Research On Signal Processing Methods For Improving Fiber Bragg Grating Sensing Performance

Fiber Bragg grating(FBG)sensors have considerable advantages,such as high sensitivity,high accuracy,immunity to electromagnetic interference,stable chemical properties,compact size,and light weight.FBG sensors are mainly used to monitor physical quantities such as temperature,strain,pressure,and vibration.As the application range of FBG sensors expanding,the requirement for measurement accuracy and stability have also increased.FBG sensor's demodulation has four parts:noise reduction,area segmentation,peak detection and calibration.The performance of the FBG sensor system is improved by studying the signal processing algorithm of each part.Noise sources,noise categories,and traditional noise reduction algorithms are analyzed,and this paper focuses on the impact of pink noise on FBG sensing signals.A noise reduction algorithm that combines complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and Savitzky-Golay(S-G)filter is proposed,which is used to effectively reduce noise and preserve the details of the FBG signal.The performance of the noise reduction algorithm in this paper in the case of FBG wide-peak and FBG narrow-peak with signal-to-noise ratios of 20 d B,25 d B,30 d B,and 35 d B has been simulated and verified,respectively.The maximum 62 pm wavelength drift caused by noise is reduced to 2 pm after noise reduction.FBG quasi-distributed sensing uses wavelength division multiplexing.Therefore,it is necessary to divide the FBG reflection spectrum for accurate demodulation of the FBG sensing system.The integrity of divided FBG reflection spectrum affects the accuracy of peak detection,directly.So,a divided method based on Mexican transformation is proposed in this paper.This method can obtain the longest possible sub-signal length without losing spectral peak information.This paper establishes FBG distortion spectrum model according to the causes of distortion,realizing accurate demodulation of FBG distortion spectrum.The spectral characteristics of asymmetric distortion and broadening distortion of FBG reflection spectrum are analyzed,as well as the advantages and disadvantages of traditional peak detection algorithms.And a resolution enhancement peak detection algorithm with distortion spectrum correction is proposed.The simulation analysis and experimental verification of asymmetric distortion spectrum and broadened distortion spectrum are carried out,and the results show that the goodness of linear fit between temperature and wavelength obtained by the algorithm demodulated in this paper can reach 0.9999,which has achieved high accuracy and stability.In order to make up for the shortcomings of field application,that is difficult to use laboratory measurement calibration methods to modify the calibration parameters.A dynamic calibration method utilizing the online sequential extreme learning machine(OS-ELM)is presented in this paper,which realizes online compensation of calibration parameters and improves the stability of FBG sensing system.Comparison with calibration methods such as polynomial fitting,back propagation(BP)neural network and radial basis function(RBF)neural network,the results indicate the dynamic method not only has better generalization but also has faster learning speed.During the measurement,the dynamic method continuously updates calibration network instead of retraining,which can reduce tedious calculations and improve the predictive speed.