Professional Documents
Culture Documents
DFG CNRS Introduction 2
DFG CNRS Introduction 2
1
Table of Contents
1. Preliminary Remarks ..............................................................................................................3
2. General Information ...............................................................................................................4
2.1 Applicants of the Research Group....................................................................................4
2.2 List of Subprojects............................................................................................................8
3. Motivation ..............................................................................................................................9
4. Main Achievements in the First Funding Period..................................................................10
4.1 Scientific Achievements .................................................................................................10
4.2 General Activities ...........................................................................................................13
4.3 Guest Program and PhD Student Exchange ...................................................................14
5. Objectives and Structure of the Project ................................................................................15
5.1 Overview ........................................................................................................................15
5.2 Jet Noise .........................................................................................................................16
5.3 Numerical Prediction of Fluid Flow...............................................................................17
5.4 Numerical Methods for Noise Propagation ....................................................................18
5.5 Fluid/Acoustic-Coupling ................................................................................................18
5.6 Validation and Verification ............................................................................................19
5.7 Structure of the Project ...................................................................................................19
6. Cooperation of FOR 508 ......................................................................................................20
6.1 Flow Problems................................................................................................................20
6.2 Methodology...................................................................................................................21
6.3 Cooperation and Communication Structure within the Research Group FOR 508 .......23
6.4 Cooperation with DFG-CNRS Research Group FOR 507 “LES of Complex Flows” ..23
7. Cost Statement......................................................................................................................24
2
1. Preliminary Remarks
Within this research group novel noise simulation concepts for turbulent flow are considered.
Since the analysis of the acoustic field strongly depends on the quality of the resolution of the
turbulent flow field we use Large Eddy Simulations (LES), hybrid RANS-LES approaches
and Direct Numerical Simulations (DNS) as a basis to predict noise generation and
propagation. The aim of this research is to re-investigate, compare, and improve aeroacoustic
approaches to find proper ways into the LES/DNS age of noise calculations, to get insight into
the basic noise generation mechanisms, and to find novel noise reduction concepts.
Furthermore, the research aims at intensifying the link between the different disciplines of
Computational Fluid Dynamics (CFD) and classical acoustics via the relatively new domain
of Computational Aeroacoustics (CAA).
Jet flows at different subsonic and supersonic flow regimes will be the common class of flow
problems to develop and improve. The prediction of noise generated by a turbulent jet
remains a very challenging problem for numerical simulation methods. The fact that turbulent
jets are widespread among different industrial applications reinforces the choice at different
regimes as basic problem for the research group. The requirements of sharper limits for noise
radiation, e.g., of turbine engines will be one of the key challenges for the aircraft and engine
industry in the near future and even decide on a companies success in the next years.
An important aspect of this French-German research group is the exchange of knowledge and
the intensification of cooperation between scientists in a more European spirit. These French-
German links will continue to be strengthened in the second funding period, when an even
stronger cooperation between both DFG-CNRS research groups FOR 507 “LES of Complex
Flows” and FOR 508 “Noise Generation in Turbulent Flows” will be established. The
cooperation in this joint research project is based on intensive data and information exchange
and mutual validation of the different approaches for common reference benchmark problems.
This DFG/CNRS program also aims at supporting and bringing together PhD students from
France and Germany. Also in the second funding period of the research group the French and
German students are again supposed to work for a limited time at the institutes of their mutual
partners. The objective of noise reduction, which is an important issue in industry and our
society, offers the PhD students involved in the DFG/CNRS research groups interesting and
promising perspectives for their future careers. Furthermore, the French-German cooperation
will fortify the scientific value gained from the research. That is, a greater progress can be
expected to be achieved due to the mutual motivation within each group and hence, this two-
country program will be more efficient than solely national projects.
3
2. General Information
German Applicants
Prof. Dr. Claus-Dieter Munz (Coordinator of the research group, German side)
Institut für Aerodynamik und Gasdynamik
Universität Stuttgart
Pfaffenwaldring 21
70550 Stuttgart
Tel.: +49 (0)711 / 685-3433
Fax.: +49 (0)711 / 685-3438
e-mail: munz@iag.uni-stuttgart.de
4
Prof. Dr. Ulrich Rist
Institut für Aerodynamik und Gasdynamik
Universität Stuttgart
Pfaffenwaldring 21
D-70550 Stuttgart
Tel.: +49 (0)711 / 685-3432
Fax: +49 (0)711 / 685-3438
e-mail: rist@iag.uni-stuttgart.de
French Applicants
Prof. Dr. Daniel Juvé (Coordinator of the research group, French side)
Centre Acoustique
Ecole Centrale de Lyon
36, av. Guy de Collongue
F-69134 Ecully, France
Tel. : +33 4 72 18 60 12
Fax. : +33 4 72 18 91 43
e-mail: daniel.juve@ec-lyon.fr
5
Dr. Christophe Airiau
Institut de Mécanique des Fluides de Toulouse (IMFT)
1, Allée du Professeur Camille Soula
F-31400 Toulouse
Tel.: +33 5 61 55 62 53
Fax.: +33 5 61 28 59 33
e-mail: christophe.airiau@imft.fr
6
Prof. Dr. Pierre Sagaut
LMM/UPMC, Boîte 162
4, place Jussieu
F-75252 Paris cedex 5, France
Tel. : +33 01 44 27 54 68
Fax. : +33 01 44 27 52 59
e-mail : sagaut@lmm.jussieu.fr
7
2.2 List of Subprojects
8
3. Motivation
Today the reduction of noise plays an important role for human wellness und health. The
aeroacoustic noise is the central environmental issue for a wide range of technical
applications, spanning the spectrum from aircraft with helicopter, propeller, and
turbomachinery noise, noise from automobiles and trains to noise generated by cooling fans
of electronic equipment and air-conditioning devices.
