Research on fine coal classification based on a dual-modal spectroscopic detection system

As a critical front-end step in the clean coal utilization process, the accuracy of coal-sorting technologies directly affects the efficiency of subsequent processes and resource utilization. At present, coal preparation plants commonly rely on X-ray fluorescence (XRF) and near-infrared spectroscopy (NIR) for mineral sorting. However, these techniques face significant technical bottlenecks when dealing with complex coal matrices: XRF exhibits low sensitivity to light elements, while NIR tends to suffer from spectral overlap when analyzing organic – inorganic hybrid components. This is particularly problematic when processing high-ash coal, gangue-coal mixtures, and multi-mineral associated coal samples, where existing unimodal detection systems struggle to maintain absolute errors below 5%, thereby hindering the realization of high-precision sorting. To address these challenges, this study establishes a dual-modal detection system that integrates terahertz time-domain spectroscopy (THz-TDS) with Fourier-transform infrared spectroscopy (FTIR). By leveraging the complementary capabilities of these two techniques, the system effectively overcomes the dimensional and precision limitations of conventional detection methods. Methodologically, a dual-pellet preparation process was developed using a KBr/PE dual-matrix system, tailored to capture the characteristic response regions of organic components (3500–400 cm−1) and inorganic minerals (0.2–2 THz), thus mitigating signal interference caused by traditional single-matrix pelletization. Experimental results demonstrate that this technique achieves high classification accuracy for four representative coal types: bituminous coal, anthracite, lignite, and coal gangue. Its application within the dense medium separation system of coal preparation plants is expected to significantly alleviate issues related to medium density control caused by coal quality fluctuations, thereby providing critical data support for the development of intelligent, closed-loop sorting control systems.