Thermal Plasma Medical Waste Treatment: Data-ML Driven System Performance and Product Prediction

Hao-yang Shi,Ping-yang Wang

DOI: https://doi.org/10.1007/s12649-024-02593-2

2024-01-01

Waste and Biomass Valorization

Abstract:Thermal plasma technology, a cutting-edge solution for managing medical waste, has gained prominence following the recent global health crisis, which notably increased medical waste volumes. This study explores the application of machine learning (ML) algorithms to predict performance and product outcomes in thermal plasma treatment systems for medical waste. Our approach involved collecting extensive experimental data to create a comprehensive database, focusing on variables like waste composition, operating parameters, gas composition, and system performance. We utilized ML models such as support vector regression (SVR), Gaussian process regression (GPR), gradient boosting regression (GBR), and random forest regression (RFR) for data analysis, aiming to develop a predictive model that accurately predict the composition of the treated gas products and system performance metrics based on the components of the medical waste to be treated and the operating conditions. We selected representative medical waste samples for work parameter optimization and model validation. Notably, the RFR model demonstrated a high coefficient of determination (R2 > 0.90) and low error rates (RMSE < 1.57 and MAE < 1.21) in output prediction. Importance analysis revealed that working conditions and the proportion of inorganic materials in the waste significantly influenced the treatment outcomes. The four medical waste samples were processed under the best working conditions based on the PSO optimization, and the predictions of the RFR model closely matched the experimental values, confirming the accuracy and robustness of the model. The scope of this study entails the development of a thermal plasma system for medical waste treatment alongside the creation of a predictive machine learning (ML) model. Nevertheless, owing to constraints in experimental conditions and model intricacy, not all factors impacting waste treatment were addressed. Nonetheless, the findings underscore the potential of machine learning (ML) in intricate industrial and environmental engineering processes, offering novel insights into the factors influencing product yield and system efficiency, thereby fostering a deeper comprehension and optimization of the medical waste treatment process.