日立、AIとSEMイメージングで性能予測

日立製作所と日立ハイテクは2025年6月、生産ラインの立ち上げを支援し、歩留まりを向上させるプロセス情報化技術を開発したと発表。開発した技術をリチウムイオン電池の試作ラインに導入して検証したところ、製品中間段階で製品の性能を高精度に予測することに成功。

In the manufacturing industry, there is a need to quickly launch production lines and further streamline manufacturing processes in order to respond to rapid changes in market needs. Hitachi and Hitachi High-Tech have proposed a “manufacturing process improvement solution” that utilizes a unique database and generative AI to achieve highly efficient manufacturing processes. However, achieving this requires a large amount of learning data, which poses challenges in terms of development time and cost.

Therefore, we focused on intermediate products during manufacturing and combined structural feature extraction technology with informatics technology to develop a “process informatics technology” that can predict product performance with high accuracy even with limited learning data.

- 広告

こちらもお読みください： ラティス、三菱電機と産業オートメーションで提携

One of the technologies developed this time is a “performance prediction model using structural features of intermediate products.” In this study, structural features of “electrode sheets,” an intermediate product in the lithium-ion battery manufacturing process, were extracted and applied to a machine learning model. As a result, it was confirmed that battery performance can be predicted with high accuracy even with limited data. In addition, the correlation between structural features and the objective variable, battery performance, was clarified.

The other is “Structural feature extraction technology using SEM image analysis.” Using an SEM developed by 日立 ハイテク、電極シート表面の画像データを取得・解析。これにより、電極内の凝集やボイド構造、成分分布を反映した構造的特徴を抽出する技術を開発。この技術を用いることで、電池製造の中間製品である電極シートの製造段階で、電池性能の良否を判定することが可能になるそうです。

ソース EETimes