See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/268351439

A Generic Microstructure-Explicit Model of Creep in Nickel-Base Superalloys

Article · January 2004

DOI: 10.7449/2004/Superalloys_2004_897_906

CITATIONS READS
35 497

4 authors, including:

Hector Basoalto Sanjay Sondhi

The University of Sheffield General Electric
60 PUBLICATIONS   1,733 CITATIONS    17 PUBLICATIONS   238 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The Centre for Advanced Simulation & Modelling for Manufacturing (CASiM2) View project

Next Generation Modelling Project View project

All content following this page was uploaded by Hector Basoalto on 14 May 2015.

The user has requested enhancement of the downloaded file.

 Superalloys 2004
Edited by K.A. Green, T.M. Pollock, H. Harada,
T.E. Howson, R.C. Reed, J.J. Schirra, and S, Walston
TMS (The Minerals, Metals & Materials Society), 2004

A GENERIC MICROSTRUCTURE-EXPLICIT MODEL OF CREEP

IN NICKEL-BASE SUPERALLOYS

H.Basoalto1, S.K.Sondhi1, 2, B.F.Dyson1 and M.McLean1

1
Department of Materials, Imperial College London, Exhibition Road, London, SW7 2AZ, United Kingdom.
2
National Metallurgical Laboratory, Jamshedpur, India.

Key Words: creep, damage, modelling, microstructure, cavitation, coarsening

Abstract of the DARPA Accelerated Insertion of Materials Programme to

which the authors have contributed. [3]
Past approaches to modelling the creep behaviour of engineering
alloys have been either totally empirical or, while having The most damaging effect of creep deformation is not simply, or
functional forms consistent with current understanding of even predominantly, through the time to rupture of macroscopic
deformation and fracture mechanisms, have been calibrated by turbine blades or discs under load at temperature. It also plays an
comparison with an available creep database. They have not important role in stress relaxation and redistribution during
specifically included quantitative measures of the microstructural variable stress/temperature and displacement-controlled
features that are thought to impart creep resistance to the alloys. deformation of these components. It is important to be able to
The present paper reviews and extends a microstructure-specific assess the contribution that creep makes during the complex
model of creep in particle-strengthened alloys in which the model loading histories of superalloys in service. The effective
parameters are directly related to measurable characteristics of the application of sophisticated computer-aided design methods
microstructure of nickel-base superalloys. The model accounts for requires that materials behaviour can be represented by reliable
three previously identified changes in microstructure that occur constitutive equations that are compatible with the numerical
codes used to simulate service response of blades and discs.
during the creep of superalloys: (i) coarsening of the Jc
precipitate; (ii) the progressive increase in mobile dislocation There have been many approaches to representing the strain-time
density with accumulated creep strain; and (iii) the development evolution of engineering alloys in general and superalloys in
of grain boundary cavities. The relative dominance of each will particular. These have ranged from the purely empirical t1/3
be demonstrated using a set of commercial alloys with specific equation for primary creep proposed by Andrade [4] to the fitting
alloy microstructures. It is emphasised that the model-generated of complete creep curves using the expanded power law
curves are true predictions using input microstructural representation of Graham and Walles [5] and the sum of
characteristics and are not empirically fitted to the creep data.
exponential functions in the T-projection approach. [6] Extension
Introduction to cyclic loading has been achieved through a number of visco-
elastic formulations, such as those of Chaboche [7] and
Superalloys for application as gas turbine blades have been Cailletaud.[8] There have also been several attempts to develop
developed over a more than sixty years with the primary aim of equations that reflect the mechanisms influencing creep
increasing their resistance to creep deformation and fracture. deformation. Kachanov [9] and Rabotnov [10] expressed creep
When coupled with control of other properties, such as corrosion rate as a function of an undefined damage parameter; others
and thermal fatigue resistance, this has led to a progressive extended this approach to the development of equations that
increase in temperature capability with concomitant reflected specific types of damage that were material and
improvements in engine performance and efficiency. The load/temperature specific. [11] All of these approaches lead to
improved creep performance of precipitation-hardened multi-parameter representations, which are calibrated against
superalloys has resulted from control of the chemistry and of the mechanical property data.
grain and precipitate structures.[1] In particular, the volume
fraction and solvus temperatures of the Jc precipitate have Although it is very clear that the progressive enhancement of
increased steadily over the years and the grain morphology has creep strength of superalloys has been associated with control of
evolved from equiaxed to columnar and eventually to the single critical microstructural features, the parameters in the above
crystal form.[2] These have been the dominant factors in the models have no immediate quantitative relationship to
improved creep performance and must ultimately be incorporated microstructure. Indeed, there seems to be no means of inferring
in any understanding of the mechanisms of creep. the effect of even a small change in composition and/or heat
treatment of a given alloy on the creep response without repeating
Although creep has not been an important factor in turbine discs, the extensive model calibration procedure. Nor do the models
which are also predominantly produced from nickel-base recognise that superalloys, and other engineering alloys, have
superalloys, there is a growing concern that as engine microstructures and thermo-mechanical response that change
temperatures increase further there will be transient periods of during exposure at service temperatures and stresses. Indeed, the
operation where creep will significantly influence the service lives microstructures of superalloys often vary spatially and temporally
of discs. This has been a problem that has been addressed as part as a result of thermal history in heat treatment and service.