Currently, we are still in the initial stages of development of computational fluid dynamics
(CFD) based tools for the numerical simulation of flow generated noise. With increasing
confidence in the use of CFD, numerical simulation plays a greater role as a tool for research
and development to predict the mean flow field. Detailed direct numerical simulations (DNS)
are limited to relatively low Reynolds numbers and simple flow geometries even with today’s
high power computers. However, large eddy simulations (LES) with their elegant
characterization of flows will become increasingly attractive for industry in the near future.
Computational aeroacoustics (CAA) can be considered a new discipline that tries to close the
gap between CFD and classical aeroacoustics. The opportunity of fully exploiting the great
advances in scientific computing offers a realistic hope of substantial improvements in this
area. CFD and CAA have the potential to identify the significant fluid dynamic events that
contribute to the generation of sound, to guide the way to noise control techniques, and
thereby to reduce noise radiation.
The overall objective of the projects proposed is refined investigation of pioneer novel noise
calculation concepts based on LES/DNS flow calculations and the simultaneous simulation of
noise generation and its propagation from the near to the far field. This is neither theoretically
nor numerically a straightforward extension of CFD into the acoustics area. It is
interdisciplinary research in mathematics, fluid dynamics, and acoustics. With the
investigation of noise generation of subsonic jets classical aeroacoustics has been started. In
the following projects jets in different subsonic and supersonic flow regimes have also been
chosen as the common problems for research. Although the growth of the bypass ratio of
turbofan engines has resulted in substantial reductions in total engine noise with improved
performance, there is no doubt that additional noise reductions, whether or not further
increase in bypass ratio is achieved, will depend on additional reductions in jet exhaust noise.
There is a large demand in research for the control of jet noise, because the limits for the noise
generation of aircraft will be further reduced in the future.
The classical approaches focused on very special and generally simplified model problems.
For this reason, it is a must from a practical stand point to reinvestigate theoretically and
numerically noise generation mechanisms, i.e., differing noise sources, especially when
broadband noise plays a major role. Various approaches will be pursued in this research
initiative ranging from two-step CFD/CAA methods to direct simulations of noise. An
especially close interaction of all the considered methods with the direct numerical simulation
of noise will be the central linking element of this jet noise research group.
9
4. Main Achievements in the First Funding Period
In the following some highlights of the research projects are briefly discussed. It is fair to
state that the key objectives of the first funding period of the research group FOR 508 have
been achieved. More details and a more profound presentation of the results of each
subproject are summarized in the project reports.
4.1.1 Subproject 1
The aim of mechanistic understanding of jet noise generation has been realized by low-
dimensional Galerkin modeling of LES data from SP1 and SP6. A systematic modeling
procedure is proposed and comprises four key enablers: a detailed modal energy flow
diagnostic, a novel pressure-term representation, a refined “subgrid” turbulence representation
and a new scheme to construct 'missing' dynamic modes. Thus, the dynamic bandwidth of
POD Galerkin models is significantly increased for computationally inexpensive parametric
studies and for numerically accessible flow control design. An analytical reduced-order
framework of the far-field acoustics has been developed and will soon be refined in
collaboration with SP3 and SP7.
