International

Journalof
International Journal of Fatigue 29 (2007) 81–94
Fatigue
www.elsevier.com/locate/ijfatigue

Analysis of the fatigue strength of a stainless steel based

on the energy dissipation
G. Meneghetti *

Department of Mechanical Engineering, University of Padova, Via Venezia, I-35131 Padova, Italy

Received 6 June 2005; received in revised form 18 January 2006; accepted 13 February 2006
Available online 19 April 2006

Abstract

Determination of fatigue limit under uniaxial tests based on the experimental measurement of material thermal increments (typically
by means of infrared cameras) is well documented in the literature. Anyway the energy dissipated in a unit volume of material as heat
seems to be a more promising parameter for fatigue characterisation rather than the surface temperature. In fact for a given material,
loading and mechanical boundary conditions the former parameter depends only on the applied stress amplitude and load ratio in a
constant amplitude fatigue test, while the latter depends also on the specimen geometry, test frequency and the thermal boundary con-
ditions that determine the rate of heat transfer from the material to the surroundings. Then it is expected that the fatigue strength of both
smooth and notched specimens can be rationalised in terms of the thermal energy dissipated in a unit volume of material per cycle. The
first aim of this paper is to define a theoretical model in order to derive the specific heat loss per cycle from temperature measurements
performed during the fatigue test. The model has been applied to analyse the fatigue strength of smooth and notched specimens made of
AISI 304 L stainless steel. Then, it has been verified to which extent the proposed approach holds true while varying the notch tip radius.
Finally, it has been analysed the material response in terms of energy released as heat in two-level fatigue tests.
 2006 Elsevier Ltd. All rights reserved.

Keywords: Notch effect; Thermography; Energy methods

1. Introduction increase at the beginning of the test and then it tends to a

stationary value. The higher the applied stress amplitude
The thermographic approach able to estimate the fati- ra, the higher the stationary temperature increase DTstat,
gue limit of materials under uniaxial tests [1–4] and compo- as schematically depicted in Fig. 1a, where the observed
nents [5,6] is an experimental procedure well documented temperature increase DT is plotted versus the number of
in the literature. It is based on the experimental observa- cycles. Indeed temperature oscillates around the ‘‘station-
tion that a material undergoing a fatigue test is subjected ary’’ value with the same frequency of the load cycles due
to an increase of the surface temperature, such that the to the thermoelastic effect [9], which is superimposed to
higher the applied stress amplitude the higher the tempera- the mean temperature evolution shown in Fig. 1a. Anyway
ture increase [7,8]. In the published papers on this subject, if the adopted load frequency is sufficiently high, then the
temperatures are in most cases measured by means of infra- amplitude of temperature oscillations due to thermoelastic-
red cameras, being a non-contact, relatively simple to be ity are negligible with respect to the stationary temperature
applied technique, able to measure the temperature field increase, at least for the material considered in this study,
on the specimen surface. In summary, surface temperature as it will be seen later.
of the material undergoing a fatigue test is seen to rapidly According to what proposed in [1], the fatigue limit of
the material estimated by means of the thermographic tech-
*
Tel.: +39 049 8276751; fax: +39 049 8276785. nique r0,th can be determined by extrapolating down to
E-mail address: [email protected]. zero the best fitting curve that correlates the applied stress

0142-1123/$ - see front matter  2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijfatigue.2006.02.043
82 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

Nomenclature

c material specific heat [J/(kg K)] TQ energy ratio defined as QA,10%/QA,90%

Ep,f critical energy per unit volume at fracture [J/m3] Tr stress ratio defined as DrA,10%/DrA,90%
f load test frequency DT temperature increment
H power released as heat by the material in a unit DTstat stationary value of the temperature increment
volume [W/m3] Dr stress range (maximum value tos minimum value)
k inverse slope of the Woehler curve Drn stress range referred to the net section
Ktn elastic stress concentration factor referred to the R stress ratio (ratio between the minimum and the
net section maximum applied nominal stress)
Nf number of cycles to failure DrA,50%, DrA,10%, DrA,90% stress range at a reference
Q energy released as heat by the material in a unit number of cycles equal to 2 · 106 with a survival
volume per cycle [J/(m3 cycle)] probability of 50%, 10%, 90%, respectively.
QA,50% characteristic value of Q at a reference number q material density [kg/m3]
of cycles equal to 2 · 106 with a survival proba- r0 fatigue limit in terms of stress amplitude
bility of 50% r0,th fatigue limit estimated by means of the thermo-
T temperature metric method
t time ra stress amplitude
T* temperature at a given point of the specimen at rA,50% stress amplitude at a reference number of cycles
time t* equal to 2 · 106 with a survival probability of
t* time at which the fatigue test is stopped 50%

amplitudes ra and the stationary temperature increments The relation existing between the thermal evidence and
DTstat, as depicted in Fig. 1b. Since it is not necessary to the physical phenomenon of the fatigue limit of low carbon
run the fatigue test up to failure in order to achieve the sta- steels has recently been deepened [11]. It has been found
tionary thermal rise, the thermographic technique was pre- that in a fatigue test where the stress amplitude is increased
sented as a rapid tool able to estimate the fatigue limit. An step by step, the slope of the curve that fits the experimen-
application of such an experimental method will be given in tal data DTstat versus ra starts to change at the stress level
the next section. According to a slightly different procedure that generates the first Persistent Slip Band (i.e., at the
[5,10], the fatigue limit is to be determined from the inter- onset of localised microplasticity). Indeed it is well known
section of the DTstat versus ra lines obtained by separately [12–14] that the fatigue limit of steels corresponds to a
fitting the experimental data for low and high stress ampli- slightly higher stress level, at which an initiated micro-
tudes, respectively (that is for stress levels below and above crack stops propagating at the first microstructural barrier.
the expected fatigue limit). In fact it has been observed that Nonetheless the thermometric methods proved to be able
the slope of such lines is lower in the former case and stee- to estimate the fatigue limit in fairly agreement with that
per in the latter case so that the stress amplitude at the knee determined by means of the traditional stair-case
point can be quite easily defined. These thermometric procedure.
approaches are of phenomenological type and have been As pointed out by Luong [15], a clear distinction has to
validated by means of a considerable number of experi- be made between the physical nature of fatigue damage,
mental results. consisting in crack nucleation followed by micro/macro

