• Aucun résultat trouvé

The motion of the shell is describ ed by the classicaltheory of thins shells

N/A
N/A
Protected

Academic year: 2022

Partager "The motion of the shell is describ ed by the classicaltheory of thins shells"

Copied!
9
0
0

Texte intégral

(1)
(2)

In this pap er we investigate the inuence of p erio dicwater waves on the deection of a

thin exibleshell,with large radius, mounted at the sea b ottom at depth h and piercing

through the free surface. The uid is assumed to b e inviscidand rotationless, hence the

velo city p otential ob eys the Laplace equation. The motion of the shell is describ ed by

the classicaltheory of thins shells. The equationsare then reducedto simplied"two di-

mensional" equationsinvolvingthe normal,tangential andlongitudinal motionu;v;wof

the middlesurface. This theory requiresthat the shellthicknessdis smallincomparison

withshellradiusR,andthewaterdepthh. Theotherclassicalassumptionscanb efound

in Markus [1]. We will fo cus on the p erio dic deections only and we pay no attention

to thestaticdeectionduetohydrostaticpressure. Bylinearity,thiseectscanb eadded.

The harmonic water waves diracted by a rigid cylinder will b e describ ed by means of

the wellknowneigenmo deexpansion. These mo desare solutions ofthe Laplace equation

and t the b oundary conditions at the free surface and the sea b ottom. To solve the

coupledequationsforthe refractedp otentialand the motionofthe shellweuse the same

set of orthogonal functions. Due to the b oundary conditions of the variables of motion

of the shell this can not lead to a converging series. To assure convergence we expand

these highest derivatives, as they o ccur in the equations of motions, in series of these

eigenmo des. The lower order derivativesand the functions themselvesfollowfrom these

expansions. Thisgivesrisetoasetofp olynomialswith unknownco eÆcients. Makinguse

of the orthogonality relationsof the eigenmo des andthe expansionof the p olynomials in

these mo des leads to a set of equations with a sparse blo ck, structured by the orthogo-

nality relation, combined with full rows due to the expansions of the p olynomials. The

b oundary conditions of the variables describing the motion of the shellgive an extra set

of equations. The nal 'square' matrixequationcan b e solvedby astandard metho d.

In section 6, an application of this mathematical approach is treated for the case of a

exibleb eamloadedbyadistributiondescrib edbytherst eigenmo deofthe water-wave

problem. Convergenceof the deection and itsderivativesis shown. In the next section,

the mathematicalmo delfor the shellis givenand the metho dis tested forseveral values

of the exural rigidity of the shell.

2 Derivation of the main equations

Westudy the b ehaviorof the thin cylindricalshellin p erio dicwaves in asea of constant

depth h. The shell,emergingfromthe sea surface isxed on the sea b ottom.

We assume the uid to b e p otential and intro duce the velo city p otentialV = r(x;t)

whereVistheuidvelo cityvector. Wegetforthep otential( x;t)theLaplaceequation

=0 inthe uid domain.

The linearized freesurface z =0, the linearized condition g

z +

tt

=0 , at the b ottom

we have

@

@z

= 0 and for the linearizedb o dy b oundary condition

@

@n

= u where u is the

normaldeection ofthe shell.

(3)

W

U V

Following Love'stheory, we derivethe equations of motion of the shell.

R

@u

@z +

R

2

(1+)

@ 2

v

@z@

+R 2

@ 2

w

@z 2

+ 1

2

(1 )

@ 2

w

@ 2

0

E

(1

2

)R 2

@ 2

w

@t 2

=0 (1)

@u

@ +

@ 2

v

@ 2

+ R

2

2

(1 )

@ 2

v

@z 2

0

E

(1

2

)R 2

@ 2

v

@t 2

+ R

2

(1+)

@ 2

w

@z@

=0 (2)

R 4

@ 4

u

@z 4

+2R 2

@ 4

u

@z 2

@ 2

+

@ 4

u

@ 4

u

0

E

(1

2

)R 2

@ 2

u

@t 2

+

@v

@

(3)

+R

@w

@z

=

(1

2

)R 2

Ed

P(r;z; ;t)

with = 1

12 (

d

R )

2

and P the dynamicpressuredue to water waves.

