Key points are not available for this paper at this time.
Energetic multi-principal-element alloys (EMPEAs) have garnered considerable attention for their reactive characteristics and compositional flexibility in pyrotechnic applications. However, the complex fragmentation behavior under impact loading presents challenges for optimal design. This study established a machine learning framework to predict the mean particle size (MPS) of EMPEAs under ballistic impact conditions and achieve end-to-end material design. Using ballistic gun experiment data from 110 EMPEA samples, multi-stage feature dimensionality reduction through Pearson correlation analysis and genetic algorithm identified five key predictors from 28 initial material descriptors. Among six evaluated algorithms, the AdaBoost model demonstrated optimal performance ( R 2 = 0.851). Model interpretability analysis revealed that atomic radius mismatch (Δ R ), mixing entropy (Δ S mix ), thermodynamic parameter (Δ TE ), bulk modulus mismatch (Δ K ), and impact velocity ( v ), all exhibit negative correlations with MPS, with Δ R identified as the most critical factor. Symbolic regression confirmed Δ R 's exponential negative relationship with MPS. Ti 15 Zr 50 Ta 35 EMPEA was designed and experimentally validated, achieving a quasi-static overpressure of 0.22 MPa at 1300 m/s impact velocity with size-dependent oxidation behavior in fragments, outperforming most existing energetic structural alloys. This work demonstrates the effectiveness of machine learning in understanding impact fragmentation mechanisms and facilitating the design of high-performance EMPEAs with enhanced energy release characteristics.
Zhang et al. (Fri,) studied this question.