Department of Engineering, University of Cambridge, Trumpington Street, Cambridge CB2 1PZ, UK

2004 Elsevier Ltd. All rights reserved.

consequences. It is estimated that 10% of all power-plant breakdowns are caused by creep fractures of

boiler tubes. In general, 30% of all tube failures in boilers and reformers are caused by creep. This paper

gives details of four case studies in which internally pressurised tubes failed by creep bulging and rupture(two boilers, one superheater and one reformer). The conditions of temperature and time under which the*Keywords: Boiler failures; Boiler tube; Chemical-plant failures; Creep; Creep rupture

1. Introduction

Internally pressurised tubes are critical components in heat-exchanger applications, such as boiler water

tubes, steam superheater elements and chemical plant reformer tubes. Tubes in such applications are

vulnerable to temperature excursions: as a consequence the material may enter the creep regime, and creep

deformation (bulging) and even fracture (longitudinal rupture) may subsequently occur, with seriousReceived 1 February 2004; accepted 8 March 2004

Available online 21 July 2004

Abstract

Internally pressurised tubes are critical components in heat-exchanger applications, such as boiler water tubes, steam

superheater elements and chemical plant reformer tubes. Tubes in such applications are vulnerable to temperature

excursions: as a consequence the material may enter the creep regime, and creep deformation (bulging) and even

fracture (longitudinal rupture) may subsequently occur, with serious consequences. It is estimated that 10% of all

power-plant breakdowns are caused by creep fractures of boiler tubes. In general, 30% of all tube failures in boilers and

reformers are caused by creep. This paper gives details of four case studies in which internally pressurised tubes failed

by creep bulging and rupture (two boilers, one superheater and one reformer). The conditions of temperature and time

under which the failures occurred are deduced from the morphology of fracture and the changes in microstructure, and

are correlated with the deformation-mechanism and fracture-mechanism maps for the tube materials.Creep failures of overheated boiler, superheaterand reformer tubes

D.R.H. Jones *

Engineering Failure Analysis 11 (2004) 873893

www.elsevier.com/locate/engfailanalTel.: +44-1223-332600; fax: +44-1223-332662.

E-mail address: drj1000@eng.cam.ac.uk (D.R.H. Jones).

1350-6307/$ - see front matter 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.engfailanal.2004.03.001

failures occurred are deduced from the morphology of fracture and the changes in microstructure, and are

correlated with the deformation-mechanism and fracture-mechanism maps for the tube materials.

2. Case study 1: creep rupture in a water-tube boiler

2.1. Description of boiler

The rst case study describes the investigation of the creep rupture of a water tube in a large water-tube

boiler, which resulted in considerable consequential damage and down-time. As part of the investigation, it

was considered necessary to establish the temperature at which the failure occurred, to estimate the time it

took for the tube to burst, and to correlate these ndings with the operating records of the boiler.

Fig. 1 is a schematic of the boiler. For simplicity, the gure shows only two banks of hot tubes (risers)

and two banks of cool tubes (downcomers). In the actual boiler, there were 12 banks of risers and 12 banksof downcomers. Each bank consisted of a parallel array of 90 tubes spaced regularly along the length of the

drums. The total number of tubes in the boiler was 2160. The drums were 17-m long and 1.7 m in diameter.

The average length of the water tubes was 10 m. Technical data are as follows:

2.1.1. Boiler operating parameters

Maximum evaporation rate: 500 tonnes h1,Maximum heat ux through riser walls: 20 kWm2,Gas temperature at inlet: 940 C,

874 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893Fig. 1. Schematic of the water-tube boiler.

Gas temperature at mid-space: 650 C,Gas temperature at outlet: 490 C,Operating pressure: 4.8 MPa gauge,

Normal operating temperature: 264 C.

2.1.2. Water-tube specications

Internal diameter: 80 mm,

External diameter: 90 mm,

Wall thickness: 5 mm,

Operating hoop stress: 38 MPa,

Material: carbon steel,

Fabrication: single longitudinal diusion weld,

Composition (wt%): 0.17 max C, 0.35 max Si, 0.400.80 Mn, 0.045 max S,Mechanical properties (at room temperature): yield stress 235 MPa/min, tensile strength 360480 MPa,

tensile ductility 25%/min.

