Machine-learning-enhanced femtosecond-laser machining: Towards an efficient and deterministic process control

Abstract

Femtosecond laser nanomachining represents a frontier in precision manufacturing, excelling in micro-and nanopatterning across diverse materials. However, its wider adoption is hindered by unintended surface damage or modifications stemming from complex non-linear laser-material interactions. Moreover, traditional effective process optimisation effort to mitigate these issues typically necessitate extensive and time-consuming trial-and-error testing. In this scenario, machine learning (ML) has emerged as a powerful solution to address these challenges. This paper provides an overview of ML’s contributions to making femtosecond laser machining a more deterministic and efficient technique. Leveraging data from laser parameters and both in-situ and ex-situ imaging of processing outcomes, ML techniques—spanning supervised learning, unsupervised learning, and reinforcement learning—can significantly enhance process monitoring, process modeling and prediction, parameter optimisation, and autonomous beam path planning. These developments propel femtosecond laser towards an essential tool for micro-and nanomanufacturing, enabling precise control over machining outcomes and deepening our understanding of the laser machining process.