Berner Fachhochschule: ARBOR

Stahel, Andreas

Proceedings of the Royal Society of Edinburgh, 106A, 307-314, 1987
A remark on the equation of a vibrating plate
Andreas Stahel
Mathematisches Institut, Universitat Zurich, Ramistrasse 74, CH-8001 Zurich,
Switzerland
(MS received 25 August 1986. Revised MS received 3 November 1986.)
Synopsis
We consider the von Karman equations, which describe a vibrating plate either with a clamped
boundary or with completely free boundary. In both cases we obtain a unique, classical solution. As
the main tool we use a set of integral equations, which we deduce from the well known "variations of
constants" formula.
Let Q be a bounded smooth domain in U2 with boundary T. We will prove that
the following system of equations has classical solutions:
u + A2u = [u, x], in QxUJ
Du = 0 on TxR (1)
u(x, 0) = u0, u(x, 0) = u0, for x in Q J
and
A2x = -[u,u], in Q x l
Dx = 0, on T x
where
[M, V] = (3nu)(S22iO + (32 2M)(3UU) - 2(31 2M)(31 2U)
and
Du = [u r , — u r
\ dv ,
denote Dirichlet boundary conditions. This system describes the transversal
displacement u and the Airy-stress function x of a vibrating plate, whose
boundary is clamped in transversal direction but free in horizontal direction (see
Landau & Lifschitz [3]). The best results so far known are due to von Wahl ([6],
[7]). He used rather complicated techniques, involving analytic mappings in
Banach spaces and nonlinear interpolation theory to prove the existence of strong
solutions.
The purpose of this paper is to give a simple method for solving the above
problem, which has the advantage of being applicable to other situations where
the nonlinearities are not analytic. In addition we can prove the classical
solvability of (1) (2), which has not been done before.
First we reformulate the problem by solving (2) and using the solution % in (1).
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
s
o
u
r
c
e
:
 
h
t
t
p
s
:
/
/
d
o
i
.
o
r
g
/
1
0
.
2
4
4
5
1
/
a
r
b
o
r
.
1
3
0
4
1
 
|
 
d
o
w
n
l
o
a
d
e
d
:
 
