• No results found

A well-posedness result for a shear wave propagation model

N/A
N/A
Protected

Academic year: 2020

Share "A well-posedness result for a shear wave propagation model"

Copied!
18
0
0

Loading.... (view fulltext now)

Full text

(1)

Propagation Model

H.T. Banks

, H. Tran y

,S. Wynne z

December 14,2001

Abstract

We consider a nonlinearmodel for propagation of shear wavesin

viscoelastic tissue. Existence anduniqueness resultsfor solutionsare

established.

1 Introduction

Inthisnotewe examinethewell-posednessofaone-dimensionalshearwave

propagationmodel that arises in inverse problems related to the detection

and characterization ofcardiac artery stenoses. Ina previouspaper [6 ],we

deneda basicmodelto emulate shear waves propagating from acoronary

stenosis through a homogeneous, soft-tissue like medium. The medium is

considered viscoelastic, and the model uses internal strain variables (see,

e.g., [1 ], [3], [4 ], or [5 ]) to capture the nonlinear stress-strain relationship.

Anidealizedgeometry(basedonexperimentalprotocolstotestpiezoceramic

basedsurfacesensors) isdepictedin Figure1.

Asoutlinedin[1 ],theevolutionequationforone-dimensionalshearwave

propagationthrougha homogeneous, viscoelasticmediumis

@ 2

@t 2

u(t;x) @

@x

(t;x)=F(x;t); R

1

<x<R

2

; (1)

CenterforResearchinScienticComputation,Box8205,NorthCarolinaState

Uni-versity,Raleigh,NC27695-8205

y

CenterforResearchinScienticComputation,Box8205,NorthCarolinaState

Uni-versity,Raleigh,NC27695-8205

z

CenterforResearchinScienticComputation,Box8205,NorthCarolinaState

(2)

Sensor array

Acoustic disturbance

R

Gel mold

Tube

Figure1: The 1Dhomogeneous viscoelasticmodel.

whereurepresentsthesheardisplacement,representstheshearstress,and

F representsa bodyforcingterm. Forboundaryconditionsa pureshearing

force on the left boundary and a free surface on the right boundary were

assumed;hence,

(t;R

1

)=f(t); (t;R

2

)=0: (2)

Theinitial conditionswere u(0;x)=u

0

(x), and u

t

(0;x)=u

1 (x):

The focus of [1 ] concerns the choice of an eective constitutive

equa-tion forthis model. In that paper, the authors investigated internal strain

variablemodelsasalternativestothecomputationallyintensivequasi-linear

viscoelastic model proposed by Fung [9 ]. Specically, they assumed the

stress isgiven asa sumof internalstrain variables,

(t)= N

X

j=1

j

(t): (3)

Thedynamics ofeach internal strainvariable ismodeled dynamicallyas

d

j (t)

dt

=

j

j +C

j d

e

dt (u

x

(t));

j

(0;x)=0; j=1;:::;N; (4)

where

e

is the elastic response function dened in ([9 ],x7), and may be

givenas

e (u

x

(t;x))=+e ux

(3)

this formulation, with linear internal strain variable models, is equivalent

to Fung's formulation with a sum of exponential terms approximating the

relaxation function. More generally however, the internal strain variables

mightbe modeled by nonlineardynamics ofthe form

d

j (t)

dt =g

j (

j

(t))+C

j d

e

dt (u

x

(t));

j

(0;x)=0; j =1;:::;N: (6)

Allof these models correspond to a viscoelastic body under either loading

orunloading. Eachconstitutiveequation expressesthestressnonlinearlyin

termsof theinnitesimalstrainu

x .

The authors of [1] investigated three particular internal strain variable

modelsasconstitutiveequations: aone linearinternalstrainvariablemodel

( =

1

), a two linear internal strain variable model ( =

1 +

2 ), and

onepiece-wise linearinternalstrainvariablemodel. Numerical experiments

veried theeectiveness of the internal strain variable models and

demon-strated good agreement with simulated data in the case of the two linear

internalstrain variable model.

In thisnote, we focus, forsimplicity, on theoretical foundationsfor the

one linear internal strain variable formulation. The case of multiple linear

internalstrainvariablesisreadilytreatedinthesameway. Theshearstress,

,isgiven by theequation

=

1 +C

D u

tx ;

where we have added a Kelvin-Voigt damping term withC

D

>0 asa rst

approximation to damping present inviscoelastic materials. The terms

1

and

e

areassumed given by

d

dt

1

+

1 =c

d

dt

e (u

x

(t;x));

1

(0;x)=0 (7)

e (u

x

(t;x))=+e ux

: (8)

We thenanalyze thesystem

u

tt C

D u

txx

1x

=F inV

(9)

u(0;x)=u

0

2V (10)

u

t

(0;x)=u

1

2H; (11)

where H = L 2

(), and V =H 1

(), and = [R

1 ;R

2

]. The inner product

(4)

embeddedin H,and H is continuously embedded inV

,the dualspace of

V.

