Vous êtes sur la page 1sur 19

Accepted Manuscript

Testing and assessment of cracking in P91 steels under creep-

fatigue loading conditions

N. Ab Razak, C.M. Davies, K.M. Nikbin

PII: S1350-6307(17)30566-6
DOI: doi: 10.1016/j.engfailanal.2017.08.007
Reference: EFA 3261
To appear in: Engineering Failure Analysis
Received date: 30 April 2017
Revised date: 5 August 2017
Accepted date: 21 August 2017

Please cite this article as: N. Ab Razak, C.M. Davies, K.M. Nikbin , Testing and
assessment of cracking in P91 steels under creep-fatigue loading conditions, Engineering
Failure Analysis (2017), doi: 10.1016/j.engfailanal.2017.08.007

This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.

Testing and Assessment of Cracking in P91 Steels under

Creep-Fatigue Loading Conditions
N. Ab Razaka,b, C.M. Daviesa, K.M. Nikbina

Department of Mechanical Engineering, Imperial College London, London SW7 2AZ, United Kingdom
Department of Mechanical Engineering, University Malaysia Pahang, Pekan Pahang 26600, Malaysia


Future green energy options dictate that renewable energy source must be utilized when available. This poses a
challenge to conventional power plant, which has generally been designed for base-load conditions, as they now

need to operate in a ‘flexible’ manner. This flexible operation, in high temperature power plant components, could
lead to a combination of fatigue and creep crack growth failure. Thus a better characterization of interactive creep-

fatigue crack growth behavior is required especially for assessing long term failure in plant. The industrial codes,
such as R5 or BS7910, treat this interaction using linear accumulation of damage. However, this does not consider
the degradation of properties and reduction in creep ductility in long term operations and their effect on the
subsequent creep/fatigue behaviour. In this work creep, fatigue and creep-fatigue crack growth (CFCG) tests were
performed on compact tension specimen of P91 steel in the as received and ex-service conditions at temperatures
ranging between 600°C to 625°C, with the hold-times ranging from static to 600 seconds. The experimental
results in addition to appropriate data from the literature have been analysed using stress intensity factor range, ΔK
appropriate under fatigue control and the creep fracture mechanics parameter C* relevant under creep control.

Scanning electron macroscopy (SEM) analysis confirms the influence of frequency on the mode of cracking.
Within the scatter of experimental data for the present short term accelerated tests a linear cummulative damage
rule can still predict the creep/fatigue interaction. However the effects due to low frequency cyclic loading as well
as degraded the steel under ex-service conditions tending to reductions in creep ductility show factors of two or
more faster cracking rates compared to as received static testing. Unavailability of long term tests at low stresses

may pose additional problems under creep since creep is stress state controlled but fatigue in not. However using
plane strain predicitons of crack initiation and growth data using the NSW multiaxial ductility creep crack growth

model suggest that conservative predictions of long term cracking can still be made under creep/fatigue and when
there is a marked reduction creep failure ductility.

Keywords: Creep fatigue; Crack Growth behaviour; P91 steel


1. Introduction

Many conventional power plants are now required to operate in a ‘flexible’ manner in response to

energy demand and availability of renewable energy, though designed to operate at base loads. This
flexible operation implies that the mechanical and thermal loads on high temperature components are
cyclic. This cyclic operation may lead to interactive creep-fatigue failure mechanisms taking place that
is known to accelerate failure compared to static creep loads alone. Therefore, it is imperative that the
creep-fatigue crack growth behavior of power-plant components are characterized and understood.
Many codes of practice such as R5 [1] and BS7910 [2] assume a linear accumulation of creep and
fatigue damage whilst not taking into account any material degradation at long terms. Defects may
already exist in the component which under static load could have grown due to creep process. The
long term damage which can occur due to formation, growth and coalescence of cavities at various
microstructural sites such as in grain boundaries, cavities, dislocations, coarsening precipitates, second
phase regions and other inclusions [3, 4] effectively degrades the material which will may have an
additional effect on creep/fatigue interaction.

