Anda di halaman 1dari 8

International Journal of Fatigue 26 (2004) 849856

Optimisation of the fatigue resistance of 2024-T351 aluminium

alloys by controlled shot peeningmethodology, results
and analysis
C.A. Rodopoulos a,, S.A. Curtis b, E.R. de los Rios b, J. SolisRomero c

Materials Engineering Research Institute, Sheeld Hallam University, City Campus, Howard Street, Sheeld S1 1WB, UK
Structural Integrity Research Institute of the University of Sheeld (SIRIUS), Department of Mechanical Engineering, The University of Sheeld,
Sheeld S1 3JD, UK
Instituto Technologico de Tlalnepantla, Tlanepantla Edo. de Mex., 54070, Mexico
Received 12 June 2003; received in revised form 23 December 2003; accepted 19 January 2004

A methodology dedicated to the optimisation of the fatigue properties of aluminium alloys by controlled shot peening is presented. Selection of the peening conditions is made out of the use of the Design of Experiment and the Eects Neutralisation
Model. Both techniques allowed the optimisation both in terms of life and crack growth rates. Experimental determination and
further analysis of the residual stress relaxation patterns revealed that at high stress levels, low cycle fatigue, life improvement is
predominantly due to slow crack growth rates, while in high cycle fatigue the extension of life is attributed to a prolonged period
of crack arrest.
# 2004 Elsevier Ltd. All rights reserved.
Keywords: Controlled shot peening; Fatigue life; Residual stresses; Surface roughness

1. Introduction
Engineering components and structures are regularly
subjected to cyclic loading and they are consequently
prone to fatigue damage. In most cases, fatigue damage will initiate at the surface due to localised stress
concentrations caused by machining marks, exposed
inclusions or even due to the contrasting movement of
dislocations [1]. Evidently, control over the initiation
and early propagation of surface cracks is paramount
for prolonging the fatigue life of components.
Shot peening is extensively used for the above purpose as it produces near surface plastic deformation
leading to the development of work-hardening and
high magnitude compressive residual stresses [24].
Work hardening is expected to increase the ow resistance of the material and thus reduce crack tip plas
Corresponding author. Tel.: +44-114-225-2457; fax: +44-114-2255390.
E-mail address: (C.A. Rodopoulos).

0142-1123/$ - see front matter # 2004 Elsevier Ltd. All rights reserved.

ticity, while, the residual stresses can act as: a) mean

stress modulatorsin the case of the onset of crack
propagation or b) closure stress in the case of crack
growth [57].
For many years shot peening was commercialised in
terms of the residual stresses. Hence, the numerous
parameters controlling the residual stress prole were
selected in such a way that the prole would reach
maximum levels of depth and magnitude. The above
complexity, which in most cases was treated via the
empirical use of large databases, along with the lack of
a scientic quantication of the eect of the peening
parameters on fatigue, have caused many unexpected
results, including degradation of the fatigue properties
of the material. In an attempt to minimise and therefore better control the reproducibility of shot peening
(controlled shot peening, CSP), Eckersley [8] suggested
the close monitoring of: a) the peening media; b) the
energy of the shot stream (intensity) and c) the percentage of indentation (coverage).


C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856



Notch depth
Normalised notch depth
Notch half width
Normalised notch half width
Far-eld applied stress
Crack closure stress
Dynamic crack closure stress
Flow stress
Crack arrest stress
Dynamic crack arrest stress
Endurance limit
Crack length
Average transverse grain size
Number of half grains
Elastic stress concentration
Grain orientation factor
Vallelanos notch fatigue parameter

Fig. 1.

Dimensions (in mm) and design of the dogbone specimen [12].


Based on the above and considering the lack of

enough information regarding the eect of shot peening on aluminium alloys [9], SIRIIUS, Airbus UK and
the Royal Academy of Engineering joined on a major
research eort to provide a clear denition of the
optimum levels and tolerances of the peening process
for a number of aerospace components. The work
represents part of this endeavour and is dedicated to
the optimisation of the 2024-T351 aluminium alloy.
2. The material, the peening parameters and fatigue
2.1. The material
Material, supplied by Airbus UK, was received in
the form of a 30.0 mm thick rolled plate. A full chemical composition and mechanical properties are given in
Tables 1 and 2, respectively. The material was solution
Table 1
Chemical composition of 2024-T351 in wt% [10]


















heat-treated at 435 C and naturally aged to a substantially stable condition. Microstructural analysis according to ASTM E112 [10] revealed that the material has a
pancake shape grain structure with an average grain
size of 220, 80 and 52 lm in the longitudinal, transverse and thickness direction, respectively.
Prior to the machining of the fatigue specimens, a
10 mm layer was removed from each side of the plate
by low speed milling. This was done to avoid contamination of the specimens from loose or broken inclusions, porosity, clusters etc due to hot rolling. The
dogbone specimens (Fig. 1) were milled to an aerospace standard of average surface roughness
Ra  0:8 lm with the crack plane surface being perpendicular to the rolling direction. All tests were carried
out using specimens oriented with the tension axis parallel to the rolling direction and the fatigue surface
perpendicular to the rolling plane (PRP).
2.2. The peening conditions
The CSP conditions were selected through an experimental database created during the course of this programme. The database [13] contains data of residual
stress proles, work-hardening proles and surface
roughness measurements from a selection of shot
media, coverage and intensity combinations. Two
dierent approaches for the selection of the peening
conditions were used, the Design of Experiments
(DoE)/Robust Design Methodology and the Eects
Neutralisation Model (ENM).
The DoE is an experimental approach to studying
the interaction eects between two or more factors in
an experimental programme known as factorial design
[1316]. In brief, the steps and order characterising the
DoE are: a) designation of critical factors aecting a

Table 2
Mechanical and physical properties of 2024-T351 at room temperature [10,11]
Elastic modulus (GPa)
Shear modulus (GPa)
Poissons ratio
Mon. yield stress (MPa)
Cyclic yield stress (MPa)


Hardness (HB)
Fr. toughness (MPam1/2)
Ult. tensile strength (MPa)
Endurance limit (R=1)
Shear strength (MPa)


C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856

dependent variable, i.e. coverage, shot type, angle of

impact on surface roughness; b) designation of an
impact level charactering each of the critical factors.
The level range (c  1) along with the number of critical factor; c) dene the number of necessary tests in a
full factorial experiment (bc); d) performance of experiments with a certain number of replications. The accuracy of the DoE strongly depends on the experimental
error or noise which is found to decrease with the
number of experiments; e) introduction of a predictive
model which relates the critical factors and dependent
variables to a certain fatigue property, i.e. life and f)
analysis of variance (ANOVA) to evaluate the impact
of each factor on the system. The method can provide
accurate statistical projections, especially when dealing
with vast amounts of data. However, it is important to
note that the designation of the impact factor requires
specic experimental arrangements to provide a
realistic grading. For example, the eect of the surface
roughness on the fatigue properties of a component
varies with the mechanical properties and the microstructure of the material.
The ENM is based on a theoretical model developed
by the authors in order to draw a boundary condition
which would allow the neutralisation of the detrimental
eects of CSP (surface roughness) by the benecial
(work-hardening and residual stress prole) in terms of
the fatigue life. The proposed methodology, a detailed
analysis is given in [17], predicts a crack closure stress
prole which fully neutralises the eect of the stress
gradient provided by the surface roughness and thus
delivering identical fatigue lives to the ground and peened material. The ENM is later plotted against the
experimental closure stress proles from the database
to provide solution by comparison. Such a comparison
plot is shown in Fig. 2. Cases where the comparing
closure stress proles intercept are considered as
unacceptable. In this work, only a brief description of
the ENM would be given.
Li et al. [18] suggested that the elastic stress concentration due to the roughening of the surface by CSP is
given by,

