Redu tion of Vibration and Noise
by
Kristen Lynn Clements
B.S., Me hani alEngineering (2002)
Massa husetts Institute of Te hnology
Submitted tothe Departmentof Me hani al Engineering
in partialfulllmentof the requirements for the degree of
Master of S ien e inMe hani al Engineering
at the
MASSACHUSETTS INSTITUTE OF TECHNOLOGY
June 2005
Massa husetts Instituteof Te hnology 2005. All rights reserved.
Author...
Departmentof Me hani al Engineering
May 19,2005
Certied by...
Steven R.Hall
Professor of Aeronauti s and Astronauti s
Thesis Supervisor
Read by...
Martin Culpepper
Ro kwell International AssistantProfessor of Me hani al Engineering
Thesis Reader
A epted by...
LallitAnand
Vibration and Noise
by
KristenLynn Clements
Submitted tothe Department of Me hani al Engineering
on May19,2005, inpartialfulllment ofthe
requirementsfor thedegree of
Master ofS ien e inMe hani al Engineering
Abstra t
A new method for ontrolling road noise transmitted through the suspension system of
an automobile was developed, using a Lin oln LS automobile asthetarget vehi le. In this
vehi le,roadsurfa eroughnessgeneratesvibrationsthataretransmittedintotheautomobile
primary through a single bushing (the point 4 bushing) on ea h of the front suspension
ontrol arms. An ele tromagneti a tuator was designed, built, and tested on a Lin oln
LS with simulated roadnoise. The a tuator applies a for e a ross the point 4 bushing, in
response to a elerations of the vehi le frame, just inboard of the bushing, with the goal
of redu ing the net for es transmitted into the vehi le frame, whi h ultimately produ e
unwanted interiornoise. Several tonal ontrollerswere developed,ea h designed to operate
inanarrowfrequen yband,andtoeliminate rossmember(frame)vibrationjustinsidethe
point4bushing. Thetonal ontrollerswereabletoeliminate rossmembervibration atthe
desired frequen y. Eliminating the ross member vibration resulted in modest redu tions
ininterior sound levels. A su essfulvibration ontrol system(in this vehi le) wouldneed
to eliminate ross member vibrations overthe frequen y range 100 to 200 Hz. However, a
broadband ontroller with this ele tromagneti a tuator system proved to be di ult, due
to undesirable non-minimumphase dynami s.
Thesis Supervisor: Steven R.Hall
Iwouldliketothankmyadvisor,Prof.StevenHall,forhissupportandguidan e,espe ially
his patien eand willingness to explain on epts asmanytimes as ne essary. I alsoextend
manythankstoDaveRobertsonfortheuseofhisele troni equipment,aswellashisadvi e,
en ouragement,andhumor. Di kPerdi hizzihelpedmendthevariousandsundrysupplies
ne essary for the ompletion of this proje t. DonWeiner tried to tea h me self-defenseas
wellashowtousethevariousma hinesintheGelbLaboratory. IamgratefultoProf.Martin
Culpepperfor takingthetimefromhis busys hedule toserve asadepartmental reader for
mythesis.
JoeS hmidtofFordprovidedmu hoftheautomobileexpertisene essaryforthisproje t
and pro ured Sneezy and the Lin oln LS used. Dieter Giese, also of Ford, was willing to
answeranyquestion,nomatterhowtrivial, that ameupduringthe ourseofmyresear h.
Ialso thankDr.JoeSaleh, theEx utive Dire torof theFord-MIT Allian e,andKristin
and Steve S hondorf, alsoof the Ford-MIT Allian e,for their assistan ewiththis proje t.
Dr. Kyung-yeol Song helped get me started on this proje t and was always willing to
tryto answeranyquestions Ihadinthe later stages. JorgeFeu htwanger assistedwiththe
useoftheele troni dis hargemilling(EDM)ma hine. JoshChambersprovidedadvi eand
te hni al assistan e.
Iwouldalsoliketothankmyfriends: Chrisforhisendless omputerassistan e,Christina
for stressreliefand ookies,Rohan for helpwithallthings MIT,Vi torfor omputer
hard-ware and editorial advi e, and Zoa for onverting what I wrote into what I wanted to
say.
Finally,Iwant to thankmy familyfor their ontinual support,parti ularly my mother.
This work was sponsored by the Ford-MIT Allian e. Additional funding was provided
1 Introdu tion 9
1.1 NoiseandAutomobiles . . . 9
1.2 Previous Methodsof Redu ing Interior Noise . . . 10
1.2.1 Interior Loudspeakers . . . 10
1.2.2 Power Steering . . . 11
1.3 A tiveSuspensionSystem Resear h . . . 11
1.4 Analysisof Ford Data . . . 13
1.5 Overview . . . 16
2 A tuator Design 19 2.1 ProblemDes ription . . . 19
2.2 A tuator Requirements. . . 22
2.3 Piezoele tri vs. Ele tromagneti A tuator . . . 23
2.3.1 Piezoele tri Cerami Sta kA tuator. . . 23
2.3.2 Ele tromagneti A tuator . . . 25
2.4 Ele tromagneti A tuator Spe i ations . . . 26
2.5 A tuator Amplier . . . 27
3 Experimental Setup 29 3.1 Lin oln LS. . . 29
3.2 Simulated Road NoiseEx itation . . . 33
3.3.2 Mi rophones . . . 37
3.3.3 ShakerFor e . . . 39
3.4 Dynami Signal Analyzer . . . 39
3.5 Ele troni Control Unit . . . 40
4 Experimental Results 41 4.1 Open Loop Results . . . 41
4.2 TonalController . . . 42
4.3 Expe tedPerforman e . . . 44
4.4 ClosedLoop Vibration . . . 48
4.5 Interior SoundLevels . . . 51
4.6 Summary . . . 52
5 Broadband Controller 57 5.1 Di ultieswithBroadband Control . . . 57
5.2 TonalControllersat Various Other Frequen ies . . . 60
5.2.1 TonalControl at 140 Hz . . . 60
5.2.2 TonalControl at 160 Hz . . . 61
5.2.3 TonalControl at 180 Hz . . . 63
6 Con lusion 69
Introdu tion
1.1 Noise and Automobiles
Unwanted noise is undesirable in many environments, among them the workpla e, home
and automobile. Automobile ustomers typi ally onsider a la k of interior noise to be a
desirable hara teristi when pur hasing a vehi le. Interior road noise an be generated by
numerous sour es throughout the automobile, with the engine being the major sour e of
noise [30℄. Generated engine noise is transmitted to the interior of the ar both as sound
that radiates from the engine ompartment and as vibrations thatare transmitted though
theengine mounts to the frame [14℄. Other me hani al systems inthe vehi le, su h asthe
power steering system [28 ℄, an also produ e audible noise in the interior. Additionally,
noises from outside the automobile an be heard inside the vehi le. Interior noise is also
aused byframe vibrations, whi h are theresult of tire onta t withvarious road surfa es
and potholes[33 ℄. Thisroad-indu ednoise transmitted throughthesuspension systeminto
theframe isthe sour e ofnoise thatwill beprimarily onsidered inthis proje t.
Undesirednoiseinsidethe abinofanautomobile anrangefrommerelyannoying
(mak-ingitmoredi ultforpassengerstolistentomusi ortohavea onversation)to dangerous
(preventing a driver from hearing important signals from outside the vehi le, su h as an
emergen y siren). Constant noise on extended drives (even at low levels) an also redu e
redu tion an be a omplished through three general approa hes. First, the sour e of the
noise an be eliminated or redu ed. Se ond, the paths that the vibration follows an be
modiedtoredu ethevibrationtransmission. Third,thesoundattheuserendofthepath,
inthis asethe interiorof the automobile, an be modiedto redu e theapparent noisein
theinteriorofthevehi le. However,manyte hniques employed to redu einteriornoise an
also have a detrimental ee t on other aspe ts of automobile performan e. For example,
the following parameters an be negatively impa ted by attempts to redu e interior noise
by modifyingthesuspensionsystem:
1. Handling. Handling is the pit h and roll of the vehi lebodyas a result of ornering
and brakingmaneuvers.
2. Road holding. Roadholding is the onta t for e between thetires andtheroad.
3. Suspension travel. The allowable limit of suspension travelin any vehi le design will
ae t the performan e a hievable fromthe suspensionsystem.
