Sei sulla pagina 1di 10

EEL3105 Fall2011

Homework6
Considermatrices

1 2
1 2 3
1 3
4
0 1
0 1 0
0 0 1
A
k
B
b b
C
c c c
(
=
(

(
=
(


(
(
=
(
(

1. ForA,choosek=3.Findtheeigenvaluesoftheresultingmatrix.Aretheyreal?Why?
Solution:ThecharacteristicpolynomialofAis

2
1 3
det( ) det 5 5
3 4
I A

(
= =
(



Itsrootsaregivenby5.85and0.85whicharetheeigenvaluesofA.Theyarereal.Thereasonisthat
thematrixAisarealsymmetricmatrixfork=3.Also,notethatthesumofeigenvaluesis5whichisthe
traceofAandtheirproductis5whichisthedeterminantofA.
2. FindaformulaforeigenvaluesofAasafunctionofk.Supposekisadesignparameter
whichrangesfrom100to+100.UsetheformulatoplottheeigenvaluesofAasafunction
ofkinthecomplexplane.UseMATLABtoverifyyouranswer.Suchaplotiscalledroot
locus.Commentonyourresults.
Solution:Nowwewillletkremainadesignparameter.Thecharacteristicpolynomialis

2
1 3
det( ) det 5 (4 3 )
4
I A k
k

(
= = +
(



ItsrootswhicharetheeigenvaluesofAaregivenby:

5 9 12
2
k +

Weneedtoplotthesetworootsasafunctionofk.First,letusanalyzethebehavioroftheseroots.Their
sumwillalwaysbe2.5.If9+12Kisnonnegative,i.e., 0.75 100, k s s ,thentherootswillbereal
addingupto5.Atk=100,theeigenvaluesare14.89and19.89.Atk=0.75,theeigenvaluesare2.5and
2.5,i.e.,wehaverepeatedeigenvalues.Inproblem1,wesawthatfork=3,theeigenvaluesare5.85and
0.85.Atk=0,theyare1and4.Thus,askdecreasesfrom100to0.75,theeigenvaluesgetclosertoeach
otheruntiltheymeetat2.5.Now,whathappensfortheremainingrangeofk, 100 0.75? k s <
Well,thefirstthingwecanconcludeisthattheeigenvaluesarecomplexasthequantityunderthe
squarerootsignisnegative.Inthisrangeofk,theeigenvaluesare
2.5 ( 9 12 ) / 4 j k
Fork=100,theeigenvaluesare 2.5 17.25 j .Askincreasesfrom100to0.75,thesetwocomplex
rootsgetclosertoeachotheralongtheverticaldirectionandatk=0.75,theymeetat2.5.This
completesthepictureoftheeigenvaluesaskvariesfrom100to100.Theplotisshownbelow.

Figure1

WecanuseMATLABtomakethisplotaswell.[Pleaseseemycommentsonthecolorcodeforfiguresin
MATLABinProb4solution.HereIhaveplayedaroundwiththecolorcode!]
k=-100:.01:100;
EA=zeros(20000,2);
for n=1:1:20000
b1=k(n);
A=[1 3;b1 4];
v= eig(A);
EA(n,1)=v(1);
EA(n,2)=v(2);
end
plot(EA)

Figure2

3. FindcharacteristicpolynomialsofBandCintermsofb
1
,b
2
,c
1
,c
2
,andc
2
.Doyouseeany
patterninyouranswer?Canyouthinkofawaytogeneralizethistoannxnmatrix?
Answer:ThecharacteristicpolynomialofBandCaregivenby

2
2 1
1 2
2
3 2 1
1 2 3
3 2
3 2 1
1
det( ) det
1 0
det( ) det 0 1 ( ) ( 1)
I B b b
b b
I C c c c
c c c
c c c


(
= = + +
(
+

(
(
= = + +
(
( +

= + + +

-15 -10 -5 0 5 10 15 20
-20
-15
-10
-5
0
5
10
15
20
Thepatternappearstobeasfollows:matriceshave1sonthesuperdiagonal[diagonalabovethe
diagonal]andthecoefficientsinthelastrowappearascoefficientsofthecharacteristicpolynomial.We
canseeifthefollowingguessforannxnmatrixiscorrect:

1 2 3
1 2 2
1 3 2 1
0 1 0 0
0 0 1 0
0
0 0 0 1
det( ) ...
n
n n n
n n
A
I A
o o o o
o o o o o

(
(
(
( =
(
(
(


= + + + + + +

Theeasiestwaytotrytodothisisproofbyinduction.So,supposethisformulaholdsfor(n1)x(n1)
matricesofthisform.Canwethenshowthatitholdsfornxnmatricesofthesameform?First,we
write

