1American Institute of Aeronautics and Astronautics

DEVELOPMENT OF A NON-INTRUSIVE TEMPERATURESENSOR IN A MODEL GAS TURBINE COMBUSTOR

T. P. Jenkins*, P. A. DeBarber†, MetroLaser Incorporated, Irvine, CAT. Yoshimura, J. Shen, and V. G. McDonell, University of California, Irvine

* AIAA member, Senior Scientist† AIAA senior member,Technical Director

ABSTRACT

A tunable diode laser temperature sensor was tested in a model gas turbine combustor at measurement rates of 200 Hz, enabling transient combustion phenomena to be resolved. Fiber optics were used to transmit laser light to the combustor. The sensor has the potential for use at high pressures typical of gas turbine combustors, and may provide a means to monitor combustion stability for active control in future lean burn combustors. In principle, the temperature is obtained from the ratio of peak absorbances of two H2O lines near 1500 nm, given accurate knowledge of the temperature dependence of the ratio. However, the measured values of absorbance peak ratio were found to extend beyond the range of predicted values obtained from a simulation based on the existing spectroscopic database. This discrepancy suggests that there are errors in the existing database for this line pair, and that the correct spectroscopic constants must be measured in order to properly calibrate the temperature sensor.

INTRODUCTION

Emissions of NOx from aircraft engines are regulated because it impacts the warming of the earth's climate. NOx can destroy ozone in the upper atmosphere that protects life from the harmful effects of solar UV radiation. The emissions can also produce ozone in the lower atmosphere around airports, which appears as smog and causes breathing problems in humans. One accepted strategy for reducing the formation of NOx in these systems is to move towards lean, fully premixed operation. Because NOx formation chemistry is highly temperature dependent, a lean, fully premixed fuel/air mixture, which produces a relatively cool reaction with a more even temperature distribution, can greatly reduce thermal NOx formation. A serious problem with fully premixed operation is that it places the combustor at the edge of stability where the reaction is in danger of blowing out, CO and hydrocarbon emissions can rise, and excessive noise and vibration can occur. These factors are driving the development of advanced sensors needed to provide real-time combustion monitoring that might be used in active control to stabilize lean combustors.

Tunable diode laser absorption spectroscopy (TDLAS) has arisen as an attractive approach to temperature measurement in combustion systems. Recent experimental studies in laboratory1 and industrial settings2,3 have shown that rapid, non-intrusive temperature measurements can be reliably obtained with a rugged, relatively inexpensive tunable diode laser sensor. Thermocouples suffer from slow time response, tend to disturb the flame being measured, and fail at high temperatures. Diode laser sensors do not have any of these shortcomings, and have been shown to be capable of withstanding very harsh environments.

The present paper seeks to apply a diode laser sensor to a model gas turbine combustor to investigate the potential for real-time combustion monitoring in turbine engines.

EXPERIMENTAL APPARATUS

Model Combustor

The combustor used for these experiments is a single can type model gas turbine combustor. Figure 1shows a dimensioned cross section of the combustor can as well as the premixing section before the can and the chimney after the can. The steel can has an ID of 80mm and a length of 160mm. A pressure tap for a dynamic pressure sensor is located halfway up the can wall. A geometrically identical quartz liner was also utilized to enable optical measurements. The combustor has a quarl at its inlet with a minimum diameter of 32mm. The exit of the can has a contraction with a final diameter of 46mm. The chimney section located after the exit contraction has an inner diameter of 46mm and an overall length of 343mm.

A diffusion pilot centerbody configuration was selected to represent current practical lean premixed combustion devices. The centerbody has an OD of 25.4mm. The diffusion pilot has a 12.7mm radius dome at its tip. Fuel is injected axially through a 0.56mm diameter hole at the tip of the dome. The diffusion pilot was adjusted to a point 5mm upstream of the contraction.

42nd AIAA Aerospace Sciences Meeting and Exhibit5 - 8 January 2004, Reno, Nevada

AIAA 2004-546

Copyright © 2004 by the authors. Published by the American Institute of Aeronautics and Astronautics, Inc., with permission.

2American Institute of Aeronautics and Astronautics

160m m

80mm

190m m

1

2

5

3

6

7

480 mm

TDLAS line

Figure 1. Cross section of combustor and flow preparation hardware including: 1)Main air plenum, 2) Axial Swirler, 3)Pilot Centerbody with installed monolith, 4)Quarl venturi, 5)Combustor Can, 6)Dynamic pressure sensor, and 7)Water cooled emissions extraction probe.

