Download as pdf or txt
Download as pdf or txt
You are on page 1of 33

Accepted Manuscript

The influence of the load frequency on the high cycle fatigue behaviour

Giovanna Fargione, Fabio Giudice, Antonino Risitano

PII: S0167-8442(16)30304-4
DOI: http://dx.doi.org/10.1016/j.tafmec.2016.12.004
Reference: TAFMEC 1793

To appear in: Theoretical and Applied Fracture Mechanics

Received Date: 30 September 2016


Revised Date: 22 November 2016
Accepted Date: 18 December 2016

Please cite this article as: G. Fargione, F. Giudice, A. Risitano, The influence of the load frequency on the high cycle
fatigue behaviour, Theoretical and Applied Fracture Mechanics (2016), doi: http://dx.doi.org/10.1016/j.tafmec.
2016.12.004

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.
The influence of the load frequency on the high cycle fatigue behaviour

Giovanna Fargionea*, Fabio Giudicea, Antonino Risitanoa

a
DICAR - Department of Civil Engineering and Architecture, Mechanical Division

University of Catania

Via Santa Sofia 64, 95123 Catania, Italia

* Corresponding author

Phone +39 095 7382419

Fax +39 095 337994

Email: gfargion@dii.unict.it

1
Nomenclature

b Width of the specimen

E Young modulus

Eh Heat quantity released to breakage per unit volume

Ec Critical energy to breakage

f Load frequency

K1 Coefficient of thermal conductivity

K2 Coefficient of thermal exchange by convection

K3 Coefficient characterising the component of heat transmission due to irradiation

Kc Coefficient characterising the sum of the components of heat transmission due to

conduction and convection

h step size of numerical integration

l Longitudinal length of the gauge portion of the specimen

n number of numerical integration steps

N Number of cycle

Nf Number of cycle to failure

p Perimeter of specimen section


Q Total heat quantity per time unit released by gauge volume of the specimen

 cd
Q Heat quantity per time unit released by gauge volume of the specimen (conduction)

 cv
Q Heat quantity per time unit released by gauge volume of the specimen (convection)

i
Q Heat quantity per time unit released by gauge volume of the specimen (irradiation)

R Load ratio

R* Load ratio with overstress due to thermal expansion

s Thickness of the specimen

2
S0 Applied load corresponding to fatigue limit 0

Sc Area of specimen section

Sp Area of outer surface of the gauge portion of the specimen

Tm Mean value of TS and T∞

TS Surface temperature of the specimen

T∞ Ambient temperature

  Coefficient of thermal expansion

T temperature increment (gap between mean surface temperature of the specimen and

the environmental temperature)

TS Stabilised temperature increment

 Surface emissivity

0 Fatigue limit

a Amplitude of fatigue stress

a* Amplitude of fatigue stress with overstress due to thermal expansion

 Amplitude of overstress due to thermal expansion

m Fatigue mean stress

m* Fatigue mean stress with overstress due to thermal expansion

min Fatigue minimum stress

max Fatigue maximum stress

n Stephan-Boltzman constant

r Rupture strength

y Yield strength

Nf
 Integral of T-N curve (Φ =  ΔT dN )
0

3
Abstract

The energetic methods to determine the fatigue limit, proposed by various researchers in the last

decades, are based on the evidence that fatigue is an energy dissipation process, and most of the

dissipated energy is converted into heat, which manifests itself in the form of temperature change.

The study of fatigue behaviour through the determination of the heat quantity released during the

fatigue tests, and the analysis of the temperature variation of material under fatigue loading, by

mean of thermography-based methods, constitutes a well-established approach.

Reaching very high temperature in the specimen is a critical aspect of this phenomenon of

conversion of the dissipated energy into heat, particularly with regards to materials characterised by

high thermal releases, that could be subject to microstructural changing and behaviour decay.

This effect can be exalted by the use of ultrasonic fatigue testing systems for the characterization of

very high cycle fatigue behaviour of material.

To highlight the influence of the fatigue test frequency on the specimen temperature that already

occurs at low frequencies, in this paper a study of the thermal behaviour of AISI 304 stressed by

fatigue loads in high cycle fatigue testing at different frequency values, based on energetic method

to determine the fatigue curve, is presented.

Keywords

High cycle fatigue; Thermography-based analysis; Load history; Test frequency

4
1. Introduction

The study of high cycle fatigue behaviour through the analysis of the temperature variation of

material and the determination of the heat quantity released during the fatigue tests is common to

the various experiences in the energy-based approach to fatigue, primarily aimed at the

development of fast methods to determine the dynamic characteristics of the materials.

Many scholars have performed research in this field by means of thermography-based analysis

methods. Luong [1, 2] observed the temperature increment in metallic specimens surface for

applied stresses below the fatigue limit, and utilized two curves to interpolate experimental data,

whose intersection was indicated as the fatigue limit, thus developing the “Two Curve Method”.

Curà et al. [3] proposed an iteration procedure for the determination of the fatigue limit of mild

steel, based on the Luong graphical representation.

Amiri and Khonsari [4,5] studied the initial rate of temperature rise as a function of time during

fatigue testing to predict the fatigue life.

Fan et al. [6] analyzed the temperature evolution due to localized microplasticity to evaluate the

fatigue damage status, and established an energetic damage model to predict the residual fatigue

life.