Kt 1 2:1
where Rtm is the mean to peak valley heights and Sm is
the spacing of adjacent peaks in the roughness prole.
Eq. (1), in terms of a single geometry notch can be
rewritten as,
Kt 1 2:1
where a is the notch depth and b is the notch half
width. In [19] Vallelano et al. described the eect of a
stress concentration on fatigue damage through a para-


Fig. 2. Comparison plots between the ENM predictions and experiv

mental data for a) S230/90 /100% and b) S110/90o/100%. The peening designation follows the order shot type and size/angle of impact/

meter Zi,

1 1

31 =2

i 6b
4 q 5

a b ki

and the parameter ki is,


2 =22 a2 b2  ba iD=2

a2  b 2
The parameter i represents the number of half grains
which constitute the fatigue damage (crack and crack
 are dimensionless forms
tip plastic zone). a and b
of the notch depth and half width respectively, i.e.
 2b=D, and D is the average grain
a 2a=D and b
size. Using Eq. (3) and the NavarroRios model for
microstructural crack growth [20], Curtis et al. [17]
proposed that the necessary closure stress (representing
the eect of the residual stress prole on the crack
wake) able to fully neutralised the eect of surface


C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856

roughness is given by,

r2 i1  Zi r

Zi mi =m1 rFL ir2  r

where mi/m1 is the grain orientation factor, r2 is the

ow resistance of the material and rFL is the fatigue
limit of the as-received material. A typical grain orientation distribution for aluminium alloys is [21],
1 0:35lni
Assuming small scale yielding conditions, the parameter i can be expressed in terms of crack length
through [22],

Fig. 4. Distribution of microhardenss with depth for the two

optimum CSP conditions.

The relationship between the closure stress and the

residual stress is given by [6,17],

1 a
f Residual stress da
a 0
Each approach concluded into dierent optimum peening conditions. The DoE suggested: a steel shot desigv
nated as S110 ( 0.279 mm), angle of impact of 45
and a coverage rate of 200%. The ENM concluded that
the optimum conditions are: S110, angle of impact of
90 and a coverage rate of 100%. Prior to peening, the
peening machine was calibrated in terms of ow rate
for the selected shot. The input parameters were: pressure 345 MPa, nozzle stand-o distance 152.4 mm, ow
rate 4.54 Kgr/min and Vernier value 60%. According
to the calibration, the 100% coverage requires two
passes of the nozzle while the 200% requires four
passes. Figs. 3 and 4 show the residual stress and
microhardenss prole of the selected peening parameters. Residual stress measurements were performed

according to the incremental hole drilling E837-95 [22].

Table 3 shows roughness measurements taken from the
peened surfaces using the Mitytoyo SV stylus prolometer. The detrimental eect of shot peening on surface
roughness can be appreciated by comparing the roughness proles depicted in Fig. 5. The results indicate that
the ENM shows a superior residual stress prole compared to DoE. On the other hand, the DoE exhibits
better work-hardening than the ENM. Work-hardening
has been found to be mainly aected by the angle and
coverage rate [13]. Both conditions showed similar
values of elastic stress concentration due to surface
roughening (Table 3).
In addition to residual stress, work hardening and
surface roughening, shot peening has also been reported to produce distortion of the near surface microstructure. In [23], de los Rios et al. reported a decrease
in the average grain size of the near surface grains in
the region of 20%. In theory, such distortions could
increase the frequency of the retardation events experienced by the growing crack due to the grain boundary
plasticity constrain eect and hence to result in a
slower propagation rate. However, the eects of such
phenomenon on fatigue damage can be more complex
since a decrease in the grain size can: a) increase the
notch sensitivity of the material (see Eq. (3)) and b) be
associated to near surface loss of ductility [24]. In this
work and for both techniques, the eects of microstructural distortions on fatigue damage are neglected due
to insucient information.
Table 3
Critical roughness parameters and corresponding Kt predictions

Fig. 3. Distribution of residual stress versus depth for the two optimum CSP conditions. The residual stress of the as-machined material
is also depicted for comparison.

