Available online at www.sciencedirect.

com
Procedia
Engineering
Procedia Engineering011 (2009)
Procedia Engineering (2009)000–000
213–216
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

Mesomechanics 2009

Inverse eigenstrain analysis of residual stresses in friction stir welds

H.T. Luckhooa, T-S. Juna, A.M. Korsunskya,*
a
University of Oxford, Department of Engineering Science, Parks Road, Oxford OX1 3PJ, UK
Received 02 March 2009; revised 25 May 2009; accepted 02 June 2009

Abstract

This paper presents the results of eigenstrain analysis in a friction stir welded 12mm-thick 12%Cr steel plate. The finite element
models are established for inverse eigenstrain analysis for three different model cases. As the region containing the eigenstrain
distribution expands, the accuracy of the residual strain reconstruction improves. If the eigenstrain is allowed to be present along
the entirety of the sample, good agreement can be achieved. It is also noting that eigenstrain (permanent plastic strain) represents
the consequence of numerous inelastic processes occurring during to welding. The results suggest that significant inelastic
deformation of the plate takes place even at larger distances from the weld. This conclusion requires validation by independent
means, experimental and/or modelling.
© 2009 Elsevier B.V. All rights reserved

Key words: Friction stir welding, residual stress, eigenstrain, synchrotron X-ray diffraction

1. Introduction

Friction Stir Welding (FSW), a relatively new joining process, is a solid phase welding technique. It was first
proposed and developed in 1991 at TWI (The Welding Institute, UK). The welding is accomplished using a welding
tool consisting of a pin and shoulder. The tool is rotated around its axis and inserted into the interface between the
abutting edges of the two plates being joined. The tool then translates along the interface and ‘zips’ it together [1].
So far a number of studies have been carried out in order to show and verify the advantages of friction stir
welding compared with conventional processes, such as arc welding, laser welding, electron beam welding, etc; and
the suitability of this joining method for many industries, such as aerospace, automobile manufacture, ship building,
etc. Most studies so far have been focused on light (e.g. aluminium) alloys, while the application to the materials
with higher melting temperatures, such as steel and titanium, have been limited by the difficulty of identifying
suitable welding tool materials. However, in recent years these have been continuously improved [2-4].
Polycrystalline cubic boron nitride (PCBN) and alloys based on tungsten-rhenium (W-25Re) have been the leading
candidates, but neither material has so far fulfilled al of the requirements for a friction stir weld tool material, such
as the resistance to wear and thermal loading, and toughness [2]. In spite of the tool wear, mechanical properties of
FSW joints of steel components are usually found to be acceptable. However, very limited data is available on the

* Corresponding author. Tel.: +44-1865-2-73043; fax: +44-1865-2-73010.

E-mail address: alexander.korsunsky@eng.ox.ac.uk.

doi:10.1016/j.proeng.2009.06.050
214 H.T. Luckhoo et al. / Procedia Engineering 1 (2009) 213–216
Luckhoo H.T. et al./ Procedia Engineering 01 (2009) 000–000

residual stress states. Recently, FSW of steel plates has been successfully carried out using refractory metal welding
tools in a collaborative project between HC Starck and TWI. The surface finish and the microstructural appearance
of sections through the weld showed great improvements in terms of low porosity and few bonding defects. For
industrial applications of this technique, it is crucial to understand the residual stresses thoroughly in order to control
them by deformation or thermal processing, and to account for their presence in calculations. Therefore, the purpose
of the present study was to investigate the residual stress states in these improved FSW joints between thick (about
½ inch) steel (12%Cr) plates. This was achieved by carrying out residual strain measurement using synchrotron X-
ray diffraction, followed by inverse eigenstrain analysis.
Inverse eigenstrain analysis of diffraction measurements is a relatively novel approach to residual stress
modelling that makes full use of the fact that often residual stresses arise from localised plastic deformation, or
misfit (within the thermo-mechanically affected zone (TMAZ) in the case of FSW). The rest of the component often
responds predominantly elastically to the introduction of eigenstrain. The FE eigenstrain reconstruction method [5]
then uses a parameterized distribution of unknown eigenstrain. Parameters of the distribution are found by
minimising the disagreement between model predictions and the measurements.
In this study, correlation was established between the results of residual strain measurements and the underlying
permanent inelastic strain (eigenstrain) field. The reconstructed residual strains from the three different FE models
are compared with the experimental results.