4.1.2 Subproject 2
4.1.3 Subproject 3
The objective of developing very accurate noise propagation algorithms has been achieved
during the first funding period. Memory-saving algorithms for the solution of hyperbolic
systems and wave equations with arbitrary order of accuracy in space and time on structured
and unstructured grids in two and three space dimensions have been developed. This was
attainable by combining discontinuous and continuous finite elements with a novel time
discretization technique that makes extensive use of the governing partial differential
equations. For the sake of computational efficiency, a heterogeneous domain decomposition
approach has successfully been introduced, which is able to couple different equations, grid
topologies, numerical schemes, and even different time steps maintaining a uniformly high
order of accuracy in space and time in the whole computational domain. The new acoustic
solver is ideally suited for simulating wave propagation in very complex geometries due to its
high accuracy on unstructured grids. It has been applied in SP5 to calculate the noise in the
10
near far field generated by a shear flow and will be used in several other subprojects such as
in SP6 to compute the noise of the co-axial jet in the new funding period. Beside the
development of numerical schemes for the noise propagation the coupling of the fluid flow to
the acoustics in low Mach number flow has been considered. Based on low Mach number
asymptotic analysis a perturbation method has been proposed that scales the EIF method in an
appropriate way. Here the linearization is motivated with respect to the leading order terms in
the Mach number and allows a straightforward extension to non-isothermal flow.
4.1.4 Subproject 4
As the main part of this project is to develop a sub-grid scale model, which determine the
distribution of the fluctuations that are introduced into the acoustic source terms, we started to
analyze the Lighthill´s well-known analogy with the Lie group method. Several assumptions
were made to adapt this equation to the test case that we studied. The viscosity effects can be
neglected. This can be demonstrated by the Reynolds number as a scale argument, so that the
contribution of the viscous stresses to the total acoustic source term is unimportant. In this
research period the entropy effects are not taken into account as the study focused on a cold
fluid at a constant temperature.
We applied the Lie group method to Lighthill´s analogy and studied two cases: the first one
was Lighthill´s analogy without source term and the second one with a source term. The
results showed reduced set of symmetries for the second case and that they represent only
some part of the symmetries for the Lighthill´s analogy without a source term. One of the
main difficulties in both cases is to find the physical interpretation. To achieve this, other
acoustic analogies have to be analysed and compared with the above mentioned results.
The numerical work includes a cold jet simulation at different Reynolds numbers (Re=1000
und Re=3000) each having a control volume of 400,000 cells. Here, simulations have been
done with the LES incompressible code. Next, more jet computations for high Reynolds
numbers will be performed. Furthermore, other acoustic analogies are studied and compared
with each other, to decide which subgrid scale model which is best for a direct and a hybrid
acoustic simulation.
4.1.5 Sub-Project 5
The German group was able to develop the necessary computational tools for aeroacoustic
computations of a compressible mixing layer. In addition to a DNS code, programs to
generate appropriate initial conditions and to compute eigenmodes of linear stability theory
were needed. The newly developed DNS code can be applied to a wide range of compressible
flows because it relies on domain decomposition using blocks with boundary conditions that
can be freely chosen and because it contains more grid stretching capabilities than the one
used before. A high computational efficiency and good use of the available vector computer
architecture resulted in remarkable performance rates. Using this code, benchmark problem
BP1 of the research group, a two-dimensional DNS published by Colonius, Lele and Moin
[1997], was computed. It consists of an isothermal subsonic mixing layer with Mach numbers
of 0.5 and 0.25 for the upper and lower stream, respectively and a Reynolds number of 250
based on the vorticity thickness at the inlet and the difference in streamwise velocity. The
resulting amplification rates in the initial part of the mixing layer match those predicted by
linear stability theory. Also the emitted sound field showed good agreement with Colonius et
al. Three-dimensional computations and grid refinement studies of the same flow
configuration followed. During the CEMRACS summer school in 2005 an online coupling of
the DNS code with the acoustic code from SP3 has been implemented. This will allow to
compute the acoustic far-field of the mixing layer.
11
The French partner has built a numerical tool for the adjoint 2-D Navier-Stokes equations,
and made first simulations on a standard test case consisting of a mixing layer and a
sinusoidal point source for the adjoint. The results of the adjoint analysis for the mixing layer
are qualitatively similar to the ones obtained by Cerviño et al. (2002) for a jet. In particular,
considerable wave refraction phenomena have been found, the flow is more sensitive to
forcing of the pressure than to density, and the spectra of the adjoint get broader in particular
regions on the opposite side of the mixing layer.
4.1.6 Subproject 6
During the first period several steps were successfully carried out to develop a tool for jet
noise computations exploiting a hybrid approach based on a large-eddy simulation (LES) for
the compressible flow field and acoustic perturbation equations (APE) for the acoustical field.
In a first step, a simpler form of the right-hand side of the APE system was derived containing
the formulation of one single source term for vortex sound. On the inner boundaries between
different matching blocks covering the LES and the acoustic domain, spurious noise
generated by discontinuities in the vorticity distribution was suppressed by a damping zone.
