| OCR Text |
Show - 2 - gas fired glass furnaces [2, 3]. The techniques for NOx reduction studied consisted of two primary strategies: the direct reduction of the flame temperature and by modification of the local stoichiometry during combustion. The NOx reduction techniques investigated were: variation in excess air variation in air preheat level variation in air/fuel mixing characteristics - water/steam injection - fuel cracking variation in gas nozzle design - variation in thermal input EXPERIMENTAL In the experiments, a scaled down version of a full-scale underport-fired glass tank compartment was simulated [2, 3]. The baseline experimental burners were scaled down versions of existing industrial burners. Burner scaling was accomplished using constant velocity as the primary aerodynamic criterion (equivalent to constant momentum flux density scaling). Geometr "c similarity was also maintained. Thermal similarity was achieved by using partial thermal modelling [4]. The range of input parameters studied in the experiments are tabled below: Fuel input (baseline) Combustion air thermal input Total Volumetric heat release Furnace sizes internal L, D, H Gas port injection angles Combustion air injection angles Combustion air velocity Gas velocities Combustion Air Preheat Excess Air Level 0.5 MW o . 2 5 MW (1 1 0 0 ° C A = 1.') 0.24 MW/m3 3.9 m, 0.9 m, 0.95 m 12° , 20° 12°, 20° 7 to 13 m/s 75, 125, 225 m/s 800, 1100, 1300°C -5, 10, 20% other variables included: nozzle desigr, water and stearr. injection, and preconditioning of the natural gas to initiate cracking in the early part of the flame. Baseline conditions were with air preheat of '100 C, a nominal excess air leve l of 10% a n air and gas injection angle of 20 and 12 respectively (20 /1 2) burner) and a gas injection velocity of 125 m/s. The experiments were conducted on the IFRF water cooled Furnace Number 2, consi sting of 14 independently water cooled segments. Within the furnace, variable refractory was installed to simulate the typical thermal boundary condi tions of a full scale glas melting furnace. The furnace roof and wal ls were constructed o~ composite refectory layers. Thicknesses wer such that heat fluxes in the order of 5 kW/m2 were obtained. To simulat e the glass load a thin layer of silicon carbide was cas lnto the furnace floor to give a representativ range of hea fluxe s between 50-90 kW/m2, dependent on surface temperature . The glas s load simulation was the predominant heat sink in the furnace fo r the experimental flames. Consequentl y , the axial varla t ion in |
