Selection of spectrum model in estimation of Brillouin frequency shift for distributed optical fiber sensor

Abstract

To improve real-time performance in distributed optical fiber sensor, the influence of Brillouin spectrum models on performance in estimation of Brillouin frequency shift (BFS) is systematically investigated. First, the related Brillouin spectrum models, objective functions, optimization algorithms and convergence criterion in estimation of BFS are introduced. The BFS estimation algorithms based on the Lorentzian, Gaussian, pseudo-Voigt and Voigt models are written in MATLAB. The algorithms are used to estimate BFS for numerically generated Brillouin spectra with different values of SNR (Signal to Noise Ratio) and measured Brillouin spectra with different values of pulse width and average time. The results reveal that regardless of the simplified Lorentzian and Gaussian models or more accurate pseudo-Voigt and Voigt model-based algorithms used, the estimated BFS between different algorithms is quite similar. However, the fitting error of the simplified model-based algorithms is larger than that of the more accurate model-based algorithms especially for the cases with high value of SNR. The computational burden of the former is less than that of the latter. The Voigt model-based algorithm requires the most computational effort. If the nonlinear least-square fit method is used, the selection of spectrum model has no big influence on the accuracy in the estimated BFS. Therefore, the algorithm based on simplified Lorentzian and Gaussian models should be selected in BFS estimation. Thus, not only high accuracy in BFS estimation is ensured, but also the computation time can, at best, decrease to one-tenth of the maximum value.