From Physical Principles to Data Science: A Review of Modeling Methods for Ultrafast Laser Processing

Ultrafast laser processing has become a key technology for micro/nano manufacturing due to its high precision and non-thermal ablation characteristics. However, the complex multi-physics coupling mechanisms pose significant challenges for process optimization. The evolution of modeling approaches for ultrafast laser processing was systematically reviewed. Mechanism-based models（e.g., two-temperature equations, molecular dynamics） offer physical interpretability but face issues of computational complexity and difficulty in cross-scale modeling; data-driven models（e.g., machine learning,deep learning） enhance the prediction efficiency by extracting knowledge from experimental data, yet they rely heavily on data quality and often lack physical consistency. Hybrid models integrate the strengths of both mechanistic and data-driven approaches, balances interpretability and computational efficiency through methods such as physics-constrained machine learning. The study highlights that future research should focus on key directions such as embedding cross-scale physics,developing dynamic process adaptive modeling, and advancing intelligent optimization. These efforts aim to shift ultrafast laser processing from an "empirical trial-and-error" paradigm to a "model-driven" approach, providing theoretical support for precision manufacturing in fields such as aerospace and biomedical engineering.