Large eddy simulations of the UMD line burner with the conditional moment closure method
Introduction
Thanks to the remarkable rise in the available computational power, the LES approach has become very popular in turbulent combustion simulations. A major advantage is that large-scale turbulence effects are captured and models for the sub-grid scale (SGS) eddies are universal in nature. In non-premixed combustion in particular, turbulence governs the mixing of fuel and oxidizer. This is one reason why LES has been popular for fire simulations. Related to this, the very important phenomenon of entrainment, i.e., the mixing of air into the flame zone and into the smoke plume, is much better resolved in LES than with RANS, because the large scale unsteadiness and the puffing phenomenon are captured. This explains why LES is the default approach for turbulence in the two most widely used CFD codes for fire simulations, namely FDS [1] and FireFOAM [2].
Whereas the effect of turbulence in turbulent reacting flows, and in fires in particular, is known to be very important, an important observation in fire configurations is the relatively large size of the flame region, compared to lab-scale (jet) flames. The size of fuel involved at the inlet is of the order of tens of centimeters and flames can easily be more than 1 m high. This has implications on the fineness of the computational mesh: given that computational domains are large in volume, mm size mesh cells cannot be afforded. Hence, this can pose issues with respect to the resolution of turbulence. The effect of turbulence in turbulent reacting flows, and in fires in particular, is known to be very important. In the present study, we use the Smagorinsky model [3]. Moreover, turbulence – chemistry interaction (TCI) and turbulence – radiation interaction (TRI) are important to consider. With respect to TCI, there are many options [4]. One option is the use of the conditional moment closure (CMC) method, as developed in Ref. [5]. Some examples of LES-CMC simulations of turbulent flames are [[6], [7], [8], [9], [10]]. An important advantage of this technique is the potential to include finite-rate chemistry effects. It is important to note that all configurations in Ref. [[6], [7], [8], [9], [10]] deal with jet flames, i.e., the flow field is mainly momentum-driven by a strong injected fuel (and/or air) flow and combustion takes place where mixing leads to close to stoichiometric conditions. In fires, the situation is fundamentally different: the momentum of the injected fuel and oxidizer flows is weak. The actual flow field is driven by buoyancy, which is created through the high-temperature region, as a consequence of the combustion process. This leads on the one hand to much stronger feedback loops in this overall process, but the low velocities also induce longer residence times and lower scalar dissipation rates, and hence lower mixing rates. The lower values are illustrated below for the test case under consideration (Fig. 11) and they have a very strong impact, particularly when radiation is taken into account. To the best of the authors’ knowledge, it is the first time that LES/CMC is applied to a fire related configuration.
The slow mixing rate also makes the impact of the radiation source term more pronounced, as already reported in Ref. [11]. This was also confirmed in Ref. [12], illustrating again strong coupling between the different phenomena. Through the impact on the buoyancy, this also affects the flow and mixing fields in physical space more strongly than when these are mainly determined by the momentum of the injected flows (as is the case in jet flames). This is illustrated below for the test case under consideration (Fig. 9). To the best of the authors’ knowledge, it is the first time that this aspect is studied in the context of CFD simulations for a fire related flame.