2. Experimental

The material used in the present study is 12% Cr steel which is a structural material for, in particular, first wall
and blanket of fusion reactors. The joints were manufactured by butt welding with the two chromium steel plates on
a stainless steel base plate, with a purpose of protecting the machine bed. In the process, the down-force was fixed at
32kN, rotational speed fixed at 625rpm, lateral speed fixed at 2mm/sec. The dimensions of the joints were 203.2Ý
66.2Ý12.7mm3, as can be seen in Figure 1.

Fig. 1. Schematic of the FSW specimen

Residual strain measurements on the FSW specimen were performed using high resolution energy-dispersive X-
ray diffraction on beamline ID15 at the European Synchrotron Radiation Facility (ESRF) in Grenoble, France. Twin
detector setup was used to allow simultaneous measurement of strain parallel (z direction) and perpendicular (x
direction) to the weld. Both detector were mounted at fixed scattering angles of 2θ = 5° . Beam spot size used was
0.5¯0.5mm2 and counting time was 20 seconds per point. Measurements were conducted along a line at 2mm step
along the x direction on the mid-section of the specimen (y=6.35mm and z=33.1mm). In the absence of reference
strain-free samples, an approximation to the global value of a 0 (unstrained lattice parameter) was obtained by
averaging the lattice parameter values in each direction. This leads to the determination of nominal strains, with a
typical error of 5¯10-5, corresponding to stress uncertainty of only about 10MPa for steel. The diffraction patterns
were interpreted and analysed using the procedure described previously [1]. Figure 2 shows the profiles of the
residual elastic strain components ε zz and ε xx . The former exhibits compressive strain in the region between
±16mm, i.e. in the weld zone, and becomes residual tension in the heat affected zone. These strain distributions are
similar to those found in the literature [10]. No significant strain magnitudes are found for the transverse strain ε xx ,
 H.T. Luckhoo et al. / Procedia Engineering 1 (2009) 213–216 215
Luckhoo H.T. et al./ Procedia Engineering 01 (2009) 000–000

except in the vicinity of the weld zone. This fact is often observed for other joining and welding processes: strain
components transverse to the weld direction are lower in magnitude and more localised than the longitudinal ones.
1000 1000

800 800

600 600

400 400
Microstrain

Microstrain
200 200

0 0
-100 -50 0 50 100 -100 -50 0 50 100
-200 -200

-400 -400
εxx εzz
-600 -600
x position (mm) x position (mm)

Fig. 2. Distribution of residual elastic strains εxx and εzz along the x direction (z = 33.1mm and y = 6.35mm).

3. Eigenstrain Reconstruction using FEA

The term “eigenstrain”, introduced by Mura [6], describes a permanent inelastic strain within the sample that can
be thought of as the origin of residual stress. One of the advantages of the term “eigenstrain” is that its mechanical
consequences can be studies without explicitly considering the physical mechanisms that lead to this state being
established. Moreover, the very nature of eigenstrain means that it is independent of the specimen geometry; once
established e.g. for a particular processing operation on a specimen, the same eigenstrain can be applied (with due
consideration) for different geometries, in order to estimate the new residual stress distribution.
In the present study the eigenstrain distribution within the FSW is determined using a variant of the
reconstruction method proposed in literature [7-9]. A simplified inverse approach is used in conjunction with a basic
two dimensional finite element model. The effect of limiting the size of region exposed to eigenstrain is explored.
The aim of this study is to reconstruct the residual stress state observed in the steel friction stir welds. Thermo-
elastic finite element model was used to impose an eigenstrain distribution that induces a residual stress field.
However, initially there is no knowledge of the correct eigenstrain distribution: the only data available is the residual
elastic strain measurements made using high energy synchrotron X-ray diffraction (only strains in the z-direction are
considered at this stage). It is assumed that the variation in the y-direction can be ignored, and therefore that the
eigenstrain in the z-direction, ε * = ε * zz , is a function of x-position only. Korsunsky et al [5] have presented a method
whereby the unknown eigenstrain can be represented by a truncated series with unknown coefficients ci :
N
ε * ( x) = ∑ ci Ei ( x) (1)
i =1

In their method Ei is made up from Chebyshev polynomials. In the present study, a series of overlapping
triangles is used instead. Because of the simplifying assumptions, the finite element model could be reduced to a 2D
problem. Three reconstruction attempts were carried out, with eigenstrain distributions limited to ±22mm (Case 1),
±59mm (Case 2) and ±101.6mm (Case 3) respectively, as measured from the centre of the bond line. Case 3 was
also considered in a 3D finite element model, however, with eigenstrain ε * zz only varying in the x-direction.

