Content uploaded by Philip J. Withers
Author content
All content in this area was uploaded by Philip J. Withers
Content may be subject to copyright.
Inertia Welding Nickel-Based Superalloy:
Part II. Residual Stress Characterization
M. PREUSS, J.W.L. PANG, P.J. WITHERS, and G.J. BAXTER
The next generation of Ni-based alloys for aeroengines are richer in
␥
⬘than existing alloys and are
more difficult to weld by conventional means. Inertia welding is currently being developed as a
joining technique for these alloys. Steep microstructural gradients have been observed in nickel-based
superalloy RR1000 tube structures welded by inertia friction welding,
[1]
and in this article, the
concomitant residual stresses are mapped at depth using neutron diffraction. One tube in the as-
welded and two in the postweld heat-treated (PWHT) condition have been investigated. In the case
of the as-welded specimen, it was necessary to establish the variation of the stress-free lattice parameter,
a
0
, across the weld line to infer elastic strain from lattice spacing changes. A biaxial sin
2
measurement
on thin slices was used to determine a
0
as a function of the axial position from the weld line. This
was in excellent agreement with the variation inferred by imposing a stress balance on the axial
measurements. The change of a
0
across the weld line can be rationalized in terms of the observed
variation in the element partitioning effect between the matrix (
␥
) and the precipitates (
␥
⬘). It was
found that the residual stresses in the weld and heat-affected zone generated by the welding process
are large, especially close to the inner diameter of the welded ring. The experimental results have
shown that, in order to relax the residual stresses sufficiently, the heat-treatment temperature must
be increased by 50 ⬚C over the conventional heat-treatment temperature. This is due to the high
␥
⬘
content of RR1000.
I. INTRODUCTION elastic strain (and thereby stress) is, however, not straightfor-
ward. It was demonstrated in part I
[1]
that, under the extreme
T
HE
push towards new high
␥
⬘-containing Ni-based thermomechanical history experienced by the near-weld zone,
superalloys to achieve higher operating temperatures in aer- marked changes in the local microstructure occur. Steep gradi-
oengines has led to the development of alloys that are difficult ents in the volume fraction and size of primary (1 to 2
m),
to join by conventional means. As a solid-phase joining proc- intermediate secondary (70 to 300 nm), and tertiary (8 to 75
ess, inertia welding is a promising alternative. In many ways, it nm)
␥
⬘have been recorded. As a result, some repartitioning
resembles a localized forging process. Because of the localized of the Ti/Al solute elements between
␥
and
␥
⬘can be expected.
heat generation caused by the friction process and the very This may give rise to changes in the stress-free lattice spacing
severe cooling rates, significant residual stresses are intro- (a
0
) across the near-weld region. Such changes can mistakenly
duced. In part I,
[1]
the metallurgical development of nickel- be interpreted in terms of stress if they are not corrected for.
based superalloy, RR1000, in the near-weld region has been To date, there have been few reports on subsurface residual
described. RR1000 has a high volume fraction of
␥
⬘to achieve stress measurements of inertia welds. These have been either
better high-temperature properties than existing alloys, such limited to the axial direction
[2,3]
or fail to include the a
0
as WASPALOY.* An alloy that has better high-tempera- issue.
[4,5]
Schroeder
[6]
has reported a maximum hoop stress of
about 1000 MPa in an as-welded specimen (WASPALOY
*WASPALOY is a trademark of Precision Rings, Inc., Indianapolis, IN.
joined to IN 718*), when he applied a stress balance model.
ture strength might also be expected to retain larger residual
*IN718 is a trademark of INCO Alloys International, Inc., Hunting-
stresses when it is joined and be more resistant to subsequent
ton, WV.
postweld heat treatment.