Creep fatigue crack growth (CFCG) behaviour of engineering alloys has been investigated by a


a crack length
a0, af initial crack length and final crack length measurements
Δa amount of crack growth
a crack growth rate
A creep constant
B,BN thickness and thickness with side groove
C* creep fracture mechanic parameter
da/dN fatigue crack growth rate
da/dt creep crack growth rate
D,  CCG power law coefficient and exponent
 e  , n  non dimensionless function of θ and n

f uniaxial creep failure strain
 *f multiaxial creep ductility

f frequency
H,η creep dimensionless coefficient
In dimensionless integration constant

ΔK stress intensity factor range
N number of cycle
n power law creep exponent
P applied load
rc creep process zone size
T temperature
ti creep initiation time
th hold time

tf test duration
W width
 ,p fatigue material constant
 Load line displacement rate

number of researchers [5-10]. Lu et.al [5] investigated the effect of temperature and hold time on the
CFCG behaviour of a nickel based superalloy and showed that as the hold time increase, the crack

growth behaviour changed from cycle dependent to time dependent behaviour. This transition occurred
at a smaller hold time as the test temperature was increased. Bassi et al. [7] conducted CFCG tests on

T/P91 power plant steel and employed a simple superposition approach to sum creep and fatigue
damage contributions and predict the CFCG behaviour. The results show that under creep-fatigue
loading conditions with a holding time 0.1 h, the crack growth behaviour is as a pure fatigue crack

growth (FCG) test. For hold times between 1 and 10 h, the crack growth behaves as a pure creep crack
growth (CCG) test, whereas for hold times between 0.1 and 1 h interactive CFCG behaviour was
observed. The CFCG behaviour on four type of steels namely, 316L, 1CrMoV, P91 and P22, which
were tested at a range of frequencies was examined by Mehmanparast et. al. [8] who found that for a

cyclic frequency below 0.1 Hz, the crack growth behaviour was time dependent and was correlated
with the creep fracture mechanics parameter, C*. A simple linear cumulative rule was also used to
describe the effects of creep-fatigue interaction on the crack growth rates.

In this work an experimental analysis of CFCG has been performed on P91 steel. CFCG tests have
been performed on compact tension, C(T), samples at a range of temperatures between 600°C and
625 °C for a range of hold time static to 600s. Tests have been performed on two ex-service materials
and a new material, here denoted the as-received (AR) material. The crack growth behavior has been
correlated with the stress intensity factor range, ΔK, and the creep fracture mechanics parameter C*.
The results presented are compared to a range of FCG, CCG and CFCG test data found in the literature
[6, 8, 10, 11]. Fractography has also been performed on fracture surface to identify the dominant
failure mechanism and linear cumulative rule approach employed to predict results obtained.

2. Material

The properties of the three P91 steels tested in this work are shown in Table 1. Note that hereon for
brevity, the as-received material is denoted ‘P91-A’. Of the two ex-service P91 steel materials tested,
one was previously in operation for over 110,000 hours at 590°C (denoted P91-B) and the other at
600°C for over 100,00 hours (denoted P91-C). Metallographic analysis has been performed on the as-
received material P91-A and the ex-service material P91-B, as shown in Fig. 1 where the materials was
etched using Villela agent (containing 1g of picric acid, 5ml of HCl and 100 ml of ethanol). Similar
microstructures are seen in both as received and ex-service materials where the expected lath
martensite microstructure is observed.

3. Experimental Testing

Creep-fatigue crack growth testing has been performed according to the testing standard, ASTM E-
2760 [12]. A number of ex-service and as received P91 compact tension, C(T) standard W=50mm,
B=25mm and BN =20mm samples were tested from different batches as shown in Table 2. Note that
two tests which contributed to a larger ASTM organised round robin project, [13, 14] were also tested.

The CFCG testing was performed on a creep machine using pneumatic load-lifting equipment. The
direct current potential drop (PD) technique was utilized to continuously monitor the crack length on

the specimen. A linear variable differential transformer (LVDT) was used to monitor the load line
displacement. A triangular waveform was used and loading and unloading times were held constant
(approximately 2s load and unload times) as shown in Fig. 2 (a). Hold time of duration of 600s, 60s,
and 30s were superimposed on the maximum load. All CFCG tests were performed at 600°C, 620°C
and 625°C (see Table 2) at a load ratio, R of 0.1. Fatigue crack growth testing was performed at 600°C
using a servo-hydraulic machine. The FCG test was performed at a high frequency, f of 10Hz at a load
ratio, R of 0.1. The same PD technique was employed in the FCG test and the loading form is shown in
Fig. 2(b).

All the tests were interrupted prior to full fracture and subsequently broken open at room temperature
by high frequency fatigue loading, in order to calibrate the PD crack length measurements. The initial
crack length, a0, and the final crack length, af, have been measured by averaging 9 measurements along
the crack front [12].

