Recent development of BEM/BIEM in vibration and acoustics
-
Upload
valentine-glover -
Category
Documents
-
view
27 -
download
0
description
Transcript of Recent development of BEM/BIEM in vibration and acoustics
M
1
S
V
Recent development of BEM/BIEM in vibration and
acoustics陳正宗
海洋大學 特聘教授河海工程學系
Nov. 19, 2004, NSYSU, 14:10~16:00
M
2
S
VOutlines
Introduction Exterior acoustics - adaptive BEM Interior acoustics - multiply-connected
eigenproblems Conclusions
M
3
S
VGrowth of BEM papers
M
4
S
VIntroduction
Finite difference method (FDM) Finite element method (FEM) Boundary element method (BEM) Meshless method (MM) Boundary integral equation method (BIEM)
FDMFEM
Domain discretization
BEM
Boundary discretization
MM BIEM
No meshNo discretization for circular boun
daries
No meshNo node
M
5
S
V
Adaptive BEM for exterior radiation
and scattering problems
M
6
S
VProblem statement
Non-uniform radiator problem Scattering problem
1),( au0),( au
Drruk ),( ,0),()( 22
Drruk ),( ,0),()( 22
M
7
S
VAdaptive scheme
( ) . . . ( , ) ( ) ( ) . . . ( , ) ( ) ( )B B
u x C PV T s x u s dB s R PV U s x t s dB s Solver
( ) . . . ( , ) ( ) ( ) . . . ( , ) ( ) ( )B B
t x H PV M s x u s dB s C PV L s x t s dB s Error indicator
R.P.V. is Riemann Principal Value
H.P.V. is Hadamard Principal Value
C.P.V. is Cauchy Principal Value
Singular formulation
Hypersingular formulation
M
8
S
VAdaptive mesh
Uniform mesh Adaptive mesh
-1.00 -0.50 0.00 0.50 1.00
-1.00
-0.50
0.00
0.50
1.00
-1 .00 -0.50 0.00 0.50 1.00
-1.00
-0.50
0.00
0.50
1.00
M
9
S
VRefinement scheme
h-version p-version r-version
1. Element number increasing2. Interpolation function order increasing3. Optimum nodal collocation
1 2 3
M
10
S
VMesh
BEM FEM(DtN)
- 1 - 1 0 1 1
- 1
- 1
0
1
1
Taiwan, NTOU
US Navy. Stanford Univ.
M
11
S
V
Non-uniform radiation : Dirichlet problem
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
Numerical solution: BEM Numerical solution: FEM(DtN) 64 ELEMENTS 2791 ELEMENTS
2ka
Taiwan, NTOU
US Navy. Stanford Univ.
M
12
S
V
Non-uniform radiation : Dirichlet problem
-1 .50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
Analytical solution: n=20 2ka
nkaH
krH
n
nru
n
n
ncos
)(
)(sin2),(
)1(
)1(
0
1),( au0),( au
Drruk ),( ,0),()( 22
32
5
M
13
S
VSuperposition principle
),(),(),( )()( rururu si
+
),()( ru s),( ru i cosikre∥
0),(
n
ru
n
u
n
u is
)()(
n
u i
)(
M
14
S
V
Scattering : Neumann problem
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
Numerical solution: BEM Numerical solution: FEM(DtN) 63 ELEMENTS 7816 ELEMENTS
4ka
Taiwan, NTOU
US Navy. Stanford Univ.
M
15
S
V
Scattering :Neumann problemAnalytical solution: n=20
-1 .50 -1.00 -0.50 0.00 0.50 1.00 1.50 2.00
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
-1.50 -1.00 -0.50 0.00 0.50 1.00 1.50
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
4ka
nkrHkaH
kaJikrH
kaH
kaJru n
n
nn
ncos)(
)(
)(2 )(
)(
)(),( )1(
)1(1
)1(0
)1(0
0
Drruk ),( ,0),()( 22
M
16
S
V
Fictitious frequency : Non-uniform radiation problem
ka0 2 4 6 8
-0.8
-0.4
0.0
0.4
0.8
1.2UT method
LM method
Burton & Miller method
Analytical solution
u(a,0)
1),( at0),( at
Drruk ),( ,0),()( 22
9
1),( at0),( at
Drruk ),( ,0),()( 22
9
M
17
S
V
Fictitious frequency : The scattering Dirichlet problem
t(a, 0)
ka
Drruk ),( ,0),()( 22
•
0 2 4 6 8
-12
-8
-4
0
4
8
UT m ethod
LM m ethod
Burton & Miller m ethod
Analytical solution
1
1J
1
2J
2
0J
1
3J
1
4J
1
5J
2
2J
3
0J
1
6J
M
18
S
VSummary
Fictitious frequency depends on the formulation (singular or hypersingular) instead of B.C. (Dirichlet or Neumann).
