Journal des bioprocédés et des biotechniques

Glutamate Fermentation Process Model Based on Gams with Double Penalty Approach

Abstract

Chunbo Liu1,2* and Xiangling Fang3

There are many fermentation variables which can influence fermentation product during fermentation process, including fermentation time, temperature, pH, oxygen uptake rate (OUR), carbon dioxide evolution rate (CER), dissolved oxygen (DO) and stirring speed. For understanding the relationship between fermentation product and associated fermentation variables, one current interest is the investigation of significant fermentation variables to construct an interpretable and stable fermentation model. In this study, the significant variables in the fermentation process of glutamate were selected based on the generalized additive models through a shrinkage approach. The GAM2 which is based on less fermentation variables after variable selection has the same value of adjusted Rsquare (0.972) and deviance explained (97.6%) as GAM1 which includes all the measuring fermentation variables. Results showed that the proposed approach not only can identify the significant variables in the fermentation process of glutamate, but also can improve the performance of the fermentation model.