Modeling and simulation of non-linear and hysteresis behavior of magneto-rheological

dampers in the example of quarter-car model

Sulaymon Eshkabilov
Dynamics & Control Lab
Tashkent Automotive Road Institute
Amir Temur Str. 20, Tashkent -100060, Uzbekistan
Email: sulaymon@d-c-lab.com

Abstract. This paper presents reviews of several models and numerical simulation models of non-linear and hysteresis behaviors
of magneto-rheological liquid dampers in MATLAB/Simulink in the example of quarter-car model of vehicle suspension
simulation, such as, Bingham, Dahl, LuGre and Bouc-Wen models. In addition, it demonstrates numerical simulation models
built in MATLAB/Simulink and discusses results from numerical simulation models for three different input excitations from
terrain.

Keywords: Bingham model, Dahl model, LuGre model, Bouc-Wen model, passive and semi-active suspension, magneto-rheological damper,
quarter car model, numerical simulation, Simulink model.

1. Introduction

In general, most of the natural phenomena, operational machine processes and dynamic system behaviors are of non-linear nature
that is very often linearized for the sake of simplicity in formulations and analyses. In fact, nonlinear behaviors or phenomena of
processes may create difficulties in studies and engineering design processes but considering some of those non-linear
characteristics of processes or behaviors of dynamic systems carefully could be also very beneficial and of great importance for
efficient and accurate control, and used for operational efficiency and energy preservation or dissipation depending on their
application areas. For example, nonlinear parameters and characteristics of some materials and interactions of different parts
made of different materials have a great potential to apply for dampers and shock absorbers [1]. One of the good examples for
such processes is a hysteresis loop observed in magnetic or magnetized materials and magneto-rheological (MR) liquids. In
studies [2, 3, 4, 5], the MR liquids are found to be one of the most suitable and promising in designing vibration dampers and
shock absorbers, and there are some combinatorial designs [6] of MR fluid dampers. In studies [7], feasibility of MR liquid
damper modeling by employing Bouc-Wen model in association with an intelligent self-tuning PID controller for semi-active
suspension modeling is studied numerically via computer modeling in MATLAB/Simulink. Nevertheless, identification of the
hysteresis loop parameters is rather complex and may require considerable laboratory and numerical studies in order to apply
them and get a best use of MR damper properties.

In this paper, we put some emphases on different mathematical models and formulations of the MR liquids, and their hysteresis
loop parameters and numerical simulation models designed for a semi-actively controlled feedback damper for a vehicle
suspension systems developed in MATLAB/Simulink. In addition, we shall try to analyze and compare efficiency and accuracy
of these models in the example of the quarter-car model to design a semi-active suspension system.

2. Mathematical formulation of a quarter-car model

To derive an equation of (vertical) motion of a vehicle while driving on uneven roads, we take quarter of a vehicle by assuming
that terrain roughness is evenly distributed under all wheels of a vehicle and loading from the whole vehicle body is equally
distributed across all of its axles. In addition, we consider that a tire has some damping effect. With these preconditions, we draw
the next physical model (Figure 1) of the system for passively and semi-actively controlled systems of a quarter-car model.
 a) Paassive suspensio
on design; b) Semi-acttive suspension design.

Figure 1. Vehicle
V suspensi on models.

wn in Figure 1, we can derive equations of m

From the passive and semii-active suspension design show motion of the two mass
bodies whicch as un-sprung mass (half of axle
a mass and one
o wheel) aand sprung mass (quarter car body mass) . So, the
equations off motion of the systems
s are

a) Fo
or passive suspension system:
0
(1)
b) Fo
or semi-active su
uspension system
m:
(2)

Where , and are diisplacement, vellocity and accelleration of the sprung mass (qquarter car boddy mass), respecctively;
, and are displacem ment, velocity an
nd acceleration of
o the un-sprung mass (half of axxle mass and onne wheel), respecctively;
and daamping coefficieents of suspensio on and tire; an nd stiffness oof suspension annd tire; and are terrain rouughness
(disturbancee) displacement and velocity withw respect to longitudinal
l speeed of the vehiccle; is the fforce generated by the
controller thhat takes into account terrain rou
ughness , an
nd vertical displ acement and velocity of the vehhicle. In the moddel, for
- the control force exerrted by the contrroller, we applyy several differeent hysteresis efffect models, succh as, Bingham
m, Dahl,
LuGre and Bouc-Wen
B modeels and design nuumerical simulattion models in MMATLAB/Simullink.

