Process monitoring of deep drawing using machine learning

This study proposes a new processing method by using the count rate of the acoustic emission (AE) signal and machine learning. To analyze the AE count, machine learning, using a multilayered neural networks, is implemented for a deep drawing process. In a press processing, a quality inspection is often carried out for each lot in a later process. Once a failure occurs in the process, a large number of defective products may be produced due to the fast processing speed. In order to prevent this, it is important to immediately stop the processing just after the defect occurs. The AE signal data has been often used for monitoring the condition of process. However, it is easily affected by noise and lacks repeatability. Also it is difficult to handle the number of AE data due to its high frequencies of target signals. Therefore, the improvement of the data processing method and recognition rate is required and thus the machine learning approach is applied in this study.