上海科技大学知识管理系统

Deep learning, which is a branch of machine learning, has gained increasing attention in the field of optical measurement owing to its excellent recognition ability. In this study, we applied a deep learning method to measure the phase error from the far-field interference fringe pattern in the two-beam tiled-aperture coherent combination system. We performed both numerical simulations and experiments to verify the validity of this method. The RMS accuracy of the phasing method in the simulation was λ/60, whereas in the experiments, it was λ/30. As an extension, the measurement accuracy of the four-beam simulation reached least λ/40. It satisfies the phase synchronization requirements at 95% combining efficiency in the coherent beam combination system. © 2022 Elsevier B.V.