2.2. Description of failure

Fig. 2 is a diagram of the ruptured tube. The outside diameter of the tube had increased from the original

value of 90 mm to a maximum of about 103 mm. A bulge had formed on one side of the tube and the wall

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 875Fig. 2. Diagram of the ruptured water tube. Dimensions in mm.

90 of the neighbouring tubes. Another 50 tubes had enlarged diameters. The worst aected had swollen

2.3. Causes of overheating

Th

possible heat transfer. Heat is supplied to the outer surfaces of the tubes by the ue gases and removed from

The term 1=AdQ=dt is the heat ux through the tube wall, which is at most 20 kWm2. A rough es-

timate for the thermal conductivity K of the layer is 0.5 Wm1 C1 [1]. Since dx 0:25 mm, dT 10 C. Itis unlikely that this small thermal resistance would have allowed the tubes to heat up much. A more likely

explanation for the temperature excursion is a major disruption in the circulation and boiling behaviour

aecting a large number of the riser tubes.

2.4. Metallurgy of failure

A sample of steel was removed from the tube just outside the damaged area. The sample was analysed

chemically and specimens from it were tested in tension. The results gave the following information.

Composition (wt%): 0.13 C, 0.18 Si, 0.65 Mn, 0.02 S, 0.02 P, 0.01 Cr, 0.01 Mo, 0.005 Ni.

Mechanical properties (at room temperature): yield stress 240 MPa, tensile strength 505 MPa and tensilethe inner surfaces by the water circulating around the boiler. Obviously, anything which interferes with the

cooling action of the water will lead to overheating. Deposits of hard-water scale, layers of corrosion

product and delaminations in the tube wall can provide disastrously eective thermal barriers. In many

water-tube boilers, the conditions for circulation may be marginal. Risers and downcomers may become

mixed up and there can be regions of the boiler where the water is not circulating. Water-tube boilers are

also prone to steam blanketing, where a stable layer of steam forms between the water and the innersurface of the tube. This helps to insulate the tube from the circulating water and allows it to warm up. The

situation can become unstable: as the tube warms up the steam blanket will tend to grow and the whole

tube can boil dry. Once this happens, there is almost nothing to prevent the tube from heating up to the

temperature of the surrounding gases. Worse, if a large number of tubes boil dry, the furnace gases will not

be cooled by the risers and the temperature of the gas around the dried-out tubes can in principle be as high

as the inlet temperature of the ue gases (940 C in the present case).The temperature drop across the layer of hard-water deposit can be estimated from the heat conduction

equation

dQdt KAdT

dx: 1ductie feature common to all water tubes is that they are meant to operate under conditions of the highestover a length of 1 m. The maximum diameters were 98103 mm. The tube walls had thinned down from the

original 5 mm to 4.6 and 4.3 mm, respectively.

Apart from a thin layer of oxide scale, the tubes were clean on the outside. On the inside they were

coated with a layer of hard-water deposit which was 0.25-mm thick.at the centre of the bulge had thinned down to 3.2 mm. At this position, a chisel-edged creep fracture had

occurred, giving a fracture surface 0.350.80 mm wide. The length of the rupture was 370 mm.

The burst tube was in the sixth bank of riser tubes counting in from the inlet side of the boiler. Under

normal operating conditions the gas temperature at this position would have been about 795 C. As a resultof the rupture the contents of the boiler were discharged into the space occupied by the hot tube banks.

Thirty tubes were cut out in order to gain access and it was found that the force of the explosion had bent

876 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893lity 25%.

The longitudinal diusion weld was barely visible even on a polished cross-section and consisted of avery narrow band with a ne grain size. The split did not line up with the weld. The chemical composition

and mechanical properties were consistent with the specications for the tube. Thus, there was no indi-

cation that the failure was caused by any defect in the material itself.

The Vickers hardness was measured on the cross-section shown in Fig. 2. Next to the fractured edge the

hardness was 250400 HV. Away from the fracture the hardness of the cross-section was 200250 HV.

When the tube was manufactured the structure would have consisted of grains of ferrite a plus nodulesof pearlite in the weight ratio of 88 to 12. As soon as the steel is heated above the A1 temperature (723 C),the nodules of pearlite are replaced by an equal weight of austenite c grains. As the temperature risesabove A1 the a progressively transforms to c until, at the A3 temperature (860 C for the present 0.13%carbon steel), the structure is all c.

