Process Development of Laser Polishing for Nitinol Surfaces Utilizing Machine Learning

Ultrashort pulse (USP) laser ablation is a common micromachining technology for precise material removal in applications requiring minimal thermal damage. Therefore, it can be used to locally change the wall thickness of a stent and to generate 2.5D-structures on nitinol for medical implants. Despite its advantages, cavities larger than 100 µm often exhibit rough surfaces at the bottom, potentially compromising the mechanical integrity and corrosion resistance. To overcome this limitation, a subsequent laser-based polishing process can be applied using the same laser source with a modified parameter set. By generating a thin melt film through pulse burst processing, the surface can be smoothed effectively. This technique also facilitates the post-processing of 3D-printed Nitinol components, for example to reduce friction. Burst processing opens up another set of parameters such as number of pulses per burst, spacing, and energy ramping beyond conventional ones like feedrate, repetition rate, and fluence. To find a suitable parameter set, the conventional approach is to conduct a DOE which can be very complex and time consuming. To boost the process development a machine learning algorithm (Bayesian optimization) is used to determine optimal process parameters. Furthermore, this allows the automation of the process development by integrating sensors into the setup and generating a direct feedback loop with in-line measurements.