CSP Condition

Ra (lm)

Rtm (lm)

Sm (mm)







C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856

Fig. 5.


Roughness proles taken from identical roughness proling conditions for a) as-machined and b) peened coupons (S110/45 /200%).

2.3. Fatigue data

Fatigue testing was performed under load control in
an Instron 8501 100 KN unit equipped with hydraulic
grips. The testing frequency was 20 Hz, the minimum
to maximum load ratio was set at 0.1 and the waveform followed a sinusoidal pattern. The results, in
terms of an S-N curve are shown in Fig. 6. A number
of fatigue tests for the determination of crack growth
rates was also performed. Crack length was measured
using acetone replicas. The replicas were later analysed
using an image analysis software. Typical results are
depicted in Fig. 7.

S-N results showed that both peening conditions

improve the fatigue life of the material, with the ENM
exhibiting a signicant supremacy. This is especially the
case in the area of low cycle fatigue, where the DoE
condition showed not to aect the performance of the
as-machined specimens. Fig. 7 reveals that almost 90%
of the life improvement is attributed to a prolonged
period (between 400 and 450 K cycles) of arrest of a
pre-existing fatigue crack or a crack-like defect. It
should be noted that such a nding is not conclusive,
since the propagation behaviour of the crack below the
~15 lm mark is not known. However, considering the
accuracy of even the most modern non-destructive

Fig. 6. S-N response of the as-machined and the two selected peening conditions. The arrows indicate run-outs.

Fig. 7. Crack length vs number of cycles for as-machined, DoE and

ENM specimens. The peak stress was 270 MPa.


C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856

Fig. 8. Crack growth rates for the as-received, DoE and ENM
conditions. The peak stress was 270 MPa.

techniques the above could be considered as engineering sound. In addition, examination of the crack length
distribution versus the corresponding number of cycles
reveals that the propagation of the DoE and ENM is
slightly slower to that of the as-machined specimens.
The above can be representative of the crack closure
eect provided by the residual stresses on crack propagation (see Fig. 8). The propagation rates for the DoE
and ENM appear to follow the residual stress prole
exhibit as a result of the residual stress prole with
their peak to be found around the 200 lm depth.
3. Analysis of fatigue data and discussion
From the previous analysis it is clear that the eect
of CSP on the fatigue behaviour of the 2024-T351 can
be attributed to a prolonged period of crack arrest as
well as slower growth rates. Thus, in order to incorporate the residual stress prole to damage tolerant design
a distinction between crack propagation and arrest is

Fig. 9. (a) Residual stress relaxation patterns for DoE at a

maximum stress of 240 MPa; (b) Residual stress relaxation patterns
for DoE at a maximum stress of 270 MPa.

gradual relaxation with life, while, the ENM is more

sensitive to the rst 25% of loading cycles and e) ENM
proles appear to maintain a more regular residual
stress prole with loading history. The latter can provide an explanation of the slower propagation rate
shown by the ENM.

3.1. Residual stress relaxation

3.2. Analysing the life improvement
To examine the stability of the residual stress prole
with the loading history, incremental hole drilling measurements were taken at 25% life intervals for a series
of test specimens. Some of the results are shown in
Figs. 9 and 10.
Examination of the relaxation data reveals that: a)
the relaxation pattern is faster and more severe at high
stress levels; b) the loading history appears to aect the
magnitude of the residual stresses while the basic shape
of the prole is left unaected. Discrepancies can be
found at very high stress levers (in Fig. 10b the relaxation has driven the point of maximum residual stress
from 0.2 to approximately 0.3 mm); c) in both cases
the residual stress prole is unlikely to reach saturation
within 75% of life; d) DoE proles seem to follow a

In [17] it was suggested that the eect of CSP on the

crack arrest behaviour of the material can be assessed
by introducing the relaxation dynamics into the
KitagawaTakahashi diagram,

mi rFL
p ri1
rarrest Zi
m1 i

where in this case ri1 represents the dynamic crack

closure stress, rarrest is the dynamic maximum applied
stress that allows the arrest of a crack of length a and
i 2a=D considering negligible plasticity. Hence, to
evaluate the loading history prior to the propagation of
the catastrophic crack, Eq. (7) should be set equal to the
applied stress. Typical results are depicted in Fig. 11.

