- Email: [email protected]
Residual stresses in laser welded ASTM A387 Grade 91 steel plates
Materials Science & Engineering A 575 (2013) 160–168
Contents lists available at SciVerse ScienceDirect
Materials Science & Engineering A journal homepage: www.elsevier.com/locate/msea
Residual stresses in laser welded ASTM A387 Grade 91 steel plates Santosh Kumar a,n, A. Kundu b, K.A. Venkata c, A. Evans d, C.E. Truman c, J.A. Francis e, K. Bhanumurthy a, P.J. Bouchard b, G.K. Dey a a
Bhabha Atomic Research Centre, Mumbai, Maharashtra 400094, India Materials Engineering, The Open University, Milton Keynes, MK7 6AA, UK c Department of Mechanical Engineering, University of Bristol, Bristol, BS8 1TR, UK d Institut Laue Langevin, Grenoble, France e University of Manchester, Manchester, M13 9PL, UK b
art ic l e i nf o
a b s t r a c t
Article history: Received 28 November 2012 Received in revised form 11 March 2013 Accepted 13 March 2013 Available online 27 March 2013
Residual stresses in 9 mm thick ASTM A387 Grade 91 steel plates, joined using constant power (8 kW) low and high heat input laser welding processes, are characterised using neutron diffraction. The measured longitudinal and normal components of residual stress show a bimodal distribution across the welded joint with a low tensile or compressive trough at the weld centre flanked by high magnitude tensile peaks in parent metal adjacent to the heat affected zone boundaries. The width of the central trough and spread of the outboard tensile zones are significantly greater for the high heat input weld. In both cases, the stress distributions can be explained by the strains associated with the austenite to martensite solid-state transformation as the joint cools after welding. & 2013 Elsevier B.V. All rights reserved.
Keywords: Laser welding ASTM A387 Grade 91 Neutron diffraction Residual stress
1. Introduction Ferritic-martensitic steels, and in particular Grade 91 steel (9Cr1Mo, V, Nb), find extensive applications in supercritical power generation plants, nuclear power systems and in the petrochemical industry owing to their excellent combination of elevated temperature strength, thermal fatigue resistance and immunity from stress corrosion cracking in aqueous and chloride environments [1,2]. Moreover they are candidate materials for structural components in many advanced reactor systems like the Very High Temperature Reactor and the Sodium Fast Reactor owing to their low susceptibility to irradiation embrittlement [3]. Different welding techniques such as Manual Metal Arc welding, Submerged Arc Welding and Gas Tungsten Arc Welding are used to fabricate components and assemblies from ferritic-martensitic steels of this type [2–7]. Laser welding is a versatile joining process with high power density and low thermal input. It can weld different alloys from sub-millimeter up to 18 mm section thicknesses in many joint configurations without the need for any filler material. Major advantages of the laser welding process include the narrow width of the fused zone and heat-affected zone (HAZ), and minimal distortion of the components being welded owing to the low heat input. There is also some evidence to suggest that welds made
n
Corresponding author. Tel.: +91 22 25595695; fax: +91 22 25505151. E-mail addresses: [email protected], [email protected] (S. Kumar).
0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msea.2013.03.046
with high-energy beam processes such as laser or electron beam welding may be more resistant to type IV cracking than welds made with conventional arc welding processes [8,9]. Very high values of residual stress approaching the yield strength of the parent material and weld filler are generally introduced in the vicinity of welded joints. Temperature excursions associated with the thermal cycle of a welding pass induce localized thermal dilation that is accommodated by the elastoplastic responses of the structure. In turn this is accommodated by cyclic stress–strain behaviour of the materials, the self-restraint of the structure and any applied restraint. In addition, for material systems like P91, the strains associated with the solid-state austenite to martensite transformation of the material in the fusion zone and the HAZ can have a profound impact on the residual stress profile [10–12]. Quantitative characterisation of the residual stress distribution around the welded joint is important, to understand the factors governing its evolution, to assess how such stresses may affect the life and integrity of components inservice, and also to design mitigation treatments including optimization of the post-weld heat treatment (PWHT). Residual stress around a weld joint can be measured by surface based techniques such as centre-hole drilling, magnetic methods and X-ray diffraction or volumetric techniques such as the inherent strain method, deep hole drilling, the contour method, neutron diffraction or synchrotron diffraction [13]. Surface based techniques provide limited information about the residual stress field, whilst volumetric strain relief methods are invasive or destructive.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
161
different welding speeds, 1.5 m/min (A) and 0.75 m/min (B), to produce welded joints with significantly different heat inputs. Helium was used as the shielding gas with a flow rate of 20 l/min. Throughout the paper the translational direction of welding is referred as the plate longitudinal direction. The normal direction is perpendicular to the welding direction passing through the thickness of the plate and the transverse direction is in the plane of the plate and perpendicular to the other two directions. Neutron diffraction for residual stress measurements were carried out at the ILL, Grenoble, France on the SALSA instrument. The neutron source at this facility is an experimental nuclear reactor with 58 MWth power. All measurements were made using the (211) crystal reflection of the material, with a monochromatic beam of wavelength λ ¼1.648 Å. This wavelength resulted in scattering angles close to 901, thereby giving essentially cuboidal gauge volumes. Source and detector slit combinations were chosen to define gauge volumes in the sample small enough to capture the rapid variations in strain; expected across the weld and in the through-thickness direction of the plate. For measurements of lattice spacing in the longitudinal direction, a gauge volume of 0.8 mm 0.8 mm 2 mm was chosen, and an extended volume (in the longitudinal direction) of 0.8 mm 0.8 mm 20 mm was chosen for measurements of lattice spacing in the transverse and normal directions. Measurement points were placed closely together (at 0.5 mm spacing) near the weld line and sparsely thereafter. The lines along which neutron diffraction measurements were made are shown schematically in Fig. 1 (a) and (b), respectively, for welding conditions (A) and (B).
Neutron diffraction was selected for the present work for several reasons. First, it is a non-destructive method that can provide quantitative information about three or more components of the stress tensor in structures made from steel up to a few tens of millimeters thick [14]. Second, a fine gauge volume can be selected and this is essential to resolve the short length-scale variations in residual stress expected in narrow width laser welds. Third, the neutron diffraction technique is well established having been applied widely to characterise residual stresses in welded joints [10–27]. Measurements of the residual stress fields around P91 welded joints by different methods, including neutron diffraction, have been reported by several workers [10–12]. However, there is no published work on residual stress measurements in laser welded joints in this material. The present work provides a detailed analysis of residual stress measurements by neutron diffraction. Keyhole laser welds are also of interest as they are expected to show a residual stress profile that is significantly different from that introduced by conventional multi-pass welding processes. This is on account of the high through-thickness to width aspect ratio of the fusion zone and HAZ and because they are single pass welds. The results reported here will provide a set of high quality measured residual stress data that can be used to validate modeling and simulation endeavors. Such models can be harnessed to explore the effects of different processing conditions on the residual stress profile, as well as providing the starting conditions for PWHT stress relaxation studies.
2. Experimental details Hot-rolled ASTM A387 Grade 91 steel plates, 10 mm thick, were procured from M/s Industeel, Belgium in the normalized (30 min at 1050 1C) and tempered (30 min at 770 1C) condition. The chemical composition of the as-received material, which had a tempered martensitic microstructure, is given in Table 1. Dilatometry was performed using a TD 5000S dilatometer to characterise the austenite to martensite solid state phase transformation on cooling, as the associated strains are expected to have a significant bearing on the residual stress profile. For this purpose, a cylindrical pin (length 14.91 mm and diameter 4 mm) was used as the sample and a cylindrical alumina pin was used as the reference. The sample and the reference pins were heated simultaneously at a rate of 10 1C/min from room temperature to 1060 1C; soaked at 1060 1C for 30 min and cooled to room temperature at 10 1C/min. The supplied plates were prepared for laser welding by wire electro-discharge machining (EDM) so that they measured 500 mm 70 mm 10 mm. The thickness of the plates was reduced by further machining to approximately 9 mm (8.7 mm to 9.2 mm) to remove surface undulations caused by rolling. Laser welding was carried out using a high power continuous CO2 laser with maximum power of 10 kW and beam quality factor K ¼0.52. Optics with a focal length of 280 mm were used, to produce a focused beam diameter of 500 μm. The parent material (PM) plates were arranged in a square butt configuration with the gap between the parting surfaces measuring less than 100 μm, and rigidly held using pneumatically controlled clamps. Spot welds were made along the weld seam (length 500 mm) at three equidistant locations. Laser welding was carried out (without preheating the plates) at 8 kW power and with two
Fig. 1. (a): Macrograph showing Plate A joint cross-section (welded at 8 kW and 1.5 m/min) with neutron diffraction measurement lines marked. Measurement lines are 1.5 mm and 4.5 mm below the left top surface. (b): Macrograph showing Plate B joint cross-section (welded at 8 kW and 0.75 m/min) with neutron diffraction measurement lines marked. Measurement line is 1.5 mm and 4.5 mm below the left top surface.
