Virtual Inspection of Direct Ink Write Additive Manufacturing via Digital Twins

Advanced Manufacturing (AM) platforms fabricate parts by executing machine instructions that prescribe machine motions over time. Completing the manufacturing process requires non-destructive evaluation (NDE) of the parts to confirm that they meet specification; x-ray computed tomography (CT), non-contact height probes, and mechanical testing are common inspection methods. Inspection presents a signficant bottleneck in manufacturing enterprise. On top of this, there are no known methods to directly connect machine instructions to the final performance of a part. This work develops a framework and implement necessary components to address these critical issues. We aim to complement NDE of a physical part with virtual NDE of process-scale digital twins of the physical process, using direct ink writing (DIW) as an initial demonstrator. Our process-scale digital twins run the same machine instructions as the physical AM system they model and generate part-scale digital twins of what is actually built; these part-scale digital twins are then inspected virtually, analogously to how a physical part would be, but faster and automatically. We perform virtual geometric tolerancing via data driven analysis and acquire stress versus strain measurements from HPC simulations of compression tests, thereby linking process, structure, property, and performance of AM parts. Digital twin-derived datasets will be benchmarked against empirical data collected both during manufacturing and from conventional NDE. Digital/physical comparisons will occur in a virtual reality environment, which allows for intuitive interaction of spatio-temporal data among multiple users who need not be co-located. Data driven methods will be used to leverage empirical data to improve digital twin prediction fidelity, quantify uncertainty, and aid virtual inspection. While our primary goal is to speed production by reducing the inspection bottleneck, we also anticipate that data and methods we develop will enable adaptative AM process parameters, machine health monitoring, and process-specific manufacturing constraints in design, among other benefits.