Application of desirability function based on neural network for optimizing biohydrogen production process

A fractional factorial design was carried out to investigate the effects of temperature, initial pH and glucose concentration on fermentative hydrogen production by mixed cultures in batch tests and then the experimental data of substrate degradation efficiency, hydrogen yield and average hydrogen production rate were described by a neural network, based on which the simultaneous optimization of the three responses was performed by the method of desirability function. The analysis showed that the neural network could successfully describe the effects of temperature, initial pH and glucose concentration on the substrate degradation efficiency, hydrogen yield and average hydrogen production rate of this study. The maximum substrate degradation efficiency of 95.3%, hydrogen yield of 305.3 mL/g glucose and average hydrogen production rate of 23.9 mL/h were all obtained at the optimal temperature of 39.0 °C, initial pH of 7.0 and glucose concentration of 24.6 g/L identified by the method of desirability function based on a neural network. In sum, the method of desirability function based on a neural network was a useful tool to optimize several responses for fermentative hydrogen production processes simultaneously.