The Raman scattering effect has attracted extensive attention from researchers since the publication of a paper titled “A New Type of Secondary Radiation” in Nature.1 The Raman distributed optical fiber sensing system uses the optical Raman scattering effect with temperature-sensitive characteristics to detect distributed temperature. Consequently, it is widely used in temperature safety testing of oil and gas wells, transportation infrastructures, and large linear infrastructures.2

Sensing spatial resolution is critical for high-accuracy positioning in Raman distributed optical fiber sensing systems. The existing system is based on the localization principle of optical time-domain reflection, with sensing spatial resolution that is mainly determined by the pulse width.3 In addition, the spatial resolution of a system with kilometer-scale distances is limited to a few meters or tens of meters because of fiber mode dispersion.4 In the field of pipeline leakage monitoring, the system can be used to monitor microleakage problems in pipelines carrying gas or liquids. When a long-distance pipeline of several kilometers is in a microleakage state, the length of the sensing fiber affected by the leakage area is often insufficient to the existing spatial resolution of meters, which can lead to difficulty in identifying the local temperature changes generated by tiny leakage points. Therefore, further improving the sensing spatial resolution is important.

To improve the spatial resolution of system sensing, researchers have proposed four schemes: light-source pulse-width compression, special-fiber sensing, light-source coding modulation, and probe signal reconstruction.

The light-source pulse-width compression method (narrow pulse width demodulation technique) uses picosecond or femtosecond lasers as the probe signal and improves spatial resolution by compressing the pulse width, thereby achieving a spatial resolution superior to that of the traditional meter scale. For example, Ososkov et al.5 proposed a sensing scheme based on mode-locked pulsed lasers, and it achieved a sensing spatial resolution of 0.01 m using 180-fs optical pulses. However, direct pulse-width compression reduces the incident optical flux and deteriorates the signal-to-noise ratio (SNR). Consequently, the proposed scheme achieved an effective sensing distance of 3.0 m.

The light-source coded modulation method increases the light flux coupled to the sensing fiber by coding the detected light source6^–8 while simultaneously combining the single-mode fiber transmission to avoid the effect of pulse broadening on sensing spatial resolution deterioration. Sun et al.7 proposed a genetically optimized aperiodic pulsed coding scheme. The results indicated that a sensing spatial resolution of 1.0 m was achieved at a distance of 39.0 km.

The special-fiber sensing method suppresses the intermodal dispersion effect of optical fibers through special fibers, thereby optimizing the sensing spatial resolution. Wu et al.9 achieved a sensing spatial resolution of 1.0 m at a distance of 24.0 km using a low-water peak special fiber. Liu et al.10 achieved a sensing spatial resolution of 1.1 m at a distance of 25.0 km using a few-mode special fiber with a large effective area and a low dispersion gradient refractive index.

The probe signal reconstruction method avoids the fiber scattering signal superposition effect by reconstructing the Raman scattering signal, thereby optimizing sensing spatial resolution. These include the Raman signal slope reconstruction and demodulation method,11 the temperature-variable region demodulated signal reconstruction method,12 Wiener’s inverse-convolution signal reconstruction method,13 and the scattering signal two-dimensional image processing method.14 Zhu et al.15 achieved a spatial resolution of 0.25 m over a distance of 2.1 km using the Wiener inverse-convolution method.

In summary, although the spatial resolution of the existing system can reach up to 0.01 m, its spatial resolution of sensing still cannot break through 1.0 m when the sensing distance exceeds 5.0 km.16 Compared with traditional pulsed light sources, chaotic laser signals can improve the detection and transmission performance of optical sensing and optical communication systems.17^–20 For example, in 2023, we achieved a 0.10-m spatial resolution and a 1.4-km sensing distance using chaotic pulse signals as probe signals.21 Compared with the traditional method, this scheme can break the limitation of the pulse width of the light source on the spatial resolution and make it break through the spatial resolution from meter level to submeter level at the kilometer-level sensing distance. However, due to the chaotic single pulse, its coupled luminous flux incident into the sensing fiber is limited, resulting in a low SNR of the scattering signal, which does not allow for sensing over a longer distance. To solve the problems of low SNR, this study uses chaotic pulse clusters as probe signals; it can improve the coupled optical flux of the system and thus enhance its SNR. In addition, the information entropy of chaotic pulse clusters is stronger than that of chaotic single pulse signals, which can effectively enhance the correlation between the detected signals and the chaotic Raman scattering signals to suppress the correlation between the noise and the probe signals. In addition, the multi-order time-domain differential reconstruction technique is used in this study to replace the single-order time-domain differential reconstruction technique in the previous work,21 which can further improve the SNR of the system. In the experiment, a spatial resolution of 0.3 m was achieved at a sensing distance of nearly 6.0 km, resolving and quantifying temperature variations up to 1.0°C. Most importantly, the spatial resolution of sensing is independent of the sensing distance.

