Best Practice Advice AC2-09: Difference between revisions
m (Dave.Ellacott moved page SilverP:Best Practice Advice AC2-09 to Best Practice Advice AC2-09) |
|||
(8 intermediate revisions by the same user not shown) | |||
Line 37: | Line 37: | ||
The computational domain for Sandia D Flame should extend far enough | The computational domain for Sandia D Flame should extend far enough | ||
from the nozzle outlet to capture at least the region of maximum | from the nozzle outlet to capture at least the region of maximum | ||
temperature, i.e. ''x/D'' ≥ 50. The numerical tests performed have shown that | temperature, i.e. ''x/D'' ≥ 50. The numerical tests performed have shown that | ||
in order to limit the CPU time the computational domain can be made | in order to limit the CPU time the computational domain can be made | ||
divergent in the downstream direction, and the lateral extent used in | divergent in the downstream direction, and the lateral extent used in | ||
Line 47: | Line 47: | ||
==Discretisation and Grid Resolution== | ==Discretisation and Grid Resolution== | ||
In the BOFFIN code second order discretization was applied in time and | |||
space. The grid refinement studies for the LES calculations showed that | |||
the grid resolution 80×80×160 nodes in the proposed computational | |||
domain is sufficient and further grid refinement leads to minor changes | |||
of statistically converged parameters. However, no quantitative measure | |||
of the contribution of the subgrid scale to overall flow parameter | |||
oscillations was evaluated. | |||
==Physical Modelling== | ==Physical Modelling== | ||
The results of the LES prediction of combusting flows could be | |||
affected by many factors. First could be the type of subgrid-scale | |||
model. In principle, in non-premixed combustion the mixing intensity | |||
determines the combustion rate so that the quality of the SGS model | |||
could be of importance. However, comparisons of Smagorinsky model with | |||
the results obtained with the dynamic model did not show a significant | |||
influence. | |||
On the other hand the LES calculations performed have shown that the | |||
results obtained in the flame region are significantly better than the | |||
flow behaviour predicted upstream of the flame. Another important | |||
factor is chemical kinetics. The two reduced mechanisms used in the | |||
current calculations seems to be sufficiently accurate for prediction | |||
of major species in the Sandia D flame. | |||
==Recommendations for Future Work== | ==Recommendations for Future Work== | ||
It should be kept in mind that LES of combusting flow is still a new | |||
and not mature field and thus many different research directions could | |||
be proposed to understand the associated problems and improve the | |||
modeling quality. On the other hand it is important to stress that LES | |||
with combustion requires a sufficiently fine mesh and consequently | |||
large computer resources and CPU time. Hence, the number of LES | |||
predictions leading to new insight is limited by current computer | |||
resources and required CPU time. A tentative list of future works | |||
generally in LES modeling of non-premixed combustion is as follows: | |||
*Testing high-order numerical schemes to speed up calculations allowing coarser meshes like spectral, compact differences, WENO schemes | |||
*Systematic studies of reduced kinetics for chosen flames allowing to establish minimum requirements in terms of required number of species and reaction steps for typical non-premixed flames | |||
*Treating Sandia D Flame as a first step towards extensive LES studies of non-premixed flames with more complex structure involving local extinction phenomena like Sandia E and F flames are needed | |||
*Studies of more advanced turbulence/combustion interaction models like full 1<sup>st</sup> order CMC and/or transported PDF models like Eulerian Fields PDF approach. | |||
<br/> | <br/> | ||
---- | ---- | ||
{{ACContribs | {{ACContribs | ||
|authors=Andrzej Boguslawski | |authors=Andrzej Boguslawski, Artur Tyliszczak | ||
|organisation= | |organisation=Częstochowa University of Technology | ||
}} | }} | ||
{{ACHeader | {{ACHeader | ||
Line 64: | Line 101: | ||
© copyright ERCOFTAC | © copyright ERCOFTAC 2011 |
Latest revision as of 15:43, 11 February 2017
SANDIA Flame D
Application Challenge AC2-09 © copyright ERCOFTAC 2024
Best Practice Advice
Key Fluid Physics
The non-premixed Sandia D flame is an example of a flame in the flamelet regime in which the Kolmogorov scale is significantly larger than the scales characteristic for the combustion process. In this not demanding test case the models based on flamelet assumption should lead to good agreement with experimental data as was shown in the Evaluation section, especially in the region of developed flame. However, more discrepancies were observed in the near field were only mixing of fuel and oxidizer is considered.
