Engineering Failure Analysis 18 (2011) 386393

Contents lists available at ScienceDirect

Engineering Failure Analysis

Failure analysis of a third stage gas turbine blade

Politecnico di Milano, Dipartimento di Meccanica, via La Masa, 1, 20156 Milano, Italy Laborelec, Rodestraat 125, B-1630 Linkebeek, Belgium

a r t i c l e

i n f o

a b s t r a c t
This paper investigates a third stage turbine blade failure in the 150 MW unit of a thermal power plant. This primary event caused extensive damage to the unit, i.e. rupture of all blades present in the third and subsequent stages. The blade is made of nickel-based superalloy, Inconel 738, and the blade failure occurred at approximately 22,400 operating hours (25,600 equivalent operating hours) after a major overhaul. Several examinations were carried out in order to identify the failures root cause: visual examination, SEM fractography, chemical analysis, micro-hardness measurement, and microstructural characterization. The fracture on the turbine blade is located at the top r tree root and the fracture surface exhibits two characteristic zones: the rst zone shows slow and stable crack growth with crystallographic faceted cracking and striation formation, and the second shows interdendritic fracture, typical of nal stage failure. From the examinations carried out, it was possible to identify the cause of the primary failure. The identied fracture mechanism was high cycle fatigue originated by fretting on the r tree lateral surface (i.e. fretting fatigue). 2010 Elsevier Ltd. All rights reserved.

Article history: Received 10 March 2010 Received in revised form 17 September 2010 Accepted 17 September 2010 Available online 25 September 2010 Keywords: Turbine Blade Fretting fatigue Shot peening

1. Introduction Gas turbine blades are critical components of a power plant. In the event of their failure, the power plant will shut down, which can lead to prolonged outages and economic loss. When this occurs, it is necessary to conduct a detailed failure analysis of the turbine blades in order to understand the problem and improve the turbine systems reliability [1,2]. During the last few decades, the operating temperatures of gas turbine engines have been continually increased to achieve increased engine power and efciency. For this reason gas turbine blades are made of nickel-based and cobalt-based superalloys: these are the only materials able to withstand the combination of high stress and high temperature [3], while weathering the oxidative and corrosive environment. Turbine blades are susceptible to damage and crack formation in regions of contact (serrations) exposed to both centrifugal loading and oscillatory vibrations. This phenomenon is known as fretting fatigue and often occurs in the blade and disk attachment region of gas turbines and jet engines [4]. Fretting is a phenomenon that occurs at the contact area between two materials under load and subject to slight relative movement by vibration or some other solicitations. Damage begins with local adhesion between mating surfaces and progresses when adhered particles are removed from the surface [5]. Fretting fatigue is similar to fretting contact but with the interface oscillation originating from an additional remotely applied cyclic bulk load [6]. This wear mode can cause surface micro-crack initiation within the rst several thousand cycles, signicantly reducing the component life. Additionally, cracks due to fretting are usually hidden by the contacting components and are not easily
Corresponding author. Tel.: +39 (0) 2 2399 8662; fax: +39 (0) 2 2399 8644.
E-mail address: [email protected] (S. Barella). 1350-6307/$ - see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2010.09.017

