R. Viswanathan J.

Stringer
Electric Power Research Institute, Palo Alto, CA 95070

Failure Mechanisms of High Temperature Components in Power Plants

The principal mechanisms of failure of high temperature components include creep, fatigue, creep-fatigue, and thermal fatigue. In heavy section components, although cracks may initiate and grow by these mechanisms, ultimate failure may occur at low temperatures during startup-shutdown transients. Hence, fracture toughness is also a key consideration. Considerable advances have been made both with respect to crack initiation and crack growth by the above mechanisms. Applying laboratory data to predict component life has often been thwarted by inability to simulate actual stresses, strain cycles, section size effects, environmental effects, and long term degradation effects. This paper will provide a broad perspective on the failure mechanisms and life prediction methods and their signicance in the context utility deregulation. S0094-42890000103-1

Introduction

Many utility power companies in the U.S. are aggressively preparing themselves to face a possible transition from being regulated monopoly companies to deregulated free market competitors. Reducing the cost of power production is paramount for staying competitive in the new scenario. Reducing capital costs by deferring replacement of expensive components and reducing operating and maintenance O&M costs by optimizing operation, maintenance and inspection procedures will both be key strategic objectives for utilities. This poses a signicant challenge to the technical community since two apparently opposing needs will need to be reconciled. On the one hand, the need for improved plant efciency and availability will dictate more severe and cyclic duty schedules which result in more severe creep-fatigue damage and warrant increased attention to the components. On the other hand, the need to reduce O&M costs may result in fewer, shorter, and lower quality maintenance and inspection outages; thus, placing the components at greater risk of failure. The challenge to the technical community, therefore, is to develop tools and techniques that will permit more rapid, cost-effective, and accurate assessment of condition of critical components, both offline and on-line. In addition to assessing the current condition, these tools must also be capable of evaluating the impact of alternative strategies for operation, inspection and maintenance. It is crucial therefore that the high temperature research community be more intimately familiar with the specic needs of the industry. This paper will bring out some of the industry perspectives regarding high temperature failures and illustrate them with some failure examples pertaining to creep, thermal fatigue and embrittlement. A detailed review of the failure mechanisms affecting the integrity of utility and chemical plants can be found in 1.

affects the economics of operation adversely. The blade would therefore be replaced even though it can continue to operate. Another example is the case of defective piping which could continue to be operated by decreasing the temperature or pressure but the resulting loss of output warrants replacement of the piping. Component failures are thus dened in terms of functional rather than structural failures. Most of the time, replacement of parts is based on economic considerations, rather than technical. 2.2 The Cost of Failure. Catastrophic failures of components occur rather infrequently, but when they do, they take a heavy toll on human lives in addition to the costs of repairs, replacement power, and litigation costs. The total cost of the hot reheat pipe failure at the Mojave Power Station in 1986, including litigation costs and downtime costs cleanup, pipe replacement, cost of replacement power is estimated to be in excess of 400 million dollars. This was in addition to loss of several lives. Failure of most stationary components are generally less costly than that of rotating components. For instance, a typical tube leak results in two days of forced outage, while a header or pipe leak may result in four days of outage. Major ruptures in piping may involve several months of outage. In the case of major failures, e.g., steam turbine rotor damage, the consequential damage involving the wreckage of the entire turbine can be severe. Missile generation also needs to be considered. In combustion turbines, breakage of a single rst row blade can damage all other downstream components. In such cases, the unit may be unavailable for nearly six months. Considering that one day of outage can cost $500,000 in lost revenue in a 500 MW plant, six months of downtime costs nearly 100 million dollars. Unavailability costs can skyrocket even more in the deregulated market during peak demand periods in the summer months when electricity prices may be as high as $5000 per MWh in the spot market for short periods. It is important to remember, therefore, that the major cost of failure is invariably the cost of unavailability of the unit, compared to which component replacement costs and repair costs pale into insignicance. 2.3 Relative Importance of Crack Initiation Versus Growth. The general ingredients of a remaining-life-assessment procedure for a commonly encountered failure scenario can be illustrated with the help of Fig. 1. Region I corresponds to incipient, microscopic damage events leading up to the initiation of a macroscopic crack. These events include dislocation rearrangements, coarsening of precipitate phases, and formation of creep cavities and microcracks. Region II corresponds to propagation of the above-mentioned macrocrack. Conventional NDE techniques Transactions of the ASME

2 Some Industry Perspectives on Failure of Components

2.1 Failure Denition. The industrial denition of failure is often quite different from the textbook denition. A component, in practice, is deemed to have failed when it can no longer serve its intended function safely, reliably and economically. Any one of these criteria can constitute failure. For example, a steam turbine blade whose tip has eroded, affects turbine efciency and hence
Contributed by the Materials Division for publication in the JOURNAL OF ENGINEERING MATERIALS AND TECHNOLOGY. Manuscript received by the Materials Division October 15, 1999; revised manuscript received February 15, 2000. Guest Editors: Raj Mohan and Rishi Raj.

