Вы находитесь на странице: 1из 7

International Journal of Pressure Vessels and Piping 83 (2006) 381387

www.elsevier.com/locate/ijpvp

Residual stresses measurement by neutron diffraction and theoretical


estimation in a single weld bead
John W.H. Price a,*, Anna Paradowska a, Suraj Joshi a, Trevor Finlayson b
a

Department of Mechanical Engineering, Monash University, Wellington Road, Clayton, Vic. 3800, Australia
b
Department of Physics, Monash University, Wellington Road, Clayton, Vic. 3800, Australia

Abstract
Welding residual stresses are important in pressure vessel and structural applications. However, residual stress remains the single largest
unknown in industrial damage situations. They are difficult to measure or theoretically estimate and are often significant when compared with the
in-service stresses on which they superimpose. High residual stresses lead to loss of performance in corrosion, fatigue and fracture.
In this research, a measurement of residual stress by the neutron diffraction technique is compared to an analysis of the same geometry by
theoretical finite element procedures. The results indicate good agreement but scope for further understanding of the details of modelling the
welding heat source, heat transfer and variation of material properties with temperature.
q 2006 Elsevier Ltd. All rights reserved.
Keywords: Residual stress; Neutron diffraction; Hole drilling; Welding

1. Introduction
Residual stresses are formed in weld structures primarily as
the result of differential contractions, which occur as the weld
metal solidifies and cools to ambient temperature. These
stresses can have important consequences on the performance
of engineering components [1]. Weld residual stresses have a
significant effect on corrosion, fracture resistance and
corrosion/fatigue performance [2] and a reduction of these
stresses is desirable.
There are several ways of directly measuring residual
stresses in small volumes. The most common ones involve
mechanical invasive methods (e.g. hole drilling or cutting
[3,4]) and non-destructive methods using radiation such as
X-ray (laboratory or synchrotron) or neutron diffraction [57].
Of these, only neutron beams can establish stresses in the
interior of components of a metallic material and have a small
volume of measurement (1 mm3).
In this paper, experimental measurements of weld stresses
generated by a single bead-on-plate of low-carbon steel using
MIG welding are presented. In this work, we have concentrated
* Corresponding author. Address: Department of Mechanical Engineering,
Monash University, P.O. Box 197, Caulfield East, Vic. 3800, Australia. Tel.:
C61 3 9903 2868; fax: C61 3 9903 2766.
E-mail address: john.price@eng.monash.edu.au (J.W.H. Price).

0308-0161/$ - see front matter q 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijpvp.2006.02.015

on the influence of restraint on the residual stresses behaviour


with the intention of providing key data for the validation of
design and fitness for purpose methodologies and finite element
tools.
2. Experimental work
2.1. Material and welding procedure
The material used in this study was a low-carbon steel [8].
The chemical composition of the parent material and weld
metal are shown in Table 1. The dimensions of the plates were
200!100!12 mm. Typical mechanical properties of parent
and weld metal are shown in Table 2.
Sample I was unrestrained and Sample II was fully
restrained. Restraint was achieved by welding Sample II to a
very thick steel plate, which was cut off after cool down.
Distortion of Sample I was overall approximately 18 in
transverse and 0.58 for longitudinal directions, Sample II had
no visible distortion.
The bead-on-plate welds were produced down the centre of
the plates, using a constant current UNI-MIG 375K AC/DC
power source. The specimens were mounted under an
automatic-speed-controlled, welding torch. The electrode was
Super-COR 5, 1.6 mm in diameter (conforming to AWS
A5.20) with a 20 mm contact tip to work distance. The welds
were made with an ARGOSHIELD 52 gas (with gas flow rate
18 l/min), with start and stop positions approximately 25 mm

382

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

Table 1
Chemical composition of the consumable materials (in wt%)
Composition
material

Mn

Parent
metal
Weld
metal

0.12 0.63 0.13 0.01 0.02 0.02 0.01 0.01 0.01 !0.01
0.10 1.7

Si

Ni

Cr

Mo

Cu

0.68 0.02 0.02 0.05 0.03 0.04

0.04

from each end of the plate. The welding parameters were:


current 260280 A, voltage of 2830 V and a speed of the
electrode 360 mm/min. The width of the welds bead (Fig. 1)
was 14 mm. There was no pre- or post-weld heat treatment.
The characteristic of the weld hardness and evaluation of
microstructure was described in details in previous work [9].

