A scientific benchmark for elasto-plastic constitutive modeling - Part II : blind predictions, calibration strategies, and benchmark results

Abstract: This paper presents the second part of the NUMISHEET 2025 scientific benchmark on elasto-plastic constitutive modeling. The study builds on the experimental dataset and benchmark problem definition introduced in Part I. It evaluates 19 blind prediction results submitted by 17 international teams from academia and industry. The benchmark task included calibrating constitutive models for the dual-phase steel DP800HHE, exclusively based on Part I data. The calibrated constitutive models were used to simulate the validation experiment MUC-Test, without prior access to the experimental validation results. The submitted solutions span a broad spectrum of modeling strategies, ranging from different phenomenological hardening models and yield locus formulations to crystal plasticity approaches. A systematic comparison with experimental ground truth data (comprising punch force-displacement curves and local strain distributions) was conducted using normalized error metrics. The results of this study reveal the characteristic strengths and limitations of different modeling approaches. They also highlight the role of anisotropy and hardening law selection. Furthermore, they demonstrate the sensitivity of predictions to calibration strategies. Providing an open dataset, a transparent evaluation methodology, and a comprehensive discussion of outcomes signifies an advancement in reproducibility and objectivity in constitutive modeling. This benchmark serves as a basis for further validation, comparison, and development of constitutive models. Dataset: https://zenodo.org/records/18431261