FATIGUE BEHAVIOUR OF AA6082-T6 ALUMINIUM ALLOY FRICTION STIR WELDS UNDER

VARIABLE AMPLITUDE LOADING


J.D. Costa
1
, J.A.M. Ferreira
1
and L.P. Borrego
2


1
CEMUC, Mechanical Engineering Department,
University of Coimbra, Rua Luís Reis Santos, Pinhal de Marrocos, 3030-788, Coimbra, Portugal.
E-mail: jose.domingos@dem.uc.pt.; martins.ferreira@dem.uc.pt.

2
CEMUC, Mechanical Engineering Department,
Polytechnic Institute of Coimbra, Quinta da Nora - 3030, Coimbra, Portugal.
E-mail: borrego@isec.pt.


ABSTRACT

In addition to uncertainty such as material strength, notch geometries, defect contents and residual stresses, welded
components are often subjected to variable amplitude service loads. In the case of friction stir welding of aluminium
alloys, no data is available concerning fatigue behaviour under variable amplitude loading. The objective of this
investigation is to determine the fatigue strength of friction stir welds in AA6082-T6 under constant and variable
amplitude loading and analyse the validity of Miners´s rule in this specific welding process. Fatigue tests were carried
out in a servo-hydraulic testing machine using stress ratios of R = 0 and R=-1. Typified Gassner amplitude spectra were
considered, using four values for the shape exponent. Microhardness tests were performed to characterize the Vickers
hardness profile in the vicinity of the weld area. Friction stir welding process leads to a decrease of the static
mechanical properties relatively to base material. Detailed examination revealed a hardness decrease in the thermo
mechanically affected zone and the nugget zone average hardness was found to be significantly lower than the base
alloy hardness. Welded specimens show significantly lower lives than base material specimens. For the welded
specimens tested at a stress ratio R=0 a good agreement was observed between constant and variable fatigue loading,
using the equivalent stress calculated by Miner´s rule. For R=-1, Miner´s rule seems to overestimate the fatigue life.


KEY WORDS: Friction Stir Welding, aluminium alloy, fatigue, variable amplitude loading.


1. INTRODUCTION

Friction stir welding (FSW) is a relatively new solid-
state joining process and is very energy efficient,
environment friendly, and versatile, being considered to
be the most significant development in metal joining in
a decade. Since its invention in 1991 at the Welding
Institute (TWI) of UK [1], a large amount of research
was carried out in several fields and different materials.
Aluminium alloys are the materials more often studied
and where this technique has shown a better
performance. Comparative mechanical properties
studies of base material and welded specimens,
including fatigue strength tests have been performed by
several authors [2-6].

In addition to uncertainty such as material strength,
notch geometries, defect contents and residual stresses,
welded components are often subjected to variable
amplitude service loads. The lack of Miner´s validity
accumulation rule [7] has been demonstrated in many
applications and, in consequence, its usage will
introduce uncertainties which must be compensated by
safety factors. In the case of friction stir welding (FSW)
no data is available concerning fatigue behaviour under
variable amplitude loading

The objective of this work is to study the fatigue
strength of friction stir welds under constant and
variable amplitude loading and analyse the validity of
Miners´s rule for this specific welding process.


2. EXPERIMENTAL DETAILS

2.1 Preparation of welds

This research was conducted using the AA6082
aluminium alloy with a T6 heat treatment. The T6 heat
treatment corresponds to a conversion of heat-treatable
material to the age-hardened condition by solution
treatment, quenching and artificial age-hardening. The
alloy chemical composition and mechanical properties
are shown in Tables 1 and 2, respectively.

The friction stir welds were performed in an aluminium
plate with 4 mm thickness using a tool with a 5 mm
diameter threaded pin and the shoulder had 16 mm
diameter.

Table 1
Si Mg
1.05 0.8

Table 2
Tensile stre
Yield streng
Elongation,
Hardness, H

Welding par
alloy type, p
parameters w
good surface
identified by
300 mm/min
rpm. Figure
welds concer
also a good
obtained. Th
excessive pe
could affect m

