Analysis of Near-Infrared Spectra of Coal Using Deep Synergy Adaptive Moving Window Partial Least Square Method Based on Genetic Algorithm

Publication date: April 2019Source: Chinese Journal of Analytical Chemistry, Volume 47, Issue 4Author(s): Sheng-Hao WANG, Yong ZHAO, Rong HU, Yu-Yan ZHANG, Xin-Hua HANAbstractSpectral analysis is one of the available methods for rapid measurement of the common important properties of coal. Many studies on the spectral analysis of coal have been reported; however, it is still unknown whether the optimal coal quantitative calibration mode can be obtained from full spectrum or by selecting wavelength variables. In addition, the data preprocessing methods and their parameters are always determined by trial and error or experience, making it difficult to find the optimal model. To solve these issues, a new optimization algorithm named as deep synergy adaptive-moving window partial least square-genetic algorithm (DSA-MWPLS-GA) was presented in this study. This algorithm contained two important steps. In the first step of searching for basic mode, after initializing the parameters, the basic mode could be focused at a desirable quantitative model that satisfies the criterion by simultaneously optimizing the preprocessing methods and wavelength variables. In the second step of searching for evolution mode, some better models could be acquired based on the previous result of search for basic mode. Finally, the near-infrared spectra of 100 coal samples obtained from 12 different power stations in Northeast China were analyzed. The results showed that the optimal moisture, ash, volatile...
Source: Chinese Journal of Analytical Chemistry - Category: Chemistry Source Type: research