SUMMARY

In this presentation, Andreas will outline and propose a doctoral research path focusing on understanding stochastic microstructure functions (i.e., microstructure distributions) and their theoretical involvement in several practically important tasks in Materials Informatics. In the first task, conditional microstructure generation, I will explore the construction of robust, data efficient, microstructure statistics conditioned generative models by building computationally tractable approximations to stochastic microstructure functions. First, I will focus on microstructure generating algorithms conditioned on 1- and 2-point spatial statistics. Subsequently, I will explore the incorporation of higher order statistics. In the second task, I will explore the generation of large microstructure data sets by actively and intelligently covering the parameterizing space of 1- and 2-point statistics and, subsequently, using conditional microstructure generators to instantiate large data sets. Finally, in the third task, I will explore the usage of the statistical descriptors of stochastic microstructure functions (i.e., 1- and 2-point statistics) to quantify discrete dislocation structures. Here, I intend to analyze the statistical diversity of discrete dislocation simulations and demonstrate the importance of synthetic microstructure generation protocols.