Acting in the same way, complementarily or alternatively to the approach based on the analysis of

temperature variations, other scholars focused on the microstructure evolution of material in high

cycle fatigue process, assuming to be necessary to calculate the energy dissipation for obtaining an

accurate observation of fatigue behaviour [7-9]. Among the others, Meneghetti et al. [10,11]

calculated the specific heat dissipation by measuring the cooling rate after stopping the fatigue

loading process, introduced a released heat parameter Q, and utilized it as a fatigue damage

indicator to analyze the fatigue behaviour of stainless steel specimens.

Liakat and Khonsari [12,13] introduced the concept of fatigue fracture entropy (FFE), applying the

thermodynamic entropy production during the fatigue process to the estimation of the fatigue life of

5
components stressed by cyclic loads. Guo et al. [14] proposed a method for the rapid evaluation of

high cycle fatigue behaviour, assuming the intrinsic energy dissipation as the fatigue damage

indicator of stainless steel, and eliminating the interference of internal friction causing no damage.

Among the fast methods to determine the dynamic characteristics of the materials in high cycle

fatigue, the one called Risitano Method (by the name of the first author) has been a precursor, and

showed particular effectiveness over the years, primarily because it drastically reduces the time

analysis and it also allows the use of a limited number of specimens to be test (one at least), and

subsequently for its bent for introducing energetic analysis criteria. It was born as a thermographic

method, based on the observation of the temperature increase on the surface of a specimen (or

mechanical component) stressed by a dynamic load, to rapidly estimate the fatigue limit by utilizing

the stabilization of temperature increment [15-17].

According to this method, the fatigue limit is evaluated as the highest stress value that produce no

thermal increment on the surface of a specimen or mechanical component cyclically loaded.

As it was been noted before [18], when a specimen is cyclically loaded above its fatigue limit, its

superficial temperature changes according three characteristic phases: phase I, with initial thermal

gradient; phase II, with temperature stabilization; phase III, with a final rapid increase. First of all, it

was noted that the higher were the applied stresses (above the fatigue limit), the higher was the

increase in temperature and the lower was the number of cycles for reaching the stabilization of

temperature increment. Subsequently, it was observed that the fatigue limit can be correlated to the

stabilization temperature (phase II).

Particularly, the fatigue limit can be determined by applying to the specimen different stress range

levels, and detecting the stabilized temperature increment by means of a thermographic technique.

Plotting the stabilization temperature against the applied stress, the value of the fatigue limit can be

found as the intercept of the curve on the stresses axis (thus the method is also called One Curve

Method), since this intersection corresponds to the highest stress range for which there is no

temperature variation.

6
These results have been verified and analyzed, applying the method to different materials and to

some simple components (notched specimens, butt welded joints) [17,19].

As a development of the method, the fatigue life of material can be described using the integral of

the temperature over time up to fracture, so to rapidly define the Wöhler curve [20]. With this

regard, it has been defined the energetic parameter Ec, constant for each material, as the limit value

of plastic energy corresponding to the onset of the breakage of specimens or machine components

[20-22].

This limit energy Ec is proportional to the integral  of the T–N curve, extended to the number of

cycles to failure Nf. As the energy absorbed by a unit volume of material till failure is the same

when load histories at different levels are applied, applying the stepped load history according the

Risitano Method, it is possible to evaluate the characteristic value of , to obtain the stabilized

temperature increment TS at each applied stress range level , and then to determine the

corresponding value of Nf. Thus, the fatigue S–N curve (or Wöhler curve) can be drawn on the basis

of the pairs (, Nf) obtained.

The increase of temperature is linked to the heat released by the material due to the

microplasticization phenomena that occur inside the material itself, and cause heat releasing and

damage. Therefore, testing a specimen with increasing loads and monitoring the temperature at the

same time, it is possible to determine when the first damage occurs. Correlating the temperature

increase with the quantity of the heat released during the plasticization phase, it is possible to

establish the material life dead line also [23].

In a number of previous works the validity of the methodology has been defined operating on

different materials and frequencies and obtaining some results about the fatigue limit or the Wöhler

curves, which coincide with those obtained during the traditional fatigue tests [17,19,24-26].

In the present paper the authors, basing their approach on the Risitano Method, focus on the

influence of the load frequency for those materials with high thermal releases which, during fatigue

tests, can be subject to microstructural changes, mechanical behaviour decay, and overloads
7
due to thermal stresses. Fatigue tests on AISI 304, which is characterized by this kind of thermal

behaviour, have been performed.

The evaluation of frequency influence has been considered particularly interesting in the

perspective of the spread and wide development of the characterization of fatigue behaviour of

material at very high cycle number (more than 107 cycles), that occurred in the last year [27]. The

well-established use of ultrasonic fatigue testing systems (having capability of testing the materials

up to 20 kHz frequency) to reach such high numbers of cycles exalts the effect of reaching very

high temperature in the specimen.

2. Performed tests and modalities of execution

In order to look into the possible effects of the frequency and of the loading typology on the fatigue

behaviour determined by mean of the thermographic method, some tests have been programmed,

varying both the frequency of the load application and the sequences of the applied loads. An

Instron 1332 machine with a load cell of 100 kN has been used. The tests have been carried on

materials able to release high quantity of heat during the plasticization phase. In particular, AISI

304 slick specimens with a calibrated area of 20x70 have been used, as showed in Figure 1, that

reports the specimen peculiar geometric properties, the chemical composition of the material, and

its basic mechanical properties.

