• Aucun résultat trouvé

Axisymmetric eddy current inspection of highly conducting thin layers via asymptotic models

N/A
N/A
Protected

Academic year: 2021

Partager "Axisymmetric eddy current inspection of highly conducting thin layers via asymptotic models"

Copied!
21
0
0

Texte intégral

(1)

HAL Id: hal-01214308

https://hal.inria.fr/hal-01214308

Submitted on 11 Oct 2015

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Axisymmetric eddy current inspection of highly conducting thin layers via asymptotic models

Houssem Haddar, Zixian Jiang

To cite this version:

Houssem Haddar, Zixian Jiang. Axisymmetric eddy current inspection of highly conducting thin layers

via asymptotic models. Inverse Problems, IOP Publishing, 2015. �hal-01214308�

(2)

onduting thin layers via asymptoti models

Houssem Haddar

, ZixianJiang

Abstrat

Thin opper deposits overing the steam generator tubes an blind eddy urrent probes in non-

destrutivetestingsofproblematifaultsandarethereforeimportanttobeidentied.Existingmethods

basedon shape reonstrutionusing eddy urrent signalsenounter diulties ofhigh numerialosts

duetothelayer'ssmallthiknessandhighondutivity.Inthisartile,weapproximatetheaxisymmetri

eddyurrent problemwithsomeappropriateasymptotimodelsusingeetive transmissiononditions

representingthe thindeposits. Inthese models, the geometrial information related to the deposit is

transformedintoparameteroeientsonatitiousinterfae. Standarditerativeinversionalgorithmis

thenappliedtotheasymptotimodelstoreonstrutthethiknessofthethinopperlayers. Numerial

testsbothvalidatingtheasymptotimodelandshowingbenetoftheinversionproedureareprovided.

Keywords: axisymmetrieddyurrent inspetion,asymptotimodel, thiknessreonstrution ofthin

layers.

1 Introdution

Thinlayersofopperdepositsareobservedinnon-destrutiveeddyurrenttestingofsteamgeneratortubes

forthe safetyand failure-free operatingof nulearpowerplants. Coveringthe shell sideof thetube, these

deposits lead to a signiant signal feedbak in the eddy urrent inspetion due to the high ondutivity,

despite their extremely small thikness (see Table 1for a omparison with the tube). These deposits do

notdiretlyeet theprodutivityof steamnor thesafetyof struture. Howevertheirpresene may mask

otherkindsofproblematifaultssuhasloggingmagnetitedeposits, raksintubes,et. Thisis whyitis

importantto identifythem.

tubewall opperlayer

ondutivity(in

S · m

−1)

σ t = 0.97 × 10 6 σ c = 58.0 × 10 6

thikness(in

mm

)

r t

2

− r t

1

= 1.27 0 ≤ f δ (z) ≤ 0.1

Table1: Condutivityandsaledierenesbetweentubewallandopperlayer.

Inpreviousstudies, we modeledthe axisymmetrieddy urrentproblem withan eientnite element

approximationwhih involvesartiial boundaryonditionsto ut o theomputational domainand thus

redue thenumerial ost(see [10℄). Basedon this model isdevelopedan inversionalgorithmusing shape

optimization methods to reonstrut the shape of some logging magnetite deposits [12℄. These methods

anbenaturally applied to thiknessreonstrution of thin opperdeposits. However, amajor numerial

hallenge to dealwith this aseis thehigh numerialost resulting from thefat that the omputational

domainshouldbedisretizedintoanemeshwiththesamesaletothelayerthiknessoftheopperdeposits.