3. Matheematical formullations of the MR

M dampers
3.1. Bingha
am model

To simulatee and identify paarameters of the MR liquids, Bin

ngham plastic m
model [8] was prooposed in 1985. It is formulatedd by the
following:

(3)

Where is a pistons relatiive displacementt and is its derrivative that is veelocity of a pistoon; is frictionnal force; is daamping
constant; is offset force (constant force value). The siggnum function will takee care of the dirrection of the friictional
force dep pending on the relative
r velocity of the hysteresis (internal) vvariable . Note that in our simuulation model, and
correspond to the displacem
ment and veloccity of the spru
ung mass.

The responsse of Bingham model

m correspon
nds to the follow
wing graph showwn in Figure 3 annd it can be assuumed that the shhape of
Bingham model
m force will
w be equal to Coulomb force plus friction forrce ( ). The dam mping coefficiennt (constant) will be
hip between the force and thee velocity diffferences- Figuree 3.
equal to the linear relationsh

Figuree 2. Bingham meechanical model proposed by [6]].

Figure 3. The reesponse of Binghham model.

Now we build a Simulink model

m Figure 4 using the form
mulation from thee equation (3) annd link it with thhe model expresssed for
the semi-acttively controlled
d suspension moddel from the exp
pression of (2) ass shown in Figurre 1.b.
 Figure 4. Bing
gham model emb
bedded in semi-aactive suspensionn control.

Note that thhe Coulomb frictional force ( ) is directly relatted to the yield stress. In Binghham model theree are two input ssignals,
which are and make up
u relative velocity in order to diirect the Coulommb frictional forrce with the ssignum
function, i.ee., .

3.2. Dahl model

m

This model considers quasi--static bonds in the

t origin of fricction [9]. Dahl m
model of the MR damper [8] is foormulated by:

(4)

| | (5)

Where, is exerted forcee from the MR damper,

d is thee control voltagee, is the dynaamic hysteresis ccoefficient, , ,
and are paarameters that coontrol the hystereesis loop shape.

Using the ex xpressions (4) and (5), we build

d a simulation model
m of Dahl moodel in Simulinkk as shown in Fiigure 5. In Dahll model
(Figure 5), there
t is one feed
dback coming fro om the un-spring
g mass that is veelocity annd there is one ouutput signal thatt is
going to thee un-sprung masss and sprung maass.
 Fig
gure 5. Dahl mod
del implemented
d for semi-active control of suspeension system.

In Dahl model, there are onne input signal and output sig
gnal . The innput signal is cooming from the velocity of the sprung
mass that iss a car body vellocity and feeding summing jun nction of fo
force, and summ ming junction off dynamic hyssteresis
coefficient. The output sign
nal is the controll force feeding a summing juncction of input forrces for sprung mass with (-) siign and
for un-sprunng mass with (+)) sign.

3.3. LuGree model

In modelingg the hysteresis loops, the LuGrre model is deveeloped within stuudies [10] and aapplied in workss [11] in modeliing and
simulation of
o dampers. This model [10] tak kes into accountt three types of frictions observved in dry frictioon and fluid flow
ws, viz.
Coulomb, sttick-slip and striibeck effects thatt are formulated by the followinng:

(6)

Where , , are stiffnesss, damping and viscous friction coefficients, resspectively; is the friction sttate (average defflection
of the bristlees), is the velocity
v of the frriction state, is the relative vvelocity of the spprung mass.

| |
(7)

In the abovee expression, is defineed by [12 and 13] that has been eexpressed with thhe following

(8)

Where is the Coulomb friction force, is

i the sticktion fo
orce, and is thhe Stribeck veloccity.

The simulattion model of th

he LuGre modell, as shown in Figure
F 6, is buillt in Simulink w
with one input ssignal that is a rrelative
velocity from
m sprung mass and
a one output signal
s that is con
ntrol force ffor the suspensioon system conneected with a summmation
junction of the
t sprung and un-sprung
u massees with (-) and (+
+) signs respectiively alike Binghham and Dahl mmodels shown in Figure
4 and 5.
 Fiigure 6. Simulink
k model of the L
LuGre model.