In this article, the magnitude of the residual stresses gener- However, a
0
measurements were not reported. In this article,
ated by inertia welding and the efficiency of subsequent post neutron diffraction data, corrected to account for variations
weld heat treatment in terms of stress relief is investigated for in the stress-free lattice parameter, are reported. To achieve
RR1000 by means of neutron-diffraction measurements. This this, a
0
has been assessed as a function of axial distance
technique enables the measurement of elastic strain in bulk from the weld line using a biaxial sin
2
method. The results
materials via changes in lattice parameter, a. Unambiguous from this direct measurement are compared with those calcu-
interpretation of lattice-parameter measurements in terms of lated on the basis of axial stress balance. Stress measure-
ments are reported for both as-welded and postweld heat
treated (PWHT) specimens.
M. PREUSS, Research Fellow, and P.J. WITHERS, Professor, are with
the Manchester Materials Science Centre, University of Manchester and
II. EXPERIMENTAL
UMIST, Manchester M1 7HS, United Kingdom. Contact e-mail: michael.
preuss@man.ac.uk J.W.L. PANG, Research Fellow, is with the Oak Ridge
A. Materials and Specimens
National Laboratory, Oak Ridge, TN 37830. G.J. BAXTER, Process Metal-
The RR1000 material and the joints studied in this investi-
lurgist, is with Rolls-Royce plc., Derby DE24 8BJ, United Kingdom.
Manuscript submitted November 19, 2001.
gation have been described in part I.
[1]
In summary, two
METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 33A, OCTOBER 2002—3227
Table I. List of Specimens Measured at ISIS and ILL and
Corresponding Samples from Part I;
[1]
the Postweld Heat
Treatments of the Specimens were Carried out by Rolls-
Royce plc
Specimen Diameter
Condition (mm) Condition
S1 143 as welded
S2 143 conventional PWHT
S3 143 modified PWHT (⫹50 ⬚C)
s1 50 as welded
s2 50 conventional PWHT
s3 50 modified PWHT (⫹50⬚C)
s4 50 modified PWHT (II) (⫹150⬚C)
inertia welds of outer diameter (OD) 143 mm and wall
thickness 8 mm were provided by Rolls-Royce plc (Derby, (a)
United Kingdom). Both samples were welded under the
same welding conditions. One specimen was examined in
the as-welded condition (S1), and the other one was conven-
tionally postweld heat treated (S2). After the first series of
neutron diffraction measurements, S1 was given a modified
PWHT that was 50 ⬚C higher than S2. The code for this
new condition is reported in this article as S3.
In addition to S1 through S3, four inertia-welded tubes
of 50 mm OD were provided by Rolls-Royce plc. These
specimens were welded with the same welding parameters
(surface speed, inertia, and axial pressure) as S1 through S3
and are represented by s1 through s4. The heat-treatment
conditions of all specimens are summarized in Table I.
Because of the relatively large neutron absorption coeffi-
cient of nickel, a neutron path length of 10 mm would reduce
the diffracted beam intensity by ⬇80 pct. To minimize the (b)
path length and facilitate the hoop strain measurements,
a small window of 12 ⫻12 mm was electrodischarge
machined from the weld region of S1 through S3 ata position
distant from the neutron measurement location. A window
was not feasible in s1 through s4 because of their small
diameter.
To evaluate the extent of any variation in the stress-free
lattice parameter, a
0
, with axial distance from the weld in
the as-welded and heat-treated conditions, two samples were
cut out from a 143 mm OD sample that was welded with
the same parameters as samples S1 and S2. The cutting
plane was perpendicular to the hoop direction, and the
dimensions of the slices were 15 ⫻8⫻0.5 mm
3
(axial ⫻
radial ⫻hoop). One sample was PWHT in a tube furnace
under argon atmosphere at the same temperature as S2 and
subsequently air cooled. Thin slices were then machined
from the two samples, ground, and electropolished to remove
machine-induced stresses at the surface.
(c)
B. Neutron Diffraction Measurements
Fig. 1—Alignment of the specimen with respect to the neutron beam in
order to measure the three principal axes.