As mentioned above, there are limitations to the possible fineness of the CFD mesh in fire simulations. Yet, due to the large domains, simulations are computationally expensive. Hence, in order to reduce the computational expenses, combustion modelling in fires has relied mostly on simplified global 1-step or 2-step chemistry mechanisms. These are coupled with the eddy dissipation concept (EDC), or eddy dissipation model (EDM), combustion modelling approach. On the other hand, phenomena like extinction, re-ignition, minor species or soot, are known to strongly depend on chemical kinetics. Due to the strong coupling of phenomena, these might also affect radiation. While in principle the EDC model is capable of handling finite-rate chemistry, to the best of the authors' knowledge this has not been applied yet in fires. Reaction rates are determined from the turbulent mixing time scale, while the mixed-is-burnt assumption is made for chemical kinetics [[13], [14], [15], [16], [17]]. Extinction and re-ignition models have been implemented with EDC separately, such as in Refs. [13,14]. In Ref. [13], a strong sensitivity of the results is reported on the critical temperature of the re-ignition model. This suggests it is worthwhile to investigate more advanced models. One approach to include detailed chemistry is the application of the flamelet model, such as in Ref. [18] (where a 1-step mechanism was used), and more recently [[19], [20], [21]]. Interestingly, in Ref. [18] two models have been evaluated for the conditional scalar dissipation rate (CSDR), which represents the mixing rate of fuel and oxidizer. It was determined that the CSDR model has a stronger impact in fires than in jet flames. The effect was observed for temperature and velocity fields. It is suggested in Ref. [18] that this would be due to the laminar-to-turbulent transition in fires, jeopardizing the assumption of statistical independence of mixture fraction and scalar dissipation rate (SDR). Regardless thereof, it has been illustrated in Ref. [12] that the CSDR model indeed has a strong impact, in particular for low SDR values, if radiation is taken into account in a non-local manner in mixture fraction space. This is confirmed in the study at hand when radiation is taken into account through the WSGGM model, comparing results with the classical AMC model [22] to what is obtained with a uniform CSDR profile (Fig. 4). Hence, the mentioned impact of the CSDR shape is shown to be significant in such conditions, even without the presence of the laminar-to-turbulence transition zone. To the best of the authors’ knowledge, it is the first time this is illustrated in fire related simulations.
As mentioned, recently, the steady flamelet model approach [21] has been used for the test case at hand, namely the UMD line burner [23,24], described in more detail below. A strong under-estimation of temperature has been reported. It was suggested in Ref. [21] to be due to the presumed β-PDF shape or the mixture fraction variance model, but most importantly it was stated that this requires further study. It was also concluded in Ref. [21] that an unsteady flamelet model would be more suited, because low SDR values imply long response times, which is not consistent with the steady state assumption. This approach has recently been considered in Refs. [19,20]. The use of (unsteady) flamelet models in fires is clearly work in progress, deserving further study.
This further motivates the use of the conditional moment closure (CMC) [5], as is done in the present study. LES-CMC is a high-end modelling approach that has, as mentioned above, been applied successfully for extinction and re-ignition in jet flame configurations [6,7,10]. It is capable of handling detailed finite-rate chemistry and has the potential to be used for the prediction of fire extinction. On the other hand, the conditioning procedure introduces an additional dimension, mixture fraction in non-premixed combustion, which renders the model expensive. Except for the spatial transport terms in the transport equations for the conditional quantities, the CMC equations resemble the unsteady flamelet equations. The application of CMC in fires is rare, which can be attributed to the large computational domains used in fire simulations. One example is [25], where CMC was coupled with RANS in an enclosure fire scenario. In Refs. [25], it was argued that the spatial terms in the CMC equations are required to accurately model CO concentrations in the post flame regions and that consequently the steady flamelet model does not lead to accurate results there. This further motivates the use of CMC in the present study. As mentioned above, CMC models have been evaluated mostly in jet flame configurations, with strong turbulence and high (conditional) SDR values. Consequently, it remains to be seen how well the various sub-models perform in fire applications, particularly because in Ref. [25] the impact of the CSDR shape on minor species was reported. This confirms again the importance of the CSDR shape, as mentioned above [18].
The present study aims to provide some insight into the impact of various aspects of the CMC approach, combined with LES, for fire CFD simulations, using the UMD line burner as fire related configuration. The findings of the accompanying CMC-0D study [12], which effectively focuses on the impact of the chemistry mechanism, the CSDR shape and the radiation modelling approach on extinction within the unsteady laminar flamelet concept approach for conditions representative of the UMD line burner, are used in order to reduce the number of CFD simulations and as assistance to interpret the results. However, because the case at hand does not exhibit much extinction, most results are presented first with the 1-step chemistry mechanism of [26], in order to reduce computing times. Comparison to the ARM 2 mechanism [27] confirms that the findings are not affected by the choice in chemistry mechanism here from a qualitative point of view. For radiation, two approaches are compared, imposed in mixture fraction space in the CMC energy equation, as described in Ref. [12]: 1) an approach where the experimentally measured radiative fraction is imposed, locally in mixture fraction space as a prescribed reduction of the local conditional HRR; 2) the WSGGM approach. The latter model has already been applied with EDC on the UMD line burner in Ref. [13], showing results in very good agreement with experiments. Moreover, the WSGGM is gaining popularity recently within the fire community [13,[19], [20], [21]]. Nevertheless, it is important to also assess results with prescribed radiative fraction, as reported below.
