Training an AI hyperelastic constitutive model with experimental data

TL;DR

This paper presents a physics-augmented neural network trained on experimental data to model hyperelastic behavior, demonstrating improved performance over traditional models and emphasizing the importance of mechanical data in training.

Contribution

The paper introduces a novel physics-augmented neural network for hyperelastic modeling trained on experimental data, showing enhanced accuracy and generalization capabilities.

Findings

The AI model outperforms the standard Neo Hookean model on unseen data.

Experimental data effectively trains the neural network for hyperelastic behavior.

The model generalizes well to new loadings and geometries.

Abstract

A Physics-Augmented Neural network is trained to model a hyperelastic behavior. The dataset used for the training, validation, and test are displacement-force couples obtained from two experiments on a rubber-like material. One experiment was dedicated for the test, to assess the capacity of the model to generalize on unseen loadings and geometries. The trained AI model outperforms a standard Neo Hookean model identified on the same data. Particular attention is paid to the mechanical data information contained in the different datasets.