All measurements of S1 through S3 were carried out on
the ENGIN diffractometer at the ISIS neutron spallation
source, Rutherford Appleton Laboratory (Chilton, UK). At beam is shown in Figures 1(a) and (b). The direction of the
strain measurement is parallel to the scattering vector, Q,pulsed neutron sources, neutrons of all wavelengths emerge
over a short-time pulse from the source. With detectors which is the bisector of the incident and diffracted neutron
beams. The presence of two detector banks ⫾90 deg apartplaced at a given angle, due to time of flight, the whole
diffraction profile can be recorded as a function of time.
[7]
means that the strains along two perpendicular sample direc-
tions can be measured simultaneously. The sampling gageThe orientation of the specimens with respect to the neutron
3228—VOLUME 33A, OCTOBER 2002 METALLURGICAL AND MATERIALS TRANSACTIONS A
Table II. Gage Volumes Used during the Neutron
Diffraction Measurements as a Function of the
Measurement Direction
Location Strain Direction Gage Volume (mm)
ISIS radial 10 ⫻1⫻1.5
ISIS axial 10 ⫻1⫻1.5
ISIS hoop 2 ⫻1⫻1.5
ILL hoop 2 ⫻2⫻1
volume is defined by the intersection of the incident and
diffracted beams. The lattice parameter, a, of the lattice
planes with plane normal parallel to Q, averaged over the
correctly oriented grains within the sampling volume, was
determined by Rietveld refinement of the spectra.
[8]
The lattice spacing was mapped out over a plane at a
specific hoop location for S1 through S3 between the weld
line and up to 8 mm away from it. A previous experiment
at Risø (Denmark)
[9]
had indicated that the residual stresses
in a similar weld were essentially symmetric about the weld (a)
line (z⫽0). Consequently, measurements were only made
on one-half of the coordinate system (z⫹ve). The sample
coordinate system, as defined for the experiment, is shown
in Figure 1(c).
All measurements on s1 to s4 were performed at the
Institute Laue-Langevin (ILL) high-flux reactor in Grenoble,
France.Thepurposeofthesemeasurementswas to determine
an optimum temperature for a modified postweld heat treat-
ment. Because of a long counting time (up to 3 hours/point),
only the hoop strains close to the inner diameter were studied
because this is the area where the most significant stresses
were expected. This was carried out with the same orienta-
tion of the specimens and direction of the scattering vector,
Q, as described for the ISIS experiment. The wavelength
employed at the ILL was
⫽2.99 A
˚. Using the (111)
reflection of Ni, a scattering angle of about 2
⫽92 deg
was obtained, resulting in a virtually cuboidal gage volume
(Table II), as used on ENGIN. This hkl reflection has been
shown by Stone et al.
[10]
to be essentially insensitive to
plastic anisotropy.
The strain for all measurements was calculated using
⫽d⫺d
0
d
0
⫽a⫺a
0
a
0
[1] (b)
where d
0
is the stress-free lattice spacing (ILL) and a
0
(ISIS)
Fig. 2—(a) Uncorrected nominal (constant a
0
) and (b) corrected (a
0
(z))
is the stress-free lattice parameter. The preliminary nominal,
axial residual stress field (in MPa) in S1.
stress-free lattice spacing was generally determined by a far-
field measurement for each sample.
The corresponding axial, radial, and hoop residual stress C. High-Energy Synchrotron X-ray Measurements
fields for S1, S2, and S3 were calculated using Thin radial-axial plane slices were cut from an inertia
welded 143 mm OD specimen as described in part I.