4. Stati dee tionresulting from variable payload. [12℄
Thegoalistomaketheinteriorquieterwithout ompromisingotherfa etsoftheautomobile
performan e.
1.2 Previous Methods of Redu ing Interior Noise
Be ause of onsumer preferen es for quieter ars, resear hers and automobile ompanies
are always looking for new ways to redu e interior noise, assuming that handling remains
omparable. Manyof the details of thenoise redu tion proje ts are onsidered proprietary
information. Presentedherearedes riptionsofsomeofthenon-proprietaryresear heorts.
1.2.1 Interior Loudspeakers
One previous method that hasbeen used to redu e interior noise ( aused by road-indu ed
noise sour e and the soundinside thevehi le an be determined, sothatthe interior noise
anbepredi ted. Interiorloudspeakers anthenbeusedtoprodu esoundoutofphasewith
the noise, redu ing interior sound levels. Using this method, Sutton and Elliott [31℄ were
able to redu e low frequen y road-indu ed noise inside an automobile by approximately
7 dB. They pla ed referen e a elerometers on the wheel, suspension, and parts of the
frame that onstitute theroadnoise transmission paths. Then, theloudspeaker signalwas
omposed of a linear ombination of the past and present referen e a elerometer signals.
Whenthispra ti alroadnoise ontroller(developedatLotusEngineering)wastested,noise
wasredu ed roughly7 dBatthemajor frequen y peaks intherangeof 100 Hz to200 Hz.
1.2.2 Power Steering
Anotherpotentialsour e ofnoiseisthepowersteeringsystem. Pressurewavesinthepower
steeringhoses auseuidnoise. Thisnoise anberedu edsomebyadjustingtheparameters
ofthesystem,su hasthelengthofthehosesandthe ongurationofthe omponents. One
studybySmid,Qatu,andDrew[28 ℄usedaMatlabsimulationsoftwareprogramtodetermine
theoptimal ongurationforthehose,tuner,andtubeinthepowersteeringsystem. (Tuners
omewithsomehosesandattenuatea ousti wavesintheuid.) They reatedamodelthat
al ulated the travel of the hydrauli pressure pulses (the sour e of vibration and noise).
Smid,Qatu,andDrewfoundthattheoptimallengthofhoseis
1/4
ofthewavelengthofthe pressure ripple,and theworst hose length is1/2
of thewavelength. Laterresear hin their labwas ondu ted to optimize theother omponentsof thepower steering system.1.3 A tive Suspension System Resear h
Many automobile resear hers have studied the benets of a tive suspension systems over
their passive suspension ounterparts. [15 , 16 ℄ Passive vehi le suspensions, found on most
road vehi les, are designed with two ompeting requirements: good vibration isolation to
ensure ride omfort, and good steering for e transmission for vehi le handling and safety.
m
m
s
k
k
s
us
us
b
s
Figure 1-1: The ommonlyused quarter ar model.
m
s
is the sprungmass (vehi le mass),m
us
is theunsprung mass(wheel mass),k
s
andb
s
arethe spring and damperbetween the sprungand unsprung mass,andk
us
isthe tire stiness.ride quality and vehi le handling. Suspension systemdesign onstraints in lude maximum
allowable relative displa ements between the vehi lebodyand unsprung mass omponents
(in luding wheels, bump stops, and protruding parts of the steering me hanism), overall
systemrobustness,reliability,weight,and ost. [16 ℄Also,Hrovat'sresear hshowsthat,fora
quarter armodel(des ribedbelow),ana tivesuspensionsystembasedonLinearQuadrati
optimal ontrol an substantially improve ride and handling performan e when ompared
withthe onventional passive suspensions. [16 ℄ However, LQ ontrollers do not ne essarily
have good stability robustness properties. For good a tive suspension performan e, and
robustness overalldesirable ride hara teristi s, the passivesuspension shouldbedesigned
withlowstiness anddamping. [32 ℄
To model vehi le dynami s, many a tive suspension system resear hers use the one or
two degree of freedom quarter ar models. [15, 16 , 32, 3 , 25, 21 , 10, 1℄ (See Figure 1-1.)
This model onsists of a sprung mass (
m
s
) representing the vehi le mass, an unsprung mass (m
us
) representing the wheel mass, a spring (k
s
) and damper (b
s
) representing the dynami s between the sprung and unsprung mass, and another spring (k
us
) representing thetire stiness. The onta tpointbetween thetireandthegroundisalsoallowedtomovein this model. Also used are the half ar model [16℄ and the full ar model [16, 34, 13 ℄.
Ahydrauli a tuatoristhe ommonlyuseda tuatorina tivesuspensionsystemresear h.
[34 ,3 ,25 ,21,1℄Thehydrauli a tuatorusedbyYamashita,Fujimori,Hayakawa,andKimura
operatesdierently at low and high frequen ies. At low frequen ies, where theservo valve
an followthe swit hingof the inputsignal, theowrate fromtheservo valveis onverted
to pressure bya hybrid gasspring and damping valve, andthis pressure isusedto a tively
attenuate the body's vibration. At high frequen ies, where servo valve following is not
possible,the ylinder pressure isgenerated only by thegas spring and thedamping valve.
This ensures the dynami s of the onventional suspension and the body's vibrations are
attenuated passively. This a tuator and its ontroller (designed using
H
∞
methods [4 ℄) wereimplementedbothinshakingexperimentsanddrivingexperiments. Inboth ases,thevibrations were redu ed at frequen ies lower than 8 Hz. Also in the driving experiments,
there was an improvement in handling with respe t to maneuvering. [34 ℄ A di ulty in
using servo valves (with the hydrauli a tuator) is that they arehigh ost items, and dirt
intolerant, soa tuator maintenan e ouldbe ompli ated. [27℄
Nonlinear ontrol has also been used with hydrauli a tuators (be ause hydrauli
a -tuators ome with asso iated nonlinear dynami s [25 ℄) to a ommodate and improve the
tradeobetween ridequalityand suspensiontravel. [25,21,10℄Theaddednonlinearitiesin
the ontroller make thesuspension stier nearits travellimits. Inexperiments byLin and
Kanellakopoulos[21℄,anonlinear ontrollerwasusedto redu ebodya elerationbyalmost
70%andbodytravelbyalmost80%as omparedtothepassivesuspension. Thesuspension
travel, however, isin reased slightly ina tive designs. [21℄
1.4 Analysis of Ford Data
Prior to the start of the proje t des ribed here, engineers at Ford Motor Company had
performeddynamometertestsonaLin oln LS.Theytookvibrationdataonadynamometer
with a oarse road surfa e. Two a elerometers were pla ed on the ontrol arm, one near
0
200
400
600
800
1000
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
x 10
−4
Frequency, f (Hz)
Acceleration spectrum,
Φ
(g
2
/Hz)
Near cross member
Near spindle
Figure1-2: A elerationspe trummeasuredbya elerometerspla edattwolo ationsonthe
ontrolarm,onenearthe rossmemberandonenearthespindle. Thethinlinerepresentsthe
a eleration spe trumnearthe rossmember andthe thi kline represents thea eleration
spe trum nearthe spindle. Note that for both a elerometer positions, virtually all of the
energy is between 100 Hz and 200 Hz, with a peak near 150 Hz. Data ourtesy of Ford
MotorCompany.
(SeeSe tion 2.2for a des riptionand photograph ofthese suspension omponents.)
Figure 1-2 shows the a eleration spe trum on a linear s ale measured by the two
a - elerometers in the dynamometer test. The thin line represents the a eleration spe trum
forthea elerometer onthe rossmember(nearthepoint4bushing)andthethi kline
rep-resents thea eleration spe trum for the a elerometer near the spindle. The a eleration
spe trum learly shows that, for both a elerometers, almost all of the energy is between
100 Hz and 200 Hz, meaning that any a tuator employed to redu e this vibration would
likely only be required tooperate nearthis frequen yrange.
Figure 1-3 is the same data as in Figure 1-2, plotted on a logarithmi s ale. Again,
the thin line represents the a eleration spe trum from the a elerometer near the ross
0
200
400
600
800
1000
10
−10
10
−9
10
−8
10
−7
10
−6
10
−5
10
−4
10
−3
Frequency, f (Hz)
Acceleration spectrum,
Φ
(g
2
/Hz)
Near cross member
Near spindle
Figure 1-3: The same data from Ford as in Figure 1-2, plotted on a logarithmi s ale.
