Overview

matmdl is a package implementing a Bayesian optimization framework using Gaussian process regression to parameterize the hardening models within crystal plasticity finite element method calculations, which are run through a user material subroutine integrated with Abaqus. Its main features are that it:

• Handles multiple input data sets simultaneously, e.g. mutliple orientations in single crystal deformation or multiple grain sizes
• Uses gradient-free optimization for the multi-dimensional parameter space describing slip system level material properties
• Wraps around an existing crystal plasticity implementation using Abaqus as the finite element solver
• Considers slight offset orientations in the case of single crystal plasticity, which may go unquantified and unreported in most experimental data sets in the literature

Basic Usage

Each run requires an input.toml file to specify the most common input options. The runnable modes are each modules:

• matmdl.run starts an optimization run
• matmdl.plot plots the results of a complete or incomplete optimization run

The input options are detailed in input.