In automotive manufacturing, the press process is used to form steel sheets into parts.
However, defects can occur due to variations in production conditions, which poses a significant challenge.
Although sampling inspections are commonly used, it is difficult to inspect all products, and more advanced diagnostic methods are required.
This research develops methods to detect anomalies in formed steel sheets using pressure sensors installed in press machines.
High-precision diagnosis is achieved through frequency analysis of pressure signals using a one-dimensional convolutional neural network (1D CNN).
In addition, image-based diagnostics using cameras and autoencoders are proposed to detect abnormalities based on material flow characteristics.