Table 1 Chemical composition of ASTM A387 Gr 91 steel in wt.% (Balance Fe). C
Mn
Zr
Si
P
S
Cr
Mo
Ni
Cu
Al
N
Nb
Ti
V
0.106
0.443
0.005
0.221
0.018
0.0008
8.965
0.901
0.212
0.045
0.010
0.0464
0.073
0.004
0.194
162
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
Values of the stress free lattice spacing, d0, for (211) planes were required to determine the measured residual strains and stresses. Invariably, it is difficult to determine reliable values of the stress-free lattice parameter in the vicinity of narrow weldments. Several approaches that can be used have been documented by Withers et al. [14]. In the present case, the very narrow width of the fusion zone ( 1.2 mm) and HAZ ( 0.6 mm) means that steep changes in d0 were expected. Therefore, it was decided to use a “comb sample” spanning the weld as illustrated in Fig. 2. The comb was machined from a cross-sectional slice of material approximately 5 mm thick that had been extracted from near the end of the laser welded plate. Four transverse “prongs” spanning the weld line, each about 2 mm thick, were created by making three EDM cuts through the thickness of the slice as shown in Fig. 2. Whilst such samples are often considered to be free of “macrostress”, significant remnant stresses in a similar comb extracted from an aluminium alloy plate have been reported [28]. This may be even more important in the present case because of the short residual stress length-scales involved. To mitigate this potential uncertainty, strains were measured in three orthogonal directions in the comb prongs and a state of plane stress (for the normal plane) was assumed to exist within the prongs in order to derive values for d0. Measurements for the d-spacing in the three orthogonal directions were made at different points across the weld in the top most prong. A gauge volume of 0.8 0.8 2.0 mm was used for measurements in the longitudinal and normal directions and an extended gauge volume of 0.8 0.8 5.0 mm was used for measurements in the transverse direction. Stress-free lattice parameter (d0) values were derived assuming a plane stress condition in the topmost prong as described below. The stress along a principal direction x can be written as sx ¼
E ð1−vÞεx þ vðεy þ εz Þ ð1 þ vÞð1−2vÞ
ð1Þ
where si and εi are stress and strain components in principal directions i¼x, y and z, and E and ν are the crystallographic elastic moduli for the (211) planes. If sx is relieved by cutting a thin strip normal to the x-direction; then E ð1−vÞεx þ vðεy þ εz Þ ¼ 0 ð1 þ vÞð1−2vÞ and making the substitution gives
d0 ¼
εx ¼
dx −d0 d0
ð1−vÞdx þvðdy þdz Þ ð1þvÞ
ð2Þ ð3Þ ð4Þ
This expression was used to derive d0 values for the fusion zone and the HAZ. The approach assumes isotropic material properties (i.e. low texture) and that values of d0 are direction-independent (i.e. negligible microstresses are present owing to the influence of plastic strain). For the parent metal, the average d0 value of ten different
Fig. 2. Sketch showing the comb sample used for d0 measurements.
measurements on a small piece of material cut from the far end of the plate was used. This value was 1.17241570.000055 Å. The diffraction angle, 2θ, determined from the measured position of the diffraction peak, was used to derive the lattice spacing, d, using Bragg's Law, λ ¼ 2dsinθ The lattice spacing along three orthogonal directions of the welded plate – longitudinal, transverse and normal – were calculated using the corresponding diffraction peaks from different measurement points. These values were then used to calculate strain values εx, εy and εz in the three directions using suitable d0 values as described above. Using these strain values, the residual stress along the three orthogonal directions—sx, sy and sz were calculated for all the measurement points using Eq. (1). Errors in the individual lattice spacing measurements were derived from the error in the location of the diffraction peak centre, arising from the Gaussian fit. This exercise was carried out for all the measurements on the plate as well as on the combs. The error propagation was applied in the derivation of all values for d0 (using a plane stress assumption) and for values of residual strain and residual stress along the three principal directions. A cross-section of each welded joint was prepared for metallographic evaluation by first polishing and then etching with Vilella's reagent in order to reveal different metallurgical regions—the parent metal, the heat affected zone and the fusion zone, and to reveal the detailed microstructure within each of these regions. Microhardness measurements were also carried out across the welded joint on a polished and lightly etched cross-section, using a 5 N load and a 10 s dwell time. The measurement points were 1.5 mm below the top surface of the welded plate and at a pitch of 0.2 mm, in order to capture the rapid variation in the microhardness through the heat affected zone, and also to ensure that the microhardness measurements were carried out at locations similar to those where neutron diffraction measurements were made. Optical microscopy was employed to obtain representative micrographs from the parent metal, the heat affected zone and the fusion zone, both for plate A and for plate B.
3. Results and discussion Polished and etched cross-sectional macrographs of the laser welded plates A and B (fast and slow advance rates) are presented in Fig. 1(a) and (b). In both cases there is a small geometric mismatch across the weld. The width of the fusion zone in the high speed weld (low heat input) is about 1.2 mm as compared with about 1.8 mm for the slow speed (high heat input) weld. The location of the neutron diffraction measurement lines are marked on the macrograph of these welds (Fig. 1(a) and (b)). 3.1. Microstructural evaluation The microstructures for the different regions—the parent metal, the heat affected zone and the fusion zone for plates A and B are presented in Fig. 3(a) through to Fig. 3(e). The parent metal (Fig. 3 (a)) showed a tempered martensitic structure within equiaxed prior austenite grains of 20 μm in size. The heat affected zones in both the plates A and B showed an as-transformed martensitic structure within reaustenitized grains of varying size. In the heat affected zone, reaustenitization of the tempered martensitic structure (present in the parent metal) occurs due to thermal cycles associated with the welding process provided that at some point the peak temperature exceeds the Ac1 temperature of the material. The size of the reaustenitized grains increases with an increase in the peak temperature that is experienced by the
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
163
Fig. 3. (a): Microstructure of the parent metal. (b): Microstructure of the heat affected zone in Plate A. (c): Microstructure of the fusion zone in Plate A. (d): Microstructure of the heat affected zone in Plate B. (e): Microstructure of the fusion zone in the Plate B.
respective region. The regions farther from the fusion line experience lower peak temperatures and therefore form smaller austenite grains, since there is limited scope for grain growth while the material exists as austenite. The heat affected zone in P91 steel welds can be further sub-categorised according to the extent to which austenitisation occurs and according to the peak temperature that is experienced during the welding thermal cycle. Adjacent to the unaffected parent material the heat affected zone contains a region that is subcritical (SC), and as the distance from the fusion line decreases the next sub-zone is intercritical (IC) or partially austenitised, followed by a fine grained (FG) region and finally a coarse grained (CG) region adjacent to the fusion line [12]. However, the temperature gradients that arise during laser welding are very steep, and the heat affected zones are very narrow. This places significant limitations on the extent to which grain growth can occur in the heat affected zone, since the region that experiences a high enough peak temperature for grain growth to occur is itself very narrow and such temperatures arise only for very short durations. This was evident in these laser welding experiments since no discernible coarse-grained heat affected zone was observed in plate A or plate B. The fusion zones in both the welded joints showed an as-transformed martensitic structure within columnar prior austenite grains. Columnar grains arise frequently in weld fusion zones, and they are a consequence of grains growing at 901 to the fusion boundary, in a direction that is
opposite to the direction of maximum heat flow. There were no significant differences in the microstructures of the two plates welded with different welding speeds.