4. Correlation of fatigue crack growth and creep deformation


The data collected from the experimental tests and relevant data form the literatures have been analysed

using fracture mechanics correlating parameters to identify crack initiation and crack growth rates
under different frequencies. For practical purposes these are short term accelerated test (<5000h) and
relative comparisons will need to be made. In reality components operate at long terms and upto design
lives of 30 years or more. For this purpose the analysed data are then compared with the NSW [15]

crack growth rate prediction model to identify the plane strain upperbound of crack growth rate under
cyclic conditions. This is more relevant under creep dominant conditions as creep damage is known to
be stress state controlled and fatigue on the other hand is not.

4.1 Fatigue crack growth

Fatigue cracking process at the microstructural level is based on the accumulation of local plastic
deformation to drive the crack forward. However at the macro level the crack tip stress fields are
predominantly elastic compared to the size of geometry. Therefore under fatigue dominant conditions,
the crack growth rate per cycle, da/dN can be described by the elastic stress intensity factor K using the
well known Paris law [16] which can be expressed as

da dN   K p (1)

where λ and p is a material constant. It is found that for most steels p=3-5. The stress intensity factor
range, ∆K, calculated using following equation;

P 2  a /W
K  f (a / W ) (2)
( BBN ) 1/2
W 1/2
1  a / W 3/2
where ∆P is the load range, a is the crack length, W is the width, B is the thickness, BN is the thickness
with side groove and f(a/W) is calculated using

f (a / W )  0.886  4.64(a / W ) 13.32(a / W ) 2  14.72( a / W )3  5.6( a / W ) 4 (3)

As fatigue is cycle dependent phenomenon and hardly affected by stress state at the crack tip it’s
assessment using Eq (1) is convenient and sufficiently accurate for predicting steady state cracking in

cases where little plasticity or ductile creep deformation is present. This usually the case in large
component in plant during long term operation.

4.2 Creep crack growth and creep crack initiation

Creep crack growth (CCG) is a time dependent process which is hugely affected by the state of stress at
the crack tip and therefore needs more complex analysis both to correlate the data and to predict failure.

For steady state creep dominant conditions, the crack growth rate is expected to be described by the C*

a  DC 

where D and  are the CCG power-law coefficient and exponent respectively. C* can be determined
experimentally using [17]

C*  H (5)
B(W  a )

where P is the applied load, and  is load line displacement, H and  is a dimensionless coefficient
that depends on specimen geometry. For a C(T) specimen, H = n/(n+1) and  = 2.2, where n is the

power-law creep exponent.


In static CCG tests, a period of time exists prior to the on-set of crack extension. This is often referred
to as the creep crack initiation (CCI) time. Under creep-fatigue conditions it is often assumed that the
crack grows from time zero i.e no initiation time exists. However, the creep crack initiation time may

be defined as the time for a small amount of crack extension to occur, usually 0.2 mm [18], which may
be estimated from

a a
ti    (6)


where the D and  are as defined in Eq (4).

Under steady state creep conditions, uniaxial creep ductility data is used in crack initiation and crack
growth predictions based on the multiaxial ductility NSW model [19] assuming a creep process zone rc.
For fracture mechanics samples the crack tip stress field is characterized by C* and multiaxial failure
strain controls the level of constraint controlling the stress state. The NSW model used to predict
cracking rate as a function of multiaxial ductility  *f and C* is therefore given as:

(n  1)  C*  n 1
aNSW  ( Arc )1/(n 1)   (7)
 *f  In 
 

where A is a material creep constant, n is the power law stress exponent, In is a dimensionless stress
state constant dependent on n, rc is the creep process zone which is relatively insensitive due to the
small fractional power and  *f is taken as an upperbound failure strain in short term tests under plane
stress and a fraction of 1/30 of it under plain strain lower bound condition under extreme low ductility
conditions. However a factor of  *f =  f 10 would be sufficient for engineering alloys [19]. In this
case, either the reduction in area or elongation failure strain in short term tests determined under plane

stress condition is used in Eq. (7) [16]. When material properties are not available, an approximate
solution to Eq.(7) for predicting crack growth rate for most engineering materials [15] can be given in
the form

a C*0.85 (8)
 *f

where 𝐷 ′ would be given as 3 and 90 for a lower/upper bound plane stress/strain prediction for a wide
range of alloys [16]. The upperbound can be usually reduced to a factor of 3x10 for long term cracking
in engineering alloys where a good extent of creep toughness exists.
In the same way creep crack initiation can be predicted by using the approximate NSW which can be
expressed as

a *f
ti  (9)

where a can be taken as 0.2 mm for the purpose of laboratory test when the measurement sensitivity

for crack extension is high.