Two different types of tests have been conducted: load check tests, traction-compression (R= - l)

with a constant load during the entire test (till the breakage of the specimen) (case 0) and tests with

incremental step load (case 1). In particular, during the step tests the load has been increased, after a

prefixed number of cycles (10.000).

The tests with a constant load (case 0) have been done at 225 MPa (27 kN). For the step tests (case

1), the sequence of the applied loads have been: S = {7, 11, 15, l9, 23, 27, 31, 35} kN.

8
All the load typologies have been applied at three different values of the test frequency: {10, 20,

30} Hz.

The AGEMA THV 900 thermocamera (working in the 8-12 μm band) has been used to acquire

infrared images. During the tests, the superficial temperature of the specimen has been always

recorded. The images have been elaborated using the MATLAB 8.0 software.

For each of the different specimens has been calculated both the fatigue limit σ0 and the energy

parameter Φ [20,22]. The latter has been calculated by means of a numerical integration that

approximates the integral of the T–N curve, extended to the number of cycles to failure Nf, where

T is the thermal gap between the mean surface temperature of the specimen and the environmental

temperature. The fatigue limit has been calculated monitoring the temperature increments ΔT. For

this reason it has been fixed a reference image, defined as “image zero”. The other ones have been

lined up following it and defining pixel by pixel the temperature increase; the integration of the ΔT

distribution on the surface gives back the ΔT related to the analyzed image. It is useful to remember

that the choice of the integration surface size, influences neither the fatigue limit nor the

determination of the energy parameters, as it has been revealed in previous work [22].

3. Brief reference modelling of energetic phenomena

In 1934 Taylor and Quinney [28] demonstrated that the great part of the plastic deformation energy

used for the metals deformation is converted into heat. Knowing the surface temperature of the

specimen, it is possible to define the total energy dissipated in heat and, as a consequence, the

quantity of the plastic deformation energy. According to the hypothesis that for each specimen

(characterized by a material and its geometrical properties) exists a characteristic value Ec of the

plastic deformation energy up to the breakage, it is possible to define the specimen life itself [20-

22].

9
As it is known, the quantity of heat per time unit released by the gauge volume of the specimen,

which is the thermal power exchanged with the environment, is given by the sum of the three

components due to conduction, convention and irradiation:

Q  Q cd  Q cv  Q i (1)

The component due to the irradiation Q i is proportional to the fourth power of the temperature

according to the relation:


Q i   S p  n Ts4  T4  (2)

where  is the surface emissivity, Sp is the area of outer surface of the considered portion of the

specimen (that is p·l, where p is the perimeter of the section of the specimen, l is the longitudinal

length of the considered portion of the specimen), n is the Stephan-Boltzman constant (5.67·10-8

W/m2 K4), Ts is the surface temperature of the specimen, T∞ is the ambient temperature.

Eq. (2), for values of thermal gaps between Ts and T∞ not so high (within 200 °C), can be also

written as follows:


Q i   S p  n 4 Tm3 Ts  T  (3)

where Tm is the mean value of Ts and T∞.

In the hypothesis of linear law conduction of the specimen, and introducing the approximate Eq.

(3), Eq. (1) can be written as follows:

    
Q  K1 Sc / l  Ts  T  K 2 S p Ts  T  K3 S p Ts  T  (4)

where K1 is the coefficient of thermal conductivity, Sc is the area of the section of the specimen (that

is b·s, where b is the width of the specimen, s is its thickness), K2 is the coefficient of thermal

exchange by convection, and K3 is assumed to be expressed by:


10
K3    n 4 Tm3 (5)

As a result, if the values of thermal gaps between Ts and T∞ are lower than 200°C, the quantity of

heat per time unit released by the specimen can be considered as proportional to the thermal gap:


Q  K Ts  T  (6)

where K is assumed to be expressed by:

K  K1 Sc / l   K 2  K3  S p (7)

In Eq. (4) and Eq. (7), K1, K2, K3, are the coefficients which characterize the three components of the

heat transmission, Sc/l is the geometrical parameter related to the conduction phenomenon, and Sp is

the geometrical parameter related to the convention and irradiation phenomena. The coefficient K

defined by Eq. (7) includes both the coefficients of the heat transmission and the geometrical

parameters of the specimen. Eq. (6) highlights how, for the same specimens or mechanical

elements, the heat quantity released per time unit is proportional step by step to the thermal gap.

The Eq. (4) can also be written as follows:

  
Q  Kc Ts  T  K3 S p Ts  T  (8)

where Kc is assumed to be expressed by:

K c  K1 Sc / l   K 2 S p (9)

Working with frequencies and loads such as to avoid that the specimen reaches high temperature,

replacing in Eq. (8) the data related to two different tests carried out according to constant load

procedures (case 0) with the same loads and different frequencies or vice versa, and finally solving

the equation obtained comparing the two corresponding expressions (8), it is possible to obtain Kc,

11
and then from Eq. (9), K1 as a function of K2, or vice versa K2 as a function of K1, being the latter

case particularly useful to overcome the well-known difficulty in the assessment of the coefficient

of thermal exchange by convection. With this purpose, in general terms, it would be better to

average out the values obtained by a set of at least three specimens.