3.2. Thermal dilation measurement The measured dilation curve for the parent Grade 91 material (PM) is shown in Fig. 4. During heating at 10 1C/min, the steel expands until it reaches the Ac1 temperature ( 825 1C), where significant contraction occurs due to ferrite to austenite transformation. This contraction continues until the Ac3 temperature is reached, but at 845 1C a significant part of the ferrite has already transformed to austenite and thermal expansion of austenite becomes dominant. A small expansion is evident during the thirty minute soak at 1060 1C which can be attributed to dissolution of carbides leading to an increase in the lattice spacing of the austenite. During cooling, there is a contraction of the austenite until it reaches the Ms Temperature ( 400 1C), where significant expansion arises due to solid-state transformation of austenite into martensite until Mf is reached ( 240 1C). Subsequently, the martensite contracts until room temperature is reached. These phase transition temperatures show consistency with those reported in the literature by using differential scanning calorimetry (DSC) for this material [29].
164
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
Soaking (30 min
0.014
Thermal Dilation (mm/mm)
0.012
0.01
0.008
0.006
0.004
0.002
0 0
200
400
600
800
1000
Temperature (°C)
Fig. 4. Measured thermal dilation of ASTM A387 Grade 91 steel (heating and cooling rate 10 1C/min).
0.51 for the parent metal and increases rapidly through the narrow HAZ to become as high as 1.41 for the longitudinal component and 11 for the transverse and normal components in the welded region. The regions of higher FWHM show a one-toone correlation with the regions which have undergone a martensitic transformation due to the welding thermal cycle. The correlation is so strong that the FWHM may be used as a signature for the fusion zone (FZ) and HAZ regions, which have undergone a martensitic transformation during welding. The higher values of FWHM of the diffraction peaks are due to a variation of strains on the microscopic length scale within the gauge volume. Such variations can arise from plastic deformation during cooling and/ or the solid state phase transformation. However, further experimentation and simulation work is needed to confirm and elucidate the contributing mechanisms. The FWHM of the diffraction peak in the weld region for the longitudinal component is significantly higher ( 40%) than that for the transverse and normal components. This could be due to texture in the prior austenite grains leading to the selection of different variants along the three measured directions—longitudinal, transverse and normal. This can however, be confirmed only with a detailed EBSD (Electron Back Scattered Diffraction) analysis. It may also be due to the smaller gauge volume used for the longitudinal measurements compared with that used for measurements of the other two components, noting that larger gauge volumes tend to give sharper diffraction peaks than smaller volumes assuming other parameters such as count time, material chemistry and crystal structure etc. remain the same. Whilst a sharp diffraction peak is preferred, as it implies a lower error, a small volume had to be employed for measuring the longitudinal component in order to capture the steep gradients in stress across the weldment.
3.4. Cross-weld strain and stress profiles in Plate A at 1.5 mm below the top surface
The strain associated with the martensitic transformation on cooling has very significant impact on residual stress profile for single pass welds. After the melt zone is solidified, continuity is established and the development of the residual stress field is influenced primarily by the localized contraction of the hot zone in and around the welded joint and resistance associated with geometric self-restraint of the joint and any applied restraint. But once the Ms temperature is reached in P91 steel, a significant strain ( 2600 micro strain in the absence of an external stress (Fig. 4)) occurs owing to the martensitic transformation, until the Mf temperature is reached, and this strain is expected to have a strong impact on the residual stress profile. Thus, for single pass welds, the evolution of the residual stress field is driven by competing contributions from thermal contraction during cooling and the strains associated with the solid-state phase transformation. 3.3. Diffraction peak FWHM analysis The measured variation in the full width at half maximum (FWHM) of the (211) diffraction peak across the weld for measurements of d-spacing in the longitudinal, transverse and normal directions at mid-length of Plate A is shown in Fig. 5. A cross-weld microhardness profile from a similar location is also imposed in this figure. A similar pattern was observed in all cross-weld measurement scans. The variation in FWHM is symmetric about the weld centre line and exhibits an interesting trend. It remains
The measured residual strain profile, across the low heat input weld in Plate A, at 1.5 mm depth below the top surface is shown in Fig. 6. As expected, the three components (longitudinal, transverse and normal) of residual strain show a high degree of symmetry across the weld centre-line. Interestingly the longitudinal component shows a bimodal distribution with a trough coincident with the weld centre-line and the two peaks lying 2 mm on either side of the centre in the parent metal near the HAZ–PM interface. The peak value of the longitudinal component of residual strain is 1900 micro-strain. The normal component shows a similar 2000
1400
800
Longitudinal Transverse Normal Fusion Line HAZ Boundary
Microstrains
Fig. 5. Variation of full width at half maximum (FWHM) of the (211) diffraction peaks across the laser weld joint in Plate A at 1.5 mm below the top surface. Crossweld microhardness profile is also imposed.
200
-400
-1000 -10
-8
-6
-4
-2
0
2
4
6
8
10
Distance from the weld centreline (mm)
Fig. 6. Cross-weld residual strain profile in Plate A (8 kW, 1.5 m/min) at 1.5 mm below top surface.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
profile, but with a lower peak magnitude ( 500 micro-strain) at 1.5 mm on either side of the weld centre-line. The transverse component of residual strain, however, shows the opposite trend. It has a peak ( 650 micro strain) coinciding with the weld centreline and two troughs ( −900 micro-strain) at 1.5 mm on either side of the weld centre-line. The same characteristics are seen in the residual stress profile (Fig. 7). The longitudinal component of residual stress shows a low tensile trough ( 50 MPa) in the weld centre which rises rapidly to a peak value of 500 MPa at 2 mm on either side of it. The regions with peak residual stresses are in the parent metal just outside the metallurgical HAZ, as is evident from a comparison of the crossweld microhardness profile and the residual stresses at the same location. Similar results have been reported by Mark et al. [30] for residual stress profiles across a single pass weld in SA508 steel, which also undergoes a bainitic/martensitic transformation while cooling from the austenite phase field. The normal component shows a similar trend, with a trough in the weld centre ( 100 MPa) and rising rapidly to peak values ( 200 MPa) at 1.5 mm on either side of the centre-line. The transverse component shows the opposite trend with the peak ( 200 MPa) coinciding with the weld centre and two troughs ( 710 MPa) at 1.5 mm on either side of it. The measured strain and stress profiles are almost completely symmetric. This gives confidence in the quality of results, and the experimental procedure, which involved remounting of the sample after measurements of the longitudinal component of strain, to obtain corresponding data in the other two orthogonal directions. Thus, further measurements were focused on only one side of the welded joint. The trough in the longitudinal residual strain and stress profiles at the weld centre-line can be attributed to strains associated with the martensitic transformation, which includes a volumetric expansion, during cool down of the region of the weldment which was heated to above the Ac1 temperature; that is the entire fusion zone and the HAZ. The tensile peaks are in regions heated to a temperature just below the Ac1 temperature, as these regions experience only contraction during cooling of the welded joint. This M-shaped profile contrasts with a non-transforming material where a single tensile peak would be expected at and near the weld centre-line, dropping to low compressive stresses in the far field [26]. However, in P91 steel, martensitic transformation strains have been observed to have a similar influence on the residual stresses in a multi-pass pipe girth weld, where compressive stresses were measured in the region of the last weld pass, and the proposed influence of the transformation strains was supported by the results of a Satoh test [12].
