Open Access
Add to BibSonomy
Abstract
High temperature monitoring is critical to the health and performance of vital pieces of infrastructure such as jet engine, fuel cells, coal gasifiers, and nuclear reactor core. However, it remains a big challenge to realize reliable distributed high temperature sensing system with high speed, high spatial and temperature resolution simultaneously. In this work, a Raman distributed high temperature sensing system with high temperature resolution and high spatial resolution was realized in a single-crystal YAG fiber. The sensing system demonstrated operation from room temperature up to 1400°C with a spatial resolution of 7 cm and response time of 1 millisecond in a 1m long YAG fiber. The average temperature sensitivity of the system is about 7.95 × 10−4/°C. To the best of our knowledge, this is the best spatial resolution and response time reported in literature. In this system, a 2D image restoration was used to boost the signal to noise ratio of sensor. Empowered by the algorithm, the average temperature standard deviation along the sensing fiber of 7.89 °C was obtained based on a single frame data in 1 millisecond. A new record of temperature resolution of 0.62 °C was demonstrated in only 1 second frame data traces, which enables a fast response capacity.
© 2023 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
High temperature sensing plays an important role in the areas of aerospace industry, etc., such as the temperature monitoring of the jet engine [1,2]. High-speed and high-resolution high temperature sensing can provide effective information of combustion efficiency and running healthy condition, which in turn adjust the injection of fuels and provide the feedback of operation security. The conventional methods for high temperature sensing, like sensors based on thermocouple [3], chameleon paint method [4], crystal temperature measurement [5], all have their limitations on large scale deployment, low resolution, and limited temperature range in actual applications. For now, the high temperature sensing areas still lack an effective approach to realize high performance in situ measurement.
Optical fiber-based temperature sensing techniques are well known due to the advantages of electro-magnetic immunity, corrosion resistance, small device footprint, and potential for multi-point and multi-variable sensing [6–8]. However, the optical fiber temperature sensors with the common silica-based optical fiber cannot endure in ultra-high temperature environment beyond 700 $^{\circ }$C, which is very often associated with these applications. Hence, the high temperature sensors based on specialty crystalline optical fibers have been proposed [9–11]. However, due to various material and signal processing limitations, these devices have not seen large-scale commercial deployments. In this work, we realized a distributed high temperature sensing system which is capable of operating from room temperature up to 1400 $^{\circ }$C and offering the best temperature and spatial resolution to the best of our knowledge.
Historically, the first high-temperature optical fiber sensor was based on blackbody radiation [2,12]. Normally, a small blackbody cavity was built as the sensing element. The radiance from the cavity was used to measure its temperature. However, these sensors are not suitable for continuous and wide range temperature measurement due to weak thermal intensity and low signal-to-noise ratio (SNR) at low temperature. A second type of high temperature fiber optic sensor is based on the fluorescence lifetime of dopant ion [13–15]. Ye et al. presented a sapphire fiber thermometer probe with Cr$^{3+}$ ion doped tip made by laser heated pedestal growth method [13], and demonstrated operation from room temperature to 450 $^{\circ }$C. The temperature was derived from the temperature dependent fluorescent decay time of the Cr ion in the probe. For such devices, however, the preparation of fluorescent material and the signal coupling remains challenging, such as the ion doping in the specialty fiber and the collection of fluorescence signals. More recently, the most widely researched category of fiber optic high temperature sensors are based on wavelength-selective devices like fiber Bragg grating (FBG) or Fabry-Perot (F-P) interferometers designed on specialty optical fibers [10,11,16–19]. FBGs can be written on single-crystal sapphire fibers with phase mask method or point-by-point inscription by femtosecond laser pulse. Yang et al. reported the inscription of FBGs in a single-crystal sapphire fiber via a point-by-point method by infrared-femtosecond laser pulses and demonstrated operation of the sensor up to 1400 $^{\circ }$C [20]. Based on the same inscription method, they fabricated a multi-point temperature sensor based on wavelength-multiplexed sapphire fiber Bragg gratings [21]. Xu et al. proposed a new method for inscribing FBGs in sapphire fibers using a femtosecond laser line-by-line scanning technique [22], which can create larger index modulation area resulting in a higher reflectivity of the FBGs. The sapphire FBGs could withstand a high temperature at about 1600 $^{\circ }$C. F-P based sensors in fibers are usually created by devising a cavity, usually by micro-machining, with two or more parallel optical surfaces. Wang et al. firstly reported a sapphire-fiber-based intrinsic F-P interferometer for high temperature sensing by splicing a silica single mode fiber to a length of multimode sapphire fiber [23]. Recently, Wang et al. also reported a high-temperature sensing F-P interferometer constructed based on sapphire fiber and sapphire wafer [11]. Sensors based on FBGs and F-P cavities are sensitive to temperature and strain. Hence, the cross-talk is the inherent weakness that needs to be finely resolved in the signal demodulation process to guarantee the measurement accuracy.
From the aspect of the working styles, except the quasi-distributed FBG sensors, most of the above categories of high temperature sensors can all be classified as single point sensors. These fail to exploit the extended length of the optical fiber as potential sensing points and just be utilized to relay the signal from the point of measurement. This leads to cost issues and difficulties in sensor deployment and signal acquiring and processing toward large-scale and array applications. Compared with single point optical fiber sensors, distributed temperature sensing (DTS) method uses the entire length of the fiber as the sensing unit, in which temperature variations along the fiber can be detected without being limited to pre-assigned discrete points. This is advantageous in sensing task, such as gas turbine due to the dynamic and asymmetric temperature distribution. Besides, DTS based on Raman scattering has no crosstalk from strain and vibration due to the unique dependence of Raman scattering intensity on just temperature. Owing to these features and advantages, Raman based DTS has been used in many applications such as fire alarm systems in tunnels and subways, temperature monitoring and leakage detection in oil pipelines and hot-spot detection in high voltage transmission lines [24].
However, conventional DTS technique with commonly used silica glass optical fiber cannot endure in high temperatures beyond 700 $^{\circ }$C. There is a significant degradation of the optical and mechanical properties of silica glass in the fibers in such harsh environments. As an alternate, single-crystal fibers, typically like derived from sapphire (crystalline aluminum oxide) have shown great potential in such applications. They have excellent mechanical strength, optical transparency, and high melting temperature over 2000 $^{\circ }$C, which makes them great candidates for harsh environment sensing. Nevertheless, the conventional DTS design will not work well with single-crystal fibers as they are no cladding fiber, and there is no single crystal fiber based circulator or coupler that can be used for routing the weak Raman signals from the single crystal fibers. In 2016, Liu et al. reported the first kind of single-crystal sapphire fiber based DTS [25,26]. The sensor was demonstrated up to 1200 $^{\circ }$C with a spatial resolution of 14 cm in which the average standard deviation is about 3.7 $^{\circ }$C over 180 s averaging time. Further, the system was improved to achieve a higher temperature measurement of 1400 $^{\circ }$C [27]. Yttrium aluminum garnet (YAG) fiber, as a kind of single crystal optical fiber, has similar properties with sapphire fiber, like high melting point, chemical resistance and so on., with the added benefit of ease of introduction of various dopants which can engineer the material properties of the crystal fiber significantly [28–30]. A previous work was reported for distributed temperature sensing with Raman scattering in YAG fiber from room temperature to 1000 $^{\circ }$C [30]. However, due to low signal to noise ratio in the device, time averaging was required which in turn impacted the device responsivity significantly.
In this work, a high-performance single-crystal YAG fiber based distributed high temperature sensing (DHTS) system was demonstrated for the first time from room temperature up to 1400 $^{\circ }$C with a 7 cm spatial resolution and 1 millisecond average time in a 1 m long YAG fiber. In the system, a 2D image restoration was utilized to enhance the signal to noise ratio of Raman signals by fully exploiting the high level of similitude and redundancy contained on the acquired consecutive time-domain data traces. The features and performances of 2D denoised method were validated with theoretical simulation. Benefiting from the denoised algorithm, the average temperature standard deviation along the sensing fiber based on 1 ms data trace was improved from 33.86 $^{\circ }$C to 7.89 $^{\circ }$C, giving the distributed sensing system an ability of fast response to the temperature changes. A better temperature resolution of 0.62 $^{\circ }$C could be obtained in only 1 second data traces. This work paves the way of high-speed DHTS on the applications in harsh environment requiring faster data acquisition rates, such as the aerospace industry.
2. System architecture and principles
To realize distributed high temperature sensing above 1000 $^{\circ }$C, the first challenge that needs to be addressed is thermal radiation and fluorescence. Thermal radiation exists in all materials, but increases dramatically as the temperature increases, as given by Wein’s displacement law. Fortunately, it has been found that the peak of thermal radiation is temperature dependent and a shorter wavelength system will experience less thermal radiation interference [27]. On the other hand, fluorescence in single-crystal fiber mainly originates from impurities in the material. In YAG optical fiber, the fluorescence peaks mainly exist in the wavelength of 540 nm $\sim$ 740 nm. Therefore, a simple solution is shifting the central wavelength of pulse laser from traditional near-infrared to visible 532 nm, which leads to significant lower radiation and fluorescence background and ensures minimal interference to the Raman signals. In addition, two aspects of light coupling with the fiber-under-test (FUT) need to be considered. Firstly, the Raman gain of single-crystal YAG fiber is very weak compared with that in the normal silica fiber. Hence, the Raman signal of single-crystal YAG fiber will be distorted and masked in the strong silica Raman signal if the YAG fiber is directly connected with an all silica fiber-based system. Secondly, the core diameter difference between single-crystal YAG fiber and silica fiber would cause strong round-trip coupling loss. Therefore, this distributed high temperature sensing system was built as a free-space system.
