Engineering Failure Analysis 84 (2018) 11–24
Contents lists available at ScienceDirect
Engineering Failure Analysis jo urn al hom ep age : www.elsevier.com/locate/engfailanal
Vibration analysis for failure detection in low pressure steam turbine blades in nuclear power plant Wensheng Zhaoa, , Yanhui Li a, Meixin Xue b, Pengfei Wang a, Jin Jiang a ⁎
a b
Key Laboratory of Hydraulic Machinery Transients (Wuhan University), Ministry of Education, Wuhan 430072, China 701 Institute, China Shipbuilding Industry Corporation, Wuhan 430064, China
A R TI C LE I NF O
A B S T R A C T
Keywords: Turbine blade Resonant vibration Failure analysis High cycle fatigue Fretting fatigue
This paper presents an investigation of the failure of a low-pressure steam turbine blade in a pressurized water reactor (PWR) nuclear power plant. The dynamical behaviour of the blade is analyzed theoretically and experimentally. A three-dimensional �nite element model is used to predict predict the blade resonances resonances in the operational operational speed range. Natural frequencies frequencies and mode shapes of the blade at static condition are obtained, then natural frequencies of the blade at diff erent erent rotational speeds are calculated with consideration of centrifugal force and steam � ow forces. A Campbell diagram is plotted to predict the likely operational conditions that may cause resonant vibration of the blade. Vibration tests are conducted to determine the vibration characteristic acteristic of the blade. blade. It is found that the 2nd natural frequency frequency of the blade is very close close to the 9th rotor speed harmonic. The experimental natural frequencies are in good agreement with the �nite element predicted values. Fretting wear is observed at the concave root surfaces of the blade trailing edge caused by resonant vibration. The fracture surface of the cracked blade shows typical typical fatigue fatigue patterns. patterns. The fretting fretting wear characteristics characteristics in the crack initiation initiation regions are observed. Stress distribution of the blade at the 9th harmonic frequency is analyzed using an elasticplastic �nite element model. Fretting fatigue experiments indicate that the fatigue life of the blade is greatly reduced due to fretting wear. The results of the investigation show that the failure of the blade is attributed to a combination of high cycle fatigue (HCF) and fretting wear.
1. Introduction Introduction Turbine Turbine blades are the critical components components in power plants which convert steam � owing into rotary shaft [1] shaft [1].. However, Failures of turbine blades are widely observed in power plants that will shut o ff the the power supply [2] supply [2].. The failure mechanisms of blades may include corrosion due to working �uids, low cycle fatigue caused by transient operations, high cycle fatigue induced by forced vibrations, etc. [3–5] 5].. Vibration induced fatigue is one of the predominant causes for blade failures in steam turbines [6] [6].. Blades vibrations are generally excited by the � uid � ow, and they may become severe when resonances occur. The turbine blades subjected to several sources of excitation are highly susceptible to undergo forced vibrations, which may occur at or near natural frequencies of the blades [7] blades [7].. The forced vibrations vibrations may enlarge enlarge the stresses resulting in degradations degradations of the blades, which is referred referred as high cycle fatigue (HCF) [8] (HCF) [8].. Furthermore, a large number of high-stress cycles are accumulated in the blades during startup of turbines. During the past decades years, vibration analysis for failure detection in turbine blades have been conducted by many scholars [9–11] 11].. Research on the blades vibration focused on natural frequencies and mode shapes, which were analyzed by modal analysis,
⁎
Corresponding author. E-mail address: wensheng.zhao@w
[email protected] hu.edu.cn (W. (W. Zhao).
http://dx.doi.org/10.1016/j.engfailanal.2017.10.009 Received 14 July 2016; Received in revised form 11 May 2017; Accepted 11 October 2017 Available online 16 October 2017 1350-6307/ © 2017 Elsevier Ltd. All rights reserved.
