Using artificial neural networks for the intelligent estimation of selectivity index and metallurgical responses of a sample coal bioflotation by rhamnolipid biosurfactants

Abstract

ABSTRACT The present study investigated the effects of rhamnolipid (RL) biosurfactant on the flotation performance of a coal sample. The significance of the effect of RL on the system responses, including ash content, coal recovery, yield and separation efficiency, and process kinetics and selectivity, was statistically assessed using One-Way Analysis of Variance (ANOVA) methodology. The ANOVA results revealed that adding RL to the system significantly challenged the studied metallurgical aspects. Briefly, RL biosurfactant depressed coal flotation through physical interaction with the surface of coal particles through chemical bonding between the carboxyl group in the RL structure with those on the surface of coal particles. The potential mechanisms involved were schematically proposed. The correlation between the condition of RL addition and process responses was modeled using the Artificial Neural Network (ANN) approach. Rested in the mean squared error (MSE), root mean squared error (RMSE), and percentage error as the measures of model accuracy, the Levenberg-Marquardt algorithm (LMA) with [7–16-1] structure was found to be the most reliable algorithm to predict the process response. As evidenced, the correlation coefficient values of test data were 98.09%, 96.93%, 98.37%, 98.46%, 99.50%, and 97.37% for ash content, coal recovery, yield, separation efficiency, the rate constant, and selectivity index, respectively. These values confirmed that the process could be simulated using the ANN method with an appropriate structure.