Copyright© 1998, American Institute of Aeronautics and Astronautics, Inc.
A98-30893
AIAA-98-2361
High-Order Accurate Dual Time-Stepping Algorithm for Viscous
Aeroacoustic Simulations
Chingwei M. Shieh* and Philip J. Morris
t
Department of Aerospace Engineering
The Pennsylvania State University
University Park, PA 16802, U.S.A.
The field of computational aeroacoustics (CAA) has shown great promise in the so-
lution of primarily inviscid noise generation and wave propagation problems that are
governed by the Euler equations. The present research extends these ideas to the solu-
tion of the Navier-Stokes equations in multi-dimensions where the acoustic and flow fields
may be influenced by viscous effects. In this paper, efficient acceleration techniques typ-
ical of explicit steady-state solvers are extended to time-accurate calculations. Stability
restrictions based on the grid spacing are relaxed greatly with the implementation of a
fully implicit time discretization. Some simplified problems that address the important
issues of viscous calculations are investigated. A two-dimensional, Navier-Stokes code,
written in Fortran 90 with the Message Passing Library (MPI) as a parallel implemen-
tation, is used to perform these calculations. As an example calculation, scattering of an
acoustic source by a flat plate in the presence of a mean flow including a viscous boundary
layer is presented.
Introduction
I
N aeroacoustic simulations, the phenomena of in-
terest are inherently unsteady and consist of a wide
range of frequencies and amplitudes. Hence, to repro-
duce the physics of these phenomena faithfully and to
capture all the scales in the flow field, new algorithms
have been proposed for very accurate calculations of
wave propagation. Typical CAA algorithms consist
of high-order finite difference spatial discretizations
(fourth-order
1
, sixth-order
2
) and high-order explicit
time-stepping techniques (high-order explicit linear
multistep methods
3
, Runge-Kutta
4
) to minimize both
dispersion and dissipation errors. For simple acous-
tic model problems and inviscid calculations, such
algorithms are efficient and accurate as long as the
numerical stability criterion is met. However, these
algorithms are no longer efficient when calculations of
the acoustic and flow fields for more realistic problems
are performed.
In a recent study, Lockard
5
performed parallel com-
putations of acoustic scattering by a three-dimensional
wing. The unsteady part of the calculations were
carried out with the use of a seven-point Dispersion-
Relation-Preserving (DRP) finite difference stencil and
a low dispersion and dissipation Runge Kutta scheme.
However, after 91 equivalent single CPU hours, the
acoustic pulse had not propagated out of the compu-
tational domain. This is because a very small tune
step was taken for the unsteady calculation due to
the numerical stability criterion based on the tiny grid
"Graduate Research Assistant, Student Member, AIAA
t Boeing/A. D. Welliver Professor, Associate Fellow, AIAA
Copyright ©1998 by C. M. Shieh and P. J. Morris. Pub-
lished by the Confederation of European Aerospace Societies,
with permission.
spacing at the tip of the wing. Although the algorithm
is still very accurate for this inviscid problem, it is no
longer efficient. Therefore, in order to extend CAA
methodologies to the calculation of the Navier-Stokes
equations in multi-dimensions where the acoustic and
flow fields may be influenced by viscous effects, a dif-
ferent approach needs to be taken.
Many popular algorithms for unsteady simulations
in CFD use an implicit method. Unlike explicit meth-
ods, one of the advantages in the use of an implicit
method is its ability to advance solutions with a very
large time step. This is especially important in Navier-
Stokes calculations since grid cells need to be highly
clustered in the vicinity of physical boundaries, and a
very small time step has to be employed to avoid the
violation of the stability criterion if an explicit method
is adopted. However, a major drawback with low-
order implicit methods is that numerical dispersion
and dissipation are high. The Euler implicit method
is only first-order accurate in time. When compared
with traditional CAA approaches, which are fourth-
order accurate in time if a classical Runge-Kutta time
integration technique is used, the wave propagation
properties of implicit methods are indeed inferior.
In the current analysis, a high-order accurate dual
time-stepping algorithm is proposed as a viable al-
ternative for viscous aeroacoustic calculations. This
algorithm is formally second-order accurate in time
and sixth-order accurate in space.
In order to perform acoustic calculations, all the
transients in the mean flows have to be eliminated from
the computational domain. A time-independent mean
flow solution is obtained from an efficient steady-state
solver in the current approach. This mean flow so-
lution is then used as the initial condition, and an
912