246 Vol. 122, JULY 2000

Copyright 2000 by ASME

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Table 1 Examples of circumstances governing crack-initiation and crack-propagation controlled failure

Fig. 1 Illustration of a remaining-life-assessment procedure for a common failure scenario involving crack initiation and propagation. A embrittlement phenomena. B unanticipated factors excess cycling, temperature excursions, corrosion, metallurgical degradation, improper material, excessive stresses. See text for denitions of regions I and II.

apply to crack detection and sizing in region II. In region II, the crack grows until it reaches critical size, dened as a c , at which point rapid fracture occurs. The critical crack size can be dened in a number of ways based on fracture toughness, ligament size, crack-growth-rate transitions, or other considerations as appropriate. A common denition for many heavy-section components is based on the fracture toughness of the material. The critical crack size, a c , is often not a constant value but decreases with service exposure due to embrittlement phenomena. Similarly, many adverse factors can accelerate the stage II crack-growth behavior, so that the failure point is shifted to the leftto shorter times. To perform a remaining-life analysis, information is needed regarding crack initiation, the rate of crack growth, and the failure point, as specically applicable to the component of interest. Conventional NDE techniques in stage II are based on the premise that a detectable crack will form and grow slowly enough to permit periodic inspections and retirement of the component prior to nal failure. There are many instances in which crack initiation alone constitute component failure and conventional NDE techniques and fracture-mechanics analyses serve no useful purpose. This is often a basis of contention between original equipment manufacturers OEM and owners. The OEMs recommend component replacement even when no aws are found, on the premise that the estimated value of a c is below the NDE detection capabilities. An entire vintage of turbine rotors, disks and retaining rings have been replaced en masse on this basis, without rigorous technical justication. Table 1 presents examples of the various circumstances that might dictate whether component failure is governed by crack initiation or crack growth. In the case of very brittle materials, such as the heavily segregated bore of a 30-to-40-year-old rotor, a c may be so small that it is below the limit of detection by conventional NDE techniques. Severely embrittled pressure vessels, bolts, and blades may be other examples of this. High stresses once again have the effect of reducing a c , sometimes below levels of detection. If a component has a thin cross section e.g., a blade or a tube, the remaining ligament can be so small that crack propagation is not of importance. In some instances, a c may be large but the rate of crack growth may be so high that once a crack initiates, it reaches critical size rapidly. Many environmentally induced failures in highly stressed components exhibit this behavior. For instance, in generator retaining rings and in steam turbine blades where crack growth under corrosive conJournal of Engineering Materials and Technology

ditions is encountered, the presence of a pit or pit-like defect is cause for retirement. Initiation of a crack by rapid propagation by fatigue is another example. In components where failure is governed by crack initiation, the detection of any defect during an inspection, or, more conservatively, the suspected initiation of a crack based on calculations, can be used to retire the component. Many advanced NDE techniques which can detect incipient damage evolution prior to crack initiation are under development industry wide. On the other hand, many stationary components such as casings, nozzles and headers are routinely operated with cracks. Techniques that use crack initiation as a failure criterion include calculations based on history, extrapolations of failure statistics, strain measurements, accelerated mechanical testing, microstructural evaluations, oxide scale growth, hardness measurements, and advanced NDE techniques. For crack-growthbased analysis, the NDE information, results from stress analysis, and crack-growth data are integrated and evaluated with reference to a failure criterion. The various techniques and their limitations are described in detail in reference 1. Analytical models combine operating conditions, materials properties and damage rules to estimate the total life consumption of a component. This approach inherently is inaccurate since operating conditions and material properties specic to the component are usually conservative. The damage rules are also not strictly obeyed. A greater accuracy in the calculated results is needed only when crack initiation is the governing mode of failure. On the other hand, most instances of creep fatigue failures have extensive crack propagation lives and lend themselves to periodic monitoring by conventional NDE techniques. In such cases modeling studies to improve the accuracy ad innitum are not needed 2. 2.4 Need to Know Failure Scenario. Although cracks may initiate and propagate by creep or creep fatigue the nal failures may occur by a different mechanism. In thin section components, the critical crack size may correspond to a wall thickness, ligament size or a crack size exceeding which crack propagation is rapid. In heavy section components a frequently encountered failure scenario involves crack initiation and growth under steadystate operating conditions at high temperature followed by nal brittle fracture at low temperatures under start-stop transients. Loading at low temperatures such as hydrotesting of piping and pressure vessels or overspeed testing of rotating components at low temperatures can also cause failure. To illustrate this point, a normal operating sequence, based on the analysis of the Gallatin HP-IP rotor experience is depicted in Fig. 2 3. Region A consists of a warm-up period after which the roll-off commenced Region B. During roll-off, the rotor was gradually brought up to speed. Once the synchronous speed was reached then loading began in Region C. The load was then slowly increased until steady state was reached after a few hours. The estimated values of the temperature and the tangential stresses at the bore region of the rotor at the failure location 7th row as a function of time are illustrated in Fig. 2. The transient JULY 2000, Vol. 122 247

