Zeon advances AI data-sharing in synthetic rubber collaboration

Tokyo – Zeon Corp. has completed a demonstration project to link experimental data from different companies to improve AI’s ability to predict material properties.

The collaboration involved Tokyo-based Zeon Corp., which among other products produces synthetic rubbers, and Kentucky-based Zeon Chemicals (ZCLP), also a manufacturer of synthetic rubbers and elastomers.

While part of the same corporation, the two organisations have a “completely independent” data management process as well as formulations, Zeon explained in a 3 Dec statement.

As part of the project, the companies shared databases to improve AI models for creating formulas and ingredients without breaching confidentiality.

According to Zeon, the project used "secret computing technology", a process that allows calculations while keeping data encrypted.

Zeon expects the technology to enable different companies to share data to enhance the prediction quality of AI while safeguarding proprietary information.

The demonstration project, launched in 2023, worked to improve database quality by "linking and analysing" scientific data, including ingredients and the physical properties of formulas.

The two companies were able to “interpolate” each other's "weak data areas" and improve the effectiveness of the data linkage as well as the predictive performance of AI models.

Zeon said it applied a "unique conversion programme" to reconcile differences in compound types and naming rules, to enable data sharing.

For instance, it said, some of ZCLP's unique compounding agents are not in Zeon's database but were "investigated and linked" in the project.

This resulted in the creation of a compounding database of over 7,000 items.

According to Zeon, currently, data in Japan’s materials industry is only used within individual companies’ databases.

Data-sharing, said the Japanese group, is “essential” in order to strengthen international competitiveness of the industry.

Zeon's integrated development centre is currently working on a platform to securely share data and enhance AI models while maintaining confidentiality.