165
Another feature is observed in the profile for the normal component of residual stress, which has the same shape as the longitudinal profile, but much lower magnitude. The common shape of the profiles suggests that the factors controlling the evolution of the two are similar in nature and possibly related to the strains associated with the martensitic transformation occurring under constraint. Because of the large aspect ratio (depth to width) of the laser weld fusion zone, the thickness of the weld bead (i.e. the plate) provides some restraint against the dilatational component of the strain. In the longitudinal welding direction, the effective restraint length is much longer giving a larger compressive effect. The adjacent tensile peaks in the longitudinal and normal components of the stress profile may be viewed as reactions to the compressive trough in the respective profiles at the weld centre-line. In contrast, the spatial profile of the transverse stress component is similar to that shown by the normal component of residual stress in multi-pass arc welds [12]. This is because, the aspect ratio of the weld beads (depth to width) is much smaller in multi-pass arc welds. The impact of solid state austenite to martensite phase transformation discussed here finds support from other measurements made on Plate A and also Plate B as discussed below. 3.5. Cross-weld strain and stress profiles in Plate A at 4.5 mm below the surface The cross-weld residual stress profiles for plate A at 4.5 mm below the top surface (mid-thickness plane) of the plate are shown in Fig. 8. The profiles for all three components of stress – longitudinal, transverse and normal – are similar to those at a depth of 1.5 mm but show differences in magnitudes at the troughs and peaks in the respective profiles. For example, the trough in the normal component at the mid-thickness plane is deeper ( −150 MPa) and the peaks are higher ( 350 MPa) than those in the equivalent locations in the profile at a depth of 1.5 mm (Fig. 7). These differences can be attributed to greater restraint of martensitic transformation strains at the mid-thickness plane than near the surface. The profiles of the transverse component of the residual stress from the two locations (1.5 mm and 4.5 mm) are similar. An important difference between the cross-weld residual stress profiles at mid-thickness and at 1.5 mm depth is that, the distance between the tensile peaks is narrower at mid-thickness (7 1.5 mm from the weld centre-line compared with 72 mm). This is because the width of the fusion zone plus HAZ is smaller at the midthickness plane than at 1.5 mm below the top surface, again noting
500
Residual Stress (MPa) Microhardness (HVN0.5)
500 400 300 200
Longitudinal Transverse Normal Microhardness Fusion Line HAZ Boundary
400 300 200
Residual Stress (MPa)
600
600
Longitudinal Transverse Normal Fusion Line HAZ Boundary
100
100
0
0
-100
-100 -15
-10
-5
0
5
10
15
Distance from the weld centreline (mm)
Fig. 7. Cross-weld residual stress and microhardness profile in Plate A (8 kW, 1.5 m/min) at 1.5 mm below top surface.
-200 -15
-10
-5
0
5
10
15
Distance from the Weld Centreline (mm)
Fig. 8. Cross-weld stress profile in Plate A (8 kW, 1.5 m/min) at 4.5 mm below surface.
166
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
that it is the martensitic transformation in the fusion zone and HAZ that governs the width of the residual stress profile. 3.6. Through-thickness stress profiles Through-thickness profiles of the longitudinal, transverse and normal components of residual stress at the weld centre-line in Plate A are shown in Fig. 9. It can be observed that the uncertainties associated with these weld metal measurements are much higher ( 730 MPa) than those in the parent metal region ( 710 MPa (Figs. 7 and 8)). This is on account of higher microstresses in the fusion zone and the HAZ as compared to the same in the parent metal. The transverse component of residual stress is most tensile, the normal component most compressive and the longitudinal component lies in between. The final residual stress state in the weld metal is the result of competition between martensitic transformation induced dilation and thermal contraction. In the normal direction, thermal contraction will be smaller than that in the longitudinal direction; therefore, the normal component of residual stress is more compressive than the longitudinal component. Residual stress below the mid thickness plane is more compressive than that above the mid plane. This is because, the heat source is applied from the top side, leading to less heat input below the mid-thickness plane and therefore, the dilative martensitic transformation commences first in this region, producing a more compressive region than that above. The variation in residual stresses through the thickness at the weld centre-line is most likely to be associated with differential cooling between the bottom and top of the weld. Greater heat is input by the laser welding process at the top surface; it can be noted from Fig. 1 that the fusion zone and the HAZ are slightly narrower towards the bottom side of the plate. This means that the bottom of the welded plate will cool faster than the top and lead to a resultant residual stress distribution controlled by the combined influences of differential thermal contraction, different degrees of martensitic transformation and self-constraint. Numerical weld computational mechanics is probably the best way to understand the complex mechanical behaviour involved. This sort of residual stress profile with a compressive zone below the midthickness plane and tensile zone above contrasts from that observed in the case of multi-pass arc-welded joints in this material [12] where significantly higher tensile residual stress develops below the mid-thickness plane relative to above it. This is on account of the thermal cycles associated with the succeeding passes introducing martensitic dilation and more compressive stresses in final weld capping passes.
Residual stress maps for longitudinal and normal components on the mid-length cross-section of the low heat input welded plate were generated on one side of the weld centre-line; using the through thickness measurements along the weld centre-line and along the measurement lines at distances of 2 mm, 4 mm and 10 mm from it. For distances of 15 mm and 30 mm the weld centre-line, values from cross-weld measurements at 1.5 mm and 4.5 mm below the top surface were used to generate these plots. The plots were generated using MATLAB. Figs. 10 and 11 show maps of the longitudinal and normal components respectively of residual stress on one side of the welded joint. These maps show that the cross-weld profiles for the longitudinal and the normal components represented in Figs. 7 and 8 exist through the entire thickness of the plate. It can also be seen from these maps that the cross-weld profiles of the longitudinal and normal components of stress are very similar in nature, i.e. both components show a low tensile or compressive trough in the weld centre and two tensile peaks on the either side of the joint in the parent metal adjacent to the PM–HAZ boundary. 3.7. Effect of welding speed on residual stress profile Figs. 12–14 compare cross-weld profiles of the longitudinal, normal and transverse components of residual stress respectively in Plates A and B, which were welded at the same laser power (8 kW) but with different welding speeds (1.5 m/min and 0.75 m/min, respectively) giving low and high heat input conditions. A cross-weld
Fig. 10. Map of longitudinal component of residual stress in Plate A at the midlength cross-section in the one side of the weld centre-line.
Fig. 11. Map of normal component of residual stress in Plate A at the mid-length cross-section in the one side of the weld centre-line.
300 600 Longitudinal Transverse Normal
HAZ Boundary_Plate A
500
Residual Stress (MPa)
400 100 300 0 0
1
2
3
4
5
6
7
8
200
Fusion Line_Plate B
Longitudinal Stress (MPa) Microhardness (HVN0.5)
200
HAZ Boundary_Plate B Longitudinal_Plate A Longitudinal_Plate B Microhardness_Plate A Microhardness_Plate B Fusion Line_Plate A
Distance from the top surface (mm) -100
100 0
-200 -100 -15 -300
Fig. 9. Through-thickness profiles of the longitudinal, transverse and normal components of residual stress in Plate A at the weld centre-line.
-10
-5
0
5
10
15
Distance from the weld centreline (mm)
Fig. 12. Cross-weld profile of the microhardness and longitudinal component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
200
100
Residual Stress (MPa)
300
Normal_Plate A Normal_Plate B Fusion Line_Plate A HAZ Boundary_Plate A Fusion Line_Plate B HAZ Boundary_Plate B
0
-100 -15
-10
-5
0
5
10
15
Distance from the weld centre line (mm)
Fig. 13. Cross-weld profile of the normal component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
250 Transverse_Plate A 200
Transverse_Plate B
Re 150 sid ua l 100 Str es 50 s
Fusion Line_Plate A HAZ Boundary_Plate A Fusion Line_Plate B HAZ Boundary_Plate B
3. There is little variation in residual stress through the plate thickness except in the fusion zone, where longitudinal and normal stresses are more compressive in the bottom half of the weld zone. This contrasts with conventional multi-pass welds in P91 steel where the residual stress is more tensile in the bottom side than in the top side. 4. The widths of the fusion zone (1.2 mm for the low heat input and 1.8 mm for high heat input) and the low tensile/compressive troughs in the residual stress profiles (2 mm wide for low heat input and 3 mm for high heat input) were wider and marginally deeper for the high heat input weld compared with the low heat input weld, but the peak magnitudes of stress were similar. 5. There is clear evidence of a very strong influence of the strains associated with the martensitic transformation on the magnitude and spatial profile of all three components of residual stress in both welded plates. 6. The most significant components of residual stress in the laser welds are in the longitudinal and normal directions. This contrasts with conventional multi-pass welds where longitudinal and transverse components are the most significant. This is mainly owing to the high aspect ratio (depth to width ratio) of the fusion zone in laser welds. 7. The width (FWHM) of the diffraction peaks shows one to one correlation with as transformed region on account of increased microstresses.
