In-situ process monitoring for metal additive manufacturing through acoustic techniques using wavelet and convolutional neural network (CNN)

Additive manufacturing (AM) is currently a widely used technology, even though a more profound understanding is still needed to track and identify defects during AM. The acoustic emission (AE) approach has gained a reputation in non-destructive testing (NDT) as one of the most influential and proven techniques in numerous engineering fields. Material testing through AE has become one of the most popular techniques in AM because of its capability to detect defects and anomalies, monitoring, and the progress of flaws. Various AE techniques have been under investigation for in-situ monitoring of AM. The preliminary results from AE exploration show promising results that need further investigation on data analysis and signal processing. AE monitoring technique allows finding the defects during the fabrication process so that the AM failure can be prevented or the process can be finely tuned to avoid damages or material waste. In this work, AE data recorded over the Direct Energy Deposition (DED) additive manufacturing process analyzed by machine learning (ML) algorithm for classification of different build conditions. The feature extraction method is used to obtain the required data for further processing. Wavelet transformation of signals has been used to acquire the time-frequency spectrum of the AE signals at each process condition, and convolutional neural network (CNN) image processing is used to identify the transformed spectrum of different build conditions. The identifiers in AE signals are correlated to the part quality by statistical methods. The results show a promising approach for quality evaluation and process monitoring in AM.