Figure 1 shows the proposed high temperature Raman DTS system architecture. A high power, sub-nanosecond pulsed laser at 532 nm was used to excite the Raman scattering in the FUT. Filter 1 was a narrow bandwidth clean-up filter. The cascaded attenuator was used to adjust the laser energy. A series of strong laser pulses with a repetition rate of 1000 Hz and pulse width of 500 ps was launched into the YAG fiber after the 50:50 beam splitter. Another path of the beam splitter was routed into a photodiode (PD) for power monitoring and synchronization of the system. The pulse laser excited the Rayleigh, Brillouin, and Raman photons along the sensing fiber. As the backscattered photons came back by the beam splitter, the two cascaded notch filters each with OD > 6 at 532 nm were used to block the Rayleigh and Brillouin scattered photons. Afterwards, a dichroic mirror was utilized to separate the Raman Stokes and anti-Stokes signals to the detectors. In each path of Raman Stokes and anti-Stokes signals, a corresponding bandpass filter with central wavelength of 540 nm and 514 nm was also used, respectively, marked Filter 2 and Filter 3 in Fig. 1. These filters could allow transmission of most Raman lines in the FUT and block the rest of the spectrum. The attenuator filter of OD 0.5 in the Stokes path was used to reduce the Stokes intensity and avoid exceeding the measurement upper limit of data acquiring system at high temperature. Two identical high bandwidth and sensitivity avalanche photodetectors (APDs) with bandwidth of 1 GHz were used to detect the Raman signals, namely APD1 and APD2. The three lenses in the system were used to adjust the focal length to maximize the coupling efficiency. The electrical signals were acquired and recorded by NI high-speed data acquiring FPGA system with two NI 5785 board cards with sampling rate of 6.4 GSa/s and bandwidth of 500 kHz to 6 GHz.
In the DTS system, each laser pulse excites a series of Raman scattering photons along the FUT. During the interval of each laser pulse, a time resolved Raman Stokes and anti-Stokes were measured in time domain, which is known as Optical Time Domain Reflectometry (OTDR). The Raman intensity of the anti-Stokes and Stokes components is characterized by the Bose-Einstein statistics [31]:
3. Experiments and demodulation results
In the preliminary experiments, in order to obtain a high spatial resolution of system, the pulsed laser (Changchun New Industries Optoelectronics Technology Co., Ltd, MPL-U-532) with full width at half maxima (FWHM) of 500 ps was used. The pulse energy of the laser was 30 $\mu$J. At the same time, two high-sensitivity APDs (Hamamatsu Photonics K.K, C5658) with bandwidth of 1 GHz were utilized to detect the Raman signals. Meanwhile, the high-speed data acquiring FPGA system (National Instruments, PXIe-5785) with sampling rate of 6.4 GSa/s was used. The narrow pulse duration, high bandwidth and sensitivity detectors of APDs and the high-speed data FPGA system. And the high nonlinear threshold of the YAG fiber is also the precondition to support the high peak power due to high pulse energy and the narrow pulse duration. The laser pulse repetition rate is 1000 Hz. The FUT YAG fiber was 1m long and 121 $\mu$m in diameter bought from the MicroMaterials Inc., FL, U.S.A. The YAG fiber fabricated by laser heated pedestal growth (LHPG) method is a no cladding and multimode single crystal fiber, which is usually used in high-power laser research and recently explored in high temperature sensing in harsh environments [32]. In the experiment, the sensing fiber was inserted into a ceramic tube for protection and then put into a tube furnace (Shanghai Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, SG-GL1800). The temperature was elevated from room temperature up to 1400 $^{\circ }$C with an interval of every 150 $^{\circ }$C for 60 minutes.
The time-resolved Raman anti-Stokes and Stokes signals along the sensing fiber are obtained during the heating cycle up to 1400 $^{\circ }$C, as shown in Fig. 2. Each measurement trace at different temperature was averaged based on 2000 raw traces for 2 seconds. The distance along the $x$-axis was calculated from the group velocity of the light in the sensing YAG fiber. The heating temperature profile of the furnace was relatively wide and highest at the central location of the furnace, which resulted in the wide responses of Raman signal Fig. 2. The responses at different temperature up to 1400 $^{\circ }$C can be clearly obtained as shown in Fig. 2, which indicates the distributed high temperature sensing ability of the system. As for the tail at each signal trace, it mainly originates from the Fresnel reflection of forward scattered Raman signals at the tail of the FUT. We also attached a video of dynamic demonstration of this system based on an oscilloscope by two butane flamethrowers in the supplementary materials (see Visualization 1).
