Application of artificial neural network in the modeling and optimization of humic acid extraction from municipal solid waste biochar

Abstract

This study investigated the extraction of humic substances from municipal solid waste (MSW) biochar during chemical activation for potential use of spent reagents and further application of pyrolysis products. It also determined the applicability of artificial neural network (ANN) in the modelling and optimization of the extraction process. The effects of KOH concentration (0.25–0.75M), extractant dose (10–30gL−1), contact time (1–12h) and precipitant volume (0.5–2.5mL) on the humic acid (HA) yield were evaluated and optimized using ANN. The extracted HA was characterized using FTIR, UV–vis and ultimate analyses. The optimum ANN model was obtained with 12 neurons in the hidden layer resulting to R2>0.99 for the training, validation and test sets. The optimized input parameters generated from the ANN model yielded a difference of 3.71% between the predicted HA yield of 180.57mgg−1 and experimental value of 187.52mgg−1. Calculation of the importance of each factor showed that KOH concentration (54.3%) and precipitant volume (26.9%) had greater effect on the HA yield than contact time (10.8%) and extractant dose (8.0%). The results confirmed the reliability of the ANN model in predicting the extraction of HA from MSW biochar.