Infrared thermography to detect residualceramic in gas turbine blades1 Dipartimento di Ingegneria Aerospaziale (DIAS), Università di Napoli Federico II, Via Claudio, 21,
80125 Napoli, Italy2 Europea Microfusioni Aerospaziali S.p.A., Zona Industriale ASI, 83040 Morra De Sanctis (AV), Italy
ABSTRACT A serious problem in the production of gas turbineblades is the detection of residual ceramic cores inside the cool-ing passages; in fact, the presence of even small ceramic piecesaffects turbine performance and may cause difficulties in suc-cessive manufacturing. Therefore, it is important to have a non-destructive technique that must be capable of detecting tinyceramic fragments in a fast and easy way. In this perspective,the suitability of infrared thermography was investigated withincooperation between the University of Naples and the Euro-pea Microfusioni Aerospaziali S.p.A. (EMA). Several blades ofthree different types were inspected revealing that in many casesinfrared thermography can discover small ceramic fragmentswhich were missed by X-ray inspection. In addition, infraredthermography allows gaining of information about other typesof anomalies (e.g., surface defects) during the same testing step(by eventually changing the test parameters) and then savingtime and money. The obtained results look promising in view ofintroducing infrared thermography among industrial instrumen-tation as an alternative to, or integrated with, the most currentlyutilized non-destructive techniques.
Gas turbines are employed in many industrial ap-plications for power generation (such as to drive aircrafts,trains, ships, electrical generators), and their efficiency in-creases with increasing combustion temperature. Protectionagainst thermal effects must be assured through effectivecooling systems which include air channels inside the blade.In fact, the durability of engine turbine components dependson several factors: mainly the material alloy characteristics,the efficiency of cooling systems and the defect-free initialconditions.
Blades are generally made of special Ni-based or Co-based superalloys, realized through lost wax casting pro-cesses, and include ribbed cooling passages, which are ob-tained with ceramic cores. The ceramic is normally removedthrough chemical etching. Sometimes, the ceramic cores arenot completely eliminated due to the complex geometry ofthe cooling passages; thus, the residual ceramic compromises
the in service performance and may complicate successivemanufacturing. In addition, flaws in the material alloy (de-bris, voids, etc.), surface defects, geometrical variations anddisplacement of cores may also occur during the differentproduction steps. These anomalies remaining undetected maylead to premature failures [1]. Therefore, every blade requiresaccurate inspection with effective non-destructive evaluation(NDE) techniques to ensure integrity and safe performance.
Many different NDE techniques (radiography, ultra-sounds, visual inspection and penetrant liquids) are used atEMA to gain information about the different types of anoma-lies. Dye penetrant inspection [2] is perhaps the oldest NDEmethod. It is based on the visualization under UV illumina-tion of surface discontinuities, which were previously filledwith a low-surface tension fluid. This method seems very sim-ple, but it includes surface pre-treatment and it is affectedby several factors (mainly the type of material, the surfacefinish, the flaw characteristics, the adopted procedure, and hu-man factors among others), and it is restricted only to surfacedefects visualization. Ultrasound is the most popular non-destructive technique [3, 4]. It is based on the propagation ofhigh-frequency sound waves inside solid materials. A localvariation of material characteristics (acoustical impedance)affects the energy transmission; thus, the amount of energythat arrives at the receiver probe gives information about thematerial characteristics (thickness, density, stiffness, poros-ity). At EMA it is mainly used for the blades dimensionalpurposes, while it is not very effective for detection of residualceramic.
Radiographic (X-ray) methods [5] supply a black/whitepicture that shows both external and internal features of theblade. Anomalies become visible as local change of luminousintensity because radiation decays in different ways throughdifferent materials and thicknesses (i.e., the light/shade ef-fect strongly depends on the density and thickness of thematerial). Thus, it is easy to distinguish an empty duct froma ceramic filled one, while it is difficult to detect very thin ce-ramic fragments attached to the channel wall, often betweenribs. Due to the complex blade geometry, more than one angleof view is usually necessary and this lengthens the completetesting also because of the time required for film develop-ing. In principle, the process could be shortened with the useof computed radiography. This offers the possibility to ex-amine different thicknesses by simply adjusting the imagebrightness, and reduces the developing times needed for filmX-rays; in addition, there is the advantage offered by digital
686 Applied Physics A – Materials Science & Processing
images to be enhanced and analysed via software. Unfortu-nately, this methodology has still a quite low resolution thatmakes it unsuitable for the ceramic detection.
