• Acta Photonica Sinica
  • Vol. 54, Issue 1, 0106001 (2025)
Yinggang LIU*, Fei LI, Yubo YUAN, Rui LI..., Rui ZHOU and Xinyi XU|Show fewer author(s)
Author Affiliations
  • Shaanxi Engineering Research Center of Oil and Gas Resource Optical Fiber Detection,Xi'an Shiyou University,Xi'an 710065,China
  • show less
    DOI: 10.3788/gzxb20255401.0106001 Cite this Article
    Yinggang LIU, Fei LI, Yubo YUAN, Rui LI, Rui ZHOU, Xinyi XU. FBG Wavelength Demodulation Method Based on Support Vector Regression Optimized by Sparrow Search Algorithm[J]. Acta Photonica Sinica, 2025, 54(1): 0106001 Copy Citation Text show less
    SSA-SVR algorithm model flow
    Fig. 1. SSA-SVR algorithm model flow
    Diagram of FBG wavelength demodulation system based on tunable F-P filters
    Fig. 2. Diagram of FBG wavelength demodulation system based on tunable F-P filters
    Variation of F-P transmission wavelength driven by sawtooth wave with different frequencies
    Fig. 3. Variation of F-P transmission wavelength driven by sawtooth wave with different frequencies
    FFP-TF surface temperature change
    Fig. 4. FFP-TF surface temperature change
    Variation of F-P transmission wavelength in the range of 5~8 V
    Fig. 5. Variation of F-P transmission wavelength in the range of 5~8 V
    Schematic of SVR model with window sliding update
    Fig. 6. Schematic of SVR model with window sliding update
    Compensation values and absolute errors for different algorithmic models
    Fig. 7. Compensation values and absolute errors for different algorithmic models
    Compensation results of different algorithm models
    Fig. 8. Compensation results of different algorithm models
    Comparison of results of several different demodulation methods
    Fig. 9. Comparison of results of several different demodulation methods
    ModelR2MSE/pmRMSE/pmMAPE
    SSA-SVR0.999 9951.2635.450.002 5%
    PSO-SVR0.999 9824.7865.720.003 5%
    LSSVR0.999 9883.0855.510.002 9%
    KRR0.998 08644.96211.960.010 9%
    Table 1. Evaluation metrics for training results of different algorithmic models(frequency is 2.5 Hz)
    ModelR2MSE/pmRMSE/pmMAPE
    SSA-SVR0.999 9971.2134.770.001 6%
    Train dataPSO-SVR0.999 9863.2256.800.003 1%
    LSSVR0.999 9834.1864.670.003 5%
    KRR0.999 61699.4315.280.016 4%
    SSA-SVR0.999 9950.7727.750.001 3%
    Test dataPSO-SVR0.999 9873.6460.330.003 1%
    LSSVR0.999 9843.9964.670.003 6%
    KRR0.999 623103.6321.860.017 4%
    Table 2. Error analysis of prediction results of four algorithmic models(frequency is 0.5 Hz)
    Yinggang LIU, Fei LI, Yubo YUAN, Rui LI, Rui ZHOU, Xinyi XU. FBG Wavelength Demodulation Method Based on Support Vector Regression Optimized by Sparrow Search Algorithm[J]. Acta Photonica Sinica, 2025, 54(1): 0106001
    Download Citation