OCR Text |
Show Modelling of combustion, NOx emissions and radiation of a natural gas red glass furnace George Mallouppas y, Yongzhe Zhang , Rajesh Rawatz yCD-adapco 200 Shepherds Bush Road, London, W6 7NL, UK george.mallouppas@cd-adapco.com CD-adapco 60 Broadhollow Road, Melville, NY, 11747, USA yongzhe.zhang@cd-adapco.com zCD-adapco 60 Broadhollow Road, Melville, NY, 11747, USA rajest.rawat@cd-adapco.com Abstract This work performs numerical simulations of a natural gas red glass furnace. The purpose of this work is to provide usefull insights in the design of glass furnaces and the appropriate modelling strategy for such type of simulations. The results are compared to the available experimental data and the importance of the recirculating region for the eective heat transfer at the glass wall is highlighted. The formation of NOx is a crucial parameter in the design of glass furnaces. Therefore, additional calculations for NOx are performed. The simulations, which are in agreement with the available experimental data, indicate the approximate intensity and location of NOx formation. Finally, the importance of radiation eects is highlighted by performing a simulation without radiation. The results indicating that radiation is an important physical phenomenon in glass furnaces and certainly cannot be ignored. 1 Introduction This work considers the numerical simulations of a natural gas red glass furnace. The experiments of Nakamura et al. (1990) invovle a full-scale underport-red glass tank of a 20o=20o conguration; the fuel and air inlets are inclined at 20o to the streamwise direction of the furnace. The design of the furnace is a critical parameter in fuel eciency and NOx emissions. This conguration has an important role in determining the NOx emissions as it is a critical factor in the mixing and burning of fuel with air. Typically in glass furnaces high pre-heated air is used in the combustion, which results in high peak ame temperatures (usually around 2200K). The formation of NOx depends on the ow characterisitics and forms in signicant amounts at temperatures above 1800K. This is aided by local inhomogeneities of the fuel-air mixture. This, therefore, can lead to local regions of high temperature and high O, OH radical concentrations. Numerical simulations of turbulent ows coupled with combustion can provide useful insights and can also have a signicant role in the design of glass furnaces. There have been a large number of advances in the modelling of complex reactive-turbulent systems. These models can reduce the complexity and number of degrees of freedom associated with the combustion chemical reactions in CFD calculations. The purpose of this work is to perform numerical simulations of a natural gas red glass furnace and compare the results with the corresponding data of Nakamura et al. (1990). Additionally, calculations of NOx emissions will be performed in order to determine the location and intensity of NOx in the glass furnace. Finally, the importance of radiation eects in the glass furnace will be shown. 1 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 The paper is divided into four sections. Section 2 will present the methodology and the models that will be used in this work. Section 3 will discuss the simulation set-up and meshing strategy. Section 4 will present and discuss the results inline with the available experimentlal data of Nakamura et al. (1990). Finally, Section 5 will conclude the main results of this work. 2 Numerical framework In this work the Reynolds averaged equations are solved in steady state, the mean-momentum equations are, @vjvi @xj = @P @xi + 2 @Sij @xj @v0j v0j @xj (1) where Sij = 1 2 @vi @xj + @vj @xi is the mean strain tensor and v0j v0j is the Reynolds stress tensor. Note that the SST model proposed by Menter (1994) is used. This model utilises a 'blending function' which provides the ability to use the k model away from the walls and the k ! model close to the walls. 2.1 Combustion modelling 2.1.1 PPDF model; Equilibrium model In STAR-CCM+ the mean mixture fraction (and in some cases, the mean enthalpy) are tracked on the grid for the PPDF model. This model takes into account via a statistical approach the mixing at the unresolved scales. The statitics at these unresolved scales are represented by a presumed probability density functions (PPDF). Therefore the mean species concentrations, temperature, and density are obtained as functions of the mean parametric variables, after an averaging process around a presumed probability distribution (PPDF) that is considered to represent turbulent uctuations. The model takes into account the convection eects due to mean and uctuating components of velocity. The transport equations for the mean mixutre fraction is @vkf @xk = @ @xk D + t Scf @f @xk (2) The transport equation of the variance of the mixture fraction, gf = (f f)2, @vkgf @xk = @ @xk D + t Scgf @gf @xk + 2t Scgf @f @xk !2 Cd k gf (3) is also solved. Note that Scgf is the turbulent Schmidt number for the variance and Cd is a dissipation constant, which represents the ratio of velocity and chemical species uctuations. This constant controls the intensity of the mixture fraction uctuations and typically values are 2:0 (Cant and Mastorakos, 2008), which is used in the current simulations. Note that the mixure fraction is presumed to have a PDF of beta function and in this work, the non-adiabatic PPDF is used. Therefore a heat loss ratio, , is used to indicate the amount of heat loss with the following equation = had(f) h hsens(f) (4) where had is the adiabatic enthalpy, h is the enthalpy of the computational cell and hsens is the sensible enthalpy. Enthalpy is also presumed to have a PDF of beta function. Moreover, a transport equation for enthalpy is solved for non-adiabatic conditions. It is also worthwhile mentioning that the PPDF tables are generated within STAR- CCM+. 