Student
Tailored Gaussian Process Kernels for Learning Representative Volume Elements (RVE) From EBSD Texture Data of Polycrystals
Please login to view abstract download link
To study the correlation between the physical properties and texture of polycrystalline materials, we employ homogenization simulations. Since constructing an RVE that incorporates texture information is essential, this work proposes using Gaussian Process Regression (GPR) with distance-based kernels and periodic distance to reconstruct the Pole Density Function. The application of GPR not only facilitates statistical analysis and infinite texture sampling, helping in the determination of RVE size, but also simplifies the creation of training datasets for neural networks aimed at accelerating homogenization simulations.
