Extraction of Temperature Distributions in Brillouin Optical Time Domain Analysis Sensors Using 2D Wiener Filter Based Matched Filter Detection

Abstract

In this paper, the use of 2D Wiener filter based matched filter detection (WMFD) is proposed and demonstrated for the extraction of temperature distributions in Brillouin optical time domain analysis (BOTDA) sensors. The experimental Brillouin gain spectra (BGSs) are obtained along a 38.2 km sensing fiber by adopting ten different numbers of BOTDA-trace averaging (NTA). These BGSs are first denoised by applying Wiener filter (WF) to enhance the signal-to-noise ratio (SNR) of the BOTDA-traces. The improvement of trace-SNR for using WF is quantified and analyzed experimentally. The matched filter detection (MFD) which is free from time-consuming iterative optimization procedure is then applied to the denoised BGSs for the ultrafast extraction of temperature distributions along the fiber. The measurement uncertainty, spatial resolution and temperature extraction speed provided by WMFD are also analyzed in detail and compared with that provided by widely-used curve fitting method (CFM). The results show that WF can improve the trace-SNR within the range from ~6.78 dB to ~8.28 dB depending on NTA. Consequently, WMFD can improve the measurement uncertainty within the range from ~48.70% to ~59.41% without sacrificing the spatial resolution as compared to CFM. Moreover, the speed in extracting temperature distributions from the experimental BGSs acquired with different NTA for using WMFD can be improved within the range from ~47.36 times to ~50.69 times as compared to that for using CFM. Thus, the proposed WMFD can be an effective approach for highly accurate and ultrafast extraction of temperature distributions along the sensing fiber. DUJASE Vol. 6 (2) 30-38, 2021 (July)

Total variation denoising and Euclidean distance based distributed temperature monitoring in Brillouin optical time-domain analysis sensors

“…The results in Fig. 4 indicate that the SNRs of experimental noisy BGSs vary along the 38.2 km OFC, notably mainly due to fiber attenuation (Azad, 2022). However, the use of TVD on the noisy BGSs manifests a significant improvement in SNR, especially at the end of OFC, where the SNRs of noisy BGSs are the lowest.…”

“…( 4), the peak gain of the ideal BGSs is normalized to be gB = 1. To monitor temperature within the range from 0 ºC to 100 ºC using the BOTDA sensor, υB (BFS) of each BGS in the database has been varied within the range from 10.834 GHz to 10.932 GHz (with the BFS spacing of 0.2 MHz) depending on the characteristics (i.e., intercept of ~10.83415 GHz and slope of ~0.97497 MHz/ºC) of the OFC (Azad et al, 2017;Azad, 2022). The linewidth (ΔυB) to be used in Eq.…”

“…Next, we have evaluated the suitability of using TEPR and the constructed database for monitoring temperature distributions from the local BGSs along the OFC collected by adopting different TN. The process of such a monitoring system is depicted in The denoising algorithm operation relies on eliminating the signals' high-frequency components (Azad, 2022). For such algorithm in BOTDA sensors, the BGSs along the OFC can be oversmoothed which can also deteriorate the experimental spatial resolution.…”

“…Other than employing the complex, costly and modified BOTDA setup, a viable alternative is to collect much noisy BGSs along the OFC, adopting a small number of trace averaging (i.e., smaller TN) with conventional BOTDA setup and then using an effective and time-efficient denoising algorithm for denoising such noisy BGSs. To explore this, several signal denoising techniques, such as wavelet transform (Soto et al, 2016;Azad, 2020), Wiener filter (Azad, 2022), non-local means filter (Qian et al, 2017;Soto et al, 2017) and anisotropic diffusion filter (Luo et al, 2020;Zhang et al, 2022) have been proposed and demonstrated to enhance the SNR of the BGSs along the OFC. The study Liu et al(2022) shows significant improvement in measurement uncertainty by using one-dimensional denoising convolutional neural network (1D-DnCNN) for a 6 km OFC in Raman scattering-based distributed optical fiber sensor.…”

Robust Signal Denoising Techniques for Highly Accurate and Spatial Resolution Preserved Brillouin Optical Time Domain Analysis Sensors

“…These noisy BGSs are then denoised by using suitable signal denoising techniques. To do so, different BGSs denoising techniques such as wavelet transform [18 -20], adaptive Wiener filtering [21], anisotropic diffusion [22,23], block matching algorithm [24], and non-local means filtering [18,25] are reported recently in the literature. However, the consequences of using these denoising techniques on the SNRs of the BGSs, uncertainty in temperature extraction, spatial resolution of the sensors and time required to denoise the BGSs are not investigated meticulously with sufficient experimental results and analysis.…”