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 3 Typical dissimilar-metal weld locations and failures Fig. 2 An illustration of a cold start sequence and associated variations of stress , temperature T , and critical ow size a c as a function of time from start

stresses reach a peak value of about 520 MPa 74 ksi and level off at a value of 330 MPa 47 ksi at longer times. Approximate values of a c corresponding to the temperature and stress transient depicted in Fig. 2 can be estimated using the lower bound of the KIC data found in the literature. During the warm up and cool-off periods, the stresses are small so that the a c values are large, despite the low values of KIC . Near steady-state conditions, the reduced values of the steady-state stresses compared to the peak transient stresses, coupled with the high values of KIC obtaining at the higher temperature lead once again to large values of a c . The worst combination of stress 520 MPa and fracture toughness 71 MPa m, resulting in minimum values of a c 0.73 cm, occurs at some intermediate time, four hours after loading 3, but prior to reaching steady state.

Examples of High Temperature Failures

Failure mechanisms at high temperatures include creep, thermal fatigue, corrosion, erosion, and hydrogen attack. In addition, embrittlement phenomena occurring at high temperatures, e.g., carbide coarsening, sigma phase formation, temper embrittlement, etc. can facilitate rapid brittle fracture at low temperatures during transient conditions. This section will describe issues associated with creep, thermal fatigue and embrittlement. Overviews on other mechanisms may be found in reference 1. 3.1 Creep. Creep damage can take several forms. Simple creep deformation can lead to dimensional changes that result in distortions, loss of clearance, wall thinning etc. Examples are steam turbine casings, blades, and piping systems. Localized deformation can cause swelling and eventual leaks in headers, steam pipes and superheater reheater SH/RH tubes. Long term creep failures generally tend to be brittle failures involving cavitation and crack growth at interfaces and at highly stressed regions. The cavitation form of damage has been found in SH/RH tubes, rotor serrations, occasionally rotor bores, highly stressed areas in piping systems and at weldments. The most common weld failures have pertained to dissimilar welds in superheater/reheater tubing, welds in headers and in hot reheat and mainsteam piping. 3.1.A Dissimilar Metal Weld Failures. Dissimilar metal welds are used to join ferritic steel to austenitic stainless-steel tubing and piping in many high-temperature applications in energy conversion systems. The most widespread application is in the superheaters and reheaters of fossil-red electric power generation boilers. The materials joined in this application are usually low alloy ferritic steels such as 1-1/4Cr-1/2Mo or 2-1/4Cr-1Mo to austenitic stainless steels such as 304H, 316H, 321H, or 347H. An important feature of these welds is that the materials joined have signicantly different metallurgical and physical properties. For example, the materials differ greatly in thermal expansion coef248 Vol. 122, JULY 2000

cient which imposes additional stresses. The welds may be made either by induction pressure welding the two materials directly together or by shielded metal arc welding with a gas tungsten arc root pass using austenitic stainless steel ller metals or nickelbase ller metal. Figure 3 illustrates a typical DMW failure. Welds of this type have been in use for a great many years. During the late 1950s, failures were encountered with a number of welds made using austenitic stainless-steel ller. Typically, failure occurred by the development of low ductility cracking in the lowalloy steel very close to the weld fusion line. These experiences prompted research to nd better methods of joining the dissimilar materials. This led to the use of nickel-base ller metals. The use of nickel-base llers to make DMWs effected a considerable life improvement. However, by the mid-1970s failures were also beginning to occur at a signicant rate in DMWs made with nickel-base ller. These failures were generally macroscopically similar in appearance to the failures that occurred in DMWs made with austenitic stainless steel ller metals. The forced outage costs resulting from these increased failure rates caused increased utility attention, worldwide. In response to this growing concern, EPRI Research Project 1874 was developed. As a result of this project our understanding of the failure causes and corrective actions is fairly complete. Inspections procedures and life assessments codes for DMWs have also been developed. A multi volume report describes these results 4, which have also been widely disseminated in the public domain 5. 3.1.B Weld Failures in Headers. A schematic illustration of a header is shown in Fig. 4. A header is essentially a pipe to which tubes are welded, spaced either axially or circumferentially. The spacing between the tubes is known as the ligament. In addition to the tubes, other pipe-to-pipe connections are also present, either integral with the header pipe or welded to it. These branch connections can be T-shape connections, as shown in Fig. 4, or of a Y-shape conguration. Numerous pipe-to-pipe and pipe-to-tube weldments are normally present, as shown in the gure. Initial signs of creep-related distress in headers often appear at weldswelds at stub-tube inlets, long seams, header branch connections or girth butt joints. With the exception of some cases of long seam welds, and Type IV cracks in girth welds, creep damage in welds is invariably manifested on the outside surface as cavities, cracks, or, in extreme cases, steam leaks. Except in regard to long seam welds, concern about catastrophic bursts has been minimal. Although weld-related cracking is generally detectable and repairable, and although it does not have as great an impact on the over-all component life as does header-body basemetal deterioration, it is important from a life-assessment point of view for the following reasons: Because weld failures are often the forerunners of damage in the body, they can provide an index of creep damage and remaining life in the base metal. Failure of welds at crucial and multiple locations may constitute the end of Transactions of the ASME