In the Simulink model, a fun

nction block witth three input sig
gnals, viz. , , , is em
mployed to comppute a control forrce that
is the MR force
f . The twwo input signalss, which are and , aree internal variables computed from the expressions (7)
and (8).

3.4. Bouc-W
Wen model

The MR dam mper with Bouc-Wen model is composed

c of stifffness (spring) ellement, passive damper and Bouuc-Wen hysteressis loop
elements. Thhe schematic rep
presentation of Bouc-Wen
B modeel of an MR dam mper is depicted bby the next scheematic view Fiigure 7.
The hysteresis loop has an internal
i variablee that represennts hysteretic behhavior and satisffies the next exppression (9). Thee model
equation of Bouc-Wen mod del [8] is expressed by the follow
wing.

| | | | | | (9)

Where is the evolutionary

y variable that caan vary from a sinusoidal
s to a quuasi-rectangularr function of the time dependingg on the
parameters , and .

Figure
F 7. Schem
matic representatiion of Bouc-Wen
en model of an M
MR damper.
The force exxerted by the MR R damper is the function of the relative displaceement and veloocity and the pparameter defiined by
the control voltage
v , and iss given by

(10)

In the modeel computing daamping force of the MR damperr, is the stiffn

fness of the sprinng element of thhe MR damper aand the
values of th
he parameters (co
oefficients) and ve a linearly relaationship with thhe control voltagge and determ
hav mine the
influence off the model on the final force . The force takes into accoount pre-yield sttress of the dammper. The valuess of the
parameters (coefficients)
( and arre determined from the followinng expressions:

, (11)

The best fitt parameter valu

ues of these parrameters are deetermined by fitt
tting to the expeerimentally meaasured response of the
system.

The simulattion model of the system fromm Bouc-Wen mo odel shown in F Figure 8 is buillt in Simulink bby using the equuations
expressed in
n (9), (10) and (11).
( The simulation model hass two input sourrces, viz. diisplacement andd velocityy of the
sprung masss ( ) of the system, and two output signals forr control force going to thee sprung mass ( ) with (-) minnus sign
and to the un-sprung
u ( ) mass with (+) plus
p sign. Note th hat is equual to and is equal to in the equatiion (2).
Note that in
n the MR model, there are two input
i signals andd one output siggnal. The input ssignals are and displacement
and velocity
y of the sprung mass
m and the outp put signal is the control force generated by the MR damperr.

Figure 8. Bouc-Wen mo del.

In Bouc-Ween model alike Bingham, Dahl and LuGre mo odels, the controol force feeds thhe summing junnction of forces for the
sprung masss with (-) sign an
nd for the un-sprrung mass with (+)
( sign.

Also, all off the four simulaation models aree summed up ass sub-systems (F Figure 9) to commpare their perfformances againnst each
other and a passively controolled system for four different ex
xcitation signals from the terrainn. The system response is displacement
ody from the roaad excitations.
in the car bo

Semi_Active (Bouc--Wen)

Fig
gure 9. Passively controlled systeem model vs. fouur MR models aas sub-systems.

4. ation results and discussions

Simula

The above depicted

d mathemmatical formulattions as implemented in Simulinnk models are ssimulated to com mpare performannces of
each model with respect to o its exerted dam
mping force, and d vibration and shock dampingg efficiency as a semi-active viibration
controller fo
ormulated in thee system equatio ons (2) of motion
n against passiveely controlled/daamped vibrationn damper formullated in
(1) in the ex
xample of quarteer car model sho own in Figure 1. In all of our simmulations, the coontrol force in (22) is set to be equal
to and vibration damp ping is evaluated in the sprung g mass. Displaceement values off the sprung maass with a semii-active
controller off the MR dampeer models are com mpared with thee displacement vvalues of a passivvely controlled ssuspension systeem. The
values of suuspension param meters (quarter car)
c are taken fro om the data givven in Table 1 aand all numericaal values for hyssteresis
model (Bing gham, Dahl, LuGre and Bouc-W Wen) parameterss are chosen froom the data giveen in Table 2, 3, 4 and 5. The rrational
parameter values
v of the hyssteresis models are found by traails and errors. F
For numerical ssimulations threee different signaals, viz.
random white noise, Heaviside step function and sine waves with 2.1 Hz and 20.8 Hz of oscillations, and also, a combinatorial
excitation signal, a sum of sine waves and random (Gaussian white) noises, are taken. Road excitation signals are set to have
maximum (absolute) magnitude of 0.075 m and oscillation frequencies of sine waves are taken by considering natural frequencies
of the quarter car model.

