Pulse convergence analysis and pulse information calculation of NOLM fiber mode-locked lasers based on machine learning method

Nowadays, mode-locked fiber laser has been extensively applied in various fields such as communication, biomedical, nuclear physics and material processing. By controlling the parameters of cavity elements, people can realize the output of ultrafast pulse with different states. Utilizing the traditional numerical method, the influence of different cavity parameters on the pulses can be computed accurately, which provides the simulation basis for the fabrication of mode-locked fiber lasers. However, with the development of the practical application demand, the performance requirements of the mode-locked pulse are also increasing. The traditional numerical methods have shown some defects in the calculation speed and analysis ability at this time. To improve the shortcomings of the existing numerical methods, in this work, we propose using the machine learning method to analyze the mode-locked pulse convergence of the nonlinear loop mirror (NOLM) fiber mode-locked laser and calculated its pulse information. Firstly, the mapping between cavity parameters and convergence region is established by the support vector machine (SVM) model, and utilizing it, the influence of different cavity parameters on mode-locking is investigated. Then, we train two artificial neural networks (ANN) for calculating the pulse shape and spectral shape in the mode-locked state and find out the cavity parameters under which. the spectral shape approximates to a parabola based on genetic algorithm and the ability of ANN in rapid computing. Finally, we construct the laser cavity according to the numerical result and obtain the mode-locked pulse with a parabolic shape spectrum. The authors believe that the model presented in this work will show widely application prospects in the numerical calculation of fiber lasers.