Quality Inspection of Casting Product Using CAE and CNN
Seokju Oh, Jaegyeong Cha, Donghyun Kim, Jongpil Jeong
Abstract
Along with the Industry 4.0, Smart Factory is receiving great attention worldwide. In particular, quality control is the most important element of the production system. Of the many processes in manufacturing, the early process, casting is the biggest role in modern root industry. Casting is a manufacturing process in which a liquid material is usually poured into a mold to harden for solidify. The quality inspection process of casting products is largely divided into four stages. The last inspection phase, surface inspection, is inspected directly by the person, or by a vision system. The main key to quality inspection is accuracy and speed. Quality inspection using the vision system improves the competitiveness of future industries through fast and accurate inspection. Deep learning techniques have been widely used and studied for quality inspection problems. This paper analyzes the image of the cast product extracted through the vision sensor and proposes a Convolutional Neural Network (CNN) for quality inspection of the cast product and a Convolutional Autoencoder (CAE) to improve the learning quality of machine learning using a small amount of data.