.05 0 0.05 0.1 0.15
Velocity [m/s]

z = x − a 0 x z (3) (c)
Fig. 3 Response of MR damper due to a sinusoidal input
These equations are modified LuGre dynamic friction (a) Force vs. time response (b) Force-displacement response
model described in Reference [31], which has been used (c) Force-velocity response

61
The force versus time, force-velocity, and force-
displacement characteristics of the MR damper due to a 2
Hz sinus wave excitation with amplitude of 0.1 m is
depicted in Fig. 3. As shown in Figure 3, each plot
includes five curves that correspond to the damper
behavior for various input voltages applied to the MR
damper is maintained at a constant level of 0, 2 and 4 V.
The flow properties of MR fluids depend on magnetic
field. The magnetic field is generated by input voltage in
a magnetic coil installed at rod head. The parameter
values of the MR damper model used in the simulation
are chosen as shown in Table I [11].

IV. ADAPTIVE CONTROL DESIGN


The overall structure of proposed controller is
presented in Figure 4. Combining (2) and (3), the MR Fig. 4 Block diagram of the overall adaptive controlled structure
damper force is defined as
fˆ = θˆ11 zˆ − θˆ13 x .zˆ + θˆ21 x + (θˆ12 zˆ + θˆ22 x )v (10)
f = σ a z + σ 0 zv − σ 1 a 0 x z + (σ 1 + σ 2 ) x + σ b xv (4)
where ẑ is the estimate of the internal state variable
Equation (4) can be written as a multiplication and which is obtained by an observer later. We define the
summation of vectors in the following form tracking error signal as follows

e = x − xd (11)
 σa 
σ 1 + σ 2 
f = z vz − x z  σ 0  + [x xv ] 
[ ]  (5)
 σb  Note that the desired input x d is zero for the structural
σ 1 a 0 
system. If we define a new variable r for the error

A more compact form of above equation is given as


r = e + α .e (12)

f = ρ1θ 1T + ρ 2θ 2 T (6)
Then the aim of control is that
where the known regression (or signal) vectors ρ1 , ρ 2 e=x  e, e → 0
 r →0 ⇒
and the uncertain parameter vectors θ 1 , θ 2 are defined as r = x + αx  x, x → 0
(13)

ρ1 = [z v.z − x .z ] ρ 2 = [x x .v ] (7) ~


The estimation error θ is defined as the difference
θ1 = [σ a σ 0 σ 1 .a0 ] θ 2 = [σ 1 + σ 2 σ b ] between the true value of the parameter θ and the
estimated value of the parameter θ̂ that is
The estimated force is defined as
~
θ = θ − θˆ (14)
fˆ = ρˆ 1θˆ1T + ρ 2θˆ2 T (8)
We consider the error dynamics defined as
In matrix form, equation (8) is written as
x + α x )
Mr = M ( 
 σˆ a  = M α x − Bx − Kx − Mxg − f + ˆf − ˆf (15)
σ +σ2 
ˆf =  zˆ vzˆ − x zˆ   σˆ  +
  
  0  [ x x.v ]  1  (9) Yϕ
  σˆ b 
σ1a0  where regression signal vector Y and unknown parameter
vector φ are defined as
Finally it can be written as

62
Y =  xα − 
xg − x − x  z = z − zˆ = − a0 x z (23)
T
(16)
ϕ = [M B K]
where a0 is assumed to be known. A candidate Lyapunov
Substituting (6), (8) and (9) into (15) we obtain the function is defined as
error dynamics as
1 1 2 1 ~T −1 ~ 1 ~ T −1 ~
V = Mr 2 + ~z + φ Γφ φ + θ 2 Γ2 θ 2
Mr = Yφ − θˆ11 zˆ + θˆ13 x .zˆ − θˆ21x − (θˆ12 zˆ + θˆ22 x )v 2 2 2 2

χ 1 1 ~ 2 1 1 ~ 2 1 1 ~ 2
+ θ11 + θ12 + θ13 (24)
− [+ ρ 2θ 2T − ρˆ1θˆ1T − ρ 2θˆ2T
ρ1θ1T ] (17)
2 γ1 2 γ2 2 γ3
~ 1 1 1
= Yφ + χ − (θˆ12 zˆ + θˆ22 x )v − ρ 2θ 2T
   + θ11 ( ~
z − ζ 1 ) 2 + θ12 (~z − ζ 2 ) 2 + θ13 (~z − ζ 3 ) 2
ux 2 2 2

