DOI: 10.11799/ce202512028
Portable near-infrared spectroscopy coupled with MTFD-Unet for synergistic prediction of coal volatile matter and calorific value
ZHANG Xiaoyan
ZUO Yuhao
LUO Huifeng
WANG Ning
ZOU Liang
Abstract：Portable near-infrared(NIR)spectroscopy often suffers from spectral peak overlap,high inter-variable correlation,and sensitivity to sampling conditions in coal quality analysis,which lead to insufficient modeling accuracy and stability.To address these issues,we proposed a collaborative prediction method for coal volatile matter and calorific value based on a multi-task feature decoupling U-shaped network(MTFD-Unet).Firstly,the method improves data quality through iterative outlier removal and standard normal variate transformation,then it employs a Unet architecture to extract shared features,and further introduces a feature decoupling module to separate task-specific information,thereby balancing indicator correlation and specificity.Experimental results on 600 coal samples show that the proposed model achieves correlation coefficients of 0.8086 and 0.8584 for volatile matter and calorific value prediction,respectively.The prediction accuracy and stability significantly outperform multiple baseline models,and the method maintains strong robustness under noise interference.This research provides an efficient,accurate,and highly adaptable technical pathway for portable,rapid,and multi-index online detection of coal quality.
Keywords：near-infrared spectroscopyrapid coal quality analysisdeep learningmulti task learning
Publication Date：2025-12-20
Online Publishing Date：2026-01-29(First online date of this platform, not the publication date of the document)
Pages：10( 218-227 )