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 4 Schematic illustration of an elevated-temperature header courtesy of B. W. Roberts, Combustion Engineering, Inc.

the life of the header, regardless of the condition of the base metal. The need for frequent weld repair may prove uneconomical and justify retirement of a header. Due to the above reasons, creep-damage assessment of welds has received considerable attention. Damage characteristics at various locations are briey reviewed here on the basis of the extensive information contained in reference 6. Stub-Tube Welds. Cracking at both the tube and header sides of stub-tube welds is the most common type of creep damage in high-temperature headers. Although such cracking may lead to steam leakage and forced outages, it is easily detected and repaired. The cracking may be attributed to any one of several causes, including improper seating of the stub tube, inadequate tube exibility, improper support of the header, bowing, weldfabrication defects, and locally excessive temperatures. Metallography in several instances has shown the creep damage to consist of cavitation and microcracking at the prior austenite grain boundaries in the heat-affected zone. Longitudinal Seam Welds. Plate-formed and seam-welded headers have been used in some designs. Detailed failure reports are available for only one incidence 6,7, in which a crack 864 mm 34 in. long, in the weld seam had led to a major leak in a secondary superheater outlet header made of 2-1/4Cr-1Mo steel after 187,000 hours of service. This failure occurred as a result of creep-rupture. The damage was conned to the weld metal, with no evidence of damage in the heat-affected zone or base metal. The cracking had apparently initiated just below the outer surface, broke through to the outer surface at an early stage, and then propagated to the inner surface. The reason for the subsurface crack initiation was believed to be the inferior metal properties at the location due to a lower carbon content and tempering of the weld head by subsequent passes. The most severe cracking occurred at the centerline of the weld beads, presumably as a result of impurity segregation during the last stages of solidication. Boat-shaped samples removed at locations away from the cracked area showed several degrees of cavitation. For each location, a life fraction consumed could be estimated on the basis of the model and the plots described later in this chapter. These values were borne out by subsequent isostress-rupture tests of samples from the various locations 6. More recently another failure has been reported presumably due to locally high temperatures 8. Girth Welds. Four types of creep damage and cracking associated with weldments for both headers or piping have been Journal of Engineering Materials and Technology

cataloged by Chan et al. 9. Each of the four creep damage types are identied below and shown schematically in Fig. 5. Type IDamage which is longitudinal or transverse in the weld metal and remains entirely within the weld metal. Type IIDamage that is longitudinal or transverse in the weld metal, but grows into the surround HAZ. Type IIIDamage in the coarse-grained region. Type IVDamage initiated or growing in the intercritical zone of the HAZ the transition region between the fully-transformed, ne-grained HAZ, and the partially-transformed parent base metal. Both axial and circumferential cracks have been observed in damaged girth butt welds, with cracking being found in the weld metal and/or the HAZ. The axial cracking has been attributed to internal pressure loading and pipe swelling, whereas the circumferential cracking has been associated with combined pressure and piping system loads. Several instances of girth weld cracking has been reviewed 6. In one instance, circumferential cracking along the coarse-grain HAZ was attributable to stress-relief cracking prior to service. Axial creep cracking across the weld metal has been attributed to a combination of pipe swelling and poor weld ductility. Circumferential cracking in the intercritical regions of the HAZ has also been observed in both Cr-Mo-V and Cr-Mo steels. This type of cracking, known as Type IV cracking, occurs at the end of the HAZ adjacent to the unaffected parent metal. Type IV cracking is generally attributed to localized creep deformation in a soft zone in the intercritical region under the action of bending stresses. Field experience suggests that Cr-Mo-V steels may be more susceptible to cracking than Cr-Mo steels and that operation at 565C 1050F rather than at 540C 1000F might further exacerbate the problem. Because most of the headers in the United States are made of Cr-Mo steels and operate at 540C 1000F, the problem has not been encountered to any signicant degree.