Infrared thermography (IRT), although it has proved ef-fective in many NDE applications [6–11], has received lit-tle attention in the inspection of turbine blades. In fact, itseems that it was only investigated by the German companySiemens [12], with the aim of measuring the wall thickness ofcertain types of blades and it was considered in fatigue tests ofcomposite blades [13].
The objective of the present paper was to look for the ex-ploitation of infrared thermography as an alternative to, orintegrated with, the techniques currently utilized for the in-spection of gas turbine blades. The main demand was to detectresidual ceramic inside the cooling passages of blades.
2 Non destructive evaluation with infraredthermography
Infrared thermography is a two-dimensional, non-contact technique, which can be usefully exploited in a vastvariety of industrial applications as well as research fields [11].
The analysis of subsurface features in solid objects by IRTgenerally requires heat energy to be transferred to the objectin the active mode and monitoring the material response tosuch stimulation in the transient heating (or cooling) stage.The detection of a slag inclusion in a certain material withIRT depends on many factors, such as the relative thermalcharacteristics between the material and the insert as well asthe geometry and depth of the insert. Basically, two differentapproaches are possible for nondestructive evaluation: pulsethermography (PT) and lock-in thermography (LT).
Starting with the object at constant temperature, in PT,the thermal energy is generally supplied at a constant rate fora given period of time. Lock-in thermography [11, 14] can beperformed in different modes depending on the test set-up. Inthe basic optical lock-in thermography (OLT), periodic heat,generated by halogen lamps, is delivered to the object surface.
In both PT and LT, the thermal energy propagates underthe surface by conduction while the infrared camera monitorstemperature variations over the viewed surface. For a uniformsurface heating, the temperature distribution is uniform in thecase of a homogeneous material. The presence of a defect,at a certain depth, interferes with the heat flow causing localsurface temperature variations.
In fact, thermal waves propagate inside the material andundergo reflections at the boundaries. The temperature modu-lation at the surface is modified by the thermal waves comingback from the inside of the component and an interferencepattern is produced that is captured by the infrared camera.Calculation of the phase and amplitude of these waves supplyinformation on hindered defects.
The depth range, for the amplitude image in OLT, is givenby the thermal diffusion length µ which is calculated from thematerial thermal diffusivity coefficient α = k/�c with k thethermal conductivity, � the density and c the specific heat andthe wave frequency f = ω/2π:
µ =√
2k
ω�c, (1)
while the maximum depth p, which can be reached for thephase image, corresponds to 1.8µ [15–18]. So, the mate-rial thickness, which can be inspected, depends on the waveperiod (the longer the period, the deeper the penetration) andon the material thermal diffusivity.
Results may be presented in terms of either phase, oramplitude images. In general, the phase image is preferablebecause it is much less sensitive to reflections from the back-ground and to non uniform heating. In the phase image, a localvariation of colour indicates a local variation of phase angleϕ and, in turn, a local variation of the properties of the un-derneath material [19]. The acquired images may be storedfor successive analysis to obtain information about size, depthand nature of defects.
3 Experimental tests
Blades may include different types of defects:mainly slag inclusions, local wall thinning (which may becaused by a ceramic core displacement occurring duringcasting), material modifications, surface cracks and cavities,ceramic residual inside the cooling channels. The main aim ofthe present investigation is to detect the presence of residualceramic with IR thermography.
Three different kinds of blades were consideredwhich are produced at the EMA S.p.A. (Italy) for aircraft tur-bine engines. The three types of blades with their ceramiccores are shown in Fig. 1a–c. As can be seen, the first blade,which is simply named BA and is shown in Fig. 1a, includestwo almost straight ducts, whose positions are indicated on theblade by the black vertical lines, the corresponding ceramiccores being shown on the right side. The other two blades,which are respectively named BB and BC, include continuousserpentine cooling passages.
With regard to the presence of residual ceramic, two morefrequent cases may arise which are sketched in Fig. 2.1. The duct is partially obstructed (Fig. 2a). Ceramic resides
mainly in the tip because of poor chemical etching. Some-times, the contrary may happen with ceramic accumulatednearby the outlet.