Thethinlineshows thea elerationspe trumfromthea elerometer pla edonthe ontrol
arm near the ross member, and the thi k line shows the a eleration spe trum from the
a elerometerpla ednearthespindle. Againitis learthatthereisasigni antpeakaround
150 Hz and that there is little energy at frequen ies higher than 300 Hz. Data ourtesy of
Ford MotorCompany.
near the spindle. The sharp drop beginning at 200 Hz onrms the hypothesis that the
a tuator wouldnotberequiredtooperateat higherfrequen ies. Thea elerationspe trum
essentially goes to zero at frequen ies higher than 400 Hz, so any a tuator authority at
frequen ieshigher than 400 Hz wouldbeunne essaryand wasteful.
Figure 1-4 shows the umulative a eleration spe trum. The umulative spe trum at
frequen y
f
is the total mean-square a eleration at frequen ies off
and above. The thin linerepresentsthea eleration umulativespe trumforthea elerometeronthe ontrolarmnear the rossmember and the thi kline represents the a eleration umulative spe trum
forthea elerometeronthe ontrolarmnearthespindle. Onthis umulativespe trum,the
previously observed peakat 150 Hz is learly seen asa large drop inenergy. Thissuggests
0
50
100
150
200
250
300
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
Frequency, f (Hz)
Acceleration cumulative spectrum (g
2
)
Near cross member
Near spindle
Figure 1-4: Cumulative a eleration spe trum of the data shown in Figure 1-2. The thin
line represents thea eleration umulative spe trumfromthe a elerometer nearthe ross
member, and the thi k line represents the a eleration umulative spe trum from the
a - elerometer nearthe spindle. Theresonan e near150 Hz is lear asa sharpdrop ointhe
umulative spe trum.
hanging the energy at any other frequen ies, ould signi antly redu e the total energy,
presumably also redu ing the total sound. In other words, the frequen ies around 150 Hz
arethehighestenergy,andthereforelikelytheloudestfrequen ies,soredu ingtheenergy at
thatfrequen y wouldallowaredu tion innoiseaswell, assumingthatredu ing suspension
vibration does,infa t, redu einterior noise.
1.5 Overview
Thegoalofthisproje twastoredu einteriorroad-indu ednoisewithoutnegatively
impa t-inghandling. Basedontheassumptionthatredu ingvibrationofthe rossmember(frame)
will redu e the interior abin noise, we designed a ontroller to redu e the ross member
spe i ations ofthe ele tromagneti a tuator we hose. The ele tromagneti a tuator a ts
a rossthepoint4bushing toredu eroad noisevibration transmission. Chapter3des ribes
theexperimentalsetup: theLin olnLS,shaker,a elerometers,mi rophone, dynami signal
analyzer,andele troni ontrolunitweusedfortesting. Chapter4presentsthe
experimen-talresultsweobtainedusingatonal(narrowband) ontroller enteredat150Hz. Bothopen
and losed loop vibration dataisgiven, aswell astheexpe ted ontroller performan e and
losedloopsoundlevels. Thedi ultieswithbroadband ontrolaredis ussedinChapter5,
along withadditional tonal ontrollers, similar to theone dis ussedin Chapter4, entered
at 140, 160, and 180 Hz. Broadband ontrol proved to be di ult due to two zeros in
thetransfer fun tion from the a tuator to the ross membera elerometer, at 128 Hz and
184 Hz, givingthe a tuator no ontrol authorityat thosefrequen ies, and due to the large
non-minimum phase lag between 60 and 70 Hz. Chapter 6 dis usses the on lusions and
A tuator Design
This hapter des ribesthea tuator designpro ess. First,Se tion 2.1dis ussestheproblem
des ription, namely, redu ing interior abin noise aused by road noise vibrations. Then
the requirements on the a tuator su h as required for e and displa ement are des ribed
(Se tion 2.2). Two dierent types of a tuators, piezoele tri erami sta k a tuators and
ele tromagneti a tuators, are onsidered (Se tion 2.3). Finally, in Se tion 2.4, the
ele -tromagneti a tuator sele ted for this proje t is des ribed, and the spe i ations for that
a tuator arederived.
2.1 Problem Des ription
The spe i goal of this proje t is to redu e road noise that is transmitted through the
suspension system and frame of the automobile into the passenger ompartment, without
ompromisingvehi lehandling. Bumps,potholes,andthegeneralunevennessoftheroadall
vibratethewheelasanautomobiletravelsoverroadsurfa es. Thevibrationsarethen
trans-ferred through the ontrol arm, a rubberbushing, and into the frame,eventually rea hing
thepassenger ompartment, ausing undesirednoise.
As dis ussed previously, there are three ommon approa hes to the problem of noise
redu tion. Eliminating or redu ing the noise at its sour e is not pra ti al inthis ase; the
roads annot be made perfe tly smooth to prevent the wheel from vibrating. Although
the front left suspension with the tire removed. The A shaped ontrol arm is towards the
bottomofthe photograph. The rossmember,theupsidedown Ushapedpartoftheframe
over the bushing, is near the top. The visible bolt end is the bolt that goes through the
point4 bushing.
outside the s ope of this proje t. Modifying the transmission path of the vibration is the
approa h thatwe have employed here.
The geometry of the suspension systemis shownin Figure2-1. The rossmember is a
stru tural elementoftheframe,lo atedapproximatelyon theaxes ofthefrontwheels,that
supports theengine. The ontrol armisan A-shaped omponent of thesuspension system.
It is onne ted to thesho k absorber, the wheel at the spindle, and to the frame through
two bushings,the point3 the point 4 bushing. Thepoint3 bushingis thefrontbushing on
the ontrol arm onne ting itto theframe, and thepoint 4 bushingis therear bushing on
the ontrol arm, onne tingthe ontrol armto the rossmember.
A ording to tests ondu ted byFord MotorCompany, mostof theroad noise (andthe
steering for es) is transmitted in a single dimension, parallel to the wheel's axis, through
thepoint 4 bushing [9℄. For noise redu tion purposes, a soft point 4 bushing ispreferable.
A soft bushing would transmit less for e to the frame of the vehi le, making the interior
to the wheels,sothatthe arrespondsqui klyand a uratelyto driverinputs.
Be ause of the two opposing requirements for the point 4 bushing stiness soft for
noiseredu tionandstifor vehi lehandling,apassivebushingisne essarilya ompromise.
However, the steering for e frequen iesare on the orderof a few hertz, and theroad noise
vibrationfrequen iesareontheorderofahundredhertz. Therefore,ana tivebushingthat
a ts stiat lowfrequen ies andsoft athigher frequen ies ould be usedto reate a quieter
interiorwithout negatively ae ting vehi lehandling.
Therearetwoapproa hesto reatethisa tivebushing. One ouldbeginwithaphysi ally
softbushing,and usea tive ontrol to stienitat low frequen ies; orone ould beginwith
a stibushing,and usea tive ontrol to soften itat higher frequen ies. Thelow frequen y
for es asso iated with handling are omparable to the weight of the ar (approximately
20,000 N), and therefore the amount of ontrol authority required to stien the bushing
at those frequen ies is quite large. An example of the former approa h (stiening at low
frequen ies) is the work done at Bose Corporation. The modular design added a linear
ele tromagneti motor at ea h wheel with a modied M Pherson strut arrangement. [20 ℄
To demonstratethea tive suspension,aBosemodiedLexus LS400 waspla ed atopa four
postshakertosimulatetraveling downarough road. Fromtheoutside,thewheels ouldbe
seen to be gyrating wildly,and from theinside, there wasvirtually no sense ofmotion. [6℄
CurrentlyBoseisredu ingtheweightandsele tingamanufa turer forthea tivesuspension
system. [20℄
Onthe otherhand,usingastibushing,the for esasso iatedwithroadnoisearemu h
lower, onthe order of 150 N (see Se tion 2.2) and,as a result, mu h less ontrol authority
isne essaryto soften thebushing at higherfrequen ies. For this reason,our proje tbegins
withastibushing,andwethenusea tive ontroltosoftenthebushingatfrequen iesfrom
To begina tuator sele tion,onemustrst determinethedesignrequirements ofthesystem
during normaloperation. The design parameters of this parti ular systemin lude bushing
displa ement (required to ountera t displa ements aused byroad noise vibrations), for e
a ross the bushing (that orresponds to the bushing displa ement), a tuator bandwidth,
and alsothe ee ts ofsho kloading.