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
and the causes of cracked blades were also investigated. Static and dynamic �nite element analysis were used to determine the vibration characteristics of blades, modal and harmonic analysis were conducted to analyze the dynamical behaviour of blades, and Campbell diagrams were used to reveal the critical conditions for resonances when the exciting frequencies are near the natural frequencies of blades. It was concluded that the dynamic stress coupled with the maximum static stress due to the centrifugal force may induce the blade fatigue damage. Since resonance vibration of blades is inevitable, and can a ff ect the fatigue life of blades, vibration analysis of blades is very important in operational turbines. A three-dimensional � nite element model was adopted to investigate the resonances of blades, and tip timing technique was used to assess the vibration characteristics of blades by Madhavan et al. [12]. A contact stress measurement method was used to analyze the responses of rotating blades by Robinson and Washburn [13], however, it has a few disadvantages such as harsh operational conditions and failure of gauges at severe environments. A non-contact stress measurement method was adopted to investigate the dynamical behaviour of blades by Zielinski and Ziller [14], it can overcome the disadvantages of direct measurement method and monitor the vibration of blades eff ectively. Recently, a method was used to investigate the dynamics of blades by means of monitoring casings vibrations [15 –18]. A Lagrangian and Eulerian meshing approach based on signal analysis of the casing vibration was adopted to study the dynamical behaviour of a single blade by Rao and Dutta [16 –18]. In their research, the vibration response of the casing excited by blade passing frequencies was analyzed to diagnose the status of the blade. A simple model was used to investigate the dynamics of blades with snubbing e ff ect by Pennacchi et al. [19], the eff ect of the snubbing on the blade vibration reduction was studied experimentally, it was found that snubbing was e ff ective when the blade was excited in resonance or close to resonance. In the present paper, the failure of the damaged blade is investigated theoretically and experimentally. A � nite element analysis (FEA) and vibration tests are performed to analyze the dynamical behaviour of the blade. The turbine operational conditions that could lead to resonant vibration of the blade are predicted. The stress distribution of the blade working in resonance condition is calculated using the � nite element method (FEM). Fretting fatigue experiments are conducted to analyze the eff ect of fretting on the blade fatigue life.
2. Background A nuclear power plant was shut down due to high vibration of a steam turbine, and several fourth stage blades were seriously damaged. The turbine accumulated around 60,000 operation hours to failure. The fourth stage blades with four-tooth � r-tree � xing roots are free-standing types, one of the fractured blades is shown in Fig. 1. The blade under investigation is a 3000 rpm fourth stage blade of the 650 MW low-pressure steam turbine, it is made of 0Cr17Ni4Cu4Nb stainless steel, which is a Chinese GB standard material with chemical composition of C 23.42, Cr 12.76, Ni 2.73, Cu 2.19, Si 0.68, Mn 1.49, Fe 56.73. A crack is observed between the blade root platform and the � rst tooth with the use of dye penetration, as shown in Fig. 2. The crack propagates at an angle from concave side to convex side of the blade root. The crack sizes on the concave side and on the convex side are, respectively, 22 mm and 8 mm.
3. Modeling and analysis Dynamic analysis of the blade is performed using a �nite element method. The three-dimensional model of the blade is established, and the high quality meshes are generated. The natural frequencies and the mode shapes of the blade at static condition are predicted. The natural frequencies of the blade at di ff erent rotational speeds are also calculated with consideration of centrifugal
Fig. 1. View of the blade root with crack.
12
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 2. Distribution of crack on the blade root: (a) concave side; (b) convex side.
