Machine-Learning-Assisted Fast Determination of the Combustion Characteristics of Municipal Solid Waste

Understanding the combustion characteristics of municipal solid waste (MSW) is pivotal to utilizing it as an energy resource. However, because of the diversity and complexity of MSW, rapidly determining its combustion characteristics is challenging. This study combined attenuated total reflectance–Fourier transform infrared (ATR–FTIR) spectroscopy with multiple machine-learning (ML)-assisted approaches for rapidly determining the combustion characteristics of MSW. A combined model that united preprocessing techniques, feature extraction, and ML algorithms was developed to fortify the modeling precision. Moreover, two loop approaches were applied for predicting the lower heating value (LHV); one approach directly correlated the LHV with the spectra, and the other approach predicted the LHV using the proximate and elemental compositions derived through spectral analysis. Notably, the spectral features of MSW combustion exhibit distinct fingerprint characteristics that serve as a basis for differentiating combustion behaviors among samples. Through hyperparameter optimization and comparative analysis across multiple models, this approach achieved high prediction accuracies (R2 > 0.92) for the LHV and carbon (C), hydrogen (H), oxygen (O), and nitrogen (N) elements, with slightly lower accuracies observed for sulfur (S) content (R2 = 0.83) and ash content (R2 = 0.70). This indicates that it is necessary to focus on the relationship between the predictor and the spectral expression and further study the functional group information contained in the sample, so as to further improve the accuracy of the model. Additionally, the model efficiently predicted the LHV indirectly based on organic element content and proximate analysis, yielding strong performance metrics (R2 test set = 0.92, R2 external validation set = 0.89, and RMSE = 1.07) and demonstrating rapid execution times (1.27 s). These findings underscore the feasibility of using ATR–FTIR and ML techniques for rapid characterization of MSW combustion properties, providing essential insights for optimizing waste treatment and resource recovery processes in industrial applications.