As shown in Fig. 3, if c containing 0.13% carbon is quenched to martensite, it will have a Vickershardness of 400. This is the same as the maximum hardness next to the fractured edge, so the rupture must

Fig. 3. Hardness of plain carbon martensite as a function of carbon content.

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 877have occurred above the A3 temperature of 860 C. Obviously, when the water poured out of the boilerthrough the hole in the tube it must have quenched the c to martensite. The large variation in hardness overthe tube cross-section (200400 HV) was caused by dierent cooling rates. Only at the fracture, where the

metal is very thin (less than 1 mm) was the quench fast enough to turn the c to martensite. Away from thefracture the metal was much thicker (up to 4.3 mm). The cooling rate would then have been less and the cwould have transformed to the softer bainite instead.

2.5. Creep-fracture mechanisms

The way in which a creeping material fractures can often provide information about the conditions

under which the failure took place. As shown in Fig. 4, there are three basic mechanisms of creep failure. At

low stresses the most likely mechanism is intergranular creep fracture. In this mechanism voids or wedge-like cracks nucleate at grain boundaries under the action of the applied tensile stress. The defects grow and

eventually the remaining ligaments fail. The deformation is concentrated at the grain boundaries and the

overall ductility and reduction in area at break are usually small. At high stresses failure usually occurs by

transgranular creep fracture. Voids nucleate and grow throughout the grains. Failure takes place by mi-

crovoid coalescence in a way which is similar to ductile failure at ordinary temperatures: the tensile ductility

and reduction in area at break are usually quite large. At high temperatures and stresses, dynamic

recrystallisation operates. Waves of recrystallisation sweep through the creeping material and continually

remove the microstructural damage caused by the creep process. As a result, voids do not nucleate and the

metal breaks down to a point (if the specimen were a round bar to begin with) or a chisel-edge (if the

specimen were initially a at plate). This mechanism is referred to as rupture. Note that the creep literature

often uses the term rupture to mean failure by any creep mechanism, which is confusing.

Fig. 4. Microstructural and fractographic features of creep fracture mechanisms.

878 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893As Fig. 2 shows, the reduction in area at the split is considerable. The width of the fracture surface isonly 0.350.80 mm compared to the initial wall thickness of 5 mm. Thus, the thickness of the plate has

decreased by 614 times. The fracture is very chisel-edged and rupture seems the likely mechanism.

The regimes of stress and temperature for the three mechanisms of fracture can be displayed on a

fracture-mechanism map [24]. Figs. 5 and 6 show the maps for pure iron and low-alloy CrMo steel

containing 0.13% carbon [3]. The vertical axis plots rn=E, where rn is the nominal tensile stress in a uniaxialtensile creep test and E is Youngs modulus. The horizontal axes plot either the temperature in celsius (topaxis) or the temperature in kelvin divided by the melting temperature in kelvin (bottom axis). Because Edecreases with temperature, a line of constant tensile stress will rise as it goes from left to right across thediagram. The maps are complicated by the polymorphic phase transformations that take place in iron. Pure

iron transforms from bcc to fcc at 914 C and transforms back to bcc at 1391 C. A 0.13% carbon steeltransforms from ferrite + pearlite to austenite over the range 723860 C. Each diagram is therefore like twoseparate adjoining maps, one for ferrite (plus pearlite if the metal contains carbon) and one for austenite.

Because of this, the diagrams actually contain ve creep-mechanism elds.

Creep data at very high temperatures are sparse and there is no map for a steel similar to that used for

the tubes. However, the creep behaviour of the tubes is bracketed by the behaviour of pure iron and low-

alloy steel: pure iron contains no alloying elements and will creep faster than the tube steel; low-alloy steel ismore highly alloyed than the tube steel and should creep more slowly. The hoop stress of 38 MPa has been

plotted on both Figs. 5 and 6 in order to identify the possible fracture mechanisms for the tube. As we

increase the temperature the sequence of mechanisms is the same in both iron and low-alloy steel: it is

intergranular (ferritic)transgranular (ferritic)intergranular (austenitic)rupture (austenitic). In pure iron

rupture begins at about 1000 C; in the low-alloy steel it begins at about 900 C. This is an example of how

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 879creep properties are structure-sensitive: clearly it is necessary to carry out experiments on samples of the

tube steel to obtain more reliable data for the onset of rupture.

2.6. Creep-fracture experiments

Fracture mechanisms are aected by the geometry of the specimen. Creep data are almost always ob-

tained from uniaxial tensile specimens. Until necking starts the only component of stress is the axial tension

Fig. 5. Fracture-mechanism map for pure iron.

Fig. 6. Fracture-mechanism map for a 2.25Cr1Mo steel containing 0.13 wt% C.

and there is no plastic constraint. However, the stress state in the creeping tube is one of plane strain. To

understand why this is the case, let us rst examine the stress state in the tube wall. The tube geometry is

shown in Fig. 7. The tubes are open-ended and at rst sight it would appear that the axial stress is zero.

Fig. 8 shows that this is not the case: the tubes behave as if they have end caps, and the three principal

stresses r1;2;3 are as follows:

hoop stress r1 prb r; 2a

axial stress r2 pr2b r2

; 2b

Fig. 7. Geometry of a creeping tube.

880 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893Fig. 8. Why boiler tubes behave as if they had end caps.

tempe

direct

gauge

comp

h0=hfA few tests were done using specimens with a gauge length of 19 mm instead of the usual 6 mm. The

gauge

eld

nicant dierence between the results from the two types of specimens. In nearly all cases, creep fracturestarted at the centre of the specimen and propagated out to the edges of the gauge section (see Fig. 9).

Th

ectin

To

abovesummarise, the creep-fracture data show that the tube burst at about 900 C. This is comfortablygranular elds and the much larger ductility of the transgranular eld. The rupture eld starts at around

890920 C, with reductions in thickness comparable to those seen in the split tube.

2.7. Temperature of failuree results are shown in Fig. 10. As the temperature increases, the reduction in thickness varies, re-

g the sequence of fracture mechanisms: the data show the relatively small ductility of the inter-sections in these specimens were less well constrained against lateral contraction and had a strain

that was essentially half way between plane strain and uniaxial tension. However, there was no sig-cted to a constant axial load which was calculated to apply a nominal tensile stress of 38 MPa to the

section. Creep-fracture tests were carried out at various temperatures. After each test had been

leted, the specimen was removed from the furnace and measured to nd the reduction in thickness of

.Creep tests were carried out in a vacuum furnace to stop the steel from oxidising. Each specimen wassubjerature before machining. The tensile axis of each specimen was aligned parallel to the hoop stress

ion.through-thickness stress r3 0: 2c

Here, p is the internal gauge pressure, r is the tube radius and b is the wall thickness. The strain rates_e1;2;3 along the three principal directions can be found from the LevyMises equations for strain-rate asfollows:

_e1 C r1

r2 r32

C r

r4

3Cr

4; 3a

_e2 C r2

r1 r32

C r

2

r2

0; 3b

_e3 C r3

r1 r22

C r

2

r4

3Cr

4: 3c

Here, C is the arbitrary constant of proportionality. Eqs. (3) lead immediately to

_e1 _e; _e2 0; _e3 _e: 4

This means that, as the circumference of the tube increases during creep, the wall thickness decreases so as

to conserve volume. There is no change in the length of the tube, which is necessary otherwise the creepingtubes could distort the whole boiler.

Fig. 9 shows the specimen geometry used to approximate to the plane strain conditions in the tube wall.

The gauge section was kept short relative to its width to help stop it from contracting laterally during the

test. Blanks for the specimens were cut from a nearby tube, which had bulged by amounts varying between

5% and 13%. The blanks were attened at red heat. They were then austenitised and air cooled to room

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 881the A3 temperature of 860 C and explains why martensite formed at the edge of the fracture. The

temp

were

2.8. T

Th

plotte

the c

wher

maxitertia882 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893erature of failure is also comfortably below the maximum inlet temperature of 940 C, so the ue gasescapable of heating the tubes up to the failure temperature.

ime to failure

e creep tests also gave values for the times to failure tf at the dierent testing temperatures. These ared in Fig. 11 in the form of ln tf versus T 1 , where T is in K. It can be shown as follows that these areorrect variables. Referring to the creep curve in Fig. 12, one can see that

_e D eftf

; 5

e ef is the strain to failure, and D is a constant which depends on the mechanism of creep fracture: themum value of D is obviously 1.0, but it can be much smaller than this if there is a large strain duringry creep. Eq. (5) then gives

Fig. 9. Geometry of test specimens. Dimensions in mm.

tf Def=_e 6

and

Since

where

(8) w

Withi

ThisTh

time

940 becau

ferrite

energ

Fig. 10. Results for reduction in thickness at fracture.D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 883ef tf _e=D: 7

creep is a thermally activated process,

_e ArneQ=RT ; 8

A and n are constants, Q is the activation energy and R is the gas constant. Combining Eqs. (6) ande obtain

tf DefArn

eQ=RT : 9

n a given mechanism eld, the pre-exponential terms are approximately constant, so

tf / eQ=RT : 10

means that a plot of ln tf versus T1 will be a straight line with a slope of Q=R.e data in Fig. 11 show some interesting features. In the rupture eld (above about 890 C) the failuredepends strongly on temperature: at 890 C, tf is about 40 min; at the maximum gas temperature ofC it is only 10 min. On the other hand, between 760 and 860 C the times hardly change at all. This isse austenite is much more resistant to creep than ferrite: as the temperature increases from A1 to A3 theprogressively transforms to austenite, and this compensates almost exactly for the increased thermal

y available for diusion.

884 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893The data also illustrate how sensitive creep rates are to changes in composition. Below the A1 temper-ature the tube steel resists creep fracture much better than pure iron. This is mainly due to the pinning eect

of both the dissolved carbon and the pearlite nodules. On the other hand, the low-alloy CrMo steel is

much better than the tube steel. Mo is a strong carbide former and the molybdenum carbides are very good

at pinning dislocations. Although there are not enough data points to be sure, there seems to be a hint of asharp step at the A1 temperature, which could be caused by the transformation of the pearlite nodules toaustenite. In the austenite eld, the creep strength of the tube steel is in fact comparable to that of pure iron:

carbon dissolves easily in austenite and does not give much solid-solution strengthening.

2.9. Boiler operating conditions

As has been shown above, the tube failed at about 900 C. At this temperature, the test specimens failedin 30 min. The rst stage in nding out what caused the temperature overshoot is to inspect the plantrecords: were there changes in the operating parameters (e.g., ring rates, safety-valve discharges, boiler

blow-downs) within 30 min of the failure? In this context, it is necessary to remember that the failure time

Fig. 11. Failure times for the tube steel. The data for pure iron and the 2.25Cr1Mo steel are taken from [3].

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 885of the tube is not necessarily the same as the failure time of the test specimens. The reasons for this are as

follows:

2.10. Changes in hoop load

In producing the nal bulge, the tube expanded from an initial outside diameter of 90 mm to approx-

Fig. 12. Typical creep curve for a constant load test in tension.imate

tube

where

the h

whereIn

done

iron

norm

const

but it

Thly 103 mm. As a result, the wall thickness decreased from 5 to 4.3 mm. This means that the bore of the

increased from 80 to 94 mm. As shown in Fig. 13, the hoop load in the tube wall is given by

Fh p2rl2

; 11

l is the length of the bulged section of tube. Since the diameter 2r increases by the ratio 94/80 1.18,oop load increases by 1.18 times as well. At constant temperature the creep equation is

_e Brn; 12

B is a constant. Thus, the strain rate will increase by 1:18n times.order to nd n, it is rst necessary to establish the mechanism of creep during the failure. This can beby inspecting the relevant deformation-mechanism map [5]. Figs. 14 and 15 show the maps for pure

and a 1CrMoV low-alloy steel [5]. The vertical axis in the maps is the equivalent shear stress sealised by the shear modulus G. Because the shear modulus decreases with temperature, a line ofant shear stress will rise as it goes from left to right across the maps. There is no map for the tube steel,

s creep behaviour should be bracketed by the maps for pure iron and the low-alloy steel.

e equivalent shear stress is given by the standard result [6]

se 1=6 r1nh

r22 r2 r32 r3 r12oi1=2

: 13

Fig. 13. Cross-section of tube showing the relationship between internal pressure and hoop load.

Fig. 14. Deformation-mechanism map for pure iron with a grain size of 0.1 mm.

886 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 887Fig. 15. Deformation-mechanism map for a 1CrMoV steel with a grain size of 0.1 mm.Comb

Since

map a

n forwhen

strain

Asabou

min t

2.11.