force and steam �ow forces. A Campbell diagram is plotted to calculate the likely resonant rotational speeds of the blade. 3.1. Modeling In this study, a three-dimensional model of the blade is built using an optical scanner system, and the blade model with topology is imported to the �nite element software ANSYS. The three-dimensional model of the blade is shown in Fig. 3. The blade model employs a mesh of 1,777,254 nodes and 787,828 solid hexahedral elements for this simulation. The computational grids for the �nite element model in the simulation are presented in Fig. 4. The geometric and the material parameters of blade are listed in Table 1. 3.2. Dynamic analysis of the blade The natural frequencies and the mode shapes of the blade at zero rpm are analyzed, the mode shapes for the � rst four mode of the blade are shown in Fig. 5. It can be seen that, the 1st mode shape of the blade is in bending mode; the 2nd mode shape of the blade is in bending-twisting coupling mode; the 3rd mode shape of the blade is in twisting mode; the 4th mode shape of the blade is in bending-twisting coupling mode. The four natural frequencies of the blade predicted by FEM are summarized in Table 2. The experimental �rst four natural frequencies of the blade are obtained using a roving hammer test, they are also listed in Table 2. It is seen that the � rst four natural 13
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 3. Three-dimensional model of the blade.
frequencies of the blade predicted by FEM are only within 3% diff erence compared with the experimental results. The � nite element results are in good agreement with the experimental data. In order to predict the operational conditions that could lead to resonant vibration, a Campbell diagram is plotted, as shown in Fig. 6. The inclined lines plotted from zero point to the right of the vertical line are the nth harmonic lines. The nearly horizontal lines plotted from left side to the right side are the natural frequency lines of the blade for di ff erent vibration modes. The intersections of the natural frequency lines with the nth harmonics lines indicate the possible resonances for the blade in the operational speed range. It can be seen in Fig. 6 that, the 1st vibration mode of the blade at 2042 rpm, 2502 rpm and 3482 rpm, respectively, have frequencies of 203.4 Hz, 208.0 Hz and 231.5 Hz are very close to the 6th harmonic frequency of 204.2 Hz, the 5th harmonic frequency of 208.5 Hz and the 4th harmonic frequency of 232.1 Hz; the 2nd vibration mode of the blade at 2525 rpm, 2828 rpm and 3214 rpm, respectively, have frequencies of 421.4 Hz, 423.3 Hz and 426.7 Hz are very close to the 10th harmonic frequency of 420.8 Hz, the 9th harmonic frequency of 424.2 Hz and the 8th harmonic frequency of 428.5 Hz. The � rst four natural frequencies of
14
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 4. Finite element model of the blade.
Table 1 Geometric and material parameters of the blade. Length ( L) Weight (W ) Density (ρ ) Young's modulus ( E ) Poisson ratio (ν)
479.3 mm 12.96 kg 7780 kg/m3 191 Gpa 0.27
s
the blade at corresponding speeds predicted by FEM are listed in Table 3. According to the operation guide for steam turbine given by the manufacturer, the allowed operational speed range is from 2880 rpm to 3090 rpm. Due to the uncertainty in calculating the blade natural frequencies, the 2nd blade vibration mode is more likely to be in resonance with the 9th harmonic.
15
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 5. Mode shapes of the blade: (a) 1st mode shape; (b) 2nd mode shape; (c) 3rd mode shape; (d) 4th mode shape.
Table 2 Natural frequencies of the blade at static condition. Mode
FEM frequency (Hz)
Experiment frequency (Hz)
Error (%)
1 2 3 4
191.7 412.9 645.1 795.0
188.3 408.7 629.3 778.8
1.81 1.03 2.51 2.08
Fig. 6. Campbell diagram of the blade.
16
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Table 3 First four natural frequencies of the blade at diff erent rotational speeds. Mode
Natural frequency (Hz)
Rotational speed (rpm)
1 2 3 4
2042
2502
2525
2828
3214
3482
203.4 418.8 648.9 797.4
208.0 421.1 650.5 798.0
208.2 421.4 650.8 798.2
212.6 423.3 651.6 799.1
218.3 426.7 653.7 801.2
231.5 428.5 656.8 803.1
4. Experimental analysis 4.1. Experimental set-up Sinusoidal vibration tests are conducted to determine the dynamic behaviour of the blade under di ff erent excitation conditions. The experimental set-up for measuring the natural frequencies and the vibration responses of the blade is presented in Fig. 7. The blade root is � xed with two clamps, which are pressed down by two plates on top surfaces of the clamps and are fastened by twelve bolts. Meanwhile, two pieces of metal blocks are plugged into the blade root groove, an upward vertical force is thus generated, which is used to simulate the centrifugal force caused by the blade rotation. The schematic diagram of the � xture assembly for the blade root is shown in Fig. 8. Two acceleration meters are mounted on the inner side of the blade trailing edge and the inner side of the blade leading edge. The output signals from the acceleration meters are � ltered and recorded by a data acquisition system, and the vibration frequencies and the vibration responses of the blade are analyzed by a signal analysis software. The experimental results are used to compare with the �nite element predictions.
