Академический Документы
Профессиональный Документы
Культура Документы
Design
http://sdj.sagepub.com/
Published by:
http://www.sagepublications.com
On behalf of:
Additional services and information for The Journal of Strain Analysis for Engineering Design can be found at:
Email Alerts: http://sdj.sagepub.com/cgi/alerts
Subscriptions: http://sdj.sagepub.com/subscriptions
Reprints: http://www.sagepub.com/journalsReprints.nav
Permissions: http://www.sagepub.com/journalsPermissions.nav
Citations: http://sdj.sagepub.com/content/38/4/349.refs.html
349
Abstract: Tensile welding residual stresses can, in combination with operating stresses, lead to
premature failure of components by fatigue and/or fracture. It is therefore important that welding
residual stresses are accounted for in design and assessment of engineering components and
structures. In this work residual stress distributions, obtained from measurements on a number of
ferritic steel T-plate weldments using the neutron diffraction technique and the deep-hole drilling
method, are presented. It has been found that the residual stress distributions for three different plate
sizes are of similar shape when distances are normalized by plate thickness. It has also been found that
the conservatisms in residual stress proles recommended in current fracture mechanics-based safety
assessment procedures can be signicantof yield strength magnitude in certain cases. Based on the
data presented here a new, less-conservative transverse residual stress upper bound distribution is
proposed for the T-plate weldment geometry. The extent of the plastic zone developed during the
welding process has also been estimated by use of Vickers hardness and neutron diffraction
measurements. It has been found that the measured plastic zone sizes are considerably smaller than
those predicted by existing methods. The implications of the use of the plastic zone size as an indicator
of the residual stress distributions are discussed.
Keywords: residual stresses, T-plate weldments, fatigue, fracture, ferritic steel, neutron diffraction
method, deep-hole drilling method
NOTATION
C
dhkl
dhkl
DHD
E
E110
E211
FE
FWHM
HAZ
HV
ILL
ND
PWHT
q
r0
t
w
wa
x, y, z
Ddhkl
Dyhkl
e
ex , ey , ez
Z
yhkl
l
n
n110
n211
sx , sy , sz
sYP
s1 , s2 , s3
u
350
INTRODUCTION
Fig. 1
Uniaxial stressstrain curves obtained from 25 and 50 mm plate material, BS EN 10025 S355 J3G3
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
Table 1
Source
25 mm plate
50 mm plate
BS EN 10025
sYP (MPa)
su (MPa)
E (GPa)
358
348
345/335
510
515
490630
212
212
WELD DETAILS
Fig. 2
351
Cross-sections of T-plate weld samples used in this investigation (all dimensions in mm)
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
352
Table 2
4.1
Weld type
Restraint
Passes
Current (A)
Voltage (V)
Weld area/(wwa )
25
50
100
T-llet
T-butt
T-butt
Yes
Yes
No
8
18
90
250
170240
170240
24
2123
2123
2.02.5
2.02.5
1.82.2
0.18
0.11
0.13
Fig. 3
1a
Geometry (not to scale) and welding sequence for (a) 25 mm, (b) 50 mm and (c) 100 mm T-plate
specimens. (From reference [10])
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
Ddhkl
0
dhkl
1b
s1 sx
2
where E is the elastic modulus and n is Poissons ratio.
Note that equations (2) can be used to obtain sx , sy
and sz , regardless of whether these are the principal
stresses.
The values of E and n used in equations (2) generally
depend on the type of source used. In this work the bulk
properties E 212 MPa and n 0:3 have been used for
the data from the spallation source, as strains are
averaged over a number of lattice planes. For the
monochromatic measurements, the values used were
those for the (211) plane (which is the recommended
plane for ferritic steels [11]), E211 224 MPa, n211
0:28. These values have been obtained by taking the
average of the Voigt and Reuss solutions for the linear
elastic constants [12]. [One measurement was made
using the (110) plane rather than the (211) plane but
E110 E211 and n110 n211 so the same values are used
for all the measurements.]
