Machine Learning-Aided Examination of Energy Absorption and Mechanical Properties in Steel Lattice Structures

Sari A.; Askar S.; Rodriguez-Benites C.; Rajabovich T.Y. · Transactions of the Indian Institute of Metals (2025) · DOI: 10.1007/s12666-024-03507-5

Resumen

This study endeavors to develop an optimized machine learning (ML) model using Bayesian theorem for predicting specific energy absorption (SEA) and mean crushing force (MCF) in polycrystalline-like lattice structures made of different steels. The basis of this study relied on obtaining reliable training data through numerical simulations, including test parameters, material properties, and geometric features, particularly the attributes of the lattice structure. The results were noteworthy, revealing a high level of accuracy in predicting both SEA and MCF, with a slight edge in forecasting energy absorption. Furthermore, the research exposed the substantial impact of output objective values on the contribution function of input features, emphasizing unique interconnections between each objective and its associated input features. Additionally, the elevation in SEA values is accompanied by a corresponding rise in the contribution functions of geometric features. Conversely, the impact of testing parameters slightly surpasses that of other input factors with the increase in MCF value. Ultimately, the findings from the ML model are utilized to offer a thorough discussion, providing a mechanistic viewpoint on why input features affect prediction performance. © The Indian Institute of Metals - IIM 2024.

Instituciones: Universidad César Vallejo