What problem does this paper attempt to address?

This paper aims to solve how to use experimental data to train a hyperelastic constitutive model based on Physics - Augmented Neural Network (PANN) to more accurately simulate the hyperelastic behavior of materials. Specifically, the research attempts to solve the following key issues: 1. **Training the PANN model with real experimental data**: - Most previous studies trained physically - enhanced constitutive models based on synthetic data (such as finite - element - analysis simulation data), while this paper attempts to use real experimental data for training. Real data contains more uncertainties and noise, which pose challenges to the model's design and performance. 2. **Improving the generalization ability of the model**: - Researchers hope that the trained PANN model can not only perform well under known loading conditions, but also maintain good prediction performance under unseen loading conditions and geometries. This involves the interpolation and extrapolation abilities of the model. 3. **Comparison with traditional models**: - The paper compared the trained PANN model with the traditional Neo - Hookean model and evaluated its performance under different loading conditions. The research shows that the PANN model is superior to the Neo - Hookean model in some aspects. 4. **Optimizing the mechanical information content of the data set**: - The quality of the data set is crucial for model training. Researchers emphasized the richness of mechanical information in the data set and pointed out that simply collecting a large amount of data from simple mechanical tests is not enough. It is necessary to ensure that the data contains sufficient mechanical information to support effective model training. ### Formula Explanation The loss function \(L[M(F), R]\) mentioned in the paper is used to minimize the difference between the stress \(M(F)\) predicted by the model and the actually measured reaction force \(R\). The specific optimization process can be expressed as: \[ \hat{p}=\arg\min_p L[M_{\text{PANN}}(p, F), R] \] where: - \(M_{\text{PANN}}(p, F)\) represents the stress predicted by the PANN model according to the strain \(F\). - \(R\) is the reaction force measured in the experiment. - \(p\) is the parameter of the neural network. Through optimization methods such as gradient descent, the optimal parameter \(\hat{p}\) that minimizes the loss function is finally found. ### Conclusion This research shows the potential of the PANN model in learning the hyperelastic behavior of materials, especially when dealing with real experimental data. The research results show that the PANN model not only performs well under known loading conditions, but also can provide accurate predictions under unseen loading conditions, reflecting its strong generalization ability. At the same time, the research also emphasizes the importance of mechanical information in the data set, providing a direction for further improving the model.

Training an AI hyperelastic constitutive model with experimental data

Experimental Learning of a Hyperelastic Behavior with a Physics-Augmented Neural Network

A Thermodynamics-Informed Neural Network for Elastoplastic Constitutive Modeling of Granular Materials

Neural networks meet hyperelasticity: A guide to enforcing physics

Mechanical Behaviors of Copolymer Elastomers: Experimental Characterization and Constitutive Modelling

Calibrating constitutive models with full-field data via physics informed neural networks

Viscoelasticty with physics-augmented neural networks: model formulation and training methods without prescribed internal variables

Automated model discovery of finite strain elastoplasticity from uniaxial experiments

A Physics-informed Assembly of Feed-Forward Neural Network Engines to Predict Inelasticity in Cross-Linked Polymers

Parametrised polyconvex hyperelasticity with physics-augmented neural networks

NN-EUCLID: Deep-learning hyperelasticity without stress data

Equilibrium-based Convolution Neural Networks for Constitutive Modeling of Hyperelastic Materials

Physics-augmented neural networks for constitutive modeling of hyperelastic geometrically exact beams

Accounting for plasticity: An extension of inelastic Constitutive Artificial Neural Networks

Neural Material: Learning Elastic Constitutive Material and Damping Models from Sparse Data

Learning the physics-consistent material behavior from experimentally measurable data via PDE-constrained optimization

Extreme sparsification of physics-augmented neural networks for interpretable model discovery in mechanics

Constitutive model characterization and discovery using physics-informed deep learning

Tools for multiaxial validation of behavior laws chosen for modeling hyper-elasticity of rubber-like materials