897
This paper describes the development of a new type of creep A constraint factor K CF (I p ) has been incorporated to take
model in which the constitutive equations are explicitly expressed
account of the 3-D stress fields due to the close approach of
in terms of quantitative measures of the initial microstructural
cubical particles in high volume fraction alloys. Its magnitude has
features and their subsequent rates of evolution.
been calculated using a formulation by Pollock and Argon [15] re-
Basis of the Model expressed here as a function of particle volume fraction:
The classic theory of Ansell and Weertmen [12] represented creep K CF (I p ) 1  2I 1p/ 3 3 3S (1  I 1p/ 3 ) for I p t 0.3 and
in particle-strengthened materials as being due to a sequential K CF (I p ) 1 for I p  0.3 .
climb-glide movement of dislocations past the rigid particulate
obstacles, which inhibited free glide. Although the model clearly The above set of equations naturally leads to a decreasing creep
captures most of the physics that even now we recognise must rate with increasing strain (primary creep) until a steady state is
apply in particle-strengthened materials, it only predicts a linear reached at constant stress.
relationship between creep rate and stress for stresses below the
Orowan stress. Experimentally, particle strengthened alloys, such However, the microstructure of nickel-base superalloys is not
as superalloys, exhibit strong stress dependencies of creep with stable at elevated temperature. The model can be extended to
stress exponents of 8 or above, if mapped by a power-law account for the dynamic nature of the material. [16]
function. Attempts to rationalise this discrepancy by introducing i. The equilibrium volume fraction of Jc varies with
a threshold stress into the power-law equation leads to an temperature and is strongly dependent on the solvus
increasing stress exponent with decreasing stress, whereas
superalloys exhibit a decreasing stress exponent. temperature of the alloy under consideration Ts . Its major
influence is through the magnitude of the isotropic strength
Following earlier work by Dyson and Osgerby[13], Dyson [14]
parameter V 0 .
suggested that a geometrically random distribution of dislocation-
particle intersections will always have a small fraction of
ii. The particle-dispersion coarsens during service.
dislocations in a position to "escape" after a single climb event
Assuming Ostwald ripening to occur by volume diffusion
and then to glide until again being temporarily trapped at another
with a kinetic rate constant Kp, the coarsening rate can be
particle. The probability of a successful “escape”, followed by
viscous glide until each dislocation is again temporarily trapped at shown to be given by D p K p (1  D p ) 4 3 that modifies V 0
another particle, was shown to be a function of stress, temperature to V 0 1  D p  where D p is an appropriately defined damage
and particle dispersion parameters. When the viscous glide rate is
fast, the model gives a shear creep rate, J that is a very strong state variable. [17]. Note that V 0 refers to the initial particle
(sinh) function of stress, temperature and inter-particle spacing. A configuration.
further departure from the Ansell-Weertman model is that it iii. As inelastic (creep) strain accumulates, dislocation
incorporates the suggestion of Ion et al [11] that the creeping sources will be activated leading to a progressive increase in
matrix progressively transfers load to the non-creeping particles, dislocation density. This effect has been represented by a
which are assumed in this case to deform elastically. The damage variable D d CH , where C is a constant and
proposal is that this is the sole mechanism of primary creep in
Dd U U i   1 . This leads to an increase in the pre-sinh term
precipitation-strengthened crystals far from athermal yield. For
the particular case where the microstructure of the material is by a factor of (1+Dd).
completely stable, the instantaneous uniaxial creep rate H and the
Thus the dynamic nature of the microstructure can be
normalised back stress caused by the particles H V b V can be
incorporated into the model, expressing the uniaxial creep rate H
described by the two coupled differential equations: as a function of three state variables (H, Dd and Dp) through its
expansion from two to the following four coupled differential
H H0 sinh V (1  H ) V 0  [1a]
equations:
H h V 1  H 
H H [1b]
Qd / j RT ª V (1  H ) º
There are four parameters in these equations, all of which are H H0 1  Dd e sinh « » [2a]
directly related to physical/ geometrical constants or quantitative «¬ V 0 (1  D p ) »¼
measures of the microstructure. These are the dislocation density
Qd , j H h V 1  H H H [2b]
U, jog density cj, volume diffusivity Dv D0 e  RT , particle
volume fraction Ip, the inter-particle (surface to surface) spacing, D d CH [2c]
Op, E the (assumed equal, matrix/particle) Youngs modulus and
K cp
the mean Taylor factor M , R is the Gas Constant and T is the
D p
3
1  D p 4 [2d]
temperature in K. :
This set of constitutive equations requires 10 parameters to be
Q
determined in order to predict the full variable temperature/stress
H0 2U iI 1/ 2
p M 1
1  I >S 4@
p
1/ 2
I 1/ 2
p c D
j v H0 c e
 d/ j
RT
behaviour when the hardening and two types of damage are in
V0 K CF (I p ) MkT b O p 2 operation:
hc EI p
H0c , Qd / j , V 0 , 'H s , Ts , hc, H , Q p , K cp , C .
H 
2I p 1  2I p .
898
At this point, no consideration has been given to the effects of No creep curves have been analysed at this point of the model
grain boundary cavitation. Consequently, it is appropriate to calibration. It is of interest to compare the strain-time evolution
single- and directionally solidified crystals and to polycrystalline calculated using equation-set 2 and the optimised parameters.
alloys that deform to large strains where the life is not Figure 2 shows creep data for two tests at 850oC with a stress of
significantly affected by cavitation or cracking. Development of 170MPa, to indicate the experimental scatter, together with the
the model to account for so-called creep brittle polycrystalline model prediction. The general shapes of the creep curves, which
alloys will be described in the appropriate section below. is determined by the evolution of the state variables H,.Dd and Dp,
has been well represented. These variables are determined by
Application of the basic model material parameters that are known with reasonable accuracy. It
The model has been used to represent a creep database for the is the timescale over which the curve develops that has had to be
superalloy IN738LC in the directionally solidified form in which fine-tuned, and this depends on Ui and cj, which are very difficult
strain/time plots are available at 750oC, 850oC and 950oC for lives to estimate. It is important to note that the model is being
to over 20,000 hours. The first step in calibrating the model was assessed on a form of the alloy that has high creep ductiliy where
to estimate the various constants, physical and microstructural, creep cavitation can be ignored. In the conventionally cast form
appropriate to this alloy. These are listed in Table 1. Some of the of IN738LC, lower creep strains, shorter lives and higher
quantities are known with some accuracy, since this is an alloy minimum creep rates are observed as illustrated in Figure 2. This
that has been extensively characterised [18]. These include the will be discussed further below.
volume fraction, initial size distribution and the coarsening
characteristics of the Jc precipitate. Others, such as the dislocation
density and the jog formation characteristics are subject to some 700
uncertainty. Although the Jc distribution is bimodal in this alloy, 600
calculations indicate that it is the small particles that determine 500
the magnitude of the strength parameter V 0 . 750oC
400
The physical and microstructural parameters are then used to Stress (MPa)
evaluate the values of the operational model parameters 300
850oC
associated with equation- set 2. The a priori set of parameters
calculated from the information in Table 1 is shown in the first
200
line of Table 2. The coarsening parameter is determined by the
average particle size. Using these values as initial estimates, a IN738LC d.s. data
950oC
numerical optimisation procedure was used to fine-tune the model prediction
parameters to give an optimum fit to the rupture life data for the
entire database. The optimised values of the model parameters 100
are also shown in Table 2. The major difference is in the value of 100 101 102 103 104 105
the characteristic creep rate, which is a factor of 25 lower than the Lifetime (h)
initial estimate. However, this is the parameter that is dependent
on the dislocation density and jog density, neither of which is
known with any precision. A comparison of the measured and 1e-5
computed rupture lives and minimum creep rates is shown in
Minimum Creep Rate (s-1 )

