High-Accuracy Determination of Coal Quality Parameters Using Dual-Pulse Laser-Induced Breakdown Spectroscopy (DP-LIBS) Combined with Machine Learning Algorithms

Rapid and accurate determination of quality parameters (carbon content, ash, volatile matter, and calorific value) is essential for the clean and efficient utilization of coal. In this work, an innovative strategy combining dual-pulse laser-induced breakdown spectroscopy (DP-LIBS) with machine learning algorithms is proposed. Through systematic optimization of critical parameters, including laser wavelengths (1064 + 532 nm), pulse energy distribution (50 + 20 mJ), pulse interval (0.5 μs), and sampling distance (9.9 cm), spectral signal intensity and signal-to-noise ratio are significantly enhanced. Furthermore, by integrating spectral preprocessing (wavelet denoising, baseline correction, and Z-score normalization) with partial least squares regression, random forest, support vector machine regression (SVMR), and k-nearest neighbor, a high-precision coal quality prediction framework is presented. The experimental results demonstrate that under DP-LIBS conditions, the SVMR algorithm significantly reduces the prediction error for carbon content with a mean relative error (MRE) of 3.05% and a root mean square error of prediction of 2.44%, and yields optimal performance for calorific value prediction with an MRE of 3.74%. This work provides a high-throughput, low-error solution for online coal quality detection, advancing the practical application of LIBS technology in energy quality monitoring and carbon neutrality pathways.