P(r;z; ;t)=

@

@t

(r; ;z;t) (4)

By dierentiatingequation(1), one can obtain.

R

@ 2

u

@z 2

+ R

2

(1+)

@ 3

v

@z 2

@ +R

2

@ 3

w

@z 3

+ 1

2

(1 )

@ 3

w

@z@

2

0

E

(1

2

)R 2

@ 3

w

@z@t 2

=0 (5)

The dierentiationwillallowusto expand later this equationin the sameeigenfunctions

than the two others shell equations. If we cho ose equation [1] to b e zero at z = h it is

automatically fullled everywhere.

Wesplit the p otentialand the pressure into twoparts.

The rst ones ( d

and p d

) are due to the diraction when the shell is rigid. The second

ones ( and p) are due to the exural deections (radiation).

with

P(r; ;z;t)=(p d

(r; ;z)+p(r; ;z))e i! t

(6)

(r; ;z;t)=( d

+)e i! t

(7)

(4)

p d

(r=R)= g

cosh(kh)

h+ 1

sinh 2

(kh)

2

1=2

f

0 (z)

X

n=0 2

n (i)

n+1

kRH 0

n (kR)

cos (n )

= 1

X

n=0 A

n

cos(n ) f

0 (z)

3 Expansion of the solutions in eigenfunctions

Werst expandthe p otential in the usual eigenfunctionsf

m .

(r; ;z;t)= 1

X

n=0 1

X

m=1 X

mn K

1

n (k

m

r )cos (n )f

m (z)e

i! t

(8)

+ 1

X

n=0 X

0n H

1

n

(kr )cos (n )f

0 (z)e

i! t

(9)

Instead of following the standard approachfor Sturm-Liouvilleb oundary value problems

where expansions in the eigen-function space of the op erator is sought, we assume that

the fourth derivative of u, the second derivative of v and third derivative of w can b e

expanded in the sameclass of eigenfunctions. Then, if we integrate those functions and

nd for u;v and w

u(z; ;t)= 1

X

n=0

1

X

m=0 u

mn

k 4

m f

m (z)+

1

6 C

1n

(z+h) 3

+C

2n 1

2

(z+h) 2

+C

3n

(z+h)+C

4n

cos(n )e i! t

(10)

v(z; ;t)= 1

X

n=1

v

0n

k 2

f

0 (z)

1

X

m=1 v

mn

k 2

m f

i

(z)+C

8n

(z+h)+C

9n

sin(n )e i! t

(11)

w (z; ;t)= 1

X

n=0

w

0n

k 3

g

0 (z)

1

X

m=1 w

mn

k 3

m g

m (z)+

1

2 C

5n

(z+h) 2

+C

6n

(z+h)

+C

7n

cos (n )e i! t

(12)

with

f

0 (z)=

p

2 coshk(z+h)

h+ 1

sinh 2

(kh)

1=2 g

0 (z)=

p

2sinhk(z+h)

h+ 1

sinh 2

(kh)

1=2

(13)

with k givenbythe usual disp ersionrelation =ktanh (kh)

(5)

f

n (z)=

2 cosk

n

(z+h)

h

1

sin 2

(k

n h)

1=2 g

n (z)=

2 sink

n

(z+h)

h

1

sin 2

(k

n h)

1=2

(14)

k

n

givenby = k

n tan (k

n h)

The f

n

functions are orthonormaland formacompleteset ofeigenfunctions forthe solu-

tion .