Figure 1

The penetrat
butt joint. Fr
the plate we
order to exc
due to the t
start. In orde
tunnel or kiss
cross-section
analysis in pl
were also p
accordingly
etched in o
different wel
poor quality
size of abo
parameters o
cross sectio
parameters
observed by
tend to appea
zone near th
- Chemical co
aluminium
Mn Fe
0.68 0.26
- Mechanical
alumin
ength, σ
UTS
(M
gth, σ
YS
(MPa
, ε
r
(%)
Hv,02
ameters depen
penetration de
were chosen in
e aspect and
y microscopic
n, tilt angle o
1 is an exam
rning its surfa
d penetration
his is a very
enetration wi
mainly the fat
1. Surface asp
tion depth wa
rom each wel
ere not used
clude possible
typical lower
er to analyse t
sing bond def
ning of the
lanes perpend
performed. T
to standard
order to enab
ld zones. Figu
y welded plat
out 0,6 mm
optimization,
on of a wel
above referr
microscopy o
ar in the advan
he thermo-me
omposition of
m alloy (wt%)
Cr Cu
0.01 0.04 0
l properties of
nium alloy
MPa)
a)
nd on several
epth and join
n order to obt
d where no d
observation: w
of 2º; rotating
mple of the b
ace aspect. In t
control of th
important asp
ll create a n
tigue strength.
pect of a frictio
as adapted to
lded plate, the
in specimens
e deviation fr
temperatures
that no defect
fects were pres
welds for
dicular to the w
The samples
metallograph
ble the ident
ure 2a shows
te, presenting
m, obtained
while figure
lded plate w
ed, where n
observation. T
ncing side of
echanically af
f AA6082-T6
)
Zn Ti Ot
0.02 0.01 0
f AA6082-T6
300
245
9
110
l factors, nam
nt. The follow
tain welds wit
defects could
welding speed
g speed of 1
better succee
this weld samp
he shoulder w
pect, because
notch effect t
.
on stir weld.
fully penetra
e first 40 mm
s preparation
rom steady s
s during weld
ts like root fla
sent in the we
metallograp
welding direct
were prepa
hic practice
tification of
an example o
a tunnel de
before weld
e 2b shows
welded with
no defects w
The tunnel defe
the nugget low
ffected zone
ther
.05
mely,
wing
th a
be
d of
500
eded
mple,
was
e an
that

ated
m of
, in
tate
ding
aws,
elds,
phic
tion
ared
and
the
of a
fect
ding
the
the
were
fects
wer
and
the
siz
we
ach


F

Th
lea
in
tun
wh
fat
avo
a h
lea
cre
liv
cor
avo
we

2.2

Fat
tes
loa
160
spe
sur
the

F

Sp
ide
μm
on
stre
dir
wit
wa
stre
(
(
ey are continu
ze of these def
elding proces
hieved.
Figure 2. a) w
weld withou
he majority of
ads to the pre
the retractin
nnel defects o
hich did not in
igue resistanc
oid tunnel def
higher axial fo
ads the forma
eating stress c
e. It was real
rrect optimiza
oid the prese
ere performed
2 Fatigue deta
tigue tests w
sting machine
adings. The di
0x15x4 mm
3
ecimens (Fig.
rface preparat
e thickness exc
Figure 3. Photo
pla
ecimens mac
entified, typic
m, were also te
fatigue resis
ess axis in th
rection. A sin
th the stress
as in the inte
ess level. Life
(a)
(b)
uously formed
fects tends to
s due to low
weld presenting
ut defects iden
f welds perform
sence of a tun
ng side. Alth
observed had a
nfluence static
ce was signific
fects is increas
orce. Howeve
ation of shear
concentration
lized to be ve
ation of all
ence of the tu
with two pass
ails
were carried
under consta
imensions of t
3
(length, wi
3) were mill
tion was mad
ceeding mater
o of a fatigue
ate before edg
chined of pl
cally tunnel d
ested in order
stance. The w
he S–N tests a
nusoidal load
ratio R set to
rval of 20–40
e was defined
d in the weld d
be great at th
wer temperatu
g a defect typ
tifiable by mic
med with only
nnel defect fo
hough, in som
a size smaller
c mechanical p
cantly affected
sing heat inpu
er, an excess o
lips in the u
which also re
ery difficult to
parameters. T
unnel defects
sages.
out in a se
ant and variab
the fatigue sp
idth, thickne
led at the edg
de only by gri
rial.
specimen cut f
ges machining.
lates where
defects with l
r to evaluate th
weld was tran
and to the ma
d–time functio
o 0 and -1. T
0 Hz depend
as the numbe
direction. The
he start of the
ure generally
e tunnel; b)
croscopy.
y one passage
ormed mainly
me cases the
r than 30 μm,
properties, the
d. One way to
ut by applying
of axial force
upper surface,
educes fatigue
o perform the
Therefore, to
s some welds
ervo-hydraulic
ble amplitude
ecimens were
ss). The cut
ges, and weld
ind removing
from welded
.
defects were
less than 200
heir influence
nsverse to the
aterial rolling
on was used,
The frequency
ding upon the
er of cycles to
e
e
y


e
y
e
,
e
o
g
e
,
e
e
o
s
c
e
e
t
d
g

e
0
e
e
g
,
y
e
o
failure and a total of 57 specimens were tested. For the
variable loading amplitude tests, typified amplitude
spectra [8], according to equation 1, were considered,
using four values for the shape exponent ν, namely, 1,5,
2, 4 and 5,

logE
ì
= _1 -_
S
c,i
S
c,mcx
]
v
_ logE
0
(1)

where
H
i
cumulative frequency of load cycles for level S
a,i

H
0
block size (number of cycles)
ν shape exponent
 
Equation 2 specifies an additional number, the
‘spectrum shape factor’, SSF, that represents the
distance (factor, life ratio) between an (arbitrary)
Wohler S–N curve and the associated spectrum fatigue
life curve (in Germany sometimes referred to as Gassner
curve in honour of Ernst Gassner). 