As an be seen in Figure 1-4, almost all of the energy in thespe trum is near 150 Hz,
and themeansquareda eleration is0.04g
2
. The rms a elerationisgiven by
σ
a
= 0.2
g= 1.96
m/
s2
(2.1)
Sin e almost all the vibration o urs at a frequen y of 150 Hz (
ω = 942
rad/
s), the rms displa ement is given byσ
d
≈ σ
a
/ω
2
= 2.5 µ
m (2.2)During normal ex itation of the wheel, the rms point 4 bushing displa ement is
approxi-mately
1 σ
d
.
2.5 µ
m under oarse road onditions. Allowing for3 σ
motion, thea tuator shouldbe apableof atleast7.5µ
mofa tuation. Thefor ethatthea tuatormustbeable to provide is al ulated byHooke'sLaw,f = kd
where
k
is the stiness of the point 4 bushing (about 20 kN/mm), andd
is the required displa ement (7.5µ
m). Therefore,itisne essaryforthea tuatortogenerate atleast150N of for e.As seen in Figure 1-2, the a tuator will be required to ountera t disturban es of
fre-quen ies between 100 Hz and 200 Hz. Therefore, the large signal bandwidth must be at
least200Hz,meaningthatthea tuator mustbe apableofgeneratingfor esontheorderof
150 N,up to 200 Hz infrequen y. However, to ensure thatthephaselag inthea tuator is
than 2000 Hz.
Thea tuator mustalsobeabletowithstand sho kloads,su hasthatofawheelhitting
a urb or a pothole, or those produ ed during sharp turns. The entire vehi le weighs
approximately 2000 kilograms;undernormal onditions ea hwheelsupports 500 kilograms
(5000N).Duringanextremeturnorhittingapotholewithmostofthelateralfor esrea ted
by theouterwheels, the for es might beashigh as10,000N. Therefore,thea tuator must
beableto withstandsteeringloadsofapproximately10,000 Ntothebushing. Equivalently,
thea tuatormustbeabletowithstandabushingmotionof500
µ
m(0.5mm)duetosteering loads, again al ulated using the bushing stiness. The a tuator would not be required toountera t these loads,but itmust be ableto bearthem and ontinue to operate afterthe
loading isremoved.
Be ause the a tuator must be able to withstand displa ements mu h larger thenwould
normally be produ ed by thea tuator, thea tuator should be ompliant, when ompared
to thestinessof thebushing. Ana tuator that is ompliant isee tively a for ea tuator
rather than a displa ement a tuator. The ideal a tuator for this proje t should ommand
for erather than displa ement.
2.3 Piezoele tri vs. Ele tromagneti A tuator
Afterdeterminingtherequirementsonthea tuator,twotypesofa tuatorswere onsidered.
One is an a tuator made from piezoele tri erami sta ks, one is an ele tromagnet made
fromsteel laminations.
2.3.1 Piezoele tri Cerami Sta k A tuator
Onepossibletypeofa tuatorisapiezoele tri erami sta ka tuator. Thistypeofa tuator
isanattra tive optionbe ausethe piezoele tri material hasahighenergy density,
approx-imately
80
kJ/
m3
,and a highbandwidth. However, itmaybedi ult to getthene essary
stroke,sin epiezoele tri materialsprodu everysmalldispla ements,andverylargefor es.
Instrumente (Auburn, Massa husetts), is
5
mm× 5
mm× 9
mm , and itis apable of 800 N for e and 6.5µ
m displa ement. A sta k this size is on the order of $150. [24 ℄ One of these piezoele tri sta ks has more for e than is ne essary for our needs, but not nearlyenough displa ement. Therefore, multiple sta ks and an ampli ation me hanism an be
used to meet the a tuator requirements. As a rough estimate (order of magnitude) of the
numberof sta ks required, ompare the piezoele tri sta kspe i ations withthea tuator
requirements. The for e of a single sta k is
800
N/150
N= 5.3
times larger than we need. The displa ement is500 µ
m/6.5 µ
m= 77
times smaller than we need. Thus, even with an ampli ation me hanism to mat h the impedan e of the material to the requirements, wend thatwe need at least15piezoele tri sta ksto a hieve enough for eandstroke.
The piezoele tri sta ks are too sti for this appli ation. Therefore, an ampli ation
fa tor must be hosen to mat h the stiness requirement. However, the displa ements
and for es annot be modied separately (as assumed in the above estimation). A single
ampli ation me hanism must be used to both amplify the displa ement and redu e the
for e. Tomakethispiezoele tri materialappropriatehere,theremustbesomeampli ation
me hanism. Consider a me hanism thatamplies the a tuator displa ement bya fa tor of
A, so that the ratio of original to modied displa ement is
1
A
. Then the ratio of original
to modied for e would be A,so thatthe ratio of original to modied stinessis
1/
A2
. If
the original a tuator for e is 800 N, but thea tuator is only required to provide 150 N of
for e,thenA
= 5.3
for anidealampli ationme hanism. Then15sta kswouldberequired to a hieve the ne essarystroke,and even more piezoele tri sta ksfor a real(less e ient)ampli ation me hanism.
Be ausethedrivingrequirementontheamount ofne essarypiezoele tri materialisthe
steering sho kloads, not thehigher frequen y a tuation requirement, more than 15 sta ks
would be required for a piezoele tri a tuator a ting a rossthe point 4 bushing. At a ost
of around $150 ea h, more than 15 sta ks is obviously far too many sta ks to make this
piezoele tri erami sta ka tuator feasible inthisappli ation.
piezoele -a proof mass may be a more pra ti al type of piezoele tri a tuator. At the operational
frequen yrange, between100 Hzand200 Hz,thepiezoele tri material wouldpushagainst
theproofmass(whi hideally wouldnotmove)todispla e thebushing. Atthelow
frequen- iesofthesteeringsho kloads,thistypeofa tuatorwouldpushagainstandmovetheproof
mass, making it ompliant, and redu ing the required number of piezoele tri sta ks. To
minimize the amount of piezoele tri material,the proofmassshould have about the same
dynami stinessasthestinessofthepoint4bushing(20kN/mm)at100Hz. Thedynami
stiness of the proof massis M
ω
2
, where M is themass, and
ω
is thepertinent frequen y (here100 Hz). Settingthedynami stiness equalto 20kN/mm resultsinarequired proofmass of approximately 50 kg. Unfortunately, a 50 kg proof mass added to the suspension
systemistoolargetobeafeasiblea tuationsolution. Asmallerproofmass anbeused,but
thatwouldredu ethe e ien y andrequiremorepiezoele tri material. A5 kgproofmass
would be more appropriate in size, but wouldagain require approximately 25 piezoele tri
sta ks,whi h would be too ostly for feasibility. Although a tive material a tuators, su h
aspiezoele tri ones, areattra tive insome respe ts,they areprobablyimpra ti al for this
appli ation.
2.3.2 Ele tromagneti A tuator
Another possible type of a tuator is an ele tromagneti a tuator; with a xed element
(magneti ore), oil(ex itation winding),and amovableelement (armature). [19℄
Ele tro-magnetsareused inmanyappli ations, for example:
Solenoida tuators: aps,pneumati andhydrauli valves,slidevalves,interlo ks
breaks.
Hammering a tuators: riveting,pun hing, stamping, hiseling ma hines.
Turningmagnets: throttlevalves, ontrolvalves(hydrauli ,pneumati ),material
support(forinstan e web of loth,paper), turnout intransportplants.
ing plants, os illating sieves, os illating tables, swinging and heli al
onvey-ors. [19℄
Ele tromagneti a tuators anbevery ompliant;thegapbetween therotorand thestator
allows for mu h larger sho k motion than the ontrol motion provided by the a tuator.
Another benet is that ele tromagneti te hnology is mature, ontributing to a redu ed
ost when ompared to the ost of a tive materials. After determining that an a tuator
made of piezoele tri erami sta ks is not feasible, we hose to pursue development of an
ele tromagneti a tuator for thereasonsoutlinedabove.