Fig. 7. Photograph of the experimental set-up.
17
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 8. Schematic diagram of the � xture assembly for the blade root.
4.2. Natural frequency analysis The frequency spectrum of the measured blade is shown in Fig. 9. It is seen that the 1st natural frequency of the blade is 189 Hz, the 2nd natural frequency of the blade is 418 –477 Hz, and the 3rd natural frequency of the blade is 511 Hz. The 2th natural frequency is very close to the 9th rotor speed harmonic. The experimental natural frequencies are in good agreement with the theoretical predictions.
4.3. Vibration response of the blade The vibration accelerations of the blade are measured under eight di ff erent exciting frequencies, where the frequencies are respectively f = 175 Hz, 191 Hz, 276 Hz, 313 Hz, 408 Hz, 413 Hz, 525 Hz and 657 Hz. The vibration responses of the blade root are shown in Fig. 10. Fig. 10(a) shows the vibration response of the inner side of the blade trailing edge, and Fig. 10(b) shows the vibration response of the inner side of the blade leading edge. It can be seen in Fig. 10 that, the vibration accelerations of the blade vary with increasing the exciting frequencies, when the exciting frequency is close to the 2nd natural frequency of the blade, the vibration acceleration of the blade reaches the maximum value, the exciting frequency is in the vicinity of 408 –413 Hz, which corresponds with resonance in the 9th harmonic. The blade resonant vibration is excited at the 9th harmonic frequency of 413 Hz for 50 h. The fretting wear is observed at the concave surfaces of the blade root, as shown in Fig. 11.
5. Blade failure analysis 5.1. Metallurgical investigation of the blade The cracked part of the blade root is cut, polished, cleaned, and then examined with a scanning electron microscope (SEM). The crack shape on the concave side of the blade root is shown in Fig. 12. It can be seen that there are sub-cracks exists in the crack initiation area, and the crack tip propagates to two cracks. The typical morphology of the cracked tip is shown in Fig. 13. The damage seems appear in the form of intergranular cracking. The normal tempered martensitic microstructures are observed, and other abnormal microstructures do not exist.
Fig. 9. Frequency spectrum of the blade.
18
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 10. Vibration responses of the blade root: (a) inner side of the blade trailing edge; (b) inner side of the blade leading edge.
Fig. 11. Fretting wear of the blade root: (a) fretting blade root; (b) fretting surface.
5.2. Fracture surface analysis of the blade The cracked part of the blade is extracted and opened after cooling in liquid nitrogen. After a cleaning in an ultrasonic alcohol bath the opened specimen is examined using a scanning electron microscope (SEM) with an X-ray electron dispersive spectroscopy (EDS). The fracture surface of the cracked blade is shown in Fig. 14. The beach marks in Fig. 14 indicate that high cycle fatigue (HCF) propagates the crack. The fatigue striations are observed on the fracture surface, as shown in Fig. 15. The average striation spacing varies in the range of 0.8–0.9 μ m in the crack propagation regions. The fretting wear characteristics in the crack initiation regions are observed, as indicated by a box in Fig. 16. Magni�cation SEM view of the microstructures on the indicated fretted position is shown in Fig. 17. The chemical compositions of the microstructures are determined by X-ray EDS, they are mainly composed of Fe, O and Cr, as shown in Fig. 18. Fretting wear of blade undergoing vibrations may occur when the clearance between the blade root serration and the disc slot is
19
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 12. Crack shape on the concave side of the blade root.
