Real-time monitoring of methyl orange degradation in non-thermal plasma by integrating Raman spectroscopy with a hybrid machine learning model

Abstract

Researchers have developed a hybrid machine learning (ML) model that has been integrated with Raman spectroscopy to enable real-time prediction of methyl orange (MO) degradation concentrations in a non-thermal plasma (NTP) environment. The model combines three ML algorithms, including Linear Regression (LR), Partial Least Squares (PLS), and Decision Trees (DT), and has been created and optimized to study the degradation process. The model demonstrates excellent predictive performance, achieving an even lower RMSE of 0.0209 and 0.0381 g/L and higher R² values of 0.9984 and 0.9969 for the training and test datasets, respectively. Based on this system, the impact of different plasma treatment parameters on MO degradation efficiency has been investigated. Among the results, MO (0.5 g/L) degraded completely whin 200 s when treated with an air plasma at a flow rate of 1 L/min and the applied discharge voltage of 20 kV. Furthermore, to understand the underlying mechanism of MO degradation, the individual contributions of different reactive oxygen species (ROS) to decomposition processes have been evaluated by employing effective scavengers, and the proposed degradation pathway was analyzed based on the identified intermediate products. This research introduces an innovative and highly efficient methodology for conducting online observation of pollutant degradation, optimizing the operational parameters of wastewater treatment in plasma, and enhancing their overall treatment effectiveness. Additionally, it establishes a solid foundation for integrating intelligent control systems, paving the way for fully automated decontamination processes.