Computational Chemo-Mechanics with Application to Multifunctional and High-Temperature Materials
Please login to view abstract download link
This presentation gives an overview of our current research activities in the area of chemo-mechanics-based material modeling. Building on experience in theoretical and computational methods for multi-physics problems, our interest here lies in the modeling of nonlinear, dissipative material behavior in which interacting thermo-chemo-mechanical processes play a dominant role. We aim to capture mechanisms such as stress- and temperature-biased phase transitions and chemical reaction-diffusion processes that directly influence, or even enable, the effective behaviors of modern engineering materials. In this general multiscale approach, model formulations may be considered on very different length and time scales. Recent work regarding analytical and computational scale-bridging methods is also briefly discussed. Three fundamental ingredients are addressed in this talk: (i) the theoretical model development, particularly regarding variational settings, (ii) the numerical treatment of these problems, and (iii) the calibration of thermodynamical properties via the CALPHAD method. Regarding the first aspect, we discuss the advantages and disadvantages of formulations for chemo-mechanical multifield problems through minimization and saddle-point principles. In terms of numerical solution schemes, we elaborate on the finite element implementation of such theoretical frameworks. Our particular approach builds on the flexible and quite general utilization of the UserELement interface (UEL) provided in the FE software package Abaqus. We further discuss ongoing collaborative work on the co-design of the variational model development and parallel solvers for chemo-mechanics problems, for which the MPI-parallel implementation instead is based on the software libraries deal.II, p4est and FROSch (Fast and Robust Overlapping Schwarz). Moreover, a concept is proposed in which thermodynamically informed material models are efficiently achieved via CALPHAD-trained neural networks. Representative numerical examples from a broad spectrum of technologically relevant studies --- ranging from hydrogels to multifunctional filters for the cleaning of steel melts --- are presented to demonstrate the validity and flexibility of our simulation frameworks.