ΔΤ ΔΤstat
σa3 σa1< σa2< σa3

σa2
ΔΤstat
σ 0,th
σa1
b
a

N σa

Fig. 1. Observed temperature evolution during constant amplitude fatigue tests on metallic materials (a), estimation of the material fatigue limit according
to [1] (b).
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 83

crack propagation and the phenomenon of heat dissipa- etries and test conditions different from the reference ones.
tion, which is not by nature an indicator of fatigue damage. Differently, the amount of energy dissipated in a unit vol-
In fact temperature rise can be observed in a specimen ume of material per cycle, which the temperature distribu-
cyclically loaded below the fatigue limit, due to non-dam- tion depends on, is a more promising parameter for fatigue
aging anelastic dissipative phenomena [16,17]. The strength analyses of engineering materials, because it is
approach proposed in the present paper is a contribution expected to be constant with respect to the test frequency
to the thermometric methods for an engineering analysis (at least within the range usual for standard servo-hydrau-
of the fatigue strength and is based on the energy released lic testing machines), the specimen geometry and the ther-
by the material in a unit volume per cycle (specific energy) mal boundary conditions.
estimated from experimental measurements of the surface There are a number of approaches in fatigue analysis of
temperature. This task will be accomplished by means of materials and components which are based on the experi-
the energy balance equation applied to a control volume mental measurement of the energy, each of them assuming
of material undergoing a fatigue test. a particular form of energy as a fatigue damage parameter.
The reason why specific energy is a more promising In fact, as it will be seen later on, the energy balance
parameter rather than temperature, is that the latter cannot involves three forms of energy, i.e., the mechanical energy
be adopted for specimens geometries and testing conditions expended to run the fatigue test, the energy converted into
different from the reference ones. In fact consider for exam- heat and the internal energy stored within the material.
ple a plain specimen and a notched specimen of the same The mechanical energy, i.e., the non-recoverable plastic
material undergoing a fatigue test with the same load ratio. work measured by the area of hysteresis loops, was postu-
Fig. 2 reports the typical temperature fields recorded dur- lated to be an index correlated to the accumulation of fati-
ing the fatigue test by an infrared camera. In the case of gue damage by Feltner and Morrow [18]. Halford [16]
the plain specimen the maximum temperature will occur summarised about 1400 fatigue test results for a wide vari-
at the specimen central point because the specimen ety of metals and alloys by plotting plastic hysteresis energy
behaves, as a first approximation, like a bar subjected to versus number of cycles to failure and showed that at lives
a uniform heat generation along the reduced section length on the order of 5 · 105 cycles to failure one hundred times
and heat dissipation by conduction at the specimen ends: more plastic strain hysteresis energy is required for fatigue
thus temperature distribution is parabolic along the speci- failure than for fracture in a monotonic test. The expended
men longitudinal axis. If temperature were an indicator mechanical energy was assumed also by Charkaluk et al.
of fatigue damage, then fracture should occur at the spec- [19] as a damage indicator in the context of fatigue design
imen mid-section, which is not true in general. Concerning of components under thermomechanical loadings. It is
the notched specimen shown in Fig. 2, points where tem- worth noting that the maximum value of the expended
perature is maximum and those where stresses are the high- energy per cycle was correlated to the number of cycles
est coincide. Even if we suppose that the fatigue lives of the to crack initiation. Recently, while dealing with crack prop-
plain and the bluntly notched specimens are the same, tem- agation in ductile solids, Klingbeil [20] proposed a theory
peratures at the critical points will be different in general where the fatigue crack growth rate is given as a function
because temperature distribution will depend on applied of two parameters, i.e., the plain stress or plain strain frac-
stress amplitude, test frequency, specimen geometry and ture toughness and the total plastic dissipation per cycle
all the boundary conditions that define the heat transfer which occurs inside the reversed plastic zone ahead of the
rate from the material to the surroundings. Thus tempera- crack.
ture is a parameter that cannot be easily extended to geom- Kaleta et al. [21] assumed the internal energy stored
within a material (known as the stored energy of cold
work) to be responsible for fatigue failure and by means
of accurate experimental tests measured the stored internal
energy as the difference between the expended mechanical
energy (i.e., the area of the hysteresis loop) and the energy
converted into heat. A special technique was adopted in
order to measure the expended mechanical energy, due to
the difficulties in measuring the hysteresis loops in high
cycle fatigue. Concerning the determination of heat dissi-
pation, a current-heated specimen was used in order to find
the relation between dissipated heat power and tempera-
ture distribution at the specimen surface [17]. Risitano
et al. proposed a parameter based on the measured temper-
ature, which was related to the stored energy after fracture
Fig. 2. A plain specimen and a smoothly notched specimen of the same and was assumed to be a material constant [22].
material: even if fatigue lives are the same, temperature distributions may A third possible approach assumes as a fatigue damage
be different. index that part of the total expended mechanical energy
84 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

which is converted into heat and then dissipated to the sur- conduced under load control, by adopting a test frequency
roundings. The present work deals with such a form of between 10 and 16 Hz and a load ratio R, defined as the
energy, since it is easier to be measured than both the ratio between the minimum and the maximum applied
stored internal energy and the mechanical expended nominal stress, equal to 0.1. The adopted infrared camera
energy: in fact most of the total mechanical expended was an AGEMA THV 900 LW/ST able to detect infrared
energy is converted into heat, and this is particularly true radiation in the range of wave lengths between 8 and 12 lm
in high cycle fatigue [16]. Heat dissipation was assumed with a resolution of 0.1 C. Temperatures reported in Fig. 3
as a fatigue damage parameter for example by Gamsted were measured at the specimen surface located at the mid-
et al. [23] in studying fatigue failure of composites. section between the machine grips and represent just the
In the next section the thermal phenomenon observed mean evolution of the phenomenon, since thermal images
during constant amplitude fatigue tests on smooth speci- were acquired by using a frequency much lower than that
mens in AISI 304 L steel will be presented and a practical of the load cycle. In high cycle fatigue of the tested material
application of the thermometric method will be shown. temperature tends to stabilize after about one third of the
After that, aims of the present paper are as follows: total fatigue life, while in medium and low cycle fatigue
temperature rapidly increases at the beginning of the fati-
• to develop a theoretical model able to estimate the gue test and then stabilises after about one half of the total
energy Q dissipated as heat in a unit volume of material number of cycles to failure. The stationary temperature
per cycle (specific thermal energy loss) starting from increment DTstat is also quoted in Fig. 3.
temperature data acquired during a fatigue test; Indeed the thermoelastic effect is superimposed to the
• to summarise the fatigue strength of notched specimens mean evolution of the phenomenon. As an example
of different geometries in terms of specific thermal Fig. 4 reports a more detailed acquisition of the tempera-
energy loss; ture increase during the initial phase of a fatigue test with
• to investigate the limits of applicability of the proposed
approach with respect to a reduction of the notch tip T [˚C]
31.90
radius;
• to investigate if the specific energy Q is load history
Test stop
independent.
31.70