Fig. 5 Four types of damage in girth welds in relation to microstructure 8

JULY 2000, Vol. 122 249

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 6 Creep cavitation in a T-section of a ferritic steel desuperheater header in a utility boiler

Branch-Connection Welds. Several instances of cracking in branch-connection welds have been observed. Such cracking has occurred on both the header side and the branch side 6, and in both the HAZ and the weld metal. An example of creep cavitation in the CGHAZ of a header T section is shown in Fig. 6. Summary of Creep Cracking in Header Welds. Numerous instances of cracking at various locations in header welds, as described above, have been reviewed by Ellis et al. 6. The salient facts brought out in this review are as follows. 1 Most weld failures are creep failures and are clearly evidenced by creep cavitation. 2 Cracks can occur in the weld metal, in the coarse-grain HAZ, or in the intercritical zone of the HAZ Type IV cracking. 3 Cracks in the weld metal are generally attributable to lower strength or lower ductility of weld metal. 4 Cracks in the HAZ can arise as a result of hoop stresses, system bending stresses, or residual stresses due to stress relief. 5 Frequently, the direction of alignment of creep cavities, which is normal to the tensile loading direction, gives a clue to the nature of the system stresses involved. 3.1.C Failures in Seam Welded High Energy Piping. Several categories of pipes carrying high temperature/pressure steam contain welds that may be of concern. Main steam pipes are pipes that carry steam at 538565C to the high pressure turbine. These pipes are small in diameter and do not contain seam welds. Hence, only girth welds are of concern. The mainstream pipes are however, often connected to the steam header using thick-walled seam welded piping. In addition, hot reheat pipes which carry steam at 538565C but at a lower pressure than the main steam pipe to the reheat IP turbine, and are frequently made of seam welded piping. Failure of seam welded pipes used in HRH piping as well as in header link piping has been of major concern to industry. Failure experience with respect to high energy piping has been reviewed by Wells and Viswanathan 10. There have been at least 17 major instances of seam welded pipe failures including 3 cases of catastrophic rupture, 5 leaks and 9 incidents of major cracking. An example of a catastrophic failure is shown in Fig. 7. The failures are generally brittle with a sh mouth appearance. In the cases of HRH pipes, the welds generally have a double V conguration and the pipes are generally subjected to a normalizing and tempering treatment. The cracking generally initiates subsurface at the cusp of the double V and then propagates along the fusion line toward the outside and inside, as shown in Fig. 8. In the case of the thicker walled header leak pipes, the weld generally has a U geometry and is subjected to subcritical PWHT. A variety of cracking modes, including fusion line, Type I and Type 250 Vol. 122, JULY 2000

Fig. 7 Rupture in Monroe No. 1 north hot reheat line

IV cracking have been observed. Failures of most of the seam welded piping have occurred prematurely and could not be predicted based on simple life-fraction rule calculations. Failures occur due to unique combination of operating and metallurgical variables. Some of the contributing factors have been identied to be operating temperature, pressure and cycling; system stresses; weld geometric factors such as conguration, cusp angle and roof

Fig. 8 Macrograph of cross-section at location 6LS1, counterclockwise side of weld sighting along ow; note ID-connectedcracking, located and detected by UT, and extent of cusp damage

Transactions of the ASME

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 9 Creep-life assessment based on ca?? classication

Fig. 10 Correlation between damage classication and expended creep-life fraction for 14Cr-12Mo steels

angle; welding practice employed; inclusion content and creep strength mismatch, etc. Currently two failure scenarios have been postulated. In one scenario, failure is proposed to involve crack initiation and propagation stages. In the alternative scenario cavities form and grow and eventually link up into a larger crack. Which of these is operative can determine whether NDE based monitoring is viable. A comprehensive review of the subject may be found elsewhere 11,12. A review of literature on Type IV cracking in girth welds and seam welds may be found in reference 13. Since in many of the early instances of girth weld damage, the damage consisted of evolution of creep cavities into cracks at the coarse grained heat affected zone CGHAZ, assessment of damage consisted of simply classifying the damage and then recommending an appropriate action. Damage was classied as A isolated cavities, B oriented cavities, C linked cavities and D microcracking, as per the German practice, Fig. 9 14. More quantitative correlations between the degree of cavitation and the

creep life expended have been established based on EPR1 research. The results shown in Fig. 10 have provided a clearcut basis for establishing re-inspection intervals. This approach is however valid only for Type III cracking in the CGHAZ. The evolution of damage in the other cases have not been sufciently investigated. While replication is very useful for detecting surface damage, many types of failures such as long seam weld and Type IV damage in girth welds originate sub-surface. In these cases, replication alone is not a reliable method to detect damage. In long seam welds in hot reheat piping and header link piping, high sensitivity conventional or automated UT, focused beam UT or time-of-ight diffraction UT methods are needed to ensure safety of the piping. In the case of girth welds however, conventional UT seems to be adequate. Some forms of creep damage are more manageable than others. For example, if Type I, II, or III creep damage is found, the

