Mathematical-Based Gene Expression Programming (GEP): A Novel Model to Predict Zinc Separation from a Bench-Scale Bioleaching Process

Attempts to predict metal recovery accurately have been hindered by the complexity of the solvent extraction process, nonlinear effects, and the multitude of influencing factors. Conventional computing algorithms must be improved for solving practical problems due to insufficient or noisy data and complex multidimensional scenarios. In this study, a gene expression programming (GEP) model was developed to predict zinc extraction (ZE) from the bioleaching process utilizing the most influential parameters such as stirring speed (0-800 rpm), temperature (25-45 degrees C), pH (1-2.5), Di-(2-ethylhexyl) phosphoric acid (D2EHPA) concentration (5-20%), phase ratio (1:5-10:1), saponification degree (0-40%), and contact time (0-900 s) as the input parameters. Under optimal conditions of 20% D2EHPA, 15% saponification degree, 650 rpm stirring speed, pH2, and an A:O ratio of 1:1, zinc extraction reached 98.4%. The main motivation for constructing the GEP model is to present a rational mathematical model with more accurate results than statistical models. Furthermore, a model should be applicable for future usages to predict the value of metal recovery using independent variables accurately. The developed GEP model outperformed multiple linear regression and multiple nonlinear regression models with the adjusted R2 of 0.9551 and 0.9453, RMSE of 5.5100 and 7.2727, MAE of 0.0758 and 2.8308, MARE of 0.0260 and 0.0334, and VAF of 92.3069 and 87.9199 for the respective training and testing parts. Besides, the sensitivity analysis was performed using the cosine amplitude (CA) technique, and the stirring speed and D2EHPA concentration have the highest (rij = 0.931) and lowest (rij = 0.581) impact on the predicted ZE, respectively. The conducted study proves the appropriateness of the developed GEP model for ZE prediction.