C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856


Fig. 12. Life improvement due to crack arrest (NI) normalised over
the total life (NT) for ENM peening conditions.

Fig. 10. (a) Residual stress relaxation patterns for ENM at a

maximum stress of 260 MPa; (b) Residual stress relaxation patterns
for ENM at a maximum stress of 330 MPa.

From Fig. 11 it is clear that the applied stress for

crack lengths denoted by the typical crack like defect,
drops below the crack arrest stress after approximately

50% of the total life (~615 K cycles). It is also important to note that at 75% of life, the crack arrest
capacity drops below that of the as-machined material.
The above manifests that the residual closure stress
prole is lesser than that needed to counterbalance the
eect of the surface roughness. Such behaviour can
explain the fact that the improvement caused by the
CSP on the endurance limit (in this work set at 7 M
cycles) of the material is not representative to the
residual stresses and therefore the current practice of
using the residual stress prole as a far-eld stress
modulator can lead to erroneous results. In addition,
the use of CSP to improve the fatigue limit of the
material (set at >10 M cycles) is questionable since the
relaxation of the residual stress prole shows a nonsaturation behaviour.
Using the methodology depicted in Fig. 11, predictions can be performed regarding the classication of
the life improvement to that due to the crack arrest
and crack propagation. Such a prediction is shown in
Fig. 12.
Fig. 12 indicates that the life improvement due to
crack arrest decreases with the applied stress. The
behaviour is representative of the eect of the applied
stress to control the residual stress relaxation prole.
4. Conclusions
From this work the following conclusions can be
drawn for the fatigue response of the 2024-T351 to

Fig. 11. The eect of residual stress relaxation on the Kitagawa

Takahashi diagram for peening according to ENM at 260 MPa. The
vertical line at 12 lm indicates the size of maximum internal defects
found in the aluminium plate [13].

. Both peening condition (ENM and DoE) resulted

into life improvement.
. Life improvement due to CSP is caused by the dual
eect provided by prolonging crack arrest and
reducing the crack growth rate with both mechanisms being aected by the residual stress relaxation


C.A. Rodopoulos et al. / International Journal of Fatigue 26 (2004) 849856

. The relaxation of the residual stress prole is of primary importance and should be included into damage tolerance methodologies.
. The relaxation of the residual stresses seems to
depend on the applied stress while the basic residual
stress prole remains unaected.
. Severe residual stress relaxation can cause the degradation of the fatigue performance of the material.
. The use of the residual stresses as far-eld stress
modulator according to the popular Dre
(maximumresidual stress) should be avoided since it
will overestimate the true eect of the residual stress.
. The use of CSP to improve the endurance limit
should be accompanied by extensive and thorough
analysis of the dynamics of the residual stress prole.

The authors gratefully acknowledge the nancial support from Airbus UK, the EPSRC, the Royal Academy
of Engineering and ConacyT. Special merit should also
be given to Mr. J. W. Eichler and Mr. J. M. Ordieres for
performing some of the experimental work as well as to
Mr. J. Goodllife for his technical support.

[1] Miller KJ. Materials science perspective of metals fatigue resistance. Mater Sci Technol 1993;9:45362.
[2] Dorr T, Hilpert M, Beckmerhagen P, Kiefer A, Wagner L. Inuence of shot peening on fatigue performance of high strength
aluminium and magnesium alloys. In: 7th International Conference on Shot Peening, Warsaw, 1999. p. 15360.
[3] Martin U, Altenberger I, Scholtes B, Kremmer K, Oettel H.
Cyclic deformation and near surface microstrutures of normalised shot peened steel SAE 1045. Mater Sci Engng A 1998;246:
[4] Romero Solis J, de los Rios ER, Fam HY, Levers A. Optimisation of the shot peening process in terms of fatigue resistance.
In: 7th International Conference on Shot Peening, Warsaw,
1999. p. 11726.
[5] Jaensson B. The inuence of the fatigue strength of aluminium
alloy parts of the relationship between the surface residual stress
state and the load induces stress state. In: 1st International Conference on Shot Peening, Paris, 1981. p. 43544.
[6] de los Rios ER, Trull M, Levers A. Modelling fatigue crack
growth in shot peened components of 2024-T351. Fatigue Fract
Engng Mater Struct 2000;23:70916.