The organization of thispaper is as follows. We rst deneweak

solu-tionsto system(7)-(11), and listsomeassumptionsinSection2. InSection

3 we develop the Galerkinapproximation, utilizingseveral lemmas, and in

Section4weestablishtheexistence anduniquenessofbothlocalandglobal

weaksolutions. Thisworkadaptsthetechniquesof[1 ]and[2 ]to oursystem

withlinearinternalstrain variablesbutwithnonlinearstress-strain

interac-tion.

2 Preliminaries

We will interpret system(7)-(11) in theV

sense. In developing a general

theory, we will at various times invoke several from among the following

assumptions:

(AF) Theforcing termsatisesF 2L 2

(0;T;V

)

(Af)The innerboundaryconditionsatises f 2L 2

(0;T)

(AL)Theelasticresponsefunction

e

satisesalocalLipschitz

condi-tion,

k

e

(u)

e

(v)k L

B

r

ku vk

forsome positiveconstant L

B

r

andforallu,v inB

H

(0;r),theballin

H centered at 0 ofradius r.

(AG) ThereexistsconstantsC

1 andC

2

such that

k

e

(u)kC

1

kuk+C

2

forevery u2H.

Notethat assumption(AL) canbeveriedfor

e

in[1](i.e.,

e

given in

(5)above)byrstcomputing d

dt

e

(tu+(1 t)v),integratingwithrespecttot

overtheinterval[0;1],thentakingthenormofbothsides. Assumption(AG)

isaphysicalboundonthegrowth of

e

priorto rupture. Itis satisedbya

modied versionof

e

in(5), callit~

e

,inwhich~

e

accounts forsaturation

before rupture and agrees with

e

up to this saturation. However, (AG) is

notsatisedbythe

e

of (5)asit isdenedthere.

We have the following denition of weak solutions for the one linear

(5)

Denition 2.1 LetL

T

=fw:[0;T]!H :w2C

W

([0;T];V)\L ([0;T];V)

and w

t 2C

W

([0;T];H)\L 2

(0;T;V)g. We dene u2L

T

to be a weak

so-lution of system(7)-(11) if it satises

Z

t

0 [ (u

s (s);

s

(s))+C

D (u

sx (s);

x

(s))+(

1 (u

x (s));

x

(s))]ds

+(u

t

(t);(t)) (u

1 ;(0))

= Z

t

0

[<F(s);(s)>

V

;V

f(s)(s;R

1

)]ds (12)

for any t 2[0;T] and 2L

T

, with the initial conditions u

0

2 V, u

1 2 H,

and

1 (u

x

(t;x))=c

e (u

x

(t;x)) e t

e (u

0x )

Z

t

0 e

(t s)

e (u

x

(s;x))ds

:

(13)

Notethisnotion ofweaksolutionfor system(7)-(11) agrees withtheusual

oneinthatityieldsu

tt 2L

2

([0;T];V

)withequation(9)holdinginthesense

ofL 2

([0;T];V

). Here, C

W

([0;T];V) refersto the setof weakly continuous

functionsinV on [0;T].

We rst establish existence of local weak solutions under only the

as-sumptions(AF),(Af),and(AL).Todealwiththenonlinearelasticresponse

term, we rst denetheoperatorP astheradial retractionfrom thespace

H onto the ballB

H (u

0

x

;1) of radius 1 centered at u

0x

. Then we dene a

newelasticresponse function^

e by

^

e

(u)=

e

(Pu); 8u2H:

Thus, fromassumption (AL),one can easilyargue theglobalconditions:

(AL2)k^

e

(u) ^

e

(v)k L

B

1+ku

0x k

ku vkforall u;v2H,

(AG2) k^

e

(u)kC

1

kuk+C

2

forallu2H.

We also denea modiedinternalstrain^

1

asfollows

^

1 (u

x

(t;x))=c

^

e (u

x

(t;x)) e t

^

e (u

0x )

Z

t

0 e

(t s)

^

e (u

x

(s;x))ds

:

(6)

We begin our arguments by developing the standard Galerkin

approxima-tion and establishing a priori bounds for them. Let f

i

g be any linearly

independenttotalsubsetofV. Foreachm,letV m =spanf 1 ; 2 ;:::; m g. Choose fu m 0

g and fu m

1 g2V

m

such thatu m

0 !u

0

inV andu m

1 !u

1 inH,

andlet M

0

andM

1

beconstants suchthat

ku m 0 k V M 0 and ku m 1

kM

1

: (15)

We dene the Galerkin approximation u m (t) = P m k=1 a m k (t) k as the

uniquesolutionofthe followingm-dimensionalintegro-dierentialsystem

<u m tt ; j > V ;V +C D (u m tx ; jx )+c( ^

1 (u m x ); jx

) = <F(t);

j > V ;V f(t) j (R 1 ) (16)

forj=1;:::;montheinterval[0;T]forsomeT >0. Wehavethefollowing

a priori estimateforthe Galerkinapproximation.