2.4 Ele tromagneti A tuator Spe i ations
The ore of an ele tromagnet is made of steel laminations. Steel is used be ause of its
ele tromagneti properties, butasolidsteel oreisnotase ientasalaminated ore. The
laminations redu e eddy urrentsthat resultfrom theindu ed magneti eld. Be ausethe
bandwidth of the a tuator is proportional to the inverse of the timeit takes for the eddy
urrentsto settle,redu ing theeddy urrentsin reasesthebandwidthoftheele tromagnet.
Theele tromagnetusedherehas54laminationsthatareea h0.0185in hesthi k(0.47mm),
for a total thi kness of 2.5 m. (Figure 2-2) The laminations were ut to shape by an
ele troni dis hargemilling (EDM)ma hine.
The ele tromagnet onsistsof astator, arotor, anda 2mmair gap separatingthetwo.
The stator is the approximately ir ular side of the ele tromagnet, and is on entri with
thepoint 4 bushing. Itis bolted to the rossmemberand hasa radius of 5 m. The rotor
is roughly U-shaped, and is bolted to the ontrol arm. The ends of the rotor arms are
rounded so that they onform to the urvature of the ir ular side of the ele tromagnet,
whilemaintainingthe2mmairgap. Theseroundededgesallowthemagnet torotateabout
the point 4 bushing bolt with the ontrol arm without altering the 2 mm air gap. The
magneti rotor isapproximately 15 mlong,8 mhigh, and2.5 mthi k. Figure2-3shows
laminated core with 2 poles
attached to control arm
laminated disk attached to cross member
2 mm air gap
copper windings
Figure2-2: Diagramof theproposedele tromagneti a tuator.
The for e produ ed by anele tromagnet isproportional to thesquare of the urrent in
the oil. Therefore, to get an approximately sinusoidal for e, a large bias urrent plus a
sinusoidal urrent is required. For this ele tromagnet, therotor iswoundwith120 turns of
opperwire. The maximumrequired urrent is10 A plusthebias urrent.
2.5 A tuator Amplier
To supplythe large urrentsthatarerequired fortheele tromagneti a tuator operation, a
substantialpowersupplyandamplierwereemployed. Theamplier usedwiththe
ele tro-magneti a tuatorisabrushtypepulsewidthmodulatedservoampliermadebyAdvan ed
Motion Controls (Montville, New Jersey), model 100A40. It requires a DC power supply
between 80 V and 400 V. The amplier is apable of
±
100 A peak urrent and±
50 A ontinuous urrent. [2℄ A tuator urrent an be measured using theamplier's monitoringport. A Hewlett Pa kard (Palo Alto, California) power supply, model 6479C, was used to
providethed powerforthea tuatoramplier. Thepowersupply anprovidefrom0-300V
and 0-35A ofpower.
To implement this ele tromagneti a tuator in a real ar, we learly annot use su h a
rossmember.
42V,whi h is ompatible withthe42 Vsystemslikelyto be usedinthenearfuture. [5℄In
addition, the total urrent requirements an be redu ed substantially by using a apa itor
Experimental Setup
This hapter des ribes the experimental testbed (the Lin oln LS) used in this proje t. In
parti ular, we des ribe the shaker used to indu e vibrations in the testbed; the load ell
used to measure the for es generated by the shaker; the a elerometers used to measure
thetestbedvibrations; themi rophone usedtomeasuretheinterior abinsound; thesignal
analyzer used as a virtual fun tion generator, data re order, and data analyzer; and the
ele troni ontrol unit used as D/A and A/D onverters, as well as an interfa e to the
software ontroller.
3.1 Lin oln LS
Experimentspreviously ondu tedbyresear hersatFordMotorCompanyontheLin olnLS
led to the on lusion that most of the road noise vibrations are transmitted through the
point 4 bushing, a nding spe i to this model. [26 ℄ As des ribed earlier, the point 4
bushing is the bushing that onne ts the ontrol arm to the vehi le frame. Other Ford
vehi lesdidnotexhibitthesamemajorvibrationpathway; anyte hniquefornoiseredu tion
a omplished bythis study maynot apply to otherFord vehi les.
It is ne essary to design a testbed setup in order to study the road noise transmitted
throughthe suspensionsystemofa Lin oln LS,withthegoal ofredu ing thisnoise. Inour
setup, road noise was simulated using a shaker to apply for es to the suspension. Sensors
interior to measure the sound. The testbed initially in luded the front suspension system
omponents, part of the vehi le frame, and the front wheels and tires for support. The
initial testbedwasthefrontend ofa Lin oln LSprovided byFord,ni knamed Sneezy,with
all ne essary omponents forward of the steering olumn. (Figure 3-1) In luded were the
major parts of the frame and the entire suspension system. The body panels, engine, and
other omponentsofthe engine ompartment werenotin luded. Dueto themissingweight
of these omponents, theride height wastoo high. We built a wooden box (lo ated inthe
engine ompartment)toholdapproximately700 kgofleadbri kstoa hievethe orre tride
height. Be ause the rear end of the arwasnot present, a metalframe was welded to the
rearof theframe inorderto properlypositionthe testbed.
Sneezy was found to be an inadequate testbed; results obtained using Sneezy mat hed
neither the expe ted results nor the results from previous tests ondu ted by Ford. To
onsider Sneezy an appropriate testbed, the transfer fun tions obtained from Sneezy and
those from Ford's vehi le modeling program should mat h, but they did not. The
appar-ent resonan e at 150 Hz from Ford's datawas not present inexperiments on Sneezy. (See
Figure 3-2.) The resonan es of Sneezy's transfer fun tions are at higher frequen ies.
Ad-ditionally, the result from Ford that most of the vibrations are transmitted through the
0
150
300
400
500
600
700
800
900
1000
−80
−60
−40
−20
Magnitude (dB)
0
150
300
400
500
600
700
800
900
1000
−50
0
50
100
150
Frequency (Hz)
Phase (degrees)
Figure3-2: Thetransferfun tionfromtheshakerex itationtothe rossmember
a elerom-eter (data a quired with Sneezy). No major peak is present near 150 Hz ( ompare with
Figure1-2).
After itwasdetermined thatthe front endof theLin oln LS didnot produ e the same
resultsasanentire ar,FordMotorCompanyprovideda omplete,drivable2001Lin olnLS.
Usingtheentirevehi le,thevibrationtransmissionpathsandnoise ouldbestudiedinmore
detail. Theneventually,we andesign a ontrollerto allowforaquieter drive without
om-promisingvehi lehandling. AnadditionaladvantageofusingtheentireLin olnLSisthatit
would permitmeasurement ofthe interior soundlevels,and allowtestingof thehypothesis
thatredu ing the rossmembervibration also redu estheinteriorsound. Figure3-3shows
thenewtestbed (theentire Lin oln LS).
The experimental setupusing theLin oln LSis furtheroutlined inFigures 3-4and 3-5.
During the shaker or a tuator transfer fun tion identi ation, thedynami signal analyzer
sendsthedrivingsignaltoeithertheshakerorthea tuator. (Figure3-4)Thesolidline
on-ne tingthesignalanalyzerto theshakerboxrepresentsthe onne tionthatispresentwhen
MIT.
thea tuatorrepresentsthe onne tionthatispresentwhenidentifyingthea tuatortransfer
fun tions. Theshakeris onne tedinserieswithaload ellandastingtothesuspension,so
thatwhen theshakerisbeingdriven,theapplied for e an bemeasured. Either theshaker
signalor the a tuator signal an be usedto indu evibration intheLin oln LS automobile.
Signals from the sensors on the automobile (the a elerometers and the mi rophones) are
sent ba kto thedynami signalanalyzer andtheele troni ontrol unit'spro essors,where
thesignals arere orded andanalyzed.
In the losed loop system (i.e., when the ontroller is present and operational), the
system fun tions in a slightly dierent manner. (See Figure 3-5.) The dynami signal
analyzerstill drivestheshakerthrough aload ellto vibratethe ar,and thea elerometer
and mi rophone signals are still sent ba k to thesignal analyzers and hardware ele troni
ontrolunitforanalysis. Themajordieren einthe losedloopsystemisthatthe ontroller
accels
mics
shaker
signal
analyser
actuator
Lincoln LS
electronic
control unit
load cell
Figure 3-4: The blo k diagram outlines the experimental setup during either shaker or
a tuator transferfun tion identi ation. The signalanalyzerdriveseithertheshaker(solid
line)orthea tuator(dashedline). Eithertheshakerorthea tuatorvibratestheLin olnLS.
