Deep Learning Based Automatic Porosity Detection of Laser Powder Bed Fusion Additive Manufacturing

Laser Powder Bed Fusion (LPBF) is a widely utilized additive manufacturing process. Despite its popularity, LPBF has been found to have limitations in terms of the reliability and repeatability of its parts. To address these limitations, a deep learning model based on You Only Look Once (YOLO) was adapted to automate the detection of defect areas from scanning electron microscopic images of LPBF-manufactured parts. The data on the defect areas are then integrated into an Artificial Neural Network to correlate the process parameters with defects. The results show that the development of defects is stochastic in nature with respect to the input process parameters. The high variability of defects generated from the same process parameters makes it difficult to reliably predict the quality of the parts using only a process data-driven approach. This highlights the importance of in-situ monitoring of the system for reliable prediction of part quality.