4.3 Creep fatigue crack growth


For the interaction of creep/fatigue a linear accumulation of cracking is assumed similar to the ones
adopted in the codes [1, 2]. The total crack growth per cycle is contributed by the cyclic dependent

(fatigue) component and the time dependent (creep) component, which can be expressed as a linear

 da   da   da / dt creep

 dN     (10)
 total  dN  fatigue 3600 f

where f is the frequency of load cycle in Hz. The fatigue crack growth rate in Eq (1) and creep crack
growth rate in Eq (4) can be substituted into Eq (10) and may expressed as

 da  p  DC
   K (11)
 dN total 3600 f

where the first term on the right hand side of Eq (11) gives the contribution from the cyclic (fatigue)
component and the second the contribution from the time dependent (creep) process. As these are two
distinct processes they can be determined separately and added up. It should be noted that as creep
under low cycle is stress state controlled the cracking rate will vary depending on a number of factors

such as reduction in failure ductility due to embrittlement in ex-service degradation, applied load,
geometric constraint.

5. Result and discussion

By using data from as received and ex-service P91 steel the focus of the analysis is based on
identifying the creep/fatigue interaction and the effect of material degradation on the subsequent creep
and fatigue life. Within the level of scatter of the experimental data it is important to state that any
other approach other than the linear accumulation method will not yield any improvement in the
analysis. It is also clear that the analysis of the present range of short term test might not indicate the
creep/fatigue behaviour at long term operational times under different states of stress, However it has
been shown that the with a knowledge of long term uniaxial creep ductility the NSW model can
conservatively bound the plane strain region of crack initiation and growth.

5.1. Crack growth correlation with the stress intensity factor range

Using the stress intensity factor ΔK is more appropriate for the fatigue controlled cracking where
stress state has less effect on cracking and the crack growth is effectively transgranular cycle
dependent. Fig. 3 (a) compares the crack extension, Δa, against number of cycles normalized by the

number of cycles to failure, N/Nf and Fig. 3 (b) shows the load line displacement (LLD) relationship
against time normalized by tests duration for all specimens tested. The x-axes have been normalized so
that each curve can be clearly observed. Note that the tests duration is dependent on the point where
the test was interrupted, which corresponded to a point of significant acceleration in crack growth rate.
In Fig. 3(a), at frequency of 0.0017 Hz (i.e. 600s hold time), test CT-C2, which was subjected to the
highest load and temperature, had the shortest test duration, as expected. Generally all tests show crack
growth from initial loading. Comparing frequency and hold time effects for the ex-service material at
600 °C, the test duration become shorter as the frequency increases. Note that test CT-B4 which was

tested in fatigue dominant conditions at high frequency thus had a very short test duration. Considering
the applied load and temperature, CT-A had significantly longer test duration than the ex-service
material and a large LLD and crack extension prior to test completion, which may be due to thermal
aging or creep damage accumulation effects in the ex-service material.

Fig. 4 shows the crack growth rate per cycle, da/dN, correlated with the stress intensity factor range,
ΔK. In order to investigate the effect of various frequencies on the CFCG growth behavior, data from

literature for CFCG [6, 8, 11] and FCG [10, 11] have been included in Fig. 4. The dashed line

illustrates the regression fit made to the data with a frequency less than 0.002 Hz and referred to as
CFCG line. The dotted and dotted line illustrates the regression fit made to the data with a frequency
less than 0.002Hz and larger than 0.01Hz. At high frequencies (>0.01 Hz), the CFCG data can be
regarded as a high cycle fatigue crack growth and the data for all temperatures considered fall close to

each other. At lower frequencies (< 0.002Hz), the crack growth rate progressively increases with a
decreasing frequency and an increase in temperature, due to a significant creep contribution.

5.2. Crack growth correlation with the C* parameter

The low frequency CFCG test data has been correlated with the C* parameter and compared with
available static CCG data for P91 steel in the temperature range of 600°C to 625°C taken from
literature [8, 11, 20] as shown in Fig. 5. In Fig. 5 the open symbols refer to CFCG data whilst the
closed symbol indicate the static CCG data. The regression line has been made to the CFCG test data
and CCG test data as shown in Fig. 5. It is shown that the cyclic crack growth rate accelerate by a
factor of 5 than the static CCG rate. The predictive aproximate NSW model and P91 CCG scatter data
band [20] at 580°C to 625°C are also plotted in Fig. 5. It is apparent that the CFCG data fall slightly
above the static CCG data. There is also some indication that an increase in temperature results in a
high crack growth rate, as expected. Given the level of scatter any detailed assessment of individual
tests would not yield any further information.

