Quantitative hydrogen and methane gas sensing via implementing AI based spectral analysis of plasma discharge

In this report we explore the feasibility of a quantitative gas detection system concept based on alternations in spectral emissions of a radio frequency power generated plasma in presence of a target gas. We then proceed with training a deep learning residual network computer vison model with the spectral data obtained from the plasma to be able to perform regressive calculation of the target gas content in the plasma. We explore this concept with hydrogen and methane gas present in the plasma at know quantities to evaluate the applicability of the concept as hydrogen or methane detection system. We will demonstrate that the system is well capable of quantitatively detecting either of the gases efficiently while it is challenging to estimate hydrogen content in presence of methane.