[7] Ohji K, Kubo S, Tsuji M, Ogawa H, Sakurai K. Methods of

predicting fatigue crack growth lives in residual stress elds.
Trans JSME 1992;53(492):151624.
[8] Eckersley JS. Shot peening process controls ensure repeatable
results. In: 2nd International Conference on Shot Peening,
Chicago, 1984.
[9] Sharp PK, Clayton JQ, Clark G. The fatigue resistance of peened 7075-T7451 aluminium alloy: repair and re-treatment of the
component surface. Fatigue Fract Engng Mater Struct 1994;19:
[10] Boyer HE, Gail TL, editors. Metals handbook. 9th ed. Metals
Park, OH: American Society of Metals; 1997.
[11] Callister Jr WD. Materials science and engineeringan introduction. 2nd ed. USA: Wiley; 1991.
[12] Airbus Industrie Test Method, (AITM 1-0011) Constant amplitude fatigue testing of metallic materials, AIRBUS INDUSTRIE,
Editor. Airbus Industrie, Engineering Directorate, Blagnac cedex,
France, 1996. p. 124.
[13] Romero Solis J. Optimisation of the shot peening process in
terms of fatigue resistance. PhD Thesis, The University of
Sheeld, 2002.
[14] Fisher RA. The design of experiments. 9th ed. New York: Hafner; 1974.
[15] Kackar RN. O-line quality control, parameter design and the
tagushi method. J Quality Technol 1985;17(4):17688.
[16] Baragetti S. Shot peening optimisation by means of DOE:
Numerical simulation and choice of treatment parameters. Int J
Mater Prod Technol 1997;12:83109.
[17] Curtis S, de los Rios ER, Rodopoulos CA, Levers A. Analysis
of the eects of controlled shot peening on fatigue damage of
high strength aluminium alloys. Int J Fatigue 2003;25:5966.
[18] Li JK, Yao M, Wang D, Wang R. An analysis of stress concentrations caused by shot peening and its application in predicting fatigue strength. Fatigue Fract Engng Mat Struct 1992;15(12):
[19] Vallellano C, Navarro A, Domnguez J. Fatigue crack growth
threshold conditions at notches. Part I: Theory. Fatigue Fract
Engng Mater Struct 2000;23:11321.
[20] Navarro A, de los Rios ER. Fatigue crack growth modelling by
successive blocking of dislocations. Proc Royal Soc London
[21] Curtis SA, Romero Solis J, de los Rios ER, Rodopoulos CA,
Levers A. Predicting the interfaces between fatigue crack growth
regimes in 7150-T651 aluminium alloy using the fatigue damage
map. Mater Sci Engng 2003;A344:7985.
[22] E837-95, ASTM. Standard test method for determining residual
stresses by the hole-drilling strain-gage method. Annual book of
ASTM standards, Vol. 03.01. Philadelphia: ASTM; 1985. p.
[23] de los Rios ER, Trooll M, Levers A. Improving the fatigue
crack resistance of 2024-T351 aluminium alloy by shot
peening. In: Proc On Life Extensions/Technology-Opportunities,
London, 1999. p. 2618.
[24] Rodopoulos CA, Romero JS, Curtis SA, de los Rios ER, Peyre
P. Eect of controlled shot peening and laser shock peening on
the fatigue performance of 2024-T351 aluminium alloy. J Mater
Engng Perform 2003;12(4):4149.