4. Reconstructed residual elastic strain fields

Using the reconstruction method introduced in section 3, the eigenstrain distributions for the three cases were
found by inversion [5], and then imposed upon the weld so that the reconstructed elastic strains could be extracted.
Figure 3 shows the residual elastic strain profiles plotted alongside the experimental data used to generate them. As
the region exposed to eigenstrain increases, so does the accuracy of the model’s prediction. In Case 3 (±101.6mm)
the prediction follows the measured data very closely, with the exception of an artefact around x = -100mm, mostly
216 H.T. Luckhoo et al. / Procedia Engineering 1 (2009) 213–216
Luckhoo H.T. et al./ Procedia Engineering 01 (2009) 000–000

likely due to a mismatch between the ranges of available X-ray data and the range of the model’s output. In Cases 1
and 2 there is good agreement in the region where eigenstrain is applied but not outside this region.
Excellent agreement can be achieved when eigenstrain is installed throughout the sample. However, this
conclusion contradicts the assumption that eigenstrain only arises in the vicinity of the bond, within the thermo-
mechanically affected zone (TMAZ). The impossibility of matching the complete residual elastic strain distribution
by installing a localised eigenstrain distribution suggests that a significant variation of the unstrained lattice
parameter must exist within the sample and needs to be taken into account in the data analysis.
1000
X-Ray Measured Data
Case 1
800
Case 2
Case 3
600

400
εzz (microstrain)

200

0
-120 -80 -40 0 40 80 120
-200

-400

-600
x - position (mm)

Fig. 3. Residual elastic strain field for the steel FSW: measured profile and three different reconstructions.

5. Conclusion

It is clear that the eigenstrain modelling methodology used here can lead to very accurate predictions of the
residual stresses in friction stir welds. However the various test cases have demonstrated that the size if region of
eigenstrain is extremely influential in the quality of prediction that the method yields. If physically realistic
constraints are placed on the eigenstrain then this model’s results are very limited in its range of accuracy,
suggesting that the 1-D assumption for the eigenstrain is to limiting and a more complex distribution is needed (i.e.
varying in both x and y position).

References

1. Jun TS, Zhang SY, Golshan M, Peel M, Richard D, Korsunsky AM. Synchrotron energy-dispersive X-ray diffraction analysis of
residual strains around friction welds between dissimilar aluminium and nickel alloys. Material Science Forum 2008;571-572;407-412.
2. Threadgill PL, Johnson R. Progress in friction stir welding of steels, TWI; 2004.
3. Jones RL. Developments in joining processes for structural steel applications, TWI; 1999.
4. Fujii H, Cui L, Tsuji N, Maeda M, Nakata K, Nogi K. Friction stir welding of carbon steels. Material Science and Engineering A
2006;429;50-57.
5. Korsunsky AM, Regino GM, Nowell D. Variational eigenstrain analysis of residual stresses in a welded plate. International Journal of
Solids and Structures 2007;44;4574-4591.
6. Mura T. Micromechanics of defects in solids. Boston: Martinus Nijhoff; 1987.
7. Korsunsky AM. On the modelling of residual stresses due to surface peening using eigenstrain distribution. J Strain Analysis
2005;40(8):817-824.
8. Korsunsky AM. Variational eigenstrain analysis of synchrotron diffraction measurements of residual elastic strain in a bent titanium
alloy bar. J Mechanics Matls Structs 2006;1:101-119.
9. Korsunsky AM. Eigenstrain analysis of residual strains and stresses. J Strain Analysis 2009;44:29-43.
10. Prime MB, Gnäupel-Herold T, Baumann JA, Lederich RJ, Bowden DM, Sebring RJ. Residual stress measurements in a thick,
dissimilar aluminium-alloy friction stir weld. Acta Materialia 2006;54(15);4013-4021