Sin

havio

doneruptu

tubeining Eqs. (2) and (13) we obtain

se 1=63r2=2 1=2 r=2: 14

the hoop stress is 38MPa, the equivalent shear stress is 19MPa. The line for 19 MPa is plotted on each

s shown in Figs. 14 and 15. At 900 C, the lines lie well inside the eld for power-law creep. The value ofthe power-law creep of austenite is 4.5 [5] and this is the appropriate value for the tube. The creep rate

the nal bulge forms is therefore 1:184:5 2:1 times the creep rate in the test specimen after the same. The average creep rate during the expansion is about 1.6 times the creep rate in the specimen.

shown by Eq. (6), the failure time varies inversely with the strain rate, so the failure time of the tube ist 1.6 times less than the failure time of the specimen. This reduces the time to failure at 900 C from 30o 20.

Dierences in the stress state

ce most of the creep tests were done under conditions approximating to plane strain, the creep be-

ur of the specimens and the tube should have been closely similar. In general, however, creep tests are

using uniaxial tensile specimens and a correction is needed when the test data are used to predict there times of tubes. The uniaxial stress that produces an equivalent eect to the biaxial stress state in the

can be found by using the standard result for the equivalent tensile stress [6]

re 1=2 r1nh

r22 r2 r32 r3 r12oi1=2

: 15

The creep equation

n

2 3

Since

the tu

3. Case study 2: failure of a superheater tube

Gr

with an intended operating temperature of 375 C. The tube steel was to the same specication as that

comparable to the original thickness of the tube wall. This low creep ductility is typical of intergranular

creep

granu

that tAsthat the tube encounters as it heats up depends on the value of the hoop stress. Unfortunately, this is not

given. If the hoop stress is below 20 MPa the maps suggest that the sequence should be intergranular(ferrihe cracks would form perpendicular to the hoop stress, as observed.can be seen from the fracture-mechanism maps (Figs. 5 and 6), the sequence of fracture mechanisms,pendicular to the direction of hoop stress. This conrms that the failure mechanism was intergranular creep

fracture. As shown in Eqs. (2a)(2c), the hoop stress is the largest tensile stress in the wall: the axial stress is

only half the hoop stress, and the through-thickness stress is essentially zero. It would therefore be expectedfracture (see Fig. 4). Metallurgical sections taken close to the fracture showed a maze of ne inter-

lar cracks running into the wall from the surface. The plane of the cracks was approximately per-discussed in the rst case study. The fracture was of the thick-lipped type, with a width of fracture surfaceogli [7] describes a creep fracture in a superheater tube which had been in service for about seven yearsbe is therefore twice the time to failure of the specimen.This result tells us that the tube creeps at about half the rate of the uniaxial specimen. The time to failure ofn 4:5, Eq. (20) reduces to

_e 0:45Brn: 213

p _e B2

r : 20

p !n_ee Bre 19

then givesCombining Eqs. (2) and (15) we obtain

re 1=2 r2=4 r2=4 r21=2 3p =2r: 16

The uniaxial strain rate which is equivalent to the strain-rate eld in the tube, is found from the standard

result for the equivalent tensile strain rate [6]

_ee 2=9 _e1nh

_e22 _e2 _e32 _e3 _e12oi1=2

: 17

Inserting the strain-rate components from Eq. (4) we obtain

_ee 2=9 _e2nh

_e2 4_e2oi1=2

2=3

p_e: 18

888 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893tic)intergranular (austenitic)rupture (austenitic). The fracture geometry should be thick-lipped up to

about 900 C, at which point it should change to a thin-lipped rupture. If the hoop stress is above 30 MPa,an additional (transgranular) eld will appear between about 700 and 750 C: failure in this temperaturerange should then be quite ductile. To summarise, the thick-lipped failure could have occurred anywhere

between the service temperature and 900 C, although the range 700750 C is tentatively excluded if thehoop stress is high enough.

The failure temperature can be narrowed down considerably by using microstructural evidence. To startwith, the tube steel has a microstructure which consists of ferrite grains and nodules of pearlite (see Fig. 16).

If the steel is overheated to a temperature below A1, there will be no transformation to austenite. However,the pearlite is thermodynamically unstable because of the interfacial free energy of the carbide-ferrite

Fig. 16. Schematic showing how pearlite spheroidises with time at high temperature below 723 C. (a) Original normalised structure.(b) Signicant redistribution of carbides. (c) Complete redistribution of carbides and particle coarsening.

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 889Fig. 17. How the time for spheroidisation depends on temperature.

gradually shifts from a large number of small particles to a small number of large ones.

lurgical sections next to the split: as shown in Fig. 4, these should show rounded voids forming in the grains

as the precursor to microvoid coalescence.

