A novel integrated framework for reproducible formability predictions using virtual materials testing

Abstract

Background: Formed aluminium alloy sheet materials are increasingly adopted in production processes such as vehicle manufacturing, due to the potential for weight-saving and improved recyclability when compared to more traditional steel alloys. To maximise these benefits whilst maintaining sufficient mechanical properties, the link between formability and microstructure must be better understood. Virtual materials testing is a cost-effective strategy for generating microstructure-informed formability predictions. Methods: We developed an open-source hybrid framework, combining experimental and computational tasks, for generating reproducible formability predictions. Starting with experimental texture measurements and stress-strain curves, we calibrated crystal plasticity (CP) model parameters. The framework used these parameters to perform a large set of multiaxial full-field CP simulations, from which various anisotropic yield functions were fitted. With these anisotropy parameters, we then employed a Marciniak-Kuczynski finite-element model to predict forming limit curves, which we compared with those from experimental Nakazima tests. Results: We executed the workflow with the aluminium alloy Surfalex HF (AA6016A) as a case study material. The 18-parameter Barlat yield function provided the best fit, compared to six-parameter functions. Predicted forming limits depended strongly on the chosen hardening law, and good agreement with the experimental forming limit curve was found. All of the generated data have been uploaded to the Zenodo repository. A set of Jupyter notebooks to allow interactive inspection of our methods and data are also available. Conclusions: We demonstrated a robust methodology for replicable virtual materials testing, which enables cheaper and faster formability analyses. This complete workflow is encoded within a simple yet highly customisable computational pipeline that can be applied to any material. To maximise reproducibility, our approach takes care to ensure our methods and data — and the ways in which that data is processed — are unambiguously defined during all steps of the workflow.