We built a Raman distributed optical fiber sensing experimental setup based on chaotic pulse cluster demodulation, as shown in Fig. 1.

First, we generate chaotic continuous lasing through a distributed feed-back laser (DFB, NEWPORT, LDM-4980) utilizing a single-feedback structure to generate stable chaotic lasers. To maintain the chaotic state of the laser, an isolator (ISO, OPEAK-OM-ISO) is implemented to prevent subsequent optical paths back to the laser from influencing its chaotic state. Then, it generates chaotic pulse cluster signals using a digital delayed pulse sequence generator (ASG, CIQTEK-ASG8100) to modulate the semiconductor optical modulator (SOA, OPEAK-OAM-SOA-PL-15-15-S) externally. The generation of a chaotic pulse cluster laser contains four pulses used as a probe signal, with a total pulse width of 410 ns. In addition, the signal is amplified by a pulsed erbium-doped fiber amplifier (EDFA, OPEAK, EDFA-C-PL-MB-100-S), which is then divided into the probe signal (99%) and the reference signal (1%) by a 1:99 optical coupler (OC, OPEAK- SSWC-P-1*2-1550). The reference signal is acquired by an oscilloscope (OSC, Teledyne LeCroy WavePro HD) after a variable optical attenuator (VOA, CONQUER, MVOA-1550-900-1-FA) and photoelectric signal conversion using an avalanche photodetector (APD, KEYANG PHOTONICS, KY-DTS-200M-MM). Furthermore, the probe signal is incident on the sensing fiber through a wavelength-division multiplexer (WDM, OPEAK, WDM-1*3-1550). The experimental devices are conventional products on the market; this hardware can assemble equipment used in the field. The fiber under test (FUT) is placed at the end of the sensing fiber, and the FUT is set for temperature change using a thermostatically heated platform. Finally, the temperature-sensitive chaotic Raman anti-Stokes scattering signal is captured using the 1450 nm port of the WDM, which is collected by an APD, and the distributed temperature information is demodulated in the OSC.

To investigate the signal characteristics of the experimentally generated chaotic pulse clusters, the spectra and time series of the chaotic pulse clusters were captured in the experiment, as shown in Fig. 2. Figures 2(a) and 2(b) show the power and optical spectra, respectively, of the chaotic signal with a 3 dB power spectral bandwidth of 4.2 GHz and the red lines are optical spectral 3 dB linewidth of 0.0477 nm. Figures 2(c) and 2(d) show the time series of chaotic pulse cluster signals and a zoom of the chaotic pulse cluster signal time series acquired by DC coupling, respectively, which are in a noise-like random oscillation state. As shown in Fig. 2(d), the chaotic pulse cluster signal average value is 445.8 mV, and the amplitude deviation is 247.4 mV. In principle, the time series of a chaotic pulse cluster signal contains more information regarding the oscillation sequence than conventional chaotic single-pulse signals. Therefore, the correlation trace between the reference signal and the reconstructed chaotic Raman anti-Stokes scattering signal can be further enhanced. This tends to increase the intensity of the chaotic correlation peaks formed at the FUT position to be multiplied, further improving the SNR (or sensing distance) performance. Compared with the chaotic single-pulse Raman distributed fiber optic sensing system proposed, the chaotic pulse-cluster temperature sensing system in this paper has an improved SNR of 12 dB.

