Two different CFD codes, FLUENT 5.5 and GENUS, are compared with independent
experimental data. FLUENT is a widely used commercial software package while
GENUS is a newly developed modular CFD prediction tool of Continuum
Computational Mechanics.
The GENUS turbulence model library features non-linear closures at the second
moment level for both velocity and scalar fields whereas FLUENT allows
closures at this level only for the velocity field. The GENUS code will be
used for turbulent reacting flow calculations in combustor-related geometries
at ALSTOM Power. FLUENT and GENUS are here compared on the basis of two test
cases relevant to flow in gas turbine combustors.
The first test case features an axisymmetric isothermal swirling flow with
central air injection in a dump combustor. Experimental data are provided by
the International Flame Research Foundation (IFRF) and include mean and RMS
velocities in the axial and tangential (swirl) directions at various
cross-stream planes. This case is computed using a 2D axisymmetric and a 3D
sector model with FLUENT and only the 3D sector model with GENUS. Both the
Standard k-e Model and second moment based closures are used to represent
turbulent transport. A grid dependency test has been performed in the context
of the 2D calculations and no substantial differences were observed. A
comparison between 2D and 3D calculations featuring a Reynolds Stress Model
and FLUENT revealed significant solution-convergence and prediction
improvements in favour of the latter - the k-e Model solution remained almost
unaffected. A comparison between results obtained by FLUENT and GENUS using
the k-e Model reveals significant differences in favour of the latter
package. When FLUENT and GENUS are compared using the Reynolds Stress Model
the results are similar but FLUENT is found to predict an unphysically large
internal recirculation zone.
The second test case features a range of planar, stoichiometric methane/air,
premixed flames propagating in a frozen uniform turbulence field. The latter
is in each case obtained by assigning to the turbulence intensity a constant
value for the whole domain. The integral length scale is assigned the value
of 10 mm. The calculation is transient and integration proceeds until a
steady state in flame-space is obtained. Calculated turbulent burning
velocities as a function of the turbulence intensity are compared against
experimental data provided by Abdel-Gayed et al. This case is in principle
simple to calculate and well suited for the evaluation of turbulent
combustion models. Using FLUENT with the Eddy-Dissipation concept based
closure for the mean reaction rate, extreme difficulties were observed with
respect to flame front stabilisation. In consequence a combination of
Arrhenius expressions and the Eddy-Dissipation concept were used as a means
to alleviate the above difficulties. In the latter case FLUENT was found to
yield substantially under-predicted values for the turbulent burning velocity
and over-predicted burnt gas temperat...