Further, based on the acquired Raman signals, the ratio of anti-Stokes and Stokes intensity was calculated at different temperature set both in the temperature-rise and temperature-fall periods by the tube furnace as shown in Fig. 3(a). In Fig. 3(a), the temperature rise and fall period data were marked with blue asterisk marker and black plus sign marker. The red curve was fitting curve by a third-order polynomial equation. The error between the real temperature value and fitting ones mainly results from the temperature inaccuracy of the furnace, the remaining fluorescence and thermal noise. From the Fig. 3(a), the average temperature sensitivity over the whole temperature range is calculated as about $7.95 \times 10^{-4} /{ }^{\circ } \mathrm {C}$. Based on the calibration curve, we can obtain the demodulation results of distributed temperature profiles along the sensing fiber as shown in Fig. 3(b). Before the temperature measurement, the calibration toward the response of system in advance is necessary. Then the temperature measurement can be conducted according to the calibration curves.
Fig. 2. Time resolved Raman anti-Stokes and Stokes signals along the sensing fiber (a) anti-Stokes signal (b) Stokes signal.
Fig. 3. Temperature demodulation results (a) calibration curve between the Raman ratio of anti-Stokes/Stokes and temperature. (b) distributed temperature measurement along the fiber after demodulation.
4. Spatial resolution measurement
Based on acquired Raman signal trace, the spatial resolution can be approximately measured from the tail of signal. An infinitesimal heater is not existed due to thermal diffusion. The reflection of the Raman signals at the end of fiber is equivalent to the response of the system toward an infinitesimal heater. The response bandwidth of peak corresponds to the minimum length of system response, namely the system spatial resolution. Hence, the spatial resolution can be derived from the tail of signal trace. Figure 4(a) gives the spatial resolution of system based on the anti-Stokes signal trace at temperature of 1400 $^{\circ }$C. A 7 cm spatial resolution was obtained. This would be further enhanced with higher bandwidth of detectors used in system.
Furthermore, as shown in the embedded figure of Fig. 4(b), we use a lighter to test the spatial response of system. To be mentioned here, the FUT was capsuled in the ceramic tube to protect it from contamination and damage. As usual, we acquired the anti-Stokes and Stokes signals and made demodulation over the acquired signals. The temperature information after demodulation is shown in Fig. 4(b). From the FWHM of the temperature peak, a 7.4 cm spatial resolution was obtained. It’s worth to mention that this result was obtained based on averaging of 50 frame raw data of signal traces, which corresponds to a 50 ms average time.
Fig. 4. Spatial resolution measurement of YAG fiber-based Raman distributed high temperature sensing system under temperature of 1400 $^{\circ }$C of furnace(a) and under heating of fire lighter(b).
In practical measurement, the fire temperature would change due to the different contact locations of flame core, inner flame, and outer flame on the sensing fiber as the flame in hand moves. Therefore, the denoising process through averaging multiple traces into one would lose detail information and degrade the response time, especially in low signal-to-noise ratio (SNR) scenarios. Figure 5 compares demodulation results based on consecutive single frame signal trace and averaging 50 frame signal traces together. The embedded illustration of Fig. 5 gives the corresponding anti-Stokes signals. Figure 5 shows that single frame data undergoes almost 55 $^{\circ }$C of temperature variation while averaging is an effective way to improve the SNR at the expense of longer response time. Hence, a more effective denoising method without sacrifice of response time is a critical need.
Fig. 5. Comparison of temperature demodulation results under fire lighter based on consecutive single frame signal trace and averaging 50 frame signal traces.
5. System modeling and algorithm optimization
In recent years, many specially designed methods have been proposed and demonstrated for performance enhancement in distributed temperature sensors, including optical pulse coding [33,34], deconvolution [35], Fourier transform [36] and wavelet transform [37], etc. These explorations were proven to be effective in noise removing. In this context, a 2D image restoration, non-local means (NLM) method [38–41] that usually used in image denoising was introduced in the current DTS system. This method relied on analyzing the high-level similitude and redundancy contained on the multiple frame traces of data. In the process, the consecutive raw frame signal trace was treated and constructed as a 2D image. And then, it was processed by NLM method by taking full advantage of correlated patterns of information and their degree of redundancy contained in the measured data. Finally, all the frame signal traces were denoised without discarding any trace. In principle, the NLM method replaces the value of a pixel by an average of a 2D patch which correspond to a set of pixels surrounding the centered pixel. All pixels are processed by sliding the neighborhoods or 2D patches.
However, adapting NLM method in current DHTS system still requires optimization. Firstly, the distributed high temperature sensing system was theoretically simulated according to the actual parameters in the experiment. Then, the NLM algorithm was implemented and the performance was optimized based on the model of the distributed sensing system.