Burton & Miller method and CHIEEF method can overcome the problem of fictitious frequency.
Fictitious frequency happens to be the true eigenvalues of the interior problem
(SingularDirichlet, HypersingularNeumann).
M
19
S
V
Spurious eigenvalues for multiply-connected pro
blems
M
20
S
VProblem domain
Doubly-connected domain
Multiply-connected domain
Simply-connected domain
M
21
S
VBEM&BIEM
BEM
BIEM
.
.
.
.
.
.
.
.
.
..
.
Boundary discretization Fourier series
0 2 0 2
u(θ) or t(θ)
θ θ
u(θ) or t(θ)
M
22
S
V
The flowchart to determine the eigenvalues and mode shape by BEM
Given G.E.and B.C.
Solve thefundamental solution
BIE for domain point
Moving to the boundary
BIE for boundary point
Boundary element discretization
Linear algebraic system
Solve boundary
dataPotential
SVD
M
23
S
V
The flowchart to determine the eigenvalues and mode shape by BIEM
Degenerate kernel
Fourier series
Null-field equation
Algebraic equation Fourier Coefficients
Potential
Analytical
Numerical
SVD
M
24
S
VIntegral Formulation
Null-field integral equations:
0 ( , ) ( ) ( ) ( , ) ( ) ( ), .eB BT s x u s dB s U s x t s dB s x D= - Îò ò%% % % %% % % %
0 ( , ) ( ) ( ) ( , ) ( ) ( ), .eB BM s x u s dB s L s x t s dB s x D= - Îò ò%% % % %% % % %
M
25
S
Vs
x
EU
rO
RIU
x : source point ; s : field point
s
r
x
Degenerate kernels
M
26
S
VDegenerate kernels
Degenerate kernels:
(1)
0
(1)
0
( , ) ( | |) ( | |) cos( ),| | | |2
( , )
( , ) ( | |) ( | |) cos( ),| | | |2
Im m j m j j j
m
Em m j m j j j
m
iU s x J k x c H k s c m s c x c
U s xi
U s x H k x c J k s c m x c s c
1, 0,.
2, 0,m
m
m
(1)
0
(1)
0
( | |)( , ) ( | |){ }cos( ), | | | |
2( , )
( | |)( , ) ( | |){ }cos( ), | | | |
2
m jIm m j j j
m j
m jEm m j j j
m j
H k s ciT s x J k x c m s c x c
RT s x
J k s ciT s x H k x c m x c s c
R
M
27
S
VDegenerate kernels
(1)
0
(1)
0
( | |)( , ) ( | |){ cos( ) cos( )
2
1 cos( )( | |) cos( )}, | | | |
2( , )
( | |)( , ) ( | |){ cos(
2
m jI
m m jm
m j j j
m jE
m m jm
c jj
c jj j
j
J k x cis x H k s c m
mJ k x c s c x c
L s xH k x ci
s x J k s c m
L
L
(1)
) cos( )
1 cos( )( | |) cos( )}, | | | |
2m j j j
c j
c jj j
mH k x c x c s c
(1)
0
(1)
0
( | |)( , ) { cos( ) cos( )
2
cos( )( | |) cos( )}, | | | |
2( , )
( , )2
( | |)
1
( | |) ( | |){
m j m jI
mm
m j j j
m j m jE
mm
c jj j
c jj
j
j
J k x cis x m
mJ k x c s c x c
M s xi
s x
H k s cM
R
J k s c H k x cM
R
(1)
cos( ) cos( )
( | |) cos( ) | | | |2
1 cos( )},m j j j
c jj
c jjj
m
H k s c x c s cm
M
28
S
V
Fourier series for boundary densities
Fourier series:
01
( ) ( cos sin ),j nj j nj j jn
u s a a n b n s Bq q¥
=
= + + Îå% %
01
( ) ( cos sin ),j nj j nj j jn
t s p p n q n s Bq q¥
=
= + + Îå% %
a),( Rs
M
29
S
VCollocation points
By choosing M terms of Fourier series, we select 2M+1 collocation points on the circle.
