Polyphenolic Extract Recovery from Azadirachta indica Leaves via Green Microwave-Assisted Extraction Technology: GA-ANFIS Prediction, Cuckoo Search Algorithm Optimization and ASPEN-Based Scale-Up Techno-economics

Abstract

This study presented hybrid genetic algorithm adaptive neuro-fuzzy inference system (GA-ANFIS) modelling, cuckoo search algorithm optimization (CSAO), and scale-up techno-economics of Azadirachta indica leaves’ microwave-assisted extraction (MAE). Box-Behnken design (BBD) of Design-Expert software was used to design MAE experiment using microwave power (520–1040 watts), extraction time (2–10 min), and solid–liquid ratio (0.4–1 g/ml) with two dependent variables: total phenolic content (TPC) and total extractible yield (TEY). Predictive GA-ANFIS and CSAO models for the MAE were implemented in Matlab 2019 environment. ASPEN base-case simulation and economic analysis for the phenolic extract recovery were achieved using CSAO optimal conditions. Process profitability uncertainty analyses were conducted with Monte Carlo simulation (MCS). BBD model results gave coefficient of determination (R2) 0.949 and 0.952 for TEY and TPC, respectively. GA-ANFIS results gave training R2 0.989, testing R2 0.999 and training R2 0.995, testing R2 0.992 for the TEY and TPC prediction, respectively. CSA optimal region for the extract recovery gave microwave power 1040 W, extraction time 4 min, S/L ratio 0.4, TEY 33.3%, and TPC 10 mg GAE/g dw. ASPEN base-case techno-economic results gave batch size (1.49 kg), batch time (135 min), total capital investment ($211,172), annual production cost ($42,386), net present value ($150,298), payback time (5.84 years), and rate of return (17%). The MCS profitability risk analysis revealed that the associated uncertainty for the process is less than 1%. Therefore, this work shows that Azadirachta indica leaves extract production using green technology is feasible.