Thesensors on the automobile,both thea elerometers andthemi rophones, measurethe
vibration and returnthe signals to thesignalanalyzer andele troni ontrol unit.
3.2 Simulated Road Noise Ex itation
For thepurposes of these experiments, it was not feasible to drive the Lin oln LS on road
surfa es to measure and re ord vibrations of the wheel and other suspension and frame
omponents, and theresulting soundinthe interior oftheautomobile. Therefore,theroad
noise vibrations had to be simulated in the laboratory. To a omplish this simulation, a
shakerwasatta hedbya stingto thepassenger side ontrol arm. (SeeFigure3-6.)
The shaker used was a model 420 shaker, made by Ling Ele troni s, In . (Anaheim,
California). The shakerisan ele tromagnet; urrent is suppliedto theshaker(whi h hasa
permanent magnet armature),indu ing amagneti eld and a magneti for e. The shaker
is apable of providing up to 133 N of for e in a frequen y range of 0 to 7500 Hz and a
electronic
control unit
Lincoln LS
accels
mics
shaker
actuator
signal
analyser
load cell
Figure 3-5: The losed loop blo k diagram represents the experimental setup during
on-troller operation. The signal analyzer drives theshaker through a load ell to vibrate the
ar,andthea elerometersandmi rophones measurethevibrationsandsendthemba kto
the signalanalyzer and ele troni ontrol unit. In this setup, the ontroller isdownloaded
to the ele troni ontrol unitexpansion boxand drivesthea tuator.
rms. [22 ℄ Theamplier usedwiththeshakerisaYorkville Audiopro 3400,ahighe ien y
stereopoweramplier, madebyYorkvilleSound, Toronto, Canada.
Vibration due to road roughness is simulated by the shaker, whi h is atta hed to the
ontrol armbya sting. A load ell waspla ed inseries withthesting to measurethefor e
applied. Thestingisaslenderaluminumrodwithoneend s rewedinto theload ell,whi h
isinturn onne tedto theshakerfa e,andtheotherends rewedintothe ontrolarm. The
stingisatta hedtothe ontrol armatthepointdesignatedbythearrowinFigure3-7. The
shakerpushesforeand aft on thesting, whi h pushesforeand aft onthe ontrol arm.
Wealsosometimesusedanalternativemethodformountingtheshaker,byatta hingthe
shakerthroughthe sting toawheellugbolt. Thesting wass rewed intoalugnutadaptor,
from this method of atta hment were found not to be as representative a load ase asthe
foreandaft ex itationdes ribedabove. Also,a ordingtoFord, theforeandaft ex itation
resultsinvibrationsinthevehi lethataremorerepresentativeofreal-worldvibrations. [26 ℄
3.3 Instrumentation
Inordertomeasureandre ordthevibration,vibrationtransmissionthroughtheframe,and
interior abinsound during experimentation, two types ofsensors were usedto instrument
the Lin oln LS automobile. A elerometers were pla ed in two lo ations to measure the
vibrations, one on the frame, and the other on the ontrol arm. Also, a mi rophone was
pla ed in four dierent lo ations in the interior of the vehi le to measure the sound level
throughout the automobile.
3.3.1 A elerometers
To measure and re ord the vibration of the relevant suspension omponents, two Endev o
(San Juan Capistrano, California) piezoele tri a elerometers were employed. One was
Thephotograph is taken fromthefront, looking aftat theright frontwheel.
Table 3.1: Additional Endev o piezoele tri a elerometer spe i ations
Serial Number EL89 EL92
Charge Sensitivity 96.8pC/g 97.1pC/g
Capa itan e 2743 pF 2698 pF
Max. Transverse Sensitivity 0.4% 2.0%
The pla ement of the a elerometers is shown in Figure 3-8. The small metal ylinders
atta hed to ables arethe a elerometers. The a elerometer above thepoint 4bushing in
the pi ture is atta hed to the ross member, and will be referred to as the ross member
a elerometer. The a elerometer below and to the left in the pi ture is atta hed to the
ontrol arm, andwill bereferredto asthe ontrol arma elerometer.
The a elerometers used were Endev o model 7701-100. The frequen y range of the
a elerometersis20Hzto 5kHz. Theserialnumbers forthe rossmemberand ontrol arm
a elerometersareEL89andEL92,respe tively. Additionalspe i ationsforthe
a elerom-eters are shownin Table 3.1. [7 ℄ Endev o laboratory harge ampliers, model 2721B, were
membera elerometer isthe a elerometer lo ated above thepoint 4bushing. The ontrol
arma elerometer islo atedbelow andto theleft ofthepoint 4bushing.
harge sensitivity of ea h a elerometer, as well as thedesired gain (V/g). The frequen y
range of the harge ampliers is3 Hz and 10 kHz. Theoutput has a maximumvoltage of
10.0V peak,and a maximum urrent of 2.0mA. [8℄
3.3.2 Mi rophones
The mi rophone on a sound level meter was used to measure the sound at ea h of four
positionsintheinterioroftheLin olnLS.The1982Pre isionsoundlevelmeterandanalyzer
wasmadebyGenRad(nowIETLabs, In .,Westbury,NewYork). Therangeof thissound
level meter is 30 dB to 130 dB rms (150 dB peak). [17℄ The sound level distribution
throughout the interior of the automobile is quite modal. The sound waves from all of
the numerous sour es of noise in the vehi le intera t, ausing a large number of nodes
and antinodes. At the nodes, the sound waves are out of phase and an el ea h other's
magnitude; and at theantinodes the pressure wavesare in phase and themagnitudes add
( )
Figure 3-9: The rst mi rophone position is lo ated at the driver's left ear. The se ond
mi rophone position is at the ree tion of this position at the front passenger's right ear
(a). Thethird mi rophone positionis lo ated attheright (passenger side) rearpassenger's
right ear(b). Thefourthmi rophone positionislo atedinthe enterofthedriver'sseat at
approximately hinlevel( ).
where the driver or any passengers would be able to hear the transmitted road noise are
signi ant for thisproje t, andthemi rophonepositionswere hosento measurethesound
inthese areas.
Therstmi rophone positionisat thedriver's leftear. (SeeFigure3-9(a).) These ond
mi rophonepositionisatthefront passenger'srightear,aree tionoftherstmi rophone
position, as shown in Figure 3-9(a). The third mi rophone position is at the right rear
passenger's right ear. (See Figure 3-9(b).) The fourthand nalmi rophone position
(Fig-ure3-9(b))isatapproximatelythe enteroftheseat,atthedriver's hinlevel. Therstand
abetter ideaof thesounddistribution throughout theinteriorof theautomobile.
3.3.3 Shaker For e
Aload ellwaspla edbetweenthestingandtheshakerfa etomeasureandre ordthefor es
appliedbytheshakertotheautomobilethroughthesting. Thefor emeasurementprovides
more detailed information about the load ase than what an be dedu ed from the shaker
input signal voltage. The load ell used is model SM-50 made by Interfa e (S ottsdale,
Arizona). It an measure for es up to 200 N. At 200 N of for e the dee tion is 0.08 mm
and the natural frequen yis 1550 Hz. [18 ℄
3.4 Dynami Signal Analyzer
Thedynami signalanalyzerservesmanyfun tions. Ita tsasavirtual fun tion generator,
datare order, and a dataanalyzer. The dynami signal analyzer usedhere onsistsof two
Siglabmodel20-42signalanalyzersandtheasso iatedsoftware,madebySpe tralDynami s
(SanJose,California). Ea hsignalanalyzerunithasa20kHzbandwidth, 4input hannels,
and 2output hannels, fora total of 8input hannels and 4output hannels. [29℄
The dynami signal analyzer was used asa fun tion generator to provide the required
sinusoidal signals to drive the shaker or the ele tromagneti a tuator, depending on the
experiment. Thesoftwarevirtual fun tiongeneratorsuppliedthene essarysignalssentout
through a D/A onverter and the output hannels. The two output hannels an
indepen-dently supply two dierent ex itationsignals for useduring the testing. Ea h hannel an
output20 mAand 10V maximumwithan available optionalDCoset. [29 ℄
The signal analyzersalso have analog to digital onverters, whi h were used to a quire
and re ord the various a elerometer, mi rophone, and load ell signals. The two signal
analyzer units were time syn hronized, allowing simultaneous re ording of eight dierent
signals. The voltage range for ea h input hannel is adjustable from
±
20 mV to±
10 V. [29 ℄the sensor signals, simple transfer fun tions, and the oheren e of the estimated transfer
fun tion. Theavailableanalysisfun tions(otherthan oheren e)in ludetimehistory,
auto-spe trum,transferfun tion, ross-spe trum,auto- orrelation, ross- orrelation,andimpulse
response. [29 ℄ Che king the oheren e wasespe iallyimportant to ensure that the signals
were of a highenough signal-to-noise ratio to trust the results. The re orded signals were
also exportedto Matlab, to allowfor more extensive analysis.