Furthermore, the application of the APE system to jet flow problems was validated by various
benchmarks considering the relevant properties of jet like flows, i.e., noise generation in
sheared mean flows and vortex sound. An LES/CAA interface was developed that is capable
to extract the source terms from the compressible flow field data on structured, multi-block
grids and interpolates them on a coarser acoustic grid. Subsequently, a fully three-dimensional
jet at the flow conditions of benchmark BP2 was computed. The LES simulation was shown
to be in very good agreement with experimental measurements as well as with numerical
results from a direct simulation at the same flow conditions. The acoustical near field results
matched very well those from investigations based on a direct LES approach by Bogey and
Bailly indicating that all dominant noise sources could be captured. More precisely, both
acoustic spectra, downstream and on the sideline of the jet showed the same narrow-banded
pressure spectra, the same peak frequency, and similar decaying rates for higher frequencies
as is characteristic for such low Reynolds number jets. The findings of the computation of the
fully turbulent high Reynolds number jet evince the impact of the smaller turbulent structure
on the directivity of the jet.
4.1.7 Subproject 7
Within the first funding period of the DFG-CNRS research group FOR 508 „Noise generation
in turbulent flows” the work on the German and French side of subproject 7 “Numerical
simulation of jet mixing noise associated with engine exhausts” focused on the rod-airfoil test
case. The German side performed compressible detached-eddy simulations (DES) employing
different turbulence closures of this test case. During the simulations flow field data required
to calculate the far-field sound were extracted. The extraction procedures have been extended
to match the requirements of the French integration procedure. On the French side additional
measurements were carried out. Additionally, the acoustic analogy integration procedure was
extended to include the evaluation of volume sources. The results obtained by joint
DES/acoustic analogy-procedure were found to agree well with the experiments but depend
on the employed turbulence closure. As the quality of sound prediction depends highly on the
quality of the flow simulation, further extensions to the flow solver have been designed.
These consist of DES based on an explicit algebraic stress model as turbulence closure and
the implementation of non-reflective boundary conditions to avoid spurious reflections from
the boundaries of the computational domain. The extensions made are applied in ongoing
simulations of the rod-airfoil test case and the simulation of a subsonic single stream jet.
12
4.2 General Activities
• Kick-off meeting of both research groups in Munich, together with the French partners
(2003)
• meeting of the project leaders of the German side with their French partners of the
noise group in Stuttgart (2004)
• one-week workshop for the PhD students followed by a meeting of the project leaders
of both research groups (French and German side) in Orléans (2004)
• meeting of the German sides of both research groups in Stuttgart (June 10th 2005)
• several bilateral meetings
13
4.3 Guest Program and PhD Student Exchange
Guests
Prof. Dr. T. Toro (University of Trento, Trento, Italy): 19. 4. 2005 in Stuttgart
Derivative Riemann Solvers
Prof. Dr. Joseph Liu (Brown University, Providence RI, USA): 7. 5. 2004 in Stuttgart
Aerodynamic Sound Generation Revisited
Prof. Dr. Xiaodong Li (Beijing University, China): 17.-19.5.04 und 18.8.-23.9.2005 in Berlin
Numerical Simulation of the Generation Mechanism of Axis-Symmetric Supersonic Jet
Screech Tones
Dr. Mei Zhuang (Michigan State University, East Lansing, USA): Sept. 2003 - July 2004 as
Fulbright Scolar in Berlin
Lectures in Computational Aeroacoustics; CAA: Methodology and Application
Prof. Dietmar Rempfer (Illinois Institute of Technology, USA): 11.-18.5. 2003 in Berlin
POD Galerkin modelling of free shear flows: Opportunities and challenges
Prof. Mohammad Samimy (The Ohio State University, USA): 13. 2. 2004 in Berlin
An experimental exploration of correlation of dynamics of large scale structures and far field
acoustics in high speed jets
14
Diploma and PhD Student Exchange
Michael Dumbser (Uni Stuttgart) in Strasbourg: 17. 11. 2003 - 30. 11. 2003
Sébastien Jund (Uni Strasbourg) in Stuttgart: 02. 08. 2004 - 15. 08. 2004
Andreas Gurr (TU Berlin) in Lyon: 15. 05. 2005 - 20. 05. 2005
Björn Greschner (TU Berlin) in Michigan 01. 10. 2004 - 20. 12. 2004
Bruno Spagnoli (IMFT Toulouse) in Stuttgart: 01. 04. 2004 - 30. 04. 2004
Anas Osman (TU Darmstadt) in Aachen: 19. 07. 2004 - 23. 07. 2004
Maximilian Dietrich (TU München) in Lyon: Diploma thesis 2004
Thomas Emmert (TU München) in Lyon: Diploma thesis 2004
The central focus topic of the second three-year phase is the coupling and improvement of the
tools to numerically simulate turbulent jet flow developed within the research group FOR 508
during the first funding period. The research in the second three-year period is still
fundamental research, but it also aims at closing the gap between basic analyses and industrial
challenges and needs. To accomplish this objective a means for computing even more
complex flow fields will be provided. With our knowledge gained in the first period the
activities in the second three-year phase will be directed to the simulation of more complex
and industrially relevant jet flows. Among other flow problems fully turbulent unheated and
heated jets with and without co-flow at various Reynolds and Mach numbers will be tackled.