4. Analysis of results

The trends of the temperature increments on the specimen surface, with the number of the cycles for

three different specimens (coded as 1,4,7), stressed at the three different test frequencies 10, 20, 30

Hz, respectively, with an stepped load increasing at prefixed number of cycles (case 1), have been

reported in Figures 2-4.

In Figure 5, instead, as an example, they have been reported the curves of the mean temperature

increments on the specimen surface for three specimens, stressed at each frequency (10, 20, 30 Hz)

by a constant load (27kN) up to the breakage (case 0).

The thermal behaviour of the material, considerably sensitive to the test frequencies and to the

modality of applied loads, has induced the authors to take into consideration some aspects which

could be important for those scholars who are interested in the dynamic characterization of so

particular materials.

A preliminary remark concerns an effect of the high increases of the surface temperature reached by

the chosen material, which has allowed to obtain temperatures values for which the influence of the

measure noise was negligible.

In Table 1 they have been reported the fundamental data related to the tested specimens and for

which it has been determined the fatigue limit by means of Risitano Method, using both the

modalities of load application: stepped load (case 1) and constant load (case 0).

12
Table 1. Fatigue limits determined by means of Risitano Method

CODE Frequency [Hz] Load application S0 [kN] σ0 [MPa]

AISI_304_1 10 stepped 12.19 101.63

AISI_304_2 10 stepped 13.56 113.04

AISI_304_3 10 stepped 13.37 111.46

AISI_304_4 20 stepped 13.54 112.85

AISI_304_5 20 stepped 14.19 118.25

AISI_304_6 20 stepped 12.00 100.02

AISI_304_7 30.3 stepped 12.22 101.86

AISI_304_8 30.3 stepped 13.21 110.08

AISI_304_9 30.3 stepped 12.20 101.69

AISI_304_18 10 const (27 kN) 14.50 120.83

AISI_304_19 10 const (27 kN) 12.74 106.20

AISI_304_20 10 const (27 kN) 12.67 105.63

AISI_304_21 20 const (27 kN) 11.57 96.41

AISI_304_22 20 const (27 kN) 13.62 113.50

AISI_304_23 20 const (27 kN) 12.17 101.39

AISI_304_24 30.3 const (27 kN) 13.78 114.85

AISI_304_25 30.3 const (27 kN) 12.20 101.66

AISI_304_26 30.3 const (27 kN) 14.26 118.50

In particular the first column (Code) characterizes the specimen, the other columns indicate the

frequency, the modality of load application, the value of the fatigue limit in terms of applied load

(S0) and corresponding tension (σ0). Just comparing the values of the fatigue limits obtained, and

reported in Table 1, first of all it can be noticed that they are influenced neither by the test

frequency, nor by the modality of load application. In Figure 6 the distribution of fatigue limit

data from Table 1 has been represented by showing the mean values as single data point, and

13
error bars (representing standard errors), for the three different frequencies and the two

types of load history. The stability of fatigue limit values despite the change in load frequency

and modality of load application is evidenced by the good representativeness of mean values

(in all cases, the standard errors are contained within 5% of mean values), and the substantial

overlapping of error bars.

For the load step (19 kN), higher than the fatigue limit (about 13 kN), and a number of cycles up to

10000, three specimens for each frequency (10, 20, 30 Hz) were tested. In Figure 7 it has been

reported the typical trend of the average temperature increments (before the stabilization phase) on

the surface of the calibrated area, for the three different frequencies. In any case, the difference of

the detected values for each specimens, compared to the mean values reported in the Figure 7, never

exceeded 3%. Similar performances have been noticed for the other load values during the stepped

tests; they have been not reported here for the sake of brevity. True to form, the graph points out

that the value of the temperature rise increases with the frequency of the load application, a

phenomenon which can be also observed in Figure 5.

If the coefficients of thermal transmission were constant and if the specimens had identical

behaviour, the temperature would rise in proportion to the frequency. On the contrary, the amount

of released heat up to the breakage would be approximately constant because it will be proportional

to the integral over time of the released heat, until the temperature reaches values above 200 °C.

The temperature values, which has overcome the 550°C during the breakage phase (not reported in

the Figures 2, 3, 4 only to reduce the ordinate scale), indicate that when fatigue tests by means of

more traditional methods are performed, for loading values higher than the fatigue limit, it could be

necessary to consider microstructure instability, damage and other phenomena of mechanical

behaviour decay, that are produced on the material by these high levels of temperature, and

in some cases also the overstressing induced by the linear expansion due to considerable

thermal gap and proportional to it. Otherwise, the estimates of the parameters which characterize

the fatigue behaviour of the material could be incorrect. This problem, as it’s clear, should be much

14
more evident in the fatigue testing at a very high number of cycles, in which it’s necessary to work

at very high test frequencies and as a consequence at high corresponding surface temperatures of

the specimens [29].

In general terms, the most common problem for stainless steels at high temperatures (starting

from 500 °C) is that the alloys forms a hard, brittle phase, called sigma. This microstructural

change can occur also in austenitic stainless steels (such as AISI 304), that can suffer

significant ductility loss and embrittlement due to sigma phase formation, also at medium

exposure times (minutes-hours). With regards to the specific tested material, with chemical

composition reported in Figure 1, it must be highlighted that even small amounts of silicon

and molybdenum markedly accelerate the formation of sigma phase [30].

Precipitation of carbides at grain boundaries, phenomenon called “sensitization”, in austenitic

stainless steels occurs starting from 400-500 °C and low exposure times (seconds-minutes).