There are two flow paths for all pilot configurations: main flow and pilot flow. The fully premixed main flow enters through a plenum and passes through a 45-degree axial swirler prior to entering the combustor can. The pilot flow passes through each injector as described above and can contain a portion of the air in addition to the pilot fuel. The total firing rate of the combustor is approximately 30kW thermal. All tests are at atmospheric pressure. Total airflow is 0.02kg/s. Air inlet temperatures prior to fuel injection are 700K for the main. Independent metering and electrical heating systems supply the main and pilot airflows. Both air systems utilize a pressure regulator upstream of a sonic venturi to provide flow metering. The fuel is grid supplied natural gas with a composition of 97% CH4, 1.6%CO2, 1%C2H6, 0.3%N2, and 0.1%C3H8 determined using a Total Flow Analysis BTU 8000 gas chromatograph. Brooks model 5850E mass flow controllers provide metering of the main and pilot fuels.

Temperature Sensor

Figure 2 shows a diagram of the TDLAS sensor as it was applied to measurements in the model gas turbine combustor facility. To enable optical access the quartz combustor body was used. The sensor system consisted of a single tunable DFB laser near 1500 nm, a

diode laser controller, a waveform generator, a fiber coupled lens, an InGaAs detector, and a computer based data acquisition and processing system. Collimating lenses projected the TDLAS beam through the measurement zone at the location indicated in Figure 1. After exiting the fiber, the beam was directed to a mirror on the opposite side of the combustor where it was reflected back through the flame twice, giving a total of four passes.

Diode lasercontroller

InGaAs Detector

Waveform generator

Model gas turbinecombustor

Mirror

Collimating lens

Fiber opticLaptop computer

Figure 2. Schematic of diode laser temperature sensor applied to model combustor in multipass arrangement.

A 10-kHz triangle wave produced by the waveform generator was used to drive the diode laser current. Each half cycle produced a scan in wavelength, enabling a sweep that covered two H2O absorption lines in about 40 µs. Signals from the detectors were digitized and processed to compute the peak absorbances of each of the two lines. Using the spectroscopic model, the temperature could be calculated from the ratio of measured absorbances. Although the data collection for one measurement took only 40 µs, processing and displaying the data took about 5 milliseconds, enabling a data acquisition rate of about 200 Hz.

TEMPERATURE MEASUREMENT TECHNIQUE

The technique of two-line TDLAS temperature measurements dates back a quarter century4. A light beam of intensity I0 passing through a medium of absorbing molecular species is attenuated to intensity Iaccording to the Bouguer-Lambert law,

( )να−= exp0I

I , (1)

where αν is the frequency-dependant absorption coefficient. Absorption occurs when the wavelength of the incident light corresponds to a particular energy level transition of the species. For a particular transition, i, of a gaseous species, the

3American Institute of Aeronautics and Astronautics

absorption coefficient, αν,i, is related to the pressure, P, species mole fraction, Χ, and temperature, T, through

αν,i = PΧSi(T)φν,iL (2)

where Si(T) and φν,i are the linestrength and lineshape functions, respectively of the transition and Lis the path length. Si(T) depends on the fraction of molecules in the absorbing state, and is given by

−−

−−

−−=

0

,0

,0

0

00,0

exp1

exp111"

exp)(

)(

kT

hc

kT

hc

TTk

hcE

T

T

TQ

QSTS

i

i

iii ν

ν (3)

In equation (3), the transition-specific spectroscopic constants are S0,i, Ei”, and ν0,i , representing the linestrength at 296 K, the lower state energy, and the linecenter frequency, respectively. These spectroscopic constants can be found in the HITRAN5 database. The partition function, Q(T), is specific to the molecule. For H2O, a polynomial expression for Q(T) is provided by Gamache.6

The lineshape function, φν,i, is a normalized function that accounts for spectral broadening from Doppler shifts due to molecular motion, and from collisions between molecules that reduce the lifetimes of the upper level energy states, increasing the uncertainty in wavelength absorbed. The function φν,i is typically approximated by a Voigt profile using broadening parameters from the HITRAN database.

