Additive Manufacturing Distortion Compensation Based on Scan Data of Built Geometry

Abstract Additive manufacturing (AM) processes such as direct metal laser sintering (DMLS) are highly attractive manufacturing processes due to the ability to create certain geometries which would be prohibitive or even impossible to manufacture by other means. However, with such high thermal gradients which are usually present in these processes, manufacturing distortions may result in the creation of unacceptable parts. This paper presents an approach to compensate input STL files based on registration of the point cloud from sacrificial part builds. A novel strain energy based non-rigid registration algorithm has been developed for robust registration of data points to the original computer-aided design (CAD) model. A neural network based approach is used to learn the deformation of the geometry based on the deviation of the scan geometry. This network is subsequently used to modify the STL file to generate a new compensated STL file. The compensated STL file was validated by building parts and comparing the change in the part distortion.