SSF = log
∑n
i
∑n
i
_
S
c,i
S
c,mcx
]
m
(2)

Using the loading spectrum defined by equation 1, SSF
can be more precisely calculated using the following
equation

SSF = log
H
c
]
d
dx
j- H
c
(1-(x)
u
)
[·(x)
m
dx
1
0
(3)
where, x =
S
c
S
c,mcx
(4)

SSF depends on the shape of the spectrum under
consideration (parameters H
0
and ν) and the slope, m, of
Wohler´s curve. Table 3 present SSF values for some ν
values and a description of its typical applications. This
factor was calculated using the Miner damage rule
(without endurance limit), taking H
0
=10
6
cycles and
m=5,5

Table 3. Spectrum shape parameters ν and SSF for
typified amplitude spectra: H
0
=10
6
; m=5,5.

v SSF
Description
∞ 0 Constant amplitude loading
5 1,30
4 1,56
ν>2 typical of bridge and crane
structures
2 2,61 Stationary Gaussian process
1,5 3,14
1 3,95
Typical for road roughness induced
loads
0,8 4,39
ν<1 typical for wind gusts, wave
actions, etc.
Figures 4 and 5 show the several variable loading
amplitude spectra applied during fatigue tests. Figure 4
is the typical form of spectrum presentation where the
normalized stress amplitude is plotted against the
cumulative number of cycles, while in figure 5 a more
intuitive way for spectrum interpretation is shown,
depicting the number of cycles versus the normalized
stress amplitude.
 
Figure 4. Typified amplitude spectra: normalized stress
amplitude versus cumulative number of cycles
(H
0
=10
6
cycles).
 
Figure 5. Typified amplitude spectra: number of cycles
versus the normalized stress amplitude.

2.3 Microhardness details

Microhardness tests were performed to characterize the
Vickers hardness profile in the vicinity of the weld area
using a 2 N load. Measurements were performed along
three lines: 0,5 mm far from both surfaces and at the
0
0,2
0,4
0,6
0,8
1
1E0 1E2 1E4 1E6
S
a
/
S
a
,
m
a
x
Cumulative number of cycles, H
ν=0,8
ν=1
ν=1,5
ν=2
ν=4
ν=5
ν= 

0,0E0
2,0E4
4,0E4
6,0E4
8,0E4
1,0E5
1,2E5
1,4E5
1,6E5
1,8E5
2,0E5
0,0 0,2 0,4 0,6 0,8 1,0
N
u
m
b
e
r

o
f

c
y
c
l
e
s
Sa/Sa,max
ν=1
ν=1,5
ν=2
ν=4
ν=5
Typified amplitude spectra
(H
o
=10
6
cycles):
ν=

specimen middle thickness, in successive positions with
1 mm of distance.

Figure 6. Hardness profiles measured along three lines.

This figure shows clearly a hardness decrease in the
thermo mechanically affected zone and that the nugget
zone average hardness is significantly lower than the
base alloy hardness. Hardness profiles were measured in
all plates. Microstructures hardness of the welds
typically ranged between 70 and 80 HV
0,2
, leading to a
softening up to 60% relatively to the base material
hardness (about 115 HV
0,2
). The results show similar
hardness profiles for the three measurement lines,
although bottom measurements show that the lower
hardness is restricted to a shorter zone.


3. RESULTS

Table 4 shows the mechanical properties of welded
specimens obtained in tension tests. Compared with
base material (table 2), friction stir welding process
leads to a decrease of the material mechanical
properties: yield and rupture stresses of friction welded
specimens are significantly lower than for base material.
Yield stress decreases 33% and the rupture stress
decreases 20% relatively to base material values.

Table 4 - Mechanical properties of friction stir welds
performed in AA6082-T6 aluminium alloy.
Tensile strength, σ
UTS
(MPa) 241
Yield strength, σ
YS
(MPa) 165
Elongation, ε
r
(%) 6,8

Fatigue results obtained for the stress ratio R=0 under
constant and variable amplitude loadings are plotted in
figure 7. Fatigue data for constant amplitude loading
were statistically analysed accordingly ASTM E739-91
Standard [9]. A low dispersion of fatigue data was
obtained: the standard deviation was 0,04 for stress and
0,21 for fatigue life. This is a lower dispersion when
compared with fatigue data obtained in tests of
specimens welded by other processes. Fatigue data
obtained with variable amplitude loading, using the
equivalent stress range calculated accordingly to
Miner´s rule, are in the same scatter band of the
constant amplitude loading results, indicating that
Miner´s rule gives real damage sum values near unity.
The damage sum D ranged between 0,6 and 2 for the
four spectra loadings. This conclusion is better visible in
Figure 8 where the probability of occurrence is plotted
against real damage sum values.


Figure 7. Wohler curves for R=0. Constant and
variable amplitude loadings.

