LAUSR

Abstract The rapid determination of coal ash content is of substantial significance in coal preparation plants. A new ash content prediction method that is based on images of multiple ring light sources with various incident angles is proposed in this paper. Diffuse and specular reflections were separated based on image brightness in a single light source, and the intersections of these regions in multiple light sources were calculated to increase the monotonicity of the relationship between the image brightness after classification and the ash content. Then, an ash content prediction model was established by using the Logistic deformation function Ash = a 1 + e ( - b ∗ x - c ) + f to fit the relationship between the variables that were extracted from the images and the ash content. Then, the average of the preliminarily predicted ash content values was calculated to obtain the final predicted ash content. The ash content values of coal samples from four mines were predicted via this method, and the maximum root mean square error of the predicted ash content with respect to the actual ash content was 1.25%; hence, the prediction method realized high accuracy in ash prediction.