Zeaxanthin production by Paracoccus zeaxanthinifaciens ATCC 21588 in a lab-scale bubble column reactor: Artificial intelligence modelling for determination of optimal operational parameters and energy requirements

Abstract

The operational optimization of zeaxanthin production by Paracoccus zeaxanthinifaciens ATCC 21588 in a bubble column reactor was performed by coupling genetic algorithm (GA) to an artificial neural network (ANN) model developed using experimental one-variable-at-a-time (OVAT) results. The effects of varying air flow rate (2-5 vvm) and inoculum size (4 and 8%) for different incubation time (30-80 h) were evaluated. Volumetric power input (P/V _L ) and energy input (E) to the bubble column were then correlated with the ANN-GA optimized conditions. A maximum zeaxanthin production of 13.76±0.14 mg/L was observed at 4 vvm using an inoculum size of 4% (v/v) after 60 h of incubation in OVAT experiments with corresponding P/V _L value of 231.57 W/m³ reflecting an energy consumption of 50.02 kJ during the fermentation period. The ANN based GA optimization predicted a maximum zeaxanthin production of 14.79 mg/L at 3.507 vvm, 4% inoculum size and 55.83 h against the experimental production of 15.09±0.51 mg/L corresponding to a P/V _L value of 202.03 W/m³ reflecting to a significantly reduced energy input (40.01 kJ). The proposed OVAT based ANN-GA optimization approach can be used to simulate similar studies involving microbial fermentation in bioreactors.