We are interested in thermo-mechanical problems arising in the context of a practically relevant manufacturing process called sintering. These models can be defined using a non-linear material model, namely the Skorohod-Olevsky Viscous Sintering (SOVS) constitutive model. This SOVS model is used to predict macroscopic sintering behavior, such as shrinkage and density evolution. Also, it relies on material properties such as temperature-dependent viscosity and surface tension. However, high-fidelity simulations of coupled, macroscopic, thermo-mechanical models are computationally intensive. Furthermore, developing reducedorder models addressing the non-linearities is challenging due to the history dependence and presence of internal variables. Performing parametric studies, optimization, real-time control, or parameter estimation for such problems, thus, becomes infeasible. In order to accelerate sintering simulations for such multi-query scenarios, a surrogate model is vital. Here, we present a non-intrusive reduced-order modelling framework based on proper orthogonal decomposition and Gaussian process regression. Furthermore, we discuss the performance of such a surrogate model using different metrics for the two-parameter Arrhenius-type viscosity function
Published on 02/11/23
Submitted on 02/11/23
Volume Machine learning and uncertainty quantification for coupled multi-physics, multi-scale and multi-fidelity modelling., 2023
Licence: CC BY-NC-SA license
Are you one of the authors of this document?