Response surface methodology and artificial neural network modeling as predictive tools for phenolic compounds recovery from olive pomace
Ana Rita Silva, Manuel Ayuso, Oludemi Taofiq, Alexandre Gonçalves, Bruno Melgar, Lillian Barros
Abstract
This study optimized the extraction of three major phenolic compounds (oleuropein, tyrosol, and verbascoside) from olive pomace using microwave- and ultrasonic-assisted methods. Screening factorial design (SFD) and central composite design (CCD) were employed, and response surface methodology (RSM) and artificial neural networks (ANN) were used for data modeling. The microwave-assisted method in the SFD yielded higher compound amounts, with verbascoside showing a four-fold increase compared to the ultrasonic-assisted method. Factors like vessel diameter, ultrasonic power using UAE, and solvent acidity in both techniques had minimally impacted extractability. CCD-RSM revealed temperaturés significantly affect on oleuropein, but improved tyrosol recovery, with the effect on verbascoside being influenced by the temperature range. RSM and ANN integration enhanced understanding and prediction of factor behavior. Microwave-assisted extraction at 113 °C for 26 min, with minimum ramp time of 7.7 min, yielded 67.4, 57, and 5.1 mg of oleuropein, tyrosol, and verbascoside per gram of extract, respectively, with a prediction error ranging from 0.83 to 15.19.