A Neural Network-integrated Elastoplastic Constitutive Model using Haigh–Westergaard Coordinates and Data Augmentation
This study introduces an elastoplastic constitutive model integrated with a neural network, utilizing Haigh–Westergaard coordinates and a data augmentation technique to enhance the prediction of material behavior under monotonic and cyclic loading conditions. The neural network employs the stress of...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-09-01
|
| Series: | Engineering Science and Technology, an International Journal |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2215098625001594 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study introduces an elastoplastic constitutive model integrated with a neural network, utilizing Haigh–Westergaard coordinates and a data augmentation technique to enhance the prediction of material behavior under monotonic and cyclic loading conditions. The neural network employs the stress of the trial state and the equivalent plastic strain of the current state as input variables, generating the updated stress, trial yield function, and plastic multiplier as output variables. The stress state in the 3D principal stress space is represented using Haigh–Westergaard coordinates, reducing it to three variables, which mitigates the out-of-distribution problem. Compared to a model trained on a principal stress space-based dataset, the model trained on a Haigh–Westergaard coordinates-based dataset demonstrated superior interpolation and extrapolation capabilities. Furthermore, the proposed neural network-integrated model, with optimized output variables, exhibited improved performance in stress prediction and plastic multiplier estimation compared to the non-optimized model. |
|---|---|
| ISSN: | 2215-0986 |