Lemma3.1 Let u m

(t) bethe Galerkin approximation on [0;T]. There

ex-istsa constant K >0 such that

ku m t (t)k 2 +ku m x (t)k 2 +C D Z t 0 ku m sx (s)k 2

dsK (17)

where K depends on the problem data (i.e., u

0 , u

1

, f, F, c, , , , C

D ,

andT), but isindependent of m.

Proof. Toobtainthea prioriestimatewemultiply(16)by d

dt a

j

(t)andsum

forj=1;:::;m,

1 2 d dt ku m t k 2 +C D ku m tx k 2

=<F(t);u m t > V ;V f(t)u m t (R 1 ) c(^ e (u m x );u m tx ) +ce t (^ e (u m 0x );u m tx )+c(

Z t 0 e (t s) ^ e (u m x (s))ds;u m tx ): Adding(u m x ;u m tx

) to each sideofthe above equation,wehave

1 2 d dt ku m t k 2 + 1 2 d dt ku m x k 2 +C D ku m tx k 2

j<F(t);u m t > V ;V

j+jf(t)u m t (R 1 )j +j(c^ e (u m x )+u

m

x ;u

m

tx

)j+cj(^

(7)

(AG2), estimates (15), the embedding H 1

() ,! C(), and standard

in-equalitiesto obtain

d dt (ku m t k 2 +ku m x k 2

)+C

D ku m tx k 2 C D ku m t k 2 +10(cC 1 +1) 2 C 1 D ku m x k 2 +5C 1 D kF(t)k 2 V +5c 2 3 C 1 D jf(t)j 2 +10c 2 C 2 2 C 1 D +5c 2 C 1 D (C 1 M 0 +C 2 ) 2 +5(c) 2 C 1 D (C 1 (M 0

+1)+C

2 ) 2 T 2 : (18)

Integrating from 0 to t,we obtain

ku m t k 2 +ku m x k 2 +C D Z t 0 ku m sx (s)k 2

dsku m 1 k 2 +ku m 0x k 2 + ~ K Z t 0 (ku m s (s)k 2 +ku m x (s)k 2

) ds+5C 1 D Z T 0 kF(s)k 2 V ds +5c 2 3 C 1 D Z T 0 jf(s)j 2

ds+10c 2 C 2 2 C 1 D T +5c 2 C 1 D (C 1 M 0 +C 2 ) 2

T+5(c) 2 C 1 D (C 1 (M 0

+1)+C

2 ) 2 T 3 :

ApplyingGronwall'sinequalitywecanconcludethatthesequencesfku m t )k 2 g and fku m x k 2

garebounded. Hencethere existsa positive constant

K = K(M

0 ;M

1

;c; ;T;kFk

L 2

(0;T;V

) ;kfk

L 2

(0;T)

) independent of m such

that ku m t (t)k 2 +ku m x (t)k 2 +C D Z t 0 ku m sx (s)k 2

dsK (19)

foreach t2[0;T]. Thisproves thelemma.

Lemma3.2 Letu m

(t)betheGalerkinapproximation on[0;T]. Thenfu m

g

isbounded uniformlyin C([0;T];H)L 2

([0;T];H).

Proof. Forfu m

t gL

2

([0;T];H), we have

(u m

(t

2

);) (u m (t 1 );)= Z t2 t 1 (u m s (s);)ds

for all 2 H and forany t

1 ;t

2

2[0;T]. We take a sequence ft k

1

g 2 [0;T]

such thatt k

1

!0 ask!1, then

(u m

(t

2

(8)

j(u m

(t);)j j(u m

0

;)j+ Z

t

0 j(u

m

s

(s);)jds

ku

m

0

kkk+ Z

T

0 ku

m

s

(s)kkkds

M

0 +TK

1=2

kk:

Thus ku m

(t)k M

0 +TK

1=2

, and fu m

(t)g is boundeduniformlyin H for

t2[0;T]and hence inC([0;T];H).

In thefollowing we repeatedly take subsequencesof sequencesof fu m

g.

Ineach case we againdenote thesubsequence asfu m

g.

