PSI - Issue 79

Daniel Leidermark et al. / Procedia Structural Integrity 79 (2026) 190–197

195

a)

b)

ˆ σ a MPa

ˆ σ a MPa

σ a MPa

σ a MPa

c)

d)

ˆ σ a MPa

MAPE%

σ a MPa

Epochs

Fig. 4. Results from the scatter augmentation method, showing a) training and b) test results, and c) MAPE vs epochs, and d) verification.

6. Conclusions

Based on this data augmentation study of machine learned constitutive models, the following can be concluded:

• Both the linear and scatter augmentation approaches work well, and capture the underlying material responses. • The linear augmentation technique is simple and generates best training predictions. However, the method can be seen as a pure artificial way to generate new artificial data, and high local densities surrounding certain sampled materials that gives lacking correspondence with the verification data. • The scatter augmentation method generates a slightly higher error than the linear one. Yet, more physically representative artificial data are generated giving higher degree of exploitation of the stress-strain space, which in turn leads to higher degree of compliance with the verification data compared to the linear method.

Acknowledgements

This study received funding from the Åforsk Foundation under grant agreement No 24-350, the support of which is gratefully acknowledged.

References

Long J., Shelhamer E., Darrell T., 2015, Fully convolutional networks for semantic segmentation. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Mikołajczyk A., Grochowski M., 2018, Data augmentation for improving deep learning in image classification problem, International Interdisci plinary PhD Workshop (IIPhDW), S´winous´cie, Poland, 117-122. Mumuni A., Mumuni F., 2022, Data augmentation: A comprehensive survey of modern approaches. Array, 16, 100258. Alomar K., Aysel H.I., Cai X., 2023, Data Augmentation in Classification and Segmentation: A Survey and New Strategies, Journal of Imaging, 9, 9020046. Holzapfel G.A., Linka K., Sherifova S., Cyron C.J., 2021, Predictive constitutive modelling of arteries by deep learning, J. R. Soc. Interface, 18, 20210411.