Besidesadaptivemeshreningmethod(seeforexample[7℄),anotherwidelypratiedstrategytoredue

omputational ost onsists in modeling the thin deposit with eetive transmission onditions (see for

example[19℄, [21℄). [20℄ in partiular suppliesanalysis andomparison of dierent impedanetransmission

onditionsforthinsheetsintwodimensionaleddyurrentmodel. Thesetransmissiononditionsareobtained

bymathingtheasymptotiexpansionsofthesolutionwithrespetto asmallparameterharaterizingthe

houssem.haddarinria.fr,DÉFIteam,INRIASalay,Palaiseau,Frane.

zixian.jiangpolytehnique.edu,DÉFIteam,INRIASalay,Palaiseau,Frane.

(3)

perspetive,anadditionalmajoradvantageofasymptotimodel istoavoidre-meshingat eahiteration. A

rih literature on asymptoti models has beendeveloped for dierent approahes and various appliations

(seeforexample[3℄,[2℄,[6℄,[5℄,[4℄, [18℄,[13℄,[14℄, [22℄andthereferenestherein).

In the researh note [9℄ the authors built and ompared several asymptoti models using a family of

eetive transmission onditions

{Z m,n } m,n∈{0,1,2}

for the axisymmetri eddy urrent inspetion of thin opper deposits. Roughly speaking, if the thikness of a thin deposit at vertial position

z

of the tube

writes

f δ (z) = δd(z)

, then theindex

m

of thetransmission ondition

Z m,n

denotes are-saling parameter

of the opper ondutivity with regard to

δ

:

σ c = σ m /(δ m )

, while the index

n

stands for the order of

asymptotiexpansionofthesolutionwith respetto

δ

. Numerialtests thereinonaredued1Dasewith

onstant deposit thikness showthat theasymptoti modelswith

Z 1,n

(

n = 0, 1

)onditionsgivesatisfying

approximationofthefullmodelandleadto easyonstrutionofinversionmethods.

Inthis artilewerst introduetheaxisymmetri modelforeddy urrenttesting andsomeusefulteh-

niquestodeveloptheasymptotimodels(Setion2). Usingthesamemethodpresentedin[9℄,wethenbuild

thevariationalasymptoti modelswith

Z 1,n

(

n = 0, 1

)transmissiononditionsfor generalaseswith vari- abledeposit thikness(Setion3). Byintroduing adjointstatesassoiatedtoderivativesoftheimpedane

measurements,weformulate theinverseproblem forthiknessreonstrutionastheminimizationof aleast

square ost funtional of layerthikness by gradient desent (Setions 4). Finally, numerialexamples of

modelvalidation,measurementapproximationandthiknessreonstrutionaregiveninSetion(5). Forthe

useofasymptotimodelsininverseproblemswemayite[8℄,[15℄and[16℄forvariousappliations.

2 Asymptoti approximation of axisymmetri eddy urrent model

This setion is devoted to a formal derivation of asymptoti models for eddy urrent problems with the

preseneofhighlyonduting thindeposits. Theobjetiveistogettheeetivetransmissiononditionson

theinterfaebetweenthethinlayerandthetubewithwhihthevariationalasymptotimodelhasnolonger

thevolumeintegralonthethinlayerdomain.

First let us briey introdue the axisymmetri eddy urrent problem. For more details readers may

refer to [10℄. In the ylindrial oordinates, a vetor eld

a

an be deomposed into the meridian part

a m = a r e r + a z e z

and the azimuthal part

a θ = a θ e θ

.

a

is axisymmetri if

∂ θ a

vanishes. Under the assumption of axisymmetry and the low eletri permittivity / low frequeny regime (

ωǫ ≪ σ

), the 3-D

time-harmoniMaxwell'sequationsfortheeletriandmagnetields

( E , H ) ( curl H + (iωǫ − σ) E = J

in

R 3 ,

curl E − iωµ H = 0

in

R 3 ,

withadivergene-freeaxisymmetriappliedsoure

J

(

div J = 0

)anbereduedtoaseondorderequation

ona2-Ddomain

R 2 + := { (r, z) : r ≥ 0, z ∈ R}

fortheazimuthalpartoftheeletrield

E θ

that wedenote

inthesequelby

u = E θ

:

− div 1

µr ∇ (ru)

− iωσu = iωJ θ = iωJ

in

R 2 + ,

(1)

where

∇ := (∂ r , ∂ z ) t

and

div := ∇·

are gradient and divergene operators in 2-D Cartesian oordinates.