In order to guarantee the convergence of the termwize integrated series to the integral

of the expanded function, we haveto assumethat the series of the successive derivatives

converge uniformly. This requirement can b e to o strong and is not necessary as can b e

shown by a slightly dierent approach. However, numerical results conrm the conver-

gence of the fourth derivative of u although slowly. The convergence of the series for u

is then guaranteed and fast. The eigenfunctions f

n

for are not eigenfunctions of the

mechanicalequations. It turns out that the p olynomials are convenientto b e added, the

found solution liesin theprop er function space. It fullsthe b oundary conditions and in

a weak sense the equations. It is reasonable to exp ect somelo cal convergence problems

for the series of the highest derivatives.

4 Boundary conditions

The b oundary condition on the cylinder for the p otential in frequency domain is as

follow:

@

@n

= i!u (15)

Theb oundaryconditionattheb ottomofthe uiddomainalsorepresentsthattheb ound-

ary isa rigid wall,therefore

@

@z

=0 at z = h (16)

The mechanicalb oundary conditions at the ends of the shell are then derivedbyconsid-

ering that the shell is xed at z = h but free at z = 0. Following Markus, the eight

extraequations read:

u=v =w=

@u

@z

=0 forz = h (17)

@w

@z +

R

@v

@ u

=0 (18)

@ 2

u

@z 2

+ 1

R 2

@ 2

u

@ 2

=0

1

R

@w

@ +

@v

@z

=0

@ h

@w 1 @v u

i

(1 ) @ h

1@w @v @ 2

u i

(6)

to zequals zero. This givesan extracondition for the co eÆcientsC

8n

and C

5n :

nR

2

(1+)C

8n +R

2

C

5n

=0 (19)

5 Numerical method

For the numericalresolution of the problem,we truncate the sumsto N for n,and to M

form. aftermultiplyingeachequationbyf

m

(z),weintegratetheequations[5],[2],[3]and

[15]fromz = hto z =0usingtheorthonormalityofthesefunctions. For eachequation,

we also expand the p olynomials inthe f

m

series.

Æ

0i

= Z

0

h f

i

(z)dz Æ

1i

= Z

0

h

(z+h)f

i

(z)dz (20)

Æ

2i

= Z

0

h

(z+h) 2

f

i

(z)dz Æ

3i

= Z

0

h

(z+h) 3

f

i

(z)dz (21)

This yields, for each n 2 [0N], to a linear system of 4M +13 equations, with un-

known values U

mn

;V

mn

;W

mn

;X

mn

and the C

i

co eÆcients. The force termof the system

is logicallyfound to b e the pressure due to the diractedproblem fora rigid shell.

6 Numerical results and test of convergence

Tomakeourmetho d clearlyunderstandable,letusrst solve,for illustrationasimplied

case. Let describ e the axisymmetric, radial vibration of a cylindrical shell under the

symmetricload q(z). The former system of equations can b e reducedto the dierential

equation

D

@ 4

u

@z 4

+ Eh

R 2

u=q(z) with q(z)=kf

0

(z) (22)

Weassume that we havethe following b oundary conditions

u =0 u 0

=0 at z=-h (23)

u 00

=0 u

000

=0 at z=0 (24)

the function uwrites

u(z)= M

X

m=0 u

m

k 4

m f

m (z)+

1

6 C

1

(z+h) 3

+C

2 1

2

(z+h) 2

+C

3

(z+h)+C

4

(25)

Cho osing k = 1, D = 1, h = 50 and Eh

R 2

= 10 2

, we give the numerical solutions for u,

u (1)

, u (2)

,u (3)

and u (4)

obtained for dierentvalues of M(5;10;20;50;150 ).

(7)

0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18

-25 -20 -15 -10 -5 0

m=5 m=10 m=20 m=50 m=150

0 0.002 0.004 0.006 0.008 0.01 0.012

-25 -20 -15 -10 -5 0

m=5 m=10 m=20 m=50 m=150

-0.001 -0.0005 0 0.0005 0.001 0.0015 0.002

-25 -20 -15 -10 -5 0

m=5 m=10 m=20 m=50 m=150

-0.0003 -0.00025 -0.0002 -0.00015 -0.0001 -5e-05 0 5e-05 0.0001 0.00015 0.0002 0.00025