Finally, nozzle geometries with, say, streamwise vortex generators, or other devices that
influence the development of the shear layer, such as a downstream located ring, are to be
considered. Since very high-order unstructured acoustic solvers are available from the work in
the first phase, these complex geometries do not pose any problems for the acoustic solvers.
The cooperation with our joint research group FOR 507 “LES of Complex Flows” will be
intensified by jointly computing the turbulent flow field and the acoustical field over a
complex aerodynamic geometry. More precisely, the flow around a high lift configuration
will be computed by the hybrid approach and direct methods.
For the second funding period close cooperation and interaction with experimentalists will
also acquire significance and therefore will be established within the research group.
Especially in SP7 dealing with geometrically complex engine exhausts it is intended to
include experiments. On the one hand, measurements from different working groups already
participating in this research initiative are taken into account and on the other hand, data from
groups outside the research establishment will be exploited for comparison with the numerical
simulation. The composition of the research group well established in the last three years
guarantees close cooperation.
5.1 Overview
The numerical simulation of noise generated by a turbulent flow may be divided into four
main topics. The first is the proper calculation of the turbulent flow field. In the case of jet
noise research this can be achieved only through time dependent numerical simulation. Here,
the method of choice is LES and in some cases also DNS. Even with in depth experience in
LES over the last decade there are still a number of open questions that have to be addressed,
such as the proper subgrid scale modeling, the convergence to the direct numerical
simulation, and especially, the interaction between the fluid flow and the acoustic wave
propagation.
15
The numerical prediction of acoustic wave propagation within the fluid domain up to the near
far field, i.e. over large distances forms the second care research topic. In Lighthill’s classical
acoustic analogy the noise propagation is mathematically modeled by the wave equation in a
flow at rest which is valid in the far field and with source terms concentrated in the relatively
small flow region. Often the calculations are performed in the frequency domain. In the
vicinity of the actual unsteady flow field this model has to be replaced by acoustic equations
with a non-constant background flow applied in the time domain. In acoustics numerical
schemes are designed to reproduce acoustic wave propagation over long distances with low
dispersion and dissipation errors on coarse grids and are different from those used in CFD.
The coupling of fluid flow and acoustics may be considered to be the third topic of research.
The decomposition of the flow field into a hydrodynamic and an acoustic component as used
in an analogy still has to be considered at subsonic or supersonic Mach numbers. Stable
extensions of the acoustic analogy valid in the entire flow region, i.e., including the transonic
regime, are desirable.
The final research topic is related to active noise control mechanisms. Therefore, the adjoint
linearized Navier-Stokes equations are solved in order to perform a sensitivity analysis, which
provides the information in which regions of the flow field active or passive control devices,
such as, e.g., actuators or a specially shaped geometry, respectively, must be placed in order
to be most efficient with respect to decreasing the sound pressure level at a certain point or
region. This kind of method is applied within two of the subprojects.
Within the proposed projects these different issues are addressed. In all topics substantial
improvement is necessary to simulate turbulent fluid flow together with the sound generation
and propagation. Renewed efforts in computational aeroacoustics are essential for new noise
simulation concepts, which will yield improved low noise design strategies and will be the
basis for a competitive advantage for industry.
One of the basic benchmark flows in aeroacoustics is a round turbulent jet. The first papers
dealing with this type of flow appeared in the fifties and sixties. Later on a lot of experimental
work on the jet flow was performed.
The generation of noise from turbulent jets is of significant interest for civilian jet powered
aircraft. As a result of progress in noise reduction today's commercial airplanes are as much as
20 dB quieter than their former counterparts. Much of this remarkable achievement is due to
the use of high bypass ratio turbofan engines. Although the increase of the bypass ratio has
generated pronounced reductions in total engine noise, there is a technical limit on the bypass
ratio. In other words, further noise reductions will be determined by minimizing jet exhaust
noise. To achieve this goal the large scale structures in the flow, which are often loosely
referred to as high energetic structures, have to be modified or other structures have to be
generated to enhance entrainment and mixing. Although different vortex generators have been
introduced in various applications the details of vortex generation, vortex interaction, and
mixing scales and thus of noise reduction are still not understood especially when one thinks
of heated co-axial jets.
