Design of a Multiple Beam Pyrometer forMeasurement of Temperature on Gas Turbine

BladesStefan Maurer

and Norbert LindleinUniversity of Erlangen-Nuremberg

Institute of Optics,Information and Photonics

(Max Planck Research Group)Erlangen, Germany

Email: mail@maurerstefan.de

Michael Willschand Thomas Bosselmann

Siemens AGErlangen, Germany

Andreas BaumgartTechnische Universitaet Ilmenau

Ilmenau, Germany

Abstract— In this paper, a new, non-invasive, stationary sensordesign for pyrometry measurements on several separate sensingspots on a gas turbine blade is presented. The sensor opticsseparates these spots and enables a spatially resolved temperatureacquisition. It deflects thermal radiation through a 10mm keyholeof the turbine wall into waveguides. Parameters of the opticalcomponents are optimized by ray-trace computer simulations,whereas experimental verification is done by a 4x scaled setup.Simulation and experimental results from the final design areshown. At present, a prototype is being constructed according tothe results of this work.

I. INTRODUCTION

Being aware of dwindling fossil resources, the productionof energy becomes more and more expensive. Neverthelessthe demand of energy increases steadily. One solution is tomake energy production more efficient. Focusing especiallyon gas turbines, the increases of their operating temperatureand therefore their efficiency is promising. However, themaximum operating temperature is limited by the robustnessof the material in the turbine, in particular of the materialof the turbine blades. These are the components stressed atmost. Consequently the very important question is: How hotbecome the turbine blades? Do they – especially the first rowblades – reach their operating limit? The maximum operatingtemperature of the turbine blades is well-known. It is muchmore interesting to what extent they really heat up while theyare in operation.Today it is already possible to measure the temperature at onepoint of the turbine blade by the means of pyrometry [1].Thermal radiation from the hot blade is collected and routedthrough a small ”keyhole” in the turbine wall. The intensityof the radiation is measured by a detector. Using Planck’sLaw, the temperature of the blade at the measured point canbe obtained. A disadvantage is that the sensor extends intothe inner turbine room during the measurement, as well asmost of the other available sensors do. Although few designsallow measurements without disturbances in the turbine room

Fig. 1. Side view of the design of the pyrometer. There are shown the turbineblade (1), measuring spots (2), inner (3) and outer turbine wall (4), window(5), prism (6), lens (7) and waveguides (8).

Fig. 2. Integration of the sensor, which is located in a long cylinder, in theturbine.

[2], information about the temperature distribution cannot beacquired yet by both presented designs.

There is just one further possibility described by [3], whichenables a spatially resolved temperature measurement on agas turbine blade: Thermal radiation of the complete bladeis collected via a dispersion prism and routed by a lensonto one single fiber waveguide. Spatially resolved infor-mation is obtained by using the fact that the deflection ofthe radiation is wavelength dependent and the coupled lightwith different wavelengths origins from different parts of theblade. Nevertheless, the concept exhibits two disadvantages: (i)

1-4244-2581-5/08/$20.00 ©2008 IEEE 744 IEEE SENSORS 2008 Conference

Fig. 3. Side view of the setup for the RAYTRACE software. The segmentsof the turbine blade, numbered from 1 to 7, are set as light source.

No simultaneous temperature measurement on different bladesegments, and (ii), loss of a big part of the beam intensityby the use of color filters. The aim of our work is to facethe above mentioned disadvantages and to develop a conceptthat makes the instant measurement of temperature gradientspossible. An optical setup is developed which enables to routethermal radiation from several discrete points on the turbineblade through a keyhole in the turbine wall of approx. 10 mmdiameter. The radiation from the 800◦C to 1300◦C hot blade isthen routed into several appropriate fiber waveguides. Finally itis passed on to detectors to obtain temperature information. Inorder to avoid disturbances of the gas flow within the turbine,no parts of the sensor extend into the inner turbine room.Furthermore, coupled into waveguides of thermal radiationwhich does not come from the measuring points on the blade isprevented. Figure 1 shows qualitatively the pyrometer design,fulfilling the conditions mentioned above, in a side view.Figure 2 illustrates the integration within the turbine.