The system modeling mainly refers to the simulation of the Raman intensity backscattered from the sensing fiber. It has close relationship with the configuration of laser, temperature profiles, and properties of the sensing fiber. In the simulation, we used the measured laser pulse profile. The temperature profile was also rebuilt by a Gaussian function with a FHWM of 35 cm, whose peak located at position of 0.62 m in 1 m long YAG fiber. In addition, the Raman gain that is temperature dependent to the scattered signal was calculated at different temperature according to differential cross section of Raman given by Eq. (1) and Eq. (2). The fiber and laser pulse were divided into a series of components at a fine sampling interval of 10 ps. Meanwhile, the transmission loss of sensing fiber was set as 3 dB/m. Based on the above configuration, the return power matrix was calculated to obtain the Raman signal intensity excited by each pulse component at different fiber locations. Here, only the anti-Stokes signal was shown for example. Overall, simulation results of Raman intensity by theoretical modeling well agree with experimental results, as shown in Fig. 6. The tiny difference between them was likely from the temperature variation of furnace. Based on this, the NLM can be validated to find out the influence on the signal and the optimization to the proposed system.
The standard deviation of noise level on actual experimental results was obtained, as shown in Fig. 7(a). The signal traces contained twenty raw frame traces of anti-Stokes signal with a zoom-in view of local part of the noise level. Further, random white noise with the same standard deviation from experiment was added to the twenty simulation frame signal traces of anti-Stokes as shown in Fig. 7(b).
The NLM method relies on three important variables, which are: the radius of similarity window $f$, radius of search window $t$, and degree of smoothing control filtering parameter $h$, respectively. NLM optimization is a searching process of highest SNR results with a particular set (or sets) of these three variables. In the context of distributed fiber sensing, the radius of similarity window $f$ is associated to the spatial resolution of the sensing system. Since the sampling interval of the system is 1.56 cm, and the estimated spatial resolution of this system is about 7 cm, the similarity window is set as $4\times 4$, corresponding to 6.24 cm which is shorter than the spatial resolution. Hence, the NLM algorithm has negligible negative impact on the spatial resolution. In principle, the larger the radius of search window $t$ is, the better the noise can be removed. However, a large size of search window requires more computing resource and processing time, while the enhancement of SNR approaches a steady value. Therefore, the value range of $t$ was set from 10 $\sim$ 15. The degree of smoothing control filtering parameter $h$ is usually set as ten times of the noise standard deviation $\sigma$. However, a much higher $h$ would induce over-smoothing resulting in the loss of details of frame signal traces. For example, the tail of signal trace would be mostly diminished under large value of $h$. Hence, partition NLM processing method was adopted to deal with the experimental data at different parts. Specifically, a different optimization strategy of NLM was employed at the tail of fiber in avoid of large intensity reduction of real backscattered signal due to the usage of NLM. In the test, the $h$ factor was set from 0.1 $\sigma$ to 10 $\sigma$.
Fig. 7. NLM optimization toward the YAG-based distributed high temperature sensing system. (a) experimental anti-Stokes traces for depriving the noise level. (b) simulation frame traces added with random white noise with the same standard deviation with that of experiment. (c) typical denoised signal trace with optimized NLM. (d) SNR enhancements by different NLM parameter combinations.
Based on the parameter configuration, the NLM algorithm was tested to deal with the noised signal traces shown in Fig. 7(b). Figure 7(c) shows the optimized result of a typical single denoised signal trace marked with orange color. Figure 7(d) shows that the $h$ value of $3\sigma$ is efficient to promote the SNR by about 13.5 dB.
From the discussion and test of NLM algorithm toward the noised simulation signal traces, the NLM was then used to deal with the actual experimental signal traces. In this part, the signal traces both at temperature of 1400 $^{\circ }$C and under the heating of fire lighter were tested for verification. Figure 8(a) and Fig. 8(b) shows the processed results of signal trace at the furnace temperature of 1400 $^{\circ }$C and that under fire lighter, respectively. In this NLM denoised processing, the $h$ factor was optimized as 8 $\sigma$ when $f$ is 4 and $t$ is 15. Both simulation and experiment have shown that NLM algorithm is a better denoising method as appose to the simple averaging algorithm.
Fig. 8. The performance results of NLM processing toward the signal trace at the high temperature (a) and under heating of fire lighter (b) of the sensing system.
6. Temperature resolution measurement
The direct advantage of SNR enhancement is the improvement of temperature resolution. The standard deviation of the demodulated temperature was derived from 20 groups of measurements. In the practical experiment and denoised processing, the consecutive raw frame signal trace was measured and constructed as a 2D matrix, which is equivalent to an image. Then, the NLM processing was used for noise processing, which was post-processed in this work. This procedure was continuously repeated for each new measured trace and independently for the Stokes and anti-Stokes signals.
Figure 9 shows the denoised performances of NLM algorithm toward the room temperature signal traces and the standard deviation of temperature. The attenuators of the system were removed in this room temperature test for temperature resolution measurement. Figure 9(a) and Fig. 9(b) shows good performances of typical single frame signal trace of anti-Stokes and Stokes before and after NLM denoised processing. Without any data processing, the temperature standard deviation derived from a single frame data along the sensing fiber was 33.9 $^{\circ }$C due to large noise as shown in Fig. 9(c). After applying the NLM algorithm, the average temperature standard deviation was 7.89 $^{\circ }$C. It indicates that the distributed high temperature sensing system is capable to response at 1 ms time step with temperature resolution of 7.89 $^{\circ }$C at temperature up limit of 1400 $^{\circ }$C. By the way, the response time can be further enhanced by improving the repetition rate of the high energy pulsed laser.
