Differential contrast tomography investigation of 3D microtexture for microstructure-driven high-cycle fatigue models through graph convolut

The overarching goal is to represent 3D microstructures through appropriate graphs to apply graph neural networks (GNN) approaches for high cycle fatigue damage prediction. These GNN methods facilitate comparatively physical modeling of polycrystalline materials and can potentially learn characteristic interactions at the microstructural scale, e.g. dislocation pile up at grain boundaries (GB). Especially in high-cycle fatigue, the microstructure sensitivity in fatigue is pronounced, leading to the requirement of comprehensive microstructural feature space, only attainable by 3D microtexture data. Therefore, DCT is supposed to be applied ex-situ (before and after fatigue) on eight specimens to capture the GB plane orientation and phase-contrast measurements to identify internal defects such as pores and inclusions. The multimodal data will be registered with experimental data from our tailored bending resonance fatigue apparatus and finally used for graph modeling and learning.