Application Uncertainties
The flow field in the near field is certainly influenced by inlet conditions. The mean velocity profile and fluctuating component were chosen according to the experimental data. However, in the unsteady LES calculations the fluctuations were simulated by white noise. This means that the fluctuations characteristic for developed turbulent flow were not reproduced at the inlet and this could influence the mixing features in the near field. It is well known that white noise provides a fluctuating signal with a very short time scale which is then smoothed at a short distance from the inlet plane. However, it seems that due to very low inlet turbulence level the further results in the flame region are only weakly influenced by these near field results as both models analyzed led to reasonable results.
Computational Domain and Boundary Conditions
The computational domain for Sandia D Flame should extend far enough from the nozzle outlet to capture at least the region of maximum temperature, i.e. x/D ≥ 50. The numerical tests performed have shown that in order to limit the CPU time the computational domain can be made divergent in the downstream direction, and the lateral extent used in the current calculations (5.5D and 18.3D at the inlet and outlet plane respectively), is sufficient not to influence the flame structure by the lateral boundary conditions. The coflow as in experimental conditions should be introduced at the inlet and at the lateral boundaries.
Discretisation and Grid Resolution
In the BOFFIN code second order discretization was applied in time and space. The grid refinement studies for the LES calculations showed that the grid resolution 80×80×160 nodes in the proposed computational domain is sufficient and further grid refinement leads to minor changes of statistically converged parameters. However, no quantitative measure of the contribution of the subgrid scale to overall flow parameter oscillations was evaluated.
Physical Modelling
The results of the LES prediction of combusting flows could be affected by many factors. First could be the type of subgrid-scale model. In principle, in non-premixed combustion the mixing intensity determines the combustion rate so that the quality of the SGS model could be of importance. However, comparisons of Smagorinsky model with the results obtained with the dynamic model did not show a significant influence.
On the other hand the LES calculations performed have shown that the results obtained in the flame region are significantly better than the flow behaviour predicted upstream of the flame. Another important factor is chemical kinetics. The two reduced mechanisms used in the current calculations seems to be sufficiently accurate for prediction of major species in the Sandia D flame.
Recommendations for Future Work
It should be kept in mind that LES of combusting flow is still a new and not mature field and thus many different research directions could be proposed to understand the associated problems and improve the modeling quality. On the other hand it is important to stress that LES with combustion requires a sufficiently fine mesh and consequently large computer resources and CPU time. Hence, the number of LES predictions leading to new insight is limited by current computer resources and required CPU time. A tentative list of future works generally in LES modeling of non-premixed combustion is as follows:
- Testing high-order numerical schemes to speed up calculations allowing coarser meshes like spectral, compact differences, WENO schemes
- Systematic studies of reduced kinetics for chosen flames allowing to establish minimum requirements in terms of required number of species and reaction steps for typical non-premixed flames
- Treating Sandia D Flame as a first step towards extensive LES studies of non-premixed flames with more complex structure involving local extinction phenomena like Sandia E and F flames are needed
- Studies of more advanced turbulence/combustion interaction models like full 1st order CMC and/or transported PDF models like Eulerian Fields PDF approach.
Contributed by: Andrzej Boguslawski, Artur Tyliszczak — Częstochowa University of Technology
© copyright ERCOFTAC 2011