2. Small ceramic pieces remain attached to the channel lat-eral walls, especially between ribs or at the channel tip(Fig. 2b).
3.2 Problem formulation
With reference to Fig. 2 and assuming the ductwidth and the blade thicknesses are constant along the x di-
MEOLA et al. Infrared thermography to detect residual ceramic in gas turbine blades 687
FIGURE 1 Photographs of the three inspected blade types with their ce-ramic cores; (a) blade type BA, (b) blade type BB, (c) blade type BC
FIGURE 2 Sketch of residual ceramic inside cooling passages; (a) chan-nel with partial obstruction, (b) channel with presence of small fragmentsattached to the wall
rection, the first case (Fig. 2a) can be reduced to a problemof a simple foreign inclusion, which is either air for the openduct, or ceramic for the obstructed one. The ceramic searchcould be basically addressed with either PT, or LT.
The first thing to be considered is that the blade surface isalmost bright and so affected by reflections mainly from the
heating lamp and from the surrounding. Thus, the use of PT inthe reflection mode (i.e., heating and viewing with the infraredcamera from the same side)is not recommended .
With PT in transmission (i.e., heating and viewing fromopposite sides), a relatively large residual ceramic (i.e., thechannel partial obstruction sketched in Fig. 2a) can be quiteeasily detected while small and thin ceramic pieces (Fig. 2b)are relatively difficult to discriminate.
In fact, the first case (Fig. 2a) reduces the problem to a sim-ple inclusion which becomes visible through the perturbation(local temperature difference) it induces to the surface tem-perature distribution. So, as sketched in Fig. 3a, the hollowpart (air filled) and the ceramic filled part respectively cause,on the surface opposite to the thermally stimulated one (bot-tom in figure), the temperature differences ∆Ta = Tfa −Ta and∆Tc = Tfa − Tc with respect to the full alloy. Subscripts a, faand c respectively stand for air, full alloy and ceramic. Thequantities ∆Ta and ∆Tc are influenced by the difference be-tween the thermal characteristics of the blade alloy and thoseof air and ceramic, respectively.
The second case (Fig. 2b), involves a combined effect ofair, ceramic and metal alloy thermal characteristics. The tem-perature distribution on the surface opposite to the heating issketched in Fig. 3b. This temperature will attain a maximumTfa value for the full alloy, a minimum Ta value for the airfilled channel and an intermediate Tc value where residual ce-ramic is present. In Fig. 3b, two probable positions, (1) and(2), of a small ceramic fragment are considered; in both cases,the ceramic is in contact with the metal alloy from one sideand immersed in air by the other sides. In position (1), the ce-
FIGURE 3 Sketch of temperature distribution on the side opposite to theheat source; (a) channel partially obstructed, (b) small fragments
688 Applied Physics A – Materials Science & Processing
ramic piece is located at the depth pc1, which is practicallyequal to the metal alloy thickness in the channel sal, and ischaracterized by a thermal contrast equal to the difference be-tween the temperature of ceramic and that of the surroundingair ∆Tc1 = Tc1 − Ta. For position (2), the ceramic piece hasa thermal contrast again equal to the temperature differencebetween ceramic and air ∆Tc2 = Tc2 − Ta, but ∆Tc2 < ∆Tc1
because of two main reasons. One reason is that the ceramicpiece is now located at depth pc2
∼= sal + sa which is greaterthan pc1. The other reason is that the ceramic piece is nowburied under a two-layer structure composed of a metal alloylayer of thickness sal plus a layer of air of thickness ∼= sa (so,not only under the metal layer as in the first case).
3.3 Test procedure
From the discussion of the previous Sect. 3.2 andfrom preliminary tests, it follows that small ceramic piecesmay be detected with PT in transmission if a strong heat pulseis applied. However, some problems, which complicate test-ing, may arise:
• In the case of PT in transmission with optical thermal stim-ulation, light may enter the optics and blind the detector.Then, protection against unwanted radiation is required.
• Thin ceramic, if located in position (2) (Fig. 3b), causesweak thermal contrast which may be confused with theinstrument noise.
• Reflections from the surrounding may affect data interpre-tation.
Most of these drawbacks can be overcome with the use oflock-in thermography. In fact, the phase images obtained withLT are practically not affected by reflections, non uniformheating and local variation of emissivity so that tests can beperformed also in the reflection mode.