[
− θ11z − θ11zˆ + θ12vz − θˆ12vzˆ − θ13 x z + θˆ13 x zˆ
ˆ ] where Γφ ∈ ℜ 3x3 and Γ2 ∈ ℜ 2 x 2 are positive definite
diagonal adaptation gain matrices; γ 1 , γ 2 and γ 3 are the
From the subsequent stability analysis designing the
control input u x in the form of positive adaptation gains. The derivative of candidate
Lyapunov function (24) is obtained as

u x = Kr + χ + Yφˆ − θˆ11ζ 1 − θˆ12ζ 2 + θˆ13 x ζ 3 (18) ~ ~ ~ ~


V = Mrr + ~z .~z + φ T Γφ −1φ + θ 2 T Γ2 −1θ 2
Here, the input voltage is derived as 1 ~ ~ 1 ~ ~ 1 ~ ~
+ θ 11θ 11 + θ 12θ 12 + θ 13θ 13 (25)
γ1 γ2 γ3
Kr + χ + Yφˆ − θˆ11ζ 1 + θˆ13 x ζ 3
v= (19) + θ 11 ( ~
z − ζ 1 )(~
z − ζ1 ) 
θˆ12 zˆ + θˆ22 x + θˆ12ζ 2 

  + θ 12 (~
z − ζ 2 )(~z − ζ 2 )  P
β 
+ θ 13 (~z − ζ 3 )(~
z − ζ3 ) 
where ζ 1 , ζ 2 , ζ 3 are auxiliary filters and K is a positive
constant control gain with β ≠ 0 . Closed loop error Substituting (20) and (23) into (25), the derivative of
dynamics can be written as Lyapunov function becomes
~ ~T
Mr = − Kr + Yφ − ρ 2θ 2 { ~ ~ ~
V = r − K .r + Y .φ − ρ 2 .θ 2T − θ11.( zˆ + ζ 1 )
[ ]
− θ11z − θˆ11 ( zˆ + ζ 1 ) ± θ11 ( zˆ + ζ 1 ) (20)
~
− θ11.( ~z − ζ 1 ) − θ12 .v.( zˆ + ζ 2 ) − θ12 .v.( ~
z − ζ 2)
− [θ ˆ ]
12 vz − θ12 v( zˆ + ζ 2 ) ± θ12 v( zˆ + ζ 2 )
~ ~
+ θ13. x .( zˆ + ζ 3 ) + θ13. x .( z − ζ 3 )
(26)

+ [θ ]
x z − θˆ13 x ( zˆ + ζ 3 ) ± θ13 x ( zˆ + ζ 3 ) ~ ~ ~ ~
13 z .{− a0 x ~
+~ z }+ φ T Γφ −1φ + θ 2T Γ2 −1θ 2
~ 1 ~ ~ 1 ~ ~ 1 ~ ~
where φ = φ − φˆ is the parameter estimation error. + θ11θ11 + θ12θ12 + θ13θ13 + P
γ1 γ2 γ3
Rearranging (20),

~ ~T ~ Therefore we can design the adaptation rules in the


Mr = − Kr + Yφ − ρ 2θ 2 − θ11 ( zˆ + ζ 1 ) following form
~
− θ11 (~
z − ζ 1 ) − θ12 v( zˆ + ζ 2 ) − θ12 v( ~
z − ζ 2 ) (21) ~ 
φ = −φˆ = −Γφ Y T r
~
+ θ13 x ( zˆ + ζ 3 ) + θ13 x (~
z −ζ3) ~ 
θ 2 = −θˆ2 = Γ2 ρ 2T r
~  (27)
The estimate of internal state variable ẑ is defined as θ11 = −θˆ11 = γ 1 ( zˆ + ζ 1 )r
~ 
θ12 = −θˆ12 = γ 2v( zˆ + ζ 2 )r
ẑ = x − a 0 x ẑ (22)
~ 
and the estimation error as follows θ13 = −θˆ13 = −γ 3 x ( zˆ + ζ 3 )r

63
Then substituting these adaptation rules into (26), we 4
Earthquake Excitation

have
3

V = − Kr 2 − rθ11 ( ~
z − ζ 1) 2

Accleration [m/s2]
− rθ12v( z − ζ 2 ) + rθ13. x (~
~ z − ζ 3) 1
(28)
− a0 x z + θ11 ( ~
~2 z − ζ 1 )(~
z − ζ1 ) 0

+ θ12 ( ~
z − ζ 2 )(~
z − ζ2 ) + θ13 ( ~
z − ζ 3 )(~
z − ζ3 )
-1

-2

Finally, the designed filter is as follows -3


0 1 2 3 4 5 6 7 8 9 10
Time [s]

ζ1 = − a 0 x ζ 1 − r Fig. 5 . Earthquake excitation (El centro)

