Case Study: Mitsui Chemicals achieves ±3% gas-product quality prediction in 20 minutes with NTT Communications' deep learning solution

Deep learning model predicts gas product quality and identifies quality anomalies in 20 minutes.

Mitsui Chemicals, a global manufacturer of materials for mobility, healthcare, packaging and more, needed to digitalize production to stay competitive and capture operator know‑how as experienced staff retire. After 15 years of testing early neural‑network approaches, the company still faced heavy manual data preparation and frequent model revisions that blocked practical AI adoption.

Partnering with NTT Communications, Mitsui applied deep learning to 51 types of raw sensor data to predict gas product quality within about 20 minutes. The solution achieved high accuracy (average +/-3% deviation), surfaced faulty sensors, reduced the need for prior data cleansing and operational costs, and is being expanded toward anomaly detection and maintenance optimization.

Mitsui Chemicals

Hideki Matsuo

Member of the Board Senior Managing Executive Officer Center Executive Production & Technology Center