Prediction of density and sulfur content level of high‑sulfur coal based on image processing

Abstract

The density distribution of coal is important to improve the efficiency of the coal preparation process, and the sulfur content of coal is significant to control pollution from coal burning. To estimate the density and sulfur content level of high‑sulfur coal, a novel approach based on image processing and the support vector machine optimized by Grey wolf optimization (GWO-SVM) was presented. First, 42 color and texture features were extracted from coal pictures. Best features for different coal sizes were found using the grasshopper optimization algorithm to improve the efficiency of subsequent modeling and the accuracy of the models. An approximate exponential relationship was found between coal density and sulfur content by using curve fitting and model identification. Finally, the prediction models of coal density level and sulfur content level under different particle size ranges were established by GWO-SVM, respectively. The accuracy of density level models for 3–6 mm, 6–13 mm, 13–25 mm, and 25–50 mm were 90.0%, 87.5%, 83.3% and 90.0%, respectively, while those of sulfur content level models were 90.0%, 90.0%, 86.7% and 93.3%, respectively.