A quasi-2D integrated experimental–numerical approach to high-fidelity mechanical analysis of metallic microstructures

Integrated experimental-numerical testing on bulk metal alloys with fine, complex microstructures is known to be highly challenging, since measurements are restricted to the sample surface, thereby failing to capture the effects of the 3D subsurface microstructure. Consequently, a quantitative comparison of deformation fields between experiments and simulations is hardly possible. To overcome this, we propose a novel ‘quasi-2D’ integrated experimental-numerical testing methodology that hinges on the fabrication of μm-thin specimens with practically through-thickness microstructures over large regions of >100 μm. The specimens are fully characterized from both surfaces and tested in-situ to retrieve microstructure-resolved deformation fields. Simultaneously, the full microstructure is discretized in 3D and simulated. This allows for a detailed, one-to-one quantitative comparison of deformation fields between experiments and simulations, with negligible uncertainty in the subsurface microstructure. Consequently, a degree of agreement between experiments and simulations is attained which we believe to be unprecedented at this scale. We demonstrate the capabilities of the framework on polycrystalline ferritic steel and dual-phase ferritic-martensitic steel specimens. At the mesoscale, the methodology enables quantitative comparisons of the interaction between multiple grains, while, at the microscale, it enables advancement of numerical models by direct confrontation with detailed experimental observations. Specifically, it is revealed that the individual slip system activity maps, identified with SSLIP, near a grain boundary can only be reasonably predicted by enhancing the adopted crystal plasticity simulations with a discrete slip plane model. Additionally, the experimentally observed strong anisotropic plasticity of martensite can only be captured with a substructure-enriched crystal plasticity model.