3
.
6
.
2
0
2
1
308 Andreas Stahel
Let A = AD be the L
2-realisation of the biharmonic operator A2 with Dirichlet
boundary conditions. We know that A is a strictly positive, self ad joint operator
with
domAD = H
4
D={ueH\Q) | Du = 0},
and we put
H(u, v, w) = -[u, Aol[v, w]],
F(u) = H(u,u,u),
F(1\u, v) = H(v, u, u) + 2H(u, u, v).
Now we replace (1) and (2) by the Cauchy problem
u(t)edomA, f (3)
M(0) = u0, u(0) = u0.
Using well known semigroup theory and a "variations-of-constants" formula we
reach (formally) the integral equation
ri
u(t) = cos (tA^Uo + sin (tA^A'^iio + I sin ((? - x)A>i)A~iiF(u(x))dx. (4)
•'0
Our approach consists in replacing (4) by another system of equations; the main
advantage of this is that we have A~l under the integral in place of A~^. This
simple trick allows us to solve the integral equation in the space domAa+>i in
place of domAa, so we have a more regular solution.
To show this, we argue formally as follows. By differentiating (4) and using
we obtain (7). Then we integrate (7) to obtain (6). Thus we arrive at the system
u(t) = cos (tA^)u0 + sin tA^)A~^u0 - cos (tA^)A~
1F(u0) + A~
1F(u0)
+ f (id - cos ((t - x)Ai))A~1Fw(u(x), u(x)) dx, (6)
Jo
u(t) = -sin (tA^A^Uo + cos (IA^UQ + sin (M^)A~^F(M0)
+ f sin ((t - x)A^A-^Fm(u(x), u(x)) dx. (7)
Jo
The same set of equations was used in [1] and [6]. We thank von Wahl for
drawing these papers to our attention. One of the differences in our approach to
that of von Wahl's paper is that we do not solve a "weak" problem first and
regularise afterwards, but solve (6), (7) in the "best possible" function spaces
directly.
Let us consider functions u, satisfying
u e C°(J, HAD) n C\J, H
2
D) n C\J, H°)
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
The equation of a vibrating plate 309
for some nontrivial interval J containing 0. So we have F(u(.)) e Cr(J, H°) and
(5) (use Lemma 1 below). Now it is easily verified that the above formal
operations are correct and the equations (3) and (6), (7) are indeed equivalent.
Now we shall show that the system (6), (7) possesses a unique solution (u, v) in
appropriate function spaces and that v = u, such that we have a solution to (3).
For this we put
Hsn=<
H*(Q)n{u\u\r = l
H'^nH^nlu
dv
if
if
if
if
\<s<l
If 0 < 6 < 5 we deduce from this definition and the interpolation results of
Grisvard [2]
using the well known fact that domAa coincide with appropriate interpolation
spaces (e.g. [5, 1.18.10] or [4]). Now we easily deduce the following theorem.
THEOREM. Suppose that
u0eHy-
6 and u0eH
3J6 (8)
for some 6 e [0, 5). Then there exists a maximal interval J = [T~, T+), containing
0, and a unique solution u of (3) satisfying
u e C°(J, H5D
+6) n C\J, H3D
+6) n C\J, HlD
+6) (9)
and by Sobolev embedding theorems this solution is in fact classical. If for some
e>0and some k(.) e C°(U,R+)
for all te[O, T+), (10)
then T+ = 00. As long as (8) and 6 < \ are satisfied, T+ does not depend on 6.
Similar results are true for T~.
Proof. From Lemma 1 (i) and Lemma 2 we obtain
F ( 1 ) e Cl-(H5D
+6 x H3D
+6, Hh+d)
(i.e. F(1) is locally Lipschitz continuous) and that F(1) is bounded on bounded
sets. By standard iteration and continuation arguments we obtain a solution
(«, v) of (6), (7) satisfying
(u, v) e C°(J, H5D
+d X H3D
+6)
and
lim
t-*T+
(11)
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
310 Andreas Stahel
provided T+ < ». We consider only the right endpoint T+ of /, since similar
arguments apply to T~. Now we differentiate (6) and obtain u(f) = v(t) with
u e C\J, Hjj6). This gives F(u(.)) e C\J, H1^6) and (5). Thus u is a solution to
(3). Suppose (10) and T+<oo, then we apply part (ii) of Lemma 1 and
Gronwall's inequality to the integral equation (7), considered in C°(J, H2). This
gives an a priori bound for ||M||W2, i.e.