Sensitization trigger parameters (temperature and exposure time) are much lower than the

higher the carbon content (such as 0.06 of the tested material, see Figure 1) [31]. This

phenomenon particularly implies a significant local depletion of chromium, with lower

corrosion resistance, and could favour the incidence of intergranular fracture, resulting in a

loss of toughness. Detailed descriptions of carbides precipitation process in austenitic steels,

and its effect in terms of microstructure modifications are reported in literature [32].

Grain growth phenomena, increasing creep strength but simultaneously reducing ductility,

may also occur but at higher temperatures (above 900 °C).

Focusing on the stability of mechanical behaviour, AISI 304 stress-strain curves at high

temperatures show a significant decay already just above 200 °C, with a reduction of about

25% for both yield strength y and rupture strength r, compared to the same properties at

room conditions [33]. Strength and stiffness reduction factors at elevated temperatures and

low exposure times for a range of stainless steel alloys (including the tested material) have

been detailed [34]. A recently proposed correlation between the beginning of the stress-strain
15
curve's nonlinear trend, and the fatigue limit [35] outlines in general terms an

interdependency of mechanical properties decay phenomena at static and dynamic load

conditions. The detrimental effect of temperature on fatigue behaviour of AISI 304 has been

clearly evidenced in terms of reduction of cycles to failure, and the increase in fatigue crack

growth rate, already starting from 300 °C [36].

These evidences confirm some conclusions previously outlined about the achievement of high

temperatures in fatigue tests, that can generate significant modifications of the material

microstructure, with a reduction of the fatigue resistance and, as a consequence, a reduction of the

limit energy to breakage [21].

Overstressing induced by the linear expansion due to thermal gap should not be considered

for the fatigue tests discussed here, because they have been all performed at imposed load. On

the other side, this specific thermal effect should be considered for fatigue tests at imposed

strain, being the specimen for fatigue test usually constrained by the grips. These type of tests

are generally performed for low number of cycles, which means low frequencies and

substantially negligible thermal gap. The thermal expansion overstress phenomena become

particularly important instead for real fatigue load condition of mechanical components

constrained at the extremities (as for the classic example of the connecting rods). Considering

the case of a mechanical component subjected to mono-axial fatigue load at the same

conditions of specimens subjected to case 0 load history (stress amplitude a constant, R = -1),

therefore with thermal behaviour referable to that described in Figure 5, and assuming a

reasonable local effect of microplasticizations grown in the stabilisation phase, and a

substantial elastic behaviour at global level, the amplitude of overstresses induced by the

temperature increment can be estimated as  = E T, where is the coefficient of thermal

expansion and E is the Young modulus. At the same conditions, for AISI 304 ( = 16.6

strain/°C, E = 200000 MPa), a considerable overstress amplitude around 65 MPa for each 20

°C of thermal gap could be estimated. The effect of such an overstress condition on fatigue
16
load parameters cannot be neglected. Being the overstress a compression stress of amplitude

, its effect results in a new fatigue load condition with load ratio R* = (min - )/(max - ),

unchanged stress amplitude a* = a, and a new mean stress m* = m -  (for R = -1, m* = -

).

Using the available data, it has been valued that just for a thermal increment higher than 200°C the

error obtained linearizing the term of thermal power released by the irradiation, as it has been

proposed in Eq. (4), exceed the allowable error limit of 10%, compared to the results obtained if it

wasn’t applied the above-mentioned approximation. This confirms that, within the mentioned

temperature limit, it is possible to assume the quantity of the heat released, and therefore the critical

energy to breakage Ec, that is the amount of energy to failure per unit volume, as proportional to the

Nf
area subtended the curve ΔT-N, defined by the parameter Φ =  ΔT dN , that is the cumulative
0

amount of the thermal increments per volume unit in the fracture zone, for equal test frequency

N f /f
[20,22]. For different frequency f the equivalent parameter can be estimated as Φ* =  ΔT f dt .
0

If the material chosen for the tests shows a high quantity of dissipated heat during the plasticization

phase (such as AISI 304), in the final load step the values will reach values higher than 200°C (see

the highest curves ΔT-N in Figures 2, 3 and 4). In this case the determination of the total heat

quantity using Eq. (4) brings to considerable errors, and so Eq. (2) should be used to determine the

irradiation term. In order to evaluate the error that occurs in this condition so extreme, for each

specimen it has been calculated the values of the heat quantity released to breakage per unit volume

Eh and Eh*, using the Eq. (2) and the Eq. (3) to express Q i , respectively.

The graph in Figure 8 shows, for the three frequency values and for the two loading cases, the

trends of the values of the heat quantity released to breakage per unit volume Eh and Eh*,

confirming quatitatively what it has been described before. The above-mentioned graph points out

that basically the higher is the frequency of the tests (which means higher temperatures), the higher

is the difference between Eh and Eh* values. Therefore, assuming that the induced overloads do not
17
cause microstructural changes on the material, it is possible to deduce that Eh and Eh* depends on

the loading application modality only through the increasing in temperature. It is possible to

observe also that, until no microstructural changes arise in the material, limit energy to breakage

does not depend on the modality of load application. This condition does not take place, as an

example, for the highest tests load (35 kN), because the specimen surface temperatures go beyond

550°C, with potential microstructural changes in the material.

