Online 3D Geometry Reconstruction for Direct Energy Deposition Based on Melt Pool Images

Direct energy deposition (DED) is an additive manufacturing technique developed to build and repair metallic objects with complex shapes. The deposition process involves a high-power laser that locally melts a metallic substrate, creating the so-called melt pool which is enriched by metal powder particles emitted by nozzles. Due to the printing process complexity, anomalies may occur. In the present paper, an innovative approach able to quickly detect geometrical defects is presented. A color camera is mounted coaxially to the laser beam providing images of the melt pool. The proposed algorithm combines all the images acquired for each deposited layer and estimates its actual geometrical profile. This profile can be compared to a reference one, for example, retrieved by sectioning the part CAD file or better from a thermo-mechanical simulation of the part printing process. From this comparison, the operator can evaluate the possible out-of-tolerance geometrical deviations at an early stage and without stopping the printing process, avoiding waste of energy, time, and raw materials. Estimated layer profiles could also be combined to generate a full 3D reconstruction of the built part. The proposed approach has been tested on real printed samples belonging to an experimental campaign aimed at evaluating the machine performance while realizing objects with significant overhang features. Some samples present evident geometrical deviations that helped in optimizing the algorithm parameters under variable conditions. Results have been validated by comparing the reconstructed profiles to the real ones, measured with an external high-resolution scanner.