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

387

detected. These fretting cracks are initially quite small, but may eventually lead to severe component damage. Eliminating or reducing slip between mated surfaces is the only method for preventing fretting fatigue, and it must be accomplished during the design process [5]. Another possible solution would be shot peening, thereby inducing compressive stresses under the surface to increase the fatigue strength. This paper reports on the failure analysis of a third stage gas turbine blade in the 150 MW unit of a thermal power plant. The turbine underwent a major overhaul in 2006 and the blade failure occurred after 22,400 operation hours. The failure caused extensive damage and economic loss to the unit. The turbine had been in operation for 67,500 h prior to the overhaul. The study focuses on a blade made of Inconel 738 (IN738), which was processed with casting followed by hot isostatic pressing (HIP). The blade was solution annealed and aged. The blade r tree was also shot peened and the blade airfoil coated with a chromium diffusion layer. Examination of the failed blade enables us to understand the fracture mechanism, which occurred near the r tree joint. The studied failure relates to a fretting fatigue phenomenon in the high cycle fatigue regime. 2. Experimental procedure Fig. 1 shows the remaining blade sections under investigation. These sections include the r tree root and the remaining blade (shank, platform and root of the airfoil). The r tree and the airfoil are completely separated (Fig. 1). After an extensive visual examination, the r tree root was cut into three parts in order to carry out SEM (scanning electron microscope) examination of the fractured surface. The SEM is equipped with an EDS (energy dispersive spectroscopy) probe that was used to perform local chemical analyses. Metallographic analyses were performed to characterize the r tree macro and microstructure. These analyses were realized by optical microscopy and SEM. All samples were prepared by grinding and polishing. Macro-etchings were realized using acqua regia (1 part HNO3 and 3 part HCl applied for 20 s). Micro-etching was realized by means of molybdic acid applied for 30 s. Chemical analysis was performed using the optical emission spectroscopy method. Micro-hardness Vickers tests were performed with a 300 g load and an indentation time of 15 s. Micro-indentation tests were also carried out on dened areas of the sample. The applied load was 100 mN. 3. Results and discussion 3.1. Fractographic analysis The fracture surface is located at the top of the r tree root and the fracture orientation is normal to the blade longitudinal axis. This failure can be considered a macroscopically brittle fracture because there are no visible plastic deformations (i.e. area reduction) near the fracture location (Fig. 1). Visual examination of the fractured blade clearly shows the typical fatigue fracture surface with a propagation area and a nal fracture zone (Fig. 2). Some brownish marks were detected on the root lateral side, near fatigue origin. These marks may be related to fretting damage due to contact between the blade and steel rotor disk (Fig. 3). SEM analyses permitted easy detection of the crack origin in the propagation area. Pictures of the propagation area are displayed in Fig. 4.

Fig. 1. Failed blade: airfoil (at the top) and r tree root.

388

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

Fig. 2. Fracture surface of the blade.

Fig. 3. Brownish marks on the r tree root lateral surface.

Fig. 4. Propagation crack zone fractography magnication near the crack origin: at low (a and b); and at high magnication (c and d).

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

389

Fig. 5. Final fracture zone fractography: ductile (a) and brittle (b) behavior.

Fig. 6. SEM analysis of r tree joint contact surface: at low (a) and high (b) magnication.

Apart from fretting damages (Fig. 4a), no other metallurgical defects were detected near the fatigue origin. A slant crack with a particular orientation can be noted in this area, most probably a secondary fatigue crack origin related to the surface mechanical damages previously identied. At higher magnication, fatigue striations were evident on the fracture surface; they are typical of transgranular fatigue propagation under high cycle fatigue conditions (Fig. 4c and d). An EDS analysis was also conducted near the crack origin: no contaminants are present in this region. SEM analysis of the nal fracture zone highlights the typical brittle and rocky area of the cast blade. This area is partially transgranular and, intergranular, both ductile and brittle at the microscopic level (Fig. 5). This fracture surface morphology is associated with an unstable fast fracture due to the exhaustion of the load carrying capacity of the blade cross section. Close examination of the r tree joint reveals fretting cracks, from 500 lm to 2 mm long (Fig. 6) across the root contact surfaces (rst lobe of the r tree joint). The magnied morphology of the brownish region consists of accumulated wear debris particles trapped between the contact surfaces (blade r tree and rotor disk). Wear debris EDS analysis reveals a high concentration of iron-based oxides; iron is absent from the blade composition but it is the base element of the disk material (Fig. 7). This wear debris entrapment occurs under fretting conditions when opposing local surface regions undergo a small localized relative displacement in tightly assembled components. These conditions are typically encountered in the r tree root of rotating turbine blades. In fact, when a turbine blade is in operation, two different forces act upon the r tree joint. The rst force is induced by radial strain, which results from changes in disk speed. As the rpm increases, the disk slot opens and the blade moves outward under centrifugal load, thus causing a small amount of slip along the surface of the r tree. The second is the high-frequency blade vibration, mainly from the aerodynamics of turbine operation, which causes blade oscillation. The relative motion between the blade r tree and the disk groove under these two forces can be responsible of a fretting phenomenon. This process plastically deforms the outermost layer (as revealed by the metallographic observation) and creates an adhesive/abrasive wear of the oxide lms on both surfaces, forming micro-cracks. Thus, design loads induce fatigue propagation and failure. In addition, during the last major overhaul all blades had been changed; the refurbished blades were heat-treated to obtain the optimal microstructure. The heat treatment caused changes in dimensional tolerance of the r tree, increasing the possibility of having reciprocal slips between the blade and rotor disk.

