International Journal of Pressure Vessels and Piping 165 (2018) 224–228

Contents lists available at ScienceDirect

International Journal of Pressure Vessels and Piping

journal homepage: www.elsevier.com/locate/ijpvp

A modified theta projection model for the creep behaviour of creep-resistant T

steel
Ming Songa,∗, Tong Xua, Qijiang Wangb, Weilian Wangc, Yedong Zhoub, Min Gongb, Chao Suna
a
China Special Equipment Inspection and Research Institute, China
b
Baosteel Co., Ltd., China
c
GE-Wuhan Boiler, China

A R T I C LE I N FO A B S T R A C T

Keywords: A modified theta projection model is proposed for the description of creep behaviour. The model shows very
Theta projection model good flexibility in fitting the entire test data of the creep-resistant steel 12Cr1MoV over the three creep stages.
Creep strain The capability of describing the primary to secondary stage makes it easier to precisely evaluate the steady state
Minimum creep rate or minimum creep rate using the model. The model is derived and modified from the original theta projection
Creep-resistant steel
method. It also keeps the four indeterminate coefficients and can meet the convergence condition easily during
regression by the least squares method. In this paper, the minimum creep rates of 12Cr1MoV are also obtained
and discussed.

1. Introduction
For decades, the creep behaviour of metal materials for elevated
temperature applications has attracted much research interest, as creep
may induce a sudden rupture of components during long-term service
at high temperature, sometimes without any obvious signs. Therefore,
great efforts have been ceaselessly made to find a proper way to
characterize creep behaviours and precisely predict the creep life. The
most commonly used life prediction models are the Larson-Miller [1,2],
Manson-Haferd [3] and Orr-Sherby-Dorn [4] extrapolations. These
methods are all based on the experimental creep rupture time and have
been widely adopted by the related industry for allowable stress cal-
culation and remaining life assessment. However, to improve the design
accuracy and the economic efficiency of equipment, engineers want to
obtain more information about the creep deformation during the whole
life, not only the rupture time. Increasing attention has been paid to the
interpretation of properties based on creep strain data. In the codes of
equipment design and fabrication, such as the ASME codes for boilers
and pressure vessels, minimum creep rates are required for the de-
termination of allowable stresses [5]. With the development of com-
puter simulation analysis tools (such as the finite element method), a
good model that well describes the creep rates and their evolutions with
time, loads and temperature is strongly desired in design and life as-
sessment.
However, due to the nature of the large scattering and nonlinear
trend of creep test data, a tiny deviation in the data fitting would be
largely amplified during the long term rupture assessment. Thus, from
the aspect of creep constitutive description, a proper model and a
careful data fitting process are necessary. At present, many different
models have been developed to represent the creep strain, which range
from simple phenomenological to complex constitutive models. These
models mainly have two forms. One is based on the consideration of the
relationship between the loads and the steady state creep strain at a
given temperature because steady state creep is the primary behaviour
during the whole life, such as the power law by Norton [6], which has a
very simple form and is frequently used in structure calculations by the
finite element method. Most of the time, the creep tests are accelerated
experiments, and no steady state creep could be observed before rup-
ture; thus, the minimum creep rates (MCRs) are always measured in-
stead. The other form of models is the description of the accumulation
strain with time, such as the omega model [7,8] (for secondary and
tertiary creep), MHG model [3], and theta projection model [9] (for all
three stages of creep). To further minimize the data fitting and life
assessment deviations, many derivations have also been developed
[10,11], especially on the basis of the theta projection model, as it can
completely describe the three stages of creep life and can be used to
obtain the MCR easily. Most of the modified models have a more
complicated form with an increased number of parameters to be de-
termined compared to the original model, which make the function
more flexible, but the difficulty and time cost of the fitting process are
increased as well. Sometimes, a reasonable solution is hard to obtain. In
this paper, from the original thinking of the theta projection model, a

∗
Corresponding author.
E-mail address: songming@csei.org.cn (M. Song).

https://doi.org/10.1016/j.ijpvp.2018.07.007
Received 7 March 2018; Received in revised form 19 June 2018; Accepted 9 July 2018
Available online 10 July 2018
0308-0161/ © 2018 Published by Elsevier Ltd.
M. Song et al. International Journal of Pressure Vessels and Piping 165 (2018) 224–228

