Additive manufacturing source identification from photographs using deep learning

Additive manufacturing (AM) provides agile and flexible manufacturing in modern supply chains; however, there are concerns that AM may be vulnerable to inferior materials, poor process control, or counterfeit parts. There is a need for new measurement technologies capable of monitoring AM suppliers and certifying the quality and authenticity of AM parts and materials. This study demonstrates a method to predict the source of AM parts from part photographs using a deep learning model. A total of 9192 parts were produced from 21 printers, with three unique designs and four AM processes. A 2D photo scanner captures high-resolution images of each part; these images train a deep-learning model to predict the specific printer that produced each part. A novel analytical framework is developed for analyzing the high resolution image data. This framework achieves >98% prediction accuracy identifying the origin of 1050 parts. The model can authenticate the source of parts without cooperation from the manufacturer, potentially enabling applications in part authentication and detecting changes in materials or production processes. The research demonstrates the potential for high-resolution image data to be used for deep learning in manufacturing and shows that image-based source identification can monitor the quality and authenticity of AM parts.