The case under study is the UMD line burner, developed at the University of Maryland [23,24]. It is a target case of the IAFSS working group MaCFP (‘Measurements and Computations of Fire Phenomena’) [28] to study fire extinction. The case features a buoyancy driven, methane-fueled fire with a controlled co-flow of a nitrogen/oxygen mixture. Extinction is reached by increasing the nitrogen levels in the co-flow. In this study, no extinction simulations have been conducted. Only the fully burning flame is considered, with the oxidizer being air. The aim is to evaluate the LES-CMC approach, before moving towards extinction in future work.
From the above it is clear that the study conducted in Ref. [12] is strongly connected to the present work. Therefore, the main findings from Ref. [12], used for the explanation of certain observations in the present study, are summarized first:
- -
The shape of the CSDR in mixture fraction space (uniform versus bell-shaped) does not have a strong impact on the results in adiabatic conditions, or when the radiative fraction model is applied, as long as the value for the CSDR at stoichiometry is the same for different CSDR models. This is because the reaction rate and the radiative source in the radiative fraction model are narrow spikes around stoichiometry.
- -
The shape of the CSDR in mixture fraction space has a strong impact on the results when the WSGGM and the RADCAL-based radiation model are used, even if the value of the CSDR at stoichiometry is the same for different CSDR models. The reason is that the radiative source term in these cases remains significant further away from stoichiometry.
- -
The ARM2 chemistry mechanism [27] yields almost identical results as the much more expensive GRI3 mechanism [29] for the quantities in the present study. It is therefore also studied here in one section (as the intent is to investigate extinction phenomena in future work).
Section snippets
LES
The LES equations are solved in OpenFOAM-2.3.0 [30], which include:
The mass conservation equation:
- -
The momentum conservation equations:
The mixture fraction transport equation:
The SGS model for turbulence is Smagorinsky, where the SGS viscosity is modelled as , and the SGS turbulence kinetic energy is . The model constants and .
Test case
The top view of the UMD line burner [23] is shown in Fig. 1. The fuel (pure methane) mass flow rate at the inlet is 1 g/s, which corresponds to an average velocity of 6 cm/s. The co-flow oxidizer (in this case air) mass flow rate is 80 g/s, which corresponds to an average velocity of 24 cm/s. The total heat release rate is around 50 kW. Temperature measurements have been taken using 12.7-μm-diamater Type-S (Pt/10%Rh) thermocouples. The effects of soot were found to be small [23], so soot is
Numerical set-up
The basic LES mesh is shown in Fig. 2. In order to save computational time, two regions of local mesh refinement are used. The finest region (6.5 mm × 4.2 mm x 6.25 mm cells) covers the region 0.3 m (width) x 0.6 m (height) x 0.6 m (length). This is high enough to capture the flame height (which is about 0.5 m according to the experiments) and implies 7.6 cells across the fuel port width (5 cm). The fine zone (13 mm × 8.4 mm x 12.5 mm cells) is 0.8 m high, 0.6 m wide and 0.8 m long. The rest of
Results
Unfortunately, not many experimental data are available for the test case at hand. To the best of the authors’ knowledge, only the evolution of mean and rms values of temperature along the centerline, as well as cross-stream profiles of these quantities at heights 0.25 m, 0.5 m, 0.75 m and 1 m, are available. Nevertheless, given the strong coupling between temperature fields and the buoyancy-driven flow field and mixing, it is worth-while to also discuss the mean and rms values of upward and
Discussion
The results using the radiative fraction model are in relatively good agreement with experiments, and with results found in literature, conducted using the EDC combustion model [14]. However, it must be kept in mind that this good agreement is at the virtue of using an experimentally measured radiative fraction. Hence, such simulations are not predictive. Hence, the strive towards a fully predictive approach remains. In Ref. [13], good agreement was obtained with WSGGM when combined with the
Conclusions
The LES/CMC method has been applied to a fire related flame with different model settings. To the best of the authors’ knowledge, this has not been reported before for fire related scenarios. The buoyancy-driven fires are characterized by low strain rate values, as opposed to momentum-driven jet flames, which makes the radiative effects important. Overall, the findings of [12], focusing on the conditional profiles in mixture fraction space in the absence of transport in physical space, are
CRediT authorship contribution statement
B. Kruljevic: Conceptualization, Methodology, Software, Data curation, Writing- Original draft preparation, Visualization, Investigation, Validation. I. Stankovic: Supervision, Writing - review & editing. B. Merci: Supervision, Writing - review & editing.