Using the approximate NSW model (Eq (8)) the upper/lower bound predictions have been plotted in
Fig. 5. A power-law creep exponent of n = 8.2 [6] and a uniaxial failure strain of 33% [6] has been
used in the NSW model which are the best approximations for a test temperature of 625°C. For the
short term data the plane-stress predcitions are show in for the fast rate tests. Using a factor εf /10 for
the plane strain line it is observed in Fig. 5that most data especially the slower rate longer term data
generally fall below the upperbound. Note that the NSW parameters drawn in Fig. 5 are nominal for
625°C which in reality are dependent on creep exponent, n and failure ductility which may vary with
material and temperature. Also as material degradation due to long term service as well as low stresses
tend to reduce failure strain cracking rates generally increase. By using lower failure strains to reflect
long term behaviour the NSW model gives conservative predictions. This is particularly relevant for
long term low C* predicitons where material degardation may exhibit lower creep ductiliy both due to
material degradation and due to the failure strain sensitivity to stress.

5.3. Creep-crack initiation

The crack growth can be caused by creep, fatigue or the interaction between two mechanisms under
creep fatigue loading. The crack growth is usually assumed to be due to creep and not fatigue [1],
hence the initiation time, ti is defined as the time to for crack extension Δa = 0.2mm under static and

low cycle loading. However in the initiation time model, it is assumed that the crack grows from the
time of first loading. Assuming the transition time is exceeded [21], the creep crack initiation has been
analysed in terms of C* parameter and predicted using Eq.(9). Fig. 6 shows, for available data, the
measured initiation times, ti correlation with C* for the crack extension Δa=0.2mm. Also included in
the figure is the NSW model prediction for the initiation times. It can be seen in Fig. 6 that the
experimental data fall at the plane stress bound of NSW model regardless of being a static or a low
frequency test. Based on steady state crack growth rate, the plane stress and plane strain NSW model
provides a reasonable estimate on the creep initiation time. Though this model is strictly valid for static

CCG only, the conservative assumption that the crack grows from time zero and under steady state
conditions appears to have accounted for the acceleration in crack growth due to fatigue for these

5.4. Creep-fatigue interaction


The frequency dependence of CFCG behaviour can be predicted using Eq. (11) where the constants,

determined from CFCG data are creep parameter, D = 10 and φ = 0.8 and fatigue parameter
λ =1.5×10-8, p = 3.57. Fig. 7 is constructed by identifying the pure fatigue level from the da/dN curve

at a fixed ΔK= 25MPam in Fig. 4. For the CCG band Eq. (11) is used assuming the fatigue
component to be zero. The mean cracking rates from the static CCG data band from Fig. 5 are taken at
a fixed value of C*=1.0×10-4 MPam/h which approximately corresponds to ΔK= 25MPam for the

fatigue tests. For each cyclic test in Fig. 5 a point is derived at a fixed C* and a best fit dotted line using
Eq. (11) is constructed with a slope of -1 for the CFCG tests which suggests a higher cracking rate
compared to the static tests. It should be noted in Fig. 7 that a scatter band is shown for each data point
indicating a level of uncertainty in the data. However the overall trends in the figure are clear. In

addition a line is constructed in the same fashion using the upper bound NSW line in Fig. 5.

The interaction diagram in Fig. 7 shows, therefore, that the cyclic ex-service steel at low frequencies
exhibits about a factor of four times the cracking rate of the mean CCG static data and that the NSW
line predicts a higher conservative upper bound. At the same time the only interaction with the fatigue
horizontal line is at the crossing points where both cracking rates are similar. The shift to the right of
the interaction region indicated an increase in fatigue dominance and constraint.

At intermediate frequencies (0.001< f < 0.01Hz), therefore, where the static CCG and high FCG
line intersect, both creep and fatigue processes are significant and mixed intergranular and
transgranular fracture is expected. As explained in [22] both types of process are likely to develop
intermittently through or around individual grains. Hence, at intermediate frequencies when one
mechanism becomes arrested locally, the other may take over to allow cracking to progress at a rate
equal to the sum of individual rate [22]. Thus it can be inferred that interaction point can be shifted

from static CCG towards cyclic FCG. The additional shift from left to the right could be due to other
factors which increase constraint in an inherent way. Therefore material degradation and embrittlement
in ex-service condition, low stresses and reduced creep failure ductilities will all tend to increase
cracking rate. As seen in Fig. 7 the NSW model could be used to bound these effects.

