THESE de DOCTORAT - cerfacs

5.2 Experimental configuration 73
fluctuating quantities such as the velocity or the temperature field. On the contrary, acoustic
pressure fluctuations, which are much smaller in amplitude, are much more difficult to evaluate
and their correct estimation is linked to the order of the numerical scheme and to the degree
of resolution of the computational grid. On the other hand, hybrid approaches have not been
ever considered to compute combustion noise in confined domains. As a consequence, it is
not known yet how appropriate is their use for combustion noise estimation in combustion
chambers. There are several conceptual differences between open and confined flames that
should be considered regarding this matter. An important one is that in a confined flame there
is not such a thing of ’far field pressure’. This means that fluctuations of pressure field are not
only composed of acoustics. Turbulence, and therefore hydrodynamic fluctuations, might be
of considerable importance and, as a consequence, ‘contaminate’ the acoustic field. Another
main concern is the boundary conditions: an open flame is computed considering acoustic
waves that propagate outwards the domain and for this purpose the boundaries of the computational
domain are treated to be non-reflecting; in a confined flame, acoustic waves produced
travel both inwards and outwards. Boundary conditions modeling is therefore expected to be
of crucial importance because their acoustic reflection is linked directly to the acoustic field
produced. These physical concepts should be taken into account when applying hybrid approaches
for confined flames.
5.2 Experimental configuration
Both direct and indirect computations of combustion noise are performed and applied to swirled
premixed combustor (EC2 combustor) carried out in the EM2C laboratory (École Centrale
Paris) [43, 44]. The EC2 combustor consists of two geometrically identical stages for air-fuel
injection, a premixer and a combustion chamber. Air is fed into each stage through a circular
manifold in which a swirler is inserted. This swirler has a hollow cylinder with large lateral
openings (see Fig. 5.1a), through which air is injected in the inner premixer channel. Inside
these rectangular openings, four injectors (1 mm diameter) deliver gaseous propane perpendicularly
to the air flow. This cross-flow configuration enhances fuel-air mixing. The tangential
injections create a strong swirl motion in the D = 30 mm diameter inner channel, which in
turn generates a central recirculation zone at the plenum that stabilizes the flame. The flame is
controlled by the fuel-air ratio imposed in each of the two stages and is considered premixed
and compact. Note that the LES could consider modeling the lines at each stage of both fuel
and air as shown in Fig. 5.1(a). Nevertheless, important computational costs would arise due
to the small grid cells that would be necessary to mesh the fuel lines. Since an homogeneous
air-fuel mixing is considered to be achieved before arriving to the reacting zone, the fact of
meshing fuel lines far upstream from the flame might be totally unnecessary. Therefore the
present LES will only consider the air lines (simplified model shown in Fig. 5.1b) in which a
premixed mixture with equivalent ratios φ 1 and φ 2 is injected for each stage.
This configuration features strong combustion instabilities depending on the fuel staging ratio