Graded density impactor design via machine learning and numerical simulation: Achieve controllable stress and strain rate
Yahui Huang, Ruizhi Zhang, Shanpeng Liu, Jian Peng, Yong Liu, Han Chen, Jian Zhang, Guoqiang Luo, Qiang Shen
Abstract
The graded density impactor (GDI) dynamic loading technique is crucial for acquiring the dynamic physical property parameters of materials used in weapons. The accuracy and timeliness of GDI structural design are key to achieving controllable stress-strain rate loading. In this study, we have, for the first time, combined one-dimensional fluid computational software with machine learning methods. We first elucidated the mechanisms by which GDI structures control stress and strain rates. Subsequently, we constructed a machine learning model to create a structure-property response surface. The results show that altering the loading velocity and interlayer thickness has a pronounced regulatory effect on stress and strain rates. In contrast, the impedance distribution index and target thickness have less significant effects on stress regulation, although there is a matching relationship between target thickness and interlayer thickness. Compared with traditional design methods, the machine learning approach offers a 10 4 –10 5 times increase in efficiency and the potential to achieve a global optimum, holding promise for guiding the design of GDI.