Fig. 1. Principles of the neutron diffraction technique showing Braggs


reflection from the crystal plane d (grain size greatly exaggerated for clarity).

2.2. Determination of residual stress


sxx Z
The neutron diffraction technique is capable of determining
residual stresses non-destructively within the interior of
components [1]. Diffraction techniques exploit the crystalline
lattice of the material as an atomic strain gauge. When a beam
is passed through a polycrystalline material, diffraction occurs
according to Braggs law which is given by the equation
nl Z 2di sin qi

(1)

where n is an integer, qi is the Bragg angle for crystallographic


planes, i having interplanar spacing di.
Under an applied tensile (or compressive) stress, the lattice
spacing (di) in individual crystallite grains expands (or
contracts). This change in the lattice spacing can be detected,
at constant wavelength, as a shift (Dqi) in diffraction peaks,
which is schematically shown in Fig. 1. From Braggs
equation, the strain (3i) is given by
3i Z

di Kd0
ZKcot qi Dqi
d0

(2)

where d0 is the strain-free interplanar spacing for the lattice


planes, i.
The orientation of the principal strains in any specimen is
determined from the geometry. The strains (exx, eyy, ezz) in a
solid can be converted to the three-dimensional stress (sxx, syy,
szz) state. For an elastically isotropic solid, Eq. (3) give the
stresses in three directions using the xx direction as an example.



E
1Kn3xx C n3yy C 3zz
1 C n1 C 2n

(3)

where E is Youngs modulus, and n is Poissons ratio.


2.3. Experiment details and results
Neutron diffraction measurements were undertaken on the
neutron beam stress scanner TASS (The Australian Strain
Scanner) at ANSTO, Sydney, Australia. The same parameters
of measurements were specified for both specimens. The
. Measurements were
neutron wavelength used was 1.40 A
made using the (112) reflection, at the detector angle, 2qI, of
approximately 73.58. Measurements were made with the
scattering vector (see Fig. 2) parallel to the three axes marked
transverse, longitudinal and normal as shown in Fig. 2.
The incident and diffracted beam slits for transverse normal
measurements were 1 mm wide and 20 mm high because the
stresses do not change much in the longitudinal direction.
Because the stresses are changing rapidly in the transverse
direction; for the longitudinal direction the slits can only be
1.5 mm width, 2 mm high. The slits are formed by a mask of a
neutron absorbing material (Cadmium) in the incident and
diffraction beams.

Table 2
Typical mechanical properties
Mechanical properties

Yield
stress
(MPa)

Tensile
strength
(MPa)

Elongation
(%)

Parent metal (experimental


measurements according to
AS 1391:1991)
Weld metal (as manufactured using Argoshield 52
shielding gas)

285

429

38

445

550

29
Fig. 2. The direction of the measurements (transverse x, normal y, longitudinal
would be z) using neutron diffraction on the single bead-on-plate.

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

383

Fig. 3. Unrestrained Sample I. The longitudinal, transverse and normal components of strain (a) and stress (b) measured by neutron diffraction against distance from
the weld centre line. Error bars based on uncertainty in the value of the peak diffraction angle are shown.

The stress-free parameters for the steel were measured on


eight 2!2!2 mm cuboids which had been cut and glued
together (the d0 specimen). The slits for the stress-free
parameters were the same as for longitudinal measurements
(1.5!1.5!2 mm). Scans were made along the transverse line
(Fig. 3) from xZ0 (the centre of the weld) to xZ32 mm. The
centre of the gauge volume was 1.5 mm below the top surface.
The d0 specimen required three measurements for accuracy and
yielded a stress free diffraction angle, 2q, was 73.36.
Counting times are determined by the accuracy required of
the measurements. Counts are collected until the experimenters
are confident that the peak has been clearly defined. The time
required for this varies with each geometry and direction of
measurement. The number of counts of neutrons at the
collectors used in our experiments was approximately 10,000
for each point. For each data point this required times of
approximately 25 min for transverse and normal strain
measurements and 60 min for longitudinal and d0 parameter
measurements.
The residual stresses were derived from the elastic strain
measurements (Figs. 3a and 4a) using Youngs modulus of
207 GPa, and Poissons ratio of 0.3. The maximum longitudinal tensile residual stress was found near the middle of the weld
and is shown in Figs. 3b and 4b, reaching the approximate