As it can be noted in the same Figure 8, the values of Eh (Eh*) are very different for the two case of

load application. In the case 1 of stepped loads, as said before, the temperatures reached were more

than 500 °C. Therefore, taking into account the microstructural phenomena induced on the

material by the different thermal condition, and the consequences in terms of its mechanical

behaviour described before, in this case the failure modality is completely different than in case 0.

Working with specimens which are similar for geometry and material, the definition of the Wöhler

curve refers to the parameter Φ, because for its determination is not necessary to know the

coefficients of the heat exchange [20,22]. For the calculation of Φ, reference was made to the case

of the broken specimens with constant load (case 0) in which problems resulting from the

attainment of excessive temperatures were not present (see Figure 5).

This choice reflects a clear evidence that appear from the test campaign: for each fatigue test,

the frequency must be chosen in function of the thermal properties of the material and the

load condition, so that the curve T-N and its significant parameters are clearly identifiable;

moreover, this condition is necessary for a significant estimation of parameter . As a matter

of fact, test frequencies too low imply insignificant temperature increment T; on the other

side, test frequencies too high, particularly if coupled with high loads, can imply unstable

trends of temperature increment T, such as pointed out by the highest curves ΔT-N in

Figures 2, 3 and 4, far from homogeneous behaviour which comprises the phase of

temperature stabilization, well represented by the curves in Figure 5.

18
In conclusion, selecting appropriately the frequency and the values of load steps, it is possible

to program fatigue tests in order to limit the thermal gaps. This expedient becomes essential

when high thermal release materials, such as AISI 304, are dynamically stressed, showing a

particularly significant increment of temperature, that can also carry in errors on the

evaluation of the energetic parameters, or even worse make it impossible.

Tables 2-4 reported the values of Φ determined for the different specimens at the different

frequencies (10, 20, 30 Hz). These values as been estimated approximating , that as it has

been stated before, is the integral of T-N curve in the domain [0, Nf], by numerical

integration with uniform grid and trapezoidal rule Φ ≈ h  k  0 ΔTN k 1   ΔTN k  /2 , with
n 1

integration step size (grid spacing) h = Nf/n, number of integration step n = Nf/1000, first grid

point N0 = 0, last grid point Nn = Nf (number of grid point n+1).

In the last columns it has been reported the average values related to the three specimens tested for

each load condition. In case 0 modality (constant load), the average values refer to the specimens

tested till their breakage (in this case no temperatures higher than 100°C arise), and show that also

the Φ parameter (proportional to the heat quantity to breakage) increases quite linearly with the

frequency, as it happens for the temperatures.

The results for testing modality applied in the case 1 (stepped load), show that for the tested

material, it is not possible to determine a correct value of Eh or Φ. In fact, as highlighted before,

the temperature of the specimen, during the application of the last loads, presents very high values

(see higher curves in Figures 2, 3, 4), moreover reached by unstable trends of temperature

increment T; that is a typical condition where the material changes its microstructure and

mechanical behaviour, and the curve T-N is not sufficiently regular to allow an estimation of

parameter .

19
Table 2. Values of parameter Φ for the different specimens: test frequency 10 Hz

Code Hz Load Φx103 Φx103 average

AISI_304_1 10 stepped 2118.2 1609.1

AISI_304_2 10 stepped 1164.5

AISI_304_3 10 stepped 1544.6

AISI_304_18 10 const (27 kN) 3769.1 4102.6

AISI_304_19 10 const (27 kN) 4406.5

AISI_304_20 10 const (27 kN) 4132.3

Table 3. Values of parameter Φ for the different specimens: test frequency 20 Hz

Code Hz Load Φx103 Φx103 average

AISI_304_4 20 stepped 2609.7 2664.1

AISI_304_5 20 stepped 2595.3

AISI_304_6 20 stepped 2787.3

AISI_304_21 20 const (27 kN) 9990.4 11556.4

AISI_304_22 20 const (27 kN) 14392.0

AISI_304_23 20 const (27 kN) 10287.0

Table 4. Values of parameter Φ for the different specimens: test frequency 30 Hz

Code Hz Load Φx103 Φx103 average

AISI_304_7 30.3 stepped 2697.2 2149.46

AISI_304_8 30.3 stepped 2082.5

AISI_304_9 30.3 stepped 1668.7

AISI_304_24 30.3 const (27 kN) 14709.3 15485.1

AISI_304_25 30.3 const (27 kN) 19811.0

AISI_304_26 30.3 const (27 kN) 11935.1

20
With the Φ average values related to the specimens stressed with a constant load (for which the

stabilization temperature is not higher than 90°C) and for the temperatures (loads) related to the

case 1, the number of the corresponding cycles was calculated. In particular, the Wöhler curves for

the case 1 load conditions have been determined using the average values of Φ for case 0, and the

temperatures of the case 1 related to the load for which they were under 100 C°.

Just as an example, the diagram of Figure 9 report, for the case 1, the above mentioned calculated

points for the three reference frequencies and the related Wöhler curves. In the same diagram the

Wohler curve for a AISI 304 (characterized by ultimate load, yielding load and elongation very

close to the ones obtained by the authors characterizing the material tested in this work), determined

by traditional method [37], is reported (points of TM series in Figure 9). The curves for each

frequency of the diagram are the mean curves obtained using tree specimens for each level load.

The diagram of Figure 9 shows that the fatigue curves, constructed as said before, are affected by

the test frequency. The cycle numbers to breakage are lower as higher is the test frequency (which

means also higher thermal gaps).

This confirms both experimental data reported in literature, which highlight the effect of

temperature increase in terms of reduction of cycles to failure [36], and the considerations

proposed before on the consequences of achieving high temperatures (in these cases also for

load well above the fatigue limit), that can generate modifications of the material

microstructure with a decay in mechanical behaviour, which translates, among other things,

in a reduction of the fatigue resistance and the limit energy to breakage.

5. Conclusions

Fatigue tests were performed on a material with a high heat release to evaluate the influence of the

load frequency on the results of energy-based characterization of fatigue behaviour.

21
For this purpose, the Risitano Method was adopted and the AISI 304 steel has been chosen. This

material, under stress, releases so high quantity of heat, that superficial temperatures of specimens

reach values which in some cases can influence the material structure itself.

The results of performed tests confirm that the fatigue limit is not influenced by the frequency (in

all cases the curve ΔT = f() tends to the value of σ0 for which ΔT = 0, in agreement with the

Risitano Method [20,22]). As far as concerns the load application modality, until the temperature

values are not higher than 200°C, the fatigue curve trend shows a resistance time higher than the

curves which have been obtained using loads histories that cause deterioration to the material.

The load histories which determine high increments in temperature of material have to be carefully

examined in fatigue design of the mechanical components. The phenomena due to high tension

(temperature) values in fact, even for short period of time, can cause important alterations of the

fatigue behaviour.

The energy parameters such as Φ according Risitano Method [20-22] (the same conclusion can be

reasonably extended to other energy parameters, such as the released heat parameter Q introduced

by Meneghetti et al. [10,11], or the fatigue fracture entropy FFE defined by Liakat and Khonsari

[12,13]) are suitable to assess the fatigue behaviour by means of thermal heat release analysis, until

the temperatures do not reach high values that can generate microstructural changes, mechanical

properties decay, and overloads due to thermal stresses.

To avoid these disturbance phenomena, which can alter test or functional conditions, it is

necessary to characterize the materials or to design mechanical components planning

appropriately the fatigue parameters. Particularly, the load frequency values must be adapted

to the load history and the material characteristics, so to reduce heat quantity release and

avoid critical thermal conditions.

22
References

[1] M.P. Luong, Infrared thermographic scanning of fatigue in metals, Nuclear Engineering and

Design 158 (1995) 363-376.

[2] M.P. Luong, Fatigue limit evaluation of metals using an infrared thermographic technique,

Mechanics of Materials 28 (1998) 155-163.

[3] F. Curà, G. Curti, R. Sesana, A new iteration method for the thermographic determination of

fatigue limit in steels, International Journal of Fatigue 27 (2005) 453-459.

[4] M. Amiri, M.M. Khonsari, Life prediction of metals undergoing fatigue load based on

temperature evolution, Materials Science and Engineering A 527 (2010) 1555-1559.

[5] M. Amiri, M.M. Khonsari, Rapid determination of fatigue failure based on temperature

evolution: fully reversed bending load, International Journal of Fatigue 32 (2010) 382-389.

[6] J.L. Fan, X.L. Guo, C.W. Wu, A new application of the infrared thermography for fatigue

evaluation and damage assessment, International Journal of Fatigue 44 (2012) 1-7.

[7] T. Boulanger, A. Chrysochoos, C. Mabru, A. Galtier, Calorimetric analysis of dissipative and

thermoelastic effects associated with the fatigue behavior of steels, International Journal of Fatigue

26 (2004) 221-229.

[8] F. Maquin, F. Pierron, Heat dissipation measurements in low stress cyclic loading of metallic

materials: from internal friction to micro-plasticity, Mechanics of Materials 41 (2009) 928-942.

[9] C. Mareau, D. Cuillerier, F. Morel, Experimental and numerical study of the evolution of stored

and dissipated energies in a medium carbon steel under cyclic loading, Mechanics of Materials 60

(2013) 93-106.

[10] G. Meneghetti, Analysis of the fatigue strength of a stainless steel based on the energy

dissipation, I International Journal of Fatigue 29 (2007) 81-94.

23
[11] G. Meneghetti, M. Ricotta, The use of the specific heat loss to analyse the low- and high-cycle

fatigue behaviour of plain and notched specimens made of a stainless steel, Engineering Fracture

Mechanics 81 (2012) 2-16.

[12] M. Liakat, M.M. Khonsari, Rapid estimation of fatigue entropy and toughness in metals,

Materials & Design 62 (2014) 149-157.

[13] M. Liakat, M.M. Khonsari, Entropic characterization of metal fatigue with stress concentration,

International Journal of Fatigue 70 (2015) 223-234.

[14] Q. Guo, X.L. Guo, J.L. Fan, R. Syed, C.W.Wu, An energy method for rapid evaluation of

high-cycle fatigue parameters based on intrinsic dissipation, International Journal of Fatigue 80

(2015) 136-144.

[15] G. Curti, G. La Rosa, M. Orlando, A. Risitano, Analisi tramite infrarosso termico della

temperatura limite in prove di fatica, XIV Convegno Nazionale AIAS, 1986, p. 211-220.

[16] G. Curti, A. Geraci, A. Risitano, Un nuovo metodo per la determinazione rapida del limite di

fatica, ATA Ingegneria Automobilistica 42 (1989) 634-636.

[17] G. La Rosa, A. Risitano, Thermographic methodology for rapid determination of the fatigue

limit of materials and components, International Journal of Fatigue 22 (2000) 65-73.

[18] R. Botny, J. Kaleta, W. Grzebień, W. Adamczewski, A method for determining the heat energy

of the fatigue process in metals under uniaxial stress: Part I and Part II, International Journal of

Fatigue 8 (1986) 29-38.

[19] A. Geraci, G. La Rosa, A. Risitano A, Determination of the fatigue limit of an austempered

ductile iron using thermal infrared imagery, SPIE International Conference, I995.

[20] G. Fargione, A. Geraci, G. La Rosa, A. Risitano. Rapid determination of the fatigue curve by

the thermographic method, International Journal of Fatigue 24 (2002) 11-19.

[21] G. Fargione, A. Geraci, G. La Rosa, A. Risitano, M. Grech, Determinazione del limite di fatica

in materiali sottoposti a differenti trattamenti termici mediante considerazioni energetiche, XXVI

Convegno Nazionale AIAS, 1997, pp. 1-12.

24
[22] A. Risitano, A. Geraci, G. Fargione, L. Maiolino, Evaluation of the limit energy to failure in

fatigue testing, Atti dell’Accademia Gioena, 2000.

[23] A. Risitano, G. Risitano, Cumulative damage evaluation in multiple cycle fatigue tests taking

into account energy parameters, International Journal of Fatigue 48 (2013) 214-222.

[24] V. Crupi, E. Guglielmino, M. Maestro, A. Marinò, Fatigue analysis of butt welded AH36 steel

joints: thermographic method and design S–N curve, Marine Structures 22 (2009) 373-386.

[25] X.G. Wang, V. Crupi, X.L. Guo, Y.G. Zhao, Quantitative thermographic methodology for

fatigue assessment and stress measurement, International Journal of Fatigue 32 (2010) 1970-1976.

[26] C. Clienti, G. Fargione, G. La Rosa, A. Risitano, G. Risitano, A first approach to the analysis

of fatigue parameters by thermal variations in static tests on plastics, Engineering Fracture

Mechanics 77 (2010) 2158-2167.

[27] S. Stanzl-Tschegg, Very high cycle fatigue measuring techniques, International Journal of

Fatigue 60 (2014) 2-17.

[28] G.L. Taylor, H. Quinney, The latent energy remaining in metal after cold working, Proceedings

of the Royal Society of London Ser. A 143 (1934) 307-326.

[29] C. Bathias, Coupling effect of plasticity, thermal dissipation and metallurgical stability in

ultrasonic fatigue, International Journal of Fatigue 60 (2014) 18-22.

[30] J.R. Davis, ASM Specialty Handbook: Heat-Resistant Materials, American Society for

Metals, 1997.

[31] M.F. McGuire, Stainless Steels for Design Engineers, American Society for Metals, 2008.

[32] A.F. Padilha, P.R. Rios, Decomposition of austenite in austenitic stainless steels, ISIJ

International 42 (2002) 325-337.

[33] Y. Tamarin, Atlas of Stress-Strain Curves, American Society for Metals, 2nd ed., 2002.

[34] L. Gardner, A. Insausti, K.T. Ng, M. Ashraf, Elevated temperature material properties

of stainless steel alloys, Journal of Constructional Steel Research 66 (2010) 634-647.

25
[35] A. Risitano, G. Risitano, Determining fatigue limits with thermal analysis of static

traction tests, Fatigue & Fracture of Engineering Materials & Structures 36 (2013) 631-639.

[36] H.E. Boyer, Atlas of Fatigue Curves, American Society for Metals, 1986.

[37] A. Di Schino , J.M. Kenny, Grain size dependence of the fatigue behaviour of a ultrafine-

grained AISI 304 stainless steel, Materials Letters 57 (2003) 3182-3185.

26
Figures with captions

Figure 1. Geometric and material properties of specimens

Figure 2. Temperature increments on the specimen surface, with the number of the cycles: stepped

load (case 1), test frequency 10 Hz

27
Figure 3. Temperature increments on the specimen surface, with the number of the cycles: stepped

load (case 1), test frequency 20 Hz

Figure 4. Temperature increments on the specimen surface, with the number of the cycles: stepped

load (case 1), test frequency 30 Hz

28
Figure 5. Mean temperature increments on the specimens surface, with the number of the cycles:

constant load (27kN) up to the breakage (case 0), test frequencies 10, 20, 30 Hz

Figure 6. Distribution of fatigue limit data values from Table 1 (mean values, standard errors)

29
Figure 7. Trend of average temperature increments (before the stabilization phase): test frequencies

10, 20, 30 Hz

Figure 8. Trends of the heat quantity released to breakage per unit volume Eh and Eh*: loading

cases 0 and 1, test frequencies 10, 20, 30 Hz

30
Figure 9. Wöhler curves: loading case 1, test frequencies 10, 20, 30 Hz; traditional method, data

from literature [37]

31
The influence of the load frequency on the high cycle fatigue behaviour

Highlights

 Fatigue tests were performed on material with a high heat release.

 Influence of frequency on energy-based fatigue characterization has been evaluated.

 High increments in material temperature can cause alterations of fatigue behavior.

 Fatigue test frequency must be adapted to load history and material characteristics.

32

You might also like