Assume that

J ∈ L 2 (R 2 + )

has ompatsupport, and that

µ

and

σ

arein

L

(R 2 + )

suh that

µ ≥ µ v > 0

,

σ ≥ 0

and that

µ = µ v

,

σ = 0

for

r ≥ r 0

suiently large. Then problem (1) with a deay ondition at

innity(

u → 0

as

r 2 + z 2 → 0

)hasauniquesolutionin

H (R 2 + )

whereforany

Ω ⊂ R 2 +

wedenote

H(Ω) := n

v : r

1

/

2

(1 + r 2 )

λ

/

2

v ∈ L 2 (Ω), r

1

/

2

∇ (rv) ∈ L 2 (Ω) o

with

λ

anyreal

> 1

(see[10℄). Letusindiatethatif

isboundedinthe

r

-diretionthen

H (Ω)

isequivalent

tothefollowingspaeforwhihweshallusethesamenotation

H(Ω) := n

v : r

1

/

2

v ∈ L 2 (Ω), r

1

/

2

∇ (rv) ∈ L 2 (Ω) o

.

(4)

a(u, v) :=

Z

1

µr ∇ (ru) · ∇ (r¯ v) − iωσrw¯ v

dr dz = Z

iωJ ¯ vr dr dz ∀ v ∈ H (Ω).

(2)

Fornumerial reasons, the omputational domain will be trunated in radialdiretion at

r = r

with

r

> 0

suientlylarge,andwesetaNeumannboundaryonditionon

r = r

. Thesolutiontothetrunated

problemshouldsatisfy(1)on

Ω = B r

:= { (r, z) ∈ R 2 : 0 ≤ r ≤ r

}

. Thisiswhyweshalluseinthesequelthe

generinotationforthevariationalspae

H (Ω)

with

denoting

R 2 +

or

B r

. Wealsoreallthattheproblem

on

Ω = B r

anbeequivalentlytrunated toaboundeddomain

B r

,z

= { (r, z) ∈ R 2 : 0 ≤ r ≤ r

, | z | < z

}

byintroduingappropriateDirihlet-to-Neumannoperatorsontheboundaries

Γ

±

:= { (r, z) : 0 ≤ r ≤ r

, z =

± z

}

. This would be onvenient for aelerating numerial evaluation of the solution. The analysis and

examplesofdomaintrunationusingNeumannandDirihlet-to-Neumannboundaryonditionsanbefound

in[10℄. Weremarkthatthederivationoftheasymptotimodelsandtheinversionmethodsinthesequelare

independentof the domain trunation strategy. This is whyweuse the variationalformulation(2) as our

startingpointand thenweapplydomaintrunationinthenumerialexamples.

Finally letusspeifytheproblem settingsof eddy urrenttestingfor thin layerdeposit. Weonsider a

thinopperdepositoveringaxisymmetriallytheshell sideofthetube(seeFigure1foraradialutofthe

setting in theylindrial oordinate system, the

z

-axis is the tube axis, the thin deposit shown in blue is

exaggeratedin thikness). The eddy urrent probeis omposed oftwooaxialoils, whose radialutsare

representedbytwosmallretanglesinFigure 1,that movein the

z

-diretionduring asan.

Onthedomainofproblem

,wedenote

±

:= { (r, z) ∈ Ω : r ≷ r t

2

}

. Theappliedurrent

J

issupported

bytheoils,therefore

suppJ ⊂ Ω

. Thedepositlayerisdepitedby

δ c ⊂ Ω +

withavariablethikness

f δ (z)

(exaggeratedinthikness,showninblue). Wedenoteby

u δ

±theeletrieldsoutsidethedepositlayer,with

u δ

on

and

u δ +

on

Ω + \ Ω δ c

, andby

u δ c

thein-layereletrield, i.e. on

δ c

.

