Modelling the extraction process parameters of amorphous silica-rich rice husk ash using hybrid RSM−BPANN−MOGA optimization technique

Abstract

This study was aimed to optimize process conditions for recovery of amorphous silica-rich rice husk ash (RHA) as a highly reactive precursor of silicon compounds for valorization of a non-conventional agricultural waste rice husk. The optimization was accomplished using a hybrid multi-objective genetic algorithm (MOGA) coupled with back-propagation artificial neural networks (BPANN) and response surface methodology (RSM). Herein, process conditions, namely leaching temperature (65–85 °C) & time (0.5–1.5 h) and calcination temperature (450–650 °C) & time (1−3 h) were considered as independent variables. The influence of process parameters on % crystallinity and volume % of silica in RHA were studied based on X-ray diffraction analysis and RHA recovery. In the hybrid RSM−BPANN−MOGA model, predicted data of BPANN were used as initial score and regression equations of RSM were used for development of fitness function. A set of optimal solutions were obtained as Pareto front and the final optimum design point was selected using TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) decision-making tool. The optimized process conditions were: leaching temperature: 75 °C, leaching time: 1 h, calcination temperature: 550 °C, and calcination time: 2 h. The optimized model was validated with observed results and found to be a well-fitted with absolute errors of 7.42%, 2.03%, and 4.26% for % crystallinity, RHA recovery, and vol% of silica, respectively. Further, a comparative study was performed between non-leached RHA and optimized RHA using several material characterization techniques. This study showed relative superiority of optimized treatment parameters with increased purity and decreased crystallinity and porous particle morphology.