Jennifer X. Wen

CFD modelling of confined jet fires under ventilation-controlled conditions

This paper presents field predictions for two confined jet fires under ventilation-controlled conditions. Three different combustion models, namely the laminar flamelet model, the constrained equilibrium (CE) method and the eddy break-up (EBU) model, are used and compared. The laminar flamelet model and the CE method are coupled with Lindstedts soot model, while the EDU model is coupled with the soot model of Magnussen. The global radiative heat exchange is considered by using the discrete transfer radiation method (DTRM) and the calculation of the microscopic radiative heat exchange within the flamelet is embeded in RUN-1DL. For both the flamelet and CE methods, the energy equation is solved so that the radiation calculation is coupled with the computation of the turbulent combustion processes. Comparisons are made between the experimental data and the predictions of different sub-model combinations for two propane jet fires in enclosures of 135 and 415 m3. It is found that all the sub-model combinations predict the correct trends of distributions for field variables such as velocity, temperature, soot, and CO. The predicted temperature distributions from the flamelet approach, which includes both the microscopic and global radiative heat losses, are found to be in close agreement with the experimental data. Modifications are made to the convective heat transfer coefficients between the wall and the gaseous products in the impinging region and this is found to have improved the predictions for the wall heat fluxes.

Investigation of a spectral formulation for radiative heat transfer in one-dimensional fires and combustion systems

The radiative heat transfer in variable concentration, non-isothermal and sooty combustion products is investigated with a correlated-k gas method and discrete ordinates technique for the solution of equation of transfer. The spectrally formulated model is an approach for predicting radiative intensities and fluxes in mixtures of gaseous products (H2O, CO2, CO) and scattering soot particles. The reliability of this approach for fire and combustion applications is analysed by comparing its predictions with some theoretical results and measured radiative intensities for a natural gas flame. Use of this model for more accurate calculations and evaluation of simplified fire and combustion radiation models, are among its main advantages.

Experimental study of water sprays for the attenuation of fire thermal radiation

A laboratory experimental work is carried out to investigate the attenuation ability of water sprays subjected to thermal radiation. The objective is to analyze the key parameters involved in the mitigation properties of this fire protection technique. The spectral transmittances of two types of sprayers, TG03 and TG05, are measured with a Fourier infrared spectrometer under various conditions. The wavelength range varies from 1.5 to 12 μm. The influence on the transmittance of both the flow rate and the pressure ranging from 1 to 7 bars, as well as the effect of the number of spray nozzles are considered. The results clearly show the advantage of small drops with high concentration. An investigation on the multi-ramp curtain configuration also provides valuable information on the mitigation behavior of the whole spray. Key guidelines are provided for fire protection engineering.

Development and validation of an advanced turbulence model for buoyancy driven flows in enclosures

Abstract A new buoyancy-modified turbulence model is developed on the basis of the four-equation model, k−e− θ 2 −eθ, of Hanjalic [K. Hanjalic, S. Kenjeres and F. Durst, Natural convection in partitioned two-dimensional enclosures at higher Rayleigh numbers, Int. J. Heat Mass Transfer 39(7) (1996) 1407–1427]. The strong anisotropy of Reynolds stresses due to buoyancy effects in the vertical boundary layers is considered by inclusion of the newly devised ‘return-to-isotropy’ concept in the pressure-strain correlation. The wall-reflection functions is also duly modified. The new model has been tested in buoyancy-driven cavity flows through comparison with published experimental data and the predictions from three other turbulence models [N. Z. Ince and B. E. Launder, On the computation of buoyancy-driven turbulent flows in rectangular enclosures, Int. J. Heat and Fluid Flow 10(2) (1998) 110–117; K. Hanjalic, S. Kenjeres and F. Durst, Natural convection in partitioned two-dimensional enclosures at higher Rayleigh numbers, Int. J. Heat Mass Transfer 39(7) (1996) 1407–1427]. It has demonstrated significant improvements in capturing the non-isotropy of Reynolds stresses and turbulent heat flux in vertical boundary layers.

