Statistical modeling and optimization of ultrasound-assisted biodiesel production using various experimental designs

Abstract

The present study compares the performances of the regression models developed by the response surface methodology combined with the full factorial, Box-Behnken or face central composite designs applied for the ultrasound-assisted KOH-catalyzed methanolysis of sunflower oil. While all models led to similar optimal reaction conditions, the models based on the simpler designs had the smaller corrected Akaike information criterion values, the insignificant lack of fit and the more favorable statistical criteria than the model based on the full factorial design. Including fewer experiments, the Box-Behnken design can be recommended for the optimization of ultrasound-assisted biodiesel production processes.

U radu se porede performanse regresionih modela razvijenih na osnovu kombinovanja metodologije površine odziva sa punim faktorijelnim, Boks-Benken-ovim ili centralnim kompozitnim planom kada se primene na ultrazvukom podržanu bazno-katalizovanu metanolizu suncokretovog ulja. Iako su svi modeli dali slične optimalne uslove reakcije, modeli zasnovani na jednostavnijim planovima imali su manje vrednosti Akaike-ovog informacionog kriterijuma, neznačajnu vrednost odstupanja od modela i povoljnije statističke kriterijume u odnosu na model zasnovan na punom faktorijelnim planu. Boks-Benken-ov plan, koji zahteva manji broj eksperimenata, može da se preporuči za optimizaciju proizvodnje biodizela u ultrazvučnom reaktoru.