Figure 1.
1e-6
Table 1 Estimates of physical constants and microstructural
parameters appropriate to IN738LC. 1e-7 950oC
10 -2
Initial dislocation density 10 m
1e-8
Burger’s vector 2.5u10-10 m 850oC

750oC
Jc Volume Fraction and mean size large 20%-/ 250nm 1e-9
small 25%-/90nm
Jog density (pre-exponential coefficient) 1 1e-10
-1 100 200 300 400 500 600 700
Volume diffusion (pre-exponential 1.9 u10-4 m2s
coefficient) Stress (MPa)

Volume diffusion activation energy 284kJmol-1 Figure 1 Comparison of IN738LC DS stress rupture and
minimum creep rate data with model predictions using the
Coarsening activation energy 305kJmol-1 optimised model parameters
Coarsening rate (pre-exponential 1.6 u10-14 m3s-1
coefficient)

899
 Table 2 Macroscopic model parameters for IN738LC theoretically calculated from the data in Table 1 and optimised to give
the best fit to the creep data.

H0c Qd / j V0 'Hs Ts hc K cp Qp
H* C
1 (kJ/mol) (MPa)
(s ) (kJ/mol) (MPa) (K)
( s 1 )
(kJ/mol)

Theoretical 1.01 x105 311 33 49.83 1394 7.2x104 0.47 67 106 305
3 4 6
Optimised 4.44 x10 316 33 49.83 1394 7.2x10 0.47 67 10 305

for Nimonic 90 is more problematical. Either the effect of

20 particle coarsening has been over-estimated or the degree of
dislocation- softening has been attenuated by a thermal recovery
In738LC c.c. process not considered in the model. Nevertheless, the ability of
In738LC d.s.
15 c.c. model prediction the model to encompass different generations of alloys is very
Creep Strain (%)

d.s. model prediction encouraging.

