Engineering Failure Analysis 18 (2011) 1511–1519

Contents lists available at ScienceDirect

Engineering Failure Analysis

journal homepage: www.elsevier .com/locate /engfai lanal

Failure analysis of a steam turbine rotor

S. Barella ⇑, M. Bellogini, M. Boniardi, S. CinceraDipartimento di Meccanica, Politecnico di Milano, Italy

a r t i c l e i n f o

Article history:Received 4 October 2010Received in revised form 26 April 2011Accepted 10 May 2011Available online 19 May 2011

Keywords:Steam turbineCrackShaftFatigue

1350-6307/$ - see front matter � 2011 Elsevier Ltddoi:10.1016/j.engfailanal.2011.05.006

⇑ Corresponding author. Address: Dipartimento dfax: +39 (0)2 2399 8644.

E-mail address: silvia.barella@polimi.it (S. Barell

a b s t r a c t

A large and growing portion of electricity is being produced by aging thermal power plants,and although steam turbines are being constructed with excellent high quality materialssuch as CrMoNiV steel, varying forms of metallurgical degradation due to creep and/or fati-gue could still affect the parts and components during long-term operation at high temper-atures [1]. Moreover, the de-regulated electricity market, which has existed forapproximately 15 years, has led to energy companies operating their power plants in aflexible manner, as opposed to continuous operation, in order to maintain profitability ina very competitive commercial environment [2].

This paper investigates the rotor turbine failure of a 60 MW unit of a thermal powerplant. The rotor was made of CrMoNiV steel, and the failure occurred after approximately10 years in operation. Several different analyses were carried out in order to identify thefailure’s root cause: visual examination, SEM fractography, micro-hardness measurement,and microstructural characterization.

The fracture of the shaft was located at the first stage [2] and the fatigue fractureextended over roughly 75% of the initial cross section. Primary failure causes were identi-fied by the analyses performed, and the observed fracture mechanism was traced back tohigh cycle fatigue damage. The origin of the fatigue phenomenon can be traced to the stressfield generated on the rotor surface by both the frequent startup cycles and the blade fixingmethod.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Steam turbine rotors are among the most critical and highly stressed components in modern power plants [3]. The con-sequences of a rotor failure are severe in terms of both safety and economic impact. For this reason electric power utilitiesand manufacturers quantify and limit the risk of such failures using the concept of ‘‘rotor life’’, which is the maximum num-ber of service hours and hot and cold starts to which a rotor can be subjected [3].

However, de-regulation of the electricity market has led energy companies away from continuous operation to a moreflexible operating schedule in order to maintain profitability in a competitive commercial environment. The principal con-sequence of de-regulation is an increase in annual startup cycles, which enhances the degradation rate of rotor material.

These rotors are often made of CrMoNiV steel and have a limited lifespan due to creep and thermal fatigue.Rotor surface thermal stresses are greatest in areas of high mechanical stress concentration and where a high bore to

periphery temperature differential exists. The most burdensome combination of these conditions is found at notches suchas the heat relieve grooves of the glands at the inlet end of the rotor, fillet radii at the base of disks, balance holes in the disksand blade grooving, in reaction-type rotors.

. All rights reserved.

i Meccanica, Politecnico di Milano, Via La Masa 1, 20156 Milano, Italy. Tel.: +39 (0)2 2399 8662;

a).

1512 S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519

In order to prevent these failures, many non-destructive test procedures have been developed. However, in some cases,cracks (i.e. rotor surface cracks) are not visible with ultra sonic testing and can only be detected using vibrations monitoring[4].

The object of this study was the rotor of a 60 MW steam turbine with a 3000 run/min top speed, and 480 �C inlet steamtemperature. The shaft was a forged component made of 30CrMoNiV4 11 (equivalent to ASTM A470 grade D class 8).