 
Figure 8. Real damage sum distribution.
50
60
70
80
90
100
110
120
130
-20 -15 -10 -5 0 5 10 15 20
H
a
r
d
n
e
s
s

H
v
0
,
2
Distance from weld centre (mm)
top
middle
bottom
50
100
200
1E4 1E5 1E6 1E7
S
t
r
e
s
s

r
a
n
g
e
,

Δ
σ
[
M
P
a
]
Number of cycles, N
f
R=0; CA
R=0; VA
95% confidence bands for
the median S-N curve
S
N
=0.21
S
σ
=0.04
R
2
=0.87
B
m
=-5,52
B
Stdv
=0,35
t
p
B
Stdv
=0,71
B
lower
=-6,22
B
upper
=-4,81
A
m
=16,87
A
Stdv
=0,72
t
p
A
Stdv
=1,45
A
lower
=15,41
A
upper
=18,32
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0,5 1 1,5 2 2,5
P
r
o
b
a
b
i
l
i
t
y

Damage sum
Figure 9 compares fatigue results obtained with constant
and variable amplitude loadings for the stress ratios R=0
and R=-1, plotting stress range against the number of
cycles to failure. Data for variable amplitude loading and
R=-1 were restricted to the shape exponent ν=2.


Figure 9. Wohler curves for R=0 and R=-1. Constant
and variable amplitude loadings.

As expected, a significant influence of the mean stress
was observed: for the same stress range, specimens
tested with R=-1 present higher lives than specimens
tested with R=0. For R=-1 fatigue data obtained under
variable amplitude present lower lives than under
constant amplitude loading, indicating that damage sum
values are lower than unity for the majority of the tests.
Therefore, Miner´s rule seems to overestimate the
fatigue life in the case of the stress ratio of R=-1. More
tests are in course under variable amplitude loading for
R=-1 and other spectrum shape exponents in order to
confirm this trend.

Figure 10 shows the Wohler curve (stress range versus
life) and Gassner curves (spectrum maximum stress
range versus life) for R=0 and for several values of the
spectrum shape exponent, ν. The graphical meaning of
the SSF parameter is also indicated. It is clear that as v
increases the Gassner curve becomes more close to
Wohler curve. As referred before SSF=1 when ν=∞.

Figure 11 compares fatigue results obtained in this work
for R=0 under constant amplitude loading with results
obtained by other authors [1,3] in welded specimens of
the same alloy. S-N curves obtained with specimens
welded by FSW, MIG and TIG processes as well as
base material are superimposed in the figure for
comparison. Design curve for full-penetration both-
sided butt joint accordingly to Eurocode 9 (class 35-4)
is also depicted in the figure.


Figure 10. Wohler and Gassner curves for R=0.


Figure 11. S-N curves for aluminium alloy AA 6082-T6
under constant amplitude loading. R=0.

Fatigue results obtained in this work are close with data
obtained by other authors for this welding process. FSW
specimens show higher fatigue resistance than
specimens welded by the other two processes (MIG and
TIG). However, the friction stir welded specimens
presented significantly lower lives than base material.
Figure 12 shows the two typical crack initiation modes
identified by optical observation, from the tunnel defect
(Figure 12a) or at surface near stress concentration due
to shear lips (figure 12 b).

50
100
200
1E4 1E5 1E6 1E7
S
t
r
e
s
s

r
a
n
g
e
,

Δ
σ
[
M
P
a
]
Number of cycles, N
f
R=0; CA
R=0; VA
R=-1; CA
R=-1; VA
50
100
200
400
1E4 1E5 1E6 1E7
S
t
r
e
s
s

r
a
n
g
e
,

Δ
σ
[
M
P
a
]
Number of cycles, N
f
Wholer: R=0 Gassner: R=0; n=1,5
Gassner: R=0; n=2 Gassner: R=0; n=4
Gassner: R=0; n=5
SSF
50
100
200
400
1E4 1E5 1E6
S
t
r
e
s
s

r
a
n
g
e
,

Δ
σ
Number of cycles, N
f
Base material
TIG [2]
MIG [2]
FSW [2]
MIG [3]
FSW [3]
Results
obtained
Eurocode 9:
class 35-4
ν=2
ν=5
ν=1,5
ν=4


Figure 12. Fatigue crack initiation modes identified in
welded specimens.

Tunnel defects revealed to be more detrimental on
fatigue resistance than stress concentration due to shear
lips. Therefore, to reduce the difference between FSW
welds and base material fatigue strengths is crucial to
avoid the presence of tunnel defects. The higher fatigue
resistance was obtained when axial force during
welding process was well controlled, leading to a
smooth upper surface and performing two welding
passages in order to eliminate completely tunnel defects.
A small loss of hardness resulting from the second
welding passage was observed, however fatigue
resistance was not significantly affected.

Comparing with Eurocode 9 design curve, FSW welds
are clearly in the safe side, indicating that a higher class
than 35-4 can be attributed to butt joints welded by
FSW, providing that a good control of welding
parameters is performed in order to avoid severe
internal defects and that stress concentration near shear
lips is minimized.


4. CONCLUSIONS

Friction stir welding process leads to a decrease of the
material mechanical properties: yield and rupture stress
are lower than for base material. Detailed examination
revealed a hardness decrease in the thermo
mechanically affected zone and the nugget zone average
hardness was found to be significantly lower than the
base alloy hardness.

The friction stir welded AA6082-T6 specimens
presented significantly lower lives than base material
specimens. Tunnel defects and shear lips, formed
mainly in the retracting side, lead to a significantly
reduction of the fatigue lives. Tunnel defects revealed to
be more detrimental on fatigue resistance than stress
concentration created near shear lips. The higher fatigue
resistance was obtained when the axial force was well
controlled in order to obtain a smooth upper surface,
and performing two welding passages to eliminate
completely tunnel defects.

A good agreement was observed between constant and
variable amplitude fatigue tests, plotting equivalent
stress, calculated accordingly to Miner´s rule, against
the number of cycles to failure. The damage sum was
ranged between 0,6 and 2 for the stress ratio R=0 under
the four spectra loadings analysed. For R=-1 fatigue
data obtained under variable amplitude presented lower
lives than under constant amplitude loading, indicating
that damage sum values are lower than unity for the
majority of the tests. Therefore, Miner´s rule seems to
overestimate fatigue life for R=-1.


REFERENCES

[1] Thomas, W. M., Nicholas, E. D., Needham, J. C.,
Much, M. G., Templesmith, P. und Dawes, C. J.,
G. B. patent application No. 9125978.8 (December
1991)..

[2] Ericsson, M., and Sandstrom, R., Influence of
welding speed on the fatigue of friction stir welds,
and comparison with MIG and TIG. International
Journal of Fatigue 25 (2003)1379–1387

[3] Moreira, P.M.G.P, Jesus, A.M.P, Ribeiro, A.S. and
Castro, P.M.S.T. Fatigue crack growth in friction
stir welds of 6082-T6 and 6061-T6 aluminium
alloys: A comparison. Theoretical and Applied
Fracture Mechanics 50-2 (2008) 1-91.

[4] Lanciotti, A and Vitali, F. Characterisation of
friction stir welded joints in aluminium alloy 6082-t6
plates. Welding International 17-8 (2003) 624-630.

[5] Kobayashi, Y., Sakuma, M., Tanaka Y and
Matsuoka K. Fatigue Strength of Friction Stir
Welding Joints of Aluminium Alloy 6082 Extruded
Shape. Welding International 21-1 (2007) 18–24.

[6] Cavaliere, P., De Santis, A., Panella, F. and
Squillace A. Effect of welding parameters on
mechanical and microstructural properties of
dissimilar AA6082–AA2024 joints produced by
friction stir welding. Materials and Design 30 (2009)
609–616.

[7] Miner MA. Cumulative damage in fatigue. J Appl
Mech 1945;12: A159–A64.

[8] Heuler P, Klatschke H. Generation and use of
standardized load spectra and load–time histories.
Int J Fatigue 27-8 (2005) 974–90.

[9] ASTM E739-91, Standard Practice for Statistical
Analysis of Linear or Linearized Stress-Life (S-N)
and Strain-Life (ε-N) Fatigue Data. E08.04, Book of
Standards, vol. 03.01, Wset Conshohocken, USA.
a) b)

while figure 2b shows the e cross sectio of a wel on lded plate w welded with the parameters above referred.68 0. εr (%) .05 0. o avo tunnel def oid fects is increas sing heat inpu by applying ut g a higher axial fo h orce.Chemical co omposition of AA6082-T6 f aluminium alloy (wt%) m ) Si Mg Mn Fe Ti Ot Cr Cu Zn ther 1. wit the stress ratio R set to 0 and -1. tunnel or kiss sing bond def fects were pres in the we sent elds.04 0 0.01 0. wh did not in hich nfluence static mechanical properties. a) w presenting a defect type tunnel. 3) were mill at the edg and weld led ges. The frequency th o T y wa in the interval of 20–40 Hz depend as 0 ding upon the e stre level. typic cally tunnel d defects with less than 200 l 0 μm were also te m. σUTS (M MPa) Tensile stre 245 Yield streng σYS (MPa gth. on . Therefore. ested in order to evaluate th influence r heir e on fatigue resis stance. The di imensions of t fatigue specimens were the e 160 0x15x4 mm3 (length. In orde to analyse t er that no defect like root fla ts aws. cre eating stress c concentration which also re educes fatigue e live. ey uously formed in the weld direction. o e lea the forma ads ation of shear lips in the upper surface. cross-section ning of the welds for metallograp phic analysis in pl lanes perpend dicular to the w welding direct tion were also p performed. d sur rface preparat tion was mad only by gri removing de ind g the thickness exc e ceeding mater rial. Photo of a fatigue specimen cut from welded F o f pla before edg machining. pect on The penetrat tion depth wa adapted to fully penetra as ated butt joint.8 0. in ere s order to exc clude possible deviation fr e rom steady state due to the t typical lower temperatures during weld s ding start. ate ges . The d d e the are continu siz of these def ze fects tends to be great at th start of the he e we elding process due to low temperatu generally wer ure y ach hieved. where n defects w no were observed by microscopy o observation.01 0. Figure 1 is an exam mple of the b better succee eded welds concer rning its surfa aspect. to T o avo the prese oid ence of the tu unnel defects some welds s s we performed with two pass ere sages. T tunnel defe The fects tend to appea in the advan ar ncing side of the nugget low wer zone near th thermo-me he echanically af ffected zone and Fat tigue tests w were carried out in a se ervo-hydraulic c tes sting machine under consta and variab amplitude ant ble e loa adings. The w weld was tran nsverse to the e stre axis in th S–N tests a to the ma ess he and aterial rolling g dir rection.2 Fatigue deta 2 ails Figure 1 Surface asp of a frictio stir weld. r . Fr rom each wel lded plate. (a) ( (b) ( Figure 2. Figure 3. Howeve an excess of axial force er. thickness). rotating speed of 1500 n.6 mm obtained before weld m. The cut t spe ecimens (Fig. Figu 2a shows an example o a ld ure of poor quality welded plat presenting a tunnel defect y te. nam nd l mely. . 1. u . wing parameters w were chosen in order to obt n tain welds wit a th good surface aspect and where no d e d defects could be identified by microscopic observation: w y welding speed of d 300 mm/min tilt angle o 2º.02 0. because an e excessive pe enetration will create a n notch effect t that could affect m mainly the fat tigue strength. A sin nusoidal load d–time functio was used. he f med y e Th majority of welds perform with only one passage lea to the presence of a tun ads nnel defect fo ormed mainly y in the retractin side. Specimens mac chined of pl lates where defects were e ide entified. size of abo out 0. H Hv. alloy type. It was real lized to be ve difficult to perform the ery o e cor rrect optimiza ation of all parameters. of g rpm. Th is a very important asp his pect. also a good penetration control of th shoulder w d he was obtained. b) F weld g weld withou defects identifiable by mic ut croscopy.26 0. ding parameters o optimization.05 Table 2 . Life was defined as the numbe of cycles to ess e er o . wi idth.Table 1 . the c p e fatigue resistanc was signific ce cantly affected One way to d. a) 9 Elongation. 2.02 110 Welding parameters depen on several factors. Hardness. Alth ng hough. In t weld samp ace this mple.Mechanical properties of AA6082-T6 l f alumin nium alloy 300 ength. in som cases the me e tun nnel defects o observed had a size smaller than 30 μm. p penetration de epth and join The follow nt. T The samples were prepa ared accordingly to standard metallograph practice and hic etched in o order to enab the ident ble tification of the different wel zones. the first 40 mm of e m the plate we not used in specimens preparation.

failure and a total of 57 specimens were tested.5 1 0. using four values for the shape exponent ν. . etc. m.5 ν=2 ν=4 ν=5 ν= ∞   where.5 Table 3.8E5 1.8 1. (4) Number of cycles SSF depends on the shape of the spectrum under consideration (parameters H0 and ν) and the slope. 2.6 ν=0.5 mm far from both surfaces and at the   .2E5 1. 2. namely.3 Microhardness details Microhardness tests were performed to characterize the Vickers hardness profile in the vicinity of the weld area using a 2 N load. SSF. (1) Figures 4 and 5 show the several variable loading amplitude spectra applied during fatigue tests. Table 3 present SSF values for some ν values and a description of its typical applications. Spectrum shape parameters ν and SSF for typified amplitude spectra: H0=106. that represents the distance (factor. life ratio) between an (arbitrary) Wohler S–N curve and the associated spectrum fatigue life curve (in Germany sometimes referred to as Gassner curve in honour of Ernst Gassner). 4 and 5. This factor was calculated using the Miner damage rule (without endurance limit).2 0.0E0 0.56 2.5.61 3.0E4 6.39 Typical for road roughness induced loads ν<1 typical for wind gusts.  ∑ ∑ . .0E5 8.0E4 4. SSF can be more precisely calculated using the following equation · 0 Cumulative number of cycles.max 0.14 3. according to equation 1.4 (2) 0. 1. depicting the number of cycles versus the normalized stress amplitude.8 ν= ∞ Sa/Sa. 1 where Hi cumulative frequency of load cycles for level Sa.5 ν=2 ν=4 ν=5 1E0 1E2 1E4 1E6 0.0E5 1. .2 Using the loading spectrum defined by equation 1. Measurements were performed along three lines: 0. were considered. taking H0=106 cycles and m=5.6E5 Typified amplitude spectra (H o =10 6 cycles): ν=1 ν=1. 0. Figure 4 is the typical form of spectrum presentation where the normalized stress amplitude is plotted against the cumulative number of cycles.30 1. m=5. v ∞ 5 4 2 1.6 0. the ‘spectrum shape factor’.i H0 block size (number of cycles) ν shape exponent   Equation 2 specifies an additional number. 1 . 2.0 0. Typified amplitude spectra: number of cycles versus the normalized stress amplitude.0E4 2.95 4. Typified amplitude spectra: normalized stress amplitude versus cumulative number of cycles (H0=106 cycles). wave actions. H (3) Figure 4.4 0. of Wohler´s curve. Description Constant amplitude loading 1.4E5 1.8 SSF 0 1.0E4 0.max Figure 5. typified amplitude spectra [8]. while in figure 5 a more intuitive way for spectrum interpretation is shown.5. For the variable loading amplitude tests.8 ν=1 ν=1.0 ν>2 typical of bridge and crane structures Stationary Gaussian process Sa/Sa.