353
4.2
354
Fig. 4
Table 3
T-plate weld with measurement directions indicated, where the origin of the axes is at the weld toe
Experimental parameters for measurement on T-plate welds. The sampling volumes are dened as input width 6 input
height 6 output width
Site
Plane hkl
Wavelength A
E (GPa)
25
25
25
25
25
50
100
ISIS
ILL
Studsvik
ILL
ILL
ILL
ILL
Rietveld
211
211
211
110
211
211
26261:4
26261:2
26262
26161
2:56161
26261:2
16161:2
90
109.4
93.5
109.4
119.95
109.4
109.4
Polychromatic
1.91
1.71
1.91
3.50
1.91
1.91
212
224
224
224
224
224
224
0.3
0.28
0.28
0.28
0.28
0.28
0.28
5
4.2.1
5.1
Neutron measurements
355
The average distributions from the ve sets of measurement are presented in Fig. 6. The individual data sets are
shown in Fig. 7. A modied Bayesian average of the
data was used [15, 16], which reduces the inuence of
outlying data points within a data set and therefore
provides a more realistic estimate of the true distribution. An estimate of the uncertainty of each data set
(residual stress distribution) can be obtained by comparing the data set with this average distribution to provide
an estimate of the overall uncertainty in the data. In
Table 4 the standard deviations of each individual data
set about the average are presented. The standard
deviations obtained from the neutron Bragg peak tting
routines are also provided for each data set (this is the
average for all points in each data set). It appears that
the uncertainty obtained from the peak tting routines is
Fig. 6
356
Fig. 7
Individual data sets of 25 mm T-plate weld residual stresses in (a) the transverse direction, (b) the
longitudinal direction and (c) the normal direction
Table 4
Laboratory
Standard deviation of
data set about average
(MPa)
ISIS
ILL
Studsvik
ILL
ILL
Rietveld
211
211
211
110
+14
+20
+16
+18
+26
Fig. 8
Fig. 9
5.2
357
358
Fig. 10
Transverse stresses in the 100 mm T-plate (a) obtained by DHD, (b) ND measurement on a 12.5 mm
slice and DHD measurement
6.1
Fig. 11
6.2
7.1
Fig. 12
Fig. 13
359
It is well known that there is an approximate relationship between yield strength (or ow stress) and Vickers
hardness (see, for example, reference [21]). By comparing the hardness (HV) and yield strength sYP of a wide
range of carbon steels (see, for example, reference [22]),
it was found that the ratio sYP =HV was in the range
2.4 + 0.4 sYP in MPa and HV in kgf/mm2 ). Vickers
hardness tests on the as-received plate indicate that the
appropriate sYP =HV ratio for the steel under examination in this work is 2.0.
To assess the variation in material ow stress around
the weld toe as a result of the welding procedure,
Vickers hardness tests have been carried out. Microstructural examination of the 25 and 50 mm specimens
indicates that the size of the heat-affected zone (HAZ),
where microstructural changes are most signicant, is
approximately 5 mm. Outside this region it has been
assumed that changes in ow stress are associated with
plastic deformation only and therefore any increase in
hardness may be interpreted as due to plastic deformation.
Vickers hardness measurements were taken on the
surface of the T-plate specimens along line YY in Fig. 4.
The measured hardness values were scaled by 2.0 to give
the apparent material yield strength and will be
compared with the results from the ND estimates of
plastic zone sizes, described in the next section.