[1]
These
axial
⫽E
(1 ⫹v)(1 ⫺2v)[(1 ⫺v)
axial
[2] slices were measured in transmission on beam-line ID11
at the European Synchrotron Radiation Facility (Grenoble,
⫹v(
radial
⫹
hoop
)], etc. France). Monochromatic X-rays of 60 ke V (0.203 A
˚) were
used. A slit size of 150
m in the axial and 3 mm in the
radial direction was used, centered at R⫽0 and scannedwhere E is the bulk Young’s modulus, and vis Poisson’s
ratio. As only the strain in the hoop direction of s1 to s4 along the zdirection. The (111), (200), (220), and (311)
primary reflections of the
␥
matrix and the
␥
⬘phase werewas measured, the residual stress in these specimens were
not determined. measured together with the (100) superlattice reflection of
METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 33A, OCTOBER 2002—3229
Fig. 3—The axial lattice parameter measured using four diffraction peaks Fig. 4—The variation in axial lattice parameter with distance from the
((111), (200), (220), and (311)) for a thin slice in the as-welded condition weld line of the (100)
␥
⬘superlattice reflection and the (200) (
␥
⫹
␥
⬘)
(S1) on ID11. reflection of the as-welded sample (S1) measured on beam line ID11.
the
␥
⬘phase. The measurement time of a primary reflection III. RESULTS AND DISCUSSION
was in the range of seconds, whereas the measurements of
the
␥
⬘superlattice peak took up to 45 minutes per point. A. Stress Balance
Because of the 0.5 mm thickness of the slices, the beam Figure 2(a) shows a contour map of the preliminary axial
bathed a statistically sufficient number of grains during the residual stress field calculated using Eq. [1] and [2], assum-
scanning of the samples. ing a nominal stress-free lattice parameter a
0
(constantacross
the weld line). The contour map was calculated from 12
D. Laboratory X-ray Measurements measurement points. The average uncertainty in the calcu-
Thelatticeparameter,a
3
(scatteringvector,Q, perpendicu- lated residual stresses was around ⫾60 MPa. Equilibrium
lar to the surface), across the weld line was measured with in a tubular structure requires that the axial stress should be
a small step size of 0.25 mm and at three radial positions balanced when considered over the wall thickness (the radial
(R⫽⫺2.5, 0, and 2.5) using a laboratory X-ray source. In direction). Figure 2(a) shows that the uncorrected nominal
addition, sin
2
measurements were performed for R⫽0axial stress in the area close to the weld line (z⫽0) is in
in the two principal in-plane stress directions. The X-ray tension close to the inner diameter (R⫽⫺4) and that this
diffractometer used was a Bruker AXS D8 Discover (Bruker is not counterbalanced by a comparable compressive stress
AXS, Ltd., Congleton, U.K.), which has a polycapillary close to the OD (R⫽4). Further from the weld line (z⫽5),
optics X-ray beam and a two-dimensional area detector col- stress balance is more nearly observed in the uncorrected
lecting backscattered X-ray Debye–Scherrer cones. An iron- values. This is to be expected considering the more modest
anode X-ray tube was used, and measurements were made thermal excursion experienced there. Clearly, the stress-free
on the (311) diffraction peak. The sample was correctly lattice spacing near the weld has been underestimated
positioned for each measurement point by using of a video resulting in an apparent net tensile stress. By adjusting a
0
microscope and a locating laser, which is positioned 45 deg as a function of axial position, it is possible to ensure the
to the microscope. The (311) diffraction peak was recorded balance of axial stress, and the corrected results are shown
at a 2
angle of about 128 deg with an irradiated area of in Figure 2(b). To achieve this, an inferred variation in a
0
approximately 1 mm
2
. The time taken for one measurement equivalent to an 800 microstrain error is required. This has
point was 5 minutes. resulted in essentially horizontal stress contours. Failure to
At the low surface penetration typical of lab full stage X- correct the stress-free lattice spacing in the vicinity of the
rays into nickel (15
m), an in-plane biaxial stress field (
3
weld would also lead to large errors in the inferred radial
⫽0) can be assumed. Using the sin
2
technique to measure and hoop stresses. In fact, on the weld line, the error in the
the principal inplane stresses,
1
and
2
, then allows determi- hoop direction is around ⫹600 MPa, which is a significant
nation of d
and d
⬘
with
⬘⫽
⫹90 deg and
⫽fraction of the yield stress (⬃1050 MPa).