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
This research has been funded by Ghent University (Belgium) through GOA project BOF16/GOA/004.
References (36)
- et al.
Turbulent combustion modeling
Prog. Energy Combust. Sci.
(2002) - et al.
Conditional moment closure for turbulent combustion
Prog. Energy Combust. Sci.
(1999) - et al.
Large Eddy Simulation/Conditional Moment Closure modeling of swirl-stabilized non-premixed flames with local extinction
Proc. Combust. Inst.
(2015) - et al.
Numerical simulation of oxy-fuel jet flames using unstructured LES–CMC
Proc. Combust. Inst.
(2015) - et al.
Structure and extinction of methane-air flamelet with radiation and detailed chemical kinetic mechanism
Combust. Flame
(1998) - et al.
Large eddy simulation of flame extinction in a turbulent line fire exposed to air-nitrogen co-flow
Fire Saf. J.
(2016) - et al.
Modeling flame extinction and reignition in large eddy simulations with fast chemistry
Fire Saf. J.
(2017) - et al.
Radiative emissions measurements from a buoyant, turbulent line flame under oxidizer-dilution quenching conditions
Fire Saf. J.
(2015) - et al.
Augmented reduced mechanisms for NO emission in methane oxidation
Combust. Flame
(2001) - et al.
WSGG correlations based on HITEMP2010 for computation of thermal radiation in non-isothermal, non-homogeneous H2O/CO2 mixtures
Int. J. Heat Mass Tran.
(2013)
Reduced storage matrix methods in stiff ODE systems
Appl. Math. Comput.
Towards predictive simulations of gaseous pool fires
Proc. Combust. Inst.
General circulation experiments with the primitive equations: I. The basic experiment
Mon. Weather Rev.
Conditional moment closure/large eddy simulation of the Delft-III natural gas non-premixed jet flame, Flow
Turbul. Combust.
Modelling of non-premixed turbulent combustion with conditional moment closure (CMC)
Eur. Phys. J. E
LES/CMC simulations of swirl-stabilised ethanol spray flames approaching blow-off, Flow
Turbul. Combust.
Cited by (7)
Numerical investigation on turbulence-radiation interaction in the UMD turbulent line fires
2023, Fire Safety JournalOn the importance of the heat release rate in numerical simulations of fires in mechanically ventilated air‐tight enclosures
2023, Proceedings of the Combustion InstituteInfluence of bottom wall on characteristics of jet diffusion flames under cross-wind
2021, FuelCitation Excerpt :Many relevant studies have been conducted, but research has mainly focused on the jet diffusion flame without a bottom wall [2,3]. In practical situations, the fuel nozzle of industrial burners is typically flush with the surrounding wall [4] (for example, a line burner [5] and the combustion of an industrial burner in the furnace [6]). Moreover, gas leakage fire accidents always occur on the ground [7,8], and the bottom wall around the leakage point affects the flames.
Systematic study of the impact of chemistry and radiation on extinction and carbon monoxide formation at low scalar dissipation rates
2020, Fire Safety JournalCitation Excerpt :Hence, a comparative study is performed for results obtained with the uniform scalar dissipation rate model and with the AMC model, taking the same value for CSDR at stoichiometric mixture fraction, as shown in Fig. 1. This sheds light on the possible importance of the shape of the CSDR results in 3D LES-CMC simulations (see, e.g. Ref. [2]). Here we show the results for the ARM2 mechanism, although the impact of the CSDR on extinction (unless adiabatic) was also observed with the other mechanisms.
Exploratory Study of the Impact of the Turbulence Model on Flame Extinction with an EDM and EDC/Finite-Rate Approach for a Line Burner Configuration
2024, Flow, Turbulence and Combustion