Acknowledgements
0 -50 -100 -15
167
-10
-5
0
5
10
15
The research was supported by the Indian Department of Atomic Energy and by grant EP/101215X/1 from the UK Engineering and Physical Research Council. The authors gratefully acknowledge the award of neutron beam time by the ILL Grenoble, France.
Distance from the weld centerline (mm)
Fig. 14. Cross-weld profile of the transverse component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
microhardness profile is also superimposed. The shapes and peak magnitudes of the profiles from the two plates are almost identical. However, it can be seen that the lower welding speed has led to both broadening and deepening of the trough in the cross-weld profiles of residual stress in both the longitudinal and normal directions. The increased width of the trough correlates with the wider fusion zone in Plate B which experienced nearly double the heat input of A; compare the fusion zones in Fig. 1(a) and (b). These results further illustrate how the martensitic transformation in the fusion zone and HAZ of a laser welded joint in P91 steel leads to reduced levels of residual stress in the weld region.
4. Conclusions 1. Neutron diffraction has been used to measure as-welded residual stresses in 9 mm thick Grade 91 steel plates that were joined using constant power (8 kW) low (320 J/mm) and high (640 J/mm) heat input laser welding procedures. 2. Cross-weld profiles of the longitudinal and normal components of residual stress in both of the laser welds show a low tensile or compressive trough (−250 MPa to 100 MPa) in the welded fusion zone and high tensile peaks (500 MPa to 600 MPa for longitudinal and 200 to 350 for normal) in the parent metal near the HAZ–PM interface on the either side of the welded joint.
References [1] A. Shibli, F. Starr, Int. J. Press. Vessels Pip. 84 (2007) 114–122. [2] K. Laha, K.S. Chandravathi, P. Parameshwaran, K.Bhanu Sankara Rao, S. L. Mannan, Metall. Mater. Trans. A 38 (2007) 58–68. [3] S. Pillot, Z. Zhao, S. Corre, C. Chauvy, L. Coudreuse, P. Toussaint. Proc. of ASME 2010 Press. Vessel. and Pip. Div. Conf. (PVP 2010) 1–14. [4] R. Mythili, P.V. Thomas, S. Saroja, M. Vijaylakshmi, V.S. Raghunathan, J. Nucl. Mater. 312 (2003) 199–206. [5] R.K. Shiue, K.C. Lan, C. Chen, Mater. Sci. Eng., A 287 (2000) 10–16. [6] P.K. Ghosh, U. Singh, Sci. Technol. Weld. Joining 9 (3) (2004) 229–236. [7] H.O. Andrén, G. Cai, L.E. Svensson, Appl. Surf. Sci. 87–88 (1995) 200–206. [8] S.K. Albert, M. Tabuchi, H. Hongo, T. Watanabe, K. Kubo, M. Matsui, Sci. Technol. Weld. Joining 10 (2) (2005) 149–157. [9] Z. Xu, Report No. ANL/NE-12/12, Argonne National Laboratory, Chicago, Illinois, 2008, September. [10] K. Seok-Hoon, K. Jong-Bum, L. Won-Jae, J. Mater. Process. Technol. 209 (2009) 3905–3913. [11] D. Dean, M. Hidekazu, Comput. Mater. Sci. 37 (2006) 209–219. [12] S. Paddea, J.A. Francis, A.M. Paradowska, P.J. Bouchard, I.A. Shibli, Mater. Sci. Eng., A 534 (2012) 663–672. [13] N.S. Rossini, M. Dassisti, K.Y. Benyounis, A.G. Olabi, Mater. Des. 35 (2012) 572–588. [14] P.J. Withers, C. R. Phys. 8 (2007) 806–820. [15] G.A. Webster, R.C. Wempory, J. Mater. Process. Technol. 117 (2001) 395–399. [16] W. Wanchuck, E.M. Vyacheslav, M. Pavel, A. Gyu-Baek, S. Baek-Seok, Mater. Sci. Eng., A 528 (2011) 4120–4124. [17] M.J. Park, H.N. Yang, D.Y. Jang, J.S. Kim, T.E. Jin, J. Mater. Process. Technol. 155– 156 (2004) 1171–1177. [18] A. Paradowska, J.W.H. Price, R. Ibrahim, T. Finlayson, J. Mater. Process. Technol. 164–165 (2005) 1099–1105. [19] J.W.H. Price, A. Paradowska, S. Joshi, T. Finlayson, Int. J. Press. Vessels Pip. 83 (2006) 381–387. [20] M. Hofmann, R.C. Wimpory, Int. J. Press. Vessels Pip. 86 (2009) 122–125. [21] C. Ohms, R.C. Wimpory, D.E. Katsareas, A.G. Youtsos, Int. J. Press. Vessels Pip. 86 (2009) 63–72. [22] S. Pratihar, M. Turski, L. Edwards, P.J. Bouchard, Int. J. Press. Vessels Pip. 86 (2009) 13–19.
168
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
[23] D. Thibault, P. Bocher, M. Thomas, M. Gharghouri, M. Côté, Mater. Sci. Eng., A 527 (2010) 6205–6210. [24] W. Wanchuck, E.M. Vyacheslav, C.R. Hubbard, H.J. Lee, P.Kwang Soo, Mater. Sci. Eng., A 528 (2011) 8021–8027. [25] M. Rogante, P. Battistella, F. Rustichelli, J. Alloy. Compd. 378 (2004) 335–338. [26] W. Suder, S. Ganguly, S. Williams, A.M. Paradowska, P. Colegrove, Sci. Technol. Weld. Join. 16 (3) (2011) 244–248.
[27] D.W. Brown, T.M. Holden, B. Clausen, M.B. Prime, T.A. Sisneros, H. Swenson, J. Vaja, Acta Mater. 59 (2011) 864–873. [28] S. Ganguly, L. Edwards, M.E. Fitzpatrick, Mater. Sci. Eng., A 528 (3) (2010) 1226–1232. [29] B.J. Ganesh, S. Raju, K.R. Arun, E. Mohandas, M. Vijayalakshmi, K.Bhanu Sankara Rao, R. Baldev, Mater. Sci. Technol. 27 (2) (2011) 500–512. [30] A.F. Mark, J.A. Francis, H. Dai, M. Turski, P.R. Hurrell, S.K. Bate, J.R. Kornmeier, P. J. Withers, Acta Mater. 60 (2012) 3268–3278.
Contents lists available at SciVerse ScienceDirect
Materials Science & Engineering A journal homepage: www.elsevier.com/locate/msea
Residual stresses in laser welded ASTM A387 Grade 91 steel plates Santosh Kumar a,n, A. Kundu b, K.A. Venkata c, A. Evans d, C.E. Truman c, J.A. Francis e, K. Bhanumurthy a, P.J. Bouchard b, G.K. Dey a a
Bhabha Atomic Research Centre, Mumbai, Maharashtra 400094, India Materials Engineering, The Open University, Milton Keynes, MK7 6AA, UK c Department of Mechanical Engineering, University of Bristol, Bristol, BS8 1TR, UK d Institut Laue Langevin, Grenoble, France e University of Manchester, Manchester, M13 9PL, UK b
art ic l e i nf o
a b s t r a c t
Article history: Received 28 November 2012 Received in revised form 11 March 2013 Accepted 13 March 2013 Available online 27 March 2013
Residual stresses in 9 mm thick ASTM A387 Grade 91 steel plates, joined using constant power (8 kW) low and high heat input laser welding processes, are characterised using neutron diffraction. The measured longitudinal and normal components of residual stress show a bimodal distribution across the welded joint with a low tensile or compressive trough at the weld centre flanked by high magnitude tensile peaks in parent metal adjacent to the heat affected zone boundaries. The width of the central trough and spread of the outboard tensile zones are significantly greater for the high heat input weld. In both cases, the stress distributions can be explained by the strains associated with the austenite to martensite solid-state transformation as the joint cools after welding. & 2013 Elsevier B.V. All rights reserved.
