Published: November 7th, 2016
We demonstrate use of a fiber optic distributed sensor for mapping the temperature field of mixing air jets. The Rayleigh scattering-based sensor generates thousands of data points along a single fiber to provide exceptional spatial resolution that is unattainable with traditional sensors such as thermocouples.
The reliability of computational fluid dynamics (CFD) codes is checked by comparing simulations with experimental data. A typical data set consists chiefly of velocity and temperature readings, both ideally having high spatial and temporal resolution to facilitate rigorous code validation. While high resolution velocity data is readily obtained through optical measurement techniques such as particle image velocimetry, it has proven difficult to obtain temperature data with similar resolution. Traditional sensors such as thermocouples cannot fill this role, but the recent development of distributed sensing based on Rayleigh scattering and swept-wave interferometry offers resolution suitable for CFD code validation work. Thousands of temperature measurements can be generated along a single thin optical fiber at hundreds of Hertz. Sensors function over large temperature ranges and within opaque fluids where optical techniques are unsuitable. But this type of sensor is sensitive to strain and humidity as well as temperature and so accuracy is affected by handling, vibration, and shifts in relative humidity. Such behavior is quite unlike traditional sensors and so unconventional installation and operating procedures are necessary to ensure accurate measurements. This paper demonstrates implementation of a Rayleigh scattering-type distributed temperature sensor in a thermal mixing experiment involving two air jets at 25 and 45 °C. We present criteria to guide selection of optical fiber for the sensor and describe installation setup for a jet mixing experiment. We illustrate sensor baselining, which links readings to an absolute temperature standard, and discuss practical issues such as errors due to flow-induced vibration. This material can aid those interested in temperature measurements having high data density and bandwidth for fluid dynamics experiments and similar applications. We highlight pitfalls specific to these sensors for consideration in experiment design and operation.
Computational fluid dynamics (CFD) codes are used to simulate a wide variety of fluid systems, from airflow around planes and automobiles down to arterial blood flow. The scope and fidelity of such simulations have grown with the availability of computing power. However, despite the sophistication of advanced simulations, their accuracy and reliability are often hard to quantify. In practice, the accuracy of CFD codes is assessed by comparing simulations with experimental data in a process called code validation.
A typical experimental data set consists chiefly of velocity and temperature measurements, both ideally of high spatial and temporal resolution to facilitate rigorous code validation. Velocity fields can be mapped at high resolution using particle image velocimetry (PIV), a well-established optical technique1,2. In contrast, it is difficult to map temperature fields with resolution comparable to that of PIV. Optical techniques such as laser-induced fluorescence are available3,4, but they require cameras and relatively high-power lasers, and are unsuitable for opaque fluids.
An alternative is available in the relatively new technique of distributed temperature sensing based on Rayleigh scattering and swept-wavelength interferometry (SWI)5-7. Thousands of temperature measurements can be acquired along a single optical fiber. A distributed temperature sensor (DTS) can span large flow fields and function in environments that are unsuitable for image-based techniques8. There are also DTSs based on Raman and Brillouin scattering9,10, but sensors based on Rayleigh scattering and SWI provide spatial and temporal resolution more suitable for typical fluid dynamics experiments.
Though DTSs offer data density far beyond that of traditional sensors such as thermocouples (TCs), sensors based on Rayleigh scattering respond to strain as well as temperature11. If the fiber coating is hygroscopic, sensors also respond to humidity changes12,13. Absorption of water vapor swells the coating while desorption shrinks it14, which strains the underlying glass fiber and alters the signal. As a result, accuracy is affected by handling, vibration, and shifts in relative humidity. This is quite unlike traditional sensors and so unconventional installation and measurement methods must be observed to obtain accurate data. This paper demonstrates the use of a DTS in a thermal mixing experiment, presenting a protocol and guidelines to ensure accuracy.
The DTS used here is based on detection and analysis of Rayleigh scattering within a fiber optic waveguide. A random distribution of impurities and structural variations along the fiber core gives rise to a backscatter pattern that is unique to the fiber and generally stable. The spectrum and amplitude of this pattern can be read to serve as a fiber signature. Physical changes such as temperature shifts or strain alter the signature in a repeatable way, and detecting signature variations is the basis for using the fiber as a sensor.
Figure 1 illustrates the principle components of the optoelectronic sensing device, called an optical distributed sensor interrogator, and denoted here simply as "interrogator". In a technique known as swept-wavelength interferometry, a low power tunable laser launches a narrow band signal into the fiber for the purpose of registering resultant backscatter5-7. The laser is swept across an interval of several nanometers and the signal split between reference and measurement legs. Scattered light from the sensor is combined with the reference signal to generate interference signals at the detectors. Detector output is digitized and analyzed to retrieve the Rayleigh scattering signal. The Rayleigh signature of the sensor shifts in wavelength where sensor temperature (strain, or humidity) changes. The magnitude of this wavelength shift is related to sensor sensitivity, which is a physical constant associated with the fiber type, which has a calibration factor analogous to the Seebeck coefficient of a TC.