0 corresponding to the direction of
1
. A more detailed
description of the biaxial sin
2
technique and the determina- B. Lattice Parameter Determination using High-Energy
tion of d
0
in a biaxial stress field can be found in References X-rays
11 and 12. The term d
0
can then be determined using Given the large error that would be introduced by a failure
to compensate for a variation in the stress-free lattice spacing
d
0
⫽v
1⫹v⭈d
⫹d
⬘
⫺2d
⬜
sin
2
⫹d
⬜
[3] as a function of position, it is important to corroborate the
inferred a
0
(z) variation determined in Section II, D. Inwhere Poisson’s ratio, v, is 0.27 for the (311) plane.
3230—VOLUME 33A, OCTOBER 2002 METALLURGICAL AND MATERIALS TRANSACTIONS A
(a)
Fig. 5—The a
3
profiles measured at three different radial positions for (S1)
on a laboratory X-ray machine.
part I,
[1]
the behavior of major and superlattice reflections
was reported from thin weld slices using synchrotron X-ray
diffraction. While the slices are not strictly stress-free, varia-
tions in the stress-free lattice parameter, a
0
, would be expec-
ted to be evident in the results. Figure 3 shows the lattice
parameters determined from four diffraction peaks in the
as-welded condition measured at the synchrotron beam line
ID11. The slices were measured in transmission with the
radial and axial direction of the slices perpendicular to the
incident beam. It can be seen that the profiles of all four
reflections are very similar, which indicates that intergranu-
lar stress is not responsible for the overall shift in a.Inthe
area between 0.5 and 2 mm from the weld line, a steep
increase in the lattice parameter is observed that is of the
same order of magnitude as the change required to achieve
stress balance (as previously discussed). However, exactly (b)
at the weld line, the lattice parameter falls to a value close
to the values measured in the parent material.
In addition to the main
␥
reflections, the (100)
␥
⬘super-
Fig. 6—Correction factor a
0
(z)⫽c(z)a
0
(parent) calculated on the basis
lattice reflection was measured (Figure 4). The as-welded
of imposing stress balance on the axial measurements for the as-welded
condition and as determined from laboratory X-ray measurement for (a)
sample displays a dramatic rise in the (100)
␥
⬘lattice param-
the as-received weld (S1) and (b) the conventional PWHT (S2).
eter close to the weld line, which is equivalent to 4000
microstrain. When compared to the (200) reflection, the
(100) reflection clearly varies in the opposite direction. This
suggests the presence of interphase microstresses or radical 4). In part I,
[1]
it was shown that, at the weld line, all
␥
⬘
had been in solution during the welding process. Only verychanges in element partitioning in the near-weld region.
However, they are not exactly balanced because the (200) fine, tertiary
␥
⬘was observed, and this supports the assump-
tion that significant supercooling took place before
␥
⬘beganreflection consists of the (200)
␥
and (200)
␥
⬘peak, which
cannot be deconvolved. Because of the steep temperature to precipitate. Consequently, in the as-welded condition,
␥
⬘
at the weld line has a different chemical composition to thatgradients introduced in the near-weld region,different maxi-
mum temperatures are experienced as a function of axial of the parent material and, thus, a different lattice parame-
ter, a.position during the welding process. With rising maximum
temperature, an increasing element partitioning effect
between the matrix and the coherent precipitates might be C. Conventional Laboratory X-ray Measurements
expected, which could explain the shift of a. Indeed, Blavette
et al.
[13]
investigated the precipitation of
␥
⬘phase in nickel- The same slices used for the synchrotron measurement
were also used for conventional X-ray measurements. Thebased superalloys by atom-probe techniques and showed
that coarse and finer
␥
⬘formed at different temperature lattice parameter, a
3
, determined from the (311) reflection
using a laboratory X-ray diffractometer, is plotted for theranges and have different levels of aluminum and titanium
content. Exactly at the weld line, a
100
rises sharply (Figure as-welded sample against the axial position in Figure 5.
METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 33A, OCTOBER 2002—3231
Fig. 7—Map of residual stress field (in MPa) of S1 (as-welded) with an axially variant stress-free lattice parameter.
Fig. 8—Map of residual stress field (in MPa) of S2 (conventional PWHT) using an axially invariant stress-free lattice parameter.
The measurements were performed at three different radial drop of the lattice parameter at the weld line in the former
is hardly detectable in the laboratory X-ray measurementpositions to detect any changes in lattice parameter in the
radial direction (this was assumed to be negligible for the because of the larger spot size in the axial direction of
the X-ray machine (1 mm
2
) compared to ID11 (150
m).stress balance approach). As can be seen, the observed varia-
tion is essentially independent of radial position. Conse- However, the X-ray beam size of the laboratory machine is
comparable to the gauge size of the axial direction duringquently, a
0
can be taken to vary only as a function of axial
position and validates the assumption that underpins the neutron diffraction experiments at ISIS.
The lattice parameter, a
3
, should not be used to determinestress balance approach. This is not totally unexpected
because the thermal history experienced by the weld region the stress-free lattice parameter across the weld line.
Although small slices were used for these measurements,is known to vary sharply with zbut weakly with R. The
lattice parameter calculated from the (311) reflection here the slices are not stress free. However,because of the shallow
penetration of the X-rays and the limited thickness of thehas a similar profile across the weld line to that observed
during the synchrotron X-ray measurements. It has to be slices (0.5 mm), a biaxial stress field (
3
⫽0) can be
assumed.noted that, in the 60 keV high-energy synchrotron X-ray
experiment (through slice thickness), the lattice parameters Totake the biaxial stress field into account while determin-
ing a
0
, a rigorous biaxial sin
2
measurement was carriedwere measured in the axial (in-plane) direction, whereas
using the 8 keV (⬇15
m penetration) laboratory X-ray out. The a
0
measurements were used to calculate a correction
factor, (a
0
(z)⫽c(z)●a
0
(parent)), by normalizing the meas-source, the hoop (out-of-plane) direction was measured. The
3232—VOLUME 33A, OCTOBER 2002 METALLURGICAL AND MATERIALS TRANSACTIONS A
ured values by the stress-free lattice parameter obtained in can be directly calculated from the stress-free lattice parame-
ter measured on the parent material during the neutron dif-the parent material. Figure 6(a) shows the variation in the
correction factor against axial distance from the weld line fraction experiment.
obtained from the biaxial sin
2
measurement and that
inferred from the stress-balanced model. As can be seen, D. Residual Stresses
excellent agreement for the as-welded sample was obtained Given that a reliable point-by-point measure of the stress-
between both a
0
correction methods, which justifies the sim- free lattice parameter is obtained, the residual strains and
ple method of balancing the axial stress to account for the stresses in the as-welded and PWHT samples can be calcu-
axial variation in a
0
when determining the stress state. lated. Figure 7 shows the residual stress field of S1 with the
The change of a
0
across the weld line in S2 was also axial stress forced into balance by correcting a
0
for each
measured by the biaxial sin
2
method. Figure 6(b) compares axial position. The maximum residual stress is in the range
the correction factor against the axial position obtained from of 1500 MPa and can be observed in the hoop direction
the biaxial sin
2
measurement and the forced stress balance close to the inner diameter (R⫽⫺4). The axial residual
model. Both methods suggest that no significant chemical stress field exhibits a significant bending moment, which
(stress-free) variation of a
0
is detectable with distance from indicates that either the tooling or uneven heating also plays
the weld line. This is perhaps to be expected given the a key role during the welding process.