Table 1
Theta parameters of the fitting curves in Fig. 2.
T/°C σ/MPa θ1 θ2 θ3 θ4

525 180 1.18E-04 1.00E+04 5.17E-02 5.51E-05

525 200 1.33E-04 8.15E+04 5.48E-02 1.96E-04
525 230 4.79E-04 1.88E+01 6.66E-02 8.87E-04
550 180 1.66E-04 2.99E+02 6.33E-02 4.30E-04
550 200 3.31E-03 2.75E-02 6.29E-02 1.63E-03
550 230 5.02E-04 4.55E+01 7.82E-02 8.78E-03
575 100 3.18E-04 5.82E-02 1.04E-02 7.27E-05
575 120 3.46E-03 1.27E-03 1.88E-02 2.10E-04
575 150 5.80E-05 4.17E+04 3.93E-02 2.65E-03
600 75 2.75E-04 1.85E+00 5.33E-03 1.23E-04
600 90 8.86E-05 5.44E+00 1.69E-02 1.36E-04
600 110 1.31E-04 7.66E-01 2.17E-02 2.92E-04
600 150 1.44E-04 2.98E-01 2.27E-02 8.38E-04

2. Modified theta projection model

Fig. 1. Modified theta projection model.
modified model is proposed. The difference is that the model keeps four
theta parameters but shows good flexibility for the creep data, espe-
cially for the primary and secondary stages.
As we all know, the creep deformation of a metal under a constant
load and at a certain temperature includes three stages: the primary
stage with a decreasing creep rate, the secondary stage with a steady-
state creep rate, and the tertiary stage with the creep rate accelerating
with time until the material's rupture. Usually, under a certain condi-
tion, the secondary or steady-state creep stage may be absent such that
after the primary creep stage, the tertiary creep stage begins im-
mediately. In this case, only the minimum creep rate can be observed

Fig. 2. Creep data of 12Cr1MoV and fitting by modified theta projection model; (a) 525 °C; (b) 550 °C; (c) 575 °C; (d) 600 °C.