5.5. Fractography

The fractographic assessment was used to confirm the experimental findings. Thus prior to breaking
open the tested samples, a 3 mm slice was extracted from the mid thickness of the sample to examine
the fracture path. Fig. 8(a-c) shows an optical microscopy of the fracture paths of samples CT-B, CT-
C1 and CT-A, respectively. The crack path in the ex-service material sample CT-B (Fig. 8(a)) is
relatively straight fronted, with a number of small branches from the main crack. In Fig. 8(b) the

cracking behaviour of CT-C1 ex-service material, the initial fatigue pre-crack is relatively straight,
however in the CFCG region the crack grows at an angle. However for the as-received specimen CT-A
(Fig. 8(c)), a large crack opening displacement is observed, which is consistent with Fig. 3(b), and the
crack shows some discontinuous branching, signifying that the crack growth is creep dominated. High

magnification images of CT-B1 (Fig. 8(a)) in region (i) and (ii) are shown in Fig. 9. Fig. 9(a) shows a
secondary crack where the crack deviates from the main crack and creep cavities near the secondary

crack and Fig. 9(b) show the main crack propagating accompanying by creep voids near the main

In order to investigate the effect of frequency on CFCG, the fracture surface of creep fatigue crack
growth region has been examined in more detail using the scanning electron microscope (SEM). Figure
10 (a) to (d) shows the CFCG region for frequencies of 0.0017 Hz, 0.0015 Hz, 0.027 Hz and 10 Hz.
Although the surfaces were oxidised the figures highlight important cracking features that confirm the
effect of frequency on the mode of cracking at elevated temperatures. A high magnification SEM

image of CFCG region in Fig. 10(a) is evidently intergranular indicating the creep dominance. The
cracking mode become more transgranular as the frequency become higher and the fracture surfaces
become more flat in appearance as see in Figures 10 (b) to (d). There was also no apparent difference
between the failure modes in the as received and ex-service P91 steels.

6. Conclusions

The creep fatigue crack growth behaviour of P91 steel in as-received and ex-service material conditions
has been examined. The crack growth data was characterized using fracture mechanics parameters ΔK

and C*. The results from short term accelerated tests showed that at high frequency (> 0.01 Hz), the
CFCG behaviour tend to that of high cycle transgranular fatigue crack growth and is best correlated
with the ΔK parameter whereas at lower frequencies, creep mechanisms have been found to dominant

as intergranular cracking and best correlated with the C* parameter. The correlation between crack
growth rate and C* parameter, shows that most of the CFCG tested at 600°C to 625°C fall on the upper
shelf of the CCG P91 scatter-band. Scanning electron macroscopy (SEM) analysis has confirmed the
influence of cyclic frequency on the fracture behaviour. Generally low dwell times and static loading

develops into intergranular cracking and high frequency in transgranular controlled featuring a flatter
fracture surface. This also confirms the fact that stress state controls creep cracking whereas fatigue is
cycle dependent and much less affected by stress state.

Within the scatter of experimental data a linear cummulative damage rule can still predict the
creep/fatigue interaction result regardless of the degradation of the steel under ex-service conditions.
This confirms the approach taken by the codes of practice at least for short term tests. However it is
clear that the mean CFCG rates for ex-service steels are faster by a factor of 4 compared to the mean
CCG data for as-received material. Unavailability of long term tests (>10,000h) at low stresses and
long dwell periods may pose additional problems under creep control due to state of stress where lower
creep ductilities and and high multiaxial stress states prevail. It is found that for low stresses low
ductility and increase in constraint under plane strain predicitons of crack growth rate data using the
NSW creep crack growth model can conservatively bound the experimental data at long terms which is
a more appropriate prediciton for component operational times.

It is clear that further detailed testing is needed to confirm the prediciton lines in Fig. 7, however the
present findings can confirm that firstly the linear cummulative damage is sufficiently accurate for
lifing assessment as long as there is appropriate low dwell cyclic tests are available for the material.
Secondly the NSW model can conveniently predict the upperbound cracking rate under creep/fatigue.


The authors would like to acknowledge the support by the Engineering and Physical Sciences Research
Council (EPSRC) through grant number EP/K021095/1.


1. R5: Assessent Procedure for High Temperature Response of Structures, in Issue 3. 2012: EDF
Energy Nuclear Generation Ltd.
2. British Standard, in Guide to methods of assessing the capability of flaws in fusion welded

structures 1999, BSI: London.
3. Choudhary, B.K. and E. Isaac Samuel, Creep behaviour of modified 9Cr–1Mo ferritic steel.

