Achieving optimal process design for minimizing porosity in additive manufacturing of Inconel 718 using a deep learning-based pore detection approach

Among metal additive manufacturing (AM) techniques, directed energy deposition (DED) has the advantage of being able to stack various materials, including difficult-to-cut materials such as Inconel 718. However, the pores generated during the DED process may reduce the hardness or abrasion resistance of the product and cause cracks or serious damage to the product. In this study, a method for detecting and quantifying pores using machine learning to detect pores in Inconel 718 products fabricated by DED was proposed. A pore detection model with YOLOv5 was established to detect and quantify porosity by learning 2448 pore images of Inconel 718 fabricated by DED. To minimize the porosity of Inconel 718 fabricated by DED, the effect of the DED process parameters on the porosity was analyzed using the design of experiments (DOE). Using the DOE, a prediction equation for predicting porosity was established, and the optimal conditions were obtained. The product was manufactured under optimal conditions, and porosity verification experiments and hardness test were performed. The resulting trends of the predicted values and the experimental results were consistent, and the hardness increased by 6.73% in the workpiece fabricated by DED compared to the workpiece fabricated by casting.