Sunteți pe pagina 1din 11

International Journal of Fracture 97: 125135, 1999.

1999 Kluwer Academic Publishers. Printed in the Netherlands.

Creep damage assessment and remaining life evaluation


I. LE MAY1 and H.C. FURTADO2
1 Metallurgical Consulting Services, Box 5006, Saskatoon, S7K 4E3, Canada
2 Centro de Pesquisas de Energia Eltrica, C. P. 2754, 20001-970, Rio de Janeiro, Brazil

Received 24 March 1997; accepted in final form 23 April 1998


Abstract. The approach of Kachanov and Rabotnov to the assessment of damage has been related to the creep of
components at elevated temperature. The relation between the damage function and void formation is discussed,
together with recent developments in which remaining life can be estimated using available materials constants
and strain rate measurements.
Key words: Creep, damage, remaining life, damage mechanics, metallography.

Nomenclature
A
Acr
At
A0
B
c
e
e0
k
K
L
m
n
p
r
t
tb
tr
trem

parameter indicating the number fraction of cavitated grain boundaries


critical value of A at failure
the value of A at time t
material constant
material constant in the damage accumulation relation at time t = 0
constant to account for deficiencies in the Norton exponent and microstructural factors
associated with stress change during creep
true strain
true strain at t = 0
exponent in the damage accumulation relation at time t = 0
material constant in the Norton equation
m/( 1)
exponent in the Norton equation
power law exponent in the Kachanov damage accumulation equation
material constant in the Norton equation modified for damage
exponent in the Rabotnov relation for damage accumulation rate
time
time for brittle failure
time to rupture
remaining life
a proportion of the difference between yield and ultimate stress values
a factor depending on the level of initial stress with respect to the average stress at
rupture
[1 + (u /y 1)]
engineering strain

126 I. Le May and H.C. Furtado


0
r
s
m

0
r
t
u
y

cr
0
p

strain at t = 0
strain at fracture
m tr
secondary (minimum) creep rate
(1 + r)/(1 + r p)
r /s
stress
stress at t = 0
average stress at tr
stress at time t
ultimate tensile strength
yield stress
= r
Kachanov damage parameter
Rabotnov damage parameter (= 1 )
critical value of for failure
initial rate of damage accumulation at t = 0
m+p+c

1. Introduction
There is increasing interest in estimating the remaining life of components that have operated
at elevated temperature for times in excess of that used in design. Microstructural changes and
creep may have occurred. In order to estimate the remaining life an assessment is required of
the accumulated damage.
Various methods are in use to estimate damage, but it is appropriate to review the ideas
of Kachanov (1986). The premise was that materials suffer a loss in strength as a result of
exposure to stress, temperature and resulting deformation and the parameter decreases from
1 at the outset to 0 at failure. Damage does not need a direct physical model, although it
can be represented by loss in cross-section from the formation of voids and cracks on grain
boundaries. This causes an increase in the true stress t under constant load. The rate of
damage accumulation can be considered as a function of the initial stress 0 and the damage
and, using a power law:
d/dt = A0 (0 /)n ,

(1)

where A0 and n are constants depending on temperature, although A0 may also depend on
aging of the material. Integrating,
n+1 = 1 A0 (n + l)0n t.

(2)

At failure, = 0 and t = tr , hence tr is given by


tr = 1/A0 (n + 1)0n .

