Two-phase fermentation for β-phenylethanol using Artificial Neural Network coupling with Particle Swarm Optimization Algorithm
Abstract: This article describes bioconversion of L-phenylalanine to β-phenylethanol with Kluyveromyces marxianus AS.2.1440 was performed in an aqueous/organic solv- ent biphasic system. At first, the appropriate organic solvent used for the bioconversion was studied. The results showed that oleic acid was the best organic solvent for establishing two-phase system. The fermentation conditions were optimized by Plackett-Burman design, artificial neural network (ANN) and particle swarm optimization (PSO). Firstly, the two-level Plackett–Burman design was applied to screen related factors from six variables that significantly influence β-PE production. The inoculum volume, temperature and the ratio of oleic acid phase to aqueous phase were identified as the most important significant factors. Subsequently, the steepest ascent experiment was used to approach the optimal region of the above three factors. Finally, ANN coupling PSO algorithm based on central composite design (CCD) and the steepest ascent experiment data was applied to predict these factors’ mutual interactions for β-PE production. The modeled maximum β-PE yield reached 1.32 g/L after a 48-h fermentation period under the optimized condition: inoculums (7.7%), temperature (31.7°C) and the ratio of oleic acid phase to aqueous phase (1.1). The optimized conditions allowed β-PE yield to reach 1.28 g/L after verification experiments test, which increased by 38 percent than 0.93 g/L in monophasic aqueous system.
Keywords：β-phenylethanol；Plackett-Burman；artificial neural network（ANN）；
particle swarm algorithm(PSO)
