Infinite Elements in Acoustics

Innite Elements in Acoustics
Peter Rucz
January 8, 2010
Problem denition
f
=
n

f
Governing equations
Helmholtz equation
2
p(x) +k
2
p(x) = 0 x (1)
Boundary conditions
p(x) n = i
0
v
n
(x) x
(2)
Sommerfeld radiation condition

lim
r
r
p
r
+ ikp
= 0 (3)
We seek p(x) for the near and the far eld,
n
and
f
for a given
circular frequency (). There are several methods for the solution.
Numerical methods for the solution
BEM (Coupled FEM/BEM)

+ Non-local boundary conditions
+ Can be applied for non-convex geometries
Full, frequency depedent system matrices
PML (Absorbing layers)

+ FE extension fast performance
+ Can be applied for non-convex geometries
Truncation of the computational domain
To obtain far eld solution postprocessing is needed
Innite elements (FEM+IEM)

+ Extension of the FE method
+ Far eld results obtained straightforwardly
+ Frequency independent system matrices
Can only be applied for convex geometries
(?) Application for CAA (Computational AeroAcoustics)
Asymptotic form of a radiating solution
General expression of the solution of the 3D Helmholtz equation:
p(x, ) =
n=0
n
m=0
h
(2)
n
(kr)P
m
n
(cos) {A
nm
sin(m) +B
nm
(cos(m))}
+
n=0
n
m=0
h
(1)
n
(kr)P
m
n
(cos) {C
nm
sin(m) +D
nm
(cos(m))} ,
where h
(1)
n
, h
(2)
n
and are P
m
n
Hankel and Legendre functions.
Inwardly propagating waves can be omitted and by expanding
Hankel functions we get:
p(x, ) = e
ikr
n=1
G
n
(, , )
r
n
(4)
The Atkinson-Wilcox theorem
The AtkinsonWilcox theorem states that the sound eld at any
point which lies outside a circumscribing sphere, which itself
encloses all radiating and scattering sources, can be written as a
multipole expansion of type (4).
(Point A satises this condition while point B does not.)
Meshing with innite elements
To fulll the conditions of the
AtkinsonWilcox theorem, the
enclosing sphere of the radiating
object must be built by nite
elements. This ensures the
convexity but costs additional
DOFs for the computation.
Innite elements are joined to the
boundary of the surface.
n
Element mapping
A mapping transformation projects the innite element into a standard,
nite parent space. Points at are mapped to the s = 1 line in the
parent element.
x =
4
i=1
M
i
(s, t)x
i
M
1
=
(1 t)s
s 1
M
2
=
(1 + t)s
s 1
M
3
=
(1 + t)(1 + s)
2(s 1)
M
4
=
(1 t)(1 + s)
2(s 1)
Mapping nodes element geometry
Variable nodes pressure assigned and computed
Radial order number of nodes, order of approximation

Formulation I. Solution scheme
The weak form in the FE scheme
A()q() = f() A
e
(w
k
2
w
)d (5)
Conjugated AstleyLeis formulation
Shape functions: N(s, t)
Test functions: (s, t) = N(s, t)e

ik(s,t)
Weight functions: w(s, t) = D(s, t)N(s, t)e

+ik(s,t)
where
AstleyLeis weighting: D(s, t) = (1 s)

2
/4
Phase functions: (s, t) = a(s, t)(1 +s)/(1 s)

AstleyLeis weights ensure the niteness of the integral (5). Phase
terms cancel each other out and the resulting system matrices (K,
M and C) are frequency independent, non-hermitian and sparse.
When the matrices are assembled the solution can be carried out
by direct or iterative formulas.
Formulation II. Shape functions
Lagrangian shape functions

N
l
(s, t) =
1 s
2
L
p
l
(s)S(t),
where L
p
l
(s) is the standard Lagrange interpolating polynomial and
S(t) is the standard shape function for a nite element.
Hence the solution is sought (corresponding to (4)) in the form
l
=
p
i=1
i
r
i
e
ik(ra)
Alternatively Jacobian polynomials can be used as shape functions,

as they have the ortogonality property:
+1
1
(1 s)
(1 +s)
J
,
i
(s)J
,
j
(s)ds =
i
ij
, (6)
which results in a lower condition number of the system matrix. In
this case a transformation matrix must be introduced to return to
the Lagrangian space.
Demo application A transparent geoemtry (I.)
In this demo application a point
source radiates in free space. A
transparent FE geometry is built
around the source. The solution
is carried out by a coupled
FEM/BEM and a FEM/IEM
method and their accuracy is
compared to the analytic
solution. In the IEM case, innite
elements with radial order of 5
are attached to the outer
boundary of the FE geometry. In
the FEM/BEM case an
impedance boundary condition is
computed for that region.
Demo application The system matrices (II.)
The IEM/FEM system matrix
Assembling time: 29.14 s.
The FEM/BEM system matrix
Assembling time: 434.19 s.
Demo application IEM solution (III.)
Relative L
2
error norm of IE/FE solution: 0.0108
Demo application BEM solution (IV.)
Relative L
2
norm of FE/BE solution: 0.0172
Real application Organ pipe modeling
Comparison with previous results
Pipe: 4/16 Measurement Indirect BEM Coupled FE/BE IEM/FEM
Harmonic F [Hz] Stretch F [Hz] Stretch F [Hz] Stretch F [Hz] Stretch
1. (Fund.) 129.87 1.000 131 1.000 128 1.000 126 1.000
2. (Octave) 261.76 2.016 263 2.008 253 1.977 255 2.024
3. 396.45 3.053 397 3.031 388 3.031 387 3.071
4. 536.98 4.135 531 4.053 522 4.078 512 4.135
5. 677.62 5.218 667 5.092 660 5.156 658 5.222
1. (Fund.) 131.22 1.000 130 1.000 128 1.000 125 1.000
2. (Octave) 262.44 2.000 262 2.008 252 1.969 253 2.024
3. 400.38 3.051 394 3.025 387 3.023 384 3.072
4. 547.08 4.169 529 4.056 521 4.070 519 4.152
5. 680.99 5.190 664 5.095 660 5.156 655 5.240
1. (Fund.) 131.22 1.000 130 1.000 126 1.000 125 1.000
2. (Octave) 265.12 2.020 262 2.007 255 2.024 253 2.024
3. 401.73 3.061 395 3.024 388 3.079 384 3.072
4. 543.71 4.143 529 4.053 524 4.159 519 4.152
5. 679.64 5.190 665 5.095 662 5.254 656 5.248

Infinite Elements in Acoustics

Uploaded by

Copyright:

Available Formats

Infinite Elements in Acoustics

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Infinite Elements in Acoustics

Uploaded by

Copyright:

Available Formats

Innite Elements in Acoustics

Sommerfeld radiation condition

BEM (Coupled FEM/BEM)

PML (Absorbing layers)

Innite elements (FEM+IEM)

Mapping nodes element geometry

Variable nodes pressure assigned and computed

Radial order number of nodes, order of approximation

Shape functions: N(s, t)

Test functions: (s, t) = N(s, t)e

Weight functions: w(s, t) = D(s, t)N(s, t)e

AstleyLeis weighting: D(s, t) = (1 s)

Phase functions: (s, t) = a(s, t)(1 +s)/(1 s)

Lagrangian shape functions

Alternatively Jacobian polynomials can be used as shape functions,

You might also like