Creating a Actionable Digital Twin: Real-time, In-situ 3D Scanning of Additive Manufacturing Processes

Conference Abstract: Additive Manufacturing’s (AM) layer-wise construction is susceptible to intrinsic process variation and fabrication errors. We present a novel methodology to rapidly capture and characterize part geometry error using a digitally integrated Structured Light Scanning (SLS) system and Signed Distance Function (SDF) approach. In-situ monitoring techniques have been proposed to detect build errors, but limit inspection to a localized deposition or final part state, and fail to capture how formed layers change throughout processing. Embedded 3D scanning can address these limitations; by scanning every layer, one can create a true, complete digital twin of the history of a part during its fabrication.
However, conducting layer-wise evaluation of high-resolution scans presents significant data management and optical challenges for compatibility with a production environment. With the use of GPU accelerated SDFs, millions of scan points can be analyzed within seconds over relatively large areas (250 cm2) to quantify geometric deviation of parts from the idealized design. Using a coordinated in-situ scanning array, a complete 360-degree representation of printed structures can be collected at a high spatial resolution (0.3mm) for every layer.
The proposed method is process agnostic, and can be easily integrated into any AM process where the part can be observed. In this talk we demonstrate its utility in both hybrid wire-arc AM and large-scale multi-axis robotic extrusion AM. With the use of in-situ scan data from both processes, part defects are captured, identified, and classified. Geometric deviation is recorded, and the z-height of subsequent depositions are adapted to compensate for the error, resulting in more accurate part. In the future, such data could be used to drive hybrid toolpathing and/or closed-loop additive control to correct in-process error.