Hostname: page-component-7b7567c8c-2n9c9 Total loading time: 0 Render date: 2026-11-16T13:49:27.612Z Has data issue: false hasContentIssue false

Linear global stability of two incompressible coaxial jets

Published online by Cambridge University Press:  13 July 2017

J. Canton
Affiliation:
Linné FLOW Centre and Swedish e-Science Research Centre (SeRC), KTH Mechanics, Royal Institute of Technology, Stockholm, SE-100 44, Sweden
F. Auteri*
Affiliation:
Dipartimento di Scienze e Tecnologie Aerospaziali, Politecnico di Milano, via La Masa 34, 20156 Milano, Italy
M. Carini
Affiliation:
ONERA – The French Aerospace Lab, F92190 Meudon, France
*
†Email address for correspondence: franco.auteri@polimi.it

Abstract

The linear stability of two incompressible coaxial jets, separated by a thick duct wall, is investigated by means of both a modal and a non-modal approach within a global framework. The attention is focused on the range of unitary velocity ratios for which an alternate vortex shedding from the duct wall is known to dominate the flow. In spite of the inherent convective nature of jet flow instabilities, such behaviour is shown to originate from an unstable global mode of the dynamics linearised around the axisymmetric base flow. The corresponding wavemaker is located in the recirculating-flow region formed behind the duct wall. At the same time, the transient-growth analysis reveals that huge amplifications (up to $20$ orders of magnitude) of small flow perturbations at the nozzle exit can occur in the subcritical regime, especially for high ratios between the outer and the inner velocities.

Information

Type
Papers
Copyright
© 2017 Cambridge University Press 
Figure 0

Figure 1.Sketch of the computational domain $\unicode[STIX]{x1D6FA}_{c}$ (azimuthal plane) employed for the investigation of the coaxial jet flow. All lengths are made non-dimensional using the inner pipe diameter $\widetilde{D}_{i}$. The black thick line is used to represent the solid wall boundary $\unicode[STIX]{x1D6E4}_{w}$ while the blue shaded area denotes the region of mesh refinement. The outer-to-inner pipe diameter ratio is fixed to $R_{D}=2$, while the thickness of the duct wall is 0.1.

Figure 1

Table 1.Characteristics of the meshes used in the present study. All the meshes are named $M_{z_{max}}$ so as to quickly recall the most important mesh parameter; where duplicates are present the changing parameter is reported as a second subscript. $N_{e}$ denotes the total number of triangular elements while $N_{dofs}$ denotes the corresponding total number of degrees of freedom. All the meshes are characterised by a grid size $\unicode[STIX]{x0394}\ell$ varying from $\unicode[STIX]{x0394}\ell _{max}=0.4$ close to the outflow border $\unicode[STIX]{x1D6E4}_{o}$, down to a minimum value of $\unicode[STIX]{x0394}\ell _{min}=0.009$ within the refinement zone (blue shaded area in figure 1); these sizes of the elements have been determined following a mesh refinement independence study. See figure 1 for the definition of the other parameters.

Figure 2

Figure 2.Sketch of the computational domain employed for the stability analysis of the single jet configuration.

Figure 3

Figure 3.Modal and non-modal stability analyses of the single jet flow for $Re=1000$ and $m=0$. (a) Global eigenspectrum. (b) Gains associated with the optimal perturbations as a function of the time horizon $\unicode[STIX]{x1D70F}$: present results (round circle) compared with those reported by Garnaud et al. (2013) for the same configuration (continuous line).

Figure 4

Figure 4.Variation of the base flow velocity profiles through the nozzle for $R_{u}=1$ and different values of $Re$. (a) Axial velocity profile imposed at the inflow and resulting velocity profiles close to the nozzle exit ($\bar{z}=0.1$). (b) Radial velocity profiles close to the nozzle exit ($\bar{z}=0.1$).

Figure 5

Figure 5.Base flow axial velocity field in the proximity of the nozzle for $Re=1000$ and different values of $R_{u}$. White lines are used to represent the flow streamlines while the grey shaded area corresponds to the duct wall separating the two coaxial fluid streams. (a) $R_{u}=0.5$. (b) $R_{u}=1$. (c) $R_{u}=1.5$. (d) $R_{u}=2$. The same colour scale is employed for all the panels and the colour bar is reported in figure (d).

Figure 6

Table 2.Outlet–inlet ratios of the maximum axial velocity component characterising the base flow for $R_{u}=1$ and different values of $Re$. See also figure 4 for $\bar{z}=0.1$.

Figure 7

Figure 6.Axial and radial velocity profiles of the computed base flow for $Re=1000$ and $R_{u}=1$ at different stations along the $z$-axis. (a) $U_{z}(r,z)$. (b) $U_{r}(r,z)$.

Figure 8

Figure 7.Global stability analysis of the coaxial jet flow. Panel (a) depicts the neutral curve associated with the leading axisymmetric mode in the $Re$–$R_{u}$ plane. The minimum critical Reynolds number, $Re=1356$ for $R_{u}\approx 1.27$, is highlighted with a (red) dash-dotted line. The round markers, $P_{i}$, highlight the values at which the spectra in figure 8 have been computed. The dashed line represents the neutral curve as a function of the bulk Reynolds number $Re_{b}$, reported on the top $x$-axis. Panel (b) reports the Strouhal number of the marginally stable mode along the neutral curve, plotted as a function of the critical velocity ratio $R_{u}$. The continuous line corresponds to the frequency scaled with the inner jet velocity, $St_{s}=s\unicode[STIX]{x1D714}_{c}/(2\unicode[STIX]{x03C0}U_{i})$, while the dashed line shows the same frequency scaled by the bulk velocity, $St_{s,b}=s\unicode[STIX]{x1D714}_{c}/(2\unicode[STIX]{x03C0}U_{b})$, reported on the top $x$-axis.

Figure 9

Figure 8.Global eigenspectrum computed at four different points along the neutral curve for $m=0,1,2$. (a) $P_{1}$, $Re=5000$ and $R_{u}=0.505$. (b) $P_{2}$, $Re=1420$ and $R_{u}=1.0$. (c) $P_{3}$, $Re=1386$ and $R_{u}=1.5$. (d) $P_{4}$, $Re=1559$ and $R_{u}=2.0$.

Figure 10

Figure 9.Direct global mode for $m=0$. Panels (a) and (b) represent the axial and radial velocity components, respectively, of the leading global mode at criticality for $Re=1420,R_{u}=1.0$ (point $P_{2}$ in figure 7). The same quantities are depicted in panels (c) and (d) for the critical mode at $Re=1559,R_{u}=2.0$ (point $P_{4}$ in figure 7). Note that the figures only reproduce a portion of the computational domain which is always characterised by $z_{max}\geqslant 40$. Mesh $M_{60}$, with $z_{max}=60$, was used for these particular fields. The complete fields can be found at https://doi.org/10.1017/jfm.2017.290.

Figure 11

Figure 10.Critical adjoint global mode for $m=0$ represented by the magnitude of the associated velocity field. (a) $Re=1420$, $R_{u}=1.0$. (b) $Re=1559$, $R_{u}=2.0$. Note that the figures only reproduce a portion of the computational domain which is always characterised by $z_{min}\leqslant -5$. Mesh $M_{60}$, with $z_{min}=-5$, was used for these particular fields. The complete fields can be found at https://doi.org/10.1017/jfm.2017.290.

Figure 12

Figure 11.Structural sensitivity map ${\Vert \unicode[STIX]{x1D61A}_{0}(r,z)\Vert }_{F}$ associated with the critical global mode at four different points along the neutral curve. White lines are used to represent the base flow streamlines. (a) $P_{1}$, $Re=5000$ and $R_{u}=0.505$. (b) $P_{2}$, $Re=1420$ and $R_{u}=1.0$. (c) $P_{3}$, $Re=1386$ and $R_{u}=1.5$. (d) $P_{4}$, $Re=1559$ and $R_{u}=2.0$. (e) $P_{4}$, $Re=1559$ and $R_{u}=2.0$ (extended view). Note that the figures only reproduce a portion of the computational domain which is always characterised by $z_{max}\geqslant 40$ and $z_{min}\leqslant -5$. Mesh $M_{60}$, with $z_{max}=60$ and $z_{min}=-5$, was used for these particular fields. The complete fields can be found at https://doi.org/10.1017/jfm.2017.290.

Figure 13

Figure 12.Gains associated with the optimal perturbations as a function of $\unicode[STIX]{x1D70F}$ for $m=0$ and different values of $R_{u}$. (a) $Re=1000$. (b) $Re=1250$. The dashed (blue) line shows the amplification of the same optimal perturbations attained on the nonlinear flow $R_{u}=1$.

Figure 14

Figure 13.Spatial structures associated with the evolution of the optimal perturbations at $Re=1000$ depicted by means of the kinetic energy distribution $|\boldsymbol{u}|^{2}/2$ for different velocity ratios: (a,b) $R_{u}=0.5$, (c,d) $R_{u}=1.0$ and (e,f) $R_{u}=1.5$. In (a,c,e) the optimal initial condition is displayed while in the (b,d,f) the resulting linearised flow response at the maximum amplification time (black filled symbols in figure 13a) is illustrated. Note that the figures only reproduce a portion of the computational domain which is always characterised by $z_{max}\geqslant 40$ and $z_{min}\leqslant -5$. Meshes $M_{60}$ and $M_{80}$, with $z_{min}=-5$ and $z_{max}=60$ and $80$ respectively, were used for these particular fields.

Figure 15

Figure 14.Flow snapshot of the fully developed unsteady coaxial jet flow obtained from three-dimensional DNS at $R_{u}=1$ and $Re=1500$: iso-surfaces of the $\unicode[STIX]{x1D706}_{2}$ criterion ($\unicode[STIX]{x1D706}_{2}=-0.5$) coloured by the azimuthal component of the vorticity field. A planar cut of the azimuthal vorticity field is also illustrated. The light grey shaded structure corresponds to the nozzle geometry.

Figure 16

Figure 15.Time history and frequency content of the radial velocity component extracted from the DNS of the axisymmetric coaxial jet flow for $R_{u}=1$ and two different Reynolds numbers, i.e. $Re=1000$ (a,d,g) and $Re=1500$ (b,e,h). The velocity probes are located at $z=2.0$ and $r=0.25,0.55,0.8$ corresponding to (a–c), (d–f) and (g–i), respectively. The third column shows the spectral content of the velocity signals for the case at $Re=1500$, for which a periodic solution is established.

Figure 17

Figure 16.Branch $b_{2}$ of the global eigenspectrum computed using meshes characterised by a different streamwise extent of the domain.

Figure 18

Table 3.Computed leading eigenvalue for $R_{u}=1$ and $Re=1420$ using computational domains of different length.

Figure 19

Table 4.Variation of the computed gain factor $G_{0}^{opt}(\unicode[STIX]{x1D70F})$ for $R_{u}=1$ and $Re=1000$ obtained using different computational domains, with respect to the reference $M_{100,z_{i}15}$ grid. The variation is also normalised with respect to the values obtained with $M_{100,z_{i}15}$.

Supplementary material: File

Canton supplementary material

Canton supplementary material

Download Canton supplementary material(File)
File 44.9 MB