To obtain temperature, the absorbance at

linecenter, i,0να , is measured for each of two

transitions. Using equation (2), the ratio of linecenter absorbances, R, is given by

)()(

)(

)(

2

1

2,2

1,1

2,

1,

0

0

0

0

TS

TS

LTSP

LTSPR =Χ

Χ==

νν

νν

φφ

αα

. (4)

The pressure, mole fraction, and path length cancel out. For a line pair in which the broadening parameters are similar, the lineshape functions tend to cancel out as well. Thus, the ratio R is equal to the ratio of linestrengths. Substituting Equation (3) into Equation (4), we have

( )

−−−=

0

21

2,0

1,0 11""exp

TTk

EEhc

S

SR . (5)

Equation (5) shows that a measurement of Ryields the temperature. Additional details of the method can be found elsewhere7,8.

1500-nm Line Pair

For temperature measurements in a gas turbine combustor, a line pair near 1500 nm was chosen that is ideally suited. The two lines are near enough each other that they can both be covered in a single scan with one laser, yet they are isolated enough that they remain separate despite broadening at higher pressures. Also, both lines are relatively weak at room temperature so absorption by ambient water vapor is minimized, and the ratio shows good sensitivity to temperature in the range found in gas turbine combustors.

Figure 3(a) shows a calculated H2O absorbance spectrum for the selected line pair. Measured absorption scans can only provide absorbance relative to the baseline, so a curve fit of the baseline was calculated and is shown in Figure 3(a). To obtain the curve fit, the two strong lines were omitted from the simulated data and the resulting simulated baseline was fit to a fourth-order polynomial using the least squares method. The portions omitted each had a width of twice the full width at half maximum of the line, centered on the corresponding absorption line. To approximate the absorbance spectrum that would be measured by the diode laser sensor, the baseline fit was subtracted from the calculated absorbance spectrum, and is shown in Figure 3(b). As the figure shows, the relative peak heights are slightly different for the baseline-subtracted spectrum in part (b) than for the absolute spectrum in part (a). Thus, when using the measured absorption peak ratio to obtain temperature, an error would be introduces if the absolute predicted absorbance spectrum were used to interpret the data.

-0.002

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0.018

0 0.5 1 1.5 2 2.5 3

ν - ν0 (cm-1)

α = -

1/l*

ln(I

/I 0)

T = 1700 K P = 1 atm XH20 = 0.05 L = 32.0 cmν0 ~ 6700 cm-1

Curve fit to baseline

(a)

4American Institute of Aeronautics and Astronautics

-0.002

0

0.002

0.004

0.006

0.008

0.01

0.012

0.014

0.016

0 0.5 1 1.5 2 2.5 3

ν - ν0 (cm-1)

α min

us b

asel

ine

(b)Figure 3. Calculated H2O absorption spectrum showing (a) absolute absorbance, and (b) absorbance relative to the baseline.

A theoretical calibration curve was generated from the HITRAN predictions by plotting the temperature against the peak ratio, R, and is shown in Figure 4. This curve includes the correction for the baseline as explained above. At the extreme low and high temperatures the curve doubles back, producing an ambiguous temperature for R values less than about 0.1 and greater than about 1.5. However, there is a useful range from about 900 K to about 2500 K over which the temperature can be obtained unambiguously from R. It should be noted that the highest predicted value of R is about 1.6. Thus, any experimental value of R greater than 1.6 would indicate a problem with this prediction.

0

500

1000

1500

2000

2500

3000

3500

0 0.5 1 1.5 2

R

T (

K)

Usable Range

Figure 4. Theoretical calibration curve for TDLAS sensor.

To estimate the behavior of a sensor using this line pair at higher pressures, spectral calculations were made for 10, 15, and 20 atm, and are plotted in Figure 5. As the figure shows, the lines are separated enough that they do not overlap even at 20 atm. Thus, accurate temperature measurements should be possible at pressures at least up to 20 atm. Further experimentation

is needed to verify these spectra experimentally at high pressures.

0

0.001

0.002

0.003

0.004

0 0.5 1 1.5 2 2.5 3

ν - ν0 (cm-1)

α

T = 1500 KXH2O = 0.90L = 1 cmν0 ~ 6700 cm-1

10 atm

15 atm

20 atm

Figure 5. Calculated H2O absorption spectra at high pressure.