value of 350360 MPa for Sample I and 470490 MPa for


Sample II (Fig. 5).
3. Finite element modelling
The objective here was to perform three-dimensional finite
element modelling of the bead-on-plate experiment to calibrate
the welding procedure. A relatively uncommon but powerful
commercial FEA package called SysweldC was used in this
attempt. The parent and the weld material were assumed to
have the same mechanical and thermal properties, as was
provided in the software database for the material S355J2G3
with chemical composition as follows: C%0.20%,
Mn%1.60%, Si%0.55%, S%0.035%, and P%0.035%. The
solidus temperature was 1440 8C, the liquidus temperature was
1505 8C, and the latent heat of fusion was 270,000 J/kg. The
temperature dependent properties were measured and tabulated
by extensive experimentation and supplied with the software.
Three-dimensional meshes of the substrate plate and the bead
were constructed as illustrated in Fig. 6.
For the sake of geometric convenience, the volume of the
bead was modelled as an elliptical pyramid (i.e. the front and
the back faces of the bead were half-ellipses). Differential
element sizes were used in meshing, with the density being

Fig. 4. Fully restrained Sample II. Data for comparison to Sample I.

384

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

Fig. 5. Change in stress in fully restrained sample in comparison to unrestrained sample.

higher near the centreline along which the bead was deposited,
and progressively reducing towards the edges of the substrate
plate. Similarly, the mesh density on the top face of the bead
was made higher than that on the model. Special care had to be
taken to ensure that the mesh discretisation specified at different
lines and edges, especially, at the juncture of the bead and the
substrate plate, was such that the nodes fell on top of each
other. The volume mesh was created with brick elements.
In order to generate the convection and radiation boundary
conditions, skin elements (two-dimensional surface mesh with
quadrilateral elements) was constructed on all the exposed
domains of the piece. As before, the mesh density of surface
mesh was specified such that the skin element nodes were
coincident with the volume element nodes under them. A
combined convective and radiative heat transfer coefficient of
25 W/m2 was assumed. The initial temperature was assumed to
be 20 8C (ambient temperature).
The program required that the weld trajectory be explicitly
specified along the direction and position of the moving heat
source with linear, one-dimensional elements. The trajectory
was chosen to be along the centreline of the substrate plate,
with concordant discretisation of the weld line to ensure that
the nodes would coalesce with those on the skin and volume
elements. In totality, a total of 13,043 nodes with 16,208
elements were created. The simulation was run only for the
unrestrained model.

3.1. Calibration of the heat source


A welding heat source is usually an arc plasma radiating
intense heat outwards with decreasing temperature. The
shape and the size of the heat plume are modified by
ionisation of the gas, which adds energy to the radiation.
Rosenthals mathematical model of heat source can be
simulated by considering a Gaussian distribution. According
to Nguyen [10,11], the most appropriate model for the heat
source for TIG and MIG welding procedures is the double
ellipsoidal heat source. A double ellipsoidal heat source
consists of two different single ellipsoids, and is suitably
considered top be a more advanced heat source than the
single ellipsoidal due to its greater flexibility in modelling
realistic shapes of the moving heat source. The heat
density at an arbitrary point (x, y, z) within each one-half
ellipsoid is described by the following equations (refer
Fig. 7)
p
6 3rf Q K3x2 =c2hf Cy2 =a2hCz2 =b2h
p e
Qx;y;z Z
(4)
ah bh chf p p
p
6 3rb Q K3x2 =c2hbCy2 =a2hCz2 =b2h
p e
Qx;y;z Z
ah bh chb p p

(5)

where ah, bh, chf, chb are the ellipsoidal heat source
parameters, QZarc heat inputZhVI, h being the efficiency

Fig. 6. Finite element mesh: (a) single bead on plate; (b) magnified area.

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

385

Fig. 7. (a) Double ellipsoidal heat source, (b) double ellipsoidal parameters as employed in SysweldC.