2.2 Radiation modelling Radiation is expected to have a signicant impact hence in the current simulations. In this work the Discrete- Ordinates Method (DOM) with the Weighted Sum of Gay Gases Model (WSGG) are used to model the radiation. The DOM method solves transport equations for the radiation intensity with a specied direction throughout the 2 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 domain. In this work a fourth order ordinate set (S4) is used. The WSGG model provides a simlpied method for calculating the absorption coecient. This method uses the Hottel charts (Hottel, 1954) and sum-of-gray-gases models (Hottel and Sarom, 1964) for a mixture containing CO2 and/or H2O gases only. In the WSGG method, the medium is assumed to consist of dierent fractions of gray gases with dierent absorption coecients. The overall absorption coecient is computed as k = 1 L ln (1 T ) (5) where L is the optical length and T is the total emissivity. The optical length is either dened by the cell dimensions or if the medium is optically thin and the combustion gases occupy the entire volume, which is given as L = 3:6 Vdomain Adomain (6) In this work the later expression has been used to dene the optical length (i.e. L = 0:736m). 2.3 NOx modelling 2.3.1 NOx Zeldovich model Partial equilibrium approach: This approach computes the oxygen concentration, [O], and OH concentration, [OH], via: [O] = Kp p [O2] (7) where Kp = 1:16 p T exp 27123 T ! (8) and [OH] = 6:7331T0:57 exp 4595 T ! p [O] [H2O] (9) The reaction rate for NO is then estimated via equation 10, RNO = A B[NO] + A C=D [NO] + C=D [NO]2 B [NO] + C=D (10) 3 Simulation set-up The operating conditions of the glass furnace as reported by Nakamura et al. (1990) are: Natural gas at 283 K at 'Fuel inlet' (see Figure 1) 10% excess air at 1373K at 'Air inlet' (see Figure 1) Figure 1 illustrates the approximate region of mixing and combustion of fuel with air (red region). Hence at this region there are large temperatures due to combustion. Therefore, near this region (shown in light blue in Figure 1), NOx formation is expected. Note, that the conguration of the glass furnace is 20o=20o; meaning that the air inlet and fuel inlet (or underport fuel injector) are inclined at 20o to the z-axis. It is worthwhile mentioning that half of the geometry is simulated and as shown in Figure 1 this is done by dening a symmetry plane at the centre of the domain. The experimental data of Nakamura et al. (1990) report measuring points at x = 0:6m, x = 0:9m, x = 1:2m and x = 1:8m along the vertical direction (y-direction); also shown in Figure 1. Note that the boundary conditions are summarised in Table 1. `Crown wall' refers to the top wall which is assumed to be at 1500 K. 'Glass wall' refers to the bottom wall which is assumed to have a high heat loss, i.e. a heat ux of 90; 000W=m2 is specied. 'Chamber walls' refer to the other walls of the combustion chamber, whiich are assumed to have a low heat loss, i.e. 2; 000W=m2 is specied. 3 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 Figure 1: Figure illustrates fuel and air inlets, the experimental measuring points at x = 0:6m, x = 0:9m, x = 1:2m and x = 1:8m along the vertical direction (y-direction). Additionally the gure shows the approximate region of mixing and combustion of fuel with air (red region) and formation of NOx (light blue). Table 1: Boundary conditions. Quantity Fuel inlet Air inlet Crown wall Glass wall Chamber walls Temperature (K) 283.0 1373.0 1500.0 { { Velocity (m/s) 125.0 10.0 { { { Heat ux (W=m2) { { { 90,000 2,000 The furnace has dimensions: 3:8m 0:88m 0:955m. The diameter of the fuel inlet is 1.2 cm and the air inlet is rectangular with dimensions 27:8 cm 27:2 cm. As shown in Figure 2, a trimmer mesh with renement near fuel inlet to resolve mixing. The reference base size of the mesh is 0:02m. The mesh near the fuel inlet is rened by 40% to ensure mixing of the fuel with the air is adequately resolved, resulting in 600; 000 number of cells. Additionally, in order to capture the high temperature gradients near the wall, especially at the glass wall, 8 prism layers are added. Figure 2: Trimmer mesh with renement near fuel inlet to resolve mixing; 600; 000 cells. 