10
700
600
5 500
Stress / MPa 400

0 300
0 5000 10000 15000 20000 25000
Time (h) 200
Figure 2 Comparison of simulated and experimental creep
curves for IN738LC. Nimonic 90 data
IN738LC d.s. data

Extrapolation to other alloys 100 CMSX4 <001> data

90 CMSX4 <111> data
80 model prediction
As indicated above, the historical development of nickel-base 70
superalloys has largely been achieved by control of the Jc solvus 100 101 102 103 104 105
and particle volume fraction and size. Since the basis of the
Lifetime / h
model is to relate the creep behaviour explicitly to quantitative
measures of these aspects of microstructure, it is of interest to Figure 3 Comparison of model predictions with measured creep
explore the extent to which the creep behaviour of different lifetimes at 850qC for three alloys with different J' volume
generations of superalloys can be represented by the model. fractions and particle size
Figure 3 compares the 850oC stress rupture data for IN738LC
superalloy with those of a wrought polycrystal (Nimonic 90) and Effect of Grain Boundary Cavitation
a single crystal (CMSX4). The original formulation of continuum damage mechanics by
Nimonic90 is a wrought alloy containing 25% volume fraction Kachanov [9] to account for creep strain trajectories never
of Jc with an initial mean diameter of about 28nm. CMSX4 is a explicitly identified the damage mechanism responsible for
2nd generation single crystal superalloy, which has 65-70% tertiary creep, although it was interpreted as cavitation for many
volume fraction of Jc of initial size of 250 Pm. Simulations were years. In that first treatment by Kachanov, failure was assumed
run for these alloys using the parameters listed in Table 1 with to occur when a dimensionless damage parameter increased
the exception of material specific values for Jc volume fraction from 0 to 1. The damage did not affect the creep rate, but
and size calculated at the appropriate temperature (850oC). determined the (potentially varying) magnitude of the creep
strain at failure. Kachanov does not predict tertiary creep for
The simulated and measured stress rupture behaviour of the conditions of constant stress. Rabotnov's contribution [10] was
three alloys is shown in Figure 3. Although there are details that to couple the development of damage with the creep rate
are not represented by the model, it clearly captures the through an increase in stress associated with a reduction in load
substantial improvement in creep strength of these different bearing area as the damage, by implication cavitation,
generations of superalloy. There are special factors that will developed. This leads to an accelerating creep rate (i.e. tertiary
have to be taken into account to provide a fuller representation creep).
of creep in the individual alloys. For example, the data for
CMSX4 for different crystallographic orientations show There have been many studies of creep cavitation, both
significant differences and no account is taken of the role of experimental and theoretical; it is beyond the scope of this paper
rhenium, which is thought to suppress dislocation climb- and to provide a detailed review. Rather, we summarise the relevant
particle coarsening-rates. The underestimate of long-term lives aspects that can be incorporated into the present model. Creep

900
cavities develop on grain boundaries that are subject to a constrained and unconstrained cavity growth as stress increases,
mechanical traction; i.e. those normal to a tensile stress. making the approximations in equation-set 3 inappropriate.
Consequently, inelastic deformation within the alloy can become
highly spatially heterogeneous. The total strain in the Dyson [20] has modelled the transition between unconstrained
neighbourhood of such transverse boundaries is composed of and constrained cavity growth in terms of a two-bar structure
dilatory displacements due to stress-directed diffusional growth between sound, uncavitated alloy D and cavitated alloy E (see
of the cavities, inelastic displacements due to creep and spatially Figure 5). Region D of volume fraction ID is taken as being
heterogeneous elastic displacements due to stress redistribution. described by equation-set 2. The cavitated region, of volume
Remote from the transverse boundaries, displacements are fraction (1-ID), experiences two contributions to strain: inelastic
elastic plus inelastic. There are two extremes of behaviour: flow as in the D region, and a dilatory contribution due to the
diffusive growth of grain boundary cavities. Using the kinetic
a. Unconstrained cavity growth in which the creep equation derived by Speight and Beeré [21], it was possible to
displacement rates within the uncavitated regions are much model the transition between constrained and unconstrained
faster than the displacement rates due to cavity growth. cavity growth in a commercial aluminium alloy.
The cavities are then free to grow at the rate determined by
the (essentially applied) stress directed diffusion. This is Basoalto and Dyson [22] are currently improving and using this
equivalent to the Kachanov formulation where the approach to model stress rupture behaviour for conventionally
macroscopic creep deformation is unaffected by cavitation, cast IN738LC. Figure 6(a) shows the predicted variation of
but fracture occurs when the transverse grain boundaries ductility with lifetime and Fig.6(b), the corresponding stress/
are fully cavitated and link up to cause fracture. lifetime plot. The model’s prediction of an increasing failure
strain for short life tests, beyond a critical stress (due to
b. Constrained cavity growth occurs when the creep increasing loss of constraint) correlates with the stress/lifetime
displacement rates within the uncavitated regions are data of the two alloy-variants converging.
slower than those due to cavity growth. In this condition,
stress is off-loaded from the fast growing cavitated In Figure 6a the model overestimates the experimental ductility.
boundary regions to the uncavitated zones. This increases This is almost certainly due to the fact that it currently excludes
the creep rate by reducing the effective stress controlling the possibility of faster cavity growth rates in the unconstrained
creep and is thus more consistent with the proposal of regime due to matrix strain rate effects. [23,24,25]
Rabotnov. In previous versions of this [16] and other [5,6,7,8] models of
Dyson and Gibbons [19] proposed a simplified method for creep curve shapes, deformation has been terminated at a
dealing with the case when grain facets cavitate at a constant predetermined strain that has been obtained experimentally. We
rate (nucleation controlled) but with individual cavity growth believe that this is the first model that predicts failure strain
constrained. An additional damage term Dc that couples with the whuich varies with stress and temperature.
creep rate equation was introduced (Equation 3e). . The
geometrical constant kc|0.33 is for equiaxed grains and H f ,min is
the material-specific minimum creep failure strain. The 600
expanded equation set is given by Eq.3: 500 unconstrained