Table 2 Fossil power plant components involving creep-fatigue as a common failure cause

Journal of Engineering Materials and Technology

JULY 2000, Vol. 122 251

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

subsequent action can range from record and monitor to some form of repair depending on the severity of damage. Advanced Type IV damage is characterized by profuse intergranular cavitation in the creep weak area of the HAZ. It has been suggested that the evolution of damage from the observation of cavitation by replication to macro-cracking can be swift and cannot be dealt with using the German system. In the absence of enough experimental evidence regarding damage evolution, the current approach is to replace completely the affected weldment, if any stage of Type IV damage is conrmed 13. 3.2 Creep-Fatigue Failures. Creep-fatigue damage induced by thermal stresses is of major concern with respect to the integrity of many high temperature components. The concern has been exacerbated in recent years due to cyclic operation of units originally designed for base load service. Table 2 is a sample list of fossil plant components in which creep-fatigue has been a dominant failure mode. The list is by no means complete, since many components not included here may also become subject to creep-fatigue if more severe cycling conditions were imposed upon them. The purpose of the table is to make several key points as follows: 1 Creep fatigue damage is generally the result of thermal stresses induced by constraint to thermal expansion during transient conditions. The constraint may be internal such as in the case of heavy section components e.g., rotors, headers, drums, casings where thermal gradients arise between the surface and the interior or vice versa. Internal constraint may also arise from internal cooling of components subject to rapid surface heating such as in combustion turbine CT blades. The constraint may be external such as in the case of joining of thick sections to thin sections or of materials of different coefcients of thermal expansion dissimilar metal welds. Since stresses are always thermally induced, and since crack initiation occurs in less than 103 cycles, this form of creep fatigue damage is also referred variously as thermal fatigue, thermomechanical fatigue and low-cycle fatigue. These terms will be used interchangeably in this paper. 2 This form of creep fatigue damage may involve large plastic strains achieved locally at stress concentrations such as rotor grooves, header bore holes, etc. It may also involve primarily elastic strains combined with stress relaxation, as occurs for combustion turbine blades. 3 Table 2 also shows that the industry view of what constitutes

failure is different for stationary components such as headers and casings and rotating components such as blades/rotors. In the former case, cracks are tolerated and crack initiation is believed to occur early 1020 percent life in life-component retirement is therefore based on economics of repeated repairs and growth of a crack to a critical allowable size. Hence, excessive concern with rening the damage rules is unwarranted in such cases. In rotating components, such as CT blades and rotor grooves, crack initiation denes failure since upon crack initiation other failure modes such as high cycle fatigue may intervene and cause rapid failure. In these cases, more rened prediction of damage evolution and crack initiation would be useful. Detailed review of literature shows that there are divergent opinions regarding which damage approach provides the best basis for life prediction. It is quite clear that a number of variables, such as test temperature, strain range, frequency, time and type of hold, waveform, ductility of the material, and damage characteristics, affect the fatigue life 2. The conclusion drawn in any investigation may therefore apply only to the envelope of material and test conditions used in that study. The validity of any damage approach has to be examined with reference to the material and service conditions relevant to a specic application. Broad generalizations based on laboratory tests, which often may have no relevance to actual component conditions, do not appear to be productive. Thus, one should use a tailored, case-specic approach for any given situation. In most instances of fatigue, the temperature varies along with the strain, giving rise to what is known as thermomechanical fatigue TMF. Depending further on when the hold time is superimposed, various cycle shapes are possible. In the past, thermal fatigue traditionally has been treated as being synonymous with isothermal LCF at the maximum temperature of the thermal cycle. Consequently, life-prediction techniques have evolved from the iso-thermal LCF literature. The assumed equivalence of isothermal LCF tests and TMF tests has been brought into questions as a result of a number of studies. High tensile strains at high temperature IP would favor creep, whereas high tensile strains at low temperature OP would favor cracking of oxide and hence accelerated environmentally induced damage during subsequent high-temperature exposure. Hence, Kuwabara et al. rationalized that in case of materials where damage is driven by creep, IP cycles would be more damaging than

Fig. 11 Ligament cracking at a tube bore hole viewed from the ID of a header

252 Vol. 122, JULY 2000

Transactions of the ASME

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 12 Oxide notching at ligament cracks