Figure 2 shows the glass tank that serves as the test section used in this study. The camera behind the tank gives a sense of scale. Air enters through two hexagonal ducts and mixes before exiting through a vent. To highlight the jets, one flow stream was seeded with oil mist while the other remained pure air. The tank lid has a window covered with a black polymer screen. Though not visible in the photo, the DTS is suspended below the black screen.
A 50 m long DTS was mounted below the tank lid as shown in Fig. 3. It was fashioned from 155 µm diameter polyimide-coated optical fiber and hung on 127 µm diameter steel wire strung between tank end panels. The sensor was woven through the wire in an alternating pattern and looped back and forth across the tank 49 times. It spans a 0.5 x 0.8 m plane and generates 1,355 independent data points at 4 Hz and spatial resolution of 30 mm, 4,067 data points when oversampled with 10 mm spacing. Such high density temperature data complements velocity data and increases the value of data sets for CFD validation. The protocol outlines the process of sensor selection, fabrication, and configuration while focusing on the particular concerns in using the DTS in a fluid dynamics experiment.
1. Select Optimal Sensor Type for Application
2. Install Optical Fiber in Test Section
3. Splice Connector and Termination to Fiber
4. Sensor Configuration
5. Map Sensor Position within Test Section
6. Sensor Baseline: The Link to Absolute Temperature
7. Run Test
8. Data Analysis
Raw DTS data is plotted in Fig. 6 showing measured ΔT from the baseline temperature (roughly 20 °C) versus distance along the sensor. The data is "raw" in the sense that it has neither been converted to absolute temperature nor mapped to physical positions within the test section. Data is based on a 30 mm gage length, which provides 1,666 independent measurements over the full sensor length of 50 m. The 30 mm gage was applied at 10 mm intervals in an oversampling mode that increases the number of data points to 5,000. Such data density is not practicable with conventional sensors such as TCs.
At x=0 in Fig. 6 the sensor is at the east end of the tank, and as x increases it loops back and forth towards the west end. Peaks occur where the sensor passes over the hot east jet and then fade where it is over the cold west jet. The plot illustrates how even the raw signal from a single DTS can provide a basic portrayal of temperature over a rather wide region. Note the signal noise towards the west end of the fiber, which is due to flow-induced vibration. Though vibration was not visible to the naked eye, it was sufficient to degrade the signal and we see this problem most often with long sensors (> 10 m).
The raw data is mapped onto the test section in Fig. 7, which shows temperature across the 0.5 x 0.8 m measurement plane formed by the DTS array. The point of view is from above the tank looking down onto the lid. Outlines of the hexagonal channels are included as an orientation aid. The contour is based on 4,067 data points since the loops taped under the lid are excluded. Linear interpolation between adjacent sensor segments was used to create the 2D contour.
The contour provides a clear sense of the thermal pattern beneath the lid with a warm region over the east jet, but not centered around it. Also evident is a rough symmetry around the tank midplane, which is y=0 on the plot. This sort of temperature data is a useful compliment to velocity data in fluid dynamics studies involving thermal mixing and heat transfer. Rigorous code validation requires such high resolution data for both the temperature and velocity fields.
The same sensor data can be processed to reveal the magnitude of temperature fluctuations. The RMS (root mean square) of the 2,000 scan data set is plotted in Figure 8. Magenta marks the region where temperature fluctuations are relatively high. This is also a region of high turbulence where the two rising jets interact as impinge on the lid. RMS data is useful for turbulence modeling in the context of thermal mixing.
Figure 1. Interrogator schematic. Principle components of optical distributed sensor interrogator for temperature measurements. The system is based on swept-wavelength interferometry, which characterizes the sensor's Rayleigh backscatter signature. Please click here to view a larger version of this figure.
Figure 2. Test section. Air jet mixing experiment: air enters tank through base via two hexagonal ducts and mixes before exiting through top vent. The black screen covering the lid window is 3 mm above the DTS (not visible). Please click here to view a larger version of this figure.
Figure 3. DTS mounting configuration. Top view of tank showing DTS woven between steel support wires strung across the long axis of the tank. Please click here to view a larger version of this figure.
Figure 4. DTS close-up. Close-up photo of DTS with view from inside tank upward at lid to highlight sensor loops, attachment, and location of first test point to be mapped with soldering iron. Please click here to view a larger version of this figure.