3.5 Ele troni Control Unit
The signal analyzers have analog to digital and digital to analog onverters that an be
used to transfer the shaker and a tuator driving signals, as well as to a quire the sensor
signals. Anotherpie eofhardware,theele troni ontrolunit,madebydSPACEIn (Novi,
Mi higan), was used for the ontroller signals. The ele troni ontrol unit onsists of an
expansionbox(modelPX10)andtwoI/Ods1003pro essorboardsmountedintheexpansion
box. The ontrolfeedba ksignals(a elerometerandmi rophonesignals)werebroughtinto
the omputer using this ele troni ontrol unit. This ele troni ontrol unit has, among
other things, hardware A/D and D/A onverters. It was also used as the interfa e for
the software ontroller. The ontroller was assembled in the Matlab modeling appli ation
Simulink, and then downloaded and run from this ele troni ontrol unit. An asso iated
program alledControl Deskworkswiththeele troni ontrol unit, andallowssome ofthe
ontrol parameters to be hanged whilethe ontroller is operating. Thisfeature allowed us
Experimental Results
This hapter presents and explains the experimental data taken on the Lin oln LS. First
is the baseline data, the open loop transfer fun tions from the shaker and a tuator to
the ross member and ontrol arm a elerometers, in Se tion 4.1. Next (Se tion 4.2), the
designof tonal ontrollersisdis ussed. Thisisfollowedbytheexpe tedperforman e. Both
the predi ted performan e from the Bode plot of the ontroller transfer fun tion and the
performan e omparison of the al ulated theoreti al transfer fun tion to the measured
experimental transferfun tion (showing thatthe experimental results mat h thepredi ted
theory) is des ribed. Finally, the losed loop data is presented, in luding the losed loop
vibration (Se tion4.4) andthe interior soundlevels(Se tion 4.5).
4.1 Open Loop Results
The initial baseline results aretheopen loop transferfun tions. The two ex itation inputs
are theshaker (des ribed in Se tion 3.2) and the a tuator (des ribed in Se tion 2.4). The
two important open loop outputs arethe rossmembera elerometer and the ontrol arm
a elerometer (des ribedinSe tion3.3.1). Figure4-1showstheopenlooptransferfun tion
fromtheshakerinputtothe rossmember(thinline)and ontrolarm(thi kline)
a elerom-eter outputs. Clearly,both the rossmember a elerometer and ontrol arma elerometer
transferfun tionsarequitemodal. Also,the150Hzpeakthatwaspresent inthedatafrom
10
1
10
2
10
3
10
−2
10
0
10
2
Magnitude
10
1
10
2
10
3
−4000
−3000
−2000
−1000
0
Phase (degrees)
Frequency (Hz)
cross member
control arm
Figure 4-1: Open loop transfer fun tions from the shaker to the ross member (thin line)
and ontrol arm (thi kline) a elerometers. Both transfer fun tionsare quitemodal. The
150 Hzpeakthatwaspresent inthedatafromFord(Figure 1-2)isalsodi ultto observe.
fun tion. Figure 4-2 shows the open loop transfer fun tions from the a tuator ex itation
to both the ross member (thin line) and ontrol arm (thi k line) a elerometers. Again,
both of the open loop transfer fun tions are quite modal. Control ould be di ult due
to themany modespresent inthe rossmembera elerometer transferfun tion,aswell as
thelarge phase roll-o present inthe rossmembertransfer fun tion. Thedi ulties with
broadband ontrol will be des ribed further inSe tion5.1.
4.2 Tonal Controller
After examining the open loop transfer fun tion from the a tuator to the ross member
a elerometer, the form and details of the ontroller must be determined. Be ause the
transfer fun tionsfrom the shakerto thea elerometers and also from the a tuator to the
a elerometersarehighlymodal(Figure4-2), thedesign ofabroadband ontrollerthathas
10
1
10
2
10
3
10
−2
10
0
10
2
Magnitude
10
1
10
2
10
3
−500
0
500
Phase (degrees)
Frequency (Hz)
cross member
control arm
Figure4-2: Open looptransferfun tionsfrom thea tuator tothe rossmember(thinline)
and ontrol arm (thi kline)a elerometers. Both transferfun tionsarevery modal, whi h
ould make ontrol di ult. Also the rossmember a elerometer transfer fun tion has a
signi ant phase roll-o that an auseproblems with ontrol.
broaden the ee tive frequen y range on ethe simpler tonal ontroller is fun tional. This
tonal ontroller will demonstrate whether ontrolis ee tive,i.e., ifredu ing rossmember
vibration redu es interiornoise.
The tonal ontroller will be a feedba k ontroller. Here, the logi al variable to be fed
ba k isthe ross membera eleration. In theunmodiedbushing, with no a tuator, road
noisevibrationistransmitteda rossthepoint4bushing. Thea tuatoraddedtothesystem
will be used to supply a for e that an els the road noise vibration the ross member
a eleration will be fed ba k into the ontroller and driven to zero. If the vibration of the
rossmember anberedu ed,thenthesoundsintheinterior ausedbythatvibrationshould
also be redu ed.
The tonal ontroller employed hereis designed toeliminate the rossmember vibration
at a singlefrequen y. The initial frequen y onsidered was 150 Hz, hosen be ause itfalls
ontrol (HHC) originally designed to redu e heli opter vibrations, developed by Hall and
Wereley [11 ℄. Their ontinuous timeHHC ompensator is
K(s) =
2k(as + bΩ)
s
2
+ (Ω)
2
(4.1) wherea =
Re[G(jΩ)]
|G(jΩ)|
2
(4.2)b =
Im[G(jΩ)]
|G(jΩ)|
2
(4.3)k =
1
T
(4.4)and
T
isthedesiredthesettlingtimeofthe losedloopsystem. Inthis ase,Ω
,thefrequen y (in rad/s) of the harmoni to be redu ed, isΩ = 2π × 150
Hz , andG(jΩ)
is the transfer fun tion fromthea tuator to the rossmembera elerometer evaluatedat 150 Hz.This ontroller (Equation 4.1) eliminates vibrations at the frequen y
Ω
be auseK
is innite there. The onstants,a = −0.099
andb = 0.2732
, determined from the transfer fun tionG(jΩ)
, generally result in good phase margins at the rossover frequen ies, just aboveandbelow150Hz. However,itissometimesne essarytoadjusta
andb
togivebetter phasemargins at one ofthe two rossoverfrequen ies(above or belowΩ
).4.3 Expe ted Performan e
Classi al ontrol theory di tates that for the most basi feedba k ontrol system, su h as
theone showninFigure4-3, the losed loop transferfun tion isgiven by
H(s) =
K(s)G(s)
1 + K(s)G(s)
(4.5)where
H(s)
representsthe losedlooptransferfun tionfromr
toy
,G(s)
istheplanttransfer fun tion, andK(s)
represents the ontroller transfer fun tion. In Figure 4-3, theinput orG(s)
u
e
r
+
-K(s)
y
Figure4-3: Classi al ontroltheoryblo kdiagram.