1 2 3
1 0 0
0 1 0
0
0 0 0 1
n
I A

o o o o
(
(

(
( =
(

(
(
+

Letuscomputethedeterminantofthismatrixbyexpandingitviathefirstcolumn.Nowthecofactorof
1,1entryispreciselyan(n1)x(n1)matrixofthesameform.Sincewehaveassumedthattheformula
holdsfor(n1)x(n1)matrices,wecanuseitforthecofactorofthe1,1term.Now,theonlyothernon
zeroentryinthefirstcolumniso
1
.Itcofactoris(1)
(n1).
Thus,wegetthecharacteristicpolynomialofA
tobe

( 1) ( 1)
1
1 2 3 2
1 3 2 1
1 2 2
1 3 2 1
det( ) ( (1,1)) ( 1) ( 1)
[ ... ] ( 1)
...
n n
n n n n
n n
n n n
n n
I A cofactor o
o o o o o
o o o o o
+

= +
= + + + + + +
= + + + + + +

Thisshowsthattheformulaholdsfornifweassumeitholdsfor(n1).Sincewehaveshownitforn=2,
byinductionwehaveshownitforanyn.
Thesetypesofmatricesarecalledcompanionmatrices.Bytheway,wecantransposethesematrices
andgetanotherformofcompanionmatriceswith1salongthesubdiagonalandtheosalongthelast
column.
Companionmatricesplayaroleinthetheoryofdigitalfilters.Digitalfiltersareimplementationsof
analogfiltersusingdigitalsignalprocessingchips.[Intheolddays,analogfilterswereconstructedusing
R,L,Candoperationalamplifiers.]Digitalsignalprocessingisaverylargebusiness.TexasInstruments,
STMicroelectronics,etc.havelargebusinessesindigitalsignalprocessingchips.Suchfiltersareusedin
consumerelectronicsdevicessuchastelevisions,musicplayers,DVDplayers,etc.See
http://en.wikipedia.org/wiki/Digital_signal_processing,http://www.ti.com/forinterestinginformation
thesechips.
4. Fixb
2
=1.Supposeb
1
beadesignparameterthatrangesfrom0to100.Itmayrepresentthe
valueofaphysicalcomponent(e.g.,massorresistance).FindtheeigenvaluesofBasa
functionofb
1
andplottheminthecomplexplane.Commentonyouranswer.Supposefor
certaindesignconsiderations(havingtodowithtransientresponse),wewanttheangleof
theeigenvaluestobe 3 / 4. t Choosethevalueofthedesignparameterb
1
tosatisfythis
constraint.
Solution:WestartbyevaluatingthecharacteristicpolynomialofB.
2
2 1
1 2
1
det( ) det I B b b
b b

(
= = + +
(
+


Nowb
2
=1.Wecansubstitutethisvalueintothecharacteristicpolynomialandsolveforrootsof
2
1
0. b + + =
Wecancomputetherootsofthisquadratic.Forb1between0and0.25,theyarerealandaregivenby

1
1
1
2
1 1 4
2
1 1 4
2
b
b

+
=

=

Forb
1
>0.25,theyarecomplexandaregivenby

1
1
1
2
1 4 1
2
1 4 1
2
j b
j b

+
=

=

Atb
1
=0,theyare0and1.Asb
1
increasesfrom0to0.25,theeigenvaluescomeclosertoeachother
suchthatatb
1
=0.25,theycoalesceat0.5.Forb
1
>0.25,theybecomecomplexwithrealpartbeing0.5.
Wecanplotthisonthecomplexplaneasbelow:

Figure3
Nowweareaskedtofindthevalueofb
1
forwhichtheanglesoftheseeigenvalues(ascomplex
numbers)will3/4.First,notethatthelinewithangleof3/4fromthepositiverealaxisattheorigin
goesfrom0bisectingthe2
nd
quadrant.Likewise,thelineinthecomplexplaneofcomplexnumberswith
angle3/4goesfromtheoriginbisectingthe3
rd
quadrant.Fromageometricviewpoint,wearelooking
fortheintersectionoftheselineswiththeplotoftheeigenvalues.Thisisshownthefigurebelow.
Wecansolveforthispointalgebraically.Notethattheangleoftheeigenvalue

