ANALYSIS OF THE STATE CHARACTERISTICS OF SAKE BREWING WITH A NEURAL NETWORK

The state characteristics of sake brewing were quantitatively defined as a numerical value, i.e. an SD value, expressing the time required for the alcohol concentration to reach 10 to 18 v/v%, assuming the temperature of the sake-mash (moromi) to be maintained at 15-degrees-C during brewing. A previously reported kinetic model of alcohol production was applied for this definition. The SD values varied from batch to batch, indicating that the target Baume reference pattern for the fuzzy control system reported previously should be selected in accordance with the characteristics of brewing. A procedure for estimating the brewing characteristics in the early stages was developed by applying error back-propagation learning with a neural network. This procedure enabled the fuzzy control system to generate the target Baume reference pattern according to the brewing characteristics.