-25 -20 -15 -10 -5 0

m=5 m=10 m=20 m=50 m=150

-0.00015 -0.0001 -5e-05 0 5e-05 0.0001

-25 -20 -15 -10 -5 0

m=5 m=10 m=20 m=50 m=150

Figure1: convergence

(8)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

-25 -20 -15 -10 -5 0

Uabs

’Uabs’

0 0.02 0.04 0.06 0.08 0.1 0.12

-25 -20 -15 -10 -5 0

Wabs

’Wabs’

Figure 2: shelldeection

The convergence of u app ears very fast and we only need a couple of terms to have a

very go o d approximation of our solution. The series of the fourth derivative seems to

have a slow converge. Esp eciallyto obtain a right value at the p oint z = 0. This is due

to the fact that the f

m

functionsdo not ob ey the b oundary equations of the mechanical

problem.

For the complete mo delit turns out that the convergenceof the series for the deection

shows similar bahaviour. Dep ending on the physical parameters we cho ose the upp er

b ounds of the series. In general the choice M = 50 is reasonable. We present here some

numerical resultsfor a cylindricalshellwith a diameterof 6m, in a25m depth sea. The

thickness of the shell is 1cmand itsdensity is 2500Kg=m 3

. The rigidityequals E = 10 9

and the incoming wave frequency ! = 0:3r ad=s. The chosen value for the rigidity is

rather low,so the thin structure is very exible

Examples of the computation of the amplitude of the deection for u and w against z,

for =0 are shown in gure [2]As we suggested we have chosen 50 termsfor the series

although, moretermsareneededto obtaina go o d convergenceofthe highestderivatives.

In gure[3],weshowthe cylinderdeplacementat twodierent depths(z=-15mand z=-

5m) and in ve dierents time moments. The main contribution to the deplacement is

due to the b ending of the cylinder.

7 Conclusion

It willb e shown that for certainvalues of the wave frequencythe cylinderb ecomes reso-

nant. This means that our homogeneousb oundary value problem has no solution at all.

This phenomenon is well known for forced Sturm-Liouville systems with homogeneous

b oundary conditions. For the structural parameters chosen inour examplethis happ ens

intheregion ofthefrequecy bandofarealisticsea-state. Ifwecho osethe thicknesstob e

larger,forinstance5cm,the resonancephenomenono ccursat afrequencylargerthanthe

frequenciesof a sea-sp ectrum. We also willshow results for a largervalue of the rigidity

parameter. Ifwe takea value for asteelstructure we can use athicknessof 1cmleads to

no problem ina sea-state. It turns out that this approach can b e used for a wide range

of parameters

(9)

4 2 0 2 4

-15m -5m

4 2 0 2 4

-15m -5m

4 2 0 2 4

-15m -5m

4 2 0 2 4

-15m -5m

4 2 0 2 4

-15m -5m

Figure 3: shelldeection

References

[1]

Stefan

Markus . The Mechanics of Vibrationsof CylindricalShells (1988).

Studies in Applied Mechanics.

Références

Documents relatifs

As all apparatus is at rest in the laboratory’s reference frame, the fact that the measured fringe shift coincides with the theoretical prediction that uses a velocity

des données a été séparée en deux parties. D’une part, les sentiments lors de l’admission du patient dans le service de soins intensifs, d’une autre part la religion pour

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

This interaction is used in a Hartree-Fock calculation of 40Ca to study the density, the single particle energies and the shell-model potential for various compressibility

In order to better understand how to use the technologies of motion detection devices for the elderly to enjoy and benefit from active aging, bringing benefits in the treatment

The NP morphology evolves accordingly, from a rounded or Wulff shape for small Au shell thickness towards a highly faceted shape formed of a cubic Fe core surrounded by six truncated

The parameters of the signal include the accuracy and speed of information transfer. By processing the input signal properly, you can significantly simplify the task of

Nonholonomy will be the main topic then of section 5, where some implications of this phenomenon for the loco- motion of articulated systems are put to light, and where propositions