A better strategy is to combine the average algorithm and NLM, namely adopting NLM algorithm after averaging processing. Figure 10 shows the performance comparison between solely average algorithm and the combined strategy. The $x$ axis "Average Frame Quantity", namely different average frame number, represents different average time in consideration that one frame equals one millisecond in this system.
Fig. 9. Denoised performances by NLM algorithm toward the room temperature signal traces and the standard deviation of temperature. Performances of typical single frame signal trace of anti-Stokes (a) and Stokes (b) before and after NLM denoised processing. (c) The temperature resolution derived from a single frame data along the sensing fiber.
Figure 10 shows that the temperature resolution under the $x$ axis ’Average Frame Quantity’ of 100 was improved from 3.35 $^{\circ }$C to 1.08 $^{\circ }$C under this new strategy, which indicates that the combination strategy has a better performance in temperature resolution than that of only average algorithm. Further, as shown in Fig. 11, a better average temperature resolution of 0.62 $^{\circ }$C along the sensing fiber was obtained via adoption of both averaging and NLM processing with 1000 frame signal traces, and the temperature resolution at the location of 0.61 m is 0.60 $^{\circ }$C from the indicator label in the figure. Moreover, we can see that the value of the two data points marked with cross shape is almost equal, which is 3.33 $^{\circ }$C corresponding to the $x$ axis of 6 based on the combined strategy and 3.35 $^{\circ }$C corresponding to the $x$ axis of 100 by only averaging processing. From this, it can be clearly declared that the usage of NLM processing can reduce the number of averages but achieve almost the same temperature resolution with only averaging processing. This outcome has two direct huge benefits. On one hand, this would shorten the acquiring and refresh time, on the other hand, it would keep more frame signal traces than only averaging.
Fig. 11. Demonstration of the whole temperature resolution results along the sensing fiber based on 1 second frame data
7. Conclusions
We presented a DHTS system with a single crystal fiber up to 1400 $^{\circ }$C in a YAG fiber for the first time. The system realized a high spatial resolution of 7 cm in a 1 m long YAG fiber. To enhance the SNR and improve response speed, a 2D image restoration method was utilized. The algorithm was simulated and further optimized to adapt the use in the proposed distributed sensing system. The average temperature resolution was about 8 $^{\circ }$C in 1 ms data. A better temperature resolution of 0.62 $^{\circ }$C was realized through the joint algorithm of averaging and NLM with 1000 frame signal traces, which corresponds to only 1 s data. The fast response capability with high-temperature resolution and high-spatial resolution is especially important in extreme detection environment, such as the instantaneous high temperature sensing in rocket firing tests and so on. This work laid a solid foundation for the applications of distributed high temperature fiber sensors in actual harsh and extreme environment.
Funding
Zhejiang Lab (2021ME0PI02, 2022ME0AL02); National Natural Science Foundation of China (62275234).
Disclosures
The authors declare no conflicts of interest.
Data Availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. M. Pakmehr, J. Costa, A. R. Behbahani, and K. Semega, “Extreme environment sensor system development: optical turbine inlet temperature (TIT) sensing for jet engine,” in AIAA Propulsion and Energy 2019 Forum, (American Institute of Aeronautics and Astronautics, Indianapolis, IN, 2019).
2. M. Willsch, T. Bosselmann, P. Flohr, R. Kull, W. Ecke, I. Latka, D. Fischer, and T. Thiel, “Design of fiber optical high temperature sensors for gas turbine monitoring,” in 20th International Conference on Optical Fibre Sensors, vol. 7503 (SPIE, 2009), pp. 1117–1120.
3. L. B. Zhang, Y. Q. Peng, J. B. Zhou, R. H. Zhang, S. Chen, and X. F. Yang, “Tungsten-rhenium thermocouple sensor for in-situ ultra-high temperature measurement of hypersonic aircraft surface,” in IET & ISA 60th International Instrumentation Symposium 2014, (2014), pp. 1–5.
