Neural Networks Assisted Reconstruction Wavefront of Laser Radiation Based on Intensity Profile near the Focus of Converging Lens

期刊论文
发表日期：

2025年8月1日

查看全文

N. V. Iakushkin,
A. V. Mitrofanov,
A. V. Vasiliev,
A. V. Larichev,
N. G. Iroshnikov,
D. A. Sidorov-Biryukov

摘要

Abstract

The architecture of the ConvFormer S18 neural network has been shown to enable reconstruction of the wavefront shape of a laser beam based on the intensity distribution near the lens focus. Incorporating a standardization method for sparse data into the wavefront reconstruction procedure reduced the root mean square error by a factor of two. Experiments have demonstrated the ability to use a neural network to reconstruct the wavefront of laser radiation based on the intensity profile near the focus of the collecting lens, even in the presence of multiple aberrations. Errors in determining wavefront based on intensity distribution measured before focus, after focus, and based on the data from both positions relative to experimentally measured values were found to be \(\lambda/13\) , \(\lambda/18\) , and \(\lambda/31\) , respectively. The method’s accuracy is proven by the match between the intensity distribution calculated using the data obtained by the neural network and the measured experimental values.