Journal of Nuclear Materials, 2011. 412(1): p. 82-89.
4. Shrestha, T., et al., Creep rupture behavior of Grade 91 steel. Materials Science and
Engineering: A, 2013. 565: p. 382-391.
5. Lu, Y.L., et al., Effects of temperature and hold time on creep-fatigue crack-growth behavior
of HAYNES® 230® alloy. Materials Science and Engineering, 2006. 429(1–2): p. 1-10.
6. Narasimhachary, S.B. and A. Saxena, Crack growth behavior of 9Cr−1Mo (P91) steel under
creep–fatigue conditions. International Journal of Fatigue, 2013. 56(0): p. 106-113.
7. Bassi, F., S. Foletti, and A.L. Conte, Creep Fatigue Crack Growth and Fracture Mechanism

at T/P91 Power Plant Steel. Materials at High Temperatures, 2015(32:3): p. 250-255.

8. Mehmanparast, A., C.M. Davies, and K.M. Nikbin, Evaluation of Testing and Analysis
Methods in ASTM E2760-10 Creep-Fatigue Crack Growth Testing Standard for a Range of
Steels. 2011.
9. Holdsworth, S.R., Creep Fatigue Interaction in Power Plant Steels. Materials at High

Temperatures, 2011(28:3): p. 197-204.

10. Granacher, J., et al., Creep fatigue crack behavior of two power plant steels. International

Journal of Pressure Vessels and Piping, 2001. 78(11–12): p. 909-920.

11. Speicher.M, A. Klenk, and Coleman.K, Creep Fatigue Interactions in P91 Steel. Proceeding

of 13th International Conference on Fracture June 16-21 in Beijing,Chine, 2013.

12. ASTM E2760-10 in Standard Test Method for Creep Fatigue Crack Growth Testing. 2010.
13. Saxena, A. and S.B. Narasimhachary, Round robin on creep fatigue crack growth testing for
verification of ASTM standard 2760-10. Materials at High Temperatures, 2014. 31(4): p. 357-

14. Kalyanasundaram, V., et al., ASTM Round-Robin on Creep Fatigue and Creep Behaviour of
P91 Steel. Jounal of ASTM International, 2011. 8(4).
15. Nikbin, K.M., D.J. Smith, and G.A. Webster, An Engineering Approach to the Prediction of

Creep Crack Growth. Journal of Engineering Materials and Technology, 1986. 108(2): p. 186-
16. Paris, P. and F. Erdogan, A Critical Analysis of Crack Propagation Laws. Journal of Fluids
Engineering, 1963. 85(4): p. 528-533.
17. Davies, C., et al., Experimental Evaluation of the J or C* Parameter for a Range of Cracked
Geometries. Journal of ASTM International, 2006. 3(4).
18. ASTM E1457-01, in Standard test method for measurement of creep crack growth rates in
metals. 2002.
19. Tan, M., et al., Comparison of creep crack initiation and growth in four steels tested in HIDA.
International Journal of Pressure Vessels and Piping, 2001. 78(11–12): p. 737-747.
20. Maleki, S., Long Term Creep Deformation and Crack Growth Predictions for Grade 91 steels
and Risk Based Methods in Their Component Life Assessment. 2015, Imperial College
21. Davies, C., et al., Analysis of Creep Crack Initiation and Growth in Different Geometries for

316H and Carbon Manganese Steels. 2006.

22. Webster, G.A., High Temperature Component Life Assessment. 1994.


Table 1

Material properties of P91 material at room and elevated temperature

Material σUTS % Tensile
Material ID Temperature (°C) σYS (MPa) E (GPa)
Condition (MPa) Elongation
A As-received 25 570 663 203 25
B Ex-service 25 490 665 233 26
C Ex-service 25 533 708 220 26
A As-received 620 340 360 127 30
B Ex-service 600 287 308 - -
C Ex-service 625 325 344 125 33

Table 2
Test loading conditions and durations
Specimen T th Max.load Initial ΔK Initial a Final ΔK Final a tf Nf

ID (°C) (s) (kN) (MPa√m) (mm) (MPa√m) (mm) (h) (cycle)
CT-A 620 600 15.0 22.5 22.5 38.4 30.5 1125 6750
CT-B1 600 600 13.0 22.6 25.0 28.9 28.7 311 1865

CT-B2 600 60 12.0 20.9 25.0 30.1 30.2 167 9105
CT-B3 600 30 12.0 20.9 25.0 30.8 30.5 133 13333
CT-B4-FCG 600 0 13.0 22.6 25.0 29.8 29.1 0.34 12316
CT-C1 625 600 7.5 20.5 20.8 31.6 27.9 408 2450
CT-C2 625 600 9.0 25.9 21.8 37.9 27.9 240 1438

(a) (b)

Fig. 1. Microstructure of P91 in a) as-received condition (P91-A) ;b) Ex- service condition (P91-B)