4. L. Yang and L. Zhi-min, “The research of temperature indicating paints and its application in aero-engine temperature measurement,” Procedia Eng. 99, 1152–1157 (2015). [CrossRef]
5. M. Perrut, P. Caron, M. Thomas, and A. Couret, “High temperature materials for aerospace applications: Ni-based superalloys and γ-TiAl alloys,” C. R. Phys. 19(8), 657–671 (2018). [CrossRef]
6. P. Xu, D. Ba, W. He, H. Hu, and Y. Dong, “Distributed Brillouin optical fiber temperature and strain sensing at a high temperature up to 1000 °C by using an annealed gold-coated fiber,” Opt. Express 26(23), 29724–29734 (2018). [CrossRef]
7. M. Wang, H. Wu, M. Tang, Z. Zhao, Y. Dang, C. Zhao, R. Liao, W. Chen, S. Fu, C. Yang, W. Tong, P. P. Shum, and D. Liu, “Few-mode fiber based Raman distributed temperature sensing,” Opt. Express 25(5), 4907–4916 (2017). [CrossRef]
8. J. Li, Q. Zhang, Y. Xu, M. Zhang, J. Zhang, L. Qiao, M. M. Promi, and T. Wang, “High-accuracy distributed temperature measurement using difference sensitive-temperature compensation for Raman-based optical fiber sensing,” Opt. Express 27(25), 36183–36196 (2019). [CrossRef]
9. Q. Guo, S. Liu, X. Pan, B. Wang, Z. Tian, C. Chen, Q. Chen, Y. Yu, H. Sun, and H. Sun, “Femtosecond laser inscribed helical sapphire fiber Bragg gratings,” Opt. Lett. 46(19), 4836–4839 (2021). [CrossRef]
10. X. Xu, J. He, J. He, B. Xu, R. Chen, Y. Wang, Y. Yang, and Y. Wang, “Efficient point-by-point Bragg grating inscription in sapphire fiber using femtosecond laser filaments,” Opt. Lett. 46(11), 2742–2745 (2021). [CrossRef]
11. B. Wang, Y. Niu, S. Zheng, Y. Yin, and M. Ding, “A high temperature sensor based on sapphire fiber fabry-perot interferometer,” IEEE Photonics Technol. Lett. 32(2), 89–92 (2020). [CrossRef]
12. W. Wei, S. Xiaotian, and W. Ying, “Sapphire fiber-optic temperature sensor based on black-body rRadiation law,” Procedia Eng. 99, 1179–1184 (2015). [CrossRef]
13. L. Ye, Y. Zhang, W. Zhao, and Y. Shen, “Sapphire fiber thermometer based on fluorescence lifetime measurement,” Proc. SPIE 4920, 25–30 (2002). [CrossRef]
14. R. Bao, L. Yu, L. Ye, X. Zhang, and L.-G. Wang, “Compact and sensitive Er3+/Yb3+ co-doped YAG single crystal optical fiber thermometry based on up-conversion luminescence,” Sens. Actuators, A 269, 182–187 (2018). [CrossRef]
15. S. Wang, K. Zhu, T. Wang, H. Zhou, N. Zhang, J. Zhang, L. Ye, Z. Jia, and X. Tao, “Sensitive Ho3+,Yb3+ co-doped mixed sesquioxide single crystal fibers thermometry based on upconversion luminescence,” J. Alloys Compd. 891, 162062 (2022). [CrossRef]
16. J. He, X. Xu, C. Liao, C. Wang, and Y. Wang, “Femtosecond-laser-inscribed fiber bragg gratings for high-tempertature sensing,” in 2018 Asia Communications and Photonics Conference (ACP), (2018), pp. 1–3.
17. Q. Guo, Y.-S. Yu, Z.-M. Zheng, C. Chen, P.-L. Wang, Z.-N. Tian, Y. Zhao, X.-Y. Ming, Q.-D. Chen, H. Yang, and H.-B. Sun, “Femtosecond laser inscribed sapphire fiber bragg grating for high temperature and strain sensing,” IEEE Trans. Nanotechnol. 18, 208–211 (2019). [CrossRef]
18. W. Li, T. Liang, P. Jia, C. Lei, Y. Hong, Y. Li, Z. Yao, W. Liu, and J. Xiong, “Fiber-optic Fabry Perot pressure sensor based on sapphire direct bonding for high-temperature applications,” Appl. Opt. 58(7), 1662–1666 (2019). [CrossRef]
19. Z. Wang, J. Chen, H. Wei, H. Liu, Z. Ma, N. Chen, Z. Chen, T. Wang, and F. Pang, “Sapphire Fabry-Perot interferometer for high-temperature pressure sensing,” Appl. Opt. 59(17), 5189–5196 (2020). [CrossRef]
20. S. Yang, D. Hu, and A. Wang, “Point-by-point fabrication and characterization of sapphire fiber Bragg gratings,” Opt. Lett. 42(20), 4219–4222 (2017). [CrossRef]
21. S. Yang, D. Homa, H. Heyl, L. Theis, J. Beach, B. Dudding, G. Acord, D. Taylor, G. Pickrell, and A. Wang, “Application of sapphire-fiber-bragg-grating-based multi-point temperature sensor in boilers at a commercial power plant,” Sensors 19(14), 3211 (2019). [CrossRef]