Table 1. Data for suspension system (quarter car model).

Parameter name Parameter notation Parameter value
Sprung Mass 2500
Un-sprung mass 320
Stiffness of suspension 80000 [ / ]
Stiffness of un-spring mass (tire) 500000 [ / ]
Damping coefficient of sprung mass 320 [ / ]
Damping coefficient of un-sprung mass 15020 [ / ]

Table 2. Data for Bingham model simulation.

Parameter name Parameter notation Parameter value
Damping coefficient in Bingham model 320 [ / ]
Offset force 10
Frictional force 100

Table 3. Data for Dahl model simulation.

Parameter name Parameter notation Parameter value
Control voltage 5[ ]
Hysteresis parameters , , , 350, 800, 250, 25

Table 4. Data for the LuGre model simulation.

Parameter name Parameter notation Parameter value
Coulomb friction force 10 [ ]
Sticktion force 25 [ ]
Stribeck velocity 0.04 [ / ]
Stiffness coefficient 500 [ / ]
Damping coefficient 10 [ / ]
Viscous friction coefficient 0.6 [ / ]

Table 5. Data for Bouc-Wen model simulation.

Parameter name Parameter notation Parameter value
Parameters of the Hysteresis shape , , ,n 1, 0, 1.5, 2
Stiffness of the spring element 300 [ / ]
Input voltage 5[ ]
Other parameters , , , 4400, 442, 10872, 49616
Pre-yield stress 0[ ]

From the numerical simulations of hysteresis loop models with Bingham, Dahl, LuGre and Bouc-Wen models for the semi-active
suspension system it is clear that all of the semi-active system models outperform passively controlled system model for four
different excitation signals from road. Figure 10 and 11 demonstrate system responses (displacement of the car body) of the
passively and semi-actively controlled models from random (Gaussian white) noise with the magnitude of 0.075 m (in the range
of -0.0375 m +0.0375 m) and from the simulation results it is clearly seen that all hysteresis models outperform passively
damped system model in damping undersigned excitations from the terrain. Out of these four semi-active models, Bingham and
Bouc-Wen models demonstrate much higher damping than the other two models, viz. Dahl and LuGre models.
 -3
x 10 Passive vs. S emi-active hyste
eresis models

8 Input signall (road): Gussia

an noise

Car body displacement, [m]

4

-2
Passive
P
-4
B
Bingham
-6 D
Dahl
L
LuGre
-8
B
Bouc-Wen
0 2 4 6 8 10 12
time, [sec]
Figurre 10. Model ressponses on rando
om (Gaussian whhite noise) excitaation from road..

Figurre 11. Model ressponses on rando

om (Gaussian whhite noise) excitaation from road..

In another excitation
e with Step
S (Heaviside) function Figu ure 12 and 13, thhe hysteresis moodels outperform
m in damping undesired
excitation in
n the car body in
n comparison with
w passively co ontrolled model. In this case, LLuGre and Bouc--Wen models peerforms
considerablyy better than thee other two models and dissipaate the step exciitation with the magnitude of 00.075 m in less than 2
seconds. Wh hereas Binghamm model damps th he excitation in about
a 8 seconds and Dahl modell in about 13 secconds.
 Passive vs. Semi-active
S hysterresis models

0
0.14 Input signa
al (road): Heavisid
de Step function

0
0.12

Car body displacement, [m]

0.1

0
0.08

0
0.06
Passive
0
0.04 Bingham
Dahl
0
0.02 LuGre
Bouc-Wen
0
0 2 4 6 8 10 12
time, [sec]

Fiigure 12. Passivee vs. semi-activee suspension moddels on Step inpput excitation.

Fiigure 13. Passivee vs. semi-activee suspension moddels on Step inpput excitation.