(3)

Rabotnov (1969) modified the approach using the parameter = 1 , so that = 0 at


the start and = 1 at the end of life. In many ways this is more satisfactory in that damage

Creep damage assessment and remaining life evaluation 127


may be thought of as a concept without a strict physical meaning (loss of cross-section) as
is implied by .
2. The physical nature of creep damage and the creep cavitation model
There has been much emphasis on grain boundary cavity formation as providing an indication of creep damage, following the pioneering work of McLean (1957). Voids are classified as being of two types, namely wedge cracks originating at triple points, and round or
elliptically-shaped cavities. Formation of the former is facilitated by grain boundary sliding,
while there has been much analysis of the growth of the latter by diffusion, involving vacancy
condensation at grain boundaries.
Metallographic methods have been used to determine the degree of void and microcrack
formation in high temperature plant, the first systematic approach being that of Neubauer
and Wedel (1983). Quantification has been made of cavity formation based on a model for
constrained cavity growth, the following relationship being predicted (Cane and Shammas,
1984)
A = 1 (1 t/tr )(1)/m ,

(4)

where A is a parameter defined as the number fraction of cavitated grain boundaries, and
= r /s , where r is the strain at rupture and s is the secondary creep strain = m tr (the
MonkmanGrant constant), m the minimum creep rate, and m is the exponent in the Norton
equation. The remaining life trem is
trem = t (tr /t 1).

(5)

From measured values of A and (4) and (5)


trem = t{1/[1 (1 A)m/(1)] 1}.

(6)

Taking conservative values of m and as 3 and 1.5, respectively (Viswanathan, 1989)


trem = t{1/[1 (1 A)9 ] 1}.

(7)

When the remaining life becomes zero t = tr and A = 1.


Various relationships between creep damage and the A-parameter have been derived,
the simplest procedure being to assume (Cane and Shammas, 1984; Viswanathan, 1989)
= A.

(8)

Thus
/cr = A/Acr

(9)

cr and Acr being the critical values at failure. These values are < 1, which is the limiting value
of or A, failure occurring when the loss of load carrying area produces an unacceptably high
local strain rate, leading to fracture. Murakami et al. (1992) note that this relation can only be

128 I. Le May and H.C. Furtado


considered valid when the cavity area fraction of grain boundary facets becomes large, that is
in the last stages of life. Accordingly, they derived more precise relations between and A.
For the situations where nucleation is complete on loading and for continuous nucleation
the relations derived are (Murakami et al., 1992)
/cr = (A/Acr )2

(nucleation at t = 0)

/cr = (A/Acr )2 /(/r )

(continuous nucleation)

(10)
(11)

where is the strain and r is the strain at rupture. Equation (11) is likely to be more realistic than either (9) or (10). Murakami et al. (1992) eliminated the creep strain from this
relationship, as values of may not be available. The resulting expression is
(/cr ){1 [1 (/cr )]L} = (A/Acr )2 ,

(12)

where L = m/( 1). Equations (11) and (12) provide more rigorous relationships between
and the A-parameter. Murakami et al. (1992) point out that for L = 1, (12) reduces to (9),
and for L = , it reduces to (10). They note that a cavity growth mechanism of creep damage
implies a value of L > 1, so the damage relationships fall between these two extremes. For
L 1, a ductile material is indicated, while L implies a very brittle material.
Damage evolution is considered to be controlled by the relation, = 1 (1 t/tr )1/(1+r),
and specifically the value of the material constant r. This is the same relation as shown for A
for constrained cavity growth in (4). Thus for constrained cavity growth A = as in (8). The
value of r can be determined by fitting actual creep data curves. Leckie and Hayhurst (1974)
derived the relation
r = [m/( 1)] 1

(13)

and can be obtained from creep test data and the relation
= r /s = r /m tr .

(14)

For remaining life estimation we can substitute the expression for damage evolution into (9),
(10) and (12), and substitute for r from (13), to give
t/tr = 1 (1 A/Acr )m/(1),

(15)

t/tr = 1 (1 A/Acr )2m/(1),

(16)

[1 (1 t/tr )(1)m][1 (1 t/tr )1/ ] = (A/Acr )2 .

(17)

2.1. P HYSICAL

LIMITATIONS

Equations (15), (16) and (17) are plotted on the basis of life fraction versus A (Figure 1)
with test data superimposed (Shammas, 1988). Note that the value of A for CrMo steel is still
relatively small ( 1) when the life is largely used up. Liu et al. (1994) note that the failure
values Acr for a number of materials can be 0.2 or less.