ζ 2 = − a 0 x ζ 2 − v.r (29) Accleration Response


8

ζ3 = −a 0 x ζ 3 + x r 6
Uncontrolled Case
With Proposed Controller

4
Substituting these filters into (28) leads to that

Accleration [m/s2]
2

V = − Kr 2 − a0 x ~z 2 − θ11a0 x ( ~z − ζ 1 )2 − θ12a0 x ( ~
z − ζ 2 )2 0

− θ13a0 x ( ~
z − ζ 3 )2 -2

(30) -4

-6
Due to last three parameters are always negative, the
derivative of Lyapunov function becomes -8
0 1 2 3 4 5 6 7 8 9 10
Time [s]

V ≤ − Kr 2 − a0 x z 2 (31) Fig 6. Acceleration of the structure

-3
We can conclude that when the control gain K is 5
x 10 Displacement Response

selected positive, V ∈ L∞ r → 0 and 4


Uncontrolled Case
With Proposed Controller
~
z → 0 as t → ∞ hence the system is stable. 3

2
Displacement [m]

0
V. SIMULATIONS AND DISCUSSION
-1

-2
In order to verify the effectiveness of the proposed
-3
controller, the following numerical simulations are
-4
performed with MATLAB-Simulink environment. The
approximated parameters of the structure; mass of the -5
0 1 2 3 4 5 6 7 8 9 10
Time [s]
system is 2.8 kg, damping coefficient is 4.9 Ns/m and the
stiffness coefficient is 5815 N/m. In the simulation, we Fig 7. Displacement of the structure
use the NS component of the 1940 El Centro seismic data
as ground acceleration shown in Figure 5.
In Figure 6 the accelerations of the structure are
In the simulation, the performance of the proposed compared. The displacement of the structure is given in
controller was compared with uncontrolled case. Figure 7. As it can be seen from Figure 6, the proposed
Comparisons between controlled and uncontrolled case controller achieved a good performance to reduce
are evaluated in terms of acceleration and displacement acceleration of the system. Adaptive controller provided
of the structure. The results of the simulations are a large amount of decrease in displacement of the
reported in Figures 6-7. structure.

