Deep learning enabling quality improvement in rotogravure manufacturing

Advances in Computer Vision have helped the manufacturing industry achieve superior quality norms with a minimal inspection time due to optical quality surveillance systems. These inspections most often take place at the end of the value chain, insuring the quality standards of the manufactured pieces. The downside to this approach is that defective parts can still continue through the value chain. Wasting a lot of resources and increasing the lead time. To avoid this drain, the machines in the value stream should only produce error-free parts or at least detects them. An optical quality inspection system at every production step would add a high price cost. For this reason, existing sensors should detect unwanted states. With structured data, a person with specific domain knowledge could rate this. This is a tough task, as a lot of unknown factors can influence each step. Therefore, this paper proposes steps to improve quality in rotogravure manufacturing using deep learning. Further research will be conducted in the coming months to expand these results. The proposed procedures will be applied to live data of a rotogravure manufacturing site and the effectiveness of this approach will be analysed.