Experimental and predicted performance for the combustion of a low heating value gas in a swirl burner

Abstract

The combustion of a low heating value gas in a swirl burner is investigated. The investigation covers the development of a finite difference analysis of the flow and combustion processes in such a burner. In conjunction with the analytical work, an experimental program was conducted to provide detailed measurements of the three-dimensional velocity distributions within the flow field. The dissertation emphasizes the development and solution of the mathematical model.

The finite difference analysis uses the primitive variables of velocity and pressure to describe the flow field. Features of the solution algorithms of several previous authors are incorporated into the analysis. A unique feature of the current approach is the use of a non-staggered grid system. An additional feature is a very straightforward technique for handling boundary conditions which eliminates the need for special treatment of the finite difference equations at boundary points. The solution algorithm is given the acronym CENSIS, derived from CENtered-Cell-Implicitly-Staggered. To illustrate the incorporation of the algorithm into a computer code, a sample program is developed to solve a simplified problem which has a closed form solution. This program, CENCIS-T, is included. Calculations are presented for the swirl burner, and the predicted results are compared with experimental data. The program used for the calculation of swirl burner performance is a more general code called PRIMCO. The PRIMCO code includes variable density and viscosity effects and incorporates a two-equation turbulence model for the Reynolds stress terms. The PRIMCO code also uses a simplified, infinite reaction-rate combustion model. Because of the use of the non-staggered grid system, the CENCIS solution algorithm is less complicated than previous algorithms. As compared to a staggered grid system approach, the current algorithm requires approximately one- third the computations of the former approach. These advantages make this approach considerably easier to code and relatively easy to apply.