The chaotic pulse cluster signals were subsequently injected into the sensing fiber as probe signals, and distributed temperature sensing was performed. In the experiments, an FUT of length 30 cm was set at the tail end of the sensing fiber of nearly 6.0 km. The temperature of the FUT was set to 70°C, and the remaining sensing fibers were maintained at ambient temperature (25°C). The sampling rate in the experiment was 2.5GSa/s, and the repetition frequency was 5 kHz. As the Raman scattering signal is ∼80dB lower than the incident signal, it has a power on the order of nanowatts. Therefore, to prevent the effective signal from drowning in the noise in our experiments, we denoised the chaotic Raman inverse Stokes scattering signal with a cumulative average of 10⁴ times during the acquisition and further denoised the data using a wavelet denoising technique (three-layer db3 wavelet) during the data processing.

This scheme is based on the chaotic cross-correlation method, which is used to demodulate the FUT temperature signal along the sensing fiber. It considers the chaotic pulse cluster probe signal incident on the sensing fiber as a reference signal. The differential reconstruction analysis of the chaotic Raman backscattering signal strips out the chaotic light intensity superposition information of each data point. Therefore, each data point in the reconstructed signal only contains the temperature information of one position of the sensing optical fiber, which breaks through the limitation of the spatial resolution of the system due to the pulse width in principle. The reconstructed signal has a good correlation with the chaotic reference signal. When the reference signal and the chaotic Raman inverse Stokes signal after differential reconstruction are cross-correlated, a positive correlation peak is excited at the beginning of the FUT. A negative correlation peak is excited at the end position of the FUT, and the position and length of the FUT can be accurately obtained according to the delay difference between these two correlation peaks.22 In addition, the chaotic correlation peaks are linearly modulated by the temperature of the FUT, so the temperature information of the FUT region can be demodulated based on the chaotic positive correlation peaks.23 Using these techniques, the processing time added to the data analyses is less than 5 min.

The Raman anti-Stokes scattering signal based on the chaotic pulse cluster modulation is acquired, as shown in Fig. 3. Figure 3(a) shows the chaotic Raman anti-Stokes scattering signal of the entire sensing fiber. The inset is the Raman anti-Stokes scattering signal in the FUT region (where the temperature is 90°C). Figure 3(b) shows the cross-correlation results for 30-cm and 60-cm FUTs heated at 90°C. Figure 3(b) shows the correlation results in the FUT position; its positive and negative amplitudes show the correlation of the chaotic Raman anti-Stokes scattering signal after differential reconstruction at the FUT location and the chaotic reference signal. In Fig. 3(b), the FUT correlation peak distance undergoes broadening, the broadening length of which is related to the rising and falling edges of the pulse cluster. In the proposed scheme, after experimental verification, the broadening length of the chaotic correlation peak is 2.5 m. L1 and L2 in Fig. 3(b) include the length of the chaotic correlation peak broadening (constant value) and the length of the 30-cm/60-cm FUT; L1 is shown with a length of 2.8 m, whereas L2 is shown with a length of 3.2 m.

Figure 4 shows the reconstruction results in the same coordinate axis. It subtracts the chaotic correlation peak broadening length to make the chaotic correlation peak distance correspond to the actual FUT distance; it can be easy to observe the experimental results. In addition, positive and negative chaotic correlation peaks are formed at the start and end points of the FUT, and the positions of the chaotic correlation peaks can be used to localize and identify the FUT position and length. The experimental results show that it can accurately locate the position of the 30-cm and 60-cm FUTs. The spatial resolution of the chaotic pulse-cluster scheme in this paper is defined as the minimum length of temperature change that can be identified, and we can achieve spatial localization and temperature detection in the region of 30 cm of sudden temperature change for any system at any position over any transmission distance in the range of 6 km. Furthermore, the chaotic positive correlation peaks were used to extract temperature information in the FUT. The signal within this length range was compressed in the time domain to the FUT start position. Therefore, the peak of the positive chaotic correlation packs the actual temperature information of the FUT.

To further validate the temperature demodulation performance, experiments on the temperature measurement accuracy were conducted. In the experiments, the temperature of the FUT was set to 70°C, 80°C, 90°C, and 100°C. The remainder of the sensing fiber was maintained at room temperature (25°C). Chaotic correlation peak curves were obtained at each of the abovementioned temperatures during the experiments, as shown in Fig. 5(a). The red dots in Fig. 5(b) are the peaking coefficients of the chaotic positive correlation peak under the calibration temperature condition. The error bars represent the temperature accuracy; it is calculated from the difference between the demodulation temperature and the calibration temperature. The temperature measurement errors are 1.0°C.

