OCR Text |
Show 2 . to conduct an exper imental program to veri fy the model performance, and 3. to formulate NOx reduction strategies for industrial processes using oxy-natural gas burners. Research is now focused on objectives one and two. Numerical Model Air Products is currently developing a computer model called GRREK to predict NOx levels in flue gas. GRREK is an acronym for Gas flow and Radiation with Reaction Equilibrium and Kinetics. GRREK incorporates submodels for calculating fluid flow, radiation heat transfer, and chemical kinetics. The specific objectives of the model development are: 1 . to generate a model to predict flue NOx at various 02 enrichment levels in a furnace with a natural gas fired burner at one end and a flue at the other end, and 2. to identify the significant parameters in NOx production. GRREK is constructed with interconnected modules. This approach has several advantages. Parallel development of modules decreases the development and debugging time. GRREK can be easily modified by substituting more efficient routines or improved kinetics when they become available. Unneeded modules can be made inactive while solving a particular problem to reduces computation time. The fluid flow module has its source in the combustion model (COHO) developed by Babcock and Wilcox for the U. S. Department of Energy2. COMO is a numerical model for predicting furnace performance in axisymmetric geometries. The model was originally written to study pulverized coal combustion in utility boilers. COMO also consists of relatively independent modules that represent the major processes in pulverized coal combustion: flow, heterogeneous and homogeneous chemical reaction, and heat transfer. This modulari ty makes COMO readily adaptable to other types of chemical processes. The fluid flow module is designed to solve the full NavierStokes equations for steady-state, turbulent, two-dimensional axisymmetric (cylindrical) flows. The furnace is divided into control volumes. The governing equations are integrated over each control volume, assuming some distribution for the dependent variables. The assumed distribution is iteratively corrected until a converged solution is obtained. The partial differential equations are discretized using finite difference techniques3 • The module solves the resulting algebraic equations to predict the pressure and velocity fields. The k-e approximation is used to model turbulence and the overall macromixing in the furnace. |