Production of Zinc Chloride Modified Avocado Pear Seed Activated Carbon: Optimization of Preparation Conditions using ANN Modeling and RSM-Aided Box Behnken Design

Abstract

The exploration of the Zinc chloride activated avocado pear seed (ZAPS) production process employing Response surface methodology aided Box Benhken design and Artificial neural network algorithms was the main focus of this research. During the study, ANN and RSM models were deployed to assess the effect of process settings such as impregnation ratio, activation time and temperature on the measured BET surface of ZAPS. The RSM and ANN BET neural models were comparatively analyzed to ascertain the optimal process settings to produce the best response (BET surface area). The Analysis of Variance (ANOVA) outcome unveiled that the key independent variable(s) were activation temperature and impregnation ratio for ZnCl₂ modified avocado pear seed (APS) fabrication. The optimum preparation conditions for developing maximal BET surface area of 457.16 m².g^-1 were impregnation ratio (0.84), activation time (67.64) and activation temperature (813.94^oC). The ANN neural model was ascertained to be the better model with respective root-mean-square-error (RMSE) and overall regression coefficient (R) of 31.93 and 0.9838. Sensitivity analysis outcome revealed that temperature of activation had the predominant effect on ANN model performance with sensitivity (S) value of 88.9%.  This study demonstrated that ANN neural network and RSM can be applied as effective tools for optimization of the avocado pear seed (APS) alkali activation process.