Detecting Defective Products Using 'Yangpum' Data

[Asia Economy Reporter Junho Hwang] The Korea Institute of Industrial Technology (KITECH) has developed a new smart factory platform for companies struggling to improve the efficiency of the die-casting process. The platform was introduced at the 'KITECH AI Smart Manufacturing Platform Forum' held on the 13th at the Forest Hall of Phoenix Pyeongchang Hotel.

The die-casting process, a casting method advantageous for mass-producing products with complex shapes, has a defect rate averaging 5-10%. Companies must consistently overproduce by more than 10% of their order volume, which undermines the profitability of small and medium-sized enterprises. To improve this with artificial intelligence (AI), data on defective products is required. However, collecting 300 data points requires over 15,000 experiments, making data acquisition impossible in actual production sites.

This platform can filter out defective products using easily collectible data from good products. Through deep learning, it builds an asymmetric database focused on good products and detects defective products caused by differences in pressure or temperature. Currently, three domestic casting companies are using this platform on a trial basis.

Lee Nak-gyu, president of KITECH, stated, "As of the end of last year, more than 12,000 smart factories have been deployed, accumulating a large amount of manufacturing data," adding, "We plan to establish an AI-based smart manufacturing platform that analyzes and processes the accumulated data to apply it practically for process improvement."