Optimization for the Bioconversion of Succinic Acid Based on Response Surface Methodology and Back-Propagation Artificial Neural Network

Abstract

At the base of primary culture medium, single factor experiment showed that CO2 and H2 and VH were distinct factors. The response surface methodology was employed to evaluate the interaction of those factors, and the result showed that there was obvious interaction between those factors, and that 74.60 g/L succinic acid was gained when the condition was as following: 66 % CO2 and 4.9% H2 and 5.9 mmol/L VH. Then a three-layer Back-Propagation artificial neural network was employed for the simulating and predicting, and the result showed that 78.10 g/L succinic acid was gained when the condition was as following: 67% CO2 and 4.8% H2 and 5.9 mmol/L VH. Comparison with the regressive analysis of the response surface methodology, the artificial neural network had better ability of predicting, since its predicting error was 0.17% while that of response surface methodology was 0.81%.