Methodology for Neural Network-Based Material Card Calibration Using LS-DYNA MAT_187_SAMP-1 Considering Failure with...

Area of Science:

• Computational Mechanics
• Materials Science
• Machine Learning

Background:

• Material parameter identification (MPI) for Finite Element (FE) simulations is computationally intensive using traditional numerical optimization.
• Accurate material parameters are crucial for reliable FE simulations, especially in early-stage structural component development.

Purpose of the Study:

• To present and evaluate a novel neural network (NN)-based method for material parameter identification (MPI).
• To compare the performance of the developed ML-based framework against conventional optimization-based methods.
• To investigate the applicability of the NN-based method for complex material cards and failure models.

Main Methods:

• Development of a self-created, ML-based Python framework utilizing various machine learning algorithms.
• Calibration of LS-DYNA material card MAT_187_SAMP-1 and GISSMO failure model parameters using a virtually generated dataset.
• Comparative analysis between the NN-based method and a standard optimization-based method (LS-OPT) with varying settings.

Main Results:

• The NN-based method demonstrates significant time savings for repeated material parameter identification tasks.
• Successful calibration of complex material properties including elasticity, plasticity, tension-compression asymmetry, VPPR, strain rate dependency, and failure.
• The study confirms the general applicability of the NN-based procedure for calibrating intricate material cards like MAT_187_SAMP-1.

Conclusions:

• The presented ML-based approach offers a computationally efficient alternative to traditional MPI methods.
• The NN-based framework is particularly advantageous for rapid material card calibration during early phases of structural component design.
• The method shows promise for widespread adoption in FE simulation workflows requiring complex material models.