This scheme is based on the chaotic cross-correlation method for high-precision localization in the FUT region. Based on the δ-function property of the chaotic laser, the sensing spatial resolution of this scheme is related to the full width at half-maximum (FWHM) of the autocorrelation δ-function of the chaotic pulse-cluster reference signal, as shown in Fig. 6, which shows a linear relationship.

To further explore the limiting FWHM factors, we investigated the relationship between pulse width and sensing spatial resolution in the chaotic pulse cluster demodulation scheme. In the experiments, we used two photodetectors to collect the chaotic pulse cluster signals required to perform the autocorrelation function operation and obtain a bar chart of the difference error between the actual FWHM and the theoretical FWHM for each bandwidth. Of these, 200 MAPD is from KEYANG PHOTONICS, KY-DTS-200M-MM model, and 10 GPD is from KEYANG PHOTONICS, KY-PRM-10G-I-FA model. Figure 7 shows the experimental results of this study. With the same bandwidth photodetector, by changing the pulse width of the chaotic pulse cluster signal, the FWHM of the main peak of its chaotic autocorrelation function remained almost unchanged; the sensing spatial resolution in the proposed scheme was independent of the pulse width. This is because chaotic signals can be stripped of the chaotic time series oscillation information carried in each Raman anti-Stokes scattering acquisition signal point using time-domain differential demodulation. Therefore, each data point contains only the light intensity position information at the sampling rate length, instead of the traditional optical time-domain localization principle, which is the superposition of the light intensity information of all position points within the pulse width. Thus, the proposed method can overcome the limitation of the pulse width on the spatial resolution of the system sensing. In addition, the detector bandwidth determines the sensing spatial resolution. The commercially available APD bandwidth for Raman anti-Stokes scattering signal acquisition is ∼200MHz. Therefore, the sensing spatial resolution of the proposed scheme is limited to 30 cm.

In summary, this study presents a high-performance Raman distributed optical fiber sensing scheme based on chaotic pulse cluster demodulation, which improves the SNR of a conventional system and achieves a high sensing spatial resolution at long sensing distances. In addition, the spontaneous scattering threshold limiting the system-coupled optical flux can be solved, resulting in a high SNR that can achieve longer-distance sensing. Next, we will improve the performance of the system using low-loss sensing fibers (reducing transmission loss) or by further studying the multiple modulation of chaotic pulses (reducing the nonlinear stimulated threshold of chaotic Raman scattering). It will be used to further optimize the sensing distance and improve the SNR of the system. The principle can be assembled into a prototype both theoretically and technically, which is one of our plans for the future.

Jian Li is an associate professor of the School of Electronic Information and Optical Engineering of Taiyuan University of Technology, a young talent of the China Association for Science and Technology, IEEE member, OSA member, and is committed to the research of new Raman distributed optical fiber sensing technology.

Zijia Cheng is a postgraduate student at the Key Laboratory of Advanced Transducers and Intelligent Control System of Ministry of Education, Taiyuan University of Technology. Her current research interests include chaos Raman distributed fiber sensors.

Bowen Fan is a PhD candidate with the Key Laboratory of Advanced Transducers and Intelligent Control System of Ministry of Education, Taiyuan University of Technology. His current research interests include chaos Raman distributed fiber sensors.

Xin Huang is a postgraduate student at the Key Laboratory of Advanced Transducers and Intelligent Control System of Ministry of Education, Taiyuan University of Technology. His current research interests include chaos Raman distributed fiber sensors.

Mingjiang Zhang is a professor at the Taiyuan University of Technology, Taiyuan, China. He has co-authored more than 100 journal and conference papers. His research interests include nonlinear dynamics of laser diodes, photonic-integrated broadband chaotic semiconductor lasers, optical fiber sensing, and microwave photonics. He is a member of the Optical Society of America, a senior member of the Chinese Optical Society, and a member of the Council of the Chinese Instrument Society.