Creep damage assessment and remaining life evaluation 129

Figure 1. Lifetime fraction t/tr versus the A-parameter. The plots are of (15), (16) and (17), and are compared
with test data from Shammas, 1988. After Liu et al. (1994).

Figure 2. Section adjacent to the fracture and at the surface of a CrMo tube that ruptured after 70,000 h at a
nominal temperature of 490 C; (a) polished and etched to emphasize grain boundary cavities; (b) polished and
etched carefully to show grain boundaries. Nital etch.

Figure 3. Plots of the A-parameter across the wall of the CrMo tube of Figure 2. The plots are heavy etching (A)
and careful etching (B).

130 I. Le May and H.C. Furtado

Figure 4. Microhardness across the wall of the material of Figure 2.

The methodology depends on observation of voids and determination of A to estimate remaining life. But, as Acr need not be large for many steels, void formation or surface cracking
on base metal is not observed until very near the end of life (Westwood, 1994), and damage
is often very local (Le May et al., 1994), there is much uncertainty. Also, the observation of
voids and their apparent size depends greatly on polishing and etching procedures (da Silveira
and Le May, 1992; Samuels et al., 1992). The best procedures to avoid enlargement of cavities
may not disclose them, detection depending on their being enlarged. These factors make the
A-parameter approach a dubious one.
To illustrate the uncertainties in computing damage based on measured values of the Aparameter, a section of CrMo steel tube that failed at a bend by longitudinal cracking after
70,000 h at a nominal temperature of 490 C is shown (Figure 2). A specimen adjacent to the
fracture was polished and etched to emphasize the damage in the form of apparent voids (Figure 2a): the same specimen was also polished and etched carefully to show grain boundaries
(Figure 2b). The A-parameter values across the wall differ for the two cases (Figure 3). They
are dependent on preparation methods, and vary through the wall thickness.
Hardness measurements and, specifically, changes in hardness, can provide an indication
of exposure to temperature and may be correlated with damage (Viswanathan, 1989). Accordingly, microhardness measurements were also made across the wall of the CrMo tube: they
are shown in Figure 4. As with the computed A-parameter values, they vary greatly across the
wall thickness. Both sets of data (Figures 3 and 4) suggest that damage is significantly greater
nearer to the outside surface.
3. Practical approaches based on damage accumulation
3.1. T HE APPROACH

OF PENNY

Penny (1974; 1996) extended the KachanovRabotnov (KR) approach. Integrating the Rabotnov relation = 0 (1 )r for temperature T = constant, where 0 = B0k , B and k being
material constants
(1 )(1+r) = 1 B(1 + r)0k t.

(18)

Creep damage assessment and remaining life evaluation 131


The stress at time t is
t = 0 A0 /At = 0 /(1 )

(19)

taking to represent an effective loss of area. At tr , = cr and the average stress is r ( ),


reached between y (yield) and u (UTS). Thus
0 /(1 cr ) = = y + (u y ),

(20)

where is a proportion above y , and


= y ,

(21)

where
= [1 + (u /y 1)].

(22)

= 1 in the extreme case for nonhardening materials, but a reasonable estimate of a bound
for ductile materials has been given as = m/(m + 1) (Penny, 1996). The failure condition is
bounded by m/(m + 1) 6 6 u /y . In all cases cr < 1. Substituting from (20) in (18),
tr = [1/B(1 + r)0k ][1 (0 / )(1+r)]

(23)

tr = tb ,

(24)

or

where
tb = [1/B(1 + r)0k ]

(25)

which is the result for brittle failure, and


= [1 (0 / )(1+r) ]

(26)

is a factor depending on the level of 0 with respect to (Figure 5).


In Pennys (1974) approach to creep strain versus damage, we substitute for (1 ) from
(18) into the Norton equation in the form
= K0m /(1 )p

(27)

where p is a material constant relating to microstructural damage. Integrating,


/0 tb = [1 (1 t/tb )1/ ],

(28)

where
= (1 + r)/(1 + r p).

