Spectral Recognition Based on LGBM in an Ethylene Discharge Plasma under Different Gas Pressures

Spectral analysis is a commonly used effective method for plasma diagnostics. However, due to the large amount of plasma spectral data and the complex and unclear quantitative relationship with discharge parameters, the plasma diagnosis is sometimes not so straightforward and obvious to perform. In this paper, in order to realize the spectral recognition in an ethylene discharge plasma under different gas pressures, the plasma spectral recognition model based on LGBM (light gradient boosting machine) was constructed and compared with four typical recognition models. The results show that the LGBM model has the best efficiency for plasma spectra recognition and the accuracy can be as high as 99.16%. The spectral analysis on the ethylene plasma under different gas pressures shows that some specific wavelengths play the key roles in the feature recognition. In addition, this paper further analyzes the correlation between the characteristic wavelengths and the discharge gas pressures.