4 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 4 Results and discussion Figure 3 compares the temperature prole as a function of height and Figure 4 compares the O2 mole fraction prole as a function of height at (a) x = 0:9m and (b) x = 1:8m. The results are in good agreement with the experimental data. At x = 0:9m, the results show two double peaks in temperature. This illustrates the location where mixing and burning of the fuel with air occur at x = 0:9m. This also explains the almost complete consumption of O2, see Figure 4(a). Figure 5 illustrates streamlines of CO2 mass fraction prole illustrating the (a) x = 0:9m (c) x = 1:8m Figure 3: Temperature prole as a function of height at (a) x = 0:9m and (b) x = 1:8m for the PPDF equilibrium model, EBU model, PPDF amelet model and PVM model compared to the experiments. (a) x = 0:9m (d) x = 1:8m Figure 4: O2 mole fraction prole as a function of height at (a) x = 0:9m and (d) x = 1:8m compared to the experiments. recirculation of hot gases within the glass furnace. The gure shows that there is a strong recirculation region above the ame which transports the hot gases in the primary region. Figure 6 illustrates the temperature contour plot across the furnace at (a) x = 0:6m, (b) x = 0:9m, (c) x = 1:2m and (d) x = 1:2m. The temperature is sligtly bend towards the glass wall. This clearly illustrates the importance of the recirculation region as well as the inclination angle of the air inlet. The gure also shows the temperature distribution accross the glass wall. The distribution is fairly constant at 1500K. Notice the relatively hotter region near the outlet of the furnace, which is caused by the hot recirculating gas products. 4.1 NOx emissions Figure 7 compares the NO mass fraction resuls in ppm at (a) x = 0:6m and (b) x = 0:9m with the experiments. The Zeldovich approach is able to capture the appropriate trends in NO formation, especially at (a) x = 0:6m where the NOx formation oscillates due to the corresponding temperature oscillations. Additionally, Figure 7 (b) shows that the strong oscillations merge into a single peak value of 2100 ppm. Note that the formation of NOx 5 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 Figure 5: Streamlines of CO2 mass fraction prole il- lustrating the recirculation of hot gases within the glass furnace. Figure 6: Snapshots of contour plot of temperature across the furnace, (a) x = 0:6m, (b) x = 0:9m, (c) x = 1:2m and (d) x = 1:2m. The gure also shows the temperature distribution accross the glass wall. is highest between x = 0:9m and x = 1:2m as illustrated by Figure 8, which shows the NO mass fraction at the centre of the glass furnace. (a) x = 0:6m (b) x = 0:9m Figure 7: NOx in ppm prole as a function of height at (a) x = 0:6m and (b) x = 0:9m compared to the experiments. Figure 8: Snapshot of mass fraction of NO. The gure shows the location of NO formation; i.e. at the centre of the furnace just after the mixing and burning of fuel with air. 6 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 4.2 Importance of radiation modeling This section investigates the importance of including the radiation model in the simulations. Figures 9 and 10 illustrate the dierences in temperature with and without the radiation model. As expected the radiation in the glass furnace is important. The gures clearly show that by ignoring radiation the temperature in the glass furnance is over predicted. Moreover, Figure 10 illustrates that the wall temperatures of the furnace are signicantly dierent (either higher or lower) compared to the simulations that include the radiation eects, since the walls are now considered not to emmit or absorb. Most importantly, the temperature of the glass wall without the radiation model drops to very low temperatures ( 300:0K) which is a not physical result. This is also highlights the importance of radiation in the heat transfer process within the glass furnace. (a) With radiation (b) Without radiation Figure 9: Figure illustrates the dierences in the temperature eld by (a) including radiation and (b) ignoring radiation. (a) x = 0:6m (b) x = 1:8m Figure 10: Figure compares the dierences in the temperature prole at (a) x = 0:6m and (b) x = 1:8m with and without the radiation model. 5 Conclusions Purpose of work, investigate glass furnace combstion with experiments of Nakamura et al. (1990) Investigated combustion using the PPDF equilibrium model Illustated importance of recirculation region in glass furnace Investigated NOx modeling via the Zeldovich approach Results are in good agreement with the experimental data 7 AFRC 2014 Industrial Combustion Symposium Houston, Texas, USA, September 7th-10th, 2014 Illustated importance of radiation modelling in glass furnaces Without radiation glass furnace temperature is much higher at the centre of the furnace than reported experimental results and simulations that include the eects of radiation. Without radiation the temperature at glass wall is unphysically low References R. S. Cant and E. Mastorakos. An Introduction to Turbulent Reacting Flows. Imperial College Press, 2008. H. C. Hottel. Radiant-Heat Transmission. Heat Transmission, 3rd Ed., W. H. McAdams (Ed.),, 1954. H. C. Hottel and A. F. Sarom. Radiative Transfer. McGraw-Hill, NY., 1964. F. Menter. Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal, 32:1598{ 1605, 1994. T. Nakamura, W. L. Vandecamp, and J. P. Smart. Further studies on high temperature gas combustion in glass furnaces. Technical report, IFRF Doc No F 90/Y/7, 1990. 8 |