Finding the temperature at which this type of failure occurred can be dicult. If the pearlite next to the

split has spheroidised, then the failure temperature is below 723 C. Another method is to take a specimenfurther along the tube from a section which has not expanded much (and which therefore has been at a

lower temperature). If this has spheroidised but the steel next to the split has not, then the fracture has

almost certainly occurred above 723 C. If the failure takes place between the A1 and A3 temperatures, thestructure can be complicated: depending on the cooling rate after the failure, it can consist of mixtures of

ferrite and bainite, or ferrite and martensite, with hardnesses to match [6,8]. A good way of supporting theconclusions drawn from the tube microstructures is to do heat treatments in the laboratory on samples of

tube, to simulate the structures seen next to the fracture.

5. Case study 4: failure of a reformer tube

5.1. Background

The heat exchanger in a reformer plant consisted of a bank of tubes made from 1.25Cr0.5Mo steel. TheThe time that the tube steel takes to spheroidise is plotted in Fig. 17. Because the process is controlled by

diusion, it is thermally activated. This means that a plot of ln t versus T 1 is a straight line with a slope ofQ=R. The value of Q found from the slope is 166 kJmol1. This is close to the value of 174 kJmol1 for thediusion of iron in ferrite grain boundaries [5], which suggests that this is the rate-limiting diusional process.

A metallurgical section taken close to the fracture showed in fact that the pearlite in the steel had

spheroidised signicantly. This tells us that the failure took place below 723 C. However, it is not knownwhere below 723 C the tube failed. The data in Fig. 17 can in principle be used to nd the time that thetube spent at any selected temperature. However, the times given are an upper limit: experiments show [9]that the rate of spheroidisation can be increased enormously if the steel is deformed plastically at the same

time.

4. Case study 3: a second boiler tube

Thielsch [10] describes the creep failure of a carbon-steel boiler tube with a diameter of 95 mm and a wall

thickness of 6 mm. The split was 190-mm long and the wall had drawn down to a thickness of about 2 mmat the edge of the fracture. The tube had expanded noticeably on either side of the nal bulge. The tem-

perature of the failure was given as approximately 700 C.The original wall thickness is three times the width of the fracture. This suggests that the failure could

have taken place by a transgranular mechanism and this possibility is reinforced by the quite large hoop

strain that occurred before the nal failure. Presumably, the hoop stress was quite large and the tube

entered the transgranular eld on the fracture-mechanism map. This would happen at about 700 C, so thequoted failure temperature seems reasonable. The fracture mechanism can be veried by taking metal-boundaries [8]. At high temperature diusion can take place, which allows the structure to evolve in such a

way that the total interfacial energy is reduced. The process, called spheroidisation, is summarised in

Fig. 16. The carbide plates gradually break down with time and the iron carbide reforms as roughly

spherical particles. Eventually, the plates vanish altogether. The nal stage is the coarsening of the carbide

particles themselves [8]: the larger particles grow at the expense of the smaller ones, and the distribution

890 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893tubes contained hydrocarbon gas at a pressure of 4.3 MPa and were heated from the outside by furnace

gases. The tubes had an internal diameter of 128 mm and a wall thickness of 6.6 mm. Owing to a tem-

perature overshoot, one of the tubes fractured and the resulting gas leak set the plant on re.

When the heat exchanger was stripped down, it was found that the tube wall had fractured longitudi-

nally over a distance of about 300 mm. At the edge of the fracture the wall had thinned down to about

2.9 mm. Metallurgical sections cut from the tube next to the failure showed a ferrite-pearlite microstructure

with a slightly enlarged grain size and indications of recrystallisation. The iron carbide in the pearlite hadnot spheroidised.

5.2. Failure analysis

Because pearlite is stable only up to 723 C, if the steel had been held just below this temperature for areasonable length of time then the iron carbide in the pearlite would have spheroidised. Since spheroidi-

sation was not observed, it is unlikely that the tube failed below 723 C. When steel is taken above 723 C,the nodules of pearlite are replaced by grains of austenite. Above the A3 temperature, the structure is 100%austenite. If the steel is held at a high enough temperature the austenite grains will coarsen [8]. When the

steel is cooled back below 723 C the austenite will revert to a mixture of ferrite and pearlite. The graincoarsening is consistent with this scenario and suggests that the tube was heated to well above 723 C.