Tube Deposit

z

r

Coils

uδ− uδc uδ+

fδ(z) rt1 rt2

Γt2 Γc Γt1

Ωδ c

Ω− Ω+

Figure1: Representationof theeddyurrenttestingofathinlayerdeposit.

2.1 Resaled in-layer eddy urrent equation

Therststep isto rewritethein-layereddyurrentequationby resalingtheoordinatein thetransverse

diretion and theondutivity with respet to thelayerthikness. Theanalytialsolution of this resaled

equation allowsto get arelation between the boundary valuesof Dirihlet andNeumann type onthe two

longitudinalboundariesofthethinlayer.

Weassume thatthe thikness

f δ (z)

writes

f δ (z) = δd(z)

, with

δ > 0

asmall parameterand

d(z) > 0

a

dimensionlessquantity. Asweareonernedonlywiththe

Z 1,n

(

n = 0, 1

)transmissiononditions(see[9℄), weresalethedepositondutivityby

σ c = σ 1 /δ

,wheretheresaledondutivity

σ 1

isonsiderablysmaller than

σ c

. Weresalealsotheoordinateinthetransversediretionofthedepositsbythehangeofvariable

ρ = (r − r t

2

)/δ

with

ρ ∈ [0, d(z)]

,and wedenote by

u(ρ, z) := ˜ u δ c (r t

2

+ δρ, z)

the resaledin-layersolution.

Let

k 1

besuhthat

(k 1 ) 2 = iωµ c σ 1

. From(1) andtheresalingsettings,weobtain

ρ 2 u ˜ = − δ B 1 1 u ˜ − δ 2 B 1 2 u ˜ − δ 3 B 3 1 u ˜ − δ 4 B 1 4 u, ˜

(3)

(5)

B 1 1 = 2ρ r t

2

ρ 2 + 1 r t

2

∂ ρ + k 1 2 , B 1 2 = ρ 2

r 2 t

2

ρ 2 + ρ

r 2 t

2

∂ ρ − 1

r 2 t

2

+ ∂ z 2 + 2ρ r t

2

k 2 1 , B 3 1 = 2ρ

r t

2

z 2 + ρ 2

r t 2

2

k 2 1 , B 1 4 = ρ 2 r 2 t

2

z 2 .

2.2 Transmission onditions between in-layer eld and elds outside the layer

Thetransmissiononditionsbetweentheeldinside thetube

u δ

andthein-layereld

u δ c

are

u δ

= u δ c

and

µ

−1

t ∂ r (ru δ

) = µ

−1

c ∂ r (ru δ c )

on

Γ t

2

.

(4)

Thetransmissiononditionsbetween

u δ c

andtheeldoutsidethedeposit layer

u δ +

write

u δ + = u δ c

and

µ

−1

v ∂ n (ru δ + ) = µ

−1

c ∂ n (ru δ c )

on

Γ c .

(5)

Withtheunit normalandtangentialvetorson

Γ c

at thepoint

(r t

2

+ δd(z), z) n = (1, − δd

(z))

p 1 + (δd

(z)) 2

and

τ = (δd

(z), 1) p 1 + (δd

(z)) 2 ,

werewrite(5) usingvetordeompositiononthesediretions

( n , τ ) u δ c = u δ +

and

∂ r (ru δ ) = µ c /µ v + (δd

) 2

1 + (δd

) 2 ∂ r (ru δ + ) + (1 − µ c /µ v ) δd

1 + (δd

) 2 ∂ z (ru δ + )

on

Γ c .