ª Qd / j º ª V (1  H ) º 400
H H0c (1  Dd ) exp « » sinh « »
¬ RT ¼ ¬« V 0 (1  D p )1  Dc  ¼»
Stress / MPa

[3a] 300
hc § H · [3b]
H ¨1  ¸H
V © H ¹ 200
D d CH [3c] d.s. data constrained
c.c. data
Kp Upper bound predictions
D p (1  D p ) 4 [3d] Lower bound predictions
3
kc 100
D c H [3e] 100 101 102 103 104 105
Hf
Lifetime / h
Numerically integrating equation-set 3 with the optimised
Figure 4 Comparison of stress rupture lives for conventionally
parameter set determined for directionally solidified IN738LC,
cast and directionally solidified IN738LC with predictions for
together with values of kc =0.33 k = 0.33 and H f 4.9% (the constrained (Equation Set 3) and unconstrained cavity growth
lower bound to the creep strain found in IN738cc; see Fig.2), (Equation Set 2).
simulates the long term stress-rupture behaviour of the
conventionally cast alloy shown in Fig.4 very well. It is poor at
higher stress levels where conventionally-cast and directionally-
solidified data converge and which can therefore be modelled by
the original optimised parameter-set. The reason for this, as
shown below, is because there is a transition between

901
 Application to Turbine Disc Alloys
The potential of representing the creep behaviour turbine disc
D alloys using the model described above has been investigated as
D D V f part as part of the DARPA Accelerated Insertion of Materials
(AIM) programme. [3] The work has been carried out in
E conjunction with GE Aircraft Engines, Pratt and Whitney and a
consortium of university research groups. Two representative
D D powder alloys have been investigated - Rene88DT with a
supersolvus heat treatment and IN100 with a subsolvus heat
E treatment. Since creep has not been perceived in the past as
being a problem for gas turbine discs, there is very little
D D information on their creep behaviour. Also, since these alloys
have been developed with optimum yield stress, they have quite
D different microstructures from the blade alloys discussed above.
The first stage of the investigation was to experimentally
characterise the microstructure and creep behaviour of the two
V f
alloys.
The basic microstructural characteristics of the two alloys in the
as-heat treated condition are summarised in Table 3. The sub-
Figure 5 Schematic showing D and E grains unaffected and solvus IN100 has a grain size of about 5Pm and three distinct
affected by the development of grain boundary cavities populations of Jc particles with a combined volume fraction of
respectively. 0.56. Rene88DT has a larger grain size of 25Pm and a smaller
volume fraction of Jc particles of 0.36 distributed as a bimodal
distribution with none of the primary Jc particles present in
(a) 24 IN100. However, in both cases the grain sizes are much smaller
22 IN738LC c.c. data than in turbine blade alloys. Although the particle sizes can be
IN738LC d.s. data measured with some precision, there are no reliable techniques
20 model prediction
available for measuring the volume fractions of the tertiary
18
Failure Strain / %

precipitate. Consequently, the tertiary Jc volume fractions have

16 been estimated as the difference between the total Jc content,
14 determined by Thermocalc calculation, and the sum of the
unconstrained constrained
12 measured primary and secondary Jc volume fractions.
10 The shapes of the creep curves for the two alloys produced with
8 uniaxial tensile stresses were very similar, and quite different
6 from those of turbine blade alloys. Figure 7(a) is quite typical,
4 showing no primary creep, imperceptible creep over about 30%
2 of the life followed by a progressive increase in creep rate to
failure at a strain of 10% or more. Close examination of the
101 102 103 104 105
initial stages of the longer-term creep curves showed that there
(b) Lifetime / h was a period of negative creep as shown in Figure 7(b). The
creep rupture lives of the two alloys at 649 and 704oC are shown
500 in Figure 8. In spite of the large difference in Jc volume
fractions in the two alloys, they display similar creep lives at
400 649oC. More surprisingly, Rene88DT, which has a smaller Jc
content than IN100, has significantly superior stress rupture
300 lives at 704oC.
Stress / MPa

Before attempting to apply the model to these data, it is

200
necessary to understand why the creep behaviour of the disc
d.s. data
alloys is so different from that of blade materials. Also, the
c.c. data retention of creep resistance of Rene88DT relative to IN100
d.s model predtions must be rationalised. To clarify these issues, a series of
c.c. model predictions
diagnostic tests have been undertaken in order to guide the
development of the model to represent the creep behaviour of
100 the disc alloys. The rationale behind this study is described in
101 102 103 104 105 detail by Sondhi et al. [26]
Lifetime / h

