Abstract:Constitutive models play a crucial role in materials science as they describe the behavior of the materials in mathematical forms. Over the last few decades, the rapid development of manufacturing technologies have led to the discovery of many advanced materials with complex and novel behaviors, which in the meantime, have also posed great challenges for constructing accurate and reliable constitutive models of these materials. In this work, we propose a data-driven approach to construct physics-consistent constitutive models for hyperelastic materials from experimentally measurable data, with the help of PDE-constrained optimization methods. Specifically, our constitutive models are based on the physically augmented neural networks~(PANNs), which has been shown to ensure that the models are both physically consistent but also mathematically well-posed by construction. Specimens with deliberately introduced inhomogeneity are used to generate the data, i.e., the full-field displacement data and the total external load, for training the model. Using such approach, a considerably diverse pairs of stress-strain states can be explored with a limited number of simple experiments, such as uniaxial tension. A loss function is defined to measure the difference between the data and the model prediction, which is obtained by numerically solving the governing PDEs under the same geometry and loading conditions. With the help of adjoint method, we can iteratively optimize the parameters of our NN-based constitutive models through gradient descent. We test our method for a wide range of hyperelastic materials and in all cases, our methods are able to capture the constitutive model efficiently and accurately. The trained models are also tested against unseen geometry and unseen loading conditions, exhibiting strong interpolation and extrapolation capabilities.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is how to construct a physically - consistent material behavior model from experimentally measurable data, especially for hyperelastic materials. Specifically, the author proposes a data - driven method based on partial differential equation - constrained optimization for constructing a physically - consistent hyperelastic material constitutive model. This method utilizes physics - enhanced neural networks (PANNs) to ensure that the model is not only physically consistent but also well - defined mathematically. ### Background of the Paper In materials science, constitutive models play a crucial role as they describe the behavior of materials in a mathematical form. With the rapid development of manufacturing technologies, many advanced materials with complex and novel behaviors have been discovered, which also poses a great challenge to the construction of accurate and reliable constitutive models for these materials. ### Solutions To meet this challenge, the author proposes the following solutions: 1. **Data - Driven Method**: Use experimentally measurable data (such as full - field displacement data and total external load) to train the model. 2. **Physics - Enhanced Neural Networks (PANNs)**: Ensure that the model is both physically consistent and well - defined mathematically. 3. **Partial Differential Equation - Constrained Optimization**: Define a loss function to measure the difference between the data and the model's predictions, and use the adjoint method to iteratively optimize the parameters of the neural network. ### Method Details - **Data Generation**: Use finite - element simulation to generate training data, simulating samples with non - uniform deformation, such as two - dimensional thin plates with randomly distributed holes. - **Model Architecture**: Adopt a modified PANN architecture, with inputs including the principal invariants of the right Cauchy - Green tensor and the determinant of the deformation gradient. - **Optimization Process**: Through the partial differential equation - constrained optimization method, use the adjoint method to calculate the gradient and iteratively optimize the model parameters. ### Results - **Training and Validation**: Tests were carried out on multiple hyperelastic material models, and the results show that this method can construct constitutive models efficiently and accurately. - **Generalization Ability**: The optimized model also shows strong interpolation and extrapolation abilities under unseen geometries and loading conditions. ### Key Formulas - **Strain Energy Density Function**: \[ W(F)=P(F, \text{cof}F, \det F) \] where \( P \) is a convex function. - **Volume Growth Condition**: \[ W_{\text{growth}}=\theta_g(J - \ln J), \quad \theta_g\geq0 \] where \( J = \det F \). - **Stress - Free Condition**: \[ \frac{\partial W_{\text{NN}}}{\partial F}\bigg|_{F = I}=0 \] Through these methods, the author has successfully solved the problem of constructing a physically - consistent hyperelastic material constitutive model from experimental data.

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

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

Data-driven computational mechanics: comparison of model-free and model-based methods in constitutive modeling

Automated model discovery of finite strain elastoplasticity from uniaxial experiments

Learning Physics-Consistent Material Behavior Without Prior Knowledge

Learning solutions of thermodynamics-based nonlinear constitutive material models using physics-informed neural networks

Physics-informed Data-driven Discovery of Constitutive Models with Application to Strain-Rate-sensitive Soft Materials

Inverse Modeling of Viscoelasticity Materials using Physics Constrained Learning

Constitutive model characterization and discovery using physics-informed deep learning

Learning Invariant Representation of Multiscale Hyperelastic Constitutive Law from Sparse Experimental Data

Learning solution of nonlinear constitutive material models using physics-informed neural networks: COMM-PINN

Weak-formulated physics-informed modeling and optimization for heterogeneous digital materials

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

Derivation of the Orthotropic Nonlinear Elastic Material Law Driven by Low-Cost Data (DDONE)

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

A Physics Preserving Neural Network Based Approach for Constitutive Modeling of Isotropic Fibrous Materials

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

A generic physics-informed neural network-based constitutive model for soft biological tissues

Thermodynamically Consistent Machine-Learned Internal State Variable Approach for Data-Driven Modeling of Path-Dependent Materials

Unbiased construction of constitutive relations for soft materials from experiments via rheology-informed neural networks