You are on page 1of 7

See

discussions, stats, and author profiles for this publication at:


https://www.researchgate.net/publication/318760595

RANS simulation of over-and under-expanded


beveled nozzle jets using OpenFOAM

Conference Paper · July 2017

CITATIONS READS

0 144

3 authors, including:

U S Vevek T. H. New
Nanyang Technological University Nanyang Technological University
3 PUBLICATIONS 0 CITATIONS 131 PUBLICATIONS 786 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Jet flow control and mixing enhancements View project

Novel non-intrusive measurements and simulations of supersonic flow and noise


View project

All content following this page was uploaded by T. H. New on 29 July 2017.

The user has requested enhancement of the downloaded file.


RANS simulation of over- and under-expanded beveled nozzle jets
using OpenFOAM
Zang B., U S Vevek and New T. H.*
School of Mechanical and Aerospace Engineering, Nanyang Technological University,
50 Nanyang Avenue, Singapore 639798
*
Corresponding Author’s email: dthnew@ntu.edu.sg

Abstract The present study numerically investigates supersonic jet flows issued at M=1.45
from a beveled nozzle with 60o inclination. Utilizing rhoCentralFoam solver in OpenFOAM,
unsteady Reynolds-averaged Navier-Stokes (RANS) simulations were performed with two
nozzle pressures ratios (NPR) of 2.8 and 4, corresponding to over- and under-expanded exit
conditions respectively. The Mach distributions reveal a more organized near-field shock
formation for the over-expanded jet with smaller and periodic shock cells, as compared to
those of the under-expanded one. Moreover, the jet flows are deflected in opposite directions
for the two different NPRs, indicating a strong dependence of jet vectoring on NPR in
addition to the bevel angle. Last but not least, reasonably good agreements can be observed
for both qualitative and quantitative comparisons between simulation and experimental
results, supporting the notion that the compressible rhoCentralFoam solver in OpenFOAM is
suitable to model such supersonic jet flows.

1 Introduction

Trailing-edge modified nozzles are known to provide improved mixing characteristics and
high-speed jet noise reduction performance when designed appropriately [1]. One good
example is the chevron-shaped engine nacelles used in modern commercial airliners, which
seek to attenuate jet mixing noises for quieter flights. Thus, they have sparked extensive
interests from both research and industrial communities for their potential applications [2].
Beveled nozzles are among the several promising configurations, due to ease in their
manufacture and their effectiveness in manipulating noise radiation directivity [3]. In their
experimental investigations of beveled rectangular nozzles, Rice and Raman [1] observed
notable deflections of jet core for single beveled nozzle at under-expanded conditions.
Furthermore, their microphone measurements revealed a shift in the jet mixing noise to higher
frequencies. Viswanathan et al. [3] performed large-eddy simulations (LES) of single- and
dual-stream beveled nozzles with inclination angles of 24 and 45 respectively, and
concluded that the noise reduction of the jet flows was due to the modifications of noise
emitted from the large-scale turbulent structures. Later, Aikens et al. [4] confirmed that there
exists azimuthal variations of turbulent kinetic energy in the asymmetric beveled jets with 35
inclination, which contribute to asymmetry in the far-field jet broadband noise.

Since experiments on high-speed jet flows are not always feasible and full-scale LES are
computationally very demanding, Reynolds-averaged Navier-Stokes (RANS) simulation has
been explored as a useful alternative to evaluate and understand the fundamental flow
behaviors and noise characteristics associated with the resulting jet flows [5-7]. This is
especially the case when different combinations of trailing-edge modifications and operating
conditions are investigated as part of the optimization process or when solutions are required
over a short ‘turn-around’ time. For instances, Omais et al. [8] discussed the possibilities of
Fig 1. Definition of the nozzle coordinate
systems and angle of inclination (θ=60
degree)

noise prediction using RANS results to serve as inputs and more recently, Souza et al. [9]
employed RANS simulations coupled with statistical noise prediction method to provide
relatively rapid and accurate assessment of broadband noise associated with a jet in cross-
flow at transonic speed.

However, as noted by Garrison et al. [10] the accuracy of such noise prediction techniques
relies heavily on the accuracy of the RANS simulations. Due to constraints imposed by
commercial solvers, it is often difficult to tailor the numerical schemes and solution strategies
towards specific flow conditions for improved accuracy and efficiency. On the other hand, the
open-source computational fluid dynamics (CFD) tool, OpenFOAM, has become increasingly
popular in the past decade, due to its economic advantage and its well-established code
structures and solver flexibility. Hence, it is timely for the present research work to explore
the use of this tool in the numerical simulation of supersonic flows. It will not only allow the
assessment of the solver performance and shock-capturing schemes in OpenFOAM, but also
the examination and better understanding of the flow structures and shock-associated
dynamics of beveled nozzle with relatively larger inclination angle of 60 at differently exit
conditions (i.e. over and under-expanded respectively).