II. SIMULATION AND EXPERIMENT

A. Simulation

After extensive, preliminary investigations with the ray-tracing simulation software RAYTRACE [5], the design whichcan solve the task best of all (see Fig. 1) was selected. Itconsists of a highly temperature resistant sapphire windowwhich seals the cylinder against the turbine room. A prismmade of fused silica deflects the rays in such a way thatthey get parallel to the cylinder axis. Finally, a lens made ofsapphire images the measurement spots onto the end of sevenwaveguides. The material of the optical components can onlybe sapphire or fused silica, because only those two glassesresist temperatures up to 2000◦C and 1000◦C respectively.Since the temperature in the turbine is at about 1100◦C, thewindow has to be made of sapphire. The lens is also made ofsapphire, because it has a greater refractive index (nSapphire =1.75, nFusedSilica = 1.46) and secures that the curvature andtherefore aberrations of the lens are kept small. As on theother hand, sapphire is birefringent (no

Sapphire − naoSapphire =

0.00787), the prism is made of fused silica in order to keep theglobal drift between ordinary and extraordinary rays as smallas possible. The radiation finally couples into the waveguideswith 62.5µm core diameter and numerical aperture of 0.22or 0.12 and is routed to detectors. Using RAYTRACE, thisdesign then was examined more deeply. Different segmentsof the blades were set as independent light sources. Thenraytracing was performed with the optical elements window,prism and lens. For all simulations the used wavelength was

Fig. 4. Experimental setup of the sensor. Dimensions are 4x scaled up dueto better handling.

1m. Fig. 3 shows a side view of the simulated configuration.The following issues were examined within this work: (i)Optimum waveguide positions and coupling efficiencies, (ii)Positioning tolerances of the waveguides, (iii) Fault couplings,(iv) coupling power and (v) Origin of coupled radiation on theturbine blade. Results are shown in section III.

B. Experiment

After simulation was done, the results were verified by anexperimental setup. As dimensions of the optical elementsare very small, the design was built up 4x scaled. Fig. 4shows the scheme of the experimental setup. The prism wasfixed on an immovable, but rotary stage (around the x-axis).The turbine blade, the lens and the holding device for thewaveguides were fixed on translation stages. Additionally, thewaveguides could also be rotated around the x-axis by agoniometer. At first the waveguides were set as visible lightsources to roughly examine the path of rays. This means thatthe problem was tested inversely. Consequently, results will beslightly different for the not inverse but direct measurement.Afterwards, different segments of the blade were set as visiblelight source. This was achieved by illuminating a small,properly dimensioned ceramic disk with a bright halogen bulb.By connecting a photometer to one or several waveguides, thepositions with the best coupled conditions could be found andcompared to simulation results.

III. RESULTS AND DISCUSSION

A. Simulation

1) Optimum waveguide positions and coupling efficiencies:In order to guarantee a signal with enough power as wellas a clear separation of the signals from the different bladesegments, it has to be made sure that the waveguides arelocated in their optimum position. The coupling efficiency ofthermal radiation from the appropriate blade segment has tobe maximum, the coupling efficiency of neighboring segmentsshould be zero. Using the simulation setup in RAYTRACE de-scribed above, the foci of the different turbine blade segments,i.e. the optimum position for the waveguides were calculated.Afterwards, the so called ”Corrected Fiber Efficiency” (CFE)was calculated for every blade segment at every waveguideposition. The CFE is defined as the ratio between the numberof coupled rays and the number of focused rays behind thelens (see Fig. 3). The optimum waveguide positions are shownin Fig. 8. It can be seen easily that the foci of the differentsegments lie, due to aberration errors of the third order, on aPetzval shell [4]. Fig. 5 depicts the CFE’s in every waveguide.

745

Fig. 5. Corrected Fiber Efficiency (CFE) of all waveguides for all turbineblade segments.

TABLE IACCEPTABLE TOLERANCES (±) FOR POSITIONING THE OPTICAL

ELEMENTS (BIREFRINGENCE EFFECTS INCLUDED).

