Framework for Hyperelastic Materials Using Neural Networks

This article presents a technical report on a novel framework for modeling hyperelastic materials using constitutive neural networks. The primary objective is to address the challenges in determining hyperelastic models by introducing a data-driven approach that simultaneously discovers suitable invariants and constitutive models for isotropic incompressible hyperelastic materials. The framework integrates the discovery process into a single neural network architecture, allowing for flexible adaptation to various material behaviors. The report details the methodology, which includes the identification of generalized invariants and the corresponding strain energy function directly from experimental data. The effectiveness of this approach is demonstrated through benchmark datasets for rubber and brain tissue, showcasing improved predictive accuracy and interpretability compared to traditional models. The study emphasizes the importance of accurately characterizing hyperelastic materials for applications in engineering and biomechanics, particularly in fields requiring high fidelity in stress-strain behavior under diverse loading conditions.