01
( ) ( cos sin )M
n nn
u x a a n b n
2M+1 terms
M
30
S
VIntegral representation
2 ( ) ( , ) ( ) ( ) ( , ) ( ) ( ),B B
u x T s x u s dB s U s x t s dB s x Dp = - Îò ò% %% % % %% % % %2 ( ) ( , ) ( ) ( ) ( , ) ( ) ( ),
B Bt x M s x u s dB s L s x t s dB s x Dp = - Îò ò% %% % % %% % % %
Integral equation formulation:
M
31
S
VNumerical examples
Example 1
2 2( ) ( ) 0,k u x x D
1 0.5r
2 2.0r
1B
0u 2B
0u
M
32
S
V
The eigenfrequenies by using singular equation
More accurate
( ) Numerical
[ ] exact
0.0 1.0 2.0 3.0 4.0 5.0
1E-004
1E-003
1E-002
1E-001
1E+000
0.5 2.0
P resent m ethod[U c] kerne l
BEM[U c] kerne l
J 1 0
(4 .83)[4.81]
[4.81]
Contaminated by spurious eigenvalues
M
33
S
V
Relation of spurious eigenvalue and true eigenvalue
( 0.5) 0 4.81nJ k k
0.5 True
Spurious eigenvalue using singular formulation happens to be the true eigenvalue of the associated interior Dirichlet problem.
M
34
S
V
The eigenfrequenies by using hyper-singular equation
( ) Numerical
[ ] exact
More accurate
More accurate
0.0 1.0 2.0 3.0 4.0 5.0
1.0E-003
1.0E-002
1.0E-001
1.0E+000
1.0E+001
J '1
0
(0.35)[0 .00]
J '2 0
(3 .68)[3.68]
J '2 0
(3.77)[3 .68]
[0.00][3.68]
0.5 2.0
Present m ethod[L c] kerne l
BEM[L c] kerne l
Contaminated by spurious eigenvalues
M
35
S
V
Relation of spurious eigenvalue and true eigenvalue
( 0.5) 0 0, 3.68nJ k k
0.5 True
Spurious eigenvalue using hypersingular formulation happens to be the true eigenvalue of the associated interior Neumann problem.
M
36
S
V
The spurious eigenvalues are filtered by Burton&Miller method
Only true eigenvalues appear
0.0 1.0 2.0 3.0 4.0 5.0
1E-002
1E-001
1E+000
1E+001
0.5 2 .0
B urton&M iller m ethod(P resent m ethod)
B urton&M iller m ethod
(BE M )
M
37
S
V
The former five eigenvalues of Helmholtz eigenproblem with an eccentric d
omain
1 2 3 4 5
FEM [Chen et. al.] 1.73 2.13 2.45 2.76 2.95
BEM[Chen and Zhou] 1.75 2.14 2.47 2.78 2.97
BEM [Chen et. al.] 1.74 2.14 2.47 2.78 2.98
Present method 1.74 2.14 2.46 2.78 2.96
M
38
S
V
The former five eigenmodes for eccentric case using present method, FEM a
nd BEM
M
39
S
VNumerical examples
Example 2
R=1
c1=0.3c2=0.4
e=0.5
2 2( ) ( ) 0,k u x x D
M
40
S
V
Extraction of the spurious eigenvalues by using SVD updating docu
ment
0 1 2 3 4 5 6 7 8 9
0.001
0.01
0.1
1
10
100
1000
The m in im um singular va lues 1 for the updating docum ent by BEM (90 e lem ents)
The m in im um singular va lues 1 for the updating docum ent by B IEM (63 co llocation po ints)
( ) : e x ac t so lu tio n
6.01 8.02More accurate
More accurate
M
41
S
V
The former five eigenvalues for a multiply-connected problem with two unequal holes using different approaches
Method ki
k1 k2 k3 k4 k5
Burton & Miller method 4.82 4.82 6.72 6.72 7.82
Direct BEM + SVD Updating 4.81 4.81 6.73 6.73 7.81
Null-field BEM + SVD Updating 4.81 4.81 6.73 6.73 7.82
Fictitious BEM + SVD Updating 4.80 4.80 6.72 6.72 7.79
Direct BEM + CHIEF method 4.81 4.81 6.73 6.73 7.82
Null-field BEM + CHIEF method 4.83 4.83 6.74 6.74 7.84
Fictitious BEM + CHIEF method 4.77 4.77 6.68 6.68 7.88
FEM 4.790 4.801 6.619 6.634 7.797
Present method 4.85 4.85 6.77 6.77 7.91
M
42
S
V
The former five modes for a circle domain with two unequal holes using present method, BEM and FEM
M
43
S
VSummary
Spurious eigenvalues depend on formulation (singular or hyper-singular).
Spurious eigenvalues are independent of B.C. (Dirichlet or Neumann).
Spurious eigenvalues happens to be the true eigenvalues of the interior problem
(Dirichletsingular, Neumannhypersingular).
To overcome the spurious eigenvalues Burton&Miller, SVD updating term, SVD upd
ating document…….
M
44
S
VConclusions
Exterior acoustic problems (radiation and scattering) were solved by using adaptive BEM.
Good accuracy and efficiency of the present method were obtained in comparison with those with FEM.
Spurious eigenvalues embedded in the BIEM/BEM were examined and filtered out in this study.
Both the fictitious frequency and spurious eigenvalue depend on the formulation instead of B.C. .
M
46
S
V