Fujitsu And Asahi Shuzo Launch Trial To Brew Sake With Predictive AI

Tuesday, May 1st, 2018 | 881 Views

Asahi Shuzo Corporation Limited, Fujitsu Limited, and Fujitsu Laboratories Limited, announced on 19 April 2018 that they will conduct a joint trial to brew DASSAI sake—a rice wine brewed and marketed by Asahi Shuzo—using a Fujitsu Laboratories-developed predictive Artificial Intelligence (AI) model that supports the brewing of sake.

As Japan’s society is ageing, however, labour shortages are expected in sake brewing facilities, and there are concerns in maintaining a stable supply and high quality of DASSAI. To address this issue, the company has been working for many years to record and systematise its experience in sake brewing.

Fujitsu and Asahi Shuzo have now launched a joint field trial to systematise the experience and knowhow of each employee with regard to brewing high-quality sake by applying Fujitsu’s artificial intelligence technology—Fujitsu Human Centric AI, Zinrai—to sake brewing.

By combining a mathematical model defining the process of sake brewing with machine learning that uses data obtained in the brewing of DASSAI, this predictive AI model technology will provide data to support an optimised sake brewing process.

In this field trial, by using such technology in actual DASSAI brewing operations, Fujitsu and Asahi Shuzo seek to improve both the validity and the accuracy of the predictive AI model, while evaluating how practical AI can be in sake brewing.

Details Of The Field Trial

The trial is planned to be conducted from April to June 2018. It will review and evaluate the applicability of the predictive AI model, which is developed with data accumulated by Asahi Shuzo—such as temperature and the proportion by weight of different ingredients—by calculating predicted values throughout the brewing process.

In addition, the companies aim to improve the accuracy of the predictive AI model by incorporating actual data obtained on-site during brewing, and the activities of employees based on their experience and intuition.

Conducted as part of Asahi Shuzo’s DASSAI brewing process, the elements that make up sake and other data will be measured by the company. Based on this data, the predictive AI model will be used to provide information supporting an optimal sake brewing process.

The company will undertake its sake brewing process with the goal of creating optimal sake, based on supporting information provided by Fujitsu and Fujitsu Laboratories. By conducting the trial for two cycles, the companies will improve both the accuracy of the predictive AI model, and evaluate its effectiveness, as well as the practicality of AI in sake brewing.

About The Predictive AI Model

Developed by Fujitsu Laboratories, it is a technology capable of computing information to support optimised processes in sake brewing by combining a mathematical model defining the procedure of sake brewing with machine learning, using measured values for components found in sake. This is based on Asahi Shuzo’s past brewing data and knowledge, including the biological process of fermentation.

Using the supporting information computed by the model, Asahi Shuzo can optimally control the devices used in the brewing process. By incorporating the data gained from this field trial and the feedback from the sake producer, the companies will improve prediction accuracy and the sophistication of the supporting information, thus enabling the system to assist with even more optimised sake brewing.

The three companies plan to use the results of this field trial to further the practical usage of AI in sake brewing operations, and will endeavour to consistently produce a stable supply of high quality sake.