Autonomous parameter optimization for femtosecond laser micro-drilling
Keiichi Bamoto, Haruyuki Sakurai, Shuntaro Tani, Yohei Kobayashi
Abstract
There is a strong need for a highly efficient method to find the optimal conditions to achieve a desired result in laser processing, oftentimes from a multidimensional parameter space. In this study, we adopted Bayesian optimization as an efficient statistical optimization method robust to the inherent variations observed in typical laser processing results. Specifically, the intensity and spatial beam profile of a femtosecond laser processing system were optimized according to results obtained from an in situ optical microscope observation. In this way, we show that the optimum set of parameters to achieve a desired shape can be obtained autonomously and more than an order of magnitude faster than with a simple grid-search.