Real-Time Diagnosis of Plasma Molecular Temperature Based on OES and Elastic-Net Regression Analysis

The diagnosis of atmospheric pressure low-temperature plasma parameters is one of the research hot spots in plasma science. As important parameters, rotational temperature and vibrational temperature need to be measured in many application fields, such as plasma medicine, spraying, and combustion technology. Nowadays, as nonimmersion measurement technology, the plasma temperature diagnostic method based on optical emission spectroscopy (OES) shows greater advantages because it causes no pollution and interference to the plasma. However, the current measurement of molecular rotational and vibrational temperatures by OES is based on manual comparison and fitting, which has the disadvantages of cumbersome operation, heavy workload, and low efficiency. Therefore, an efficient method based on linear regression to process emission spectrum data has been proposed in relevant literature, but there is still room for improvement in accuracy and applicable range of temperature prediction. In this article, a theoretical spectral dataset with uniform and extensive temperature distribution is established, and some improvements and optimizations are made to the training variables and elastic-net regression model to obtain better fitting results. The fitting
$R^{2}$
of the experimental test dataset reaches 0.99427 and 0.96435 for these two temperatures, and the prediction time of a single spectrum data is less than
$5~\mu \text{s}$
, almost real time, which contributes new insights into the intelligent and efficient processing of plasma emission spectrum data.