Temperature control of Ginjo sake mashing process by automatic fuzzy modeling using fuzzy neural networks

Automatic fuzzy modeling using fuzzy neural network (FNN) was attempted for use for temperature control of Ginjo sake mashing process. Data for 25 Ginjo sake mashings obtained from a commercial fermenter were used for the modeling. Models were constructed in four control regions. The acquired models could precisely output a set temperature, and the acquired rules coincided well with those experiences of Toji. The models were applied to temperature control of commercial scale mashing. Time course data were similar to those from a conventional control based on the decision of Toji. The Ginjo sake obtained was analyzed and quality assessed by expert sensory evaluation. The concentrations of chemical components and evaluation scores were confirmed to be similar to those obtained from the conventional control.