Pulverized coal is widely used in industrial energy supplies and chemical production. Its combustion efficiency and reactivity are closely related to its particle size distribution. Therefore, precise, fast, and reliable measurement methods are required to determine the particle size of pulverized coal. This study investigated the near‐infrared (NIR: 939.909–1671.329 nm) reflectance and visible light/near‐infrared (VIS/NIR: 350–1000 nm) reflectance of 11 particle size grades of pulverized coal. The K‐nearest neighbor algorithm based on metric learning was used with this noncontact, rapid, and simple spectral method to classify different particle sizes found in one type of coal. Experiments were conducted to compare the classification results and bands of other classifiers. Image processing technology was used to calculate the average particle sizes corresponding to different particle size levels. The average accuracy of classifying pulverized coal using NIR spectroscopy (939–1183 nm) was approximately 85.59%, whereas that of classifying pulverized coal using VIS/NIR spectroscopy (611–763 nm) was approximately 80.46%. The correlation between the spectral reflectance and pulverized coal particle size was clear. Even if a misclassification occurred, the misclassified label was close to the corresponding particle size of the correct label. The classifier based on metric learning adopted in this study had the best classification results when using NIR spectral reflectance (939–1183 nm). The spectral reflectance method can identify the particle size of pulverized coal quickly and without contact, thus providing a new method for the determination of pulverized coal particle size.
Wang et al. (Thu,) studied this question.