Keywords: Laser welding ASTM A387 Grade 91 Neutron diffraction Residual stress
1. Introduction Ferritic-martensitic steels, and in particular Grade 91 steel (9Cr1Mo, V, Nb), find extensive applications in supercritical power generation plants, nuclear power systems and in the petrochemical industry owing to their excellent combination of elevated temperature strength, thermal fatigue resistance and immunity from stress corrosion cracking in aqueous and chloride environments [1,2]. Moreover they are candidate materials for structural components in many advanced reactor systems like the Very High Temperature Reactor and the Sodium Fast Reactor owing to their low susceptibility to irradiation embrittlement [3]. Different welding techniques such as Manual Metal Arc welding, Submerged Arc Welding and Gas Tungsten Arc Welding are used to fabricate components and assemblies from ferritic-martensitic steels of this type [2–7]. Laser welding is a versatile joining process with high power density and low thermal input. It can weld different alloys from sub-millimeter up to 18 mm section thicknesses in many joint configurations without the need for any filler material. Major advantages of the laser welding process include the narrow width of the fused zone and heat-affected zone (HAZ), and minimal distortion of the components being welded owing to the low heat input. There is also some evidence to suggest that welds made
n
Corresponding author. Tel.: +91 22 25595695; fax: +91 22 25505151. E-mail addresses: [email protected], [email protected] (S. Kumar).
0921-5093/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.msea.2013.03.046
with high-energy beam processes such as laser or electron beam welding may be more resistant to type IV cracking than welds made with conventional arc welding processes [8,9]. Very high values of residual stress approaching the yield strength of the parent material and weld filler are generally introduced in the vicinity of welded joints. Temperature excursions associated with the thermal cycle of a welding pass induce localized thermal dilation that is accommodated by the elastoplastic responses of the structure. In turn this is accommodated by cyclic stress–strain behaviour of the materials, the self-restraint of the structure and any applied restraint. In addition, for material systems like P91, the strains associated with the solid-state austenite to martensite transformation of the material in the fusion zone and the HAZ can have a profound impact on the residual stress profile [10–12]. Quantitative characterisation of the residual stress distribution around the welded joint is important, to understand the factors governing its evolution, to assess how such stresses may affect the life and integrity of components inservice, and also to design mitigation treatments including optimization of the post-weld heat treatment (PWHT). Residual stress around a weld joint can be measured by surface based techniques such as centre-hole drilling, magnetic methods and X-ray diffraction or volumetric techniques such as the inherent strain method, deep hole drilling, the contour method, neutron diffraction or synchrotron diffraction [13]. Surface based techniques provide limited information about the residual stress field, whilst volumetric strain relief methods are invasive or destructive.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
161
different welding speeds, 1.5 m/min (A) and 0.75 m/min (B), to produce welded joints with significantly different heat inputs. Helium was used as the shielding gas with a flow rate of 20 l/min. Throughout the paper the translational direction of welding is referred as the plate longitudinal direction. The normal direction is perpendicular to the welding direction passing through the thickness of the plate and the transverse direction is in the plane of the plate and perpendicular to the other two directions. Neutron diffraction for residual stress measurements were carried out at the ILL, Grenoble, France on the SALSA instrument. The neutron source at this facility is an experimental nuclear reactor with 58 MWth power. All measurements were made using the (211) crystal reflection of the material, with a monochromatic beam of wavelength λ ¼1.648 Å. This wavelength resulted in scattering angles close to 901, thereby giving essentially cuboidal gauge volumes. Source and detector slit combinations were chosen to define gauge volumes in the sample small enough to capture the rapid variations in strain; expected across the weld and in the through-thickness direction of the plate. For measurements of lattice spacing in the longitudinal direction, a gauge volume of 0.8 mm 0.8 mm 2 mm was chosen, and an extended volume (in the longitudinal direction) of 0.8 mm 0.8 mm 20 mm was chosen for measurements of lattice spacing in the transverse and normal directions. Measurement points were placed closely together (at 0.5 mm spacing) near the weld line and sparsely thereafter. The lines along which neutron diffraction measurements were made are shown schematically in Fig. 1 (a) and (b), respectively, for welding conditions (A) and (B).
Neutron diffraction was selected for the present work for several reasons. First, it is a non-destructive method that can provide quantitative information about three or more components of the stress tensor in structures made from steel up to a few tens of millimeters thick [14]. Second, a fine gauge volume can be selected and this is essential to resolve the short length-scale variations in residual stress expected in narrow width laser welds. Third, the neutron diffraction technique is well established having been applied widely to characterise residual stresses in welded joints [10–27]. Measurements of the residual stress fields around P91 welded joints by different methods, including neutron diffraction, have been reported by several workers [10–12]. However, there is no published work on residual stress measurements in laser welded joints in this material. The present work provides a detailed analysis of residual stress measurements by neutron diffraction. Keyhole laser welds are also of interest as they are expected to show a residual stress profile that is significantly different from that introduced by conventional multi-pass welding processes. This is on account of the high through-thickness to width aspect ratio of the fusion zone and HAZ and because they are single pass welds. The results reported here will provide a set of high quality measured residual stress data that can be used to validate modeling and simulation endeavors. Such models can be harnessed to explore the effects of different processing conditions on the residual stress profile, as well as providing the starting conditions for PWHT stress relaxation studies.
2. Experimental details Hot-rolled ASTM A387 Grade 91 steel plates, 10 mm thick, were procured from M/s Industeel, Belgium in the normalized (30 min at 1050 1C) and tempered (30 min at 770 1C) condition. The chemical composition of the as-received material, which had a tempered martensitic microstructure, is given in Table 1. Dilatometry was performed using a TD 5000S dilatometer to characterise the austenite to martensite solid state phase transformation on cooling, as the associated strains are expected to have a significant bearing on the residual stress profile. For this purpose, a cylindrical pin (length 14.91 mm and diameter 4 mm) was used as the sample and a cylindrical alumina pin was used as the reference. The sample and the reference pins were heated simultaneously at a rate of 10 1C/min from room temperature to 1060 1C; soaked at 1060 1C for 30 min and cooled to room temperature at 10 1C/min. The supplied plates were prepared for laser welding by wire electro-discharge machining (EDM) so that they measured 500 mm 70 mm 10 mm. The thickness of the plates was reduced by further machining to approximately 9 mm (8.7 mm to 9.2 mm) to remove surface undulations caused by rolling. Laser welding was carried out using a high power continuous CO2 laser with maximum power of 10 kW and beam quality factor K ¼0.52. Optics with a focal length of 280 mm were used, to produce a focused beam diameter of 500 μm. The parent material (PM) plates were arranged in a square butt configuration with the gap between the parting surfaces measuring less than 100 μm, and rigidly held using pneumatically controlled clamps. Spot welds were made along the weld seam (length 500 mm) at three equidistant locations. Laser welding was carried out (without preheating the plates) at 8 kW power and with two
Fig. 1. (a): Macrograph showing Plate A joint cross-section (welded at 8 kW and 1.5 m/min) with neutron diffraction measurement lines marked. Measurement lines are 1.5 mm and 4.5 mm below the left top surface. (b): Macrograph showing Plate B joint cross-section (welded at 8 kW and 0.75 m/min) with neutron diffraction measurement lines marked. Measurement line is 1.5 mm and 4.5 mm below the left top surface.
