Accurate and Efficient PINN Method for Higher-Order Soliton Compression In Optical Waveguides
Introduction:
We present a higher-order soliton compression model utilizing a Physics-informed Neural Network (PINN) to address the challenges posed by high computational demands and slow simulation speeds in conventional numerical simulations. At the same time, by integrating physical knowledge, we can achieve the desired prediction effect with less training data, thus significantly reducing the training cost compared to purely data-driven neural network methods.
Instruction：
Step1：Selecting trained models
For the fiber optic parameters β2 = -5.23×10-3 ps2m-1, β3 = -4.27×10-5 ps3m-1 and nonlinear Kerr parameter γ = 18.4×10-3 W-1m-1 [1] at 1550 nm, we provide two models that have already been trained as examples for testing.
Model_1 : we set the initial pulse width varying from 1 to 2 ps and peak power varying from 1.9 to 3.8 W, respectively. This yields a variation in soliton number from 2.6 to 7.3.
Model_2 : we set the initial pulse width varying from 0.4 to 0.8 ps and peak power varying from 12 to 24 W, respectively. This yields a variation in soliton number from 2.6 to 7.3.
Step2: Inputing the pulse width and peak power to initiate the pulse
Step3: Predicting
References：
[1] L. Salmela, N. Tsipinakis, A. Foi, C. Billet, J. M. Dudley, and G. Genty, “Predicting ultrafast nonlinear dynamics in fibre optics with a recurrent neural network,” Nat. Mach. Intell., vol. 3, no. 4, pp. 344–354, 2021, doi: 10.1038/s42256-021-00297-z.
Download:
https://web.pkusz.edu.cn/wp-content/uploads/2023/12/PINN-method.zip