City Research Online

Abstract

Crystal plasticity simulations are widely used to capture anisotropic deformation, slip driven hardening, and texture evolution in crystalline metals. Their ability to represent microstructurally informed constitutive behaviour has made them valuable in both academic research and finite element based engineering analysis. However, their wider application remains limited by the high computational cost of local constitutive integration, particularly when nonlinear evolution equations must be solved repeatedly over many time increments. This thesis addresses that bottleneck by developing a deep learning based surrogate model for the constitutive update within a crystal plasticity framework. The study is deliberately restricted to a single material point representing a body centred cubic ferrite single crystal with fixed crystallographic orientation, so that the constitutive behaviour can be examined in a controlled setting without the additional complexities of polycrystalline aggregation or full structural simulation. A long short term memory neural network is trained to predict the history dependent constitutive evolution that would otherwise be obtained through the conventional iterative integration procedure. A dataset of 16,000 loading histories was generated to provide broad coverage of the response space and to support robust model training and evaluation. Hyperparameter tuning was carried out using Bayesian optimisation, and the structure of the dataset was further examined using dimensionality reduction and neighbourhood based analysis. In addition, a hybrid correction framework was introduced in which the machine learning predictions are combined with physics based constitutive updates to improve robustness and predictive accuracy. The results show that the proposed approach can reproduce constitutive responses with good accuracy while offering a computationally efficient alternative to the conventional local integration procedure.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA76 Computer software T Technology > TJ Mechanical engineering and machinery
Departments:	School of Science & Technology > Department of Engineering School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads