You are on page 1of 12

ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

JAN KOHOUT1

ANOMALIES IN TEMPERATURE SHIFT


OF S-N CURVES

Analyzing several families of finite-life fatigue curves of various materials obtained at


various temperatures, the following rough regularities have been observed: (i) a de-
crease in temperature leads to a shift of the curves towards higher strength values,
(ii) the slope of the curves at different temperatures is approximately constant, and
(iii) if the curves are drawn in log-log scales and absolute Kelvin temperature is used,
the same ratio of temperatures leads approximately to the same shift in fatigue strength
values. On the basis of these regularities the Basquin equation describing finite-life
fatigue curves can be generalized for various temperatures, which approximately ex-
presses inverse law between fatigue strength and the square to fourth root of absolute
temperature. In log – logT fit it is represented by straight lines with slope equal to
temperature sensitivity parameter. Deviation from these straight lines is evidence that
additional degradation mechanisms are effective besides fatigue, whose temperature
dependences differ from the mentioned temperature dependence of fatigue strength. In
high-temperature region it is most often creep depending on temperature according to
the Arrhenius equation, in low-temperature region athermal processes of plastic de-
formation can play significant role in fatigue failure.

Introduction
Fatigue tests at lower and higher temperatures are relatively complicated
and expensive. Deeper insight into the temperature dependence of fatigue
behaviour could save a part of fatigue tests, which need not be performed
in studied temperature interval so densely. On the other hand, fatigue proc-
esses based purely on temperature-dependent cyclic plastic deformation
can be observed only in a limited temperature region. At higher tempera-
tures fatigue is accompanied by creep and enhanced diffusion [3], at very
low temperatures some athermal processes of plastic deformation occur [9].
Stress-fatigue lifetime curves (S-N curves or Wöhler’s curves) are
the classical way of describing fatigue behaviour of structural materials.
A decrease in testing temperatures shifts S-N curves towards higher
1
Department of Mathematics and Physics, Military Technology Faculty, University of De-
fence, Kounicova St 65, CZ-612 00 Brno, Czech Republic. E-mail: jan.kohout@unob.cz.

97
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

fatigue strengths and vice versa, provided no phase or morphological


changes occur in the materials under investigation. In the literature, the
fatigue results obtained at various temperatures are mostly presented by
families of strain-lifetime curves (e.g. GH4049 nickel base superalloy
[1] or 15G2AFDps structural steel, [2]), and rarely by families of stress-
lifetime curves. In spite of this a sufficient number of S-N curves for
various materials are available (authentically measured or recalculated):
polycrystalline copper [6, 7], 11423.0 carbon steel [11], SAE2330 struc-
tural steel in two different states of thermal treatment [12], 15G2AFDps
structural steel [15], X22CrMoV121 creep-resisting steel [10, 14] etc.
However, no explicit (i.e. analytically expressed) temperature depend-
ence of S-N curves, with universal validity and generally accepted, has
yet been published.

Phenomenological derivation
The dependence of the applied fatigue stress amplitude a on the number
of cycles to failure Nf is usually described by the Basquin equation [17]
 a ( N f )  a N bf (1)

where a and b are material constants, positive and negative respectively.


On log-log scales, the function represents a straight line, i.e.
log  a ( N f )  log a  b log N f (2)

with slope b. Families of fatigue curves obtained at various tempera-


tures [6, 7, 10, 12, 15] show the following characteristics:
1. a decrease in temperature leads to a shift of the curves towards
higher strength values;
2. the slope of the curves at different temperatures is approximately
constant;
3. if log-log scales and absolute Kelvin temperature are used, the
same ratio of temperatures leads approximately to the same shift
in fatigue strength values.
All the above characteristics can be taken into consideration if Eq. (2) is
generalized to the form
log  a ( N f , T )  log a*  b log N f  c log T (3)

98
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

where c is also a material constant and its value is negative according to


item 1 above. The constant c, named the temperature sensitivity pa-
rameter, can be defined by the equation

 log  a ( N f ,T )
c (4)
 log T N f  const .

By considering Eq. (3), the following expression is obtained

 a ( N f , T )  a*N bf T c (5)

Constant a* differs from the constant a in Eqs. (1) and (2). If these
equations are valid for a certain reference temperature T0, then Eq. (5)
can be rewritten in the form [4, 5]
c
T 
 a ( N f , T )  a   N bf (6)
 T0 
thereby containing the original constant a.

Experimental materials and results

Decreased temperatures
For verification of Eq. (6) validity the results of push-pull (tension-
compression) strain-controlled fatigue tests performed on commercially
pure copper (99.98%) at temperatures of 83, 173, and 295 K (–190, –100,
and 22 °C) performed by Lukas and Kunz [6, 7] were used. They are pre-
sented as S-N curves (experimental points together with regression lines)
in Fig. 1. The dashed lines represent separate fits of experimental data for
each temperature using Eq. (1), whereas the full lines represent one
common fit for all temperatures using Eq. (6). In the first case the total
sum of least squares is equal to 3751, in the second case to 4094. The
increase of the sum by 9% when the number of fitted parameters de-
creased from 6 (2 parameters for each of the 3 temperatures) to only 3
shows the suitability of the proposed equation. The difference between
the full and the dashed lines is always less than 5 MPa, which represents
a very low value, far less than the scatter of experimental results (disre-

99
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

garding the two results for lives of about 2105 and 3105 at a temperature
of 83 K, the accuracy would be better still). The value and standard de-
viation of the newly introduced parameter are c = (0.4090.014).

Figure 1. Dependence of the fatigue strength of finite lifetime on cycle number


to failure for copper tested at mentioned temperatures [6, 7]. Fits to Eq. (1)
(separately for each temperature, dashed lines) with a fit
to Eq. (6) (common for all temperatures, full lines) are compared

Elevated and high temperatures


While the experimental results obtained at low temperatures were fitted
without any substantial problems, the results obtained at elevated tem-
peratures can be difficult to fit because creep can play a part within the
high temperature region. This is the case for the results obtained else-
where, i.e. from fatigue tests performed by Eifler et al. on X22CrMoV121
creep-resisting steel, a material used for turbine blades [2, 10]. Stress-
controlled uniaxial push-pull fatigue tests of heat treated samples (for

100
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

further particulars see [10]) were carried out at temperatures in the


range from 20 to 530 °C (note that the number of reversals instead of
the number of cycles is used in the original paper). The results (stress
amplitude vs. number of cycles to failure) and their fits are presented in
Fig. 2. Dashed lines indicate separate fits for each temperature using Eq.
(1). Full lines are the result of fits using Eq. (6) for the experimental
data obtained at temperatures from 20 to 400 °C. The dot-and-dash line
is the predictive curve for a temperature of 530 °C, drawn using Eq. (6).
This line lies (567) MPa above the fit by Eq. (1) (dashed line) for the
same temperature. However, the Eq. (1) and Eq. (6) fits for all the other
temperatures are very close together (see dashed lines and full lines).
This shows that a new failure process decreases the fatigue resistance,
which starts to have remarkable influence somewhere between 400 and
530 °C. It is natural to assume that it should be creep which is con-
firmed in further paper [2] concerning the same X22CrMoV121 steel.
Also in paper [13] concerning A470 class 8 steel used for steam turbine
rotors working at approx. 420 °C, i.e. at comparable conditions, creep
was unambiguously identified. The value of the temperature sensitivity
parameter for the X22CrMoV121 steel [10] is 0.3120.018.
Fig. 3 presents the same results given in Fig. 2, but the temperature
dependence of fatigue strength is plotted here for a given number of
cycles to failure (103, 3103, 104, 3104, 105, and 3105). Since experi-
mental points for the above cycles to failure are not available, values
obtained by separate fits for each temperature using Eq. (1) (see dashed
lines in Fig. 2) were used. Logarithmic scales for fatigue strength and
absolute temperature are used, with the temperature in degrees centi-
grade, too. Fig. 3 shows two important aspects: firstly, the linear de-
pendence, in a bi-logarithmic scale, of fatigue strength for temperatures
lower than 400 °C demonstrates the validity of Eq. (6) [see also Eq.
(3)]; secondly, the substantial decrease in fatigue strength well below
the straight lines at the temperature of 530 °C indicates that creep be-
gins to play a significant role at this temperature. In general terms, the
deviation from the linear dependence of fatigue strength on absolute
temperature (on the log-log scale) is undoubtedly a symptom of addi-
tional degradation process contributing to purely fatigue failure.

101
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

Figure 2. Dependence of fatigue strength of finite lifetime on cycle number to failure


for X22CrMoV121 steel at temperatures 20 to 530 °C [10]. Separate (dashed lines)
and common (full lines) fits are compared.

Figure 3. Dependence of fatigue strength of finite lifetime on testing temperature


for X22CrMoV121 steel [10], for chosen numbers of cycles to failure

102
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

Very low temperatures


McCammon and Rosenberg [8] studied fatigue behaviour of copper in
temperature range from room temperature to temperature of liquid he-
lium. Verification of the validity of Eq. (6) for this case shows that it
can be applied for temperatures 20 to 293 K (separate fits for all three
temperatures, see dashed lines in Fig. 4, and common fit using Eq. (6),
see full lines there, are very well comparable). But at 4.2 K the stress
amplitude is substantially lower than should be according to forecast
based on Eq. (6), see dot-and-dash line in Fig. 4. The difference be-
tween the actual and the forecast stress is decr = (938) MPa [rela-
tively it represents decrease from forecast 100% to actual (752)%].
This stress difference can be also expressed as temperature shift towards
higher temperatures Tincr = (10.41.2) K. The temperature sensitivity
parameter is c = (0.2310.007) in this case, which is the smallest abso-
lute value of this parameter determined up to now.