390

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

Fig. 7. Brownish marks on the r tree joint contact surface: SEM (a) and EDS (b) analysis.

The above ndings, which include the formation of oxide wear products on r tree contact surfaces, the presence of wear tracks, the crack nucleation in the wear tracks vicinity, and the fatigue crack slant orientation in the nucleation region, are indicative of the strong fretting fatigue mechanism inuence on the blade failure [7,8]. The typical aspects of the fretting phenomenon were found also on different other blades, even if the damages were observed to be restrained at their initial stage. 3.2. Chemical and microstructural analyses Chemical analyses were performed on two specimens: one cut from the r tree region beneath the fracture surface and one from the airfoil. The results are shown in Table 1 and they comply with the standard chemical composition for IN738 [9]. A macro-etch was carried out on a transverse section of the r tree root. The macrostructure shows a typical coarse dendritic equiaxed cast structure (Fig. 8). Micro-etches were performed on the r tree transverse section both beneath and at a distance from the fracture surface. The samples show a dendritic microstructure with grain boundary carbides and ne gamma prime particles. Carbides (MCtype) are homogeneously distributed at grain boundaries (Fig. 9a). Distribution of ne gamma prime particles and interdendritic carbides in the nickel matrix can be seen in the higher magnication SEM micrograph in Fig. 9b. Cubical gamma prime precipitates are homogenously distributed and appear slightly rounded: the mean diameter is between 1 and 1.5 lm and inter-particle distance is less than 10 lm. Aging treatment of IN738 also induces the formation of small disk-like coherent precipitates, called secondary gamma prime, which improve high-temperature creep resistance. This precipitated phase was present in the studied material.

Table 1 Blade r tree and airfoil chemical composition (wt.%). C 0.094 0.091 Co 8.52 8.62 Cr 15.76 15.48 Mo 1.86 1.99 W 2.55 2.63 Ta 1.73 1.77 Nb 0.92 0.95 Al 3.69 3.81 Ti 3.36 3.17 Fe 0.08 0.07

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

391

Fig. 8. Fir tree root macrostructure (acqua regia macro-etching).

Fig. 9. Fir tree root microstructure: carbide distribution (a); primary and secondary gamma prime distribution (b).

Fig. 10. Gamma prime depleted zone at low (a) and high magnication (b).

Its shape and distribution appear normal. The microstructure quality rating is between good and acceptable for this application taking into consideration the components creep resistance and long-time operating service.

392

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

Neither metallographic defects nor metallurgical discontinuities were observed in samples taken from the root at a distance from the fracture surface. However, a gamma prime depleted zone near the surface (depth about 20 lm) was detected all around the r tree surface (Fig. 10). In-depth examinations are required to better understand this problem and its effect on the failure mechanism. The gamma prime depleted zone may be an unexpected side-effect of the re-heat treatment (solution annealing and aging) performed during the major overhaul. It most likely depends on a synergistic effect between the re-precipitation kinetics of gamma prime phase (occurring during re-heat treatment) and the remaining compressive stresses (induced by shot peening during blade fabrication). High compressive stresses located near the surface may inhibit gamma prime re-precipitation and only secondary gamma prime phase may form, thus inducing a softer layer near the surface. This phenomenon denitely took place during the major overhaul heat treatment [10]. Micro-indentation testing was conducted in this particular zone: the results reported in Fig. 11 show a signicant reduction in micro-hardness values with respect to the r tree centre. Refs. [5,1113] suggest that a reduction in surface hardness decreases the resistance of the material against fretting fatigue, increasing the materials susceptibility to crack initiation and thus to fretting fatigue. Metallographic observations carried out on transverse sections beneath the fracture surface show the same features as those exhibited in the samples taken at a distance from this region. The microstructure is homogeneous and without defect. Some interdendritic secondary micro-cracks emanating from the fracture surface are present. Finally, several micro-hardness proles were compiled for different blade sections. The averages are reported in Fig. 12. Micro-hardness numbers show a slight increase when moving from the top of the r tree to the bottom, which is due to the operational thermal gradient typically found within the blade root.