(6)

Togetherwiththepartialdierentialequationforthein-layereld(3), thesetransmissiononditionson

bothsidesofthethin layerasboundaryonditionswillservetoestablishCauhyproblems whosesolutions

linktheeldsonbothsidesofthethin layer,thatisarelationshipbetween

u δ

on

Γ t

2 and

u δ +

on

Γ c

. Toget

theeetivetransmissiononditionson

Γ t

2, wefurtherapplyaTaylorexpansionof

u δ +

.

2.3 Extension of

u δ +

using Taylor's expansions

Togeteetivetransmissiononditionslinkingup

u δ

±ontheinterfae

Γ t

2 betweentubeandthindeposit,we

shouldextendtheeld

u δ +

,originallydened onlyon

Ω + \ Ω c

,tothewholedomain

Ω +

. Anaturalapproah

would be to assumethat the extendedeld, still denoted by

u δ +

, satises the sameeddy urrentequation

(1)as

u δ +

on

Ω + \ Ω c

, i.e. with oeientsof thevauum

µ = µ v

,

σ = σ v = 0

. We remark that theright

hand side of equation(1) vanishesin

Ω +

sine

J

hassupport onlyin

. Usingthe variable substitution

ν = r − r t

2,onerewrites(1)in

Ω +

as

X 4 j=0

ν j A j (ν∂ ν , ∂ z ) u δ + = 0,

(7)

where

A 0 (ν∂ ν , ∂ z ) = (ν∂ ν ) 2 − ν∂ ν , A 1 (ν∂ ν , ∂ z ) = 2 r t

2

(ν∂ ν ) 2 − 1 r t

2

ν∂ ν , A 2 (ν∂ ν , ∂ z ) = 1

r 2 t

2

(ν∂ ν ) 2 − 1

r t 2

2

+ ∂ z 2 , A 3 (ν∂ ν , ∂ z ) = 2 r t

2

z 2 , A 4 (ν∂ ν , ∂ z ) = 1 r 2 t

2

z 2 .

Theasymptotiexpansionof

u δ +

withrespetto

δ

isintheform

u δ + (r, z) = P

n=0 δ n u + n (r, z)

. Obviouslyeah

term

u + n (r, z)

veriesthesameequation(7). With Taylorseriesexpansion,onehas

u + n (r t

2

+ ν, z) =

X

k=0

ν k u + n,k (z)

where

u + n,k (z) = 1

k! ∂ ν k u + n

(r t

2

, z).

(6)

Sine

ν∂ ν (ν k u + n,k (z)) = k(ν k u + n,k (z))

, we an indeed write

A i (ν∂ ν , ∂ z )

as

A i (k, ∂ z )

while it is applied to

k u + n,k (z))

. Thus,from(7)onehas

P 4

j=0

P

k=0 A j (k, ∂ z )(ν k+j u + n,k ) = 0.

Theequalityatorder

O (ν k )

gives

A 0 (k, ∂ z ) u + n,k = −

X 4 j=1

A j (k − j, ∂ z ) u + n,k−j ,

with

u + n,−1 = u + n,−2 = u + n,−3 = u + n,−4 = 0

. Nowweonsider

A 0 (k, ∂ z ) = k 2 − k

. For

k ≥ 2

,

A 0 (k, ∂ z ) 6 = 0

,

thus invertiblewithitsinverse

A

−1

0 (k, ∂ z ) = k

2

1

−k. Sowehave

u + n,k = −A

−1

0 (k, ∂ z )

 X 4 j=1

A j (k − j, ∂ z ) u + n,k−j

 , k ≥ 2.

(8)

Nowweindutivelydenetwofamiliesofoperators

{S k 0 (∂ z ) , S k 1 (∂ z ) }

:

S 0 0 := Id, S 0 1 := 0, S 1 0 := 0, S 1 1 := Id,

k ≥ 2

 

 

 

 

 

 

 

S k 0 := −A

−1

0 (k, ∂ z )

 X 4 j=1

A j (k − j, ∂ z ) S k−j 0 (∂ z )

 ,

S k 1 := −A

−1

0 (k, ∂ z )

 X 4 j=1

A j (k − j, ∂ z ) S k−j 1 (∂ z )

 .

