Crystal’s Self-Alignment for High Power Laser Facility Based on Machine Learning

Online alignment of harmonic conversion crystal in high-power laser facilities is a challenging and labor-intensive task. An automated technique for self-alignment of crystals on these facilities is proposed based on machine learning. The crystal alignment beam is sampled using grating diffraction. This method employs a machine learning algorithm running on a Raspberry Pi to automatically locate the reflective spot from the crystal’s back surface and adjust its position to achieve alignment. The proposed scheme comprises two modules: a rectangular spiral spot scanning search method module and an automatic spot aligning method module based on the open-source Machine-Learning Online Optimization Package (M-LOOP) algorithm. M-LOOP employs Bayesian optimization based on Gaussian process probabilistic agent model. The combination of these two modules enables automatic adjustment of the laser spot to align with the reference center, thus achieving crystal alignment. The hardware system comprises a crystal alignment optical setup, motors, a CCD camera and a Raspberry Pi. Multiplexed experiments conducted on the SG-II upgraded laser facility demonstrate that the method can complete automatic search and alignment of the crystal’s reflected spot within approximately 10 minutes. This solution addresses the limitations of traditional approaches that require manual search and adjustment of the crystal’s reflected spot for alignment.