In-situ Density Prediction in Metal Binder Jetting Using Powder Bed Imaging

Abstract

Metal binder jetting promises cost-effective end-use parts, but quality hinges on green part density. Traditional density measurement methods (e.g., Archimedes, geometric) require extra effort and equipment. This paper presents an in-situ density prediction tool using process images to reduce cost and time. The powder bed is photographed layer by layer with an integrated camera system. Process images are then analyzed using semantic pixel coloring in Python. Subsequently, the layer contours are approximated by unit cells, which are assigned a relative density by counting colored pixels indicating binder infiltration. Although predicted and actual green part densities have a weak linear correlation (R² < 30%), a significant linear relationship (R² > 96%) was found between predicted density and the drift from the geometric density, allowing a reliable forecast of the green part density with an average accuracy of 98.28%.