In sine wave excitations with 2.1 Hz of freqquency shown in n Figure 14 andd 15, semi-activeely controlled m
models except forr Bouc-
Wen modell have demonstrrated slightly better in damping magnitude of eexcitation oscillaations in compaarison with a paassively
controlled system
s model an nd frequency of excitation from the road is presserved clearly aas a periodic signnal with all moddels. In
this case, Bouc-Wen
B modeel has outperformed all other models in term ms of damped oscillation maggnitudes. In sinee wave
excitations with 20.8 Hz off frequency show wn in Figure 16 6 and 17, all hyysteresis models have dissipatedd magnitude of excited
n car body moree than passively controlled model by preservingg periodic oscillaations with respect to road excittations.
vibrations in
Performances of all modelss after about tw wo seconds of sim mulation time hhave reached to very similar steeady state valuee in the
range of 5 of displaccement in car bo
ody. In this case, Bouc-Wen moodel has perform med slightly poorrer than the otheer three
MR damperr models in terms of damped exccitation.
 Passive vs. Semi-active
S hysteressis models

n(2*f*t); f=2.1 Hz
Input siignal (road): 0.75sin
0.0
05

0.0
04

0.0
03

Car body displacement, [m]

0.0
02

0.0
01

-0.0
01

-0.0
02 Passsive
Bingham
-0.0
03
Dahhl
-0.0
04 LuG
Gre
Bouuc-Wen
-0.0
05
0 2 4 6 8 10 12
time, [sec]

Figure
F 14. Passiv
ve vs. semi-activ n sinusoidal wavve: 0.075 sin 2
ve suspension on , 2.1 excitation.

Figure
F 15. Passiv
ve vs. semi-activ n sinusoidal wavve: 0.075 sin 2
ve suspension on , 2.1 excitation.
 Passive vs. Semi-active hysteresis models

Passive
0.01 Bingham
Dahl
LuGre
Bouc-Wen
Car body displacement, [m] 0.0
005

-0.0
005

Input signal (road):

( 0.75sin(2*f*t);
* f=20.8 Hz
-0.01

0 2 4 6 8 10 12
time, [sec]

Figure
F 16. Passiv
ve vs. semi-activ n sinusoidal wavee: 0.075 sin 2
ve suspension on , 20.8 excitation

Figure
F 17. Passiv
ve vs. semi-activ n sinusoidal wavee: 0.075 sin 2
ve suspension on , 20.8 excitation

A fourth exxcitation signal from

f road used tot simulate the models
m is sine w
wave with 20.8 H
Hz of frequencyy plus white noisse. The
performancees of the semi-aactive models for this excitationn Figure 18 annd 19 have been similar to the pprevious case wiith sine
wave excitaation with 20.8 HzH frequency fo or some extent and
a in this case, the two MR moodels, viz. Binghham and Dahl m models,
have not reaached to a stable steady-state vallue.
 Passive vs. Se
emi-active hysterresis models

015
0.0 Input sig sin(2*f*t)+Gussia
gnal (road): 0.75s an noise

0.01

Car body displacement, [m]

0.0
005

-0.0
005
Passive
P
B
Bingham
-0
0.01
D
Dahl
L
LuGre
0 2 4 6 8 10 B
Bouc-Wen
12
time, [sec]

Figure 18
8. The system responses of passiive and semi-acttive on sinusoidaal wave ( 20
0.8 ): 0.075 sin
n 2 + Gausssian
white noise excitationn.

Figure 19
9. The system responses of passiive and semi-acttive on sinusoidaal wave ( 20
0.8 n 2
): 0.075 sin + Gausssian
white noise excitationn.

5. Summ
mary

The develop ped simulation models

m of the hyysteresis or non--linear system bbehaviors of the MR liquids useed in dampers byy using
mathematical formulations ofo Bingham, Dahl, LuGre and Bouc-Wen
B modells in MATLAB//Simulink in the example of quaarter car
model havee showed adequaacy of these MR R dampers for designing
d vibratiion and shock ddampers. The sim mulation resultss of the
semi-activelly controlled dam
mper models wiith Bingham, Daahl, LuGre and B Bouc-Wen modeels have demonsstrated superiority over
passively coontrolled damperr model in the ex
xample of four different
d excitatiion signals mimiicking terrain rouughness for the quarter
car suspension system mod del. Amongst th hese semi-activeely controlled mmodels, Bouc-W Wen has outperfoormed other moodels in
terms of thee damped vibrations and steady-sstate response tim
me in three excittation signals, viiz. step, white nooise and low frequency
sine wave. In case of high her (>20.8 Hz) frequency (puree periodic) excittation from roadd, Bingham, Daahl and LuGre m models
perform better than Bouc-W Wen model.
Further stud
dies will be aimeed to develop maathematical (emp pirical) formulattions and experiimental validatioons to compute ooptimal
parameters of MR hysteresis based damp pers with Dahl, LuGre and Boouc-Wen modelss with respect tto suspension aand tire
parameters. In addition, it iss planned to deveelop an adaptivee PID controller in association w
with these MR daamper models.

dgements. This research study is supported by

Acknowled m the State Science and Techhnology
y the state grannt # A-3-54 from
Committee of Uzbekistan.