The studied spindle was overhauled on a regular basis every 10 years. Approximately 6 months after the last major over-haul, the vibration levels started to steadily increase and the maximum vibration level was achieved after 1–2 h of operationat full load. Ultrasonic tests on the steam turbine rotor were conducted to verify the possibility of shaft cracks, however nocracks were found using non-destructive tests.

The turbine was restarted after cursory maintenance operations, however the vibration and eccentricity measurementsdisplayed values above the alarm limits, and thus the rotor was taken out of service.

The turbine blades were removed and a circumferential crack was discovered along the first stage dovetail slot on thesteam turbine rotor.

2. Experimental procedure

Fig. 1 shows the failed rotor shaft sketch. The rotor was cut in order to open the crack and analyze the fracture surface. Aforging hammer performed the crack opening.

Chemical analyses were performed using the optical emission spectroscopy method. These analyses were performedproximal and distal to both the fracture surface and the rotor bore.

Several mechanical tests were conducted on the shaft: tensile and resilience tests were performed in different positions(proximal and distal to both the fracture surface and the rotor bore) and on different orientations in relation to the shaft axis(circumferential, radial and longitudinal).

After the visual examination, the fracture surface was cut into 24 parts in order to carry out SEM (scanning electronmicroscope) study on the fracture surface. The SEM was equipped with an EDS (energy dispersive spectroscopy) probe forlocal chemical analyses.Metallographic analyses were performed to characterize the rotor microstructure in different posi-tions (proximal and distal from both the fracture surface and shaft bore). These analyses were performed using opticalmicroscopy and SEM. Microstructure samples were prepared by grinding and polishing and they were etched in Nital 2%(2 ml HNO3 and 98 ml ethanol applied for 10 s).

Micro-hardness Vickers tests were carried out applying a 10 mN load for 15 s on the fracture surface.

3. Results and discussion

After the blades removal, the crack appeared in the first stage groove. The rotor was examined extensively prior to thecutting operation, and the crack was circumferential along the first stage and with a width of approximately 2–3 mm. Nonon-destructive tests (NDT) were necessary for detection.

Liquid penetrant inspection was likewise carried out on the entire shaft length in order to locate other non-visible cracks,however no other cracks were detected.

The crack was present along the entire circumference of only the first stage dovetail slot. Multiple equally-spaced inden-tations were present on the dovetail slot (Fig. 2). These indentations were very likely due to the blade fixing technique andmeasured approximately 1.5 mm long and 0.4 mm deep.

This type of defect was also present on the other dovetail slot stages.

Fig. 1. Failed rotor sketch.

Fig. 2. Aspect of the dovetail slot: visible crack and indentations.

Table 1Chemical composition of the rotor near the bore and the fracture surface.

El C Si Mn P S Cr Ni Mo Cu V

wt.% 0.308 0.0490 0.72 0.0056 0.0041 1.53 0.640 1.067 0.0770 0.314

S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519 1513

The chemical composition was quite similar in all analyzed zones. An example of these tests is reported in Table 1. Thesevalues fully comply with the standard ASTM A471.

The mechanical tests results are reported in Tables 2 and 3. Mechanical properties were homogeneous on the entire shaftsection (proximal and distal to the fracture surface) and in different directions (longitudinal, circumferential and radial direc-tion). In addition, these values are fully compliant with the ASTM A470 standard.

After the crack opening operations, the fracture surface was clearly visible. The crack propagation area was widely ex-tended and the residual cross sectional area was about 25% of its original size. Moreover, the crack propagation depthwas irregular along the circumference (Fig. 3) and the direction of maximum propagation was at an approximate angularposition of 75�.

The fracture surface showed a typical fatigue rupture, with a clearly detectable fatigue propagation zone and a final frac-ture area (due to the crack opening operation). Many typical fatigue beach marks were clearly visible on the fracture surface(fatigue propagation zone). The final fracture zone showed a brittle morphology. This brittle behavior is related to the equi-triaxial stress field, generated during the crack opening operation [5].

