3. All fiber Raman optical fiber distributed temperature sensor with dynamic self-calibration

Temperature sensors are ubiquitous devices that permeate our daily lives. Many areas of temperature measurements require a large area of coverage with high localization accuracy. Raman optical fiber-based distributed temperature sensors (ROFDTSs) are equipped with the ability of providing temperature values as a continuous function of distance along the fiber. In an ROFDTS, every bit of fiber works as a sensing element as well as data transmitting medium, to substitute the role played by several point sensors, thus allowing reduced sensor network cost. ROFDTSs have attracted the attention as a means of temperature monitoring and fire detection in power cables, long pipelines, bore holes, tunnels, and critical installations like oil wells, refineries, induction furnaces, and process control industries. The basic principle of temperature measurement using ROFDTS involves

Raman scattering [10] in conjunction with OTDR. The ratio of Raman anti-Stokes (AS) and Stokes (St) intensities is used for determination of unknown temperature. The AS signal is strongly dependent on temperature, while the Stokes signal is slightly dependent on temperature. Based on time of flight and intensity of Stokes and anti-Stokes signals, location and temperature information can be retrieved. The backscattered light has many spectral components as shown in Figure 3 [24]. For temperature measurements, Raman components are analyzed.

The OTDR principle allows estimation of the location of hot zone whereas Raman scattering permits measurement of temperature of the hot zone. Sensing fiber is coupled to short interrogating laser pulses, and backscattered AS and St components are monitored for signal changes. Unknown temperature of hot zones can be estimated from the ratio (R) of AS and St using the following expression [11]

$$R = \frac{I\_{at}}{I\_s} = \left(\frac{\lambda\_s}{\lambda\_{as}}\right)^4 \exp\left(-\frac{hc\nabla}{kT}\right) \tag{1}$$

where λ<sup>s</sup> and λas are the Stokes and anti-Stokes optical signal wavelengths, ∇ is their wave number separation from the pump laser wavelength, h is Planck's constant, c is the velocity of light, and k is Boltzmann's constant. AS is the main signal which carries the signature of temperature variation whereas St provides reference and eliminates a number of effects common to both the signals. One can simplify Eq. (1) by replacing known terms by B where,

$$B = \frac{hc\nabla}{k} \tag{2}$$

Here, the calibration zone is kept at some known absolute room temperature (ϴ). Eq. (3) can be deduced after taking the quotient of the ratio profile at unknown temperature (RT) for an arbitrary zone and the ratio value at the calibration zone (Rϴ) and solving it for T. Parameter (Rϴ) is the ratio value of AS to St signal (AS/St) for the calibration zone (at temperature ϴ) of length 1 m chosen from sensing fiber at the laser end. In Eq. (3), parameters B and ϴ are known. Therefore, Eq. (3) will yield temperature profile (T in °C) for complete fiber length, provided that profiles

Figure 4 shows the block diagram of Raman optical fiber distributed

Further, Figure 5 gives an idea of averaged anti-Stokes Raman signal for a 2.5 m long zone heated by a proportional intergral derivative (PID) controlled heating oven. To determine the unknown temperature profile with certain accuracy for complete fiber length by using Eq. (1), appropriate measures are to be devised and implemented to address several error-causing issues. These issues are described as below [21, 23]. The author's laboratory has successfully solved these issues and

The first issue is the difference in theoretical and experimental values of the ratio (R) at various temperature values. For example, at room temperature (25°C, say), the theoretical and experimental values of R are 0.1693 and 0.55, respectively. At 50°C, the theoretical and experimental values of R are found to be 0.1995 and 0.658, respectively. On the other hand, the theoretical and experimental values are 0.2415 and 0.8279, respectively, at 85°C. The reason for this difference is explained

At 25°C (for example), obtaining a theoretical value of 0.1693 for R requires that the optoelectronic conversion using photomultiplier tube (e.g. PMT-R5108, Hamamatsu) detectors, the beam splitting, and the subsequent light coupling into AS and St detectors are in such a way that the relation St = 5.906 AS is maintained for backscattered AS and St signals while traveling the path from fiber to the final stage of detection. However, due to nonideal behavior of various optical components in

of RT and value of R<sup>θ</sup> are available.

Distributed, Advanced Fiber Optic Sensors DOI: http://dx.doi.org/10.5772/intechopen.83622

designed a field portable unit.

temperature sensor.

3.1 Issue no. 1

below [23].

Figure 4.

59

Block diagram of Raman optical fiber distributed temperature sensor.

Since values of h, c, and k are known, the numerical value of B is found to be 631.3 for silica fiber having ∇ = 440 cm�<sup>1</sup> .

One can simplify the profile analysis by referencing the ratio profile at unknown temperature to the ratio value at known temperature of a pre-selected calibration zone of fiber. The temperature of a given zone T (°C) is then given by the following expression [25].