Figure 6 Comparison of (a)stress rupture lives and (b) creep

failure strain for conventionally cast and directionally solidified
IN738LC with predictions for a transition between constrained
and unconstrained cavity growth.
902
 ii. A comparison was made of the creep response in axial
12 tension and compression loading. Figure 9 is typical of the
o
IN100 704 C 400MPa results obtained on both alloys. In compression there is a
10
distinct primary creep that is absent in tension, with up to 1%
8 strain being accumulated before there is any perceptible strain in
tension. If this tension/compression asymmetry were the linear
Strain / %

6 superposition of creep strain and shrinkage due to an unspecified

4
microstructural change then the difference between the two
curves would be twice the shrinkage as a function of time. It is
2 clear that the tension-compression asymmetry is much greater
than the sum of the two effects (Figure 9).
0
iii. The effect of prior heat treatment at the test
-2
temperature is shown in Figure 10 for both tension and
0 200 400 600 800 1000 1200
compression loading. There are two quite distinct effects in the
Time / h short and the long term. In the case shown for a prior heat
treatment of 600 hours at 704oC, the initial tension compression
0.3 asymmetry has been completely eliminated. Thus the initial
o
IN100 704 C 400MPa creep strength in compression is increased, whereas that in
tension has been reduced. However, in the later stages of creep
0.2 prior heat treatment leads to increased creep rates relative to the
as received alloys.
Strain / %

0.1 iv. The results of ageing heat treatment on the alloy

microstructures have also been examined. The only significant
change observed was a coarsening of the tertiary Jc particles.
0.0 This is consistent with previously published work on the
coarsening of J', which is reflected in the coarsening constants
incorporated in the model and listed in Table 2. Assuming that
-0.1 the normal r3 LSW coarsening law applies [18,29], r for the
0 100 200 300 400 500 tertiary particles will be 102 and 104 faster than for the secondary
Time / h and primary particles respectively. It would be expected that the
tertiary J' would eventually dissolve and reprecipitate on the
Figure 7 Creep curve for IN100 tested in tension at 704oC and secondary particles.
400MPa (a) full creep curve and (b) enlargement of the early
part showing negative creep. IN 100 704°C 400 MPa
1.6
As Received: Tension
1.4 As Received: Compression
Htension + 2 Hshrinkage
1.2
1.0
Strain / %

1000
900 0.8
800
700 0.6
Stress / MPa

600 0.4
500 0.2
400 0.0
IN100 - 704°C
Rene88 - 704°C
IN100 - 649°C 0 100 200 300 400 500
300 Rene88 - 649°C
model prediction for IN100 Time / h
model prediction for Rene88

200
Figure 9 Comparison of the initial stages of creep curves
10 100 1000 10000 obtained in tension and compression for IN100 tested in constant
load at 704oC and 400MPa. The curve computed from the
Lifetime / h
Thermal ageing of the alloys has two important effects:
Figure 8 Comparison of creep lives in tension as a function of
stress for IN100 and Rene88DT at 649 and 704oC. a. The volume misfit between particle and matrix can be
relaxed by a combination of loss of coherency and mass
The diagnostic tests undertaken were: transport. Complementary to the matrix compressive stress, the
particle will be in tension. The initial value of H in the model
i. Creep specimens were monitored in the absence of an
will no longer be zero, and will evolve with both accumulated
applied stress at 704oC. Both alloys exhibited progressive linear
creep strain and with recovery of the internal stress. In the
shrinkage strains of |0.1% over periods of several hundreds of
present analysis we assume that the stress is relaxed only by
hours. Henry [27] has shown that the density of Rene88DT
increases during heat treatment.
903
diffusional mass transport. The effect of relaxation of the stress with increasing temperature, which is determined by the slope of
will be to attenuate the tension/compression asymmetry. the solvus. In the present analysis we assume that at elevated
temperatures the dissolution of Jc is entirely accommodated by a
measured tensile strain and twice the axial contraction in the
reduction in the volume fraction of the tertiary precipitate. The
absence of stress is shown for comparison.
temperature dependence of volume fraction is calculated using
The model has been expanded to take this effect into an empirical relationship for the J-Jcequilibrium in nickel-base
consideration giving the expanded set of Equations: superalloys. This allows the model to be used to calculate the
intrinsic creep behaviour associated with the secondary and
H Hcreep  Hthermal [4a] tertiary particles to be determined. Sondhi et al. [26] have

ª V (1  H ) º
calculated the values of V0 associated with each population as
H0 1  Dd e
Qd / RT
Hcreep j
sinh « » [4b] function of temperature (Table 4) and the other parameters are
¬«V 0 (1  D p ) ¼» listed in Table 5. As the tertiary particles coarsen, the relative
 QD
contributions of the different populations of particles to the
Hthernal Ae RT
V in [4c] creep resistance will change. This leads to the stress-rupture
plots having a sigmoidal shape, the two bounds corresponding to
H h V 1  H

H Hcreep  EHthermal [4d] the creep performance associated with small and the large
precipitate particles. The separation of these bounds depends on
D CH [4e]
d
the values of V0 for the two populations and the transition
K cp between them will be determined by the coarsening kinetics of J'
D p
3
1  D p 4 [4f] at that temperature.

