Machine learning method for calculating mode-locking performance of linear cavity fiber lasers

•Pulse convergence in linear cavity fiber laser is studied by ANN.•The pulse shape is predicted by ANN.•The genetic algorithm reversely calculates the laser parameters.•ANN in this paper has high computational efficiency. As a commonly used pulse light source, fiber mode-locked lasers are widely use...

Full description

Saved in:
Bibliographic Details
Published in:Optics and laser technology 2022-05, Vol.149, p.107883, Article 107883
Main Authors: Ma, Xuexiao, Lin, Jiaqiang, Dai, Chuansheng, Lv, Jialiang, Yao, Peijun, Xu, Lixin, Gu, Chun
Format: Article
Language:English
Subjects:
Artificial neural network
Artificial neural networks
Convergence
Fiber lasers
Fourier series
Genetic algorithm
Genetic algorithms
Laser mode locking
Lasers
Learning theory
Light sources
Machine learning
Nonlinear optics
Parameters
Pulse duration
Pulse shape
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Pulse convergence in linear cavity fiber laser is studied by ANN.•The pulse shape is predicted by ANN.•The genetic algorithm reversely calculates the laser parameters.•ANN in this paper has high computational efficiency. As a commonly used pulse light source, fiber mode-locked lasers are widely used in communication, frequency comb, nonlinear optics and other fields. Although people can accurately predict and analyze the performance of mode-locked pulse with given parameters of fiber laser by existing theories and calculation methods, it is difficult to analyze the mode-locked performance of fiber laser with multi-parameter changes by traditional methods. In this paper, we use machine learning technology to put forward a new view on this problem. Firstly, we use an artificial neural network (ANN) to judge whether a small noise pulse can evolve into a stable mode-locked state in the linear cavity fiber laser. Then, the time domain function of sample pulse is expanded by Fourier series. We use the Fourier coefficients as the network output and train another ANN that can predict the pulse shape quickly and accurately. Finally, we demonstrate how to use the genetic algorithm and the trained network to compute the parameters of the fiber laser when the given pulse width is known. The authors believe that the theoretical ideas and computational models presented in this work have great potential in the dynamics research and fabrication of fiber lasers.
ISSN:0030-3992
1879-2545
DOI:10.1016/j.optlastec.2022.107883