2 Numerical Methodology and Procedures

In the present study, a beveled convergent-divergent (CD) nozzle with 60 inclination angle
was numerically investigated using unsteady RANS approach in OpenFOAM. The nozzle
undergoes perfect expansion at nozzle pressure ratio (NPR) of 3.4 at a designed Mach number
of 1.45, within the range of operating conditions typical for supersonic jet engines [4]. More
design and test details of present CD nozzles can be found in Wu and New [11]. Figure 1
shows the Cartesian coordinates used as well as the definition of inclination angle θ. Jet flows
at two different NPRs of 2.8 and 4, corresponding to over- and under-expanded conditions
respectively, were freely exhausted from the beveled nozzle into a quiescent ambience and
rhoCentalFoam solver was employed to capture the development of the flow fields.

The compressible solver, rhoCentralFoam, is based on a central upwind flux schemes


proposed by Kurganov and Tadmor [12]. Essentially, flux of any conservative flow field
quantities (such as density and pressure) can be evaluated by splitting the flux in two
directions, namely the incoming and outgoing fluxes to the face of an owner cell, which yield:
[U ]dV fψf [a f ψ f (1 a) f ψ f w f (ψ f ψ f )] (1)
V f f

where φf is the volumetric flux, being splitted into ‘+’ and ‘-’ sides, ψf represents any field to
quantities to be evaluated with a, the weighting coefficient. Readers are advised to refer to
Kurganov and Tadmor [12] for details on the evaluation of the fluxes. Subsequently, the
Fig. 2 (a) Mesh topology and
boundary conditions for the present
three-dimensional flow field and
(b) two-layered structured mesh
construction with increased density
along shear layers

positive (i.e. ‘+’) and negative (i.e. ‘-’) fluxes at each cell can be reconstructed by a limiter
function that satisfies total variation diminishing (TVD) criteria. In particular, Gutiérrez
Marcantonia et al. [13] have benchmarked rhoCentralFoam against several well-defined
problems, such as supersonic flow over a wedge and past a slender diamond-shaped airfoil
and concluded that van Leer flux limiter was able to provide most balanced performance with
reasonable accuracy. Thus, the field variables were reconstructed using van Leer scheme in
the present simulations, which results in a second order accuracy in spatial discretization for
smooth regions.

Figure 2(a) shows the simulation domain of the jet flows, which extends up to 40D (i.e. D is
the jet diameter) downstream of the jet exit in the streamwise direction and 10D in the cross-
stream direction respectively. In order to improve upon computational efficiency while not
compromising on accuracy, the fully structured grid was partitioned into two regions, a dense
core region and a coarse outer region, as shown in Fig. 2(b). The grid density in the core
region closely follows that employed by Shen and Tam [14] in their numerical studies of jet
screech tones, and hence was deemed sufficiently fine to resolve the present flow fields.
Moreover, a factor of three was subsequently applied to the outer region with reference to the
core, which gave rise to a total number of approximately 24million cells. Lastly, additional
refinement was imposed upon the grids adjacent to the nozzle wall and shear layers to achieve
y+≈10, such that the flow turbulence could be captured satisfactorily by k-ω shear stress
transport model. Time-averaged results were then obtained through approximately 100,000
time steps after the jet flows became stabilized and fully developed.

3 Results and Discussion

3.1 Near-field shock structures

It is useful to begin the discussion by first examining the dominant shock structures of the 60
beveled nozzle at two differently expanded conditions. Figure 3 shows the isometric
Fig. 3 Near-field shock cell structures of (a) over-expanded NPR2.8 and (b) under-expanded
NPR 4 beveled jet flows

Fig. 4 Comparisons of the jet centerline flow development between over- and under-expanded
jet flows along both (a) Asymmetric and (b) Symmetric planes

view of the Mach number contour maps cut along both asymmetric (i.e. beveled x=0) and
symmetric (i.e. y=0) planes. Significant differences can be observed between the over- and
under-expanded jet flows. When the jet is over-expanded at NPR=2.8, the resulting shocks
cell structures are well-organized and exhibit some form of periodicity, which is characteristic
of imperfectly expanded axisymmetric jets [15]. Such periodicity is even clearer when the
Mach number developments along the centerline are extracted and shown in Fig. 4, where the
spatial intervals between the peak and trough remain relatively constant for the over-expanded
jet (i.e. solid lines) along both planes. On the other hand, the dominant shock structures of the
under-expanded jet at NPR=4 are essentially irregular in their formation as seen from Figs.
3(b) and 4. Such shock cell formations suggest that pressure correction and relief of the
beveled jet against the ambience is likely to be more gradual for the over-expanded jet. In
fact, a closer examination of the flow development in Fig. 4(b) reveals that the Mach number
decays much faster for the under-expanded jet along the symmetric centerline, indicating a
faster and possibly more drastic process in achieving pressure equilibrium with the ambience.
Fig. 5 Comparison of
the shock structures
between RANS and
experimental results
taken at (a) NPR=2.8
and (b) NPR=4
(Schlieren images
courtesy of Lim H.D.
and Wei X.)

