Physics-informed machine learning model for peak stress prediction generated from cylindrical charges in concrete

Libin Wang, Bingwen Qian, Zhun Bai, Yutao Hu, Kexin Di, Gang Zhou, Haoqing Ding,
Physics-informed machine learning model for peak stress prediction generated from cylindrical charges in concrete,
Engineering Structures,
Volume 344,
2025,
121361,
ISSN 0141-0296,
https://doi.org/10.1016/j.engstruct.2025.121361.
(https://www.sciencedirect.com/science/article/pii/S0141029625017523)
Abstract: With the widespread application of explosive technologies in civilian and military fields, predicting the distribution of stress waves generated by cylindrical charges is critically important for both understanding damage mechanisms and designing protective structures. To investigate the peak wave systematically, a set of experimental studies was conducted. Subsequently, a numerical model was developed to further investigate the effects of aspect ratio and buried depth of cylindrical charges on stress waves. Specifically, a series of test points were arranged at 10° interval to analyze the peak stress from cylindrical charges with various aspect ratios ranging from 1 to 5 and dimensionless buried depths ranging from 0 to 10.83. Based on the analysis presented in this paper, the radial peak stress generated by cylindrical charges exhibited highly irregular patterns. Finally, a physics-informed machine learning model was employed to achieve the prediction of peak stress rapidly. By introducing a physical loss term in the loss function, the proportion of average relative error less than 20 % in the testing domain achieved 97.68 %. The application of the machine learning model not only predicts the propagation of stress wave but also guides the mechanism research.
Keywords: Cylindrical charges; Peak stress; Physics-informed machine learning; Buried depth; Aspect ratio