References

1. Mitu
M A.M., Popeescu I., Siretean nu T., (2012) Mathematical
m
modeling of sem mi-active controol with applicaation to
bu
uilding seismic protection,
p BSG
G Proceedings, Vol.
V 19, 2012, ppp. 88-99.
2. Saapinski B., (20099) Magneto-rheeological damperrs in vibrational control of mechhanical structurees, Mechanics V Vol. 28,
No 1. 18-25 pp.
3. Brraz Cesar M., R. Carneiro dee Barros R., 20 012, Propertiees and Numericcal Modeling oof MR damperss, 15th
nternational confference on experimental mechanics, Porto, Portuugal.
in
4. Zhhang H., et al. (2004),
( Study on the design, test
t and simulattion of a MR daamper with twoo-stage electromagnetic
co
oil, www.paperr.edu.cn viewed d on March 13, 2016.
2
5. Esshkabilov S., Grrimheden M., 2015, Car seat daamper controller design with maggneto-rheologicaal fluids, Int. C Conf.,
Nov., 2015, Navo oi, Uzbekistan.
6. Leee T.Y., Kawash hima K., Chen P.C.
P (2008) "Exp perimental and A Analytical Studyy on a Nonlinearr Isolated Bridgee under
Seemiactive Contro ol", 14th World Conference on Earthquake
E Engiineering, Octobeer 12-17, 2008, BBeijing, China.
7. Mat
M Hussin Ab Taalib, Intan Z. Maat Darus, 2013, "Self-tuning
" PIDD Controller withh MR damper A And Hydraulic Acctuator
Foor Suspension Sy ystem", Fifth Intternational Confference on Compputational Intelliigence, Modellinng and Simulatioon,
IE
EEE-Computer Society,
S DOI 10..1109/CIMSim.2 2013.27, pp. 1199-124.
8. Gugliemino, et al.. 2008, Semi-acttive Suspension Control, Springeer, pp. 192-196.
9. Laampaert V., Al-B Bender F., 20033 A generalized d Maxwell slip frfriction model apppropriate for coontrol purposes, IEEE
Physics Confereence, St. Petersburg, Russia, pp. 1170-1177.
10. Canudas de Wit C, Olsson H.J.,, Astrom K.J., Lischinsky
L P (11993), Dynamiccs friction modeels and control design.
American
A Controll Conference, Saan Francisco, US SA, pp. 1920-19226.
11. Nguyen B. D., Aldo A. F., Olivierr A. B., 2007, Efficient Simulatiion of a Dynamiic System with L LuGre Friction, JJournal
off Computational and Nonlinear Dynamics,
D Vol. 2, pp. 281-289.
12. B. Armstrong-Heelouvry, Controll of Machines wiith Friction. Bosston, MA: Kluweer, 1991.
13. Canudas de Wit C., Olsson H., AstromA K. J., and
a Lischinsky P P., 1995, A neew model for coontrol of system ms with
friction, IEEE Trrans. Autom. Co ontr., vol. 40, no.. 3, pp. 419425.

Suulaymon L. ES SHKABILOV. He got ME from m Tashkent Auttomotive Road IInstitute (TARI))

in
n 1994, earned MSc
M from Rocheester Institute of Technology in 22001 and PhD frrom Cyberneticss
reesearch Institute of Academy Scciences of Uzbeekistan in 2005. He was a visiting professor inn
20010-2011 at Mecchanical Engineeering departmennt of Ohio Univeersity, OH, USA A. He is currentlyy
ann associate proffessor at TARI and holds a ppart-time professsor position at Tashkent-Turinn
Poolytechnic Univ versity. His reseaarch areas are m
mechanical vibraations and appliccations of modall
annalysis, system identification, co
ontrol and modelling of dynamic systems.