All the types of blades, described in the previous Sect. 3.1,were tested with lock-in thermography (Fig. 4). Thermal stim-ulation was performed with halogen lamps and, in particular,two lamps were used:
• A large one of 1000 W for simultaneously heating of a cer-tain number of blades; this may be useful for a fast dis-crimination of blades with obstructed channels.
• A smaller one of 500 W positioned close to the blade atabout 10 cm; this, coupled with close-up view, being moreeffective to detect small ceramic fragments.
It is worth noting that the smallest blade (Fig. 1c) is about3 cm long and 2 cm wide with a serpentine cooling passagehaving a section of about 1×3 mm2. Thus, the problem wasto detect very thin ceramic fragments (e.g., 0.1 mm) under1 mm of the metal curved surface. This is a difficult task toachieve and required searching for optimal test conditions. Aneffective test configuration for the smaller blade type BC wasdevised to include, as sketched in Fig. 4, heating with a halo-gen lamp of 500 W positioned close to the blade (only 10 cmaway), with sinusoidal modulation, close-up view lens anda frame rate of 60 Hz. Then, preliminary tests were carried outto find out the blind frequency [10]. For all the blades a a blindfrequency of about 0.2 Hz has been found.
The blades of type BA (Fig. 1a), which include straightcooling channels, were also inspected with pulse thermogra-phy by thermally stimulating them by injecting water vapour(steam) in the channels (Fig. 5). This procedure is calledPTJV. The advantage of using steam is that vapours in con-tact with a cold wall have a very high convective heat transfercoefficient so that the wall surface contacting the vapourgets suddenly very close to the vapour condensing tem-perature, thus determining a strong heat pulse. Furthermore,the first drawback listed at the beginning of this section isovercome.
In PTJV the vapour fills the channel and induces heattransfer through the wall thickness sal, as sketched in Fig. 6(arrows in figure). Of course, the heat flows in any directionstowards the channel outside; the sketch in Fig. 6 indicates onlythe heat flux directed towards the surface viewed by the in-frared camera. Therefore, the presence of a ceramic fragmentattached to the channel wall can be detected through the per-turbation it causes to the temperature distribution Tfa over theblade surface; such perturbation is accounted by the tempera-ture difference ∆Tc = Tfa − Tc.
However, the proposed PTJV method is effective only if itis performed in a fast way during the transient heating phase;
FIGURE 5 Setup for PTJV tests
MEOLA et al. Infrared thermography to detect residual ceramic in gas turbine blades 689
FIGURE 6 Sketch of temperature distribution on the surface viewed by thecamera during PTJV tests
this means that the channel must be quickly filled with vapourand thermal images acquired in real-time at high frequencyduring transient heating.
All the images shown in this section are taken with theblade viewed from the leeward (convex) side and having thetrailing edge on the right (e.g., the BA blade in Fig. 1a).
Phase images of two BC blades obtained with OLT areshown in Fig. 7. It is possible to see phase angle discontinu-ities (inside the rectangles) in the first channel on the right siderespectively in the tip of blade 1 (Fig. 7a) and at about halfway in blade 2 (Fig. 7b). Such discontinuities are caused bythe presence of residual ceramic there.
Figure 8 shows the OLT phase image of a BA blade typein which channel 2 is free from ceramic, with the phase anglealong the channel pathway being almost constant. The pres-ence of the small white area, pointed out by an arrow, indicatesa local variation of wall thickness there, or better, the probablepresence of a cavity.
Phase and thermal images of a BA blade type, in whichthere is a presence of small ceramic fragments, are shownin Fig. 9. More specifically, in the blade total view (OLT phaseimage of Fig. 9a) it is possible to see local phase variationsclose to the tip and to the root of channel 2. Another OLTphase image taken with a close-up view lens is shown inFig. 9c; it is now possible to clearly distinguish the presenceof residual ceramic. From a comparison between Fig. 9a andc, it is possible to perceive how weak the thermal contrast as-sociated with small fragments is.
To examine the variations of phase angle observed in thechannel tip, the blade was tested again with pulse thermogra-phy and vapour injection (the vapour was injected as close aspossible to the tip region). A PTJV thermal image, acquired12 s after injection started, is reported in Fig. 9b. As can beseen, the tip region appears non uniformly warm, the colderzone indicating the presence of a ceramic fragment.