22. X. Xu, J. He, C. Liao, K. Yang, K. Guo, C. Li, Y. Zhang, Z. Ouyang, and Y. Wang, “Sapphire fiber Bragg gratings inscribed with a femtosecond laser line-by-line scanning technique,” Opt. Lett. 43(19), 4562–4565 (2018). [CrossRef]
23. A. Wang, S. Gollapudi, K. A. Murphy, R. G. May, and R. O. Claus, “Sapphire-fiber-based intrinsic Fabry-Perot interferometer,” Opt. Lett. 17(14), 1021–1023 (1992). [CrossRef]
24. A. Ukil, H. Braendle, and P. Krippner, “Distributed temperature sensing: review of technology and applications,” IEEE Sens. J. 12(5), 885–892 (2012). [CrossRef]
25. B. Liu, Z. Yu, Z. Tian, D. Homa, C. Hill, A. Wang, and G. Pickrell, “Temperature dependence of sapphire fiber Raman scattering,” Opt. Lett. 40(9), 2041 (2015). [CrossRef]
26. B. Liu, Z. Yu, C. Hill, Y. Cheng, D. Homa, G. Pickrell, and A. Wang, “Sapphire-fiber-based distributed high-temperature sensing system,” Opt. Lett. 41(18), 4405–4408 (2016). [CrossRef]
27. B. Liu, M. P. Buric, B. T. Chorpening, Z. Yu, D. S. Homa, G. R. Pickrell, and A. Wang, “Design and implementation of distributed ultra-high temperature sensing system with a single crystal fiber,” J. Lightwave Technol. 36(23), 5511–5520 (2018). [CrossRef]
28. G. Lim, J. Wuenschell, and M. P. Buric, “High temperature sensing using raman-DTS with single-crystal fibers for nuclear application,” Proc. SPIE 12105, 121050B (2022). [CrossRef]
29. S. Hu, C. Lu, X. Liu, and Z. Xu, “Optical temperature sensing based on the luminescence from YAG:Pr transparent ceramics,” Opt. Mater. 60, 394–397 (2016). [CrossRef]
30. S. Bera, J. Wuenschell, J. Thapa, M. Buric, and B. Liu, “Raman scattering in single-crystal YAG for application in distributed temperature sensing,” Proc. SPIE 11682, 1168217 (2021). [CrossRef]
31. M. Höbel, J. Ricka, M. Wüthrich, and T. Binkert, “High-resolution distributed temperature sensing with the multiphoton-timing technique,” Appl. Opt. 34(16), 2955–2967 (1995). [CrossRef]
32. B. Liu, D. Karki, S. Bera, and P. O. Jr, “Fabrication and application of single crystal fiber via laser heated pedestal growth system,” Proc. SPIE 11739, 1173903 (2021). [CrossRef]
33. G. Bolognini, J. Park, M. A. Soto, N. Park, and F. D. Pasquale, “Analysis of distributed temperature sensing based on Raman scattering using OTDR coding and discrete Raman amplification,” Meas. Sci. Technol. 18(10), 3211–3218 (2007). [CrossRef]
34. M. A. Soto, G. Bolognini, and F. D. Pasquale, “Analysis of optical pulse coding in spontaneous Brillouin-based distributed temperature sensors,” Opt. Express 16(23), 19097–19111 (2008). [CrossRef]
35. J. P. Bazzo, D. R. Pipa, C. Martelli, E. Vagner da Silva, and J. C. Cardozo da Silva, “Improving Spatial Resolution of Raman DTS Using Total Variation Deconvolution,” IEEE Sens. J. 16(11), 4425–4430 (2016). [CrossRef]
36. M. K. Saxena, S. D. V. S. J. Raju, R. Arya, S. V. G. Ravindranath, S. Kher, and S. M. Oak, “Optical fiber distributed temperature sensor using short term Fourier transform based simplified signal processing of Raman signals,” Measurement 47, 345–355 (2014). [CrossRef]
37. M. K. Saxena, S. D. V. S. J. Raju, R. Arya, R. B. Pachori, S. V. G. Ravindranath, S. Kher, and S. M. Oak, “Raman optical fiber distributed temperature sensor using wavelet transform based simplified signal processing of Raman backscattered signals,” Opt. Laser Technol. 65, 14–24 (2015). [CrossRef]
38. A. Buades, B. Coll, and J.-M. Morel, “A non-local algorithm for image denoising,” in 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), vol. 2 (2005), pp. 60–65 vol. 2. Journal Abbreviation: 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05).
39. A. Buades, B. Coll, and J. M. Morel, “A review of image denoising algorithms, with a new one,” Multiscale Model. Simul. 4(2), 490–530 (2005). [CrossRef]
40. A. Buades, B. Coll, and J.-M. Morel, “Non-local means denoising,” Image Processing On Line 1, 208–212 (2011). [CrossRef]
41. M. A. Soto, J. A. Ramírez, and L. Thévenaz, “Intensifying the response of distributed optical fibre sensors using 2D and 3D image restoration,” Nat. Commun. 7(1), 10870 (2016). [CrossRef]
