Enhanced lignin extraction and optimisation from oil palm biomass using neural network modelling

Lignin from industrial crops is the most promising feedstock which can be used to function modern industrial societies. However, it is very challenging to separate lignin from lignocellulosic biomass effectively. Commercial application of lignin faces many challenges concerning practical applications and sub-optimal extraction approaches. Investigating one factor at a time is a significant limitation in standard experimental protocols. The current processing conditions need to be improved, which can be performed by modelling the processing conditions and identifying the most appropriate process conditions to suit the market demands. In this study, both the response surface methodology (RSM) and an artificial neural network (ANN) model was developed for the enhanced lignin extraction from the available experimental data of our previous work. The effect of various operating parameters such as; extraction temperature, time, particle size range and solid loading affecting the lignin extraction efficiency was optimally analyzed. Likewise, this is the first study reporting a detailed comparison and prediction of lignin extraction using RSM and ANN. The models were evaluated through the coefficient of determination (R²), Root Means Square Error (RMSE) Mean Average Deviation (MAD) and Average Absolute Relative Error (AARE) showing that the ANN was superior (R² = 0.9933, RMSE = 1.129) to the RSM model (R² = 0.8805, RMSE = 4.784) for lignin extraction efficiency predictions using various species of oil palm biomass. The results showed the accuracy of the ANN model in the prediction of lignin extraction from empty fruit bunches (EFB), palm mesocarp fibre (PMF) and palm kernel shells (PKS), as compared to the RSM model.