Translation Rotation

x y z α β γ

Window - - - 1◦ 5◦ -

Prism - 100µm 500µm 0.05◦ 0.5◦ 0.5◦

Lens 50µm 50µm 200µm 1◦ 1◦ -

Waveguide 100µm 10µm 450µm 1◦ 1◦ -

The diagram confirms that there is coupled only for radiationfrom the appropriate, but not from adjacent segments intothe waveguide. In the case that the CFE’s from neighboringsegments were not negligible, i.e. imaging didn’t work thatwell, it would have been necessary to replace the lens by anasphere. This element would feature much less aberrations.

2) Positioning tolerances of the waveguides: Another im-portant question is: How precise have the optical elements(waveguides, prism, lens and window) to be positioned? Tol-erances for each degree of freedom were defined by applyingsmall shifts or rotations to the elements, and monitoring thedecline of the CFEs. Concretely, translations in x, y and zdirection, as well as rotations around these axes, described bythe angles α, β, and γ were examined. The maximum toleranceis reached when the CFE into the appropriate waveguidedecreases by 20% or when the CFE into a neighboringwaveguide increases up to 10% of the original signal. Takinginto account that the ordinary and the extraordinary beam splitup by birefringence up to 15µm, the resulting allowances areshown in Tab. I. The positioning of the lens, prism and windowaccording to the mentioned tolerances is easily achievable.On the other side, the realization of the positioning of thewaveguides guaranteeing especially the low y−tolerance ofjust 10µm is a challenge. Nevertheless, a solution for thisproblem is in progress. As shifts of the waveguides in y-direction are extremely sensitive, the CFE of the first threewaveguides, which lie together most closely, were examinedmore extensively. Moving the waveguide in steps of 20µm,the CFEs from both the appropriate and the adjacent segments

Fig. 6. Corrected Fiber Efficiency (CFE) for waveguide number 2 of radiationfrom the appropriate and the direct neighboring turbine blade segments asfunction of the y-position of the waveguide. The CFE is given relative to themaximum value in the appropriate segment. Birefringence is not consideredhere.

TABLE IIRADIANT FLUX φc , FRACTION Fw OF THIS FLUX, WHICH COUPLES INTO

THE WAVEGUIDES, AND THE APPROPRIATE ABSOLUTE VALUE φw .

Segment 1 2 3 4 5 6 7

φc / µW 248 297 360 443 555 709 914

Fw / 10−3 2.76 2.17 1.74 1.38 1.24 1.44 0.91

φw / µW 0.69 0.64 0.62 0.61 0.69 1.02 0.83

were logged and are depicted exemplarily for waveguide No.2 in Fig. 6. Birefringence effects are not considered here.

3) Coupling power: Until now, no absolute values for cou-pled fluxes were given. This was done in a further calculationaccepting the following assumptions: Temperature of turbineblade: Tb = 1000◦C, sensitive wavelength range for detector:800nm < λ < 1800nm. Integrating Planck’s radiation law,the radiation flux φb, emitted by a turbine blade segment withan area A = 4mm x 4mm, was calculated:

φb = A ·∫ 1800nm

800nm

2πc2h

λ5(e

chλkTb − 1

)dλ = 35, 5mW (1)

This power is distributed within the half space above the blade.Just a small fraction, given by the ratio of the spherical’scap’s area (2πr2 (1 − cosχ)) and the hemisphere’s area hitsthe cylinder. By weighting the result according to Lambert’slaw, we get the radiant flux φc, entering the cylinder (see Tab.II). The fraction Fw of φc which couples into the waveguideswas evaluated by RAYTRACE. Finally, the absolute radiantflux, coupled into the waveguides, is given as φw in Tab. II.φw lies between 0.7 and 1µW and is detectable by the foreseendetector. If by any reason the coupled radiant flux was too low,the core diameter of the waveguides could be increased. But,as the waveguides are already positioned very close together, arepositioning of the waveguides could be necessary then. Thiswould be linked with a decrease of the spatial resolution.

