Surrogate Modelling of Crystal Plasticity Simulations for Austenitic Stainless Steels

Abstract

Understanding the effect of the microstructure of a material on the resulting macroscopic material properties and performance is important in predicting how a material or component will perform in service. To address this, numerous complex simulation techniques spanning different lengthscales and physical processes are typically used. Crystal plasticity modelling is an example of a complex computational technique that accounts for the microstructure of a polycrystalline material, making it well suited to the microstructurally informed analysis and design of components. However, crystal plasticity modelling is computationally expensive, making component-scale analysis infeasible, and limiting the opportunities for robust uncertainty quantification and propagation. Machine learning techniques, such as Gaussian process regression, have been emerging as a promising avenue to act as surrogate models to methods such as crystal plasticity at a fraction of the computational cost.
In this thesis, the application of surrogate modelling techniques to challenges within the field of crystal plasticity is explored. The first part of this thesis introduces a strain gradient crystal plasticity model before exploring the use of surrogate models to aid the development of the crystal plasticity models through sensitivity analysis and parameter calibration. The second part provides two example applications of using surrogate models to emulate the crystal plasticity
response: the first accounting for differences in the microstructural texture of a material, and the second considering both reversible and non-reversible cyclic loading applied to the crystal plasticity model. For each application, the quantification and propagation of uncertainties are considered, with discussion of how the methods presented can facilitate a probabilistic approach to bridging structure-property relationships across different lengthscales.

Date of Award	20 Jan 2026
Original language	English
Awarding Institution	University of Bristol
Supervisor	Matthew J Peel (Supervisor) & Mahmoud Mostafavi (Supervisor)