OCR Text |
Show 7¥^\ MULTIVARIATE TESTING FOR MECHANISTIC INSIGHT IN A M O D E L INDUSTRIAL, NATURAL GAS-FIRED BURNER M. M. Miyasato and G. S. Samuelsen U C Irvine Combustion Laboratory Irvine, C A 92697-3550 Phone: (949)824-5950 ABSTRACT A statistical design methodology, commonly referred to as multivariate optimization or design of experiments, is applied to optimize the N O x and C O emissions from a model industrial, natural gas-fired burner. The multivariate approach incorporates testing of several input variables, or factors, at two levels (high and low) to determine the effect on the selected emissions, or responses. The factors for this optimization are the fuel jet injection velocity, number of fuel jets, air cross-flow velocity, air swirl intensity, and fuel jet axial location. The measured responses are N O x and C O emissions. In order to keep the dilution rate constant, all of the experiments were conducted at a constant excess air level of 10%. The results indicate that N O x emissions are predominantly affected by the cross-flow velocity, premixing distance, and swirl intensity. The C O emissions are affected by an interaction between the cross-flow velocity and swirl intensity. In this study, the jet velocity did not show a statistical affect on either the N O x or C O emissions. INTRODUCTION Oxides of nitrogen (NOx) are being regulated to lower and lower levels, with concurrently low C O emissions, due to their role in photochemical oxidant, i.e., "smog," formation. This is especially true in the South Coast Air Basin in southern California, where current regulations require less than 30 p p m of N O x for existing industrial burners ( S C A Q M D , 1998), with a Best Available Control Technology requirement of 5 ppm N O x for new installations greater than 4 MMBtu/hr but less than 35 MMBtu/hr. These requirements necessitate a new generation of low-NOx, industrial burners which can not only meet but beat these regulations in order to maintain economic vitality in the region. A s part of this goal, increased understanding of N O x formation and C O burnout in the burner system is required. To achieve this understanding, a statistical method of optimization, called multivariate testing or design of experiments, is applied to a model industrial, natural gas-fired burner. This technique is based on the premise that physical phenomena will obey statistical behavior, and if a population of samples is taken according to the structured multivariate protocol, the results can be statistically evaluated and mathematically modeled. Several iterations and refinements following this protocol provide a final system optimization based on the selected input |