You are on page 1of 3

Proceedings of Eurotherm83 — Computational Thermal Radiation in Participating Media III

15–17 April 2009, Lisbon, Portugal

Influence of the TRI on thermal radiation in

LES of turbulent flows

by Maxime ROGER(∗) , Pedro J. COELHO(∗) and Carlos B. DA SILVA(∗)

(*) Mechanical Engineering Department, Instituto Superior Tcnico/IDMEC, Technical University

of Lisbon, Av. Rovisco Pais, 1049-001 Lisboa, Portugal

In turbulent reactive flows, radiative transfer plays a significant role, and turbulence-
radiation interaction (TRI) effects are relevant in most cases [1]. Among the various
approaches used to model turbulence, the direct numerical simulations (DNS) is a pow-
erful tool to provide fundamental insight on turbulent flows but it is too computational
expensive for practical applications. With the increase of computer power, large eddy
simulation (LES) is becoming an attractive approach to achieve high accuracy at an af-
fordable computational cost for solving complex problems of practical usefulness in the
CFD and combustion communities. To our knowledge, few work has been done so far to
model the TRI in LES. Poitou et al. [2] have made an a-priori study from DNS and tested
models based on Taylor development for the TRI emission in a flame. Coelho [3] has pro-
posed TRI models based on an assumed subgrid-scale probability density function. He
generated a time-series of turbulent scalar fluctuations along optical path in a Sandia flame
D and solved the filtered RTE along this path by applying one-dimensional filtering oper-
ation. In these studies, it is shown that the TRI is less relevant in LES than in Reynolds
Averaged Navier-Stokes (RANS) approaches. Thermal radiation has been coupled to LES
in combustion systems in various studies where the influence of the subgrid-scale fluctu-
ations over the resolved radiative heat transfer has been neglected. In this study which
is a continuation of previous work [4] where the filtered RTE was studied, this assump-
tion is evaluated. The filtered RTE is calculated without TRI models and is compared to
DNS and filtered DNS (or a-priori results, in order to evaluate the influence of the LES
non-resolved scales of the LES on the radiation.

1. Theory and computational details

1.1.The filtered RTE
The filtered RTE is obtained by applying the spatial filtering operation to the RTE:

dI ν
= −κν Iν + κν Ibν
= −κν I ν − κν Iν − κν I ν + κν I bν + κν Ibν − κν I bν (1)
The terms in parentheses, κν Iν − κν I ν and κν Ibν − κν I bν need to be modelled in
order to close the filtered RTE. They represent the effects of the unresolved-scales of the
LES on the monochromatic radiative intensity.
The simplest way to close Eq. (1) is to suppose
 that the subgrid-scale
  fluctuations of
thermal radiation are negligibles, i.e., the terms κν Iν − κν I ν and κν Ibν − κν I bν
are neglected in the filtered RTE. In this assumption, the filtered RTE may be written as
dI ν
≃ −κν I ν + κν I bν (2)
In the case of a medium composed by a mixture of H2 O and CO2 , it is further assumed
that κν ≃ κν (T , X CO2 , X H2 O ) and I bν ≃ Ibν (T ), i.e., the temperature self-correlation
is neglected as well as the correlation between the absorption coefficient and the temper-
ature/chemical composition.

1.2.Computational details
The simplest possible turbulent configuration flow, statistically steady (forced) ho-
mogeneous isotropic turbulence without combustion, have been chosen for clarify the
analysis. The turbulence and the radiative heat transfer calculations are decoupled in this
study. The radiative transfer has been calculated by using a ray-tracing/correlated-k dis-
tribution method. The DNS is carried out using a standard pseudo-spectral code in which
the temporal advancement is made with an explicit 3rd order Runge-Kutta scheme. More
detail on the DNS solver and on the thermal radiation calculations can be found in [4].
For the LES, a spatial filtering operation is applied to the temperature and the species con-
centration fields. A new coarser grid is defined according to the filter size. The filtered
RTE is then calculated in this new grid. In 1, results for a filter size of ∆ = 16δ (where
∆ is the filter size and δ is the mesh size in the DNS grid) are presented because it is a
representative size of filter used in engineering applications. This filter size is located in
the inertial range region of the kinetic spectrum.