225
M. Song et al. International Journal of Pressure Vessels and Piping 165 (2018) 224–228

Table 2 The accumulated creep strain proposed here is

Z values estimations of different fitting models on 12Cr1MoV.
εt = θ1 ln (θ2 t + 1) − θ3 ln (1 − θ4 t ) (1)
Model Z values for times to specified strain (%)
where εt is the accumulated creep strain, and θi is a parameter that is
0.2 0.5 1.0 2.0 5.0 temperature and stress dependent and can be described empirically as
Theta projection model (5 5.9 2.9 1.3 1.2 1.0 logθi = a + bσ + cT + dσT (2)
parameters) [9]
MHG model* [3] 7.2 6.1 4.4 2.2 2.3 The creep strain function (Eq. (1)) has a similar form to that by
Modified theta projection 2.7 1.6 1.4 1.2 1.0 Evans [9]. The difference is that the exponential terms are replaced by
(proposed here) the logarithm forms. The first logarithm term represents the primary to
secondary stage, while the second logarithm term represents the sec-
Note*: The multilinear combinations of σ and ε is of the form
ondary to tertiary stage (Fig. 1). Usually, the creep of metal is con-
F (ε, σ ) = a0 + a1 logσ 2 + a2 logε 3 .
sidered a thermally activated procedure, and so is the model proposed
here.
For the primary to secondary stage, the immediately high strain
reduces rapidly with time. This process is mostly contributed by dis-
location movement and strengthening at a constant load. According to
the work hardening explanation of creep [9], the creep rate decreases
gradually with an increase in the activation energy for deformation
brought about by work hardening. After immediate deformation, all
elements have a critical stress σy equal to the applied stress. As the
strain accumulates, the internal stresses of all elements are assumed to
rise to σy′ . For the coefficient of work hardening, h,
σy′ − σy = hε (3)
The activation energy is
u = c(σy′ − σy )m = c(hε )m (4)
where c and m are constants that depend on the micro-mechanism of
the dislocation process. If there are N such elements per unit volume,
the creep rate is:
dε
Fig. 3. Creep rates of 12Cr1MoV at 575 °C,100 MPa, and the minimum value = AvNexp [−c(hε )m /kT ]
dt (5)
determined by Eq. (17).
Where A is the strain arising from an element with activation en-
ergy, u; v is a slowly varying function of u. A, v and N can be regarded
as constant [9].
For a number of dislocation processes, m will be close to 1, so in-
tegration of Eq. (5) gives
kT ⎛ ch
ε= ln AvNt + 1⎞
ch ⎝ kT ⎠ (6)
kT ch
Let θ1 = ch
and θ2 = kT
AνN ; then,
ε = θ1ln(θ2 t + 1) (7)
Eq. (7) shows the logarithm creep from the first to secondary stage.
For the secondary to tertiary stage, the deformation is attributable
to three causes: (1) microstructural instability, such as grain growth,
recrystallization and changes in particle dispersion; (2) the develop-
ment of cavities to a sufficient size such that they affect the deformation
rate; and (3) the occurrence of necking, which causes a localized re-
duction in the stress-bearing area. The second phase particles pre-
cipitating at grain boundaries offer sites for vacancy nucleation. Then,
Fig. 4. The minimum creep rates show power function relationship with the
the vacancies accumulate to form cavities under diffusion and loading.
external loads, witch coincide with the Norton's power law. (data of 525 °C, At a lower experimental condition (a lower temperature and a lower
550 °C, 575 °C were from tubes, data of 600 °C was from pipe). load), if necking is ignored, the stages could be considered as Coble
diffusion creep [12,15]. The concentration of vacancies C0 in a metal at
the thermal equilibrium state should have the follow relationship:
instead of the steady-state creep. However, regardless of whether the
secondary stage exists, all the creep processes can be considered as a C0 ∝ exp (−Qf / RT ) (8)
competition between two types of mechanisms. One is the strain
where Qf is the energy required to form a vacancy. Each time a vacancy
hardening or a decrease in free or mobile dislocations that contribute to
is produced at the grain boundaries experiencing a tensile stress σ , the
a decrease of creep; the other is the mechanism of the loss of resistance
work done by the stress is on the order of Ωσ , where Ω is the volume of
to creep, which usually comes from microstructure degradation and
a mol. vacancy. Therefore, the energy required to create a vacancy at
vacancy formation. All these phenomena are related to the external
the grain boundary is reduced from Qf to (Qf −Ωσ) [12]. Then, the
load.
concentration of vacancies under loading changes to

226
M. Song et al. International Journal of Pressure Vessels and Piping 165 (2018) 224–228

Then, integration of Eq. (12) gives

1
ε=− ln (−K 4 K2 t − K 4 C )
K4 (13)

When t = 0 and ε = 0 ,
1
ε=− ln (1 − K 4 K2 t )
K4 (14)
1
Let θ3 = K4
and θ4 = K 4 K2 ; the creep strain at the tertiary stage is
ε = −θ3 ln (1 − θ4 t ) (15)

Eq. (15) has the same form of the Omega model, which is derived by
taking into account damage accumulations due to the load and tem-
perature with time [7].

3. Creep strain test and fitting process

12Cr1MoV is a Chinese creep-resistant enhanced ferrite steel for the

Fig. 5. Monkman-Grant relationships between the minimum creep rate with fabrication of super heaters, reheaters and headers of power boilers
rupture time. with an outlet temperature below 580 °C. It was designed with a very
good balance between the performance and cost. 12Cr1MoV has che-
mical composition and microstructure similar to those of the steels
Gr.22 and Gr.23 and long term mechanical property between that of
these two grades. With the alloy element content less than 1 wt%, its
weldability is much better than that of Gr.22 and Gr.23. With such
advantages, 12Cr1MoV is expected to be incorporated in the design and
construction codes of ASME BPVI in the near future. To determine the
allowable stress values at elevated temperature for the ASME codes,
creep tests were carried out.
The tests at 525 °C, 550 °C and 575 °C were carried out by Baosteel,
and the tests at 600 °C were carried out by GE Wuhan Boiler. Fig. 2
shows the creep data and the model fitting of 12Cr1MoV, and theta
parameters are listed in Table 1. The curves show a very good fitting
performance for the modified theta projection model proposed here,
especially in terms of the capabilities of describing the details of the
primary stage (insets of Fig. 2 (a) to (d)), which is beneficial for the
precise determination of the minimum creep rate.
The European Creep Collaborative Committee (ECCC) proposed a Z
parameter to provide a means of quantifying model-fitting effectiveness
Fig. 6. Creep strain with similar lifetime at different temperature and load in specific creep strain regimes, which is defined as [11].
couples. Z= 102.5SA − RLT (16)