4) Origin of coupled radiation on the turbine blade:The key feature of the pyrometer is its ability to measure

746

Fig. 7. The CFE into the appropriate waveguide is shown for every turbineblade segment from segment 1 at z = −42 to segment 7 at z = −18 asfunction of the emission position.

temperature spatially resolved. Thus, it was examined whereexactly the coupled radiation comes from on the turbine blade.This was done by shifting a small light source on a gridof 16x16 sub segments on every segment of the blade andlogging the CFE (see Fig. 7). The results show that the coupledradiation origins from a quite small area of 1mm x 1mm.Therefore the temperature on the blade is not measured byaveraging over a big area, but on a relatively sharp spot. Thusexact temperature information is obtainable.

5) Fault couplings: Fault couplings occur when light,which was not emitted from the turbine blade but for examplefrom the cylinder walls, couples into waveguides. At first, theorigin of the fault radiation was located. Assuming furtherthat the temperature of the turbine blade is Tb = 1000◦C,the fraction of the coupled radiation, which origins from thefour most intense areas together, i.e. the fraction with faultradiation was calculated. The temperature of the cylinder, Tc

was assumed to be 900, 800, .. . or 400◦C. Depending onTc, the fraction of fault coupled lies in a range from 0.01%to 5%. Including results of a temperature simulation [6], areasonable value for Tc is 800◦C. The fault fraction then liesat about 2%. This would be acceptable, as it would result inan absolute temperature error of only 4◦C.

B. Experiment

Each optical element was positioned according to the sim-ulation results. Then, on the waveguide mount, one singlewaveguide, coupled to a laser light source, was positioned.Its position was adapted in such a way that the image ofthe waveguide end was centered sharply on each of theseven turbine blade segments successively. The waveguidepositions (in y- and z-direction) were noted and are depictedin Fig. 8. This procedure was done for both a waveguide withnumerical aperture NA = 0.22 and one with NA = 0.12. Thecore diameter was 62.5µm. In the same Fig. 8 the optimumwaveguide positions according to the simulation results are

Fig. 8. The diagram displays the y− and z− position of each of theseven waveguides according to simulation results and experimental results.The simulated positions are shown with tolerance bars, which are 4x scaleddue to the experimental setup.

presented as well. Within the tolerance range of 40 µm theexperimental y− positions of the NA = 0.12 waveguide arein an excellent agreement with the simulated ones. Also thez− positions are in good agreement, although the experimentalpositions of waveguide 1 and 2 are slightly out of the tolerancerange of 1.8mm. This is due to the big focus, which was noteasy to identify. As this procedure was even more difficult forthe NA = 0.22 waveguide, the corresponding results do notfit as well as those of the proceeding one.The very motivating simulation results as well as the experi-ment, confirming these results, indicate that the prototype inconstruction will work successfully.

IV. CONCLUSION

The prototype construction, adapted to small industrialturbines, is currently in progress. First test runs will soonbe done. As the complete design of the multiple beam py-rometer is easily scalable, it will subsequently be adaptedto bigger turbines with powers of up to 200MW . With theknowledge about the exact temperature distribution on theirturbine blades, it will be possible to adapt the turbine designin such a way that an increase of the operating temperatureand eventually of the efficiency can be achieved.

REFERENCES

[1] R.A. Rooth, Blade temperature monitoring using boroscope holes, FirstInternational Conference on Gas Turbine Instrumentation, 28th & 29thSeptember 2004

[2] R.A. Rooth, Dual wavelength temperature monitoring of TBC coatedAlstom 13E2 turbine blades, Proceedings of ASME/IGTI Turbo Expo2003: Power of Land, Sea & Air, June 16-19, 2003, Atlanta, Georgia,USA

[3] US Patent 4240706, Optical Probe, Joseph Douglas, 23.12.1980, Rolls-Royce Limited, London, England

[4] E. Hecht, Optics, 3. ed., Reading, Mass., Addison-Wesley 1998[5] N. Lindlein, F. Simon, M. Lano, O. Stolz, A. Mitnacht, RAYTRACE

Version 0.9(15), University of Erlangen-Nuremberg, Copyright 2006[6] A. Baumgart, Vielstrahlpyrometer fuer Temperaturmessungen an Gastur-

binenschaufeln, Diploma Thesis, Technische Universitaet Ilmenau, 2008

747