Constant and variable amplitude loadings.   .2).specimen middle thickness. although bottom measurements show that the lower hardness is restricted to a shorter zone.71 Blower=-6. The results show similar hardness profiles for the three measurement lines. This figure shows clearly a hardness decrease in the thermo mechanically affected zone and that the nugget zone average hardness is significantly lower than the base alloy hardness.5 2 2. leading to a softening up to 60% relatively to the base material hardness (about 115 HV0. The damage sum D ranged between 0. Yield stress decreases 33% and the rupture stress decreases 20% relatively to base material values.52 BStdv =0. 3.22 Bupper=-4.6 and 2 for the four spectra loadings. Wohler curves for R=0. indicating that Miner´s rule gives real damage sum values near unity. Hardness profiles measured along three lines. Fatigue data for constant amplitude loading were statistically analysed accordingly ASTM E739-91 Figure 8.5 Damage sum Fatigue results obtained for the stress ratio R=0 under constant and variable amplitude loadings are plotted in figure 7. Fatigue data obtained with variable amplitude loading. A low dispersion of fatigue data was obtained: the standard deviation was 0. This conclusion is better visible in Figure 8 where the probability of occurrence is plotted against real damage sum values. Stress range. Compared with base material (table 2). σYS (MPa) Elongation.04 for stress and 0.Mechanical properties of friction stir welds performed in AA6082-T6 aluminium alloy. This is a lower dispersion when compared with fatigue data obtained in tests of specimens welded by other processes.2 200 Am=16.21 for fatigue life. Hardness Hv 0.45 Alower=15. are in the same scatter band of the constant amplitude loading results. σUTS (MPa) Yield strength.32 Bm=-5.35 tp BStdv =0. 130 120 110 100 90 80 70 60 50 top middle bottom Standard [9]. εr (%) 241 165 6.87 50 1E4 1E5 1E6 1E7 Number of cycles. friction stir welding process leads to a decrease of the material mechanical properties: yield and rupture stresses of friction welded specimens are significantly lower than for base material. Hardness profiles were measured in all plates. Real damage sum distribution.41 Aupper=18. RESULTS Table 4 shows the mechanical properties of welded specimens obtained in tension tests. Microstructures hardness of the welds typically ranged between 70 and 80 HV0. Δσ [MPa] -20 -15 -10 -5 0 5 10 15 Distance from weld centre (mm) 20 100 R=0.8 100% 90% 80% 70% Probability 60% 50% 40% 30% 20% 10% 0% 0. Tensile strength. VA 95% confidence bands for the median S-N curve SN=0.04 R2=0. using the equivalent stress range calculated accordingly to Miner´s rule. Nf Figure 7.2.87 AStdv =0.21 Sσ=0.81 Figure 6.5 1 1. in successive positions with 1 mm of distance.72 tp AStdv =1. CA R=0. Table 4 .

n=5 200 Stress range. Miner´s rule seems to overestimate the fatigue life in the case of the stress ratio of R=-1. Data for variable amplitude loading and R=-1 were restricted to the shape exponent ν=2. Wohler and Gassner curves for R=0. a significant influence of the mean stress was observed: for the same stress range. n=1. Constant and variable amplitude loadings. Δσ [MPa] ν=1. VA R=-1. However. Wohler curves for R=0 and R=-1. 400 200 SSF 100 Wholer: R=0 ν=2 Gassner: R=0. indicating that damage sum values are lower than unity for the majority of the tests. the friction stir welded specimens presented significantly lower lives than base material. Δσ 200 FSW [3] Results obtained As expected. 100 Eurocode 9: class 35-4 MIG [3] 50 1E4 FSW [2] MIG [2] TIG [2] 1E5 Number of cycles. S-N curves obtained with specimens welded by FSW. Figure 10 shows the Wohler curve (stress range versus life) and Gassner curves (spectrum maximum stress range versus life) for R=0 and for several values of the spectrum shape exponent. It is clear that as v increases the Gassner curve becomes more close to Wohler curve. FSW specimens show higher fatigue resistance than specimens welded by the other two processes (MIG and TIG). Therefore. Figure 12 shows the two typical crack initiation modes identified by optical observation. n=2 ν=5 Gassner: R=0. Nf Base material Figure 9. Fatigue results obtained in this work are close with data obtained by other authors for this welding process. Nf 1E6 Figure 11. Δσ [MPa] Figure 9 compares fatigue results obtained with constant and variable amplitude loadings for the stress ratios R=0 and R=-1. CA R=-1. More tests are in course under variable amplitude loading for R=-1 and other spectrum shape exponents in order to confirm this trend. CA R=0. plotting stress range against the number of cycles to failure. n=4 100 R=0. 400 50 1E4 1E5 1E6 1E7 Number of cycles. Nf Figure 10. S-N curves for aluminium alloy AA 6082-T6 under constant amplitude loading. VA 50 1E4 1E5 1E6 1E7 Number of cycles. . Stress range. from the tunnel defect (Figure 12a) or at surface near stress concentration due to shear lips (figure 12 b). ν. MIG and TIG processes as well as base material are superimposed in the figure for comparison. R=0.Stress range.5 ν=4 Gassner: R=0. For R=-1 fatigue data obtained under variable amplitude present lower lives than under constant amplitude loading. The graphical meaning of the SSF parameter is also indicated. As referred before SSF=1 when ν=∞. Design curve for full-penetration bothsided butt joint accordingly to Eurocode 9 (class 35-4) is also depicted in the figure. Figure 11 compares fatigue results obtained in this work for R=0 under constant amplitude loading with results obtained by other authors [1. specimens tested with R=-1 present higher lives than specimens tested with R=0.3] in welded specimens of the same alloy.5 Gassner: R=0.

