Investigation of boundary layer flashback for non-swirling premixed hydrogen/ammonia/nitrogen/oxygen/air flames

Hydrogen and ammonia are attracting more attention as a carbon free energy carrier for gas turbines and other combustion systems. Hydrogen/ammonia fuel blends allow the flexible adjustment of the flame properties over a broad range, as the reactivity can be controlled with the hydrogen/ammonia ratio content. In this study, the boundary layer flashback (BLF) is investigated for various H2/NH3/N2/O2/air mixtures in a non-swirling premixed burner at normal pressure (101 kPa) and partly increased temperature (293 K to 557 K). The effect of nitrogen addition is studied by varying H2 to N2 ratios from 1:9 to 1:1 in H2/N2/NH3 mixtures. A H2 to N2 ratio of 3:1 is used to simulate dissociated ammonia. The results show that for an increasing nitrogen content and increasing dissociated ammonia content the flashback propensity decreases and increases, respectively. It was further found that the content of ammonia in the fuel mixture reduces much stronger the flashback propensity than the nitrogen content. The impact of oxygen enrichment and increased unburned mixture temperature are also studied for two selected H2/NH3/air mixtures. By increasing the oxygen content from 21 vol% to 31 vol% or the unburned mixture temperature from 293 K to 556 K, the mean flow velocity at flashback onset is found to double approximately. Furthermore, the boundary layer flashback of potential methane substitute mixtures are investigated as these mixture might allow to substitute methane in gas turbines. The substitute mixtures with a methane like laminar burning velocity show a higher flashback propensity compared to methane due to flame instabilities. By increasing the ammonia content and therefore reducing the laminar burning velocity, new substitute mixtures with a methane like flashback propensity were found. In order to describe the flashback propensity in terms of laminar burning velocity and other relevant parameters, a previously proposed model for the critical velocity gradient is extended for the new experimental data. The extended model is able to predict the measured data with good accuracy.