2. Constant amplitude fatigue tests with thermal monitoring

31.50
Fig. 3 reports as an example the absolute temperature T
recorded by means of an infrared camera during constant
amplitude axial fatigue tests on 2-mm thick bluntly 31.30
notched specimens in AISI 304 stainless steel [24], charac-
terised by a stress concentration factor of 1.5. Tests were
31.10
0 5 10 15 20 t [s] 30
T [˚C]
80
T [˚C]
31.90
70
250 MPa
31.80
60 230 MPa

200 MPa
50 180 MPa 31.70
σa=170 MPa Final fracture

40 σa=150 MPa 31.60

σa=130 MPa, run out ΔΤstat

30
Room
31.50
temperature
20
0 2E5 4E5 6E5 8E5 31.40
N. cycles 15 16 17 18 t [s] 20

Fig. 3. Temperature vs number of cycles recorded during constant Fig. 4. Thermoelastic effect superimposed to the mean temperature
amplitude fatigue tests on smooth specimens in AISI 304 stainless steel evolution at the initial stage of a constant amplitude fatigue test with
[24] (load ratio R = 0.1). ra = 150 MPa. Measurements performed by means of thermocouples.
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 85

a stress amplitude equal to 150 MPa and a load frequency σ a [MPa]

of 2 Hz, where the thermoelastic effect makes the tempera- 400
ture to oscillate around the mean level with a range of R σA, 50% k
PS 10% 0.1 104 5.27
about 0.15 C. In order to be able to estimate such a ther- 300
moelastic effect, copper–constantan thermocouples (T-
type) having a resolution of 0.01 C were adopted together
200
with an acquisition unit HP 34970A operating at a sample
PS 90%
frequency of 22 Hz. As Figs. 3 and 4 show and also by
analysing the temperature data acquisitions for the other 150
stress amplitudes, it could be concluded that for this mate- σ0,th=128 MPa

rial and testing conditions temperature variations during

100
one load cycle due to the thermoelastic effect are at least
one order of magnitude lower than the stationary temper-
ature increase DTstat due to the mean evolution of the 1E4 1E5 1E6
phenomenon. N. cycles
Fig. 3 highlights that for medium and high stress levels, Fig. 6. Comparison between the Woehler curve and the fatigue limit r0,th
temperature at the beginning of the test increases beyond estimated by means of the thermographic technique.
the subsequent stationary value, while for low stress levels
temperature monotonically increases and then achieves the between the trend of the Woehler curve and the estimated
stationary value. This phenomenon may be associated with fatigue limit r0,th. As aforementioned, since it is not neces-
the microstructure changes leading to material hardening, sary to run each fatigue test up to failure in order to achieve
as it was already observed for other construction steels [25]. the stationary thermal rise, as illustrated by the previous
Since stationary temperature increments DTstat are seen Fig. 3, the thermographic technique was presented as a
to be of only few degrees for stress amplitudes close to the rapid tool able to estimate the fatigue limit [1].
fatigue limit and on the order of tens of degrees in the med- As a final remark, the author of the present work could
ium cycle regime, it was proposed [1] that the fatigue limit of verify [26] that the thermometric method able to estimate
the material can be estimated by extrapolation down to zero the fatigue limit is not suitable for all type of metals and
of the experimental curve that correlates the applied stress available testing facilities. In fact difficulties were fre-
amplitudes ra and the stationary temperature increments quently encountered, due to the observed very small tem-
DTstat. Fig. 5 shows the result of the extrapolation proce- perature increments, when (i) testing high strength steel
dure, while Fig. 6 compares the experimental Woehler curve or metals having high thermal conductivity, (ii) performing
with the fatigue limit r0,th estimated by means of the ther- rotating bending fatigue tests where a small volume of
mographic technique [24]. Even if the fatigue limit was material is subjected to the maximum applied stress, (iii)
not experimentally determined by means of the classic adopting too low load frequencies, (iv) using severely
stair-case procedure, Fig. 6 highlights the fairly agreement notched specimens.

3. Theoretical model based on the energy balance

ΔΤstat [˚ C]
50
In order to derive the energy Q released by the material
in a unit volume per cycle, a theoretical model was defined.
Let us consider a material undergoing a fatigue test and
40
consider a control volume V with the outward normal n, as
depicted in Fig. 7. The external surface S of the control vol-
ume V can be divided into three parts, namely Scv, Scd and
30
Sir through which the heat Q is transferred to the surround-
ings by convection, conduction and radiation, respectively.
Moreover k is the thermal conductivity of the material, a is
20
the heat transfer coefficient by convection, j is the surface
σ0,th=128 MP a emissivity, q is the density, c is the specific heat, T1 is the
room temperature, f is the load test frequency and T(x,y,z;t)
10
Load ratio R=0.1 is the time-dependent temperature field. The statement of
Test frequency 10÷16 Hz energy balance (the first law of thermodynamics) is given by

0 input mechanical energy ðW Þ

120 140 160 180 200 220 σ a [MPa] 260
¼ dissipated thermal energy ðQÞ
Fig. 5. Estimation of the fatigue limit r0,th by means of the thermographic
method [24]. þ variation of the internal energy ðDU Þ
86 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

Indeed this hypothesis does not hold true any longer when
cracks have initiated at the specimen surface. Anyway, as it
will be seen later, the energy dissipated in a unit volume of
material will be estimated at a number of cycle between one
third and one half the total fatigue life of plain and
smoothly notched specimens, when macro-cracks are not
supposed to be initiated. Moreover temperature increments
are supposed to be limited during the fatigue test, so that
thermo-physical properties do not vary with time.
Fig. 7. Energy balance for a material subject to fatigue loadings.
4. Estimation of Q in the one dimensional case
and can be written in terms of power as follows:
Z I  Z If we consider a smooth specimen the energy balance
rij  deij  f  dV ¼ k  grad~ T ~n  dS cd equation can be applied to the reduced section length, as
V S cd
Z shown in Fig. 8, where the problem can be assumed as
þ a  ðT  T 1 Þ  dS cv one-dimensional, i.e., the temperature variation in the spec-
ZS cv imen cross section can be neglected, at least as far as the
þ j  rn  ðT 4  T 41 Þ  dS ir presence of any discontinuity inside the material (such as
S ir defects, inclusions, pores), which could disrupt temperature
Z  
oT distribution, is neglected. In fact the heat transfer capacity
þ qc _
þ Ep  dV ;
ot by convection is much lower than the heat transfer capacity
V
by conduction in metallic material usually adopted in engi-
ð1Þ
neering structures and then temperature inside the material
where rn is the Stephan–Boltzmann constant equal to can be considered to be uniform in the cross section. This
5.67 · 108 W/(m2 K4), grad~ T is the gradient of the tem- fact is demonstrated by the very low Biot numbers (much
perature field. The rate of variation of internal energy lower than 1) that can be calculated for specimens com-
has two terms: the first one depends on the temperature monly used in fatigue tests. As an example, take a specimen
variations of the material in the volume V, while the second with a 10-mm-thick rectangular cross section fabricated in
one is the rate of variation of the so-called stored energy of cast iron having k = 20 W/(m K). A reasonable heat trans-
cold work, which is that part of the mechanical input en- fer coefficient under the hypothesis of natural convection
ergy responsible for changes in material microstructure along the specimen surface is of the order of a = 10 W/
(re-arrangement of crystal imperfections, persistent slip (m2 K). Then we have [27]:
band formation, etc.) leading to the initiation of fatigue mi- as
cro-cracks [21]. By applying Green’s theorem, the surface Bi ¼ ¼ 7  104
k
integral of the conduction term can be transformed into a
volume integral: being s the specimen thickness.
Z Z The elementary volume and its lateral surface are
P given
k  gradT  n  dS cd ¼ k  divðgrad~
~ ~ TÞ  dV ; by: dV = A Æ dz and dS = dScv = dSir = p Æ dz = i li Æ dz,
S cd V

where
o2 T o2 T o2 T
divðgrad~TÞ ¼ r2 T ! r2 T ¼ 2 þ 2 þ 2
ox oy oz
the energy balance equation in integral form can be written
as
Z I  Z
rij  deij  f  dV ¼ k  r2 T  dV
V VZ

þ a  ðT  T 1 Þ  dS cv
ZS cv
þ j  rn  ðT 4  T 41 Þ  dS ir
S ir
Z  
oT
þ qc þ E_ p  dV .
V ot
ð2Þ
In this section the material is treated as an homogeneous
continuum so that all thermo-physical properties are Fig. 8. A plain specimen with a rectangular cross section having area A
thought of as constants with respect to position (x,y,z). and perimeter p.
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 87

respectively, where A is the cross section area, p is the not be verified in high cycle fatigue since the measurement
perimeter of the cross section and li is the length of the of the expended mechanical energy is a critical issue and, in
ith side of the section. second place, most of the expended mechanical energy is
The energy balance Eq. (2) applied to the elementary converted into heat, so that estimations of Ep,f as difference
volume dV can be written as follows: between experimentally measured values of Wf and Q Æ Nf
I  are affected by great uncertainties. Conversely, in low cycle
rij  deij  f  A  dz ¼ k  r2 T  A  dz fatigue Wf is of the same order of Ep,f. In a previous work
[28], Ep,f was estimated as difference between Wf and Q Æ Nf
þ a  ðT  T 1 Þ  p  dz and the results were in fairly agreement with Feltner and
X
þ ðji  ‘i Þ  rn  ðT 4  T 41 Þ  dz Morrow’s hypothesis. As aforementioned, accurate experi-
i mental measurements of the energy stored in a specimen
 
oT based on an electrical analogy have already been per-
þ qc _
þ Ep  A  dz. ð3Þ
ot formed [21], but in the present work Q was assumed as a
fatigue damage parameter because temperature is a more
Or in differential form as simple parameter to be measured at the critical point of a
I 
p specimen or a component, provided that the unfavourable
rij  deij  f ¼ k  r2 T þ a  ðT  T 1 Þ  circumstances recalled at the end of section two are not
A
X rn met.
þ ðji  ‘i Þ   ðT 4  T 41 Þ The parameter Q (=H/f) can be derived by means of Eq.
i
A
  (6) from experimental measurements of the surface temper-
oT _ ature along the specimen z-axis (see Fig. 8) and of the room
þ qc þ Ep . ð4Þ
ot temperature, provided that the heat transfer coefficient a
and the surface emissivities ji are known. Eq. (6) enables
Since we do not consider the thermoelastic effect, which
one to calculate also the different contributions of conduc-
is not responsible for energy dissipation in one load cycle,
tion, convection and radiation to the total energy dissipa-
we can apply Eq. (4) after the mean temperature evolution
tion: it has been found [24] that in standard axial fatigue
has achieved stabilisation. Under such steady-state condi-
tests using specimens having rectangular cross section
tions the time derivative of temperature T is null. Then,
and made of stainless steel or nodular cast iron the three
according to Eq. (4), the amount of expended mechanical
contributions are all of the same order of magnitude and
energy which is not stored inside the material is totally con-
that conduction is the prevailing mechanism for heat
verted into heat and dissipated to the surroundings. Under
transfer.
these conditions we can write:
I  The major drawbacks when using Eq. (6) are the need
for knowledge of the heat transfer coefficient a and the sur-
rij  deij  f ¼ H þ E_ p ð5Þ
face emissivities ji. Moreover Eq. (6) is limited to only one
dimensional problems, which rules out the case of notches,
being H equal to
where temperature distribution near to the notch tip is at
p X rn least two dimensional. In the next paragraph a more gen-
H ¼ k  r2 T þ a  ðT  T 1 Þ  þ ðji  ‘i Þ   ðT 4  T 41 Þ
A i
A eral approach and a simpler technique is presented in order
ð6Þ to experimentally evaluate Q.

and represents the thermal power dissipated in a unit vol- 5. Estimation of Q in the general case
ume of material due to conduction, convection and radia-
tion, respectively. The energy dissipated as heat can be evaluated by exper-
By supposing that a significant part of the fatigue test is imentally measuring the cooling rate of the material after a
under steady-state conditions, as demonstrated by the sudden interruption of the fatigue test. Recalling Eq. (4),
experimental results (see Fig. 3 and Ref. [1]), Eq. (5) can under steady-state conditions it can be written:
be easily integrated over the time length [0,tf] of the fatigue I 
test. Recalling that f Æ dt = dN, it can be written: rij  deij  f ¼ ðH cd þ H cv þ H ir Þ þ E_ p ; ð8Þ
W f ¼ Q  N f þ Ep;f ; ð7Þ
where the terms H represent the thermal power dissipated
where Wf is the total mechanical energy expended per unit in a unit volume of material due to conduction, convection
volume of material until failure, Q = H/f is the heat and radiation, respectively, according to Eq. (6).
amount dissipated in a unit volume of material per cycle Suppose now to suddenly stop the fatigue test at the
and Ep,f is the critical energy per unit volume, which has time t* when the surface temperature at the considered
been supposed by Feltner and Morrow to be a material point has reached the stationary value T* = Tstat, as shown
constant [18] and equal to the energy expended in a unit in Fig. 9: the mechanical input power and the rate of accu-
volume of material after a static test. This hypothesis could mulation of plastic strain in Eq. (8) will drop to zero, so
88 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

