• Acta Optica Sinica
  • Vol. 40, Issue 7, 0706001 (2020)
Hao Jiang1、2, Yougang Wang1、2, Jing Chen1、2、*, and Xinyu Huang1、2
Author Affiliations
  • 1College of Electrical Engineering and Automation, Fuzhou University, Fuzhou, Fujian 350108, China
  • 2Research Institute of Power System & Power Equipment, Fuzhou University, Fuzhou, Fujian 350108, China;
  • show less
    DOI: 10.3788/AOS202040.0706001 Cite this Article Set citation alerts
    Hao Jiang, Yougang Wang, Jing Chen, Xinyu Huang. Wavelength Detection of Overlapping Spectra in FBG Sensor Network Based on Gated Recurrent Unit Network[J]. Acta Optica Sinica, 2020, 40(7): 0706001 Copy Citation Text show less
    Structure of series-parallel WDM fiber Bragg grating sensor network
    Fig. 1. Structure of series-parallel WDM fiber Bragg grating sensor network
    Overlapping types of FBG reflection spectra. (a) Non-overlapping; (b) partly-overlapping; (c) fully-overlapping
    Fig. 2. Overlapping types of FBG reflection spectra. (a) Non-overlapping; (b) partly-overlapping; (c) fully-overlapping
    Wavelength demodulation architecture of GRU model
    Fig. 3. Wavelength demodulation architecture of GRU model
    Unit structure of GRU network
    Fig. 4. Unit structure of GRU network
    Training process of GRU model
    Fig. 5. Training process of GRU model
    Three typical cases of overlapping spectral wavelength detection. (a) Non-overlapping; (b) partly-overlapping; (c) fully-overlapping
    Fig. 6. Three typical cases of overlapping spectral wavelength detection. (a) Non-overlapping; (b) partly-overlapping; (c) fully-overlapping
    Effect of training samples on detection performance
    Fig. 7. Effect of training samples on detection performance
    Effect of batch size on detection performance
    Fig. 8. Effect of batch size on detection performance
    Effect of numbers of GRU hidden units on detection performance
    Fig. 9. Effect of numbers of GRU hidden units on detection performance
    Cumulative probability distribution of RMS for different methods
    Fig. 10. Cumulative probability distribution of RMS for different methods
    CaseλB1 /nmλB2 /nmWavelength interval Δλ /nmRMS /pm
    Non-overlapping1550.59961551.2996-0.70000.2629
    Partly-overlapping1550.89971551.0999-0.20020.4158
    Fully-overlapping1551.20041551.19920.00120.6338
    Table 1. Wavelength detection results of typical overlapping spectra
    MethodsTraining time /sRMS /pmCumulative probability of RMS less than 1 pmCumulative probability of RMS less than 3 pm
    GRU3128.70.83250.8820.949
    LS-SVR17290.01.20630.7490.968
    DE1.91220.2670.887
    PSO2.40370.2640.846
    Table 2. Performance comparison of different methods
    Hao Jiang, Yougang Wang, Jing Chen, Xinyu Huang. Wavelength Detection of Overlapping Spectra in FBG Sensor Network Based on Gated Recurrent Unit Network[J]. Acta Optica Sinica, 2020, 40(7): 0706001
    Download Citation