where SA − RLT is the standard deviation of residual log time for specific
C0 exp [Ωσ / RT ] (9)
creep strains, and Z ≥ 4 is unacceptable for single-cast assessments. The
As a consequence, a vacancy concentration gradient will exist be- Z values for 12Cr1MoV obtained by the model fitting here are listed in
tween the high and low stress areas, which will cause vacancy and atom Table 2. The fitting approaches are a unified global search with the least
diffusion and then contribute to the deformation. According to Fick's squares method.
law, the diffusive flux is
C0 exp [Ωσ/RT] − C0 4. Minimum creep rate
Jv = −Dv
d (10)
Usually, the MCR is obtained directly by the drawing method on the
where Dv represents the vacancy diffusion coefficient, and d represents differential diagram of the creep strain. For high quality experimental
the average path length. In a unit time, the tensile boundaries move data, there is no problem in the accurate evaluation of the MCR.
apart by a distance Jv Ω. Then, the strain rate should follow However, for a set of data with too much fluctuation (the zigzag curve),
dε JΩ D ΩC the magnification effect of the differential will induce a large deviation,
= −K1 v = K1 v 2 0 [exp (Ωσ /RT ) − exp (−Ωσ /RT )] which is sometimes several orders of magnitude. Thus, a calculation
dt d d (11)
method for the curve of data fitting is recommended [9]. The creep tests
Where K1 is a constant. With a stress level below yielding, σ≈ K3 ε . in the laboratory always undergo a shorter primary and steady creep
Usually, for diffusional creep, the vacancy concentration under stress is stage than in practical service, and the point of the MCR is usually very
much higher than that at the thermal equilibrium state [12]. That is, close to the origin of the coordinates. Thus, a good data fitting process
D ΩC
exp (Ωσ /RT ) ≫ exp (−Ωσ /RT ) , so if let K2 = K1 v 2 0 and K 4 = ΩK3/ RT , of the primary to secondary stage is necessary to precisely estimate the
d
then the strain rate transforms to MCR.
For the modified theta projection model proposed here, the creep
dε
= K2 [exp (K 4 ε )] rate (CR) can be derived from Eq. (1) by the differential with respect to
dt (12)
t:

227
M. Song et al.
θ1 θ2 θ3 θ4
CR = +
θ2 t + 1 1 − θ4 t (17)
Fig. 3 shows the creep rate changes with time for 12Cr1MoV at
575 °C and 100 MPa. The MCR is found at 2245 h. It is an early stage
compared to the rupture time of 12236 h, and the fitting process suc-
cessfully averts the large data fluctuations at this stage. In Fig. 4, a
power function relationship between the MCRs and external loads is
illustrated clearly, which coincides with Norton's power law and verify
the activation mechanism of the creep expressed by Eq. (11). It should
be note that, the specimens for 600 °C tests are sampled from pipes, but
the others are sampled from tubes. Because of the difference in wall
thickness between these two types of production, the heat treatment of
pipes and tubes are different. As a consequence, the final micro-
structures are ferrite and pearlite in tubes, but tempered bainite in
pipes. For low alloy steel 12Cr1MoV, comparing to ferrite and pearlite,
tempered bainite will enhance the short term creep resistance at higher
load by multi-mechanism strengthen, but weaken the long term creep
resistance at lower load because of its lower structure stability. In
Fig. 5, the Monkman-Grant relationships [13] between the MCR and
lifetime can be observed. With the rupture time increasing, the MCR
decreases fast at first and then more slowly as the lifetime is extended.
Additionally, with the increase of temperature, the MCR data tend to
distribute at a lower level. These two trends indicate the coupling of
two factors, the load and the temperature, which can change the me-
chanism of creep gradually from dislocation movements and granular
slides to diffusions. As we all know, diffusion is thermally activated; a
higher temperature induces greater element and vacancy diffusion that
contributes to the creep (Fig. 6), and a longer lifetime (lower load) gives
the diffusion more time (Fig. 5). From the morphologies of the ruptured
materials, the features of diffusion-controlled creep can be dis-
tinguished easily, such as the decrease of the elongation (sometimes
approaching zero), the increase of the distribution area of the cavities
along the tensile axis, and an entirely intergranular fractography.
5. Conclusion
The modified theta projection model proposed in this paper showed
good fitting qualities to the creep data of 12Cr1MoV. Although it has
just four parameters, this model still has better performance than the
modified 5 parameter theta projection model and the MHG model. It
maintains excellent flexibility for the test data from several hundred
hours to over ten thousand hours under different experimental condi-
tions, especially for the primary to secondary stage, which is beneficial
for the accurate calculation of the MCR. Because many secondary creep
models, such as the power law [6], Monkman-Grant relationship [13]
and Sinh law [14], are based on the relationship of the external load
and MCR, the significance of a precise estimation of the MCR is evident.
228
International Journal of Pressure Vessels and Piping 165 (2018) 224–228
As the lifetime increases, the creep will be controlled more by diffusion,
which may induce an ultra-low creep rate at the secondary stage. When
the creep is fully controlled by diffusion (an extreme condition [15]),
the creep rate might approach zero and will not show any creep ac-
celeration stage before rupture, for which θ3 and θ4 of Eq. (1) will be
zero. Until now, there has not been a clear and practical mathematical
expression for the theta parameters, even in the model by Evans, and
most of the life prediction processes are based on empirical formulas
such as Eq. (2), which are not able to distinguish the different con-
trolling mechanisms of the creep between the short and long terms.
Thus, it is quite difficult to predict the pure diffusion creep via ac-
celerated creep tests within only hundreds to thousands of hours. The
only way at the present to improve the accuracy of creep strain and
lifetime prediction for material engineers is to make an effort to accu-
mulate creep data as long as possible.
Acknowledgement
This work was supported by The National Key Research and
Development Program of China [2016YFC0801901].
References
[1] F.R. Larson, J. Miller, A time-temperature relationship for rupture and creep
stresses, ASME Trans. 5 (74) (1952) 765–775.
[2] K. Maruyama, F. Abe, H. Sato, J. Shimojo, N. Sekido, K. Yoshimi, On the physical
basis of a Larson-Miller constant of 20, Int. J. Pres. Ves. Pip. 159 (2018) 93–100.
[3] S.S. Manson, A.M. Haferd, A Linear Time-temperature Relation for Extrapolation of
Creep and Stress-rupture Data, Technical Report Archive & Image Library, 1953.
[4] R.L. Orr, O.D. Sherby, J.E. Dorn, Correlations of rupture data for metals at elevated
temperatures, Transactions of the American Society for Metals 46 (7) (1953)
113–156.
[5] ASME, ASME BPV Code, Section II: Materials, (2017).
[6] F.H. Norton, The Creep of Steel at High Temperatures, (1929).
[7] M. Prager, Development of the MPC omega method for life assessment in the creep
range, J. Pressure Vessel Technol. 117 (2) (1995) 95–103.
[8] H. Chen, G.R. Zhu, J.M. Gong, Creep life prediction for P91/12Cr1MoV dissimilar
joint based on the omega method, Procedia Engineering 130 (2015) 1143–1147.
[9] R.W. Evans, B. Wilshire, Creep of Metals and Alloys, The Institute of Metals,
London, 1985.
[10] W.D. Day, A.P. Gordon, A modified theta projection creep model for a nickel-based
super-alloy, ASME Turbo Expo 2013: Turbine Technical Conference and Exposition,
2013, pp. V7A–V26A.
[11] S.R. Holdsworth, M. Askins, A. Baker, E. Gariboldi, S. Holmström, A. Klenk, et al.,
Factors influencing creep model equation selection, Int. J. Pres. Ves. Pip. 85 (1–2)
(2008) 80–88.
[12] M.T. Perez-Prado, M.E. Kassner, Chapter 3-Diffusional Creep. Fundamentals of
Creep in Metals and Alloys, third ed., Butterworth-Heinemann, Boston, 2015, pp.
103–108.
[13] G. Sundararajan, The Monkman-Grant relationship, Mater. Sci. Eng. 112 (1989)
205–214.
[14] A. Nadai, The Influence of Time upon Creep: the Hyperbolic Sine Creep Law,
(1938).
[15] F. Abe, T. Kern, R. Viswanathan, Creep Resistant Steels, CRC, USA, 2008.