Design Of Panda Polarization-Maintaining Few-Mode Fiber Based On Neural Network And Research On Brillouin Characteristics

With the advent of the era of cloud computing and big data,highly centralized data and mobile services have grown rapidly,and the demand for services from the Internet traffic carried by the underlying optical network infrastructure continues to increase,which stimulates the exploration of network capacity expansion.Given that the existing multiplexing technology has reached its scalability limit,space division multiplexing technology has shown great potential.Compared with the single-mode fiber with limited transmission capacity,the few-mode fiber can be used as the space channel of the space division multiplexing network and provide multi-mode transmission.At the same time,panda-type polarization-maintaining fiber has been widely researched and applied because of its good polarization retention characteristics,simple manufacturing process and low loss.Therefore,this paper mainly studies the structural design and Brillouin characteristics of panda polarization-maintaining few-mode fiber with multiple transmission modes and high effective refractive index difference.The main work of this paper is as follows:(1)The research status of polarization maintaining fiber and fiber brillouin characteristics is introduced,and the finite element analysis method and stress-type polarization maintaining fiber characteristics used in this research are emphasized.The difference between spontaneous brillouin scattering and stimulated brillouin scattering is analyzed.It provides a theoretical basis for the design of panda polarization-maintaining few-mode fiber and the study of brillouin characteristics.(2)A panda polarization-maintaining few-mode fiber structure based on an elliptical ring core is designed.The fiber structure has an elliptical ring core and two stress-applying rods.The fiber structure can support 14 polarization modes,which can effectively separate adjacent modes.By quantitatively analyzing the influence of various structural parameters of the optical fiber such as the outer major axis radius,the ratio of the elliptical ring core radius,the ellipticity,the stress-applying rod radius and the distance between the stress-applying rod and the elliptical core,etc.,the minimum effective refractive index between modes is obtained by adjusting structural parameters.The minimum effective index difference between modes is 2.0×10-4 the broadband characteristics,dispersion and bending loss of the fiber are analyzed.(3)An inverse design method based on neural network and polarization-maintaining few-mode fiber is proposed to optimize the parameters of the designed panda elliptical ring core polarization-maintaining few-mode fiber structure.Mainly use the constructed data set to train the neural network,and optimize the neural network model by adjusting the activation function,optimizer,hierarchical structure and learning rate,save the trained model,and compare the actual with data simulation and simulation analysis The accuracy between the data and the predicted data shows that the correlation coefficient between the actual value and the predicted value is higher than 0.98,which proves that the network model can fit the relationship between the fiber structure and its performance well.Finally,the optimized fiber structure parameters are obtained,and the minimum effective refractive index difference of each mode of the fiber is increased to 3.06×10-4,which has a significant performance improvement effect.Finally,the performance of the designed fiber structure is compared to confirm the advantages of the fiber structure.(4)The research on the brillouin characteristics of the few-mode fiber based on neural network is proposed.The difference between the single mode of the few-mode fiber and the Brillouin characteristics of mode superposition is analyzed.And use the designed panda elliptical ring core polarization-maintaining few-mode fiber to carry out actual verification research and analysis.The brillouin performances of each transmission mode,including brillouin frequency shift,brillouin scattering width and brillouin scattering gain peak are simulated and analyzed.At the same time,the brillouin scattering parameters of the adjacent modes on the intermodal spectrum are analyzed,and finally the brillouin scattering parameters of the full mode superposition are obtained.In order to optimize the research process of the brillouin characteristics of the fiber,the neural network and the simulation process of the brillouin characteristics are combined.The result shows that the prediction correlation coefficient of the neural network model is higher than 0.99,which indicates that the network can accurately and efficiently simulate the brillouin scattering characteristics of optical fiber and obtain the brillouin scattering parameters.In summary,this paper designs a panda polarization-maintaining few-mode fiber structure based on an elliptical ring core,and uses neural networks and reverse design to optimize the structural parameters.At the same time,the neural network is used to simulate and verify the brillouin scattering parameters of the panda few-mode fiber.The design methods and performance analysis methods proposed in this paper show new ideas for the simulation design process of polarization-maintaining few-mode fiber and have good generalization.