Predicting emulsion breakdown in the emulsion liquid membrane process: Optimization through response surface methodology and a particle swarm artificial neural network

To anticipate emulsion breakdown in the ELM process, the Box-Behnken design was used with an artificial neural network (ANN) and a metaheuristic approach, namely particle swarm optimization (PSO) and response surface methodology (RSM). Membrane stability testing began with an experimental component to collect data. The following parameters were used to estimate membrane breakdown: emulsification time (3-7 min), surfactant loadings (2-6% v/v), internal phase concentration ([Na2CO3]: 0.01-1 mg L-1), external phase to w/o emulsion volume ratio (1-11), and internal aqueous phase to membrane volume ratio (0.5 to 1.5). The PSO algorithm was used to determine the optimal ANN parameter values. The hybrid ANN-PSO model outperformed the RSM in identifying optimal ANN parameters (weights and thresholds) and accurately forecasting emulsion breaking percentages throughout the ELM process. The hybrid ANN-PSO method may be a valuable optimization tool for predicting critical data for ELM stability under various operating conditions.

Automated summary

The combination of Box-Behnken design, artificial neural network, particle swarm optimization, and response surface methodology was utilized to predict emulsion breakdown in the ELM process, with the hybrid ANN-PSO model outperforming the RSM in identifying optimal ANN parameters and accurately forecasting emulsion breaking percentages. This hybrid approach may serve as a valuable optimization tool for predicting critical data for ELM stability under various operating conditions.