Coal-Fine Beneficiation Studies of a Bench-Scale Water-Only Cyclone Using Artificial Neural Network

Abstract

Coal fines generated in Indian coal preparation plants account for 25%–30% of run-of-mine (ROM) coal. Coal cleaning is receiving increasingly greater attention of process engineers in view of the increase in amounts to be handled as well as the difficult washability characteristics of high-ash-content Indian coals. Froth flotation is usually practiced in Indian coal washeries for washing the coals to bring down their ash content to acceptable limits. Because of the supply of feed coal from multiple sources, their different characteristics and composition, viz., varying mineralogy and ash content, presence of microfines, and their varying oxidation levels, the fine coal circuits, more often than not, fail to deliver consistent product quality and desired yields. Water-only cyclones have been used in most of the western countries for treating coal and mineral fines below 3 mm. However, the industrial use of these cyclones in India has not yet been put to practice in the coal-washing industry; the primary reason for this being that their design is unsuitable for high-ash content coals and therefore needs to be suitably modified according to the feed material characteristics. Highlighted in the present paper are the results of a case study of beneficiation of high-ash fine coal, using a water-only cyclone. The influence of two of the critical design variables, viz., cyclone length and solid concentration, on which the cyclone performance and the process yield (%) depend to a great extent, is described. Further, based on the experimental data of a water-only cyclone of varying lengths used for below a 3 mm coal beneficiation study, an attempt has also been made to develop a three-layer feed-forward artificial neural network (ANN) model, which is inherently trained using an error-back propagation algorithm. The results evince that the predictions from the ANN model are in good qualitative and quantitative agreement with the experimental observations, thereby validating the applicability and accuracy of the developed ANN model.