RESULTS AND DISCUSSION

Measurements were taken in the model gas turbine combustor by scanning the laser current, and thus wavelength, in a repeating ramp waveform such that a single ramp covered the two absorption lines in 40 µs. Figure 6 shows three typical scans taken at an equivalence ratio of φ = 0.45. The absorption coefficient, α, was calculated from the measured intensity, I, from α = -ln(I/I0), where I0 is the reference intensity obtained by curve fitting the portions of the intensity scan that do not include the strong lines. Note that the abscissa in Figure 6 is time, which can be roughly taken as the optical frequency of the laser. Each of the three scans was taken at the same operating condition about two minutes apart so they are statistically independent. The heights of the peaks, located at about 10 µs and 35 µs, stay constant to within around 10 percent of the peak value, which is on the order of the noise in the absorption measurement as seen by the fluctuations in the flat part of the spectrum between 24 and 33 µs.

-0.005

0

0.005

0.01

0.015

0.02

0 10 20 30 40

Time from start of scan (µs)

α

Scan 1Scan 2Scan 3

φ = 0.45

Figure 6. Three example scans from the model gas turbine combustor at φφφφ = 0.45.

5American Institute of Aeronautics and Astronautics

Turbulent flames are characterized by time scales on the order of the large eddy turnover time, δ/U, where δ represents the size of the largest eddies and Urepresents the mean axial velocity.9 Taking the largest eddies in the combustor to be on the order of the centerbody diameter, or 0.025 m, and using CFD simulations to estimate the highest velocities in the combustor, which are around 100 m/s, the large eddy turnover time is about .025/100 = 0.00025 s, or about 250 µs. For accurate temperature measurements, the duration of the scan needs to be short enough that the flow does not change appreciably while scanning from one absorption peak to the other. Initially, a scan time of 100 µs was tried. However, this resulted in excessive noise in the ratio of absorption peaks, indicating that the flow was changing too rapidly. Reducing the scan time to 40 µs greatly improved the result, as shown by the reproducibility in Figure 6.

Figure 7 shows a time series of the ratio, R, which is the height of the first peak divided by the height of the second. Data rates were about 200 Hz. However, the data are running averages of 10 points, so the effective data rate of Figure 7 is about 20 Hz. At a time of about 25 seconds, a change occurs in the combustor operating conditions from φ = 0.45 to φ = 0.58. The inset plots show how the lineshape changes from one condition to the other, in which the ratio of peaks goes from about 1.2 to about 1.6. These measurements reveal the transient response, showing a steady state before the change, a rapid transition, and then some oscillations until a steady state is again reached at the new condition. The ability of the sensor to capture these transients could be exploited in active control applications.

0

0.5

1

1.5

2

2.5

0 10 20 30 40 50 60

time (s)

R =

αα αα pea

k,1/αα αα p

eak,

2

Lineshapeφ = 0.58Lineshape

φ = 0.45

ν

α

ν

α

Transition tohigher φ

Figure 7. Absorption peak ratio of the two H2O lines in a model gas turbine combustor. The data are running averages of 10 points each.

Theoretically, temperatures can be obtained from the ratios in Figure 7 by using the calibration curve of Figure 4. However, a problem is apparent by

comparing Figure 4 and Figure 7. While the theoretical calibration curve of Figure 4 shows that the ratio would never go above 1.6, the data in Figure 7 show the ratio consistently going beyond this, as high as 2.0. Thus, the calibration curve must be incorrect, probably due to errors in the HITRAN database. Other researchers10,11

have found errors in HTIRAN linestrengths, line positions, or broadening parameters, especially for lines that have not been widely investigated in the past, such as these. Nevertheless, all that is required to produce an accurate calibration curve is to carefully measure these spectroscopic constants for the lines in this part of the spectrum in a controlled environment, such as a heated cell at low pressure. An alternative approach is to construct an experimental calibration curve by making a series of measurements with the system at known temperatures by, for example, using a flat flame burner to produce a flame and a thermocouple as the calibration standard.

Temperatures in the test rig can be estimated from thermocouple measurements and model predictions. A type R thermocouple was placed in the combustion zone near the location of the TDLAS measurements, yielding temperatures of 1398 K and 1477 K for φ = 0.48 and 0.58, respectively. An analysis of radiation losses of the thermocouple suggests that the indicated temperature would be about 220 K lower than the gas temperature for φ = 0.48, and 250 K lower for φ= 0.58, resulting in gas temperatures of 1618 K and 1727 K for φ = 0.48 and 0.58, respectively. Temperatures were obtained for the same conditions from a steady state axisymmetric numerical model, using a commercially available code (CFDACE+ by CFD Research Corporation). The model predicted 1678 K and 1915 K for φ = 0.48 and 0.58, respectively. While the corrected thermocouple measurement for φ = 0.48 is in reasonably good agreement with the model prediction, the corrected thermocouple measurement for φ = 0.58 is about 200 K lower than the model prediction. The model does not include heat losses to the environment, which could account for this discrepancy.