[9]
As one would
extendedtime at the elevated temperature for chemical repar- expect, the radial stress component is negligible across the
titioning between
␥
and
␥
⬘. As a result, the strain and stresses weld line because of the short distance between the inner
and outer diameter. The uncorrected residual stress maps
have been published previously
[5]
and comparison of the
results shows the significance of the a
0
correction. When
the chemical shift of a
0
is not taken into account, residual
stresses close to the weld line are overestimated (by 600
MPa).
In Figure 8, the residual stress field is plotted for S2.
The contour maps were calculated from 20 measurement
positions with an average stress uncertainty of ⫾60 MPa.
The variations in the three components of the residual stress
field are very similar to S1, except that the stress magnitudes
are reduced by about 30 pct. The maximum stress in the
hoop direction is around 1000 MPa in S2. As already shown,
the axial stresses in radial direction are close to stress balance
in the PWHT specimen without invoking an axially variant
lattice parameter. This justifies the direct use of the stress-
free lattice parameter obtained from the far-field measure-
ments during the neutron diffraction experiment.
The measurement of S2 has shown that the conventional
postweldheattreatmentisnot completely effective.InFigure
9, the hoop strain against the axial position near the key
inner-radius location (R⫽⫺2.5) is plotted for specimens
Fig. 9—Hoop strain (in 10
⫺6
) recorded at R⫽⫺2.5 of the 50 mm diameter
variously heat-treated specimens measured at ILL.
s1, s2, s3, and s4 measured at the ILL in Grenoble. The
Fig. 10—Map of residual stress field (in MPa) of S3 (modified postweld heat treatment).
METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 33A, OCTOBER 2002—3233
measured strain for s1 is plotted using the constant and the 4. The residual stresses introduced during welding are very
large. The maximum tensile stress close to the inneraxially variant corrected a
0
values. For the correction of a
0
,
the c(z) factors from the 143-mm diameter specimen were diameter in the hoop direction is ⬇1500 MPa. Tooling
also plays a key role in the development of residualused. When the strains of s1 (a
0
corrected) are compared
with s2, a 25 pct reduction in strain is apparent and in good stresses as inferred from the large axial stresses in the
welds.agreement with the measurements of S1 and S2. A 50 ⬚C
rise in PWHT (s3) reduces the hoop strain significantly 5. The conventional postweld heat treatment relieves the
residual stresses only to a limited extent. The maximum(75 pct). When heattreated ata temperature of 150⬚C above
the conventional heat-treatment temperature, s4, the hoop stress is still in the range of 1000 MPa for the hoop
component.strain is completely relieved. The pronounced reduction in
hoop strain in s3 led to the ⫹50 ⬚C heat treatment of the 6. Increasing the postweld heat treatment temperature by
50 ⬚C led to a pronounced reduction of the axial and143 mm specimen (S3) to measure all three principal direc-
tions and to calculate the residual stress field. Figure 10 hoop stresses. The maximum hoop stress observed does
not exceed 400 MPa. The large differences between theshows the contour maps for S3 calculated from 15 measure-
ment points with an average stress uncertainty of ⫾60 MPa. inner and outer diameter stresses of the hoop component
have vanished.It can be seen that the stresses in the axial and hoop direction
are significantly lower than in S1 and S2. The maximum
stress in the hoop direction is less than 400 MPa. The axial
stresses, which were significant in S1 have also been largely ACKNOWLEDGMENTS
relieved. It is also interesting to note that the hoop stresses The authors thank Dr. M.R. Daymond and J. Wright (ISIS,
inS3attheinner and outer diameter are very similar,whereas ENGIN), Dr. A. Terry and Dr. G. Vaughan (ESRF, ID11),
in the other two conditions this was not the case. Dr. T. Pirling (ILL), and Judith Shackleton (MMSC) for
experimental assistance and Peter Wilson, Kevin Bass, and
IV. SUMMARY AND CONCLUSIONS A. Thompson (Rolls-Royce plc.) for advice. These experi-
The residual stresses in the as-welded (S1), conventional ments were performed under EPSRC Project No. GR/
(S2), and modified heat-treated (S3) RR1000 inertia welds M68704 and are financially supported by EPSRC and Rolls-
have been mapped. RR1000 is a high-strength nickel-based Royce plc.