G(s)
istheplant,K(s)
isthe ontroller,r
is the input(also alled the referen esignal),e
is theerror signal,u
is the ontrol signal, andy
isthe outputsignal.G(s)
u
e
+
-K(s)
y
+
+
0
d
Figure4-4: Disturban ereje tionblo kdiagram. Thesignalsarethesamesignalsdes ribed
inFigure4-3.
referen e signal,
r
,is the ommand signal. The error signal,e
, thedieren e between the referen e signal and the output, is also the ontroller input. The ontrol signal,u
, is the outputfrom the ontroller. Theoutput signal,y
,isthemeasurement of interest.In disturban e reje tion problems (Figure 4-4), su h as this one, the input referen e
signal is set to zero, sin e the desired output,
y
, is zero. The transfer fun tion fromd
toy
, whi h measures the attenuation of the disturban e by the ontroller, is alled the sensitivity transfer fun tion,and isgiven byS(s) =
1
1 + G(s)K(s)
(4.6)TheLin oln LS systemisonly slightlymore ompli atedthanthestandard disturban e
reje tion systemdes ribed above, be ause the disturban e signal,
w
, does not add dire tly to theplantG(s)
output. The disturban e signal is modied by the shakertransfer fun -tion, makingd = wG
w
(s)
the signaldire tly added to the plant output. A blo k diagram representing the signi ant parts of theLin oln LS systemis shown in Figure 4-5.G
w
(s)
u
+
-K(s)
y
+
+
0
G (s)
G (s)
a
w
w
d
Figure4-5: Blo kdiagramrepresentingthepertinentpartsoftheLin olnLSsystem.
G
w
(s)
istheshakertransferfun tion,G
a
(s)
isthea tuatortransferfun tion,K(s)
isthe ontroller transfer fun tion,w
is the signal into the shaker,u
is the ontrol signal, andy
is the measurement signal.representsthe shakertransfer fun tion,
G
a
(s)
isthe a tuator transferfun tion,K(s)
isthe ontroller transferfun tion,w
isthe signalinto theshaker,u
is the ontrol signal,andy
is themeasured rossmembera eleration. Thereferen esignalinthis ase,asinthedistur-ban ereje tion ase, iszerobe ausethe ontroller issupposedtoreje ttheshakervibration
disturban e. The fra tionalattenuation dueto ontrol inthis s enariois
T
yd
=
1
1 + K(s)G
a
(s)
(4.7)
and the losed looptransfer fun tionfrom the shakerto thea elerometer is
T
yw
=
G
w
(s)
1 + K(s)G
a
(s)
(4.8)
Some ofthe hara teristi s oftheexpe tedperforman e an be seenintheBode plotof
the a tuator to ross member a elerometer transferfun tion times the ontroller transfer
fun tion,
G
a
(s)K(s)
. (See Figure 4-6.) Clearly, the Bode plot shows that there is a large gainat150Hz,andlowergainsatfrequen iesfartherawayfrom150Hz. Also,theBodeplotshowsthattherearetwo rossoverfrequen ies,147 Hzand153 Hz. First, onsiderthelower
rossoverfrequen y,at147Hz. Atthisfrequen y,thephaseis-67degrees. This rossoveris
10
1
10
2
10
3
10
−5
10
0
Magnitude
10
1
10
2
10
3
0
500
1000
Phase (degrees)
Frequency (Hz)
Figure 4-6: Bode plot of
G
a
(s)K(s)
, the a tuator to ross member a elerometer transfer fun tion times the ontroller transfer fun tion. There is a large gain at 150 Hz, and tworossovers, one at147 Hzand theother at153 Hz,with orrespondingphasesof-67and 72
degrees.
a orresponding phase of 72 degrees. Again, the rossoveris stable, and thephase margin
is 108 degrees. This Bode plot demonstrates that the ontroller designed above should be
stableandattenuatedisturban esatfrequen iesnear150Hzandhavelittleee telsewhere.
Theexpe tedperforman edemonstrated bytheBodeplotis showninFigure4-7.
Con-sider the losed loop transferfun tion from the shaker to the ross member a elerometer
of the ontroller used here. The measured experimental transfer fun tion should losely
resemblethe transferfun tion al ulatedfromtheory. Figure4-7showsa omparison of
ex-perimental and predi ted results. The experimental losed loop transferfun tion from the
shakerex itationto the rossmembera elerometer ismeasuredusing thesignalanalyzer,
and is represented by the thin line. The thi k line represents the theoreti al losed loop
transfer fun tion al ulated using Equation 4.8. The experimental and theoreti al results
aresimilar, although thebandwidth of theexperimental transferfun tion is narrowerthan
pro-100
125
150
175
200
−80
−60
−40
−20
0
Magnitude (dB, g/V)
100
125
150
175
200
−600
−400
−200
0
Frequency (Hz)
Phase (degrees)
Experimental
Calculated
Figure 4-7: Experimental and al ulated transfer fun tions from the shaker ex itation to
the ross member a elerometer. The thin line is the experimentally determined transfer
fun tion and the thi kline is the al ulated transferfun tion basedon ontrol theory. The
theoreti al al ulatedtransferfun tionandtheexperimentaltransferfun tionmat h losely.
essingdone on the experimental data. The signalanalyzera quires theexperimentaldata
at a nite rate, and italso averages several streams of data. The lose mat h between the
experimental and theoreti al transfer fun tions indi ates that the experimental ontroller
does what it was expe ted to do, and the ross member vibration is signi antly redu ed
near150 Hz.
4.4 Closed Loop Vibration
The purpose of the tonal ontroller is to eliminate ross member vibration over a narrow
band offrequen ies. Asa result, themagnitudeof thetransferfun tion from theshakerto
the ross member a elerometer should de rease signi antly at 150 Hz when the ontrol
loop is losed. This drop an be seen by omparing the open loop transfer fun tion from
100
125
150
175
200
−40
−20
0
Magnitude (dB, g/V)
100
125
150
175
200
−300
−200
−100
Frequency (Hz)
Phase (degrees)
Open Loop
Closed Loop
Figure 4-8: Open and losed loop transfer fun tions from the shaker to the ross
mem-bera elerometer. The thin line shows the open loop transfer fun tion and the thi k line
shows the losed loop transfer fun tion. At 150 Hz the ross member vibration de reases
dramati ally, asexpe ted.
theshaker to the ross membera elerometer. (Figure 4-8) The thin line is theopen loop
transferfun tionandthethi klineisthe losedlooptransferfun tion. Asexpe ted,the ross
member vibrations (open and losed loop transfer fun tions) aresimilar at all frequen ies,
ex ept at those around 150 Hz, where the vibration goes to nearly zero. The vibration
de reaseextends for onlya fewhertz oneitherside of 150Hz, roughly148 Hz to 152Hz.
Unlikethe rossmembervibration,themagnitudeofthe ontrolarmvibrationat150Hz
would not be expe ted to de rease when the ontrol loop is losed. The ontrol arm
a - elerometer is on the opposite side of the point 4 bushing, relative to the ross member
a elerometer. A ontroller thatredu es the a eleration on one side of thebushing ( ross
member)isnotexpe tedtoredu ethea elerationontheothersideofthebushing( ontrol
arm). Figure4-9 shows theopen (thin line) and losed loop (thi k line) transferfun tions
from theshakerto the ontrol arm a elerometer. Nosigni ant dieren e isseen between
100
125
150
175
200
−30
−20
−10
0
Magnitude (dB, g/V)
100
125
150
175
200
−600
−400
−200
0
Frequency (Hz)
Phase (degrees)
Open Loop
Closed Loop
Figure 4-9: Open and losed loop transfer fun tions from the shaker to the ontrol arm
a elerometer. The thin line is the open loop transfer fun tion and the thi k line is the
losed loop transfer fun tion. As expe ted, there is no signi ant de rease in vibration at
150 Hz. At 142Hz, thereis ade rease inthe magnitudeof vibration.
shiftupwardsofmagnitude ofvibrationat frequen iesabove 150Hz,probablydueto small
experimentaldieren es. Also, thereis a de reaseof roughly6 dBinthemagnitude ofthe
vibration at 142 Hz. This result isdi ult to explain. Withthe addition of the ontroller
( losedloop),the onlypartoftheloopthat hangesistheee t ofthe ontroller on ontrol
arm vibration. Therefore, we would expe tthe ontrol armvibration to hange only when
the ontroller signal is large, but without a more detailed examination of the ontrol arm
transferfun tions, itis di ult to predi t what ee t the ontroller wouldhave. However,
how the ontrol arm vibration hanges is not very important, assuming that ontrol arm