The fatigue fracture surface was heavily oxidized and differing grades of oxidation were clearly visible. The oxidized layerwas thick, black and compact near the crack initiation (shaft circumference), while it appeared reddish and thin near the finalfatigue fracture zone. As the oxidized layer rendered observation of the damaged surface difficult, different attempts at oxideetching were made in order to remove the layer. Unfortunately, none of the tested etchants removed the oxide: it was com-pact and did not react to chemical agents.

Four different zones on the fatigue fracture surface are visible (moving from the external boundaries to the final brittlefracture) in Fig. 4a:

1. a thin annulus (2–3 mm), where several fracture plane changes were present (Fig. 4b);2. a smooth zone (50–60 mm) in which beach marks were not easily visible;3. a beach marks zone in which several beach marks were clearly visible;4. a secondary crack zone intermediate adjoining the final rupture surface

In spite of several fracture surface parts having been analyzed by SEM, only the analysis conducted on the 135–150� sec-tor has been reported. The fracture surface appeared to be better preserved in this sector.

Although the fracture surface was oxidized, the SEM analysis revealed important crack features.In the first zone, being the most external, many fracture plane changes were evident. All of these changes corresponded to

the indentations present on the dovetail slot. Near the shaft edge the fracture surface appeared slightly more even than that

Table 2Mechanical properties near the fracture surface in different direction, the value in brackets is the standard deviation.

Set (MPa) Rp0.2 (MPa) Rm A (%) Z (%) KV (J)

Circumferential 542 (1.3) 682 (2.4) 21 (2.2) 50 (3.2) 16.2 (1.6)Longitudinal 542 (1.2) 683 (2.9) 23 (0.6) 65 (1.3) 14.7 (2.3)Radial 541 (3.1) 680 (2.4) 20 (1.1) 54 (2.8) 16.6 (1.6)

Table 3Mechanical properties far from the fracture surface in different direction, the value in brackets is the standard deviation.

Set Rp0.2 (MPa) Rm (MPa) A (%) Z (%) KV (J)

Circumferential 544 (1.4) 694 (0.9) 21 (0.9) 55 (2.9) 17.3 (1.9)Longitudinal 547 (1.8) 689 (1.5) 24 (0.7) 67 (0.8) 23.5 (2.0)Radial 545 (2.8) 692 (3.1) 21 (0.5) 58 (1.5) 21.0 (1.5)

Fig. 3. Fracture surface after the crack opening. A cut along the diameter was necessary to open the crack.

Fig. 4. Fatigue fracture surface zone (a). Detail of the external zone (b): plane changes were evident.

1514 S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519

of the inner parts. The crack plan changes were evident in this zone (by visual examination). Fig. 5a shows crack planechanges at high magnification. They appeared smooth, shiny (with QBSD probe) and not oxidized. This is probably due tothe relative movement of the two sides of the fracture surface while it was in operation.

Fatigue striations were also present in this zone and they were parallel to the shaft edge but they turned in proximity tothe crack plane changes (Fig. 5b).

In the second fatigue fracture zone SEM analysis did not distinguish any particular features.In the third fatigue fracture zone morphological analysis showed beach marks on the oxidized layer (Fig. 6).River patterns were detectable in the last fatigue fracture zone. These radial stepwise lines indicated the proximity

of unstable crack propagation, and were related to the impending catastrophic failure of the component [6]. SEM anal-ysis was also conducted on the final brittle fracture and showed a typical brittle transgranular morphology (cleavagefracture) due to the particular tri-axial stress field generated during the opening crack operation. Various decohesion

Fig. 5. Plane change detail: (a) QBSD analysis; (b) striation near the plane change.

Fig. 6. SEM analysis on fatigue fracture surface third zone: (a) low and (b) high magnification.

S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519 1515

zones were present, however these are typically found in the brittle fracture of a forged component, especially in thebore area.

A lateral section of the fracture surface was prepared and analyzed after the fracture surface examination. In this section,oxide thickness was measurable at approximately 60–70 lm thick. The oxide thickness decreased starting from the shaftedge to the bore (Fig. 7a) as a result of the gradual opening of the crack during the component’s service life.

