Gaseous Emission Spectroscopy for Equivalence Ratio Determination in a Thermoacoustic Combustor

This study explored an estimation method for equivalence ratio in a thermoacoustic combustor by combining emission spectroscopy and a random forest (RF) model. Spectra signals under different equivalence ratios were experimentally acquired using a spectrograph. The measured signals were primarily processed and five obvious spectral components were observed, including OH $\ast $ (309.348 nm), CH $\ast $ (430.482 nm), $\text{C}_{{2}}\ast $ (516.192 nm), K (766.188 nm), and $\text{H}_{{2}}\text{O}$ (927.119 nm). Characteristic peaks of the spectral components at various equivalence ratios were extracted to establish the raw features. Before regression modeling, the feature importance of the spectral components was analyzed and taken as a reference to further optimize the raw features. A nonlinear regression model was then established based on the optimized features and RF algorithm. Results demonstrate that the equivalence ratio can be predicted by the proposed model with an average determination coefficient higher than 98%. The spectral information proved an effective method for the prediction of the equivalence ratio in the thermoacoustic pulse combustor.