T Table 1
Mechanical properties of the AISI304L steel

T* Tensile Yield Elastic Elongation

strength strength modulus after fracture
∂T rR [MPa] rp0,2 [MPa] [MPa] [%]
∂t t = ( t*)+ 650 300 200500 60

having a diameter equal to 20 mm centred in a rectangular

2-mm-thick plate, as reported in Fig. 10. An additional
notched series was tested at R = 0.1 having a hole diameter
t* t equal to 10 mm. Fatigue tests were conduced on a Schenck
Hydropuls PSA 100 axial servo-hydraulic machine
Fig. 9. Determination of Q by experimental measurement of the cooling
rate at t = (t*)+.
equipped with a 100 kN load cell. The adopted test fre-
quencies varied between 5 and 20 Hz. Fatigue tests were
that just after t* (that is at t = (t*)+) Eq. (4) can be written run until specimen complete separation. Strictly speaking,
as the method proposed here is concerned with fatigue life
 up to crack initiation. Anyway it is believed that for this
oT 
q  c   ¼ ðH cd þ H cv þ H ir Þ; ð9Þ kind of geometry almost all the fatigue life is expended
ot  þ
t¼ðt Þ for initiating a crack, similarly to what found in [29], where
where the thermal power H dissipated per unit volume of for bluntly notched specimens fatigue life up to crack initi-
material does not change just before and just after t*, be- ation was determined and resulted about 95% of the total
cause it depends on the temperature distribution and on life evaluated at specimen separation. Surface temperatures
the room temperature, which do not change through the at the central point of smooth specimens were monitored
time t*. Thus the thermal power dissipated per unit volume by means of the infrared camera AGEMA THV 900
of material at a given point of a specimen (i.e., LW/ST. At the beginning of the test, temperature evolu-
H = Hcd + Hcv + Hir) can be estimated by means of the tion was monitored and, after temperature stabilisation
experimental measurement of the cooling rate at that point was achieved, the test was suddenly stopped in order to
just after a sudden interruption of the fatigue test according measure the cooling rate of the material, according to the
to Eq. (9). After that, the thermal energy dissipated per sketch reported in Fig. 9. The maximum sampling fre-
unit volume and per cycle Q can be simply derived by quency of the thermal images allowed by the available mea-
knowing the test frequency f: suring system was 7 Hz. After processing the recorded
images by means of the IRWIN Research 2.01 software
H
Q¼ . ð10Þ of Agema Infrared System, temperatures versus time data
f were derived. At least five cooling curves were measured
Similarly to the mean temperature evolution, it has been for each specimen during the fatigue test and then the spe-
verified that, for the material analysed in the present work, cific heat loss Q was calculated from each cooling curve by
the energy dissipation Q rapidly increases at the beginning means of Eqs. (9) and (10). Finally the mean value was con-
of the test and then stabilizes between one third and one sidered. For some specimens many more cooling rates were
half the total fatigue life. measured during the whole fatigue test in order to investi-
In the next sections, estimations of Q by means of Eqs. gate the evolution of the specific heat loss Q. It could be
(6) and (9),(10) will be compared. After that, only Eqs. verified that during the phase when temperature is stabi-
(9),(10) will be considered, being the related experimental lised at T = Tstat, the measured values of Q for each spec-
technique and calculations much simpler to be performed. imen were within ±15% with respect to the mean value.
It will be shown that the parameter Q can be used in order The stationary temperature increments DTstat were on the
to rationalise the fatigue behaviour of smooth and notched order of some tens of degrees for the smooth specimens
specimens in AISI 304 L stainless steel. and then were compatible with the resolution of 0.1 C of
the available infrared camera. On the contrary, much lower
6. Material and methods temperature increments were measured for notched speci-
mens, that resulted in the worst case on the order of tenths
Fatigue tests with a nominal stress ratio R (ratio of degree. Then in this case temperatures were measured by
between the minimum and the maximum applied nominal means of copper–cosntantan thermocouples, fixed at the
stress) equal to 0.5 and 0.1 on specimens in AISI 304 L notch tip by means of an epoxy glue, having 0.127-mm-
stainless steel were performed. Material properties were diameter wires and a resolution of 0.01 C. The data logger
experimentally evaluated and are reported in Table 1. HP34970A operating at a 22 Hz sampling rate was used in
Two test series were prepared for each load ratio: one order to measure the cooling curve of the material after test
smooth and one notched, the latter consisting of a hole stopping. The material density q, the specific heat c and the
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 89

50 50

45

100

Geometry for R=0.1

38 44 200 200

20

45

Thickness = 2 mm

50
50

45

100 140 Geometry for R=-0.5

Tabs
20
45

Fig. 10. Geometry of the smooth and notched specimens used for fatigue tests at R = 0.1 and 0.5.

thermal conductivity k were taken from the literature [30] Fig. 12a and b report as an example the typical temper-
and resulted equal to 7900 kg/m3, 500 J/(kg K) and ature versus time acquisitions for R = 0.5 and 1, respec-
16 W/(m K), respectively. tively, after the test has been suddenly stopped. For each
curve specimen’s geometry, the nominal stress range
7. Constant amplitude fatigue test results adopted in the fatigue test, the load frequency, and the
number of cycle to failure are detailed. Fig. 12 highlights
The constant amplitude fatigue test results are reported that, differently from the notched specimens case, for
in Fig. 11a and b in terms of net section stress ranges. The smooth specimens the cooling curve is practically linear
elastic stress concentration factor Ktn (referred to the net within many seconds after stopping the fatigue tests at
section) for the plate having the 20 mm diameter hole t = t*: this fact made the calculation of cooling rates easier
was 2.24. The ratio between the fatigue strength of smooth with respect to the notched specimens, for which the time
and notched specimens at the reference number of cycle of derivative of the temperature had to be calculated within
2 · 106 resulted equal to 1.38 for R = 0.5 and 1.51 for a time window of about one second, or less, after t*. Some-
R = 0.1. Then the fatigue strength cannot be rationalised times a problem arose in recognising t* from the tempera-
either in terms of net section stresses or in terms of elastic ture versus time curve, reported as an example in the
peak stresses. previous Fig. 4a. In fact the available control software
90 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

