Phenolic compound extraction from Nigerian Azadirachta Indica leaves: Response surface and neuro-fuzzy modelling performance evaluation with Cuckoo Search multi-objective optimization

Abstract

This study is aimed at modelling and optimization of phenolic compound extraction from Azadirica Indica Leaves (AIL) using Response Surface Methodology (RSM), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Cuckoo Search algorithm (CSA). The extraction experiments were conducted at Extraction Temperature (ET): (33.79–76.21oC), Process Time (PT): (2.79–4.21 ​h) and Solid-Liquid Concentration (SLC): (0.007929–0.018355 ​g/ml) with Total Phenolic Content (TPC) and Total Flavonoid Content (TFC) as dependent variables. Predictive Regression models (RM) for AIL extraction were developed from RSM in Design Expert software and compared with ANFIS model in Matlab environment. Multi-objective optimization of AIL extraction conditions was performed using CSA and Numerical Desirability Function (NDF) techniques; while sensitivity analysis of the process was performed using Monte Carlo Simulation (MCS). RM correlation coefficients (R2) of TFC and TPC are 0.988 and 0.949 respectively; whereas ANFIS model R2 for TFC and TPC gave 0.997 and 0.982 accordingly. MCS sensitivity analysis show the contribution of input variables (SLC: -56.3%; PT: 39.2% and ET: -4.59%) on TPC and (SLC: -0.9%; PT: -78.6% and ET 18.5%) on TFC respectively. The CSA optimum conditions gave 2.79 ​h, 40.540C, 0.01 ​g/ml with TFC 27.7 and TPC 1.06; while NDF optimum results gave 2.79 ​h, 40.540C, SLC 0.01 ​g/ml, with TFC 26.09 and TPC 1.272. The RSM and ANFIS models are satisfactorily predicted the process. The optimum conditions results obtained from the two methodologies are analogous. Therefore, the optimum results from this study could be used for AIL extract production plant design and techno-economic evaluation.