Lemma3.3 There exist functions u 2L 2

([0;T];V) and u^ 2L 2

([0;T];V),

and a subsequencefu m

g such that

u m

!u weakly in L 2

([0;T];V)

u m

t

!u^ weakly in L 2

([0;T];V) :

Proof. From Lemma 3.1 and Lemma 3.2 we have that fu m

g is bounded

uniformlyin C([0;T];V) L 2

([0;T];V). We also have, from Lemma 3.1,

thatfu m

t

gisboundeduniformlyinL 2

([0;T];V). Wethenapplythe

Banach-AlaogluTheoremto obtainthedesired resultsinthelemma.

Lemma3.4 Theset fu m

g is an equicontinuous andbounded subsetof

C([0;T];V), moreover,

u m

(t)!u(t) weakly in V

uniformlyin t2[0;T], i.e.,u m

!u weakly in C

W

([0;T];V).

Proof. The boundedness result follows from Lemma 3.1 and Lemma 3.3.

To prove theequicontinuity,we have

u m

(t+t) u m

(t)= Z

t+t

t u

m

s (s)ds

fort;t+t2[0;T]. UsingLemma3.1, we obtain

ku m

(t+t) u m

(t)k

V

Z

t+t

t

ku m

s (s)k

V ds

t

1=2

(TK+C 1

D K)

1=2

(9)

Thus, forany >0 and t2[0;T], 9 Æ(;t) =(=K(T +C

D ))

2

such that

jt 0

tj<Æ impliesku m

(t 0

) u m

(t)k<.

For theconvergence, we usea version ofthe Arzela-Ascolitheorem(see

[2 ] or [10 ], Thm. 3.17.24). Let Y = B

V (0;K)

V

, the closure in V of the

ball centered at zero with radius K taken with the weak topology. Y is

a complete metric space. Let F = fu m

g C([0;T];Y). By the above

estimate,equicontinuityintheV senseimpliesequicontinuityintheY sense.

Also,foreach t2[0;T], theset fu m

(t);u m

2Fgis relativelycompactin Y

(Banach-Alaoglu Theorem). Then F is relatively compact in C([0;T];Y),

i.e.,thereexistsasubsequence,denotedfu m

gagain,suchthatu m

(t)!u(t)

inY uniformlyint2[0;T],i.e.,

u m

(t)!u(t) weakly inV uniformlyint2[0;T]:

Lemma3.5 The derivative u

t

exists in the V sense and u

t

= u^ a.e. in

[0;T].

Proof. Now u

t 2L

2

([0;T];V)if thereexists av2L 2

([0;T];V) such that

u(t

2 ) u(t

1 )=

Z

t2

t

1

v(s)ds; t

1 ;t

2

2[0;T]

where R

t2

t

1

v(s)dsor R

t2

t

1 u

s

(s)dsisdened usingduality,i.e.,

<u(t

2 ) u(t

1 );>

V

;V =

Z

t2

t

1 <u

s

(s);>

V

;V ds

forall 2V and fort

1 ;t

2

2[0;T]. Now

<u m

(t

2 ) u

m

(t

1 );>

V

;V =

Z

t

2

t1 <u

m

s

(s);>

V

;V ds

for all 2 V and for t

1 ;t

2

2 [0;T]. By Lemma 3.3 u m

t

! u^ weakly in

L 2

([0;T];V), hence

Z

t2

t

1 <u

m

s

(s);>

V ds!

Z

t2

t

1

<u(s);^ >

V

ds 82V:

ByLemma 3.4u m

(t)!u(t)weakly inV uniformlyin t2[0;T], hence

<u(t

2 ) u(t

1 );>

V =

Z

t

2

t1

<u(s);^ >

V ds

forall 2V and fort

1 ;t

2

2[0;T]. Thereforeu

t

(t)=u(t),^ a.e. t2[0;T]in

(10)

Lemma3.6 Thesequence fu

t

g converges to u

t

strongly in L ([0;T];H).

Proof. Use Aubin's Theorem ([7 ], Lemma 8.4). We have fu m

t

g bounded

uniformlyinL 2

([0;T];V). If we can show thatfu m

tt

g is bounded uniformly

inL 2

([0;T];V

), then, sinceV ,!H compactly and H,!V

continuously,

there existsasubsequence,denoted fu m

t

gthat converges inL 2

([0;T];H).

