Ash content prediction of coarse coal by image analysis and GA-SVM

Abstract

Ash content is one of the most important indexes of coal quality, and fast prediction of ash content is urgent and important for coal processing industry. The aim of this paper is to propose a method of ash content prediction of coarse coal by the use of image analysis and GA-SVM. Coal particles on the surface were randomly selected to measure the ash content, and a semi-automatic local-segmentation algorithm was proposed to identify the corresponding coal particle regions. Thirty-eight features were extracted, and selected by GA. Ash content prediction model was established by SVM, and K-CV method is used to determine the hyper-parameters (c, g) of SVM. RMSE and R-square were used to measure the prediction effects of ash content. Results indicated that the prediction effects of narrow size fractions are better than wide size fraction, and larger size fraction is more accurate than smaller size fraction in ash content prediction.