indicating that damage sum values are lower than unity for the majority of the tests. Characterisation of friction stir welded joints in aluminium alloy 6082-t6 plates. P. [6] Cavaliere. vol. Book of Standards. The higher fatigue resistance was obtained when axial force during welding process was well controlled. against the number of cycles to failure.a) b) ranged between 0. C. A...M.. The higher fatigue resistance was obtained when the axial force was well controlled in order to obtain a smooth upper surface. For R=-1 fatigue data obtained under variable amplitude presented lower lives than under constant amplitude loading. J. De Santis. M. Theoretical and Applied Fracture Mechanics 50-2 (2008) 1-91.. F.. Therefore. [7] Miner MA. Generation and use of standardized load spectra and load–time histories. J. Fatigue crack growth in friction stir welds of 6082-T6 and 6061-T6 aluminium alloys: A comparison.. Sakuma. W. and comparison with MIG and TIG. formed mainly in the retracting side. and performing two welding passages to eliminate completely tunnel defects.M. Fatigue Strength of Friction Stir Welding Joints of Aluminium Alloy 6082 Extruded Shape.. Standard Practice for Statistical Analysis of Linear or Linearized Stress-Life (S-N) and Strain-Life (ε-N) Fatigue Data. Wset Conshohocken. P. Tunnel defects revealed to be more detrimental on fatigue resistance than stress concentration created near shear lips. [8] Heuler P. providing that a good control of welding parameters is performed in order to avoid severe internal defects and that stress concentration near shear lips is minimized. Welding International 21-1 (2007) 18–24. 03. D. P. A. M.04. Much. M. The damage sum was . G. Y. calculated accordingly to Miner´s rule. A small loss of hardness resulting from the second welding passage was observed. patent application No. The friction stir welded AA6082-T6 specimens presented significantly lower lives than base material specimens. [2] Ericsson. Panella. Effect of welding parameters on mechanical and microstructural properties of dissimilar AA6082–AA2024 joints produced by friction stir welding. and Castro. P.S. 9125978. B. FSW welds are clearly in the safe side. USA. Materials and Design 30 (2009) 609–616. REFERENCES [1] Thomas. A. Cumulative damage in fatigue. Tunnel defects revealed to be more detrimental on fatigue resistance than stress concentration due to shear lips. Welding International 17-8 (2003) 624-630. International Journal of Fatigue 25 (2003)1379–1387 [3] Moreira. [4] Lanciotti. CONCLUSIONS Friction stir welding process leads to a decrease of the material mechanical properties: yield and rupture stress are lower than for base material. Miner´s rule seems to overestimate fatigue life for R=-1. und Dawes.01. Jesus. J Appl Mech 1945. plotting equivalent stress. Therefore. Tanaka Y and Matsuoka K. leading to a smooth upper surface and performing two welding passages in order to eliminate completely tunnel defects. A good agreement was observed between constant and variable amplitude fatigue tests. indicating that a higher class than 35-4 can be attributed to butt joints welded by FSW. and Sandstrom.G.. E08.12: A159–A64. Int J Fatigue 27-8 (2005) 974–90. F.T. Fatigue crack initiation modes identified in welded specimens. Templesmith..S. E. C.8 (December 1991).. A and Vitali. [9] ASTM E739-91. Needham. R. however fatigue resistance was not significantly affected. [5] Kobayashi. Detailed examination revealed a hardness decrease in the thermo mechanically affected zone and the nugget zone average hardness was found to be significantly lower than the base alloy hardness.6 and 2 for the stress ratio R=0 under the four spectra loadings analysed. lead to a significantly reduction of the fatigue lives.P... Figure 12. to reduce the difference between FSW welds and base material fatigue strengths is crucial to avoid the presence of tunnel defects. Ribeiro. Comparing with Eurocode 9 design curve. Influence of welding speed on the fatigue of friction stir welds.P. 4. and Squillace A. Klatschke H. G. Tunnel defects and shear lips. Nicholas.M. M.