### Figure 3.

Backscattered laser light from optical fiber in the case of Raman and Brillouin scattering (source: https://www.g oogle.co.in/search?q=BACKSCATTERED+LASER+LIGHT&source=lnms&tbm=isch&sa=X&ved=0ah UKEwiCvf-iq9jcAhWKT30KHdJ4CjYQ\_AUICigB&biw=1093&bih=530) (images) [24].

### Distributed, Advanced Fiber Optic Sensors DOI: http://dx.doi.org/10.5772/intechopen.83622

Here, the calibration zone is kept at some known absolute room temperature (ϴ). Eq. (3) can be deduced after taking the quotient of the ratio profile at unknown temperature (RT) for an arbitrary zone and the ratio value at the calibration zone (Rϴ) and solving it for T. Parameter (Rϴ) is the ratio value of AS to St signal (AS/St) for the calibration zone (at temperature ϴ) of length 1 m chosen from sensing fiber at the laser end. In Eq. (3), parameters B and ϴ are known. Therefore, Eq. (3) will yield temperature profile (T in °C) for complete fiber length, provided that profiles of RT and value of R<sup>θ</sup> are available.

Figure 4 shows the block diagram of Raman optical fiber distributed temperature sensor.

Further, Figure 5 gives an idea of averaged anti-Stokes Raman signal for a 2.5 m long zone heated by a proportional intergral derivative (PID) controlled heating oven.

To determine the unknown temperature profile with certain accuracy for complete fiber length by using Eq. (1), appropriate measures are to be devised and implemented to address several error-causing issues. These issues are described as below [21, 23]. The author's laboratory has successfully solved these issues and designed a field portable unit.

### 3.1 Issue no. 1

Raman scattering [10] in conjunction with OTDR. The ratio of Raman anti-Stokes (AS) and Stokes (St) intensities is used for determination of unknown temperature. The AS signal is strongly dependent on temperature, while the Stokes signal is slightly dependent on temperature. Based on time of flight and intensity of Stokes and anti-Stokes signals, location and temperature information can be retrieved. The backscattered light has many spectral components as shown in Figure 3 [24]. For

The OTDR principle allows estimation of the location of hot zone whereas Raman scattering permits measurement of temperature of the hot zone. Sensing fiber is coupled to short interrogating laser pulses, and backscattered AS and St components are monitored for signal changes. Unknown temperature of hot zones can be estimated from the ratio (R) of AS and St using the following expression [11]

> <sup>¼</sup> <sup>λ</sup><sup>s</sup> λas � �<sup>4</sup>

where λ<sup>s</sup> and λas are the Stokes and anti-Stokes optical signal wavelengths, ∇ is their wave number separation from the pump laser wavelength, h is Planck's constant, c is the velocity of light, and k is Boltzmann's constant. AS is the main signal which carries the signature of temperature variation whereas St provides reference and eliminates a number of effects common to both the signals. One can

<sup>B</sup> <sup>¼</sup> hc<sup>∇</sup>

Since values of h, c, and k are known, the numerical value of B is found to be

. One can simplify the profile analysis by referencing the ratio profile at unknown temperature to the ratio value at known temperature of a pre-selected calibration zone of fiber. The temperature of a given zone T (°C) is then given by the following

> <sup>θ</sup> � ln RT þ ln R<sup>θ</sup> � � " #

Backscattered laser light from optical fiber in the case of Raman and Brillouin scattering (source: https://www.g oogle.co.in/search?q=BACKSCATTERED+LASER+LIGHT&source=lnms&tbm=isch&sa=X&ved=0ah

UKEwiCvf-iq9jcAhWKT30KHdJ4CjYQ\_AUICigB&biw=1093&bih=530) (images) [24].

exp � hc<sup>∇</sup> kT � �

<sup>k</sup> (2)

� 273 (3)

(1)

temperature measurements, Raman components are analyzed.

Applications of Optical Fibers for Sensing

<sup>R</sup> <sup>¼</sup> Ias Is

simplify Eq. (1) by replacing known terms by B where,

631.3 for silica fiber having ∇ = 440 cm�<sup>1</sup>

T ° C

� � <sup>¼</sup> <sup>B</sup>: <sup>1</sup> B: <sup>1</sup>

expression [25].

Figure 3.

58

The first issue is the difference in theoretical and experimental values of the ratio (R) at various temperature values. For example, at room temperature (25°C, say), the theoretical and experimental values of R are 0.1693 and 0.55, respectively. At 50°C, the theoretical and experimental values of R are found to be 0.1995 and 0.658, respectively. On the other hand, the theoretical and experimental values are 0.2415 and 0.8279, respectively, at 85°C. The reason for this difference is explained below [23].

At 25°C (for example), obtaining a theoretical value of 0.1693 for R requires that the optoelectronic conversion using photomultiplier tube (e.g. PMT-R5108, Hamamatsu) detectors, the beam splitting, and the subsequent light coupling into AS and St detectors are in such a way that the relation St = 5.906 AS is maintained for backscattered AS and St signals while traveling the path from fiber to the final stage of detection. However, due to nonideal behavior of various optical components in

Figure 4. Block diagram of Raman optical fiber distributed temperature sensor.