3.2 Jet vectoring

Vectoring of the beveled jet flows occurs naturally as direct result of the non-uniform
pressure distribution and existence of strong pressure gradient between the short and long
nozzle tips. The phenomena have been reported by a number of studies [1, 3-4], nevertheless
relationships between nozzle pressure ratio as well as inclination angle and the extent of the
jet deflections remain unclear. The present simulations aim to provide some quantification on
the jet deflection angles at both over- and under-expanded conditions. First of all, Fig. 5
compares the RANS pressure field results with the experimental Schlieren images along the
asymmetric (i.e. beveled) plane. Initially, two distinct shocks can be seen to reflect off close
to the exit of the longer nozzle length, which later appear to interact and merge as a single
shock propagating downstream, regardless of the exit flow condition. This good qualitative
agreement between the simulation and experimental results reassures the validity and
accuracy of the present RANS simulations. Subsequently, the streamwise evolution of the jet
Mach number was extracted along several downstream locations, as shown in Fig. 6, and a
reasonable estimate of the angle of jet deflection can be determined from the collection of
Mach number peaks, thus jet velocity profiles. The results are tabulated and compared with
experimental measurement in Table 1. Interestingly, the jet core deflects in opposite
directions between over- and under-expanded jets. Note that positive angle corresponds to
upward deflection along y-axis. Though some uncertainties are expected in the method, there
are satisfactory agreements between the simulation and experimental results. It is noteworthy
that much more significant deflection of the under-expanded jet could possibly arise from the
high asymmetry in the shock structures, as observed from Fig. 3(b) along the beveled plane.

4 Conclusions
Unsteady RANS simulations were performed on beveled supersonic jet flows with relatively
large nozzle inclination angle of 60 , using rhoCentralFoam solver in OpenFOAM. The near-
field shock structures are in good qualitative agreement with the experimental Schlieren
images and are found to be relatively smaller and well-organized for over-expanded jet than
those for under-expanded one. Moreover, jet vectoring can be observed for both
Fig. 6 Streamwise evolution of the jet Mach number extracted along several downstream
locations for (a) over- and (b) under-expanded beveled jet flows

Table 1 Comparison of the jet deflection angles between RANS simulation and experiments
NPR RANS simulations Experiments
2.8 -0.7 degree -0.7 degree
4.0 2.9 degree 2.8 degree

cases, while the exact direction and extent of the deflection are highly dependent on the exit
conditions. The deflection angles determined from the simulations compare relatively well
with the experiments, which provide confidence in this open source CFD solver and
encourages further investigation into the beveled supersonic jet flows, such as the
computation of the aeroacoustics field.

Acknowledgment
The authors would like to acknowledge Lim H. D. and Wei X. for the Schlieren flow images,
as well as the support for this study by a Singapore Ministry of Education AcRF Tier-2 grant
(Grant number: MOE2014-T2-1-002).

References
[1] E.J. Rice, G. Raman, NASA Tech Report N94-19484 (1993)
[2] H. Xie, P.G. Tucker, S. Eastwood, Int J Heat Fluid Fl. 30, pp.1067-1079 (2009)
[3] K. Viswanathan, M. Shur, P.R. Spalart, M. Strelets, AIAA J, 46(3), pp. 601-626 (2008)
[4] K.M. Aikens, G.A. Blaisdell, A.S. Lyrintzis, 53rd AIAA Aero Sci Meeting, AIAA 2015-
0509 (2015)
[5] G. Lehnasch, P. Bruel, Int J Numer Meth Fluids, 56, pp. 2179–2205 (2008)
[6] S. Arunajatesan, 50th AIAA Aero Sci Meeting, AIAA 2012-1199 (2012)
[7] R. Chaplin, 31st AIAA Appl Aerodyn Conference, AIAA 2013-2911 (2013)
[8] M. Omais, S. Redonnet, B. Caruelle, E. Manoha, J Acoust Soc Am, 123(5), 2008
[9] P.R. Souza, O. de Almeida, C. R. Ilário da Silva, 55th AIAA Aero Sci Meeting,
AIAA 2017-0233 (2017)
[10] L. Garrison, A.S. Lyrintzis, G. Blaisdell, 12th AIAA/CEAS Aeroacoustics
Conference, AIAA 2006-2599 (2006)
[11] J. Wu, T.H. New, Aero Sci Tech, 63, pp. 278-293 (2017)
[12] A. Kurganov, E. Tadmor, J Comput Phy, 160(1), pp. 241-282 (2000)
[13] L.F. Gutiérrez Marcantonia, J.P. Tamagno, S.A. Elaskar, Mecánica Computacional Vol
XXXI, edited by Cardona et al., pp. 2939-2959 (2012)
[14] H. Shen, C.K.W. Tam, AIAA J, 36(10), pp. 1801-1807 (1998)
[15] C.K.W. Tam, J.A. Jackson, J Fluid Mech, 153, pp.123-149 (1985)

View publication stats

You might also like