In recent times there have been a number of numerical studies on supersonic jet noise using
LES or even DNS. These CFD calculations are coupled to those for acoustic wave
propagation into the far field using the Kirchhoff method or another approach based on
classical acoustic analogies. The simulation of noise generation by subsonic jets is more
difficult than that in the supersonic regime. Subsonic jets are relatively quiet and the effect of
16
computational noise is far more critical. The ratio of acoustic to mechanical energy flux scales
at low Mach numbers with M5 and at higher Mach numbers with a lower exponent. Since the
small eddies may generate an important part of the noise an LES becomes very sensitive with
respect to the size of the grid cells and the subgrid modeling. Lighthill’s theory has been
developed for this case of fluid flow in the 60s and has been the dominant theory almost up to
the present day. Recent papers, however, indicate that the prediction of the overall sound
pressure level is very susceptible to the size of the source region, especially the downstream
located arc. In other words, the maximum sound pressure values are extremely user dependent
and as such this acoustic analogy can be used to predict trends but is no basis to analyze noise
generation mechanisms.
The numerical simulation of subsonic and supersonic jet noise is still a computational
challenge for high performance computing. The jet noise of supersonic jets consists of
different components. One component radiates principally in the downstream direction and is
consistent with Mach wave radiation from the large turbulence structures of the jet flow.
These structures correspond to instability waves, which generate wave packets. Another
component with uniform directivity is the noise from the fine-scale turbulence. Additional
noise may be generated by shocks in an imperfectly-expanded jet due to the interaction of the
large scale jet turbulence with the shock-cell structure, called jet-screech when feedback is
present. Since there are still many open questions in the context of noise reduction for
turbulent jets and it is of general environmental interest, jet noise is the core problem of this
research group.
Direct numerical simulation of turbulent flow, which numerically resolves all the scales of the
flow, is limited to relatively low Reynolds numbers and simple flow geometries even with
today’s high power computers. Engineering flow computations for complex geometries are
usually based on the Reynolds-averaged Navier-Stokes equations (RANS) calculating time or
ensemble averaged mean quantities in the flow. Relatively coarse grids can in general be
used, because the influence of all turbulent scales is approximated via a statistical turbulence
model. But RANS calculations only provide local statistical properties and do not give direct
information about the acoustic source term. A good compromise between the detailed time
dependent DNS calculations and the mean flow RANS calculations is given by the large-eddy
simulation. A spatial filtering procedure is applied and a model is needed to account for the
influence of the subgrid scale motions. The advantage of LES is that the time dependent large
scale motions of the flow are resolved, not modeled. Further increase in computational power
or improvement of simulation methods will lead to better predictions of turbulent flow. At the
moment, however, LES cannot be considered an engineering design application in CFD. A lot
of insight has been gained in the past and significant advances have been made, but there is
still a good deal of research to be done. To mention just a few topics, efficient high-order
methods on arbitrary meshes are necessary and quality measures as well as theoretical results
still have to be provided that indicate convergence to a DNS for finer and finer grid
resolution.
If large eddies, i.e., large in the sense of the inertial subrange, generate the main sound, noise
calculations based on LES seem to be well motivated. However, when the details of small
scale turbulence influence the sound production, the impact of subgrid scale modeling on
sound generation has to be considered.
With respect to approximation errors aeroacoustics imposes other severe restrictions on CFD
calculations. Since the amplitudes of the acoustic fluctuations are often much smaller than
17
hydrodynamic fluctuations, there is the danger that the noise generated by the numerical
approximation errors is larger than the physical noise, if appropriately accurate numerical
methods are not used.
Within this French-German research group, some projects focus on the aspects of the CFD
calculations to provide their essential findings to partner projects that deal with the prediction
of the acoustical field. As in the first period, the whole range of research issues is supposed to
be covered in the second phase: DNS of the flow field together with sound generation and
propagation as well as large-eddy simulations being coupled to acoustic calculations.
Furthermore, theoretical considerations with respect to the modeling of small scale turbulence
complete the research.
The next challenge of aeroacoustic simulation is noise propagation within the time domain.
Numerical schemes for acoustics are designed to predict wave propagation over long
distances with low dispersion and dissipation errors on coarse grids. They are different from
those used in CFD. For instance, acoustic pressure fluctuations can be 60dB below those of
interest to fluid dynamicists, with frequencies in the 100 Hz to 10,000 Hz range.
A standard numerical method for the prediction of wave propagation is the dispersion relation
preserving (DRP) method, a finite difference method where the coefficients are chosen to
preserve the basic wave properties. The tendency to instability, which can be observed in
practical calculations, is simply suppressed by artificial dissipation. New concepts such as
ADER finite volume or discontinuous Galerkin finite element methods as well as continuous
finite element approaches, which are more appropriate for arbitrary grids, are under
consideration by this research group. These schemes can be designed to maintain the positive
properties of the DRP approach and yet be more robust on highly clustered meshes. This is
definitely of interest, when strong shear layers such as in high Reynolds number co-axial jets
occur in the flow field or complex geometries.
5.5 Fluid/Acoustic-Coupling
18
extension of the acoustic analogy to the flow region may be based on perturbation methods.