Fourier filtering causes spatial inaccuracy in locating the hot-zones which in turn

The amplitude of AS and St signals varies with time due to slow variations/drifts in laser power and laser-fiber coupling. Also, the temperature of calibration zone itself may change unless it is controlled by a dedicated setup. Therefore, any previously stored reference values of AS and St signals and calibration zone temperature can no longer be used as a reference for temperature measurement at a later stage.

Stoddart et al. [20] proposed to use Rayleigh instead of St from the backscattered spectrum to avoid the temperature measurement error in hydrogen-rich environments due to differential attenuation caused by the optical fiber for AS and St signal wavelengths. This resulted in better results but could not eliminate the error caused by the differential attenuation completely. The dual-ended (DE) configuration [26] (i.e. both ends of sensing fiber are connected to ROFDTS unit) and dual laser source schemes [27, 28] have also been proposed to take care of the difference in attenuation between AS and St. These schemes have resulted in improvements but add complexity and need double length of fiber, extra distributed temperature sensor (DTS) with an optical switch, and two costly lasers. A correction method to take care of the difference in attenuation for AS and St signals has been proposed with only one light source and one light detector but requires attachment of a carefully designed reflective mirror at the far fiber-end of the sensing fiber [29]. Recently, a more sophisticated correction technique [30] based on detection of AS signal alone in combination with DE configuration has been investigated. ROFDTSs based on the above schemes are important and to a certain extent become mandatory in situations where sensing fiber is exposed to the severe radiation environment or hydrogen darkening in oil wells. Requirements for less demanding situations like temperature measurement in steam pipelines of turbines, electrical cables and temperature profiling of big buildings, gas pipelines and mines etc. can be met by the

In order to address the above issues satisfactorily, a discrete wavelet transform

The DWT-based technique is simpler, more automatic, and provides a single solution to address all the above issues simultaneously. The DWT technique takes care of the difference in optical attenuation for AS and St signals by using their trend and also de-noises the AS and St signals while preserving spatial locations of peaks. Also, this technique requires just 1 m long calibration zone which is much less than the 100 m required in the previous technique. Moreover, the dynamic measurement of calibration zone's temperature eliminates the requirement of keeping the calibration zone at a constant temperature, and thus, complicated heating arrangement is avoided. Actual wavelet transform-based processed signal profile is shown in Figure 6. Table 1 presents the comparison of error in temperature measurement at various zones using Eq. (3) with unprocessed and processed Raman signals. Both absolute errors and percentage errors (in brackets) are

(DWT)-based dynamic self-calibration and de-noising technique is used and implemented by the authors as given in detail [23]. Briefly, wavelets are mathematical functions that can be used to segregate data into various frequency components. Each component can then be studied with a resolution matched to its scale. In DWT, a signal may be represented by its low frequency component and its high

yields erroneous information about the location of hot zones [18].

3.5 Approaches to solve the problematic issues

technique based on digital signal processing.

frequency component.

61

3.4 Issue no. 4

Distributed, Advanced Fiber Optic Sensors DOI: http://dx.doi.org/10.5772/intechopen.83622

Figure 5. The anti-Stokes Raman signal profile at various temperatures.

the path and band nature of AS and St signals, the above relation does not hold. The relation gets deteriorated at every stage in the path. For example, the cathode radiant sensitivity of a PMT for AS wavelength (1018 nm) and St wavelength (1109 nm) is 0.95 mA/W and 0.2mA/W, respectively, which causes St current to be approximately 5 times less compared to AS current. Nonideal performance of beam splitters and optical filters also does not support the above ideal relation. As a result, the cumulative effect of various components makes experimental values of R to be different from the theoretical one.

Direct use of experimental values of R in Eq. (1) will yield highly erroneous and unacceptable temperature profile (T). Hence, Eq. (1) needs to be modified to obtain correct values. Modification is done by referencing the experimentally obtained ratio values with respect to the ratio value at some known temperature of calibration zone which is chosen from sensing fiber itself.

### 3.2 Issue no. 2

The second issue is the nonidentical fiber attenuation along the fiber length for Raman AS and St signals due to difference in their wavelengths [9, 11]. In a typical system using 1064 nm excitation laser, the difference between two wavelengths is 90 nm. The lower optical wavelength signal (AS) experiences higher attenuation in comparison to higher optical wavelength signal (St) while traveling in sensing optical fiber. This attenuation difference results in an unwanted downward slope in ratio (R) profile and finally in unknown temperature (T) profile with respect to fiber length. It may be noted that downward slope in ratio (R) profile causes additional errors in unknown temperature (T) profile of fiber and should be corrected.

### 3.3 Issue no. 3

While de-noising Raman AS and St signals for better signal-to-noise ratio (SNR), conventional finite impulse response/infinite impulse response (FIR/IIR)-based

Fourier filtering causes spatial inaccuracy in locating the hot-zones which in turn yields erroneous information about the location of hot zones [18].