Fig. 11. Micro-hardness numbers in the gamma prime depleted zone (numbers are in HV).

Fig. 12. Micro-hardness numbers on the r tree root.

S. Barella et al. / Engineering Failure Analysis 18 (2011) 386393

393

4. Conclusion The studied turbine blade failure must be attributed to fretting fatigue phenomena as a result of contact between the nickel-based blade r tree and the steel-based rotor disk. Moreover, the absence of shot peening at the time of refurbishment is a relevant cause of failure. Based on fractographic observations, three distinct stages (crack initiation, crack growth and nal fracture) were identied on the surface of the fractured blade, which failed at the top r tree lobe in the blade/disk joint. Debris formation and wear products on the contact surfaces of the r tree root, presence of wear tracks, crack initiation near wear damages, and the slant orientation of fatigue crack origin, present primary evidence of the mechanism of fretting fatigue to the blade failure process. In conclusion fretting fatigue must be considered as the unique cause of the component failure. Different mechanical phenomena may have promoted the observed fretting fatigue damage, but since relative movements are inherent to blade operation, it suggests that after the refurbishment the mechanical strength in the surface of the components were not high enough to mitigate the occurrence of fretting fatigue crack initiation. References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] Wang W, Xuan F. Failure analysis of the nal stage blade in steam turbine. J Eng Fail Anal 2007;14:63241. Vardar N, Ekerim A. Failure analysis of gas turbine blades in a thermal power plant. J Eng Fail Anal 2007;14:7439. Bhaumik SK, Sujata M. Failure of a low pressure turbine rotor blade of an aeroengine. J Eng Fail Anal 2006;13:120219. Garcia DB, Grandt AF. Fractographic investigation of fretting fatigue cracks in Ti6Al4V. J Eng Fail Anal 2005;12:53748. AA, VV, ASM handbook, vol. 19. Fatigue and fracture. Materials Park (Ohio): ASM International; 1997. Farhangi H, Fouladi Moghadam AA. Fractographic investigation of the failure of second stage gas turbine blades. In: Proceedings of 8th international fracture conference, 79 November 2007, Istambul, Turkey. Rajasekaran R, Nowell D. Fretting fatigue in dovetail blade roots: experiment and analysis. J Tribol Int 2006;39:127785. Jin O, Mall S. Effects of slip on fretting behavior: experiments and analyses. J Wear 2004;256:67184. Donachie MJ, Donachie SJ. Superalloy, a technical guide, 2nd ed. Materials Park, OH: ASM international; 2002. p. 6. Cai D, Nie P, Shan J, Liu W, Gao Y, Yao M. Precipitation and residual stress relaxation kinetics in shot-peened Inconel 718. JMEPEG 2006;15:6147. Lykins CD, Mall S, Jain V. An evaluation of parameters for predicting fretting fatigue crack initiation. Int J Fatigue 2000;22:70316. Mall S, Kim H-K, Porter WJ, Ownby TraylorAG. High temperature fretting fatigue behavior of IN100. Int J Fatigue 2010;32:128998. Kubiak K, Fouvry S. Introduction of a fretting-fatigue mapping concept: development of a dual crack nucleation crack propagation approach to formalize fretting fatigue damage. Int J Fatigue 2009;31:25062.