At low Mach numbers the perturbations are the compressible corrections of the
incompressible base flow.
At moderate or higher Mach numbers the decomposition of the flow field into a
hydrodynamic and an acoustic part becomes difficult. Acoustic waves are simultaneously
transported by the CFD solver and the acoustic equations. An alternative to decomposition or
hybrid approach is to perform highly accurate CFD calculations including the generation and
propagation of sound waves in the compressible flow region without any analogy. The far
field sound levels can then be deduced from surface methods that relate the acoustical field to
integrals over a closed surface that surrounds all acoustic sources (Kirchhoff method). Both
approaches are studied by the research group. Different flow decompositions are proposed
and will be compared with theoretical, DNS, and classical methods based on acoustic
analogies.
The methods and tools developed in the first funding phase will be validated using apart from
basic benchmark tests also problems in a parameter regime of technical relevance. These
problems are a mixing shear layer and subsonic and supersonic jets at laminar and fully
turbulent conditions in the engine exit. Experimental and numerical results are available for
these problems from different working groups of this research program as well as from the
literature such that in depth analyses of the fluid mechanical and acoustical phenomena are
feasible.
The research project FOR 508 “Noise Generation in Turbulent Flows” contains seven
subprojects, which are each investigated by a French-German research group, who have
jointly formulated the proposal of each subproject for the next phase under one common title.
The different research topics of the French and German partners are highly synchronized. To
cover a broad range and many aspects of turbulent jet flow, various research issues are
investigated by several project teams. Thus, there is a need for a pronounced cooperation
among all French and German groups. Three groups focus on jet flow simulation. The flow
regimes considered range from subsonic to supersonic. Four groups perform DNS of subsonic
mixing layers, and subsonic and supersonic jet-related flows to provide fundamental solutions
of the proposed common test problems for the other groups. Some of these partners consider
LES simulations with some fluid/acoustic coupling and a standard scheme for the acoustic
wave propagation. Further investigations focus on the understanding of noise generation,
numerical modeling of the fluid/acoustic coupling and acoustic wave propagation. The proper
acoustic wave propagation into the far field is considered using finite element and finite
volume schemes with good wave simulation properties. The immediate exchange of the flow
data gives these groups the possibility to scrutinize and compare different acoustic
approaches. Furthermore, the results of the “acoustic research groups” such as algorithmic
developments and boundary condition formulations will be directly used within other projects
to improve their computational acoustics method.
19
6. Cooperation of FOR 508
As outlined in the structure of the research program different groups deal with various topics
of research that cover numerous aspects of noise generation in turbulent jet flow. The inner
and outer cooperation, which has been established in the first period, will be continued into
the next phase. The direct collaboration with the corresponding French partner within each
subproject represents the inner cooperation. This inner cooperation is strongly emphasized by
the common French-German definition of the research topics. The mutual exchange of data
and CFD / CAA codes within the complete research group FOR 508 "Noise Generation in
Turbulent Flows" is the basis of the outer cooperation.
The list of basic test problems, that have been defined and have been the common milestones
to compare and discuss the numerical results, will be extended for the next phase. The
problems of the first period have covered different flow regimes and different applications.
The team members of each subproject have chosen appropriate flow problems to establish a
sound basis for validation and comparison in various flow regimes. Some of tests are still
valid for some of the subprojects in the next phase. They are listed as follows:
- BP2: A transitional circular jet at a Mach number M=0.9 and a Reynolds number
Re=3600 is considered. Experimental and numerical data are available from the
investigations by Stromberg et al. and Freund. Furthermore, a LES is performed by the
group of Pierre Comte.
The flow problems for the new period contain on the one hand, still basic flow phenomena
that are of interest to emphasize the strengths and weaknesses of the various approaches and
on the other hand, bridge the gap to more industrially relevant configurations. Moreover, a
test problem, jointly defined with the research group FOR 507 “LES of Complex Flows” is
given to substantiate also the cooperation between the turbulence and the acoustic groups:
- BP4: The geometry of a coaxial jet defines this test problem. That is, there is a
primary or inner jet, which is covered by a bypass stream. The configuration and the
flow parameter are related to the flow problem in the European CoJeN program.
Hence, the Reynolds number will be of the order of 105 and the Mach number 0.9.
Two configurations are defined, one with a cold and one with a heated inner jet.
20
- BP5: For engine exhaust with or without chevrons the noise control by flat rings in the
mixing layer will be considered as proposed by Roger & Serafini in their AIAA paper
2005-2958. Following the experiments the Reynolds number of the jet will be 4x105.
The aim of the configuration is to examine the interaction of turbulence with walls.
The strong interaction between the partners of this research group, the data and knowledge
transfer and the benchmark problems computed within each subproject are illustrated in Fig.2.