NowL 2

([0;T];V

)=L 2

([0;T];V)

. FixM andlet

M = P M k=1 c k (t) k withc k

(t)2C 1

([0;T]). Then formM,

j<u m tt ; M > L 2 ([0;T];V) ;L 2

([0;T];V) j= Z T 0 <u m ss (s); M (s)> V ;V ds :

Usingequation (16)to substitute infor<u m ss (s); M (s)> V ;V

,wehave

j<u m tt ; M >j R T 0

j<F(s);

M (s)>

V

;V jds+

R T 0 j f(s) M (s;R 1 )jds + R T 0 j(C D u m sx (s); M x

(s))jds+ R

T

0 j ( ^

1 (u m x (s)); M x (s))jds C D ku m tx k L 2 ([0;T];H) +k ^

1 k

L 2

([0;T];H) +kFk

L 2 ([0;T];V ) k M k L 2 ([0;T];V) + R T 0 jf(s)jk M (s;R 1 )k L 1 ds;

where we used Cauchy-Schwartz and Young's inequality in the last step.

Now k ^

1 k 2 L 2 ([0;T];H) = R T 0 k ^ 1 (u m x (s))k 2

ds and, using the denition of ^ in

equation (14), we have

k

1 (u

m

x

(t))kc k^ e (u m x

(t))k+k^

e (u m 0 x )k+ Z t 0 k^ e (u m x (s))kds c C 1 (ku m x

(t)k+ku m

0

x

k)+2C

2 + Z T 0 C 1 ku m x

(s)k+C

2 ds c 2C 2 +C 1 (M 0

+K+T(C

1 K 1=2 +C 2 )

wherewehaveusedtheboundforu m

0

andLemma3.1inthelaststep. Thus,

we have R T 0 k ^ 1 (u m x (s))k 2

ds C(c;C

1 ;C

2 ;M

0

;K ;T) bounded independent

ofm. Also,usingthefactthat,inonedimension,H 1

(),!C 0

(),wehave

Z T 0 jf(s)jk M (s;R 1 )k L 1ds kfk L 2 ([0;T]) k M k L 2

([0;T];V)

: (20)

Hence,

j<u m tt ; M > L 2 ([0;T];V) ;L 2

([0;T];V) j

K+C+kFk

L 2

([0;T];V

) +kfk

(11)

Sincef

M g

M=1

forma dense subsetof L ([0;T];V),wehave

ku m

tt k

L 2

([0;T];V

)

K+C+kFk

L 2

([0;T];V

) +kfk

L 2

([0;T])

and thus fu m

tt

g isuniformlyboundedinL 2

([0;T];V

).

Lemma3.7 Thefunctionsfu m

t

gareboundedinC([0;T];H)andaree

quicon-tinuous in C

W

([0;T];H); moreover, for each t,

u m

t

(t)!u

t

(t) weakly in H:

Proof. The boundedness statement follows from Lemma3.1. The

conver-gencestatement willfollowfrom an applicationof Arzela-Ascoliin

C

W

([0;T];H) oncewe establishtheequicontinuity. To dothiswe rst note

that, forv2V,

j<u m

t

(t+t) u m

t

(t);v>

V

;V

j

Z

t+t

t

j<u m

ss

(s);v>

V

;V jds

ku

m

tt k

L 2

([0;T];V

) kvk

V t

1=2

~

Ckvk

V t

1=2

wherewehaveusedtheresultfromtheLemma3.6toboundtheku m

tt kterm

intermsof ~

C=K+C+kFk

L 2

([0;T]

V )

+kfk

L 2

([0;T])

,independentof m.

Assume nowthat2H and x>0. Forv2V and t;t+t2[0;T],

we have

j(u m

t

(t+t) u m

t

(t);)j j<u m

t

(t+t) u m

t

(t);v>

V

;V j

+j(u m

t

(t+t) u m

t

(t); v>j

~

Ct 1=2

kvk

V

+2Kk vk:

Choose v suchthat Kk vk=4and choose Æ=(=2 ~

Ckvk

V )

2

. Then

j(u m

t

(t+t) u m

t

(t);)j:

Thisprovesequicontinuity inC

W

([0;T];H). LetY =B

H (0;K)

H

with the

weak topology, and let F = fu m

t

g C([0;T];H). It follows that F is

equicontinuousinC

W

([0;T];Y)andfu m

t (t):u

m

t

2Fgisrelativelycompact

in Y for each t 2 [0;T]. We then use an application of Arzela-Ascoli in

C

W

([0;T];H) to obtainF relativelycompactinC

W

([0;T];Y). Thus,there

existsasubsequence, denotedfu m

t

g,such that

u m

t

(t)!u

t

(t) weakly inH

(12)

Lemma3.8 There existsan h2L ([0;T];H) such that

^

1 (u

m

x

)!h weakly in L 2

([0;T];H):

Proof. We needtoshowthat^

1 (u

m

x

)isboundeduniformlyinL 2

([0;T];H).

