|
The use of chemiluminescent radiation for monitoring of premixed flames has been explored in a swirl-stabilised combustor. The tests included natural gas flames and its blends with hydrogen, in order to evaluate the applicability of the techniques to cases with non-conventional fuels. A number of previous works demonstrate the existence of relationships among ratios of chemiluminescent emission from different excited radicals and, for example, equivalence ratio. The approach applied here is somewhat different. Besides studying bandfiltered radiation (due to OH∗), broadband emission as well as the magnitude of their fluctuations are also treated as variables representative of different flame states. Relationships among chemiluminescence parameters and meaningful variables were developed by means of artificial neural networks, which afforded notable versatility and generality. The potential of this approach is analysed in different situations, mainly conceived as ‘proof-of-concept’ tests. First, the possibilities of determining equivalence ratio and pollutant emissions from radiation signals are evaluated for the case of premixed flames of natural gas. Second, the properties of the fuel are modified by injecting different amounts of hydrogen, with the objective of assessing the potential of this approach for a range of fuels. Third, chemiluminescence signals are used as the only input in control tests in order to assess their applicability for advanced optimisation strategies. The good results achieved in all cases are thought to indicate that average and fluctuation of chemiluminescent emission collected with narrow- and/or broadband sensors may provide reliable indications of combustor state and performance in a wide range of combustion situations.
|