Fig. 13. Typical morphology of the crack tip.
small. The fretting action signi�cantly accelerates the successive crack propagation. It may be concluded that the failure of the damaged blade is caused by a combination of high cycle fatigue (HCF) and fretting wear. The crack initiation and propagation of the blade may be attributed to high cycle fatigue (HCF) and important fretting wear contribution, which is induced by the contact between the blade root and the rotor disc when vibrations occur. 5.3. Static stress of the blade The static stress of the blade at rotational speed 2828 rpm is analyzed, centrifugal force and steam � ow forces are considered in the elastic-plastic stress calculation. The contact between the blade and the rotor is non-linear boundary condition. The Von Mises stress distribution of the blade due to centrifugal force and steam � ow forces is shown in Fig. 19. It is seen that the maximum Von Mises stress is about 579 MPa, which is located at the concave root side of the blade trailing edge. The Von Mises stress at the crack initiation position is about 324 MPa. The maximum stress of the blade is lower than the yield strength of the blade material σ = 707 MPa. s
5.4. Evaluation of the fretting fatigue of the blade Fretting fatigue experiments are conducted to analyze the in �uence of fretting on the blade fatigue life. The experiments are conducted on a servo-hydraulic testing machine with a frequency of 15 Hz. The experimental specimens cut from the cracked blade 20
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 14. Micrograph of the cracked blade fracture surface.
Fig. 15. Typical fatigue striations on the fracture surface.
are rod structures with diameter d = 8 mm. Frictional pairs made of turbine rotor steel are used to contact with the experimental specimens, the loading force extends from 0.4 to 4 KN, which ensure that the contact stress on the contact surface is lower than the yield strength. The fretting fatigue tests are performed with a stress amplitude of 560 ± 350 MPa on the specimen after 106 cycles of fretting. The fatigue life of the fretted specimen is 166,724 cycles, while the fatigue life of the unfretted specimen is 1,074,301 cycles. The fatigue life of the specimen is reduced by 85% due to fretting wear. The high-magni�cation SEM fractography of the fretting fatigue specimen is shown in Fig. 20. It is seen that an amount of wear debris adheres to the fretted surface. The accumulated debris may induce stress concentration and cause the crack initiation, the fretting action may accelerate the crack propagation. The micro-scaled oxide particles on the wearing surface show typical fretting fatigue characteristics.
6. Conclusions In this study the � nite element analysis and the experimental investigation are conducted to investigate the failure of the turbine blade. The fracture surface of the cracked blade shows typical high cycle fatigue patterns. The fretting wear characteristics in the crack initiation regions are observed. It may be concluded that the failure of the blade is attributed to a combination of high cycle fatigue (HCF) and fretting wear, which is induced by the contact between the blade root and the rotor disc when resonant vibration 21
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 16. Wear patterns near the crack initiation position.
Fig. 17. Microstructures on the fretted surface.
22
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 18. Chemical compositions of the microstructures.
Fig. 19. Von Mises stress distribution of the blade.
occurs. The three-dimensional �nite element analysis is performed to predict the natural frequencies and the mode shapes of the blade. The � rst four natural frequencies of the blade at static condition predicted by FEM are only within 3% diff erence compared with the experimental results. The �nite element results are in good agreement with the experimental values. The Campbell diagram shows that the 2nd vibration mode of the blade at 2828 rpm is more likely to be in resonance with the 9th harmonic. The vibration tests indicate that the vibration response of the blade reaches the maximum value as the exciting frequency is in the vicinity of 408–413 Hz, which corresponds with resonance in the 9th harmonic. Fretting wear is observed at the concave root surfaces of the blade trailing edge for 50 h resonant vibration. The maximum Von Mises stress of the blade at 2828 rpm located at the concave root side of the blade trailing edge is about 579 MPa, which is lower than the yield strength. Fretting fatigue experiments show that the fretting wear has a strong in �uence on the fatigue life of the blade, which is reduced by 85%. The wear debris accumulated on the fretted surface may cause the crack initiation and propagation.