In Lemma 3.6 we computed R

T

0 k ^

1 (u

m

x (s))k

2

ds C(c;C

1 ;C

2 ;M

0

;K ;T):

Thus f ^

1 (u

m

x

)g is bounded in L 2

([0;T];H) independent of m, and there

existsa subsequenceand an h2L 2

([0;T];H) such that

^

1 (u

m

x

)!h weaklyin L 2

([0;T];H):

4 Existence and Uniqueness Theorems

We now state andprovethetheorem regardingexistence anduniqueness of

weak solutions,rstestablishing alocalresult.

Theorem 4.1 Under the assumptions (AF), (Af), and (AL), system

(7)-(11) has a unique local weak solution on [0;t

] for some t

>0.

Proof. Denote by P

M

(M =1;2;:::) the class of functions 2L

T which

can be representedintheform

(t)= M

X

k=1 c

k (t)

k

wherec

k 2C

1

([0;T]). LetP =[ 1

M=1 P

M

. Note P isdense inL

T .

We start with equation (16), multiply by c

k

(t), sum from 1 to M, and

integrateover (0;t) to obtain

Z

t

0 (<u

m

ss

(s);(s)>

V

;V +C

D (u

m

sx (s);

x

(s))+( ^

1 (u

m

x (s));

x

(s)))ds

= Z

t

0

(<F(s);(s)>

V

;V

f(s)(s;R

1 ))ds

or,integratingthe rst termbyparts, we nd

Z

t

0 [ (u

m

s (s);

s

(s))+C

D (u

m

sx (s);

x

(s))+( ^

1 (u

m

x (s));

x

(s))]ds

+(u m

t

(t);(t)) (u m

1

;(0))= Z

t

0

[<F(s);(s)>

V

;V

f(s)(s;R

(13)

M

Now x 2P

M

with M m and pass to the limit as m ! 1, using

Lemma3.3,Lemma3.6,Lemma3.7,Lemma3.8,andtheconvergenceu m

1 !

u

1

inH. Hence,onanyinterval[0;t], withtT,we obtain

Z t 0 [ (u s (s); s

(s))+C

D (u

sx (s);

x

(s))+(h(s));

x

(s))]ds+(u

t

(t);(t))(21)

(u

1

;(0))= Z

t

0

[ <F(s);(s) >

V

;V

f(s)(s;R

1 )]ds:

We nowneed to showthat

Z t 0 (h(s); x (s))ds= Z t 0 ( ^ 1 (u x (s)); x

(s))ds; 82L

T :

This is accomplished by establishing the strong convergence of u m

x (t) !

u

x

(t) in H as m ! 1. To do thiswe take u m

t

and u

t

as test functions in

equations(16)and d

dt

(21)respectively,andadda(u m

x ;u

m

tx )or(u

x ;u

tx )term

to bothsides oftheirrespectiveequations. We have

<u m tt (t);u m t

(t)>+C

D (u m tx (t);u m tx

(t))+( ^

1 (u m x (t));u m tx

(t))+(u m x (t);u m tx (t)) =(u m x (t);u m tx

(t))+<F(t);u m t (t)> V ;V f(t)u m t (t;R 1 ) and <u tt (t);u t

(t)>+C

D (u

tx (t);u

tx

(t))+(h(t);u

tx

(t))+(u

x (t);u tx (t)) =(u x (t);u tx

(t)+<F(t);u

t (t)> V ;V f(t)u t (t;R 1 ): Letz m

(t)=u m

(t) u(t)and subtract thetwo equationsto obtain

1 2 d dt kz m t k 2 +kz m x k 2 +C D kz m tx k 2 = d dt <z m t ;u t > 2C D <z m tx ;u tx >

+<F(t);z m t > V ;V <^ 1 (u x );z m tx

> <^

1 (u

m

x

) h;u

tx > f(t)z m t (R 1 )+<z

m

x ;z

m

tx

> <^

1 (u m x ) ^ 1 (u x );z m tx >:

We then use the Cauchy-Schwartz and Young's inequalityon the last two

termson theright and integrateon (0;t) to obtain

kz m t k 2 +kz m x k 2 ku m 1 u 1 k 2 +ku m 0 x u 0x k 2 +2C 1 D Z t 0 kz m x (s)k 2 ds +2C 1 D Z t 0 k ^ 1 (u m x

(s)) ^

1 (u

x (s))k

2

ds+X

m (t)+Y

(14)

X

m

(t)=2<u m 1 u 1 ;u t

(t)> 2<z m t (t);u t (t)> 4C D Z t 0 <z m sx (s);u sx

(s)>ds+2 Z

t

0

<F(s);z m s (s)> V ;V ds 2 Z t 0 <^ 1 (u x (s));z m sx

(s)>ds 2 Z t 0 <^ 1 (u m x

(s)) h(s);u

sx >ds

and

Y

m (t)=2

Z t 0 jf(s)z m s (s;R 1 )jds:

Note that X

m

(t)! 0 as m ! 1 byLemma 3.3, Lemma 3.7, Lemma 3.8,

and theconvergenceu m

1 !u

1

inH. To see thatY

m

(t)!0 asm!1, we

usetheembeddingV ,!C 0

and Agmon'sinequality(see [11 ])to obtain

Y

m

(t) 2 Z t 0 jf(s)jkz m s (s)k L 1 ds 2ckfk L 2 (0;T) Z t 0 kz m s (s)kkz m s (s)k H 1 ds 1=2 2ckfk L 2 (0;T) kz m t k 1=2 L 2 ([0;T];H) kz m t k 1=2 L 2 ([0;T];V) : Since z m t

! 0 strongly in L 2

([0;T];H) by Lemma 3.6, and since fz m

t g is

boundeduniformlyinL 2

([0;T];V),wehave Y

m

(t)!0 asm!1. Forthe

integraltermontherightside,weuseassumptions(AL)and(AL2)to show

that Z t 0 k ^ 1 (u m x (s)) ^ 1 (u x (s))k 2

ds 4L 2 Tku m 0 x u 0x k 2 +2L 2

(2+T) Z t 0 ku m x (s) u x (s)k 2 ds:

SeetheAppendixfortheproof.

We then have

kz m t k 2 +kz m x k 2 2C 1 D

(1+4L 2 +2L 2 T) Z t 0 kz m x (s)k 2

ds+ku m 1 u 1 k 2

+(1+8C 1 D L 2 T)ku m 0 x u 0 x k 2 +X m (t)+Y

m (t)

We may apply the generalized Gronwall inequality to the above equation

to obtain kz m

x

(t)k ! 0 a.e. t 2 [0;T]. Thus, ^

1 (u m x ) ! ^ 1 (u x

) strongly in

L 2

([0;T];H),and bytheuniquenessof thelimitwehave h=^

1 (u

x

(15)

Z

t

0 [ (u

s (s);

s

(s))+C

D (u

sx (s);

x

(s))+( ^

1 (u

x (s));

x

(s))]ds+(u

t

(t);(t))

(u

1

;(0))= Z

t

0

[<F(s);(s)>

V

;V

+f(s)(s;R

1 )]ds

whichholdsforanyinterval[0;t],fortT,andforall2L

T

. Thussystem

(7)-(11)with

1

replacedby^

1

hasasolutiononanyarbitraryinterval[0;T].

Uniqueness of this weak solution can be shown in the standard way (see,

e.g.,[1 ], [4],[8 ]).

To show thatsystem (7)-(11) with

1

has a local uniqueweak solution,

we notethat theweaksolutionu has thepropertythat u

x

is continuousin

t. Thus,there existsa t

with0t

T such that

ku

x (t) u

0x

k1 forall t2[0;t

];

and therefore,fromthe denitionof ^

1

,we have

^

1 (u

x

(t;x))=

1 (u

x

(t;x)) forallt2[0;t

]:

Henceuisa weaksolutionofsystem(7)-(11) on[0;t

]. Uniquenessis again

shown in the standard way. This completes the proof of local existence

underassumptions(AF), (Af),and (AL).

If we add the growth condition (AG) on

e

(recall that (8) does not

satisfy this condition unless it is modied for large u

x

), then we can use

argumentssimilarto those aboveto establish globalexistence.

Theorem 4.2 Undertheassumptions(AF),(Af),(AL),and(AG),system

(7),(9)-(11) has a unique global weaksolution on any nite interval [0;T].

Proof. Undertheadditionalassumption(AG),wecanargueasinTheorem

2 of [1] to establish that this local unique solution actually exists on any

arbitraryinterval[0,T].Essentially,oneusesthecondition(AG)toestablish

a prioriboundssimilarto(17)forapproximationsinvolving

e ,not^

e ,and

thenarguesapointwiseboundon u

x

(t). ThelocalLipschitzcondition(AL)

can then be used on

e

and arguments similar to those above carried out.