FIGURE 7 Phase images of BC blades type, f = 0.65 Hz; (a) blade BC-1,(b) blade BC-2
FIGURE 8 Phase image of BA blade type with open channel
To understand how to identify a small ceramic fragmentwhich resides in the channel tip, one may compare the twoPTJV images of Figs. 9b to 10. In the first image (Fig. 9b) it ispossible to see a colder area near to the channel tip; this is be-cause the ceramic fragment located there, hinders the vapourflow and prevents warming up of that region. Conversely, the
690 Applied Physics A – Materials Science & Processing
FIGURE 9 BA blade type with residual ceramic; (a) phase image – fullview, (b) thermal image with vapour injection, (c) phase image – close-upview
FIGURE 10 BA blade type with open channel; thermal image with vapourinjection
vapour completely fills the free channel of Fig. 10 and, asshown by the thermal image, the tip region heats up uniformly.
The OLT phase images of BB blades type are reportedin Fig. 11. The first image (Fig. 11a) shows phase angle dis-continuities (indicated by arrows) along the channel on thetrailing edge, which indicate there presence of residual ce-ramic. More specifically, two ceramic pieces, which partiallyobstruct the channel are located towards the tip while a thin-ner fragment is present near the root. The second image showsa blade with ceramic-free channels; however, a lighter zone onthe tip (pointed out by an arrow) indicates a local thinning ofthe wall.
5 Comparison with radiographic inspection
The results obtained with infrared thermographyhave been compared with the findings of X-rays examinationin Table 1. The comparison mainly regards the presence of re-sidual ceramic, and in the table details about the location of
FIGURE 11 Phase images of blades type BB; (a) with residual ceramic,(b) with ceramic-free channels
ceramic pieces (such as tip, or root, of the channel), the chan-nel number, when channels are numbered as for blades BAand BC, or the channel position in the blade (trailing edge TE,leading edge LE) are reported.
As can be seen, there is a general agreement betweenresults obtained with the two techniques. However, more re-sidual ceramic fragments seem to be detected by IRT. In-deed, these additionally discovered ceramic fragments arevery small and/or very thin and are difficult to detect with bothtechniques because of their weak associated contrast. How-ever, the thermographic images are stored in a computer andcan be retouched via-software to enhance the contrast; thus,also small local variations of phase angle can be appraised tomake the critically small ceramic fragments visible.
Vice versa, with X-rays a black/white picture of the bladeinterior is produced on a photo-sensitized film where the darkand white zones are the result of the different X-ray decay,which is caused by the variation in material density and thick-ness. In the case of blades, the absorption of radiation is
MEOLA et al. Infrared thermography to detect residual ceramic in gas turbine blades 691
Blade X-rays Infrared thermography
type
BA-1 Residual ceramic Channel 2 obstructed in the upper halfin tip channel 2
BA-3 Residual ceramic Residual ceramic in tip channel 2in tip channel 2 Local wall thinning in tip channel 2
BA-4 Residual ceramic Residual ceramic in tip channel 2in tip channel 2 Local wall thinning in tip channel 2
BA-5 Open channels Open channels
BA-6 Open channels Open channels
BA-7 Residual ceramic Residual ceramic in root channel 2in root channel 2 At different positions along channel 2
BA-8 Residual ceramic Residual ceramic in tip channel 2in tip channel 2
BA-9 Residual ceramic Residual ceramic in rootin root channel 2 and in tip channel 2
BA-10 Residual ceramic Residual ceramic in tip channel 1in tip channel 1
BB-1 Residual ceramic Residual ceramic along the channel in TEalong the channel Local anomaly in the blade central partin TE
BB-2 Residual ceramic Residual ceramic in TE tip and rootin TE tip Local anomaly in the blade central part
BB-3 Open channels Open channels, local wall thinning in TE tip
BC-1 Residual ceramic Residual ceramic in tip and in root left sidein root channel 1 channel 1 and on the root right side
of channel 2Variation of wall thickness in channel 1
BC-2 Residual ceramic Residual ceramic towards tip in channel 1in tip channel 1
BC-3 Residual ceramic Residual ceramic in root channel 2in root channel 2 Small ceramic fragment in root of channel 1
Residual ceramic in tip right side of channel 1
BC-4 Residual ceramic Residual ceramic at half way in channel 2at half way Residual ceramic in the horizontal ductin channel 2 between channels 2 and 3
TABLE 1 Data comparison X-rays – IRT
greater in the metal alloy, it decreases in ceramic and jumpsdown in air filled open channels. Therefore, open channelswill appear darker on the radiograph while shadows indi-cate the presence of residual ceramic. Discrimination betweenmetal alloy and residual ceramic is possible through localvariation of luminous and shade effects and is demanding onthe operator’s ocular acuity, as a thin piece of ceramic causesa very weak shadow effect which is difficult to resolve due tothe human eye’s physiological limits.
