Comparison of machine learning and two-temperature models for prediction of ultrafast laser ablation

Predicting the outcome of ultrafast laser ablation based on laser and system parameters, or narrowing the experimental parameters for optimization, can significantly reduce development time and enhance the efficiency of new laser applications. While the Two-Temperature Model (TTM) traditionally fits data with few adjustable parameters, it faces limitations for high ablation depth, high speed processing, non metallic materials. Machine learning (ML) techniques can predict diverse ablation results and optimize process strategies without prior knowledge of the physics involved. However, ML requires substantial data for effective training, which is often costly and time-consuming to acquire. This study uses two small datasets to compare ML and TTM, evaluating their predictive power, generalization capabilities, and the impact of dataset size. We aim to assess whether ML can serve as a viable, more adaptable alternative to traditional models.