Methusalem (M2dcR2) Advisory Board Meeting, Ghent, 24/06/2013 CFD Investigation on Fire-Side of Steam Cracking Furnace Yu Zhang, Kevin M. Van Geem and Guy B. Marin http://www.lct.UGent.be E-mail: [email protected] Laboratory for Chemical Technology Technologiepark 914, 9052 Ghent, Belgium European Research Institute of Catalysis Furnace Geometry • The furnace is a double radiant box with common convection section • Equipped with only long-flame (floor) burners, which have separated fuel and air inlet, situated at the bottom of the furnace • 44 reactor tubes are suspended in one row, each has two passes which are connected by a S-band and U-band one fourth of a radiant box Aim To investigate the flow, combustion and radiation in the fire-side of an industrial steam cracking furnace and to obtain the precise heat transfer into tubular reactor and the NOx emission Significance • The precise heat flux profile is required by the cracking reactor simulation software to better predict the yields, temperature, and coking rate inside the tubular reactor, which are essential for reactor design and optimization • The radical combustion kinetics which take NOx formation into account will give burner designer clues to reduce NOx emission in the furnace Fuel Composition • The fuel gas can be considered as CH 4 /H 2 mixture • Excess air inlet flow rate to ensure the fuel gas is fully combusted • Fuel gas flow rate : 2.22 kg/s • Fuel gas inlet temperature : 288.75 K • Oxygen excess (vol%) : 2 Models • Flow model : Re-Normalisation Group (RNG) k-ε turbulence model • Species transport model : Finite- Rate/Eddy-Dissipation model by Magnussen and Hjertager (1977), the reaction rate is governed by the minimum of Arrhenius and eddy-dissipation reaction rates • Combustion kinetics : two-step reaction mechanism based on Westbrook and Dryer (1981) • Radiation model : discrete ordinates (DO) radiation model is used to solve the radiative transfer equation (RTE) for a finite number of discrete solid angles • Flue gas absorption model : weighted-sum- of-gray-gases model (WSGGM) Magnussen and Hjertager (1977), Symp. (Int'l.) on Combustion, 16: 719-729 Westbrook and Dryer (1981), Combustion Science and Technology, 27: 31-43 Conclusions • The effects of burner geometry on the simulated flow, combustion and radiation heat transfer inside steam cracking furnace is not negligible • When using detailed combustion kinetics to predict NOx formation in the furnace ,the geometry effect must be taken into account Future Work • Perform furnace simulations in which detailed combustion kinetics is taken into account to predict the NOx emission • Evaluate the applicability of Eddy Dissipation Concept (EDC) model coupled with detailed combustion kinetics in non-premixed combustion Effect of burner structure on furnace simulation In most of previous study on fire-side of steam cracking furnace, the long-flame burners are always simplified as a plane at the bottom of the furnace In order to evaluate the geometry effect, two simulations are preformed with simplified and detailed burner structure respectively, as shown Due to the symmetry consideration, only a quarter of one radiant box is simulated, coupled with cracking reactor simulation software simplified burners detailed burners • The initial velocity of flue gas above the burners with detailed structures are significantly higher than that of the simplified burners, resulting in a stronger recirculation in the furnace • Flame shifts towards the furnace wall in the detailed burner simulation whereas remains the same place in the simplified one Flue gas velocity and temperature Heat transfer into process • The shifted flame temperature maximum towards the higher elevation and furnace wall leads to shorter distance between the two heat flux maximum of the tubular reactors • Non-uniform heat loads between 22 tubular reactors are observed No. 18 reactor tube No. 1-22 Heat flux of the tubular reactors No. 9 Heat transfer rate to all tubes Acknowledgement This work was carried out using the STEVIN Supercomputer Infrastructure at Ghent University, funded by Ghent University, the Flemish Supercomputer Center (VSC), the Hercules Foundation and the Flemish Government – department EWI