Table 1 Chemical composition of ASTM A387 Gr 91 steel in wt.% (Balance Fe). C
Mn
Zr
Si
P
S
Cr
Mo
Ni
Cu
Al
N
Nb
Ti
V
0.106
0.443
0.005
0.221
0.018
0.0008
8.965
0.901
0.212
0.045
0.010
0.0464
0.073
0.004
0.194
162
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
Values of the stress free lattice spacing, d0, for (211) planes were required to determine the measured residual strains and stresses. Invariably, it is difficult to determine reliable values of the stress-free lattice parameter in the vicinity of narrow weldments. Several approaches that can be used have been documented by Withers et al. [14]. In the present case, the very narrow width of the fusion zone ( 1.2 mm) and HAZ ( 0.6 mm) means that steep changes in d0 were expected. Therefore, it was decided to use a “comb sample” spanning the weld as illustrated in Fig. 2. The comb was machined from a cross-sectional slice of material approximately 5 mm thick that had been extracted from near the end of the laser welded plate. Four transverse “prongs” spanning the weld line, each about 2 mm thick, were created by making three EDM cuts through the thickness of the slice as shown in Fig. 2. Whilst such samples are often considered to be free of “macrostress”, significant remnant stresses in a similar comb extracted from an aluminium alloy plate have been reported [28]. This may be even more important in the present case because of the short residual stress length-scales involved. To mitigate this potential uncertainty, strains were measured in three orthogonal directions in the comb prongs and a state of plane stress (for the normal plane) was assumed to exist within the prongs in order to derive values for d0. Measurements for the d-spacing in the three orthogonal directions were made at different points across the weld in the top most prong. A gauge volume of 0.8 0.8 2.0 mm was used for measurements in the longitudinal and normal directions and an extended gauge volume of 0.8 0.8 5.0 mm was used for measurements in the transverse direction. Stress-free lattice parameter (d0) values were derived assuming a plane stress condition in the topmost prong as described below. The stress along a principal direction x can be written as sx ¼
E ð1−vÞεx þ vðεy þ εz Þ ð1 þ vÞð1−2vÞ
ð1Þ
where si and εi are stress and strain components in principal directions i¼x, y and z, and E and ν are the crystallographic elastic moduli for the (211) planes. If sx is relieved by cutting a thin strip normal to the x-direction; then E ð1−vÞεx þ vðεy þ εz Þ ¼ 0 ð1 þ vÞð1−2vÞ and making the substitution gives
d0 ¼
εx ¼
dx −d0 d0
ð1−vÞdx þvðdy þdz Þ ð1þvÞ
ð2Þ ð3Þ ð4Þ
This expression was used to derive d0 values for the fusion zone and the HAZ. The approach assumes isotropic material properties (i.e. low texture) and that values of d0 are direction-independent (i.e. negligible microstresses are present owing to the influence of plastic strain). For the parent metal, the average d0 value of ten different
Fig. 2. Sketch showing the comb sample used for d0 measurements.
measurements on a small piece of material cut from the far end of the plate was used. This value was 1.17241570.000055 Å. The diffraction angle, 2θ, determined from the measured position of the diffraction peak, was used to derive the lattice spacing, d, using Bragg's Law, λ ¼ 2dsinθ The lattice spacing along three orthogonal directions of the welded plate – longitudinal, transverse and normal – were calculated using the corresponding diffraction peaks from different measurement points. These values were then used to calculate strain values εx, εy and εz in the three directions using suitable d0 values as described above. Using these strain values, the residual stress along the three orthogonal directions—sx, sy and sz were calculated for all the measurement points using Eq. (1). Errors in the individual lattice spacing measurements were derived from the error in the location of the diffraction peak centre, arising from the Gaussian fit. This exercise was carried out for all the measurements on the plate as well as on the combs. The error propagation was applied in the derivation of all values for d0 (using a plane stress assumption) and for values of residual strain and residual stress along the three principal directions. A cross-section of each welded joint was prepared for metallographic evaluation by first polishing and then etching with Vilella's reagent in order to reveal different metallurgical regions—the parent metal, the heat affected zone and the fusion zone, and to reveal the detailed microstructure within each of these regions. Microhardness measurements were also carried out across the welded joint on a polished and lightly etched cross-section, using a 5 N load and a 10 s dwell time. The measurement points were 1.5 mm below the top surface of the welded plate and at a pitch of 0.2 mm, in order to capture the rapid variation in the microhardness through the heat affected zone, and also to ensure that the microhardness measurements were carried out at locations similar to those where neutron diffraction measurements were made. Optical microscopy was employed to obtain representative micrographs from the parent metal, the heat affected zone and the fusion zone, both for plate A and for plate B.
3. Results and discussion Polished and etched cross-sectional macrographs of the laser welded plates A and B (fast and slow advance rates) are presented in Fig. 1(a) and (b). In both cases there is a small geometric mismatch across the weld. The width of the fusion zone in the high speed weld (low heat input) is about 1.2 mm as compared with about 1.8 mm for the slow speed (high heat input) weld. The location of the neutron diffraction measurement lines are marked on the macrograph of these welds (Fig. 1(a) and (b)). 3.1. Microstructural evaluation The microstructures for the different regions—the parent metal, the heat affected zone and the fusion zone for plates A and B are presented in Fig. 3(a) through to Fig. 3(e). The parent metal (Fig. 3 (a)) showed a tempered martensitic structure within equiaxed prior austenite grains of 20 μm in size. The heat affected zones in both the plates A and B showed an as-transformed martensitic structure within reaustenitized grains of varying size. In the heat affected zone, reaustenitization of the tempered martensitic structure (present in the parent metal) occurs due to thermal cycles associated with the welding process provided that at some point the peak temperature exceeds the Ac1 temperature of the material. The size of the reaustenitized grains increases with an increase in the peak temperature that is experienced by the
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
163
Fig. 3. (a): Microstructure of the parent metal. (b): Microstructure of the heat affected zone in Plate A. (c): Microstructure of the fusion zone in Plate A. (d): Microstructure of the heat affected zone in Plate B. (e): Microstructure of the fusion zone in the Plate B.
respective region. The regions farther from the fusion line experience lower peak temperatures and therefore form smaller austenite grains, since there is limited scope for grain growth while the material exists as austenite. The heat affected zone in P91 steel welds can be further sub-categorised according to the extent to which austenitisation occurs and according to the peak temperature that is experienced during the welding thermal cycle. Adjacent to the unaffected parent material the heat affected zone contains a region that is subcritical (SC), and as the distance from the fusion line decreases the next sub-zone is intercritical (IC) or partially austenitised, followed by a fine grained (FG) region and finally a coarse grained (CG) region adjacent to the fusion line [12]. However, the temperature gradients that arise during laser welding are very steep, and the heat affected zones are very narrow. This places significant limitations on the extent to which grain growth can occur in the heat affected zone, since the region that experiences a high enough peak temperature for grain growth to occur is itself very narrow and such temperatures arise only for very short durations. This was evident in these laser welding experiments since no discernible coarse-grained heat affected zone was observed in plate A or plate B. The fusion zones in both the welded joints showed an as-transformed martensitic structure within columnar prior austenite grains. Columnar grains arise frequently in weld fusion zones, and they are a consequence of grains growing at 901 to the fusion boundary, in a direction that is
opposite to the direction of maximum heat flow. There were no significant differences in the microstructures of the two plates welded with different welding speeds.
3.2. Thermal dilation measurement The measured dilation curve for the parent Grade 91 material (PM) is shown in Fig. 4. During heating at 10 1C/min, the steel expands until it reaches the Ac1 temperature ( 825 1C), where significant contraction occurs due to ferrite to austenite transformation. This contraction continues until the Ac3 temperature is reached, but at 845 1C a significant part of the ferrite has already transformed to austenite and thermal expansion of austenite becomes dominant. A small expansion is evident during the thirty minute soak at 1060 1C which can be attributed to dissolution of carbides leading to an increase in the lattice spacing of the austenite. During cooling, there is a contraction of the austenite until it reaches the Ms Temperature ( 400 1C), where significant expansion arises due to solid-state transformation of austenite into martensite until Mf is reached ( 240 1C). Subsequently, the martensite contracts until room temperature is reached. These phase transition temperatures show consistency with those reported in the literature by using differential scanning calorimetry (DSC) for this material [29].
