Acta Optica Sinica, Volume. 45, Issue 13, 1306024(2025)

Artificial Intelligence-Assisted Design Methods of Multi-Core Fiber (Invited)

Xiaoze Tang1, Cong Xu1、*, Jiajing Tu1、**, and Zhaohui Li2,3、***
Author Affiliations
  • 1College of Information Science and Technology, Jinan University, Guangzhou 510632, Guangdong , China
  • 2State Key Laboratory Optoelectronic Materials and Technologies, Sun Yat-Sen University, Guangzhou 510275, Guangdong , China
  • 3Southern Marine Science and Engineering Guangdong Laboratory (Zhuhai), Zhuhai 519000, Guangdong , China
  • show less
    Figures & Tables(13)
    Fiber structure and data acquisition. (a) Fiber cross-section and refractive index profile of the core; (b) data-collection region; (c) traditional design process
    Schematic diagram of segmented analysis of the transmission process of randomly-coupled multi-core fibers
    Flow chart of the proposed algorithm-assisted NN forward design method
    Process of PSO-RF
    Pseudocode of the SSA-RF algorithm
    Model prediction effect. (a) Confusion matrix of model NN 1 test for Λ; (b) correlation graphs between actual data and predicted data to evaluate the design accuracy of SMD; (c) comparison between actual and predicted fiber performance
    Schematic diagram of 3-mode 19-core fiber with the circularly symmetrical arrangement. (a) Circular arrangement structure of a 19-core fiber; (b) refractive index profile of a trench-assisted graded-index core in the 3-mode 19-core fiber
    Relationships between structure parameters and κ. (a) Relationship between r1 and κ; (b) relationship between r2 and κ; (c) relationship between w1 and κ; (d) relationship between Δ1 and κ; (e) relationship between α and κ; (f) relationship between Λ and κ
    Relationships between structure parameters and both DMGD and Aeff. (a) Relationship between r1 and both DMGD and Aeff ; (b) relationship between r2 and both DMGD and Aeff; (c) relationship between w1 and both DMGD and Aeff; (d) relationship between Δ1 and both DMGD and Aeff; (e) relationship between α and both DMGD and Aeff; (f) relationship between Λ and both DMGD and Aeff
    Inverse design process of few-mode multi-core fiber. (a) Flowchart of two machine learning algorithms combining NN 2 with the PSO algorithm and NN 3; (b) trained NN 2 structure; (c) trained NN 3 structure
    Model prediction effect. (a) Pearson correlation coefficient comparison between NN 2 and NN 3+PSO algorithms; (b) bending loss calculation results of structural parameters obtained by NN 2 and NN 3+PSO algorithms at a wavelength of 1550 nm and R of 140 mm
    • Table 1. Comparison of time consumption between simulation and model

      View table

      Table 1. Comparison of time consumption between simulation and model

      ModelL /ΔL (km/m)Time taken
      Transfer matrix multiplications1/15.4 s
      1/0.0011.5 h
      10/0.00115 h
      100/0.001150 h
      Ours1/0.001 (training)120 s
      1/0.001 (testing)5 ms
    • Table 2. Selection range of structural parameters

      View table

      Table 2. Selection range of structural parameters

      ParameterRangeStepChoices
      r1 /μm6‒100.141
      r2 /μm1‒40.131
      w1 /μm2‒60.141
      Δ1 /%0.1‒10.110
      α1‒60.151
      OCT /μm25‒45121
      Λ /μm25‒50126
    Tools

    Get Citation

    Copy Citation Text

    Xiaoze Tang, Cong Xu, Jiajing Tu, Zhaohui Li. Artificial Intelligence-Assisted Design Methods of Multi-Core Fiber (Invited)[J]. Acta Optica Sinica, 2025, 45(13): 1306024

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Fiber Optics and Optical Communications

    Received: Apr. 10, 2025

    Accepted: Jun. 5, 2025

    Published Online: Jul. 21, 2025

    The Author Email: Cong Xu (yanxc@stu.jnu.edu.cn), Jiajing Tu (tujiajing@jnu.edu.cn), Zhaohui Li (lzhh88@mail.sysu.edu.cn)

    DOI:10.3788/AOS250887

    CSTR:32393.14.AOS250887

    Topics