Copyright© 1998, American Institute of Aeronautics and Astronautics, Inc.
A98-30813
AIAA-98-2221
DEVELOPMENT OF A HIGH-ORDER COMPACT ALGORITHM FOR
AEROACOUSTICS
EMPLOYING PML ABSORBING BOUNDARIES
Peter W. TenPas*, Stephen E. Schwalm
+
and Ramesh K. Agarwal**
*
>+
Department of Mechanical Engineering
University of Kansas
Lawrence, Kansas 66044-7526
** Aerospace Engineering Department
National Institute for Aviation Research
Wichita State University
Wichita, Kansas 67260-0093
ABSTRACT
A finite-difference method for solving the two-
dimensional linearized Euler equations in terms of the
acoustic variables is described. The far-field
boundary conditions are approximated by the
perfectly-matched-layer (PML) equations, which
simulate an absorbing layer surrounding the interior
of the computational domain. The spatial derivatives
are represented by high-order compact differences,
while the four-stage Runge-Kutta method is used to
integrate the equations forward in time. Several
benchmark problems have been computed with this
technique, including propagation of a linear wave, and
convection of acoustic and vortical disturbances in a
uniform flow. The results obtained with fourth-order,
sixth-order and eighth-order compact difference
methods are presented. The compact method gives
high resolution on a relatively small difference stencil.
Progression from the fourth-order to the sixth-order
and eighth-order schemes is straightforward. The
PML approximation worked well in suppressing non-
physical reflections at the boundaries of the domain.
The boundary error can be controlled or eliminated by
increasing the width of the absorbing PML layer.
Copyright © 1998 by P.W. TenPas. Published by the
Confederation of European Aerospace Societies, with
permission.
* Associate Professor, Member AIAA
"""Undergraduate Research Assistant
** Bloomfield Distinguished Professor, Fellow AIAA
INTRODUCTION
Numerical simulation of aerodynamic sound
propagation places severe requirements on the
accuracy of the computational process. Research in
computational aeroacoustics (CAA) is directed toward
the development of procedures that are capable of
high resolution and are also robust and
computationally efficient. Tarn
1
has recently
published a survey of the field. The list of technical
issues includes but is not limited to: high frequency
wave resolution on coarse grids, proper representation
of the far-field boundary conditions, integration over
many cycles or long times and acceleration of
convergence to steady-state.
Computational resources such as processor speed and
storage space place a practical limit on the number of
grid points used in a numerical simulation. To study
high frequency wave propagation on a relatively
coarse grid the numerical method used must be able to
accurately resolve waves with as few as 6 node points
per wavelength. Low-order finite-difference methods
exhibit large dispersion errors when applied on coarse
grids. Conventional high-order methods yield more
accurate results at the expense of incorporating a
larger number of nodes in the difference stencil. This
adds to the processing time, and more importantly the
wider stencils are more difficult to apply near
boundaries. The formal order of accuracy and the
number of points in the stencil are not reliable
measures of the resolution of a particular scheme.
The compact schemes described by Lele
2
achieve
higher-order accuracy without greatly expanding the
104