Since the ratios of TDLAS absorbances go outside the range of the predicted calibration curve, it is not appropriate to use this curve to convert the raw data into temperatures. Nevertheless, to illustrate the problem with the existing database, the temperature that results from the measured absorbance ratio of 1.30 at φ= 0.48 using the calibration curve of Figure 4 is about 1890 K. This is obviously in error since it is well above the model result of 1678 K that should be an upper limit since does not include heat loss. There is no corresponding predicted temperature for the absorbance ratio of 1.63 observed at φ = 0.58 since the simulated spectrum predicts that this ratio would never occur.

6American Institute of Aeronautics and Astronautics

CONCLUSIONS

Preliminary testing has begun on a real-time in situ temperature sensor in a model gas turbine combustor. The single-diode laser sensor is simple, rugged, and suitable for measurements at high pressure. The results presented here demonstrate the ability of the sensor to capture transient combustion phenonmena, and suggest that the sensor might be useful in an active control application. A discrepancy between the measured and predicted absorption peak ratio was observed, suggesting that future work is needed to establish proper calibration of the sensor.

ACKNOWLEDGEMENT

This research was supported, in whole or in part, by DOE Grant No. DE-FG03-99ER82828 and such support does not constitute an endorsement by DOE of the views expressed in this article.

REFERENCES

1 T. P. Jenkins, P. A DeBarber, E. H. Scott, V. G. McDonell, T. N. and Demayo, “A Rugged Low-Cost Diode Laser Sensor for H2O and Temperature”, Spring Meeting of the Western States Section/The Combustion Institute, San Diego (2002).

2 T. P. Jenkins, P. A. DeBarber, and M. Oljaca, “A Rugged Low-Cost Diode Laser Sensor for H2O and Temperature Applied to a Spray Flame”, AIAA 2003-0585 (2003).

3 T. P. Jenkins, P. A. DeBarber, and M. Oljaca, “Diode Laser Sensor for H2O and Temperature Applied to Measurements in an Industrial Combustion Vapor Deposition Torch”, Joint Meeting of the US Sections/The Combustion Institute, Chicago (2003).

4 R. K. Hanson and P. K. Falcone, “Temperature Measurement Technique for High-Temperature Gases using a Tunable Diode Laser”, Applied Optics 17, pp. 2477-2480 (1978).

5 L. S. Rothman, C. P. Rinsland, A. Golman, S. T. Massie, D. P. Edwards, et al., “The HITRAN Molecular Spectroscopic Database and HAWKS (HITRAN Atmospheric Workstation): 1996 Edition”, Journal of Quantitative Spectroscopy and Radiative Transfer, Vol. 60, pp. 665-710 (1998).

6 R. R. Gamache, J.-M. Hartmann, and L Rosenmann, “Collisional Broadening of Water Vapor Lines-1. A Survey of Experimental Results”, Journal of Quantitative Spectroscopy and Radiative Transfer, Vol. 52, pp. 481-499 (1994).

7 E. C. Rea and R. K. Hanson, “Rapid Laser-Wavelength Modulation Spectroscopy used as a Fast Temperature Measurement Technique in Hydrocarbon Combustion”, Applied Optics, Vol. 27, pp. 4454-4464 (1988).

8 M. G. Allen, “Diode Laser Absorption Sensors for Gas Dynamic and Combustion Flows”, Measurement Science and Technology, Vol. 9, pp. 545-562 (1998).

9 H. Tennekes, and J. L. Lumley, A First Course in Turbulence, The MIT Press, Cambridge, Massachusetts, pp. 20-21 (1972).

10 V. Nagali, “Diode Laser Study of High-Pressure Water-Vapor Spectroscopy”, Ph. D. Thesis, Stanford University, Dept. of Mech. Engr., (1998).

11 R. M. Mihalcea, “CO and CO2 Measurements in Combustion Environments using External Cavity Diode Lasers”, Ph. D. Thesis, Stanford University, Dept. of Mech. Engr., (1999).