64
VI. CONCLUSION [14] Dyke S.J. and Spencer Jr. B.F., “A Comparison of Semiactive
Control Strategies for the MR Damper”, IEEE, 1997.
In this study, the design of an adaptive control [15] McClamroch, N.H. and Gavin, H.P. “Closed Loop Structural
scheme for the vibration isolation of a SDOF structural Control Using Electrorheological Dampers,” Proc. of the
system using a semi-active MR damper is presented. The Amer. Ctrl. Conf., Seattle, Washington, pp. 4173-77 1995.
stabilization of the structural system is provided with the [16] Inaudi, J. A., ‘‘Modulated homogeneous friction: A
proposed control. The performance of the proposed semi-active damping strategy’’ Earthquake Engrg. and Struct.
Dyn., 26(3), 361–376, 1997.
controller is compared with uncontrolled case. The aim [17] Gavin H.P. and Aldemir U., “Optimal Control of Earthquake
of this controller is earthquake hazard mitigation of Response Semiactive Isolation”, Journal of Engineering
structural systems and to ensure the safety of systems. Mechanics, Vol.131, No.8, August 1, 2005.
Simulation results show that the adaptive controller [18] Ying Z.G., Zhu W.Q. and Soong T.T., “A Stochastic Optimal
scheme is able to achieve good performance in the SDOF Semi-Active Control Strategy for ER/MR Dampers”, Journal
of Sound and Vibration, 259(1), 45-62, 2003.
structural system despite the parametric uncertainty.
[19] Sakai C.,Terasawa T. and Sano A., “Integration of Bilinear
However the proposed control requires the exact H∞ Control and Adaptive Inverse Control for Semi-Active
knowledge of the internal parameter of a0 of MR damper. Vibration Isolation of Structures”, Proceedings of the 44th
IEEE Conference on Decision and the European Control
Conference 2005 Seville, Spain, December12-15, 2005.
[20] Villamizar Rodolfo, Ningsu Luo, Dyke Shirley J. and Vehi
Josep “Experimental Verification of Backstepping Controllers
REFERENCES for Magnetorheological MR Dampers in Structural Control”,
[1] J. T. P. Yao, “Concept of structural control”, ASCE J. Struct. Proceedings of 13th Mediterranean Conference on Control
Div. 98, 1567-1574, 1972. and Automation, Limassol, Cyprus, June 27-29, 2005.
[2] Housner GW et al. “Structural control: past, present, and [21] Song X., Ahmadian M., Southward S. and Miller L.R., “An
future”, J Engng Mech ASCE 1997;123(9):897–971. Adaptive Semiactive Control Algorithm for
[3] Dyke S.J., Spencer Jr. B.F., Sain M.K. and Carlson J.D., Magnetorheological Suspension Systems”, Journal of
“Modeling and Control of Magnetorheological Dampers for Vibration and Acoustics, Vol.127, 493-502, 2005.
Seismic Response Reduction”, Smart Mat. Strc. 5, 565-575, [22] Neelakatan V.J. and Washington G.N., “Vibration Control of
1996. Structural Systems using MR Dampers and a ‘Modified’
[4] Spencer B.F. Jr., Dyke S.J., Sain M.K. and Carlson J.D., Sliding Mode Control Technique”, Journal of Intelligent
“Phenomenological model of a Magnetorheological Damper”, Material Systems and Structures, Vol.0-2007.
Journal of Engineering Mechanics, ASCE, Vol. 123, No.3, [23] Zribi M. and Karkoub M., “Robust Control of a Car
pp.230-238, 1997. Suspension System using Magnetorheological Dampers”,
[5] Wereley N.M., Pang L., Kamath G.M., “Idealized hysteresis Journal of Vibration and Control, 10:507-524, 2004.
modeling of electrorheological and magnetorheological [24] Choi KM, Cho SW, Jung HJ, Lee IW. “Semiactive fuzzy
dampers”, J. Intell Mater Syst Struct, 9(8):642–9, 1998. control for seismic response reduction using MR dampers”,
[6] J Jansen M.L. and Dyke S.J., “Semiactive Control strategies Earthquake Engineering and Structural Dynamics,
for MR Dampers: Comparative Study”, Journal of 33(6):723–736, , 2004.
Engineering Mechanics, ASCE, Vol. 126, No.8, pp.795-803, [25] Yan G., and Zhou L.J., “Integrated Fuzzy Logic and Genetic
August 2000. Algorithms for Multi-Objective Control of Structures using
[7] Yang G., Spencer Jr. B.F., Carlson J.D. and Sain M.K. MR Dampers”, Journal of Sound and Vibration, 296, 368-382,
“Large-scale MR Fluid Dampers: Modeling and Dynamic 2006.
Performance Considerations”, Engineering Structures, 24, [26] Gavin, H. P., Hanson, R. D., and Filisko, F. E.,
309-323, 2002. ‘‘Electrorheological dampers, part 1: Analysis and design.’’ J.
[8] Alvarez L. and Jimenez R., “Real time identification of Appl. Mech., ASME 63, 669–675, 1996.
magneto-rheological dampers”, 15th Triennial World [27] Kamath, G. M., Hurt, M. K., and Wereley, N. M., ‘‘Analysis
Congress, Barcelona, Spain, 2002. and testing of Bingham plastic behavior in semi-active
[9] Aldemir U., “Optimal Control of Structures with electrorheological fluid dampers’’ Smart M. Str.,5, 576–90,
Semiactive-tune Mass Dampers”, Journal of Sound and 1996.
Vibration, 266, 847-874, 2003. [28] Wereley N M and Pang L., “Nondimensional analysis of
[10] J Yang G, Spencer BF, Jung H-J, Carlson JD. “Dynamic semi-active electrorheological and magnetorheological
Modeling of large-scale magnetorheological damper systems dampers using approximate parallel plate models”, Smart
for civil engineering application”, Journal of Engineering Mater. Struct. 7 732–43, 1998.
Mechanics (ASCE) 2004; 130(9):1107–1114. [29] Jimenez R. and Alvarez L., "Real time identification of
[11] Terasawa T., Sakai C., Ohmori H. and Sano A., “Adaptive structures with magnetorheological dampers", Proceedings of
Identification of MR Damper for Vibration Control”, 43rd the 41st IEEE Conference on Decision and Control, pp.
IEEE Conference on Decision and Control , Atlantis, Paradise 1017-1022, 2002.
Island, Bahamas, December 14-17 2004. [30] Jimenez R. And Alvarez-Icaza L., “LuGre friction model for a
[12] Jung Hyung-Jo, Choi Kang-Min, Spencer Jr. Billie F., and Lee magnetorheological damper”, Structural control and health
In-Won “Applicatin of some semiactive control algorithms to monitoring, 12:91-116, 2005.
a smart base-isolated building employing MR dampers”, [31] Sakai C., Ohmori H. and Sano A., “Modeling of MR Damper
Structural Control and Health Monitoring, 13:693-704, 2006. with Hysteresis for Adaptive Vibration Control”, Proceedings
[13] Leitmann, G., “Semiactive control for vibration attenuation”, of the 42nd IEEE Conference on Decision and Control Maui,
Journal of Intelligent Mat. Struct., 5, 841-846, September Hawaii USA, December 2003.
1994.

65

You might also like