Comparison of response surface methodology (RSM) and artificial neural network (ANN) modelling for supercritical fluid extraction of phytochemicals from Terminalia chebula pulp and optimization using RSM coupled with desirability function (DF) and genetic algorithm (GA) and ANN with GA

Abstract

In the present study supercritical fluid extraction parameters for extraction of phytochemicals from Terminalia chebula (Haritaki) pulp were optimized using different approaches and compared. A central composite rotatable design (CCRD) was employed with four numerical factors viz. temperature, pressure, time and co-solvent flow rate. The responses were total phenol content (TPC), total flavonoid content (TFC) and antioxidant activity (2,2-diphenyl-1-picrylhydrazyl (DPPH) radical scavenging activity) of the extracts. Two different methods viz. response surface methodology (RSM) and artificial neural network (ANN) were used for modelling from the same set of experiments. Subsequently, optimization was carried by three different approaches viz. RSM coupled with numerical optimization by desirability function (RSM-DF), RSM coupled with genetic algorithm (RSM-GA) and ANN coupled with genetic algorithm (ANN-GA). Statistical analyses indicated that the models derived using both the methods i.e., RSM and ANN can be used to predict the response precisely, but RSM (R2 = 0.9987) method was somewhat found to be superior compared to ANN model (R2 = 0.9973). When comparing the optimization approaches, it was observed that results obtained from all the approaches were close to each other, but RSM-GA and RSM-DF approaches provided higher values of TPC, TFC and DPPH radical activity compared to ANN-GA approach. The values of TPC (mg GAE/mL), TFC (mg QE/mL) and DPPH (%) for RSM-GA were 428.03, 136.58, 92.63 respectively, for RSM-DF were 432.28, 137.36, 92.54 respectively and for ANN-GA were 414.25, 135.55, 91.32 respectively. From the present study it can be concluded that both RSM and ANN can be used for modelling of the processes with good predictability and optimization can be done using different approaches which will depend upon the specific process or the problem.