Fig. 8. (a) Front view of the molten pool on plate, (b) wireframe model of the molten pool.

and V and I being the arc voltage and current, respectively,


and rf, rb are the proportional coefficients at the front and the
back of the heat source, respectively, such that (rfCrbZ2).
SysweldC, however, provides a convenient method of
estimating the coefficients to obtain the heat energy density (in
W/mm3) of each one-half ellipsoid. The heat input fitting
module of the program allows the programmer to enter the
starting value of the geometric parameters with accurate arc
energy input to give the values of Qf (heat energy density in the
front half) and Qr (heat energy density in the rear half)
iteratively. The programmer is able to iterate several times to

calibrate the heat source closely with the measured dimensions


of (i) thermal image of the molten weld pool, (ii) distortions of
the edges, or (iii) temperatures at specified points. For this
experimental set-up, the heat source was calibrated using the
perceived image of the weld pool as is shown in Fig. 8.
With an input of 7500 W for the energy input and afZ4,
arZ8, bZ7, cZ0.5, and vyZ6 mm/s (velocity of the weld
torch along the weld line), the heat density values obtained for
the above shown profile of the weld pool were QfZ
104.72 W/mm 2 and QrZ69.81 W/mm 2. The parameters
mentioned here are used in the definition of the double

Fig. 9. Temperature profiles at times (a) about 10 s (b) about 20 s.

386

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

3.3. Results of the mechanical analysis

Fig. 10. Temperature history of the midpoint of the trajectory.

ellipsoidal model as provided in the documentation of


SysweldC (Fig. 7b).
3.2. Results of the thermal analysis
The simulation was run for a time period of 7500 s, i.e. more
than 2 h, allowing for the complete thermal cycle, including
heating and subsequent air cooling to occur. The temperature
profiles obtained at about 10 and 20 s are shown in Fig. 9.
Automatic time stepping adjusted itself in such a manner that the
initial time intervals between successive recordings were quite
small during the period of deposition of the bead. As soon as the
weld torch finished depositing the bead, the time steps between
recordings became increasingly large. In either case, the time
steps were non-uniform during the entire 7500 s period.
It was observed that the maximum temperature reached
during the entire process was about 1600 8C with the above
mentioned heat source parameters. The thermal cycle for the
node at the midpoint of the weld trajectory is shown in
Fig. 10.

The mechanical analysis followed from the results of the


thermal analysis and took much longer to complete. The output
data included stresses in elements, integration points and
element nodes. Fig. 11 shows the plots of the residual stresses
at element nodes at 7500 s.
The finite element analysis results show that the transverse
stresses are tensile, though small in magnitude and nearing
zero, in the region close to the centreline, and become
compressive as we move towards the outer edges. This is in
good comparison with the measured values, which show the
same pattern. The longitudinal residual stresses, as predicted
by the finite element analysis results, are tensile in the region
near the weld line and turn more and more compressive when
moving towards the edges on either side. This is also in good
qualitative agreement with the experimentally measured values
as illustrated in Fig. 3.
4. Discussion
Transverse and normal stresses are low in the unrestrained
Sample I, because the sample deformed during welding.
However, for Sample II transverse and normal stresses are
raised at all points of measurements as shown in Fig. 5.
Longitudinal stress also generally increases but there are some
reductions around the toe of the weld.
The highest increase between the unrestrained and
restrained specimen was observed in middle of the weld
where the normal and transverse stresses change from
compressive for unrestrained sample (Fig. 3b) to tensile for
the restrained sample (Fig. 4b).
The peak stress in the weld (which is in the longitudinal
direction) is significantly higher in the weld area in both
samples, than the specified yield stress of the steel in question
(250 MPa). Hardness tests (on other samples) indicate

Fig. 11. (a) Transverse residual stress, (b) longitudinal residual stress, at 7500 s.