n (MPa) did not enabled us to impose the static load value that the
600 R Δ A,50% k machine had to maintain after the test was interrupted.
-0.5 363 13.4 Then the machine could stop either during the loading part
-0.5 262 10.6
500 or the unloading part of the load cycle. If test interruption
T = 1.2
occurred during the loading phase, then the material was
400 getting cooler before t*, due to the thermoelastic effect.
The same of course happened after t*, due to heat dissipa-
tion from the material to the surroundings. Then the time
300
t* was sometimes difficult to be distinguished from the
smooth specimens analysis of the measured temperature versus time curve.
hole =20 mm Nevertheless it was noted that the cooling curve was stee-
200 per before t* than after t* and this fact enabled us to single
a 104 105 106 N.cycles out the value of t*. Anyway the best way to conduce the
fatigue test would be to stop the test at the minimum value
n (MPa) of the load cycle, when the temperature of the material
700 R Δ k
0.1
A,50%
346 8.52
achieves its maximum value (due to the thermoelastic
600 0.1 194 7.66 effect): by so doing the temperature versus time curve
0.1 229 10.9
500 would be increasing just before test interruption and
T = 1.1 decreasing just after test interruption so that t* could be
400
more easily detected.
300
From temperature versus time data reported in Fig. 12,
time derivatives at t = (t*)+ were evaluated. Fig. 13 reports
an example of cooling curve as measured by means of the
200 smooth specimens infrared camera when testing smooth specimens loaded
hole φ=10 mm
hole φ=20 mm with R = 0.1 [31]. The higher resolution that can be
achieved by means of the thermocouples can be appreci-
b 104 105 106 N.cycles
ated by comparison with the range of values reported in
Fig. 11. Fatigue test results for smooth and notched specimens at load the ordinate of Fig. 12b. It can be seen that temperature
ratio R equal to 0.5 (a) and 0.1 (b). Scatter bands are given for 10% and variations after t*, which are adopted in order to calculate
90% survival probability. the cooling rate, are of the order of tenths of a Kelvin for
notched specimens and of some degrees for smooth speci-
0.0 2.0 4.0 6.0 (t-t*)[s] 10.0
0.0 mens. That is why only the thermocouples and not the
Hole φ = 20 mm available infrared camera could be used in order to mea-
Δσn = 333 MPa
-1.0 f = 15 Hz, Nf = 88490
sure meaningful data for notched specimens.
The specific energy dissipated per cycle Q was calculated
(T-T*) with both Eqs. (6) and (9) for the smooth specimens loaded
-2.0
[K] with R equal to 0.1 and the results were compared. The
Smooth specimen heat transfer coefficient involved in Eq. (6) was estimated
-3.0 Δσ = 460 MPa
f = 7 Hz, Nf = 96352 by means of three formulae taken from the technical liter-
ature. Details are reported in Appendix A. The specimen
R=-0.5
-4.0 surface was covered with a black paint and the emissivity
a

0 1.0 2.0 3.0 4.0 5.0 6.0 (t-t*) [s] 9.0 0 5 10 15 (t-t*) [s] 20
0 0
Smooth specimen
Δσ = 450 MPa
-0.2 f = 7 Hz, Nf = 156629 -0.5 Δσ=450 MPa, f=5.5
∂T
(T-T*) ∂t t =( t *)+
[K] -0.4 (T-T*) -1
[K]
-0.6
-1.5
Hole φ = 20 mm
-0.8 Δσn = 275 MPa Δσ=540 MPa, f=5 Hz
R=0.1
f = 20 Hz, Nf = 169157 -2
b
-2.5
Fig. 12. Temperature vs time acquisitions for R = 0.5 (a) and R = 0.1
(b) after the fatigue test has been suddenly stopped. Temperatures were Fig. 13. Temperature vs time acquisitions at R = 0.1 for smooth speci-
measured by means of thermocouples (see Fig. 9 for T* and t* definitions). mens. Temperatures were measured by means of the infrared camera.
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 91

Table 2
Comparison between Q values as calculated by means of Eqs. (6) and (9) for smooth specimens loaded with R = 0.1
Dr [MPa] f [Hz] Nf [cycles] Tstat [C] a [W/(m2K)] Q Eq. (6) [kJ/(m3cycle)] Q Eq. (9) [kJ/(m3cycle)]
300 14.0 2,000,000 33.2 4.8 16.6 17.4
400 10.0 2,000,000 50.5 5.9 50.3 47.4
450 5.5 262,330 42.9 5.6 68.2 64.5
450 7.0 156,629 45.3 5.7 53.9 58.8
450 6.0 231,572 45.5 5.7 65.5 64.8
490 5.0 86,817 45.8 5.8 71.1 79.2
520 5.5 66,603 50.9 6.1 79.6 83.2
540 5.0 51,310 49.7 6.1 87.6 89.2
400 10.0 698,007 49.1 5.9 51.7 51.9

was determined by comparison between the temperature 10 mm diameter hole tested at R = 0.1 cannot be rationa-
measured by means of a thermocouple and by means of lised in terms of dissipated energy, at least in the medium
the infrared camera. The result was j = 0.92. H was then and high cycle fatigue regime. For lives lower than
calculated by means of a trial and error procedure by 5 · 104 cycles all data fall in the same confidence band.
assigning a tentative value of H and then by solving the dif- That is probably due to the very small volume of material
ferential Eq. (6) in order to compare the temperature calcu- subjected to the maximum temperature compared to the
lated along the z-axis of the specimen reported in Fig. 8 area (approximately equal to 2 · 2 mm) where the thermo-
with that experimentally measured. Table 2 reports the couple wires were glued. Additional experimental tests are
results and highlights that the two procedures are in needed to investigate this issue.
agreement. In Fig. 15 the scatter bands valid for R = 0.5 and 0.1
The fatigue curves in terms of dissipated energy Q are are reported, which summarise the available experimental
reported in Fig. 14, where it is seen that the results of the
smooth specimens fall within the confidence band of the
notched specimen having a 20 mm diameter hole. Con- Q [kJ/(m3⋅cycle)]
versely the experimental data of the specimens having a 800 R QA,50% k
-0.5 123 3.57
600
Q [kJ/(m3 cycle)]
1000
R=-0.5 400
800

500
400
200
300

200
smooth specimens
100 hole =20 mm

100 smooth specimens 104 105 106 N. cycles

a
hole φ=20 mm

a 104 105 106 N. cycles Q [kJ/(m3 cycle)]

