Hankook uses artificial intelligence to help build a better tire

SEOUL, South Korea—Like every other tire company, Hankook Tire & Technology Co. Ltd. conducts tens of thousands of experiments every year for compound development, according to Hakjoo Kim, Hankook vice president of R&D material development.

"As we collected all of this data, an idea occurred to us," Kim said. "Wouldn't we be able to predict the properties if we made an artificial intelligence program that could learn the data?"

This was the provenance of Hankook's new Virtual Compound Design system, which the tire maker claims will predict the characteristics of a tire compound through artificial intelligence analysis, eliminating the need for actual testing.

Hankook realized by 2015 that it had the relevant data to create what eventually became the VCD system, according to Kim. The company officially launched the project in 2016, he said.

"March of 2016 was also an important milestone and the driving force behind this project," he said. "It's when Lee Se-dol, the professional Korean go player and world champion, was beaten by AlphaGo. That victory gave us confidence in the power of AI."

Hankook spent the next three years preprocessing the data, developing the AI model, and upgrading and deploying the model, according to Kim.

The program got a boost early in 2019 when it joined with the Korea Advanced Institute of Science and Technology to create the HK-KAIST Digital Innovation Center.

The purpose of the center was to adapt data in the field to AI and lead the digital transformation, he said.

"It's a win-win for both parties, as Hankook Tire has the latest AI technology and KAIST can secure well-refined, up-to-date manufacturing data," Kim said.

"The VCD system is only one of the results of this synergistic partnership, and we expect the project to continue with even more results that are mutually beneficial to both Hankook and KAIST," he said.

Developing an AI program to predict the behavior of tire compounds was a very complicated process, according to Hankook. Tire compounds are made up of more than 15 types of materials, including synthetic and natural rubber and carbon black, it said.

There are many variables surrounding these materials, such as temperature, facilities, order of combination, and pressure, as well as the ratio of combinations of each material, Hankook said.

Generally, it takes six months to three years to develop a new tire compound, but Hankook expects to cut that time in half with the VCD system.

The system operates within the cloud platform by creating a "Digital Twin," according to Hankook. The Digital Twin is a twin of real-life objects that repeat the process of reflecting results derived from virtual simulations into reality, Hankook said.

The company has analyzed tens of thousands of units of data through cloud computing platforms such as Amazon Web Services and Google's TensorFlow, it said.

With the success of the VCD system, Hankook plans to expand AI technology to the entire process of developing tires, it said. This includes material selection, design, testing and production, the company said.

Also, Hankook plans to accelerate the introduction of data-based technologies based on accumulated data throughout the tire industry, including materials supply and demand, design, research and development, distribution and customer usage, it said.

Hankook plans to extend the use of AI to tire development and finding the optimal manufacturing process, with a goal of achieving this by 2023, Kim said. It will expand the methodology to production, distribution and customer use by 2026, he said.

"We believe our VCD system is one of the first compound development AI-applied models in the tire industry," Kim said.

"Given the great synergies between Hankook and KAIST, we plan to continue to build the partnership even beyond the AI technology itself. I anticipate even more exciting news from the Hankook and KAIST collaboration in the future."