superalloy containing a volume fraction of almost 50 pct
␥
⬘, when fully annealed. In addition to the stress-field meas- REFERENCES
urements by means of neutron diffraction, we have studied
the variation of a
0
across the weld line using a combination
1. M. Preuss, J.W.L. Pang, P.J. Withers, and G. Baxter: Metall. Mater.
of synchrotron radiation and a laboratory X-ray source. The
Trans. A, 2002, vol. 33A, pp. 3215-25.
conclusions can be summarized as follows.
2. H.G. Priesmeyer and J. Schroeder: Residual Stresses—III: Science
and Technology, H. Fujiwara, T. Abe, and K. Tanaka, eds., Elsevier
1. When the residual stress field for as-welded specimens
Applied Science, London, 1992, pp. 253-58.
is calculated from strain measurements (on the basis of
3. M.E. Abdel Moniem, S.M. Serag, A.A. Nasses, and A.A. Moustafa:
Wear, 1981, vol. 70, pp. 227-41.
a constant a
0
evaluated from far-field measurements) the
4. H.U. Baron: 3rd Int. Conf. on welding in Aerospace Industry, German
tensile residual stresses close to the weld line are overesti-
Welding Society, Essen, 1993, pp. 114-17.
mated because of a chemical shift (increase) of a
0
in the
5. J.W.L. Pang, M. Preuss, P.J. Withers, and G.J. Baxter: 6th Conf. on
heat-affected zone.
residual stresses (ICRS-6), IOM Communication, London, 2000, pp.
2. The chemical-induced increase in a
0
in the as-welded
1415-21.
6. J. Schroeder: Ph.D. Thesis, Technische Universitaet Hamburg-
condition is due to an element-partitioning effect between
Harburg, Hamburg, 1993, p. 119.
␥
and
␥
⬘in the heat-affected zone. This is evidenced by
7. M.W. Johnson, L. Edwards, and P.J. Withers: Physica B, 1997, vol.
the shift of the (100) superlattice reflection and the (200)
234, pp. 1141-43.
reflection in opposite directions.
8. R.A. Young: Rietveld Method, Oxford University Press, Oxford United
Kingdom, 1993.
3. To determine the correct a
0
shift in the axial direction,
9. J.W.L. Pang, G.J. Baxter, P.J. Withers, G. Rauchs, and M. Preuss: Oak
a biaxial sin
2
measurement was carried out. This tech-
Ridge National Lab., Oak Ridge, TN, unpublished research, 2001.
nique enabled us to take into account the residual stresses
10. H.J. Stone, T.M. Holden, and R.C. Reed: Scripta Mater., 1999, vol.
in thin slices when determining a
0
. The measured shift
40, pp. 353-58.
of a
0
with position was invariant with radial position.
11. I.C. Noyan and J.B. Cohen: Residual Stress, Measurement by Diffrac-
tion and Interpretation, Springer-Verlag, Berlin, 1987, pp. 122-25.
As a result, it was in excellent agreement with the a
0
12. L. Edwards, J. Santisteban, M.E. Fitzpatrick, G. Bruno, A. Steuver,
shift calculated from the stress balance correction. A
P.J. Withers, M.R. Daymond, and M.W. Johnson: 6th Conf. on Residual
significant shift ofa
0
was only observed in the as-welded
Stresses (ICRS-6), IOM Communication, London, 2000, pp. 1239-46.
sample, whereas the a
0
shift in the PWHT sample can
13. D. Blavette, A. Bostel, and J.M. Sarrau: Metall. Trans. A, 1985, vol.
16A, pp. 1703-11.
be neglected.
3234—VOLUME 33A, OCTOBER 2002 METALLURGICAL AND MATERIALS TRANSACTIONS A