sup \\u(t)\\H2<<*>.
<e[0, r + )
If we look at the expressions under the integral in (6), we obtain an a priori
bound for HMIIH4. NOW we find a bound for HiiH^+s by Lemma 1 (i) and
Gronwall's inequality applied to (7). Thus, using (6) again, ||M||HS+« is bounded
and we have a contradiction to (11).
li 61^d2<2 and T~f (respectively T%) is the corresponding time of existence,
it is obvious that T^Tf. If T% < 7\+, then
sup | |M(OII//I,<00,
te[0, 1?]
and we obtain TJ = °o (by means of (10)). Consequently T^=TX and the
Theorem is proved. •
Remarks. (A) Von Wahl proved global existence under one of the
assumptions
HKO, "(0111^<*(')> ^ all te[0,T+), (10')
for some p > 1, or
for all te[0,T+), (10")
for some p > 1, where K is the Gaussian curvature of the bent plate (see [6,
formula V.12/13]). In [6, p. 71] the author proves that (10') and (10") are
equivalent if used to prove global existence. We show that (10) and (10') are
equivalent in this context. So it does not matter which of the conditions (10),
(10') or (10") we use. They are different formulations of the same assumption and
all of them imply global existence of the solution u.
The proof of part (ii) of Lemma 1 shows that (10) implies (10') (look at the
lower left corner of the first diagram of that proof). Let us assume (10'). If we
consider in F(u) = H(u, u, u) only the first argument of H and fix the other two,
then we obtain, by the proof of Lemma l(i) and (10'), the estimate
\\H^c \\[u, u]\\LP \\u\W^ck(t) ||«||«3.
Now we take (4) as an integral equation for u e C°(J, H3) and apply Gronwall's
inequality. Thus we obtain an a priori estimate for ||w||/f<. We see that (10) is
satisfied and so is in fact equivalent to (10').
(B) It is apparently not known whether the above result is true in the limiting
case 6 = 5. The above proof does not work, since H$ is not a subspace of dom>l|,
(c.f. [2]) and the nonlinearity does not satisfy further boundary conditions (except
the one given in Lemma 2).
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
The equation of a vibrating plate 311
(C) The basic idea of the above proof applies to more general situations. Let A
be a strictly positive, selfadjoint operator. For a fixed number ff^Owe have to
find an
pW 6 C
1~(domA1+a X domA^+a, domAa),
(where domAa is endowed with the graph norm), which is bounded on bounded
sets such that
u e C°(J, dom A1+a) n C\J, domA^")
implies
F(u(.))eC\J, dom Aa) and ^-F(u(t)) = Fw(u(t), u(t)).
at
If this condition is satisfied, by solving (6) (7) we obtain a solution u to (4) with
u e C°(J, domA1+a) n C\J, dom A^+a) n C\J, domAa).
Thus u is also a solution to (3).
This remark corresponds to von Wahl's Theorem II.2 in [6], but the conditions
on the nonlinearity are different. It is not obvious which approach is more general
(with respect to the nonlinearities). For simplicity we have not included the case
of time dependent A(t), as is done in [6].
(D) If we consider the equation for a vibrating plate with completely free
boundary we must replace the Dirichlet boundary conditions in (1) (but not in
(2)) by
Bu = (Bou, Bxu) = (0, 0),
3 \ 2 //d\2 9 \
— U + O\\ U + K U],
av/ Wdu/ dv I
Bxu = (-|-) AM + (1 -
\dv/
where o e (0, 1/2) is Poisson's ratio, a constant that depends on the plate
material, K is the curvature and fi the tangential vector of the boundary F. The
deviation of this equation is given in [3] (provided some additional calculations
are carried out). Now, in the Theorem, we replace A by AB, F by F and F
w by
F(1), where
ABu = A
2« + u,
F(u) = F(u) + u,
Fw(u, v) = Fw(u, v) + v.
We know that for 6 < \ (writing A for AB again),
domA±+6l* = H2+d, domAl+m = H%+6, d o m ^ + a / 4 = H%+d.
Together with part (iii) of Lemma 1 we are in the situation of Remark C with