The oxide morphology along the fracture surface section was not uniform: near the fracture surface it was compact andbecame porous 30 lm away (Fig. 7a).

In addition, during SEM study of the fracture section, many secondary cracks were found beneath the surface in zonethree and four of the fracture surface (Fig. 7b). These were internally oxidized and presented transgranular morphology.These cracks are normally found in the final section of a fatigue crack.

Fig. 7. Lateral section: oxide thicknesses (a) near the rotor surface (b) below the third fatigue fracture zone.

Fig. 8. Section beneath the fracture surface (QDSD): secondary cracks were visible in the region near the indentation at low (a) and high (b) magnification.

Fig. 9. Metallographic sections (a) beneath the fracture surface near the rotor surface and (b) near the rotor bore.

Fig. 10. Evidence of work hardening near the indentation on a metallographic section below the fracture surface.

1516 S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519

Fig. 11. Microhardness test on a metallographic section below the fracture surface.

Table 4Microhardness number on a section under the fracture surface.

Point (#) Microhardness (HV) Point (#) Microhardness (HV)

1 242.9 11 200.62 277.2 12 198.83 285.9 13 196.04 267.4 14 202.65 219.4 15 216.36 206.8 16 217.67 203.6 17 285.88 196.6 18 277.99 206.1 19 255.9

10 202.7 20 193.6

S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519 1517

Metallographic analysis was carried out in order to determine material microstructure both proximal and distal to thefracture surface.

The analysis showed a typical martensitic structure [7] in every observed shaft zone. The metallurgical structure was nor-mal and no abnormal defects were present. Several carbide networks were present at grain boundaries. They are typicallypresent in the microstructure of a high temperature component in operation for thousands of service hours (Fig. 8).

In order to better understand the fatigue origin, a metallographic section was obtained from the external annulus, justbelow the fracture surface.

This section was prepared by grinding a portion of the fracture surface. The SEM observation highlighted secondary inter-nally oxidized cracks perpendicular to the shaft edge (0.1 mm long). This was due to extended exposure to high temperature(Fig. 9). The origin of these cracks could be the blade fixing method.

In addition, the metallographic analysis was carried out on the section beneath the fracture surface. Fig. 10 shows thematerial work hardening [8] corresponding to the dovetail slot indentations.

The material work hardening was also confirmed by the micro-hardness test values (Fig. 11 and Table 4).Micro-hardness numbers were higher in the dovetail slot indentation zones than in the zone between the two

indentations.The blade fixing indentation method induced an altered metallurgical status on the shaft surface in terms of structure,

hardness and residual compressive stresses.A simplified FEM simulation showed the material work hardening effect (Fig. 12).FEM analysis was performed assuming a surface thermal stress of about 400 MPa (due to the transient thermal stress dur-

ing start-up and shutdown) and a residual stress of about 50 MPa due to work hardening [9,10]. Considering this input data,the stress field on the component surface reached its maximum in between the two indentations.

The simulation results also explained the particular fatigue striations figure. The fatigue crack origins were between thetwo indentations due to the high stress intensity, and then the cracks propagated parallel to the shaft surface. The work-hardened zones, due to dovetail slot indentations, acted as obstacles during the first crack growth period. These regions wereharder than the surrounding areas and the cracks did not propagate across them. When the cracks were large enough theycoalesced and the work hardened zones broke, originating crack plane changes. After this initial period, the main crack prop-agated according to a fatigue phenomenon.

The analysis performed underlined the conclusion that crack origination was neither due to corrosion nor fretting fa-tigue phenomena. The only possible explanation for the crack initiation was the thermal gradient during start-up andthe resultant presence of high stress concentration in the first groove [9,10], which was magnified by the presence ofthe indentions. However, it is worth mentioning that the specific initiation sites did not correspond to the areas contain-ing the highest stress concentration, but instead to the rim, where the work hardening was not strong enough to preventcrack initiation.

