Coal analysis based on visible-infrared spectroscopy and a deep neural network

• Published in: Infrared physics & technology 2018-09, Vol.93, p.34-40
• Main Authors: Le, Ba Tuan, Xiao, Dong, Mao, Yachun, He, Dakuo
• Format: Article
• Language: English
• Subjects: Artificial bee colony (ABC); Coal; Convolutional neural network (CNN); Extreme learning machine (ELM); Proximate analysis; Visible-infrared spectroscopy
• ISSN: 1350-4495; 1879-0275
• DOI: 10.1016/j.infrared.2018.07.013

•A coal proximate analysis method based on a combination of visible-infrared spectroscopy and deep neural networks.•This method can fate examines the moisture, ash, volatile matter, fixed carbon, sulphur and low heating value in coal.•Compared with traditional coal analysis, this method has unparalleled advantages and important value for practical applications in terms of efficiency, speed and accuracy.•We proposed a new CNN-AELM deep neural network algorithm. The proximate analysis of coal is the umbrella term for the six indexes that include the moisture, ash, volatile matter, fixed carbon, and sulphur contents and the heating value. Burning of coal creates carbon dioxide, sulphur dioxide and nitrogen dioxide which are main reasons causing air pollution. Therefore, before utilizing coal, it is indispensable to analyse coal. The traditional proximate analysis of coal mainly relies on chemical analysis, which is time-consuming and costly. Hence, a method to construct a coal analysis is introduced in this paper. By using the method to analyse moisture (%), ash (%), volatile matter (%), fixed carbon (%), and sulphur (%) contents and the low heating value (J/g). We first obtained different coal sample from different coal areas in China. Then, measured the spectral data through the spectral analysis instrument and extracted spectral features through a convolutional neural network. Finally, we applied the extreme learning machine algorithm to construct the prediction and analysis model of the spectral feature data. The experimental result shows that the model in the study can predict the components of coal. Compared with the chemical analysis method, this method has unparalleled advantages in terms of financial efficiency, speed and accuracy.