Figure 4. Dependence of fatigue strength of finite lifetime on cycle number


to failure for copper tested at temperatures 4.2 to 293 K [13].
Separate (dashed lines) and common (full lines) fits are compared

103
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

Unfortunately, fatigue tests performed in the hydrogen-helium tempera-


ture range are very rare and their results are presented only descriptively,
without major success in understanding the substance of fatigue phenom-
ena. On the other hand, temperature dependence of yield stress was studied
very deeply also at very low temperatures including the region below liquid
helium temperature. Approximately below the temperature of 20 K the
athermal processes of plastic deformation are considered, which are ex-
plained by various theories. Some of them take into account the inertial
effect on dislocation mobility, e.g. [9]. In principle, these processes can be
considered also in the case of fatigue failure because fatigue is usually pre-
sented as accumulation of plastic deformation. Assuming athermal depend-
ence of stress amplitude, the results presented in Fig. 4 in the dependence
on number of cycles to fracture can be presented in the dependence on
temperature in the way shown in Fig. 5. The points in Fig. 5 were obtained
using separate fits of results for each of test temperatures using the Basquin
function. Assumption of athermal dependence of stress amplitude at tem-
peratures near 4.2 K also allows estimating the temperature separating the
athermal range from the range of power law dependence described by Eq.
(6). According to Fig. 5 it is approximately 15 K.

Figure 5. Dependence of fatigue strength of finite lifetime on testing temperature


for copper [13], for chosen numbers of cycles to failure

104
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

Discussion
In finite life region Eq. (6) representing the Basquin function general-
ized for various temperatures seems to be very good tool for description
of fatigue stress (stress amplitude as well as upper stress of loading cy-
cle) in dependence on number of cycles to fracture and on temperature.
Its double power law dependence allows its linearization if log Nf – log
a or log T – log a fits are used. The slopes of the straight line families
are equal to parameter b or to parameter c, respectively. The values of
both b and c parameters are always negative. In all up to now studied
cases the range of c parameter is approximately –0.2 (copper at very
low temperatures [8]) to –0.6 (ferritic nodular cast iron, determined
from results of thesis [16]). Previously determined values of c parameter
very close to –0.5 led to hypothesis that  a  1 T [4, 5], but subse-
quent studies gave also values close even to –0.2. It was one of the rea-
sons, why this hypothesis has been rejected.
Standard deviations of parameter c are equal to a few per cent of its
values. This means that this parameter is a very good tool for a rela-
tively precise description of temperature dependence of fatigue curves
in finite-life region where the Basquin equation can be used. In the ex-
treme this parameter can be determined from at least two finite-life fa-
tigue curves for two different temperatures T1 and T2 and then the
curves for temperatures from the interval (T1,T2) and its near vicinity
can be predicted using Eq. (6). This equation valid for the finite-life
region can be expanded for low-cycle region down to ultimate tensile
strength Rm if the Rm(T) dependence is known, see two slightly different
ways described in [4, 5]. On the other hand, the (ultra) high-cycle region
of fatigue (Nf  107) has not been studied yet from this point of view.
Scheme containing family of practically parallel straight lines for
various numbers of cycles to fracture with common slope c, which
represents Eq. (6) in log T – log a fit, is shown in Fig. 6. This figure
shows also both limits of Eq. (6) validity: high-temperature limit Tc
connected with creep appearance and low-temperature limit Ta con-
nected probably with athermal processes of plastic deformation. Just the
application of log T – log a fit allows estimation of both temperature
limits, which are demonstrated by the deviations from straight-line
course. While the determination of Tc limit can be done only very

105
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

roughly or needs many test temperatures near this limit, the determina-
tion of Ta limit can be done more precisely on condition that plastic de-
formation processes as well as fatigue stress are really athermal below Ta
limit. Its graphical estimation 15 K for copper tested at very low tempera-
tures [8] can be compared with the sum of the lowest experimental tem-
perature 4.2 K and temperature shift Tincr = 10.4 K, i.e. 14.6 K.
Both the temperature limits of Eq. (6) validity were determined only
for different materials: the high-temperature limit for X22CrMoV121
steel and the low-temperature limit for copper. It would be very useful
to have at disposal experimental data of fatigue tests from extremely
wide temperature range including very low and sufficiently high tem-
peratures for one material. Unfortunately, no results of fatigue tests cov-
ering all three temperature regions with both temperature limits have
been found in literature and it is out of authors’ abilities to perform
them by their own means.

