Predictive visualisation of fibre laser machining via deep learning

Fibre laser materials processing is a non-contact manufacturing technique used widely across academia and industry. However, defects such as cracks and striations are generally observed on the surface of the cut material, and hence modelling of the light-matter interaction is of particular interest. Laser machining is a highly non-linear process and is challenging to model via equation-based approaches (e.g. finite element modelling), particularly as the physical origins of many effects are not fully understood [1]. Recently, deep learning has been shown to be capable of modelling femtosecond laser machining [2]. Modelling via deep learning uses a data-driven approach, where the model is created directly from experimental data. Deep learning therefore provides an excellent opportunity for simulating laser machining effects that are not fully understood, and consequently assists in parameter optimisation and even provide novel insights and understanding.