A mathematical model of fuel combustion in tubular furnaces was developed. The model allows to calculate the concentration of nitrogen oxides in the exhaust gases for different configurations of the furnace and burners, different volume flow rates and methods of air supply to the furnace. In order to confirm the adequacy of the developed mathematical model, a numerical experiment was conducted for the existing basic furnace design with fuel combustion technology according to the current regulations. A slight (2.3%) discrepancy between the calculated and experimental values of the concentration of nitrogen oxides at the outlet of the operating furnace indicates the adequacy of the developed mathematical model. Subsequent numerical experiments to study various fuel combustion schemes for the purpose of decreasing NOx emission showed that the use of two-stage combustion in burners located in the same plane is ineffective. The use of this method allows to reduce the NOx yield by no more than 10%. The method of two-stage combustion by the height of the combustion chamber is more effective. The supply of a mixture of gases with an excess air coefficient of 0.8 to the burners and the supply of the remaining air higher along the cross section of the furnace allow reducing the total output of nitrogen oxides by 2,5 times.