Figure 6. Schematic division of temperature range into athermal region, fatigue region
described by power law dependence, and region with important role of creep.

106
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

Conclusions
1. The equation suitable for describing the dependence of fatigue
strength of finite lifetime on the number of cycles to fracture and
testing temperature in the range of low as well as of elevated tem-
peratures was proposed and successfully verified.
2. At high temperatures, a deviation from this equation downwards to
lower stresses can be observed. It is evidence that another degrada-
tion process joined fatigue; in studied case it was creep.
3. Also at very low temperatures another deviation from this equation
to lower stresses appears which is probably connected with athermal
processes of plastic deformation.
4. The log T – log  fit using proposed equation consists in a family of
parallel straight lines with slope equal to the temperature sensitivity
parameter. Using this fit, the deviations from this equation can be
presented most apparently.
5. The values of temperature sensitivity parameter have been calculated
between –0.2 (copper) and –0.6 (ferritic nodular cast iron). Knowing
its value, interpolation of S-N curves in temperature and prediction
of their shift with temperature can be done.

Acknowledgements
Financial support of the Ministry of Defence of the Czech Republic
within research project MO0FVT0000404 is gratefully acknowledged.
Moreover, the author is grateful to all the authors (see [6, 7, 8, 10]),
whose original experimental results were used in presented considera-
tions and calculations.

REFERENCE
[1] CHEN, L. J.–WANG, Z. G.–YAO, G.–TIAN, J. F.: Int. J. Fatigue. Vol.
21, No. 8, 1999, 791–797. o.
[2] EIFEL, D.–RÖTTGER, D.: Proceedings of the 7th international fatigue
congress Fatigue ’99. Edited by X. R. WU and Z. G. WANG, High
Education Press, Beijing, China, 1999, Vol. IV, 2145–2151. o.
[3] KLESNIL, M.–LUKÁŠ, P.: Fatigue of metallic materials. Academia,
Prague, Czech Republic, 1992.

107
ANOMALIES IN TEMPERATURE SHIFT OF S-N CURVES

[4] KOHOUT, J.: Fatigue & Fracture of Engineering Materials & Struc-
tures, Vol. 23, No. 12, 2000, 969–977. o.
[5] KOHOUT, J.: Phenomenological description of experimental depend-
ences in materials fatigue and their regression (habilitation thesis).
Military Academy, Brno, Czech Republic, 2004 (in Czech).
[6] LUKAS, P.–KUNZ, L.: In: Basic Mechanisms in Fatigue of Metals.
Edited by P. LUKAS and J. POLAK, Academia, Prague, Czech Re-
public, 1988, 161168. o.
[7] LUKAS, P.–KUNZ, L.: Mat. Sci. Eng., Vol. A103, 1988, 233–239. o.
[8] MCCAMMON, R. D.–ROSENBERG, K. M.: Proceedings of the Royal
Society. Vol. A242, 1957, 203-211. o.
[9] MOSKALENKO, V. A., NATSIK, V. D.–KOVALEVA, V. N.: Mat. Sci.
Eng., Vol. A309-310, 2001, 173–177. o.
[10] PETERMEIER, P.–EIFLER, D.: Proceedings of the 6th international
fatigue congress Fatigue ’96. Edited by G. LÜTJERING and H.
NOWACK, Elsevier, Holland, 1996, Vol. II, 753–758. o.
[11] POLAK, J.–KLESNIL, M.: Mat. Sci. Eng., Vol. 26, 1976, 157–166. o.
[12] PUSKAR, A.–GOLOVIN, S. A.: Fatigue in materials: cumulative
damage processes, Elsevier, Amsterdam, Holland, 1985.
[13] SAXENA, A.: Proceedings of the 7th international fatigue congress
Fatigue ’99. Edited by X. R. WU and Z. G. WANG, High Educa-
tion Press, Beijing, China, 1999, Vol. IV, 2085–2092. o.
[14] TAUSCHER, H.–FLEISCHNER, H.: IfL Mitteilungen, Vol. 7, No 7,
1968, 247–259. o. (in German).
[15] TROSHCHENKO, V. T.–STRIZHALO, V. A.–RUBEL, A. P.: Problemy
prochnosti. 1975, No 1, 37. o. (in Russian).
[16] VECHET, S., Behaviour of ductile cast iron in conditions of fatigue
loading (PhD thesis), Brno University of Technology, Czech Re-
public, 1989 (in Czech).
[17] WEIBULL, W.: Fatigue testing and analysis of results. Pergamon
Press, Oxford, England, 1961.

108

You might also like