Figure 5. Rayleigh scattering signal. Typical Rayleigh scattering signal recorded with sensor configuration utility (short sensor shown here for clarity). Proper termination will generate sharp signal drop to noise floor. The slight signal step up and modest reflection at the connector is characteristic of a properly spliced connector. Please click here to view a larger version of this figure.
Figure 6. Raw DTS data. A single scan of raw DTS data with the hot east jet at 45 °C and cold west jet at 25 °C. Peaks occur where sensor is directly above hot jet. Recall that the sensor is looping back and forth between tank walls. Please click here to view a larger version of this figure.
Figure 7. Measured air temperature below lid. DTS raw data converted to absolute temperature and mapped to physical position within tank. Data based on 2,000 scans logged at 4 Hz. Data spacing 10 mm for a total of 4,067 plotted data points. Linear interpolation used to fill regions between sensor segments. Hexagons show positions of inlets. Please click here to view a larger version of this figure.
Figure 8. Root mean square (RMS) of measured temperature. RMS of data plotted in Fig. 7. Magenta indicates high temperature fluctuations and thermal mixing of hot and cold jets. Hexagons show positions of inlets. Please click here to view a larger version of this figure.
Table 1. Order of magnitude thermal response time for selected fiber types and housing configurations in cross flow at 1 m/sec and 20 °C.
Table 2. Approximate operating temperature limits and humidity sensitivities for selected coating configurations.
We have demonstrated the use of a DTS in a fluid dynamics experiment. The main advantage of these sensors is the great number of measurement points that can be obtained from a single sensor. The DTS used here generated data at 4,067 points across a 0.5 x 0.8 m plane, far beyond the practicable limits of conventional point sensors such as thermocouples. While such data density can be exceeded by optical techniques such as laser induced fluorescence (LIF), a DTS will function in opaque fluids and applications that lack optical access. The high data density of a DTS is suitable for experiments involved in computational fluid dynamics code validation.
Baselining is the critical step in the protocol and central in determining measurement accuracy. An isothermal test section is essential to ensure the entire DTS is at one temperature when baselined. If this is not possible, Tbase becomes Tbase (x), which should be mapped by multiple TCs placed in close proximity to the DTS. Though baseline quality can be improved in this fashion, it complicates the process of mapping the DTS baseline to the standards for conversion to absolute temperature.
Always be on the lookout for sources of strain after the baseline, which can introduce unpredictable signal shifts. Such sources are, for example, test section thermal expansion that stretches the sensor, movement of supports, dynamic loading from high flow rates, or flow-induced vibration. The pre- and posttest measurements under isothermal conditions will help identify such problems.
Strain sensitivity is the main shortcoming of this Rayleigh scattering-based DTS. Unlike conventional sensors like thermocouples, it is sensitive to handling, humidity, and vibration. These issues are most relevant for the bare sensor configuration demonstrated here, but far less important for sensors housed in capillaries.
Unlike conventional sensors, a DTS cannot be procured with paperwork tracing it to a recognized calibration standard such as NIST (National Institute of Standards and Technology). In-situ calibrations are necessary, preferably with an isothermal test section, which may be difficult in some applications. Vibration is of special concern for bare fiber strung across a large test section. We have had mixed success with a vertically-oriented array that spans the long axis of the tank at segment lengths of 1.7 m. A configuration with 28 m of fiber and 16 segments performed well during one study18, but attempts to extend it to 53 m with 29 segments was unsuccessful16.
In general, signal noise for any sensor length and configuration can be decreased by increasing the gage length over which the interrogator software calculates the Rayleigh signal shift, but this reduces effective spatial resolution. Each application must strike its own balance between signal noise and spatial resolution. Again, such difficulties can be largely avoided by housing the sensor in a capillary at the expense of extended thermal response time.
This relatively new temperature measurement technology requires development to reduce susceptibility to vibration. Much of this work will necessarily involve the interrogator hardware and software. The sensors themselves may also be improved to reduce sensitivity to handling and humidity changes, which are affected by the fiber coatings. Work could focus on developing coatings superior to the polyimide and acrylate-coated fibers currently commercially available.
The authors have nothing to disclose.
The authors thank Tyler Gorney and Aida Rahim at Luna Innovations for their invaluable technical insight and assistance with our application.
The submitted manuscript has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory ("Argonne"). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357. This work was supported by the U.S. Department of Energy, Office of Nuclear Energy.
|ODiSI-A and -B
|The two systems differ primarily in speed and spatial resolution
|3-hole jacket stripper
|Fiber Instrument Sales
|OFS, Specialty Photonics Division
|OFS, Specialty Photonics Division
|552 HPWR 040
Copyright © 2024 MyJoVE Corporation. All rights reserved