300
R QA,50% k
Q [kJ/(m3 cycle)] 200 0.1 29.0 2.92

200 R=0.1

100 100

20 40

smooth specimens
hole =20 mm
smooth specimens
hole φ =10 mm
hole φ =20 mm b 104 105 106 N. cycles

b 104 10 5 106 N. cycles Fig. 15. Fatigue curves and 10–90% scatter bands in terms of thermal
energy dissipated in a unit volume per cycle for smooth and notched
Fig. 14. Fatigue curves in terms of thermal energy dissipated in a unit specimens in AISI 304 L stainless steel at load ratios R equal to 0.5 (a)
volume of material per cycle at load ratios R equal to 0.5 (a) and 0.1 (b). and 0.1 (b).
92 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

results. The slopes of the curves are close to 3 for both the Q [kJ/(m3 cycle)]
R values and the diagrams highlight that the energy dissi- 140
pated at R = 0.5 is approximately 4.2 times higher than
that dissipated at R = 0.1 for the same fatigue life. When 120
comparing the scatter bands, Figs. 15 and 11 highlight that
100
they are much wider in terms of energy parameter Q than
in terms of stress amplitude. In particular Tr is equal to
80
1.2 and 1.1 for R = 0.5 and 0.1, respectively, while the
corresponding TQ values are equal to 2.2 and 1.9. Nonethe- 60
less one should note that the range of values covered by the
ordinate of the diagrams is also much higher in terms of 40
Constant amplitude σ a= 500 MPa
energy parameter Q than in terms of stress.
σa= 500 MPa after 105 cycles at 400 MPa
20 Constant amplitude σa= 400 MPa
8. Two-level fatigue test results σa= 400 MPa after 104 cycles at 500 MPa
0
0 5000 10000 15000 20000 30000
Two-level fatigue tests were conduced on smooth speci- Cycles
mens at stress ranges equal to 400 and 500 MPa with a load Fig. 16. Comparison between the specific thermal energy dissipated in
ratio R = 0.1. Both low–high and high–low sequences were two-level fatigue tests and in constant amplitude fatigue tests for smooth
adopted in order to compare the amount of energy dissipa- specimens at a load ratio R equal to 0.1.
tion at a given load level between constant amplitude and
variable amplitude fatigue tests. Additional specimens from
a new batch of material were prepared with the same smooth number of specimens used in this study. Then for the con-
geometry that was adopted in the previous tests. Two con- sidered material and testing condition, the specific heat
stant amplitude fatigue tests conduced at Dr = 400 MPa energy Q seems to be load history independent.
resulted in fatigue lives of 202,000 and 146,000 cycles,
respectively, which are lower than that expected from the 9. Conclusions
previous tests (see Fig. 11b). Anyway the amount of energy
released as heat in steady-state conditions was 60 kJ/ The fatigue tests performed at room temperature on 2-
(m3 cycle), which is higher than that released by the previous mm-thick smooth and notched specimens in AISI 304 L
specimens tested at the same stress level and it was seen to be stainless steel allowed to draw the following conclusions:
compatible with the fatigue curve drawn in Fig. 15b in terms
of energy parameter Q. The same held true at a stress level of 1. The mean evolution of the surface temperature of the
500 MPa: the fatigue lives obtained from three new speci- specimens rapidly increases at the beginning of the fatigue
mens were between 54,900 and 68,500 cycles, which are test and then tends to a stationary value. The temperature
slightly lower than the mean value of 80,000 cycles expected oscillation due to the thermoelastic effect is much lower
from the curve reported in Fig. 11b. Anyway Q resulted than the mean temperature evolution. The stationary
about equal to 100 kJ/(m3 cycle) and then the predicted fati- temperature increments are of the order of tens of Kelvin
gue life according to Fig. 15b was about 60,000 cycles, which for the point located at the mid-section of smooth speci-
agrees with the new experimental values. mens, and of the order of tenths of a Kelvin for the point
The results of variable amplitude fatigue tests are sum- located at the notch tip of holed specimens. These values
marised in Fig. 16, where the evolution of the specific are to be referred to the adopted testing conditions. In
energy dissipation Q versus the number of cycles is particular test frequencies were between 5 and 20 Hz at
reported. Each point represents an experimental estimation load ratios of R = 0.5 and 0.1.
of Q by means of the cooling rate method presented in the 2. The fatigue strength of smooth and notched specimens
present paper. The curves obtained from the previous con- having a notch tip radius equal to 10 mm can be rationa-
stant amplitude fatigue test results are plotted with filled lised for a given load ratio R in terms of energy Q
markers, while the present two-level fatigue test results released as heat in a unit volume of material per cycle.
are plotted by using open markers. Two specimens were Heat dissipation was measured at the notch tip in
fatigue tested initially at 400 MPa for 105 and 7.5 · 104 notched specimens.
cycles, respectively (approximately half the constant ampli- 3. Heat dissipation at R = 0.5 is about 4.2 times higher
tude fatigue life) and then at 500 MPa until fracture, which than that at R = 0.1 for the same fatigue life.
occurred after 10,100 and 18,900 cycles, respectively. One 4. The high cycle fatigue data relative to notched speci-
specimen was fatigue tested at 500 MPa for 104 cycles mens having a notch tip radius equal to 5 mm do not
and then at 400 MPa for 131,000 cycles, before fracture collapse in the energy-based fatigue scatter band: that
occurred. No differences were noticed between the energy is supposed to be due to the limited spatial resolution
dissipation curves between constant amplitude and variable of the adopted thermocouples having wires diameter
amplitude fatigue tests, at least on the basis of the limited equal to 0.127 mm.
 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94 93

Table 3
Heat transfer coefficient by natural convection as calculated by means of different formulae found in the literature
k [W/(m K)] a [W/(m2 K)] Pr Gr Ra
Eq. (11) Eq. (12) Eq. (13) Mean value
16 6.09 5.97 6.30 6.1 0.701 1.88 · 106 1.32 · 106

5. on the basis of the performed two-level fatigue tests, at It was verified [24] that such formulae give very similar re-
400 and 500 MPa, respectively, the heat dissipated in a sults. As an example, for one set of experimental conditions
unit volume per cycle is independent of the load sequence, comparison among Eqs. (11)–(13) is reported in Table 3.
but depends only on the applied stress amplitude.
References