(9)

From the indution (8) one gets

u + n,k (z) = S k 0 (∂ z ) u + n (r t

2

, z) + S k 1 (∂ z ) ∂ r u + n (r t

2

, z)

. Therefore we havethe

followingexpansion

 

 

 

 

 

u + n (r t

2

+ ν, z) = X

k=0

ν k

S k 0 (∂ z ) u + n + S k 1 (∂ z ) ∂ r u + n

(r t

2

, z),

∂ r u + n (r t

2

+ ν, z) = X

k=0

ν k (k + 1)

S k+1 0 (∂ z ) u + n + S k+1 1 (∂ z ) ∂ r u + n

(r t

2

, z).

Wealsodene theoperators

S e k 0 := S k 0 − r 1

t

2

S k 1

and

S e k 1 := r 1

t

2

S k 1

. ThentheaboveTaylorexpansionswrite

 

 

 

 

 

u + n (r t

2

+ ν, z) = X

k=0

ν k

S e k 0 (∂ z ) u + n + S e k 1 ∂ r (ru + n )

(r t

2

, z),

∂ r (ru + n )(r t

2

+ ν, z) = X

k=0

ν k (k + 1)

(r t

2

S e k+1 0 + S e k 0 )u + n + (r t

2

S e k+1 1 + S e k 1 )∂ r (ru + n )

(r t

2

, z).

(10)

Althoughonly asymptotiproblems oforder

0

and

1

aredisussed in thesequel, theaboveexpressionsfor anyordersouldfailitatefurtherexploitationsbeyondthesopeoftheurrentwork.

3 Asymptoti models for deposits with variable layer thikness

Inthissetion,webuild theasymptotimodelsoforder

0

and

1

forthin deposits ofvariable thikness,that

is to establishthe transmissiononditions

Z 1,n

for

n = 0, 1

basedon the materialof the previoussetion.

ThegeneralproedureonsistsinsolvinganalytiallyaCauhyproblemoftheresaledin-layereddyurrent

equationin itsasymptoti expansionform(3)withboundaryonditionson

Γ t

2 (4). The resultingsolution

yieldstheDirihletandNeumannboundaryvalueson

Γ c

whihshould mathwith transmissiononditions (6). FinallywiththeTaylorexpansionsof

u δ +

(10) ,onegetstheeetivetransmissiononditionson

Γ t

2.

Tosimplifythepresentation,espeiallytheomplexitiesintroduedbythetransmissiononditions(6)on

theurvedboundary

Γ c

,weassumethatthemagnetipermeabilityofthedepositsequalstothatofvauum

µ c = µ v

. Thisassumptionmathesthepratialappliation wehaveinmindsinetherelativepermeability ofopperis

1

.

(7)

Weintroduethenotationforjumpandmeanvalueson

Γ t

2

[v] := v + | r

t2

− v

| r

t2

, [µ

−1

∂ r (rv)] = µ

−1

v ∂ r (rv + ) | r

t2

− µ

−1

t ∂ r (rv

) | r

t2

, h v i := 1

2 v + | r

t2

+ v

| r

t2

h µ

−1

∂ r (rv) i = 1 2

µ

−1

v ∂ r (rv + ) | r

t2

+ µ

−1

t ∂ r (rv

) | r

t2

.

3.1 Formal derivation of

Z 1 , 0

transmission onditions We reall the asymptoti expansion

u δ + = P

n=0 δ n u + n

and expand

u(ρ, z) = ˜ P

n=0 δ n u n (ρ, z)

and

u δ

= P

n=0 δ n u

n

. Theeddyurrentequation (3)andthetransmissiononditions(4)on

Γ t

2 fortherstterm

u 0

yieldtheCauhyproblem

( ∂ ρ 2 u 0 = 0 ρ ∈ [0, d(z)], u 0 | ρ=0 = u

0 | r

t2

, ∂ ρ u 0 | ρ=0 = 0.