OP cycles, for a given strain range 15. In other materials, where the environmental contribution is signicant, OP cycles may be more damaging than IP or isothermal LCF cycles. In addition to environmental effects, differences also arise between cycles in terms of the relaxed mean stresses. The relative severity or the

different cycles can also change with material ductility, maximum temperature and hold time. Consequently, a simple classication of material behavior is not possible. A case in point is the ligament cracking encountered in CrMo steel header pipes illustrated in Fig. 11. Cracks initiate in the tube bore holes and are oriented parallel to the axis of the tube bore hole. Linking up of cracks between holes on the inside surface of the header leads to propagation to form cross ligament cracks. Presence of ligament cracking has been observed in a very large number of superheater headers in the U.S. The cracking mode has been identied as creep fatigue. A computer code, B oiler L ife E valuation and S imulation S ystem BLESS developed recently, incorporates two alternate approaches for predicting crack initiation; one involving an inelastic linear damage summation method, and a second approach involving repeated cracking of oxide scale and oxide notching 16. For a variety of cycle histories, the Code predicts crack initiation occurring in about 20,000 h by the oxide cracking mechanism. The creep-fatigue damage summation approach on the other hand, is inconsistent with the early initiation of cracks observed in headers. Metallography of cracked headers has shown numerous oxide spikes, see Fig. 12, indicating oxide cracking to be the crack initiation mechanism. This example clearly illustrates the need for using appropriate thermomechanical fatigue data simulative of actual component cycles in predicting crack initiation life of components. Another example of the critical need for TMF data is in the case of protective coatings. In the case of coated components such as combustion turbine blades, cracking of the coating leads to loss of environmental protection from the coating, and, eventually, to cracking of the base metal Fig. 13. The integrity of the coating depends upon both the ductility of the coating and the strain-time history of the coated blade, as shown in Fig. 14. The strain-tocracking of the coating is a strong function of temperature, often given by a ductile-to-brittle transition temperature DBTT curve. The strain-temperature cycle in the engine must lie below the DBTT curve. For the example shown in Fig. 14, coating A will not crack under normal duty, but will crack during an emergency shutdown. However, coating B will not crack under either condition. In view of the DBTT behavior exhibited by coatings, it becomes even more critical in evaluating coated components that

Fig. 13 TMF cracks in GT29 CoCrAIY crating penetrating the INCO 739 base metal

Journal of Engineering Materials and Technology

JULY 2000, Vol. 122 253

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 14 Typical thermomechanical cycle for a rst-stage blade, showing leadingedge strain and temperature variations for normal start-up and shut-down, and an emergency shutdown

thermo-mechanical fatigue tests be performed. Furthermore, simple TMF cycles, in which the maximum tensile strain is made to coincide with either the peak temperature in-phase, IP or with the lowest temperature out-of-phase, OP in the cycle, will lead to unrealistic results. If one were to compare the performance of two coatings A and B by conducting an isothermal LCF test or an IP type thermomechanical fatigue strength, one would conclude that both coatings perform well, since both coatings have high strain capability at high temperature. If the same comparison were made under more relevant TMF conditions, coating B would be chosen over coating A. Hence, TMF cycles simulative of actual blade cycles must be performed to evaluate the effect of the coating. Such test data on coated components is extremely scarce in the open literature, and is limited even in proprietary data bases. One of the major problems in evaluating the applicability of different life-prediction methods is that in many cases it is necessary to use all the available data in deriving the life-prediction method, and thus it is not possible to examine the accuracy with which a given method describes data not used for the development of the method, or outside of the range of conditions, or lives, considered. There also is a scarcity of instances in which service experience has been compared with specic life-prediction methods. In general, the available methods are utilized only to predict the lives of samples tested under laboratory conditions. Validation against component test data in the laboratory and inservice monitoring of actual equipment would lead to more condence in the use of the various rules. Results from most studies show that even the best of the available methods can predict life only to within a factor of 2 to 3. Some of the cited reasons for these inaccuracies have already been discussed. Some additional reasons are: failure of the methods to model changing stress-relaxation and creep characteristics caused by strain softening or hardening, use of monotonic creep data instead of cyclic creep data, and lack of sufciently extendedduration test data. All of the damage rules available today are at least partly, if not totally, phenomenological in nature. They all involve empirical constants that are material-dependent and difcult to evaluate theoretically. Extrapolation of the rules to materials and conditions outside the envelope covered by the specic investigation may result in unsuccessful life predictions and usually result in predictions whose accuracy is difcult to evaluate. One form of extrapolation that is especially difcult to evaluate is 254 Vol. 122, JULY 2000

the need to use short time data for long time service. For components whose service time is from 3 to 30 years, and which may operate continuously for hundreds to thousands of hours, use of test datawhich lasts as much as four months 1/3 of a year and has hold times as much as 16 hoursrequires extrapolation of one to two orders of magnitude. The statistical methods chosen to make these extrapolations signicantly affect the estimated lives of the components. Furthermore, in performing these accelerated tests, either the strain, the temperature, the frequency, or some combination must be increased over actual service conditions in order to produce failure in a reasonable amount of time. Thus, there is always the danger that the physical mechanism of failure in the laboratory test is different from that during service, or that important aspects of the service conditions are not considered during the laboratory tests. For application to service components, the stress-strain variation for each type of transient and its time dependence must be known with accuracy. The importance of using relevant TMF data cannot be overemphasized. This realization has led to several recent studies in life prediction of combustion turbine components using TMF based algorithms.