2. Results and conclusion

Figure 1 displays the radiation intensity from the DNS (IDN S ) and the filtered ra-
diation intensity obtained by LES (I LES ) and by a-priori calculations (I DN S ) along a
line of sight chosen arbitrarily. First, it is observed that the LES tends to underestimate
the filtered intensity in comparison with the filtered DNS, i.e., I LES < I DN S . This
tendency is general and observed in the whole domain. It is the combined effect of all
the subgrid-scale correlations, such as the temperature self-correlation, the absorption
coefficient-temperature correlation, or the absorption coefficient-radiative intensity corre-
lation, which is highlighted in the difference between I LES and I DN S . In the estimation
of I LES , it is supposed that T 4 ≃ T , that κν Ib,ν ≃ κν I b,ν , and that κν Iν ≃ κν I ν . It
has been observed that the temperature self-correlation tends to increase the filtered emis-
sion. It is the stronger subgrid-scale correlations which explains the difference between
the LES and the filtered DNS. However, the absorption coefficient-temperature correla-
tion decreases the emission. The opposite effect between these correlations implies that
it is better to neglect both correlations instead of modelling just one, which confirms the
results of Poitou et al. [2]. Regarding now the absorption coefficient-radiation intensity
correlation, it is observed that this correlation is weak, and confirms that the optically thin
1.12 I DN S /hIDN S i in LES grid
I DN S /hIDN S i in DNS grid

1.06 I LES /hIDN S i in LES grid




0.96 I LES /hIDN S i in DNS grid

0.2 0.4 0.6 0.8


Fig. 1 — Normalized radiative intensity profiles along a different line of sight, estimated
by DNS (IDN S ), by LES without SGS models (I LES ) and by filtered DNS (I DN S ). The
results are displayed in the DNS and in the LES grids.

fluctuation approximation can be extended to LES context, as shown in [3].

Another observation, which can be made from 1, is that the grid has an influence,
although quite small, on the radiation intensity. It can be seen that in the common nodes
of the two grids, the a-priori estimations of the filtered radiation intensity remain equal
in both grids, as expected. But, the LES results exhibit a difference in the LES grid and
in the DNS grid (for instance between s/L = 0.5 and 0.8). This difference, observed
only for non-local quantities such as the radiation intensity and the radiative absorption,
highlights the influence of the grid. The term is non-local and consequently the error due
to the lack of subgrid-scale modelling spreads and changes along an optical path.
In conclusion, this study has shown that neglecting the influence of the non-resolved
scales in the filtered RTE remain a good approximation. However, some effects should
be taken carefully in configurations where the turbulence intensity is important. The
extension of the analysis to real unsteady cases should be the purpose of future work.


[1] COELHO,P.J., Numerical simulation of the interaction between turbulence and radiation in reactive
flows Progress in Energy and Combustion Science, volume 33, issue number 4, pages 311-384, 2007

[2] POITOU, D., EL HAFI, M., CUENOT, B., Diagnosis of turbulence-radiation interaction in turbulent
flames and implications for modeling in large eddy simulation,Turkish J. Eng. Environ. Sci.,volume 31,
issue number 6, pages 371-381, 2007

[3] COELHO,P.J., Approximate solutions of the filtered radiative transfer equation in large eddy simulations
of turbulent reactive flows, Combustion and Flame,volume 156, issue number , pages 1099-1110, 2009

[4] ROGER, M., DA SILVA, C.B.,COELHO,P.J., Analysis of the turbulence-radiation interaction for large
eddy simulations of turbulent flows, volume 52, issue number 9-10, pages 2243-2254, 2009