Synergistic optimisation of fine coal oil agglomeration using response surface methodology and random forest coupled with decision trees

This study investigates the optimization of fine coal-oil agglomeration by focusing on critical parameters: pulp density, oil dosage, pH, agglomeration time, agitation rate, and oil type. The goal is to improve coal quality, specifically reducing ash value and moisture, while maximizing calorific value (CV), fixed carbon (Fixed C), and minimizing sulphur content. Using a combination of Response Surface Methodology (RSM), random forest, and decision trees, the study identifies optimal conditions for agglomeration. A pulp density of 20%wt yielded a relative 18% ash reduction and a 0.5% sulphur reduction. An oil dosage of 15% was found to be efficient, significantly enhancing the calorific value to 27.5 MJ/kg and fixed carbon to 48% and as dry (ad) moisture basis of 5.58%. The optimal pH of 4 preserves electrostatic properties crucial for effective agglomeration, while an agitation rate of 2800 rpm and agglomeration time of 12 minutes further improve coal recovery. The incorporation of advanced modelling techniques such as random forest and decision trees enhance predictive accuracy, ensuring reliable optimization of these parameters. These findings underline the potential of this approach to improve coal recovery processes. They also promote sustainable practices in the coal industry.