Acknowledgements The authors gratefully acknowledge the support by National Natural Science Foundation of China (Grant No. 51409197) and Hubei Provincial Natural Science Foundation of China (Grant No. 2015CFB253).
23
Engineering Failure Analysis 84 (2018) 11–24
W. Zhao et al.
Fig. 20. High-magni�cation SEM view of the specimen fracture surface.
References [1] H. Kim, Crack evaluation of the fourth stage blade in a low-pressure steam turbine, Eng. Fail. Anal. 18 (2011) 907–913. [2] E. Silveira, G. Atxaga, A.M. Irisarri, Failure analysis of a set of compressor blades, Eng. Fail. Anal. 15 (2008) 666–674. [3] P. Pennacchi, Mechanisms of Machine Science, Proceedings of the 9th IFToMM International Conference on Rotor Dynamics, Vol. 21 2015. [4] T. Carter, Common failures in gas turbine blades, Eng. Fail. Anal. 12 (2005) 237–247. [5] F. Xue, Z. Wang, W. Zhao, X. Zhang, B. Qu, W. Liu, Fretting fatigue crack analysis of the turbine blade from nuclear power plant, Eng. Fail. Anal. 44 (2014) 299–305. [6] J. Hou, B.J. Wicks, R.A. Antonia, An investigation of fatigue failures of turbine blades in a gas turbine engine by mechanical analysis, Eng. Fail. Anal. 9 (2002) 201–211. [7] G. Tsai, Rotating vibration behavior of the turbine blades with diff erent groups of blades, J. Sound Vib. 271 (2004) 547–575. [8] L. Witek, Numerical stress and crack initiation analysis of the compressor blades after foreign object damage subjected to high-cycle fatigue, Eng. Fail. Anal. 18 (2011) 2111–2125. [9] Z. Mazur, R. Garcia-Illescas, J. Porcayo-Calderon, Last stage blades failure analysis of a 28 MW geothermal turbine, Eng. Fail. Anal. 16 (2009) 1020–1032. [10] L. Witek, Crack growth simulation in the compressor blade subjected to vibration using boundary element method, Key Eng. Mater. 598 (2014) 261–268. [11] K.W. Barlow, R. Chandra, Fatigue crack propagation simulation in an aircraft engine fan blade attachment, Int. J. Fatigue 27 (2005) 1661–1668. –39. [12] S. Madhavan, J. Rajeev, C. Sujatha, A.S. Sekhar, Vibration based damage detection of rotor blades in a gas turbine engine, Eng. Fail. Anal. 46 (2014) 26 [13] W.W. Robinson, R.S. Washburn, A real time noninterference stress measurement system (NSMS) for determining aero engine blade stress, Proceedings of the 37th Symposium of the Instrumentation Society of America, San Diego, California, 1991. [14] M. Zielinski, G. Ziller, Non-contact vibration and tip clearance measurement system for aero engine application, 18th ISABE Conference, Beijing, China, 2007 . [15] L. Forbes Gareth, B. Robert, Randall non contacting gas turbine blade vibration measurement from casing pressure and vibration signals, Proceedings of the 8th IFToMM International Conference on Rotor Dynamics, Seoul, Korea, 2010. [16] A.R. Rao, B.K. Dutta, In situ detection of turbine blade vibration and prevention, J. Fail. Anal. Prev. 12 (5) (2012) 567–574. [17] A.R. Rao, B.K. Dutta, Vibration analysis for detecting failure of compressor blade, Eng. Fail. Anal. 25 (2012) 211–218. [18] A.R. Rao, B.K. Dutta, Blade vibration triggered by low load and high back pressure, Eng. Fail. Anal. 46 (2014) 40–48. [19] P. Pennacchi, S. Chatterton, N. Bachschmid, E. Pesatori, G. Turozzi, A model to study the reduction of turbine blade vibration using the snubbing mechanism, Mech. Syst. Signal Pr. 25 (2011) 1260–1275.
24