(29)

132 I. Le May and H.C. Furtado

Figure 5. The modified Kachanov rupture curve. After Penny (1996).

Substituting (24) in (28),


/0 tr = (/)[1 (1 t/tr )1/ ]

(30)

At t = tr ,
r /0 tr = [1 (1 )1/ /].

(31)

The term 0 tr can be considered as the MonkmanGrant constant, but 0 is not the minimum
creep rate in a creep test involving primary and tertiary creep (at least): it is the initial creep
rate upon loading with 0 .
An alternative derivation (Penny, 1996) for strain variation provides a useful result. From
constant volume considerations
dLt /Lt = dAt /At = d(1 )/(1 )

(32)

(1 + )(1 ) = 1.

(33)

or

Based on true strain e = ln(1 + ), and from (18), (24) and (25):
e = ln[1 t/tr ]1/(1+r).

(34)

Differentiating,
[(tr /) t]de/dt = 1/(1 + r).

(35)

This shows that remaining life times the current (true) strain rate is a constant. If strain rate
can be measured in service, the remaining life can be computed, as r can be determined from
creep rupture tests and can be estimated, being 1 at low stresses.

Creep damage assessment and remaining life evaluation 133


3.2. T HE MATERIALS

PROPERTIES COUNCIL

(MPC)

OMEGA METHOD

The omega method was developed from the KR approach to assess in-service components
(Prager, 1995). Starting with the KR equation for strain rate in the form
e = e0 (/0)m [1/(1 )p ],

(36)

where e is the true strain, and substituting for (/0 ) = exp(e) from constant volume considerations,
e = e0 exp(me)[1/(1 )p ].

(37)

This is not easily integrated, so is rewritten as an exponential function, noting that there are
three factors, namely increasing stress, increasing damage, and microstructural changes not
related to damage. Hence,
e = e0 exp(me)[1/ exp(pe)][1/ exp(ce)],

(38)

where c is a constant that accounts for deficiencies in Nortons exponent and other microstructural factors associated with stress change. Thus,
e = e0 exp[(m + p + c)e].

(39)

Integrating
[1/e(m
+ p + c)]{1 exp[(m + p + c)e]} = t,

(40)

e = [1/(m + p + c)] ln[1 e0 (m + p + c)t].

(41)

For values of (m + p + c)e  2 or 3, exp[(m + p + c)e] is negligible, thus at failure


1/e0 (m + p + c) = tr = 1/e0 p ,

(42)

where
p = m + p + c.

(43)

From (39),
d ln e/de

= m + p + c = p ,

(44)

p may be determined by plotting ln e versus e and taking the slope. From (40), at fracture
tr = (1/e
p )[1 exp(p er )]

(45)

and the remaining life is


tr t = (1/e
p )[exp(p e) exp(p er )]

(46)

134 I. Le May and H.C. Furtado


Now,
e = e0 exp(p e).

(47)

Neglecting er in (46) and substituting from (47)


tr t
p
= 1/e

(48)

(tr t) de/dt = 1/ p .

(49)