Using Eq. (2a), the hoop stress in the tube wall is found to be 42 MPa. Using Eq. (14), the equivalent

shear stress is found to be 21 MPa. Figs. 18 and 19 show the fracture-mechanism map for 2.25Cr1Mo steel

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 891[3,4] and the deformation-mechanism map for 1CrMoV steel with a grain size of 0.1 mm [5]. The hoop

stress is marked in Fig. 18, and the equivalent shear stress is marked in Fig. 19.

The 42-MPa line on the fracture map lies in an intergranular eld from 400 to 700 C, just nips atransgranular eld between 700 and 750 C, passes through another intergranular eld between 750 and900 C, and nally passes into the rupture eld. The considerable ductility observed at the edge of thefracture suggests that the failure took place in the rupture eld at a temperature of at least 900 C.

The 21-MPa line on the deformation map intersects with an equivalent shear strain rate of about 103/sat 900 C. The standard result for the equivalent shear strain rate is [6]Fig. 18. Fracture-mechanism map for a 2.25Cr1Mo steel.

2 2 2

Com

The h

woul

temp

At

which

6. Co

Th

preci

as red

fractu

istics

Refer

[1] K

[2] A

19bining Eqs. (4) and (22), we nd that the hoop strain rate is given by_ce 2=3 _e1 _e2 _e2 _e3 _e3 _e1 : 22

nh oi1=2Fig. 19. Deformation-mechanism map for a 1CrMoV steel with a grain size of 0.1 mm.892 D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893_e1 _ce=2: 23

oop strain rate is therefore 5 104/s. If we assume a strain to failure of say 50%, or 0.5, then failured take 1000 s, or 17 min. This is a maximum value, since failure will occur more quickly at a higher

erature.

900 C, the 21-MPa line lies inside the region of dynamic recrystallisation on the deformation map,again is consistent with the metallurgical observations.

nclusions

e four case studies have shown that it is possible to derive a great deal of information about the

se conditions under which overheated, internally pressurised tubes fail by creep. Measurable data such

uction in area at fracture, applied stress and microstructure can be correlated with information from

re-mechanism maps, deformation-mechanism maps, phase diagrams and heat-treatment character-

to determine both the temperature and timescale of the failure process.

ences

aye GWC, Laby TH. Tables of physical and chemical constants. 14th ed. London: Longmans; 1973.

shby MF, Gandhi C, Taplin DMR. Fracture-mechanism maps and their construction for FCC metals and alloys. Acta Metall

79;27:699.

[3] Fields RJ, Weerasooriya T, Ashby MF. Fracture mechanisms in pure iron, two austenitic steels and one ferritic steel. Metall Trans

A 1980;11:333.

[4] Gandhi C, Ashby MF. Fracture-mechanism maps for materials which cleave: FCC, BCC and HCP metals and ceramics. Acta

Metall 1979;27:1565.

[5] Frost HJ, Ashby MF. Deformation-mechanism maps. Oxford: Pergamon Press; 1982.

[6] Jones DRH. Engineering materials III. Oxford: Pergamon Press; 1993.

[7] Grogli A. Creep fractures on tubes from steam generating plants. Source book in failure analysis. Ohio: American Society for

Metals; 1974. p. 332.

[8] Ashby MF, Jones DRH. Engineering materials II. 2nd ed. Oxford: Butterworth-Heinemann; 1998.

[9] Chattopadhyay S, Sellars CM. Kinetics of pearlite spheroidisation during static annealing and during hot deformation. Acta

Metall 1982;30:157.

[10] Thielsch H. Why high-temperature piping fails. Source book in failure analysis. Ohio: American Society for Metals; 1974. p. 60.

D.R.H. Jones / Engineering Failure Analysis 11 (2004) 873893 893

Creep failures of overheated boiler, superheater and reformer tubesIntroductionCase study 1: creep rupture in a water-tube boilerDescription of boilerBoiler operating parametersWater-tube specifications

Description of failureCauses of overheatingMetallurgy of failureCreep-fracture mechanismsCreep-fracture experimentsTemperature of failureTime to failureBoiler operating conditionsChanges in hoop loadDifferences in the stress state

Case study 2: failure of a superheater tubeCase study 3: a second boiler tubeCase study 4: failure of a reformer tubeBackgroundFailure analysis

ConclusionsReferences