Fig. 12. Indentation effect simulation (FEM). The white arrows indicate the blades position. Stresses reported in MPa.

1518 S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519

4. Conclusion

In conclusion, mechanical fatigue shall be considered the unique root cause of the failed component and the transientthermal stresses the origin of this phenomenon: these stresses resulted from bore to periphery thermal gradient during tur-bine startup and shutdown [11,12]. Rotor surface thermal stresses were greatest in areas of high stress concentration (i.e.indentations) and where high bore to periphery temperature differentials existed [13]. The initiation and propagation ofcracks at the rotor surface resulted from low cycle high strain fatigue, due to cyclic thermal stressing from startup and loadchanging [3]. This cyclic stress field promoted small crack (0.1–0.2 mm) initiation and slow propagation at the rotor surface.The coalescence of the small surface cracks formed a wide principal circumferential crack at the shaft periphery [2]. The riskof shaft fracture due to this circumferential crack depended on its progression by thermal stresses to a size large enough toallow failure by mechanical high cycle fatigue (due to rotor ‘self weight’ cyclic bending stresses). Cracks like these are alsodangerous because their position makes them difficult to detect.

Changes in the Electricity Supply Industry (marked by the end of monopoly age) include substantial and increasing com-mercial pressure to operate the steam turbine plant on a flexible basis. The increase in annual start cycles basically enhancesthe rotor material degradation rate [3]. Moreover, dovetail slot indentations are detrimental to rotor safety because they in-duce a further stress increase on the component surface. The indentations are the resultant marks of blade fixing, which is avery common fixing procedure, but unfortunately, within the context of the new energy market, not a good practice.

New turbine projects take this problem into account, and the fixing methods have been modified.Further accurate ND tests should be performed on steam turbines in operation to determine the existence of early stage

fatigue cracks (as suggested by many authors [14–16]).

References

[1] Saito K. Recent life assessment technology for existing steam turbines. Proc ASME Power Conf 2005;49(2):597–602.[2] Johnston C. An approach to power station boiler and turbine life management. Insight-Non-Destruct Testing Condition Monit 2004;46:606–9.[3] Martin RJ. SteamTurbine management in a changing market. Ommi 2003;1.[4] Ishida Y. Cracked rotors: industrial machine case histories and nonlinear effects shown by simple jeffcott rotor. Mech Syst Signal Process

2008;22:805–17.[5] Gdoutos EE. Fracture mechanics: an introduction. 2nd ed. Netherlands: Springer; 2005.[6] Fractography. USA: ASM International; 1987.[7] Metallography and microstructures. USA: ASM International; 2004.[8] Campbell FC. Elements of metallurgy and engineering alloys. USA: ASM International; 2008.[9] Vania A, Pennacchi P, Chatterton S. Effects of thermal transients on cracked shaft vibrations. In: Proceeding of ASME turbo expo 2011 – GT2011.

Canada: Vancouver.[10] Vania A, Pennacchi P. Shaft crack detection in a steam turbine: experimental evidences and model-based simulations. In: Proceeding of SMA 2010, int.

conference on noise and vibration engineering. Belgium: Leuven.[11] Novak G, Rusin A. Lifetime deterioration of turbine components during start-ups. Ommi 2004;3.

S. Barella et al. / Engineering Failure Analysis 18 (2011) 1511–1519 1519

[12] Ghosh SJ. Cracks in steam turbine components. Russ J Nondestruct 2006;42:134–46.[13] Stubbs T. The role of NDE in the life management of steam turbine rotors. Insight-Non-Destruct Test Condition Monit 2004;46:529–32.[14] Moser R. Automated steam turbine straddle root disc head inspection. Springer Tracts Adv Robot 2008;42:513–20.[15] Mayer KH. Characterization and quantification of defects in rotors and casings of steam turbines. Mater High Temp 1998;15:249–58.[16] Bird CR. Early detection of service-induced cracks in turbine generators using ultrasonic techniques. Brit J Nondestruct Test 1992;34:453–5.