AI-based material property prediction models require an AI utilization platform for tuning and optimization

Since establishing its automotive suspension spring factory in 1939, NHK Spring Co., Ltd. has leveraged its strengths in metal processing to supply key parts supporting a variety of industries worldwide. As a global company with 54 locations in 14 countries, the company's operations include mobility-related businesses, which manufacture springs for passenger cars, motorcycles, trains, and ships, as well as automotive seats; information and communications-related businesses, which handle key parts needed for data center HDDs and semiconductor manufacturing equipment; and industrial and lifestyle-related businesses, which provide various parts needed for factories, plants, and home appliances. Currently, the company has established its 2026 Medium-Term Management Plan, which is based on the group's basic policy of "valuing people and contributing to society" and "further promoting sustainability activities." Under this plan, the company is working to create a rewarding workplace, accelerate its efforts to address global environmental issues, and strengthen its competitiveness through the promotion of digital transformation as a company-wide project.

The company is also promoting digital transformation in its research and development field, including the establishment of a department specializing in simulations and data science. As part of this, the company has long been considering creating an environment for utilizing AI, including machine learning. "The idea of using AI to shorten development lead times, optimize manufacturing conditions, and create new knowledge by making data-based decisions came from some of our young employees who had become interested in AI technology and had begun researching it ahead of time," recalls Toru Shiraishi, General Manager of the First Basic Technology Department of the Research and Development Headquarters.

The predictive model itself had been somewhat shaped through joint research with universities, but the lack of a database that could be used with AI for tuning and optimization posed an issue. "Previously, each researcher managed experimental results and verification data in Excel, which was an environment that was highly dependent on individual researchers. There was no data infrastructure that could be shared, and it was not an adequate environment for promoting the use of AI," says Keita Takahashi, chief of the department.

You can download the continuation of the case here.

Customer story 1st page sample