Prediction of cadmium removal efficiency by emulsion liquid membrane using an artificial neural network

This study presented Intelligent Network's proposed methodology for forecasting the effectiveness of cadmium removal from wastewater using emulsion liquid membrane (ELM). The research examined the removal of cadmium from an aqueous solution in the internal phase of water-in-oil emulsions. ELM was composed of kerosene as the membrane phase, Span 80 as the surfactant, di-2-ethylhexyl phosphoric acid (D2EHPA) as the extracting agent, and hydrochloric acid (HCl) as the stripping solution. Experiments were conducted to investigate the effects of five parameters: surfactant concentration, feed-phase agitation speed, internal-to-membrane phase volume ratio, emulsion-to-feed volume ratio, and stripping-phase concentration in the internal phase. More than 97.4% of cadmium can be extracted in less than 15 minutes. This study evaluates the effectiveness of various learning algorithms, including gradient descent (GD), resilient back propagation (RB), gradient descent with momentum (GDM), gradient descent with learning momentum and adaptive rate (GDX), polak-ribiére conjugate gradient (CG), and levenberg marquardt (LM), in predicting the efficiency of cadmium elimination from wastewater using liquid emulsion membrane technology. The researchers developed a neural network model with 8 input neurons, 10 hidden neurons, and 1 output neuron. This feed-forward artificial neural network (ANN) incorporated various back-propagation training algorithms to simulate the cadmium removal process using ELM. The neural network model's predictions closely aligned with results from batch experiments, as demonstrated by a correlation coefficient (R2) of 0.9723 and a Mean Squared Error (MSE) of 0.0041, calculated in MATLAB. In conclusion, the ANN models demonstrated high accuracy and reliability in predicting cadmium removal efficiency, confirming their potential as practical tools for process optimization.