(a) (b)
Pmax Pmax

Pmin Pmin

Time,t Time,t

Fig. 2. Triangular waveform a) CFCG test b) FCG test


(a) T( C) f Kmax (MPa√m)

8.0 CT-A
CT-A 620 0.0017 25.02
CT-B1 600 0.0017 25.11
CT-B2 600 0.015 23.18
6.0 CT-B6
CT-B3 600 0.027 23.18
Δa (mm)

CT-C1 625 0.0017 22.82
CT-C2 625 0.0017 28.83
CT-B4 600 10 25.11



0.0 0.2 0.4 N/Nf 0.6 0.8 1.0

(b) T( C) f Kmax (MPa√m)
CT-A 620 0.0017 25.02
CT-B1 600 0.0017 25.11
CT-B2 600 0.015 23.18
CT-B6 600 0.027 23.18
LLD (mm)


3.0 CT-C1
CT-A-1 625 0.0017 22.82
CT-A-2 625 0.0017 28.83



0.0 0.2 0.4 0.6 0.8 1.0


Fig. 3. (a) Crack extension versus normalizes number of cycles; b) Load line displacement versus

normalized time

1.0E-01 T( C) f(Hz) Kmax(MPa√m)

CT-A 620 0.0017 25.02
CT-B1 600 0.0017 25.11
CT-B2 600 0.015 23.18
1.0E-02 CT-B6,th=30sT=600C,
CT-B3 600 0.027P=12kN23.18
CT-C1 625 0.0017 22.82
CT-C2 625 0.0017 28.83
da/dN (mm/cycle)

1.0E-03 ASTM
Ref [6] T=625,
625 th=600s,
0.00177.5kN - CFCG
Ref [6] T=625,
625 th=600s,
0.0017P=9kN -
Ref [8]Ali,T=625,f=0.001
625 0.0010 10.2
1.0E-04 CFCG
Ref [8]Ali,T=625,f=0.01
625 0.01 10.3
Ref [8]Ali,625
1.0 9.7
Ref [11] 600 T=600, th=60min
0.00027 -
FCG magdalena,T=600,
Ref [11] 600 0.5 f=0.5Hz-

Ref [10] T=600C,
600 f=0.05 Hz
0.05 - FCG

FCG test data,

CT-B4 600 f=10Hz,P=13kN
10 25.11

1.0E+01 1.0E+02

ΔK (MPa√m)

Fig. 4. Crack growth per cycle for P91 steel at various cyclic frequencies


Fig. 5. Correlation of CFCG, CCG data and predictive approximate NSW models for P91 material.

εf =33% CT-A
NSWA- PS (εf )
ti (h)


NSWA- PE (εf /10) Δa=0.2mm
1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01


Fig. 6. Correlation of creep crack initiation and predictive NSW model for P91 material

CFCGtest data
Test data
FCG Test data
Ali, Delta 30 data [8]

1.0E-01 Ave CFCG CFCG

Average data test data
Speicher 60mins

CCG data [20]

da/dN (mm/cycle)




High FCG

Interaction in constraint and intreaction point

shift from static CCG towards cyclic FCG

1.0E-05 1.0E-04 1.0E-03 1.0E-02 1.0E-01 1.0E+00 1.0E+01
Frequency (Hz)

Fig. 7. Frequency interaction map for creep/fatigue at a constant ΔK=25MPam and

C*=1.0x10-4 MPam/h


Fig. 8. Cracking behaviour of (a) CT-B; (b) CT-C1; (c) CT-A


Secondary crack

Main crack

i (a) ii (b)

Fig. 9. High magnification images of CT-B region (a) i and (b) ii, showing cracks and cavities near the

Fig. 10. SEM images of fracture specimen on the creep fatigue crack growth region at different
frequencies (a) 0.0017Hz (CT-B1) (b) 0.015Hz (CT-B2) (c) 0.027 Hz (CT-B3) and (d) 10Hz (CT-B-4-



 The CFCG data has been characterized using stress intensity factor range, ΔK and the
creep fracture mechanics parameter C*.
 The correlation between crack growth rate and C* parameter, shows that most CFCG
data fall within the creep crack growth (CCG) P91 scatter band data, which may
indicate that the crack growth behaviour is dominated by creep processes.
 The CFCG rate and the time for 0.2 mm creep crack growth extension, has been
compared to the NSW-MOD CCG model’s predictions, and resonable agreement has
been seen.
 A linear cummulative rule has ben used to predict the CFCG experimental result. It is
shown that under given frequency the approach well describes the CGCF behaviour.

 Scanning electron macroscopy (SEM) analysis has shown the influence of cyclic
frequency on the fracture surface