J.W.H. Price et al. / International Journal of Pressure Vessels and Piping 83 (2006) 381387

significant increases in hardness in the welded region reflecting


higher yield stresses in the weld. The peak stress in this case
does not occur at the toe of the weld but in the middle of weld.
However, the maximum longitudinal stress at the toe is around
150 MPa which will be approximately 60% of the yield
strength.
For comparison purposes, the fitness-for-purpose code
BS7910 [12] states in clause 7.2.4.1 that the following
assumption should be made for welds not subjected to postweld heat treatment:
For a flaw lying in a plane transverse to the welding
direction (i.e. the stresses to be considered are parallel to the
weld in the terminology used in this paper longitudinal
stresses) the tensile stress should be assumed to be equal to
the room temperature yield strength of the material in which
the flaw is located.
If this procedure described in BS7910 were used, a uniform
stress of at least 250 MPa would have to be used in the
calculations. The result of this difference will be a very
conservative assessment of the failure point of the weld and the
use of different growth curves for fatigue assessment.
Our work is also currently developing methods of cross
checking these results with finite element modelling of the
welding process. Initial finite element analysis results using
SysweldC were quite promising. While the longitudinal and
normal stresses were found to be in good agreement with the
experimental values, the longitudinal stresses were a bit off
from the measured values. However, qualitatively, the nature
of the residual stresses predicted by the program is in good
agreement with the experimental observation. Continuing
effort is being made to get an accurate calibration of the heat
source and refining the mesh to locate specific nodes at exactly
those points where experimental values were obtained, so that a
perfect tally can be made.
5. Conclusions
The use of a neutron beam as a non-destructive method of
measuring residual stress due to welding has been explored.
Experimental investigation showed that restraint of the plate
during welding has a significant effect.
The analysis of the same weld geometry following the
procedures of SYSWELDC were used. The handbook
suggested method of analysing the weld pool, heat transfer
and material properties were used. The results of the

387

comparison were promising, but further improvement in the


modelling could be achieved.
Acknowledgements
This work was conducted with the assistance of an
Australian Research Council grant supported by the Welding
Technology Institute of Australia (WTIA). Other assistance has
been received from the Monash University Research Fund, the
Australian Nuclear Science and Technology Organisation
(ANSTO) and an Australian Institute of Nuclear Science and
Engineering (AINSE) grant.
References
[1] Webster GA, Wimpory RC. Residual stress in weldments. J Neutron Res
2001;9:2817.
[2] Price JWH, Kerezsi B. Potential guidelines for design and fitness for
purpose for carbon steel components subject to repeated thermal shock.
Int J Pressure Vessels Piping 2004;81:17380.
[3] Jae-il Jang, Dongil Son, Yun-Hee Lee, Yeol Choi, Dongil Kwon.
Assessing welding residual stress in A335 P12 steel welds before and after
stress-relaxation annealing through instrumented indentation technique.
Scripta Mater 2003;18(6):7438.
[4] Pang JWL, Preuss M, Withers PJ, Baxter GJ, Small C. Effects of tooling
on the residual stress distribution in an inertia weld. Mater Sci Eng A
2003;356(1/2):40513.
[5] Owen RA, Preston RV, Withers PJ, Shercliff HR, Webster PJ. Neutron
and synchrotron measurements of residual strain in TIG welded
aluminum alloy 2024. Mater Sci Eng A 2003;346(1/2):15967.
[6] Webster J, Ananthaviravakumar N, Hughes DJ, Mills G, Preston RV,
Shercliff HR, et al. Measurement and modeling of residual stresses in TIG
weld. Appl Phys A 2002;74:S1421S3.
[7] Lorentzen T, Ibs JB. Neutron diffraction measurements of residual
strains in offshore welds. Mater Sci Eng A 1995;197(2):20914.
[8] Grade 250. Structural steelhot-rolled plates, floor plates and slabsAS/NZS 3678; 1996.
[9] Paradowska A, Price JWH, Ibrahim R, Aloraier A, Finlayson T. Caroline
curfs, neutron beam measurements of residual stress in a single weld bead
in a low carbo steel. International conference on failure analysis &
maintenance technologies, Institute of Materials Engineering, Brisbane,
Australia; 2930 April, 2004.
[10] Nguyen NT, Ohta A, Matsuoka K, Suzuki N, Maeda, Y. Analytical
solution of double-ellipsoidal moving heat source and its use for
evaluation of residual stresses in bead-on-plate. International conference
on fracture mechanics and advanced engineering materials, December 8
10, 1999, University of Sydney, Australia; 1999. p. 14349.
[11] Nguyen NT. Thermal analysis of welds. Southampton, UK: WIT Press;
2004.
[12] British Standards. BS 7910: 2000 Guide on methods for assessing the
acceptability of flaws in metallic structures. London: BSI; 2000.

Вам также может понравиться