As a final point, it is worth noting that, in some cases,IRT has evidenced other types of anomalies, which could bea local wall tinning, or porosity, even if during the tests, car-ried out with the primary aim to search for residual ceramic,neither the test procedure nor the test parameters, were chosenaccordingly.
6 Conclusions and recommendations
The obtained results show that infrared thermog-raphy is capable of detecting residual ceramic inside the
cooling passages of different types of turbine blades. In par-ticular, it has been found that, by using lock-in thermog-raphy with intensive heating of blades and a fast samplingrate at 60 Hz, it is also possible to identify very small ce-ramic fragments which may be missed through radiographiccontrol. Small fragments located in the tip of straight ducts(e.g., BA blade type) can be discovered also with the PTJVtechnique.
For industrial purposes, it is suggested to use lock-in ther-mography in two steps.
• First – Search for large channels obstruction by viewingcontemporaneously a certain number of blades; this al-lows for rapid discarding of flawed blades.
• Second – Control unflawed (so from first analysis) bladesone at a time to assure they are free of residual ceramic.Due to curvature and cooling passages allocation, twoviews, for leading and trailing edges, should be consid-ered.
In addition, to detect small and thin ceramic pieces, tests mustbe performed with strong thermal stimulation and fast sam-pling rate of 60 Hz. Of course, an infrared camera of sensitiv-ity similar to the ThermaCam SC3000, or higher, should beused.
It is worth noting that, during close up views of a sin-gle blade it is possible to detect other types of anomalies,not only the residual ceramic, by simply varying the heat-ing stimulation frequency. This is a great advantage in termsof overall testing time as well as techniques and operatorsinvolved.
2 B. Larson, Report of the U.S. Department of Transportation FederalAviation Administration, Washington, D.C. (2002)
3 J. Blitz, G. Simpson, Ultrasonic Methods of Nondestructive Testing(Chapman & Hall, London, 1996)
4 J. Krautkramer, H. Krautkramer, Ultrasonic Testing of Materials, 4thedn. (Springer, Berlin, 1990)
5 J. Pitkänen, T. Hakkarainen, H. Jeskanen, P. Kuusinen, K. Lahdenperä,P. Särkiniemi, M. Kemppainen, M. Pihkakoski, 15th WCNDT 2000NDT.net
6 R.L. Thomas, J.J. Pouch, Y.H. Wong, L.D. Favro, P.K. Kuo, A. Rosen-cwaig, J. Appl. Phys. 51, 1152 (1980)
7 E. Grinzato, P.G. Bison, S. Marinetti, V. Vavilov, Res. Nondestr. Eval. 5,257 (1994)
8 D. Wu, T. Zweschper, A. Salerno, G. Busse, NDT.net 3 (1998)9 X. Maldague, Trends in Optical Nondestructive Testing (Pramond Ras-
togi, http://www.ndt.net/v06n04.htm, 2000)10 C. Meola, G.M. Carlomagno, G. Giorleo, J. Adhesion Sci. Technol. 18,
617 (2004)11 C. Meola, G.M. Carlomagno, Meas. Sci. Technol. 15, 27 (2004)12 M. Goldammer, W. Heinrich, Proc. ECNDT 2006, 11 (2006)13 A.G. Beattie, M. Rumsey, AIAA 99-0046 (1999)14 A. Dillenz, T. Zweschper, G. Riegert, G. Busse, Rev. Sci. Instrum. 74,
417 (2003)15 A. Letho, J. Jaarinen, T. Tiusanen, M. Jokinen, M. Luukkala, Electron.