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
312 Andreas Stahel
a = \ + 8/4. Therefore we have a solution u of (4) with
u e C°{J, H%+d) n C\J, H%+6) n C2(/> H2+s).
By Sobolev-imbedding theorems we deduce that u is a classical solution of
u + A2u = F{u), in QxJ
Bu = 0, on T x /
It remains to prove the lemmas which have been used above.
LEMMA 1. Fix 0 g 6 ̂  \ and e > 0. Then
He <£\XxYxY,Z)n%\YxYxX, Z),
{i.e. H is trilinear and continuous) with
(i) X = H3+S, Y = H\ Z = H1+6,
(ii) X = H2, Y = H2+£, Z = H°,
(iii) X = H4+6, Y = H6, Z = H2+d.
Proof. Let d2 be a differential operator of order 2 and denote the multiplica-
tion of two functions by °. Then we have to find appropriate function spaces,
over Q, such that in the following two diagrams all mappings are continuous:
'"}-
z,
x^^z
• E2
 52AD" > G
It is not too difficult to see that the following choices are admissible:
(i) Ex = H
2, E2 = H\ G = H
3,
(ii) £x = H
e, E2 = U for 1 - e/2 = Up, G = W
2-",
(iii) E, = H\ E2 = H
2, G = H\ D
The following lemma implies that F(u) and F(1)(w) vanish on T if u and
(d/dv)u are 0 on T. We use this fact in the Theorem to obtain solutions u e H5+d,
which is better than von Wahl's result [6] [7] (he proved u e / / 4 + 6 ) .
LEMMA 2. Ifu,ve H3^6 then [u, v] e HlD.
Proof. As in the proof of Lemma 1 it is seen that [u, v] e H1. We have to show
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
The equation of a vibrating plate 313
that this expression vanishes on the boundary. Because F is smooth, we can
choose the outer unit normal vector v and a tangential vector fi defined on a
neighbourhood of F such that
3 3 d d
— V = Kfl, —-fl = -KV, — v = 0, — jU=O,
dfi dfi dv 3v
where K is the curvature of F. If we denote by D2u the matrix of second order
partial derivatives of u, we obtain by elementary differentiation and matrix
calculations
3 3 , _ 3 3 , Tu = vDuv , u = vDufi ,
dv dv dvdfi
3 3 . _ 3 3 3 , _ 3
U = flUUfl —K U, U = flDuV + K U.
dfi dfi dv dfi dv dfi
From this and the invariance of the Karman bracket [u, v] under rotations we
obtain
[u, v) = (dnu){322v) + (32 2M)(3UW) ~ (3i2«)(32iu) - (dnu){d12v)
= (vD2uvT)(fiD2vfiT) + (fiD2ufiT)(vD2vvT)
- (vD2ufiT)(fiD2vvT) - (fiD2uvT)(vD2vnT)
(33 \( 3 3 \ (33 \( 3 3
= (— — u l — — - v) + \ — — « I — — v
\dv dv /\dfldfl I \dfidfi /\dvdv
d 3 \( d d \ ( d 3
ull u)~"\"a u
I ( d 3 \ 3 ( d \( d 3
Wdv dv I dv \3v )
( 3 3 \ 3 ( 3 \( 3 3
On F we have, because of u e H3D,
3_ _ 3 _ 3 _3_
dv dfi dfi dfi
and the same is true for v. Now it is easy to see that [u, v] | r = 0 and we have
proved the lemma. •
References
1 P. Brenner and W. von Wahl. Global classical solutions of nonlinear wave equations. Math. Z. 176
(1981), 87-121.
2 P. Grisvard. Caracterisation de quelques espaces d'interpolation. Arch. Rational Mech. Anal. 25
(1967), 40-63.
3 L. D. Landau and E. M. Lifschitz. Lehrbuch der theoretischen Physik vii (Berlin: Akademie,
1975).
4 M. Taylor. Pseudodifferential Operators (Princeton: Princeton University Press, 1981).
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the
314 Andreas Stahel
5 H. Triebel. Interpolation Theory, Function Spaces, Differential Operators (Amsterdam: North
Holland, 1978).
6 W. von Wahl. On nonlinear evolution equations in a Banach space and on nonlinear vibrations of
the clampled plate. Bayreuth. Math. Schr. 7 (1981), 1-93.
7 W. von Wahl. Corrections to my paper: . . . (see [6]). Bayreuth. Math. Schr. 20 (1985), 205-209.
{Issued 30 July 1987)
Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0308210500018436
Downloaded from https://www.cambridge.org/core. Haute Ecole Spécialisée de Suisse occidentale, on 13 Nov 2020 at 14:12:54, subject to the


Bern University of Applied Sciences

Not a member yet

A Remark on the Equation of a Vibrating Plate

3,898

12,291