7.2
Comparison of FWHM (6 1120) measurements with a typical tensile stressstrain curve of the material
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
360
Fig. 14
Comparison of variation in ow stress (yield point) estimated from FWHM and Vickers hardness
measurements: (a) 25 mm weld, (b) 50 mm weld and (c) 100 mm weld
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
Fig. 15
361
Transverse residual stress distributions measured in 25, 50 and 100 mm T-plate welds compared to
data from the literature: (a) as-measured and (b) after satisfying force balance equilibrium
Downloaded from sdj.sagepub.com at University of Limerick on July 31, 2014
362
9.1
In the R6 procedure, the recommended throughthickness transverse residual stress distribution in Tplate welded joints consists of an upper bound bilinear
function. The peak stress is at the weld toe and equal to
the parent material yield stress and reduces linearly to
zero at a distance r0 from the weld toe. The distance r0
represents the size of the yielded zone as recommended
by Leggatt [18] and can be estimated from
C Zq 1=2
r0
3
sYP u
where sYP is the parent material yield stress (or 0.2 per
cent proof strength of the parent material), Z is a process
efciency parameter, q is the arc power (in J/s) and u is
the weld travel speed (in mm/s). C is a constant that
depends on the coefcient of thermal expansion,
Youngs modulus, density and specic heat of a
material. Typical values of C and Z are included in the
R6 document for a range of materials. For ferritic steels
the values provided are C 153 N mm=J and Z 0:8,
resulting in a simplied version of equation (3):
r0
122q 1=2
sYP u
9.2
1:033C
Zq
sYP uw 0:5wa
9.3
BS 7910 provides two transverse residual stress distributions for T-plate joints. The rst follows the approach in
R6, with the distribution dependent on the size of the
plastic zone. When the plastic zone, calculated via
equation (3) or (4) is less than the base plate thickness
the residual stress is taken to be that of the parent
material yield stress level at the weld toe, reducing
linearly to zero over the size of the yielded zone (as in
Table 5
Base plate
thickness w (mm)
R6 (mm)
BS 7910 (mm)
Measured (mm)
25
50
100
23.5
29.2
27.2
25
29.2
27.2
12 + 2
20 + 5
15 + 3
363
7910(2), although the former distributions are nonconservative in the region y=w > 0:8. Finally, for the
100 mm weld (Fig. 16c), the R6 and BS 7910(1)
distributions provide a good conservative estimate for
y=w 4 0:8 with the BS 7910(2) distribution being again
somewhat overconservative.
It should be pointed out that the recommended
residual stress distributions in R6 and BS 7910 are
based on limited experimental data and indeed the heat
input, q=n, for the T-plate welds in this study is in the
range 1.8 kJ/mm 4 q=n 4 2:5 kJ= mm (see Table 2),
which is greater than the value of 1.4 kJ/mm quoted in
R6, although no such stipulation is given in BS 7910.
Furthermore, equation (6) is based on a range of
through-thickness transverse residual stress data for a
number of joints, including pipe-on-plate joints, tubular
joints and only one T-plate joint [3]. The additional
measurements presented here imply that the conservatism, in R6 and BS 7910, can be reduced for T-plate
welds.
Using a modied Bayesian approach [15, 16], the
averages of the available data before and after
equilibrium balancing are provided in Fig. 17a. Note
that the Allen et al. [4] data were not included in the
averaging procedure, as they did not follow the trends of
the other data. The magnitudes of the error bars shown
in Fig. 17a are given by +2 standard deviations about
the mean. As shown in Fig. 17a the average experimental data can be represented by a bilinear plot
starting from a stress of 0:75sYP at y=w 0, decreasing
to 0:3sYP at y=w 0:275 and increasing to 0:25sYP at
y=w 1:0 (this line captures the average distributions
both before and after equilibrium balancing and satises
force and moment balance to within +0:05sYP ). Also as
shown in Fig. 17a the upper limit of the data can be
represented by this mean curve displaced by 0:25sYP .
Comparisons of all the available T-plate residual
stress data with the parametric BS 7901(2) equation and
the new upper bound t of Fig. 17a are provided in
Fig. 17b. The R6 and BS 7910(1) distributions are not
included in this gure as they do not collapse to a single
curve when normalized in this manner (see Fig.16). It
may be seen that the BS 7910(2) distribution is
conservative in all cases but that a more accurate
estimate is provided by the upper bound line.
Fig. 16
364
Fig. 17
Transverse residual stress distributions: (a) bilinear estimations of averages and upper bounds of data
and (b) transverse residual stress distributions after satisfying force balance equilibrium compared to
upper bounds
ACKNOWLEDGEMENTS
The authors would like to acknowledge Dr M. Daymond at ISIS (UK), Dr T. Pirling at ILL and Dr R. L.
Peng at Studsvik (Sweden) for assistance with the
neutron diffraction measurements. Financial support
for the work was provided by the IMC, HSE, EPSRC
and DERA. Helpful input of the industrial sponsors, in
particular Dr R.A. Ainsworth, Dr A Stacey and Dr S.
Birley, is gratefully acknowledged.
REFERENCES
1 Milne, I., Ainsworth, R.A., Dowling, A.R. and Stewart, A.T.
Assessment of the integrity of structures containing defects.
CEGB Report R/H/R6-Rev. 4, 2001.
365