The Jc will coarsen, particularly the finest particles, and this will
affect the creep properties in both tension and compression.
This is complicated by the reduction in volume fraction of Jc

2.0 5
IN100 704°C 738 MPa
IN100 704°C 738 MPa - Compression

4 As Received
1.5 Aged 600h at 704oC
Strain / %
Strain / %

3
1.0
2
As received - Tension
Aged - Tension
0.5 As received - Compression
Aged- Compression 1

0.0 0
0 5 10 15 20 0 20 40 60 80
Time / h Time / h

Figure 10 Effect of prior heat treatment on (a) initial stages of compression and tension creep curves and (b) longer term
compression creep curves for IN100 tested at 704oC and 738MPa.

Table 3 Summary of the microstructural characteristics of the turbine disc alloys studied..

Grain size Pm Jc population: Primary Secondary Tertiary

IN100 5 size (nm) 1000 200 10 to 20

volume fraction 20 34 6
Rene88DT 25 size (nm) 120 <15
volume fraction 33 7.2

904
Table 4 Calculated values of V 0 associated with each version of the model will incorporate the effects of particle-
cutting but suffice to say here that it weakens the alloy in creep
population of Jc particles as a function of temperature. relative to the performance of an alloy containing non-shearable
particles.
V0 (MPa)
Alloy T (°C)
Secondary Tertiary

IN100 649 20.36 51.01 12

o
IN100: 400MPa / 704 C
704 19.71 44.17 10
760 19.41 31.12
Rene88DT 649 38.89 46.68 8 Data

Strain / %
Model prediction
704 39.50 42.79 6

4
The model captures the principal features of the complex creep
behaviour of the disc alloys:
2
x A comparison of experimental and simulated creep curves
(Figure 11 ) shows that the model represents the unusual shapes 0
of creep curves for turbine disc alloys. In particular, a period of
negative creep is predicted for the longer-term test at 704oC and 0 200 400 600 800 1000 1200
400MPa. Time / h
x Figure 12, which shows simulations of tension and 8
compression creep with and without prior heat treatment, should o
IN100: 850MPa / 649 C
be compared with Figure 10. Elimination of the initial internal
stress by ageing has removed the initial tension/compression
asymmetry. However, coarsening of the tertiary J' has reduced 6 Data
Model prediction
the creep resistance both in tension and compression.
Strain / %

x Calculated rupture lives for the two alloys are shown with
4
the experimental data as continuous curves in Figure 8.
The reason that IN100 shows a greater fall in creep strength
with increasing temperature than Rene88DT is not due to it
being less stable microstructurally. Rather, the secondary J' 2
in Rene88DT, being significantly finer, provides
significantly greater creep resistance than in IN100.
Consequently, as the tertiary precipitate coarsens and 0
dissolves, the sigmoidal limits in the stress rupture
0 200 400 600 800 1000
behaviour are close together in Rene88DT.
Time / h
Although the model appears to represent the creep behaviour of
Figure 11: Model predictions of individual creep curves for
the disc alloys reasonably well, the original assumption of
IN100 at two different test conditions, compared with the
plastic deformation being restricted to the matrix is not valid in
experimental data.
these materials. There is ample evidence of extensive cutting of
the J' by dislocations [28] particularly as the short-term yield
stress is approached in peak-aged or under-aged alloys. A future

Table 5 Model parameters used to simulate the creep behaviour of IN100 and Rene88DT.

H0c Qd / j A 'Hs Ts hc K cp Qp
H* C
1 (MPa)-1 (MPa)
(s ) (kJ/mol) (kJ/mol) (K)
(s 1
) (kJ/mol)

Rene88DT 200 310 16035 49.83 1394 2.0x104 0.20 0.4V 106 280
IN100 2.6 x103 310 16035 49.83 1394 2.0x104 0.20 0.05V 106 280

905
 5. A.Graham, and K.F.A. Walles, J. Iron Steel Inst., 179
(1955) 105-120.
s. 6. R.W.Evans, J.D.Parker, and B.Wilshire, An extrapolation
8 Procedure for Long-Term Creep Strain and Creep Life
704°C 738 MPa Prediction with Special Reference to 1/2Cr1/2Mo1/4V
As Recd - Tension Ferritic Steels, Recent Advances in Creep and Fracture of
Aged - Tension Engineering Materials and Structures, Wilshire, B., and
6 As Recd - Comp
Aged - Comp
Owen, D.R.J., eds., Pineridge Press, Swansea, p. 135-184,
1982.
Strain / %