(16)

We argue theinequality Z t 0 k ^ 1 (u m x (s)) ^ 1 (u x (s))k 2

ds 4L 2 Tku m 0 x u 0x k 2 +2L 2

(2+T) Z t 0 ku m x (s) u x (s)k 2 ds:

Usingequation(14) to dene,^ we have

Z t 0 k ^ 1 (u m x (r)) ^ 1 (u x (r))k 2 dr Z t 0 fk^ e (u m x (r)) ^ e (u x (r))k +ke r (^ e (u m 0 x ) ^ e (u 0x ))k +k Z r 0 e (r s) (^ e (u m x

(s)) ^

e (u x (s))dskg 2 dr 4 Z t 0 k^ e (u m x

(r)) ^

e (u

x (r))k

2

dr+4 Z

t

0 k(^

e (u m 0 x ) ^ e (u 0x ))k 2 dr +4 Z t 0 Z r 0 2 e 2(r s) k(^ e (u m x (s)) ^ e (u x (s))kds 2 dr 4L 2 Z t 0 ku m x (r) u x (r)k 2

dr+4L 2 Z t 0 ku m 0 x u 0x k 2 dr +4 Z t 0 Z r 0 e (r s) Lku m x (s) u x (s)kds 2 dr;

where L=L

B

1+ku

0x k

and we have used assumption (AL2) in thelast step.

Now Z t 0 ku m 0 x u 0x k 2

drku m 0 x u 0x k 2 T

fort2[0;T]. Also,usingCauchy-Schwartz, we have

(17)

Z

t

0 k ^

1 (u

m

x

(s)) ^

1 (u

x (s))k

2

ds 4L 2

Tku m

0

x u

0x k

2

+2L 2

(2+T) Z

t

0 ku

m

x

(s) u

x (s)k

2

ds:

5 Acknowledgments

This research was supported in part (H.T.B. and S.W.) by the U.S. Air

Force OÆceof Scientic Research under grant AFOSR F-49620-00-1-00 26.

References

[1] AAckleh,HTBanks,andGAPinter"Well-posednessresultsformodels

of elastomers",CRSCTech.ReportTR00-21, NCStateUniv.

Septem-ber2000; J.Math. Anal.Appl., to appear.

[2] HTBanks, DSGilliam,and VIShubov"Globalsolvabilityfordamped

abstract nonlinearhyperbolicsystems",DierentialIntegral Equations

10, (1997) 309{332.

[3] HTBanks,GAPinter,andLKPotter"Modelingofnonlinearhysteresis

in elastomersunder uniaxialtension",J. IntelligentMat. Systems and

Structures 10, (1999) 116{134.

[4] HT Banks,GA Pinter, andLKPotter "Existenceofuniqueweak

solu-tions to a dynamical systemfornonlinear elastomerswithhysteresis",

CRSCTech.ReportTR98-43NC StateUniv.,November1998;

Dier-ential and Integral Equations 13, (2000) 1001{1024.

[5] HT Banks, GAPinter, LK Potter, MJ Gaitens, and LC Yanyo

"Mod-eling of quasi-static and dynamic load responses of lled viscoelastic

materials", in Modeling: Case Studies from Industry, E.Cumberbatch

and A.Fitt, eds., CambridgeUniversityPress,to appear.

[6] HT Banks, JH Barnes, A Eberhardt, H Tran, and S Wynne,

"Mod-eling and computation of propagating waves from coronary stenosis",

CRSC Tech. Report TR00-20, NC State Univ., August 2000; Comp.

(18)

Chicago Press, Chicago,(1988).

[8] R.DautrayandJLLions,Mathematical Analysis andNumerical

Meth-ods for Science and Technology, Volume 5, Evolution Problems I,

Springer,New York,2000.

[9] YC Fung, Biomechanics: Mechanical Properties of Living Tissues

Springer-Verlag,New York,(1993).

[10] AW Naylor and GR Sell, Linear Operator Theory in Engineering and

ScienceSpringer-Verlag, NewYork,(1982).

[11] R.Temam, Innite-DimensionalDynamical SystemsinMechanicsand

Physics, Applied Mathematical Sciences, 68, Springer-Verlag, New

References

Related documents

Considering that technical expertise was highly appreciated in other contexts, such as Bangladeshi Big-4 firms (Spence et al., 2016), this finding highlights a need to investigate

In this chapter, we discuss the enhancements and new features of the latest release of the Microsoft database management system, SQL Server 2008, as well as the reasons that the

“Faktor yang paling esensial di dalam proses pendidikan adalah manusia yang ditugasi dengan pekerjaan untuk menghasilkan perubahan yang telah direncanakan pada anak

Buna göre, sigara içme seviyesi azaldıkça birleşik sağlık uyarıları daha fazla katılımcı tarafından etkili olarak değerlendirilmiştir.. Ancak, sigara içme

3P‑MACE: 3‑point major adverse cardiovascular events; 4P‑MACE: 4‑point major adverse cardiovascular events; ACE: angiotensin‑converting enzyme; ARB: angiotensin‑receptor

The principal strategy of HySA is to execute research and development (R&amp;D) work, with the main aim of achieving an ambitious 25% share of the global hydrogen and fuel

In order to further determine whether knockdown of Par3 inhibits PCa metastasis by activating the Hippo pathway and attenuating the nuclear translocation of non-phosphorylated YAP as