Summary and Conclusions

Creep and creep-fatigue are the principal failure mechanisms affecting the integrity of components operating at elevated temperatures. Creep damage in weldments poses major challenges both in analytically calculating it and in experimentally reproducing it. Several alternative damage locations and mechanisms have been observed which are often difcult to reproduce in laboratory tests. Fusion line cracking and ne grain heat affected zones FGHAZ cracking has led to catastrophic failure of high energy piping. Thermomechanical fatigue TMF or creep fatigue affects many heavy section components as well as internally cooled components such as combustion turbine blades. It is important that researchers focus on component specic rather than generic life prediction models with a full understanding of the applicable failure denition, failure scenario and relevant duty cycle. Future research needs to address advanced NDE techniques, on-line monitoring techniques, TMF mechanisms, and evolution of damage and growth of cracks in welds. Transactions of the ASME

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

References
1 Viswanathan, R., 1987, Damage Mechanisms and Life Assessment of High Temperature Components, ASM International Metals Park, OH. 2 Viswanathan, R., and Bernstein, H., 1996, Some Issues in Creep Fatigue Life Predictions of Fossil Power Plant Components, ASME PVP, Vol. 335, Service Experience and Design In Pressure Vessels and Piping, W. H. Barnford, ed., Book No. H01063, pp. 99119. 3 Viswanathan, R., and Jaffe, R. I., 1983, Toughness of Cr-Mo-V Steels for Steam Turbine Rotors, ASME J. Eng. Mater. Technol., 105, pp. 286294. 4 Roberts, D. I., et al., 1985, Dissimilar Weld Failure Analysis and Development Program, Final Report CS-4252, Vols. 1-7, Electric Power Research Institute, Palo Alto, CA. 5 Roberts, D. I., Ryder, R. H., and Viswanathan, R., 1985, Performance of Dissimilar Welds in Service, ASME J. Pressure Vessel Technol., 107, pp. 247254. 6 Ellis, F. V., et al., 1988, Remaining Life Assessment of Boiler Pressure Parts, Final Report RP2253-1, Vol. 1-5, Electric Power Research Institute, Palo Alto, CA. 7 Henry, J. F., et al., Failure Investigation of Longitudinal Seam Welded Elevated Temperature Header, Microstructural Science, M. E. Blum et al., eds., 15, ASM International, pp. 150169. 8 Hickey, J. J., et al., 1995, Investigation and Repair of a Failed Seam Welded Reheat Outlet Header, Proc. of Conf. Welding and Repair Technology for Power Plants, Daytona Beach, Electric Power Research Institute, Palo Alto, CA, May.

9 Chan, W., McQueen, R. L., Prince, J., and Sidey, D., 1991, Metallurgical Experience with High Temperature Piping in Ontario Hydro, ASME PVP, Vol. 21, Service Experience in Operating Plants, ASME, New York. 10 Wells, C. H., and Viswanathan, R., 1993, Life Assessment of High Energy Piping, Technology for the 90s, M. K. Au-Yang et al., eds., ASME Pressure Vessels and Piping Division, New York, pp. 179216. 11 Viswanathan, R., and Foulds, J., 1995, Failure Experience with SeamWelded Hot Reheat Pipes in the USA, ASME PVP, Vol. 303, Service Experience, Structural Integrity, Severe Accidents and Erosion in Nuclear and Fossil Plants, S. R. Paterson et al., eds., ASME, New York, pp. 187207. 12 Foulds, J. R., Viswanathan, R., Landrum, L., and Walker, S. L., 1995, Guidelines for the Evaluation of Seam Welded High Energy Piping, Report TR104631, Electric Power Research Institute, Palo Alto, CA. 13 Ellis, F., and Viswanathan, R., 1998, Review of Type IV Cracking in Welds, ASME PVP Conference, July 1998, PVP, Vol.380, Fitness for Service Evaluation in Petroleum and Fossil Plants, pp. 5976. 14 Neubauer, B., and Wedel, V., 1983, Rest Life Estimation of Creeping Components By Means of Replicas, Advances in Life Prediction Methods, D. A. Woodford and J. R. Whitehead, eds., ASME, New York, p. 307. 15 Kuwabara, K., Nitta, A., and Kitamura, T., 1985, Advances in Life Prediction, D. A. Woodford and R. Whitehead, eds., ASME, New York, pp. 131141. 16 B oiler L ife E valuation and S imulation S ystem, BLESS Code and User Manual, 1991, Report TR-103377, Vol. 4, Electric Power Research Institute, Palo Alto, CA.

Journal of Engineering Materials and Technology

JULY 2000, Vol. 122 255

Downloaded 14 Jan 2009 to 132.248.9.103. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm