Environmental Engineering Reference
In-Depth Information
ow and
acoustic codes, and are chosen in order to compute the underlying physics accu-
rately (discussed next). Both the
In general, the domain, grid and time step size are different for the
fl
flow and acoustic computations are performed in
MPI-based parallel environment. MPI-based specialised modules are added that can
perform spatial/temporal averaging and interpolation of
fl
fl
flow quantities from the
fl
flow grid to the acoustic grid and vice versa. Simple arithmetic average is followed
to perform spatial average in view of the
fineness of the grids used. Quadratic
Shepard scheme is used to perform data interpolation in 3D (Renka
1988
).
The
flow computations and acoustic computations are performed using 16 and 8
parallel processors, respectively. Additional 6 and 4 processors are used to handle
the forward and backward data transfer. A total of 34 processors spread across two
nodes (24 processors per node) are used to perform a coupled simulation.
fl
4 Results and Discussion
Flame-acoustic coupled simulations are performed for the laboratory combustor
(Chakravarthy et al.
2007
) (schematic as shown in Fig.
1
) containing a backward
facing step for
flame stabilisation with fuel injection at the step corner. The length
of the combustor is 1.4 m and the step plane is located 0.4 m from the inlet. The
step height h = 0.03 m and the expansion ratio across the step plane is 2. The span
wise extent of the duct is 0.06 m. Coupled 3D simulations are performed for Re in
the range of 9,000
fl
flow rate of 142 mg/s, for which
experimental data on acoustic pressure measurements are available for direct
comparison (Chakravarthy et al.
2007
).
A reduced domain of 0.15 and 0.4 m (5 and 13 h) upstream and downstream of
the step plane, respectively, is considered for the turbulent combustion computa-
tions. The range of scales expected in terms of non-dimensional wave number k for
Re = 18,000 and 33,000 are 1.0
59,000, at a
fixed fuel
fl
-
10
3
, respectively. Multi-
block structured mesh with a total size of 1.04 million nodes is used. Besides
resolving the near-wall regions down to a wall
10
3
and 1.0
1.5
×
2.5
×
-
-
+
3, the grid employed
resolves scales up to k = 15.2 in the regions where the mesh is coarsest.
Random inlet velocity perturbations with amplitude of 20 % of the mean inlet
velocity are superposed on the mean velocity and are prescribed at the inlet plane.
No-slip boundary condition is used at the walls. Out
y
of 2
-
ow boundary condition is
used at the exit plane of the computational domain. At the exit plane, the velocity is
corrected such that the instantaneous balance in volume
fl
ed taking
into account the dilatation that occurs in the computational domain. This improves
the global convergence and also ensures the advection of vortical structures from
the exit plane. The walls of the combustor are maintained at 700 K as the
approximation of hot walls witnessed in the experiments.
The dynamic viscosity, thermal conductivity, species diffusivity, speci
fl
flow rate is satis
c heat
capacity and gas constant for all the species are kept constant and are taken as
1.84
10
−
5
Ns/m
2
, 0.0425 W/mK, 2.67
10
−
5
m
2
/s, 1,354 J/kg K and 296.9 J/kg K,
×
×
Search WWH ::
Custom Search