6.2 Methodology
One the one hand, the inner and outer cooperation are supported by jointly investigating the
aforementioned test problems. On the other hand, this cooperation is also substantiated
through the application and exchange of likewise methods and approaches. This collaboration
on the methodological level concerns for instance the discontinuous Galerkin methods, the
acoustic perturbation equations, the sub-grid scale modelling, and the adjoint equations to
mention just a few of these common theoretical aspects.
21
SP2: Sesterhenn, Juvé, Bailly,
Schmid DNS data
BP: BP3
F: LES for rectangular SP3: Munz, Sonnendrücker,
SP1: Dillmann, Noack, Comte, nozzle, shape variation Frenod
Farge, Schneider D: DNS for BP3, nozzle BP: BP1, BP3, BP5, BP6
BP: BP2,BP4,BP5 outlet shape optimization F: High order FE schemes,
F: DNS, postprocessing and for noise reduction, multi-scale asymptotics
mathematical reduced structure identification for source term evaluation
order modelling
D: High order unstructured
D: Low dimensional models
DG and FV schemes,
for jet noise Acoustic solver / Galerkin models as source input
domain decomposition
equations, CAA mechanisms for shear F : Main research on the French side
efficiency, sound layer noise control, D : Main research on the German side
22
6.3 Cooperation and Communication Structure within the Research Group FOR 508
To ensure intensive communication and substantial mutual information exchange between the
members of the DFG/CNRS research group the successful structural and organizational
measures of the first period are carried over into the second period.
That is, regular meetings of the project directors and the participating scientists are planned to
take advantage of the expertise within the group. The website is www.iag.uni-
stuttgart.de/DFG-CNRS and will continuously be updated to provide a fast, easy, and efficient
communication interchange concerning the organization as well as the latest scientific
achievements of each subproject.
Besides regular meetings within each subproject, the following activities will be organized by
the research group in the second phase of the DFG-CNRS research program:
- Two international workshops
- Guest program
Two international workshops will provide the possibility to discuss and exchange the results
of every subproject and to receive fruitful exterior impulses. International specialists will be
invited to document the state of the art of research, to judge the findings of the research group,
and to stimulate new ideas. These workshops are open to participation from outside the
research group.
The guest program will be continued, therefore, also in the next period it will be possible to
invite experts in the field of computational aeroacoustics, who are interested in cooperating
with different subprojects of the group.
6.4 Cooperation with DFG-CNRS Research Group FOR 507 “LES of Complex Flows”
The numerical methods for fluid acoustic coupling and the acoustic wave propagation are
developed such that they may also be applied to LES data in complex geometries at the end of
the second funding period. The flow problem BP6 has especially been introduced as a
common test case together with our partner research group FOR 507 “LES of Complex
Flows”.
The high-lift configuration (BP6) is defined to be a problem for both research groups. The
high-order unstructured acoustic solvers developed in the group FOR 508 are able to deal
with complex geometries very easily since unstructured mesh generation is well understood
and can be done in an almost completely automatic manner. As far as the flow solvers are
concerned, the partners of research group FOR 507 will come up with numerical methods,
which are able to deal with this kind of complex geometries. They will deliver reliable LES
results, which can be used to feed the high-order acoustic solvers. Since the high lift
configuration poses a low subsonic flow problem it is well-suited for the hybrid approach.
That is to say, the LES data of the analyses from the FOR 507 research group will be used as
input data for the acoustic perturbation equations (APE) already developed in research group
FOR 508 and other standard acoustic equation systems such as the linearized Euler equations
or the wave equation.
23
7. Cost Statement
The following table summarizes the estimated costs on the German side per year and in total
for the next three years of the French-German research program. The costs of the French side,
are assumed to be covered by CNRS. The details are given in each subproject.
SP2
Sesterhenn, 1 BAT IIa 1 assistant 2500 ¼ 500 ¼ 3000 ¼
Juvé, Bailly, Schmid (40 h/month)
SP3
Munz, Sonnendrücker, 1 BAT IIa 2 assistants 2500 ¼ 500 ¼ 3000 ¼
Frénod (40 h/month)
SP4
Oberlack, Sagaut 1 BAT IIa 1 assistant 2500 ¼ 500 ¼ 3000 ¼
(40 h/month)
SP5
Rist, Kloker, Airiau 1 BAT IIa 1 assistant 2500 ¼ 500 ¼ 3000 ¼
(40 h/month)
SP6
Schröder, Comte 1 BAT IIa 1 assistant 2500 ¼ 500 ¼ 3000 ¼
(40 h/month)
SP7
Thiele, Jacob 1 BAT IIa 1 assistant 2500 ¼ 500 ¼ 3000 ¼
(40 h/month)
Additional costs:
Total costs for the research group over three years without personnel: 123000 ¼
24