Image-based multiscale modeling with spatially varying microstructures from experiments: Demonstration with additively manufactured metal in fatigue and fracture

This manuscript presents a novel modeling framework to predict mechanical performance in material with spatially varying microstructure. The framework combines an efficient process model for AM, a database of experimental 3D images of defects in AM metal, and a microstructure based multiscale modeling method that leverages recent advances in reduced order modeling. Thus the examples presented will explore heterogeneous and processing dependent dispersion of voids in additively manufactured (AM) metals. The method presented here allows for parametric studies with repeated instantiations of different possible configurations of microstructures (images of defects) throughout the simulated part. Two demonstrations of the method are provided using a database of synchrotron x-ray computed tomography images of porosity collected at the Advanced Photon Source for Inconel 718 built with Laser Engineered Net Shaping (R): one case is high cycle fatigue crack incubation and the other is fracture initiation. In both, we show that the model can capture the effects on performance of variability within and between builds. Although not all variability is captured or quantified, the method shows promise for application in AM metals because of its unique ability to mechanistically connect part scale performance with individual microstructures and the distribution of these microstructures throughout the part.

Automated summary

This manuscript introduces a novel modeling framework for predicting mechanical performance in materials with spatially varying microstructure. The method allows for parametric studies exploring the dispersion of voids in additively manufactured metals, and demonstrates its ability to capture variability in part scale performance due to individual microstructures and their distribution. This approach shows promise for application in AM metals.