ABSTRACT
The present paper deals with the results obtained for a 76 mm diameter Dense Medium Cyclone treating high ash Indian coals possessing difficult washing characteristics, coal crushed to −3 mm for improved liberation and better end utilization. The key influential parameters like the diameter of outlets and feed medium density were considered for analyzing their influences on yield and quality of the product. During this study, the influence of other variables such as magnetite to coal ratio, cyclone dimensions, feed pressure, medium properties, and feed rates were kept constant. The results of the test obtained thereof were analyzed for estimation of yield and ash content. In this paper, a comparative study of the parameter using an artificial neural network and the approach of response surface methodology (RSM) based on Box Behnken Design has been used. From both, the methodology separate equation has been developed for the prediction of quality and quantity of washed coal with the level of variables studied. The result shows that the artificial neural network implemented is able to provide a higher rate of accuracy than that of RSM. The yield % values predicted through ANN shows good agreement with the achieved experimental values (R2 0.9970) and demonstrated an improved correlation than RSM (R2 0.9831). Similar findings were observed for ash content (R2 0.9992) using ANN and found to be higher than RSM (R2 0.9698). Sensitivity analysis performed using the ANN model indicated that the Feed Medium Density is the most significant followed by the Vortex Finder Diameter and Spigot Diameter. A similar finding was also observed from the RSM-BBD for both the yield and ash content.