Combining Bayesian optimization and advanced modeling into a workflow for first time right production of complex microstructures using laser ablation

In the realm of ultra-short, pulsed laser processing, the production of complex microstructures with precise properties is a critical challenge. The changing surface structure after each ablation layer makes trial-and-error methods and purely data-driven approaches time-consuming and inefficient. In this work, we present a combination of Bayesian optimization with a threshold fluence model for laser ablation to find optimal process parameters for waist radius and pulse overlap. A detailed model of the ablation process is used to compute the surface structure after each layer and correct the remaining ablation depth on each ablation point. We demonstrate that this approach can be used to efficiently compute a process strategy to generate a microstructure corresponding to an arbitrary greyscale image provided as input. This study highlights the potential of Bayesian optimization in combination with models of the ablation process to realize a true first-time right production for micro structuring processes using ultra-short, pulsed lasers and demonstrates possible trade-off solutions between accuracy and processing time.