Application of data processing and machine learning techniques for in situ monitoring of metal additive manufacturing using acoustic emission data

Additive manufacturing (AM) is one of the most expanding sectors in the current industrial world because of its adoption in different industries. Prototyping was one of AM's main applications before, but now AM is also equally valuable for commercial components production in various industries such as aerospace, medical, consumer product, and others. Following its increased demand in the industry, the quality of the product is becoming a substantial concern for making sure of its safety and long-term usability. Several studies have been conducted on testing and quality inspection of the product by destructive and nondestructive testing (NDT) techniques. NDT can be used for testing without affecting the sample, which saves the material and production cost. Besides, in situ monitoring through different NDT techniques is also popular because it saves cost and time of production. Currently, in situ monitoring through acoustic emission (AE) is becoming one of the most popular techniques due to its provision of testing in the surface and subsurface of the material. In this study, the data acquired by AE is analyzes using data processing techniques, including wavelet transformation (WT). Because of the significant difference among process conditions in the graphical representation of the WT, the graphs of wavelet images are finally classified by a convolutional neural network (CNN). Proposed data and image processing techniques show that the acoustic data obtained from the AM processes can be efficiently classified for the purpose of process monitoring and quality control.