7. J.L.Chaboche, and G.Rousselier, On the Plastic and

4 Viscoplastic Constitutive Equations – Part II: Application
of Internal Variable Concepts to the 316 Stainless Steel, J.
Press.Vessel Tech., Vol. 105, p.159, 1983.
2 8. L.Meric, P.Poubanne and G.Cailletaud, Trans. ASME, J.
Eng. Mater.Tech. 113 (1991) 162-170.
9. L.M.Kachanov, On Creep Rupture Time, Izv. Ak. Nauk
0 SSSR Otdel. Tekh. Nauk 8 (1958), 26-31.
0 20 40 60 80 100 120 140 160
10. Y.N.Rabotnov, Creep Problems in Structural Members,
Proc. XII IUTAM Congress, Stanford, ed. by Hetenyi &
Time / h Vincenti, Springer, (1969)137
11. J.Ion, A. Barbosa, M.F. Ashby, B.F. Dyson, and M.
Figure 12 Simulated creep curves for IN100 tested in tension
McLean, The Modelling of Creep for Engineering Design -
and compression at 704oC and 738MPa in the as-received
I, Report DMA A115, The National Physical Laboratory,
condition and prior aged for 600 hours at 704oC.
Teddington, 1986.
Conclusions 12. G.S.Ansell J. Weertman, Trans AIME, 215 (1959) 838-843.
13. B.F.Dyson and S.Osgerby NPL Report DMA A116 (1993)
1. A model of creep deformation in particle-strengthened 14. B.F.Dyson in Creep Behaviour of Advanced Materials for
alloys that is expressed in terms of microstructural the 21st Century (1999), edited by R S Mishra, A K
parameters and physical constants has been developed. Mukherjee & K L Murty, TMS, Warrendale, USA, 3-12.
2. The model has been applied to a range of J' strengthened 15. T.M.Pollock and A.S.Argon, Creep Resistance Of CMSX-3
nickel-base superalloys taking into account the effect of Nickel-Base Superalloy Single-Crystals, Acta Metall.
dislocation accumulation, particle coarsening and grain Mater.40 (1992) 1-30.
boundary cavitation. 16. B.F.Dyson and M.McLean, Proc. of IUTAM Symposium on
Creep in Structures, 3-7 April 2000 Nagoya. ed. by
3. By accounting for the transition between unconstrained and S.Murakami and N.Ohno, pp 3-16, Kluwer Academic
constrained cavity growth, creep fracture strains can be Publishers 2001
calculated and the different creep response of superalloys 17. M.F.Ashby and B.F.Dyson, Creep Damage Mechanics and
with equiaxed and columnar grain structures can be Micromechanism, in Advances in Fracture Research in
predicted. Advances in Fracture Research, ed. by S R Valluri et al,
(1984) Pergamon Press, 1, 3-30.
4. Extension of the model to turbine disc alloys which operate
18. R.Stevens and P.E.J.Flewitt, Mater. Sci. Eng. 37 (1979)
at lower temperatures require the introduction of an initial
237-247.
matrix compressive stress which decays with time and
19. B.F.Dyson and T.B.Gibbons, , Acta Metall., 31 (1987) 17-27.
consideration of the differential strengthening associated
20. B.F.Dyson, in Creep Deformation: Fundamentals and
with multimodal J' distributions. Applications, eds. Mishra et al. TMS, (2002), 309-318.
Acknowledgements 21. M.V.Speight and W.Beeré, Metal Scienc, 9, (1975), 190-
191
The work reported in this paper was supported by EPSRC 22. H.C.Basoalto and B.F.Dyson (to be published)
(Grant Numbers GR/M93123/01 and GR/R11575/01) and by 23. F.A.McClintock, J. Appl. Mech. 4 (1968), 363
DARPA as part of the Accelerated Insertion of Materials 24. J.W.Hancock, Met. Sci. 10 (1976), 319.
Programme. The authors acknowledge with thanks materials 25. A.C.F.Cocks and M.F.Ashby, Progress in Materials
and discussions with colleagues from GEAE and Pratt & Science, 27 (1982), 189-244.
Whitney and their university support teams. 26. S.K.Sondhi, B.F.Dyson, and M.McLean, M., Tension-
References Compression Creep Asymmetry in a Turbine Disc
Superalloy: Roles of Internal Stress and Ageing, Acta
1. C.T.Sims, "Superalloys: Their Use And Requirements In Mater. In press
Advanced Energy Systems", pp1 -24, Proceedings of 27. M.Henry, GE Global Research, Schenectady, N.Y.- private
Superalloys 1976 communication.
2. D.N.Duhl in Superalloys II, ed. by C.T.Sims, W.Hagel and 28. S.Sinharoy, P.Virro-Nic and W.W.Milligan, Metall Mater
N.S.Stoloff, Wiley, New York (1987) Trans A32 (2001), 2021-2032 .
3. http://www.darpa.mil/dso/thrust/matdev/aim/index.html 29. R.A. Ricks, A.J. Porter, and R.C. Ecob, “The Growth of Jc
4. Andrade, E.N. da C., “On the Viscous Flow in Metals and Precipitates in Nickel-Base Superalloys,” Acta. metall., 3 1
Allied Phenomena”, Proc. R. Soc. London, A, 84, p.1-12, (1983), 43-53.
1910.
906

View publication stats