Spectroscopy and Artificial Intelligence-driven coal quality detection: a review

Rapid and accurate analysis of coal quality is crucial for clean energy utilization. However, current industrial analysis is hampered by long cycles of 8–24 hours and destructive sampling. The fusion of spectroscopy and AI offers a revolutionary solution, capable of shortening detection time to within 2 minutes and increasing sample volume to approximately 1 ton, which is 106 times that of traditional methods. This paper systematically reviews the progress and challenges of this field. First, principles of five core spectroscopic techniques are outlined. The critical role of spectral preprocessing algorithms is then elucidated. Subsequently, the modeling evolution from machine learning to deep learning is traced. The paradigm’s practical evolution is analyzed through single-spectrum techniques and multi-spectral fusion. Finally, key challenges in data, model robustness, and system integration are addressed, providing a forward-looking roadmap for the field.