OCR Text |
Show Analysis of Local Turbulent Reaction Rates from CFD Predictions of a 2MWt Natural Gas-Fired Turbulent Diffusion Flame Peter P. Breithaupt N.V. Nederlandse Gasunie, Gasunie Research, P.O.Box 19, 9700 MA Groningen, The Netherlands August 1997 Abstract The present study is concerned with mathematical modelling of natural gas flames. The attention is focused on CFD-predictions in the near burner zone of a 2 M W , type-2 swirling natural gas fired turbulent diffusion flame. A combustion model for turbulent diffusion flames generally calculates local turbulent reaction rates, which can be used in computational-fluid-dynamics (CFD) software packages as source terms for the species mass and the enthalpy transport equation. Unfortunately, turbulent reaction rates are difficult to measure directly, and profiles of time-averaged temperature, species concentration, velocity components and mean local velocity fluctuations are commonly used to validate the model by companson of predicted results with experimental data: "The better the agreement, the better the combustion model used". O n the other hand, measurements and/or predictions of stationary, time-mean flow and species properties do not always serve the combustion engineer's need to know "where something is happening". Therefore, the first objective of this paper is to show that turbulent reaction rates generated by means of C F D predictions may be used for analysis required for engineering purposes. The strength of the widely applied eddy-break-up (EBU) combustion model to predict swui-stabilised natural gas flames is demonstrated. However, modern gas-fired combustion technologies incorporate less intense stabilisation which results m less distinct "main-flame-zones" of high heat release and high peak flame temperatures. Consequently, the second objective of this paper is to propose a "novel" combustion model. A statistical analysis of the predicted reaction rates yields a formulation for an improved Arrhenius law-type model, called "turbulent-kinetic" (TK) combustion model. The proposed model accounts not only for the strong temperature dependency of chemical reactions at lower flame temperatures but as well for small scale turbulent fluctuations within the chemical reaction zone. The "novel" model stays attractive for use in C F D applications where computational speed is a critical consideration. At present the proposed combustion model is validated against detailed inflame measurements from a variety of semi-industrial natural gas flames. Introduction In the process leading to improved combustion and heatmg processes higher system efficiencies and present and forthcoming emission regulations are the major measures for gas-fired installations. During the development phase of new combustion equipment computational fluid dynamics (CFD) plays an increasing role. While CFD is entering the engineer's workfloor is has to combme fast solution turnaround cycles in the order of a day's run with an accuracy level suitable to generate valuable engineering information. Weber et.al. (1993) classified the engineering information required for designmg industrial flames suitable for a particular process into three classes: First-order information encompasses rough qualitative estimations of heat fluxes and flame shapes. Second-order information is characterised by improved accuracy of results concerning temperature, oxygen concentration and heat fluxes. Predicted flame temperatures and oxygen and unbumt hydrocarbon concentrations are expected to be within an accuracy of ±100K, respectively ±0.3% by volume to deliver valuable second-order information. Finally, third-order information deals with increased knowledge of detailed in-flame chemistry including identification of regions of high pollutant formation rate. To allow a confident analysis of CFD-predictions the goal of most engineering calculation is to meet the level of second-order information. To match fast solution turn-around cycles present commercially available C F D codes make use of combustion models like the Eddy-Break-Up model (EBU) by Magnussen and Hjertager (1972) and apply the k-£ turbulence model to account for the turbulent flow transport phenomena. Both models are attractive for use in C F D applications, where computational speed is a critical consideration. Moreover, recent studies have shown that, when predicting swirl-stabilised gas flames, applying the E B U combustion model and the k-£ turbulence model, predicted temperatures and species concentrations are commonly obtainable within the defined requirements for second-order information (Breithaupt et al., (1994): Peters and Weber, (1995)). Type-2 swirl-stabilized flames possess a rather intense and confined area, where local volumetric reaction rates are by order of 1 |