This problem hasaonstantsolution

u 0 (ρ, z) = u

0 | r

t2

for

ρ ∈ [0, d(z)]

. Consideringitsvalue at

ρ = d(z)

andthersttransmissiononditionof (6)for

u 0

on

Γ c

,weget

u

0 | r

t2

= u + 0 | r

t2

.

(11)

TheCauhyproblemfor

u 1

withinitialvaluesgivenby(4)is

 

 

 

 

ρ 2 u 1 = −B 1 1 u 0 = − k 2 1 u 0 | r

t2

ρ ∈ [0, d(z)], u 1 | ρ=0 = u

1 | r

t2

,

∂ ρ u 1 | ρ=0 = − 1 r t

2

u

0 | r

t2

+ 1 r t

2

µ c

µ t ∂ r (ru

0 ) | r

t2

.

Itfollowsthat

∂ ρ u 1 = − 1 r t

2

u

0 | r

t2

− µ c

µ t

∂ r (ru

0 ) | r

t2

− ρk 2 1 u

0 | r

t2

,

(12)

u 1 = u

1 | r

t2

− ρ

r t

2

u

0 | r

t2

− µ c

µ t

∂ r (ru

0 ) | r

t2

− ρ 2

2 k 2 1 u

0 | r

t2

.

(13)

Theseond transmissiononditionof (6)for

u 1

on

Γ c

implies

∂ ρ u 1 | ρ=d(z) = − 1 r t

2

u + 0 | r

t2

− µ c

µ v ∂ r (ru + 0 ) | r

t2

.

(14)

Mathing(14)and(12)at

ρ = d(z)

gives

1

r t

2

µ c

µ t

∂ r (ru

0 ) | r

t2

− k 2 1 d(z)u

0 | r

t2

= 1 r t

2

µ c

µ v

∂ r (ru + 0 ) | r

t2

.

(15)

Equalities(11),(15)andthefat that

k 2 1 = iωσ 1 µ c

implythat

[u 0 ] = 0

and

−1

∂ r (ru 0 )] = − iγ 1 h u 0 i ,

(16)

where

γ 1 = ωσ 1 d(z)r t

2

= ωσ c f δ (z)r t

2

.

3.2 Asymptoti model of order

0

If

u δ := u δ

± in

± denote anapproximationoftheexatsolutionupto

O(δ)

error,thenfrom (16),possible

transmissiononditionsanset as

[u δ ] = 0

and

−1

∂ r (ru δ )] = − iγ 1 h u δ i

(17)

Références

Documents relatifs

For the hole in the inconel slab, the agreement between the model and the experimental data is better than 6% for the amplitude and 2 ◦ in phase (Figure 7.c) whereas, for the hole

Parmi les systèmes de contrôle des émissions des moteurs diesel, la plupart des recherches et des études ont été réalisées sur la réduction des émissions des

Experimental results show that fine nanocrystalline alloy powders prepared by mechanical milling are very promising for microwave applications and it is suggested that Eddy

Effective models for non-perfectly conducting thin coaxial cables Geoffrey Beck 1,∗ , Sébastien Imperiale 2 , Patrick Joly 1. 1 POEMS (CNRS-INRIA-ENSTA Paristech),

transmission eigenvalues, asymptotic expansions, thin layers, inverse scattering problems 8.. AMS

One observes that for a given re-saling parameter m , the asymptoti models approximate better the full model as the asymptoti expansion order n inreases.. the

Again, one clearly observes that using the DtN maps for the horizontal truncation introduces a reasonably small error compared to the reference solution u exact while truncating

Monique Dauge, Patrick Dular, Laurent Krähenbühl, Victor Péron, Ronan Perrussel, Clair Poignard.. To cite