Design of high-performance photonic crystal fiber polarization filter by Grey Wolf Optimizer with convolutional neural network

Abstract

In this paper, a D-shaped photonic crystal fiber polarization filter is optimized using a novel machine learning framework. The framework incorporates the Grey Wolf Optimizer (GWO) and convolutional neural network (CNN) to optimize the optical fiber structure parameters. The designed CNN accurately classifies optical fiber modes, while the GWO selects complex fiber structure parameters. The fitness function conducts multi-objective optimization on the confinement loss and resonance wavelength of the polarization filter. The simulation results show the y-direction confinement loss of the optimized polarization filter is 1111.51 dB/cm at a wavelength of 1.55 µm. At a fiber length of 1000 µm, the crosstalk on the communication band reaches 965.3 dB, with a bandwidth exceeding 1000 µm. Numerical analysis demonstrates that the performance of the designed PCF filter is better than that of the original and also have showed the superiority of the proposed framework, which is a promising avenue for designing other photonic devices.