or

As with Pennys analysis ((35), there is constancy of the product of remaining life and (true)
strain rate during service.
4. Concluding remarks
The uncertainties of computing damage based on measured values of the A-parameter have
been discussed. It has been demonstrated that metallographic preparation is extremely important in determining the value of the A-parameter in creep-damaged material. There is not
necessarily a right or a wrong procedure, but a standardized procedure to display creep
damage is required (Samuels et al., 1992). The apparent variation in creep damage through
the wall of a steam tube is interesting, suggesting that observations at the outside surface may
overestimate the creep damage in the bulk material.
It has been shown that Kachanov based procedures exist that can be used to estimate
remaining life. The procedure of Penny (1974; 1996) and the later MPC one have strong similarities. Penny (1996) indicated how remaining life can be estimated by in-situ measurement of
the bulk strain rate of a component, and substituting this and available materials constants in
(35). Similarly, Prager (1995) discussed the application of the omega model by running tests
on damaged material samples to obtain initial strain rate and the rate of change of strain rate
with strain. This allows determination of p and from this and an estimation (or measurement)
of the strain rate in service, the remaining life is estimated. Thus, tools are available for
remaining life estimation of creeping components on a reasonable and fundamental basis.
The Kachanov based methods are also likely to provide better estimates of the damage in a
component, particularly if the section thickness is large, than can be obtained with surface
observations alone. Combining them with microstructural evaluations provides an appropriate
approach to the evaluation of aging high temperature plant.
References
Cane, B.J. and Shammas, M.S. (1984). A Method for Remanent Life Estimation by Quantitative Assessment of
Creep Cavitation on Plant: Report TPRD/L/2645/N84. Central Electricity Generating Board, Leatherhead.
Da Silveira, T.L. and Le May, I. (1992). Effects of metallographic preparation procedures on creep damage
assessment. Materials Characterization 28, 7585.
Kachanov, L.M. (1986). Introduction to Continuum Damage Mechanics. Martinus Nijhoff, Dordrecht.
Liu, Y., Murakami, S. and Sugita, Y. (1994). Identification of creep damage variable from A-parameter by a
stochastic analysis. International Journal of Pressure Vessels and Piping 59, 149159.

Creep damage assessment and remaining life evaluation 135


Le May, I., Da Silveira, T.L. and Cheung-Mak, S.K.P. (1994). Uncertainties in the evaluation of high temperature
damage in power stations and petrochemical plant. International Journal of Pressure Vessels and Piping 59,
335343.
Leckie, F.A. and Hayhurst, D.R. (1974). Creep rupture of structures. Proceedings of the Royal Society, Series A
340, 323347.
McLean, D. (1957). Grain Boundaries in Metals. Oxford University Press, Oxford.
Murakami, S., Liu, Y. and Sugita, Y. (1992). Interrelation between damage variables of continuum damage
mechanics and metallographic parameters in creep damage. International Journal of Damage Mechanics 1,
172190.
Neubauer, B. and Wedel, V. (1983). Restlife estimation of creeping components by means of replicas. ASME International Conference on Advances in Life Prediction Methods (Edited by D.A. Woodford and T.R. Whitehead),
American Society of Mechanical Engineers. New York, 353356.
Penny, R.K. (1974). The usefulness of engineering damage parameters during creep. Metals and Materials 8,
278283.
Penny, R.K. (1996). The use of damage concepts in component life assessment. International Journal of Pressure
Vessels and Piping 66, 263280.
Prager, M. (1995). Development of the MPC omega method for life assessment in the creep range. Journal of
Pressure Vessel Technology 117, 95103.
Rabotnov, Yu. N. (1969). A mechanism of a long time failure. Creep Problems in Structural Members. North
Holland, Amsterdam.
Samuels, L.E., Coade, R.W. and Mann, S.D. (1992). Prepacking structures in a creep-ruptured low-carbon
Cr-Mo steel: their nature and detection by light microscopy and scanning electron microscopy. Materials
Characterization 29, 343363.
Shammas, M.S. (1988). Metallographic methods for predicting the remanent life of ferritic coarse-grained weld
heat affected zones subject to creep cavitation. International Conference on Life Assessment. Electric Power
Research Institute, Palo Alto, 238244.
Viswanathan, R. (1989). Damage Mechanisms and Life Assessment of High Temperature Components. ASM
International, Metals Park, Ohio.
Westwood, H.J. (1994). Applications of quantitative metallography in creep life assessment. Materials Performance, Maintenance and Plant Life Assessment (Edited by I. Le May, P. Mayer, P.R. Roberge and V.S. Sastri).
The Metallurgical Society of the Canadian Institute of Mining, Metallurgy and Petroleum, Montreal, 5770.

S-ar putea să vă placă și