1
1 4 1
2
j b +

is
1
arctan 4 1 b t .Wewantthistobeequalto3/4.Therefore,

1
arctan( 4 1) / 4 b t =
Solvingforb
1
givesb
1
=0.5.
WecanuseMATLABforplottingaswellasvisualizingthegeometry.
MATLABCode:
b=0:.01:100;
EA=zeros(10000,2);
for n=1:1:10000
b1=b(n);
B=[0 1;-b1 -1];
v= eig(B);
EA(n,1)=v(1);
EA(n,2)=v(2);
end
plot(EA)

Figure4
InFigure4,IdontlikethecolorcodeMATLABchoosesbecauseitseemstoindicatethattheeigenvalue
fromthepositiveimaginarypart[blue]goestotherighttowardtheoriginwhiletheonewiththe
negativeimaginarypart[red]goesleftto1aftertheybothmeeteachotherat0.5ontherealaxis.
Otherthanthisminorquibble,theplotinFigure4isverymuchliketheplotwegotfromourown
analyticalcalculationsinFigure3above.
Wecannowzoomintothepartwherethe3t/4lineswillintersecttheplotasshownbelow.
-1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0
-10
-8
-6
-4
-2
0
2
4
6
8
10

Figure5

5. LetAbeannxnmatrix.ShowthateigenvaluesofAarethesameasthoseofA
T
.
Thisiseasy.EigenvaluesofAaretherootsofdet(IA).Now,det(IA)=det(IA)
T
=det(IA
T
).The
rootsofthelastpolynomialaretheeigenvaluesofA
T
.Thus,theeigenvaluesofAandA
T
arethe
same.WehavealsoshownintheprocessthatthecharacteristicpolynomialofAandA
T
arethe
same.
6. Considermatrices

1 2
4 5
3 1
3 6 9
2 4 8
A
B
(
(
=
(
(

(
=
(

CalculateeigenvaluesofABandthoseofBA.YoucanuseMATLABtocomputethese
answers.Pleasecommentonyouranswers.
6.
>>A=[12;45;31];
>>B=[369;248];
>>AB=A*B;
>>BA=B*A;
>>eig(AB)
ans=

0.0000
5.0000+12.2066i
5.000012.2066i

>>eig(BA)

ans=

5.0000+12.2066i
5.000012.2066i
TheeigenvaluesoftheABandBAarethesameexceptthatABalsohasazeroeigenvalueinaddition.Is
thisacoincidence?No!Thisisthenextproblem.

7. NowsupposeAandBaretwomatrices(notnecessarilysquare)suchthatABandBAare
bothwelldefined.Suppose isanonzeroeigenvalueofAB.Showthatitisalsoan
eigenvalueofBA.
Solution.SupposeisanonzeroeigenvalueofAB.Thenthereexistsanonzeroeigenvectorvsuch
that
ABv v =
WeneedtoshowthatisalsoaneigenvalueofBA.WewilldothisbyproducinganeigenvectorofBA.
NotethatBvcannotbezerosinceisnonzero.Letw=Bv.Now,
BAw BABv B v Bv w = = = =
Thus,wisaneigenvectorofBAwitheigenvalue.
==========================================================
Notforgrading[hardproblemforexplorationandchallenge]:
8. LetusreturntoProblem3.RecallItoldyouinclassthatinordertoensurestability,all
eigenvaluesshouldhavenegativerealparts.Ifwehavenumericalvaluesfortheparameters
b
1
,b
2
,(orc
1
,c
2
,andc
2
),thenwecanuseMATLABtocomputetheeigenvaluesandseeifthe
realpartsofeigenvaluesarenegative.Butiftheyareunknownparameters,thenthistaskis
harder.Thisisespeciallysoasthematrixsizegetslarger.Howcouldyoucheckifall
eigenvalueshavenegativerealpartsintermsoftheparameters?Explorethewebresources
tofindananswertothisquestion.
Solution:Stabilityofasystemrequiresthattheeigenvalueshavenegativerealparts.Inotherwords,all
theeigenvaluesmustbeinthelefthalfofthecomplexplane.Fromamathematicalpointofview,we
needtocheckifallrootsofapolynomial(thecharacteristicpolynomial)areinthelefthalfofthe
complexplane.Whentheoryofcontrolsystemswasbeingdeveloped,thiswasamajorissuesince
computationofeigenvalueswasveryhard.ThisproblemwassolvedbyE.J.Routhandtheresultingtest
iscalledRouthHurwitztest.Itisanalgorithmthattakescoefficientsofthecharacteristicpolynomialand
doessomesimplealgebraiccalculationsandtestswhetherallrootsareinthelefthalfplane.Fordetails
ofthealgorithm,seehttp://en.wikipedia.org/wiki/Routh%E2%80%93Hurwitz_stability_criterionandthe
bookscitedattheendofthatwebpage.

Potrebbero piacerti anche