Appendix A [1] La Rosa G, Risitano A. Thermographic methodology for rapid

determination of the fatigue limit of materials and mechanical
components. Int J Fatigue 2000;22:65–73.
In order to apply Eq. (6) an estimation of the heat trans- [2] Middelford K, Harig H. Thermometric investigation of the fatigue
fer coefficient by convection must be performed. It was behaviour of power-forged specimen. The future of powder metal-
assumed that the heat transfer from the surface of the spec- lurgy, P/M 86. In: Proceedings of international powder metallurgy
imen to the surroundings by convection occurs by natural conference and exhibition. Freiburg: Verlag Schmid GmbH; 1986. p.
499–502.
convection. Formulae valid for vertical surfaces subjected
[3] Dengel D, Harig H. Estimation of the fatigue limit by progressively-
to natural convection can be found in the literature and increasing load tests. Fatigue Fract Eng Mater Struct 1980;3:113–28.
are reported in [27]. Such formulae correlate the non- [4] Harig H, Weber M. Estimation of crack initiation in plain carbon
dimensional Grashof (Gr), Nusselt (Nu) and Prandtl (Pr) steels by thermometric methods. Defects, fracture and fatigue. In:
numbers, or alternatively the Rayleigh (Ra = Gr Æ Pr), Nus- Proceedings of the 2nd international symposium. The Hague:
selt and Prandtl numbers. Such non-dimensional numbers Martinus Nijhoff Publ; 1983. p. 161–70.
[5] Luong MP. Fatigue limit evaluation of metals using an infrared
are given by thermographic technique. Mech Mater 1988;28:155–63.
aL cpa  l L3  b  q2a  g  DT a [6] Blarasin A, Fracchia R, Pozzati M. Rapid determination of the
Nu ¼ ; Pr ¼ ; Gr ¼ ; fatigue limit of materials and components by means of the infrared
ka ka l2 technique. ATA – Ingegneria Automotoristica 1988;51(5):255–65.
where L is the length of the vertical surface (i.e., the length [7] Catalbiano T, Geraci A, Orlando M. Analysis of the fatigue strength
of specimens by means of the infrared technique. Il Progettista
of the reduced section of the specimen), cpa is the specific Industriale 1984:2.
heat of the air [J/(kg K)]; ka is the thermal conductivity of [8] Reifsnider KL, Williams RS. Determination of fatigue-related heat
the air [W/(m K)]; l is the viscosity of the air [kg/(m s)]; b emission in composite materials. Exp Mech 1974;14:479–85.
is the coefficient of thermal expansion at constant pressure [9] Boulanger T, Chrysochoos A, Mabru C, Galtier A. Calorimetric
[K1]; qa is the density of the air [kg/m3]; g is the gravity analysis of dissipative and thermoelastic effects associated with the
fatigue behaviour of steels. Int J Fatigue 2004;26:221–9.
acceleration equal to 9.81 m/s2; DTa is the difference be- [10] Curà F, Curti G, Sesana R. A new iteration method for the
tween the surface temperature of the specimen and the thermographic determination of fatigue limit in steels. Int J Fatigue
room temperature. All the thermophysical properties of 2005;27:453–9.
the air were evaluated at a mean temperature value Tf be- [11] Cugy P, Galtier A. Microplasticity and temperature increase in low
tween that of the specimen surface Tp and that of the air Ta: carbon steels. Fatigue 2002. In: Blom AF, editors. Proceedings of the
8th international fatigue congress, vol. I, EMAS; 2002. p. 549–56.
Tp þ T1 [12] Miller KJ. The three thresholds for fatigue crack propagation. In:
Tf ¼ .
2 Piascik RS, Newman JC, Dowling NE, editors. ASTM STP 1296;
The thermophysical properties of the air at the temperature 1997. p. 267–86.
[13] Akiniwa Y, Tanaka K, Akimura H. Microstructural effects on crack
Tf were found from the literature [32]. The following three closure and propagation thresholds of small fatigue cracks. Fatigue
different formulae reported in [27] were used in order to Fract Eng Mater Struct 2001;24:817–29.
evaluate the heat transfer coefficient a in the case of natural [14] Murakami Y. Metal fatigue: effects of defects and non metallic
convection: inclusions. Elsevier; 2002.
8 92 [15] Luong MP. Infrared scanning of fatigue in metals. Nucl Eng Des
>
< 1=6
>
= 1995;158:363–76.
0:387  Ra [16] Halford GR. The energy required for fatigue. J Mater 1966;1:3–18.
Nu ¼ 0:825 þ h  9=16 i ; ð11Þ
>
:
8=27 >
; [17] Blotny R, Kaleta J. A method for determining the heat energy of the
1 þ 0:492 Pr fatigue process in metals under uniaxial stress. Int J Fatigue
Pr 1986;1:29–33.
Nu ¼ A  ðGr  PrÞ1=4 ; A4 ¼ ; [18] Feltner CE, Morrow JD. Microplastic strain hysteresis energy as a
2:43478 þ 4:884  Pr1=2 þ 4:95283  Pr criterion for fatigue fracture. Trans. ASME, Ser. D: J Basic Eng
ð12Þ 1961;83:15–22.
 1=4 [19] Charkaluk E, Bignonnet A, Costantinescu A, Dang Van K. Fatigue
Pr
Nu ¼ 0:667   Ra1=4 . ð13Þ design of structures under thermomechanical loadings. Fatigue Fract
0:952 þ Pr Eng Mater Struct 2002;25:1199–206.
94 G. Meneghetti / International Journal of Fatigue 29 (2007) 81–94

[20] Klingbeil NW. A total dissipated energy theory of fatigue crack [26] Eusebione E, Longo F, Meneghetti G. Fatigue tests monitoring by
growth in ductile solids. Int J Fatigue 2003;25:117–28. means of an infrared camera, Internal Report CM 00/2. University of
[21] Kaleta J, Blotny R, Harig H. Energy stored in a specimen Padova; 2000 [in Italian].
under fatigue limit loading conditions. J Test Eval 1990;19: [27] Bonacina C, Cavallini A, Mattarolo L. Heat transfer. Cleup Editore,
326–33. Padova; 1992 [in Italian].
[22] Fargione G, Geraci A, La Rosa G, Risitano A. Rapid determination [28] Atzori B, Meneghetti G. Energy dissipation in low cycle fatigue of
of the fatigue curve by the thermographic method. Int J Fatigue austempered ductile irons. LCF 5. In: Portella PD, Sehitoglu H,
2002;24:11–9. Hatanaka K, editors. Proceedings of the 5th international conference
[23] Gamstedt EK, Redon O, Brønsted P. Fatigue dissipation and failure on low cycle fatigue. Berlin: DVM; 2003. p. 147–52.
in unidirectional and angle-ply glass fibre/carbon fibre hybrid [29] Lazzarin P, Tovo R, Meneghetti G. Fatigue crack initiation and
laminates. Key Eng Mater 2002;221–222:35–48. propagation phases near notches in metals with low notch sensitivity.
[24] Atzori B, Gasparini E, Meneghetti G. Thermographic analysis of the Int J Fatigue 1997;19:647–57.
fatigue strength of metallic materials. In: Proceedings of 30th AIAS [30] ASTM, Metal handbook – 9th ed., vol. 3; 1990.
national conference; 2001. p. 367–76 [in Italian]. [31] Meneghetti G, Vanzin C. Fatigue characterisation of a stainless steel
[25] Yang B, Liaw PK, Wang H, Jang L, Huang JY, Kuo RC, based on surface temperature measurements. In: Proceedings of the
Huang JG. Thermographic investigation of the fatigue behav- 31st AIAS national conference; 2002 [in Italian].
iour of reactor pressure vessel steels. Mater Sci Eng A [32] Perry RH, Green DW. Perry’s chemical engineers’ handbook. 7th ed.
2001;314:131–9. McGraw-Hill; 1997.