Sono electro-chemical synthesis of LaFeO3 nanoparticles for the removal of fluoride: Optimization and modeling using RSM, ANN and GA tools

The aim of this work was to model, optimize, and compare fluoride removal by LaFeO3-NPs using the RSM (Response Surface Methodology), ANN (Artificial Neural Network), and GA (Genetic Algorithm) techniques.The input variables considered were pH, time, temperature, LaFeO3-NPs dose, and fluoride. The CCD (central composite design) plan was exercised for the analysis of RSM, and ANN to determine their capabilities of prediction of the response. Their performances were evaluated using the regression coefficient (R-2), RMSE, SEP, and the AAD. Also, RSM and GA were used to maximize the response and their optimum conditions evaluated. Both RSM (R-2 = 0.9970, AAD = 0.00001, RMSE = 0.0037, SEP = 0.0042) and ANN (R-2 = 0.9919, AAD = 0.00044, RMSE = 0.0066, SEP = 0.0074) gave high degree of accuracy. The model equation obtained for the process through RSM was adequate. The GA and RSM gave very close values for the optimization of the fluoride reduction process; RSM gave optimum fluoride removal of 96.35% (at pH 8.6, time = 75.03 min, temperature = 34.9 degrees C, dose = 0.225 g, and concentration = 23.68 mg L-1) while the GA gave 96.30% (at pH 10, time = 120.39 min, temperature = 28.41 degrees C, dose = 1.030 g, and concentration = 16.31 mg L-1). But from the confirmation experiments, RSM and GA data gave 96.52% and 96.63%, respectively. RSM, ANN, and GA were capable of modeling and optimizing the elimination of fluoride using LaFeO3-NPs.

Automated summary

The study aimed to model, optimize, and compare fluoride removal by LaFeO3-NPs using RSM, ANN, and GA techniques. The results showed that both RSM and ANN had high prediction accuracy, and RSM and GA had similar performances in optimizing the fluoride reduction process.