Large eddy simulation of a small pool fire

Large eddy simulation has been applied to the prediction of a small pool fire experimentally tested by Venkatesh et al. The Smagorinskys eddy viscosity model is used for subgrid-scale (SGS) turbulence closure and a modified laminar flamelet model (MLFM) based on the Cook and Riley approach is used for SGS combustion modeling. The predictions have captured the unique characteristics of small pool fires such as flame anchoring and double flame, as shown in previous theoretical analysis and experiments. The existence of the premixed zone of fuel and air, which is thought to be the reason for flame anchoring, is evidenced by the low gradient of temperature and mixture fraction near the burner rim. The experimentally observed double flame can be seen from both the predicted temperature contour and velocity vectors. For the mean temperature field where experimental data are available for quantitative comparison, the predictions with the MLFM are found to be in very good agreement with the data. In line with general expectations, the study also reveals that the predicted small pool fire is nearly axially symmetric. Comparison of the predictions with the two different SGS combustion models have highlighted the importance of SGS combustion modeling in capturing the fine details of such small pool fires. Considerable discrepancies have been found in the predictions of the velocity and temperature fields, and the predictions of the mixture fraction model indicate a slightly higher and more centered premixed flame near the burner rim.

The effect of turbulence modelling on the CFD simulation of buoyant diffusion flames

Abstract For buoyant diffusion flames, both thermal and mechanical forces affect the turbulence mixing and combustion processes. The computation of individual turbulent heat flux θ′′u i ′′ and temperature variance θ″ 2 are necessary and important. This raises questions about the use of traditional two equation k–e type turbulence models in such applications. The present study is aimed at demonstrating the significant effects of turbulence modelling on the CFD simulation of buoyant diffusion flames. Two different turbulence models are used to compute McCaffreys flame data. The first model is based on Hanjalics (Proceedings of the 10th International Heat Transfer Conference, Vol. 1, 1994, p. 1) four-equation turbulence model, which has been modified by the present authors to account for turbulence anisotropy and coupled with an algebraic formulation for Reynolds stresses. The second is the low-Reynolds-number (LRN) k–e model of Ince and Launder (Int. J. Heat Fluid Flow 10 (1989) 110). The predictions of both models for temperature gradients and buoyancy generation of turbulence are examined. It is found that the generalised gradient diffusion hypothesis formula in the LRN k–e model severely under-predicts the buoyancy production of turbulent kinetic energy. Considering that the simple gradient diffusion formula in the standard k–e model with buoyancy modification predicts even lower turbulence production due to buoyancy, fire modellers are cautioned about the use of the two equation k–e type turbulence models in such applications. Furthermore, it is found that the velocity and temperature predictions by the two turbulence models differ as much as 20% along the centreline. The LRN k–e model under-predicts the radial spreading rate of the flame, the temperature rise in the persistent flaming region and the lower portion of the intermittent flaming region. For the rest of the intermittent flaming region and the thermal plume, it over-predicts the temperature rise. The modified version of Hanjalics turbulence model has achieved quantitatively good agreement with the experimental data on the predictions of velocity and temperature distributions. It has also given better predictions for the radial spreading rate of vertical flames and the flame shapes.

Dispersion of carbon dioxide from vertical vent and horizontal releases—A numerical study:

Numerical simulations of far-field carbon dioxide dispersion were conducted for a vertical vent release and a horizontal release from a shock tube. These scenarios had also been studied experimentally at field scale commissioned by National Grid. This work and the experiments both form part of the National Grid dense phase CO2 pipeLine TRANSportation (COOLTRANS) research programme. All tests involved releases of dense phase CO2 into an atmospheric flow. The dispersing plumes were subjected to transient wind conditions where both the direction and magnitude of the wind fluctuated with time. As part of the COOLTRANS research programme, the far-field dispersion simulations started from source terms derived from the near-field simulations conducted by the University of Leeds and outflow simulations conducted by University College London. The numerical model used for the far-field simulations is based on OPENFOAM, which is an object-oriented open source computational fluid dynamics toolbox. A dedicated solver CO2FOAM has been developed within the framework of OPENFOAM for simulating dispersion from dense phase CO2 releases. This has included the implementation of the homogeneous equilibrium method for fully compressible two-phase flow, treatment of the transient atmospheric boundary conditions and the time-varying inlet boundary conditions. The experimental measurements were supplied to the authors after the predictions were completed and submitted to National Grid. Hence, the validation reported here is indeed “blind.” While further fine tuning of the model and validation is still underway, the relatively good agreement between the predictions and measurements in the present study has demonstrated the potential of CO2FOAM as an effective predictive tool for far-field CO2 dispersion in the context of pipeline transportation for carbon capture and storage.

Burning rate of merged pool fire on the hollow square tray

In order to characterize fire merging, pool fires on hollow trays with varying side lengths were burned under quasi-quiescent condition and in a wind tunnel with the wind speed ranging from 0m/s to 7.5m/s. Burning rate and flame images were recorded in the whole combustion process. The results show that even though the pool surface area was kept identical for hollow trays of different sizes, the measured burning rates and fire evolutions were found to be significantly different. Besides the five stages identified by previous studies, an extra stage, fire merging, was observed. Fire merging appeared possibly at any of the first four stages and moreover resulted in 50-100% increases of the fire burning rates and heights in the present tests. The tests in wind tunnel suggested that, as the wind speed ranges from 0 m/s to 2 m/s, the burning rates decrease. However with further increase of the wind speed from 2 m/s to 7.5 m/s, the burning rate was found to increase for smaller hollow trays while it remains almost constant for larger hollow trays. Two empirical correlations are presented to predict critical burning rate of fire merging on the hollow tray. The predictions were found to be in reasonably good agreement with the measurements.

The effect of microscopic and global radiative heat exchange on the field predictions of compartment fires

This paper reports on some further results of the CFD simulations of large-scale compartment fires previously reported in Wen et al. (Proceedings of the Combustion Institute , vol. 27, 1998) and Wen and Huang (Fire Safety J 2000;34(1)). It focuses on the use of the laminar flamelet approach and highlights the effect of microscopic radiation on the field predictions of temperature and species concentrations in compartment fires. The flamelet calculations with and without microscopic radiation are performed using RUN-1DL (Rogg. RUN-1DL Manual, 1099) . Radiative properties in the flamelet are calculated by a modified exponential wide band model. Global radiation is coupled with the field calculation through the discrete transfer radiation method (Shah. Ph.D. thesis, Imperial College of Science and Technology, 1979) and Hubbard and Tiens (ASME J Heat Transfer 1978;100:235–9) mean emission and absorption coefficient concept. The soot model of Leung et al. (Combust Flame 1991;87;289–305) is used for soot predictions. Improved agreement with experimental data on temperature distributions has been achieved by including the microscopic radiation in the flamelet calculation. Microscopic radiation is also found to have significant effect on the predictions of soot and OH radical but its effect on the predictions of CO2, CO and H2O are found to be marginal. The present study recommends that radiative heat exchange at microscopic level (within the laminar flamelet) should be included when using the laminar flamelet approach to compute turbulent reacting flows.

Analysis of the two-flux model for predicting water spray transmittance in fire protection application

An investigation is carried out to assess the two-flux model for evaluating the attenuation ability of a water spray curtain in fire protection. Transmittances calculated with this model are compared with the exact discrete ordinates solutions for a range of water curtains under practical conditions. The results show the unsuitability of the two-flux method under a collimated incidence boundary condition, even if some improvements could be expected with very small droplets. Whereas the diffuse incidence type provides relatively better results, it is more reliable for transmittance calculations.