164
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
Soaking (30 min
0.014
Thermal Dilation (mm/mm)
0.012
0.01
0.008
0.006
0.004
0.002
0 0
200
400
600
800
1000
Temperature (°C)
Fig. 4. Measured thermal dilation of ASTM A387 Grade 91 steel (heating and cooling rate 10 1C/min).
0.51 for the parent metal and increases rapidly through the narrow HAZ to become as high as 1.41 for the longitudinal component and 11 for the transverse and normal components in the welded region. The regions of higher FWHM show a one-toone correlation with the regions which have undergone a martensitic transformation due to the welding thermal cycle. The correlation is so strong that the FWHM may be used as a signature for the fusion zone (FZ) and HAZ regions, which have undergone a martensitic transformation during welding. The higher values of FWHM of the diffraction peaks are due to a variation of strains on the microscopic length scale within the gauge volume. Such variations can arise from plastic deformation during cooling and/ or the solid state phase transformation. However, further experimentation and simulation work is needed to confirm and elucidate the contributing mechanisms. The FWHM of the diffraction peak in the weld region for the longitudinal component is significantly higher ( 40%) than that for the transverse and normal components. This could be due to texture in the prior austenite grains leading to the selection of different variants along the three measured directions—longitudinal, transverse and normal. This can however, be confirmed only with a detailed EBSD (Electron Back Scattered Diffraction) analysis. It may also be due to the smaller gauge volume used for the longitudinal measurements compared with that used for measurements of the other two components, noting that larger gauge volumes tend to give sharper diffraction peaks than smaller volumes assuming other parameters such as count time, material chemistry and crystal structure etc. remain the same. Whilst a sharp diffraction peak is preferred, as it implies a lower error, a small volume had to be employed for measuring the longitudinal component in order to capture the steep gradients in stress across the weldment.
3.4. Cross-weld strain and stress profiles in Plate A at 1.5 mm below the top surface
The strain associated with the martensitic transformation on cooling has very significant impact on residual stress profile for single pass welds. After the melt zone is solidified, continuity is established and the development of the residual stress field is influenced primarily by the localized contraction of the hot zone in and around the welded joint and resistance associated with geometric self-restraint of the joint and any applied restraint. But once the Ms temperature is reached in P91 steel, a significant strain ( 2600 micro strain in the absence of an external stress (Fig. 4)) occurs owing to the martensitic transformation, until the Mf temperature is reached, and this strain is expected to have a strong impact on the residual stress profile. Thus, for single pass welds, the evolution of the residual stress field is driven by competing contributions from thermal contraction during cooling and the strains associated with the solid-state phase transformation. 3.3. Diffraction peak FWHM analysis The measured variation in the full width at half maximum (FWHM) of the (211) diffraction peak across the weld for measurements of d-spacing in the longitudinal, transverse and normal directions at mid-length of Plate A is shown in Fig. 5. A cross-weld microhardness profile from a similar location is also imposed in this figure. A similar pattern was observed in all cross-weld measurement scans. The variation in FWHM is symmetric about the weld centre line and exhibits an interesting trend. It remains
The measured residual strain profile, across the low heat input weld in Plate A, at 1.5 mm depth below the top surface is shown in Fig. 6. As expected, the three components (longitudinal, transverse and normal) of residual strain show a high degree of symmetry across the weld centre-line. Interestingly the longitudinal component shows a bimodal distribution with a trough coincident with the weld centre-line and the two peaks lying 2 mm on either side of the centre in the parent metal near the HAZ–PM interface. The peak value of the longitudinal component of residual strain is 1900 micro-strain. The normal component shows a similar 2000
1400
800
Longitudinal Transverse Normal Fusion Line HAZ Boundary
Microstrains
Fig. 5. Variation of full width at half maximum (FWHM) of the (211) diffraction peaks across the laser weld joint in Plate A at 1.5 mm below the top surface. Crossweld microhardness profile is also imposed.
200
-400
-1000 -10
-8
-6
-4
-2
0
2
4
6
8
10
Distance from the weld centreline (mm)
Fig. 6. Cross-weld residual strain profile in Plate A (8 kW, 1.5 m/min) at 1.5 mm below top surface.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
profile, but with a lower peak magnitude ( 500 micro-strain) at 1.5 mm on either side of the weld centre-line. The transverse component of residual strain, however, shows the opposite trend. It has a peak ( 650 micro strain) coinciding with the weld centreline and two troughs ( −900 micro-strain) at 1.5 mm on either side of the weld centre-line. The same characteristics are seen in the residual stress profile (Fig. 7). The longitudinal component of residual stress shows a low tensile trough ( 50 MPa) in the weld centre which rises rapidly to a peak value of 500 MPa at 2 mm on either side of it. The regions with peak residual stresses are in the parent metal just outside the metallurgical HAZ, as is evident from a comparison of the crossweld microhardness profile and the residual stresses at the same location. Similar results have been reported by Mark et al. [30] for residual stress profiles across a single pass weld in SA508 steel, which also undergoes a bainitic/martensitic transformation while cooling from the austenite phase field. The normal component shows a similar trend, with a trough in the weld centre ( 100 MPa) and rising rapidly to peak values ( 200 MPa) at 1.5 mm on either side of the centre-line. The transverse component shows the opposite trend with the peak ( 200 MPa) coinciding with the weld centre and two troughs ( 710 MPa) at 1.5 mm on either side of it. The measured strain and stress profiles are almost completely symmetric. This gives confidence in the quality of results, and the experimental procedure, which involved remounting of the sample after measurements of the longitudinal component of strain, to obtain corresponding data in the other two orthogonal directions. Thus, further measurements were focused on only one side of the welded joint. The trough in the longitudinal residual strain and stress profiles at the weld centre-line can be attributed to strains associated with the martensitic transformation, which includes a volumetric expansion, during cool down of the region of the weldment which was heated to above the Ac1 temperature; that is the entire fusion zone and the HAZ. The tensile peaks are in regions heated to a temperature just below the Ac1 temperature, as these regions experience only contraction during cooling of the welded joint. This M-shaped profile contrasts with a non-transforming material where a single tensile peak would be expected at and near the weld centre-line, dropping to low compressive stresses in the far field [26]. However, in P91 steel, martensitic transformation strains have been observed to have a similar influence on the residual stresses in a multi-pass pipe girth weld, where compressive stresses were measured in the region of the last weld pass, and the proposed influence of the transformation strains was supported by the results of a Satoh test [12].
165
Another feature is observed in the profile for the normal component of residual stress, which has the same shape as the longitudinal profile, but much lower magnitude. The common shape of the profiles suggests that the factors controlling the evolution of the two are similar in nature and possibly related to the strains associated with the martensitic transformation occurring under constraint. Because of the large aspect ratio (depth to width) of the laser weld fusion zone, the thickness of the weld bead (i.e. the plate) provides some restraint against the dilatational component of the strain. In the longitudinal welding direction, the effective restraint length is much longer giving a larger compressive effect. The adjacent tensile peaks in the longitudinal and normal components of the stress profile may be viewed as reactions to the compressive trough in the respective profiles at the weld centre-line. In contrast, the spatial profile of the transverse stress component is similar to that shown by the normal component of residual stress in multi-pass arc welds [12]. This is because, the aspect ratio of the weld beads (depth to width) is much smaller in multi-pass arc welds. The impact of solid state austenite to martensite phase transformation discussed here finds support from other measurements made on Plate A and also Plate B as discussed below. 3.5. Cross-weld strain and stress profiles in Plate A at 4.5 mm below the surface The cross-weld residual stress profiles for plate A at 4.5 mm below the top surface (mid-thickness plane) of the plate are shown in Fig. 8. The profiles for all three components of stress – longitudinal, transverse and normal – are similar to those at a depth of 1.5 mm but show differences in magnitudes at the troughs and peaks in the respective profiles. For example, the trough in the normal component at the mid-thickness plane is deeper ( −150 MPa) and the peaks are higher ( 350 MPa) than those in the equivalent locations in the profile at a depth of 1.5 mm (Fig. 7). These differences can be attributed to greater restraint of martensitic transformation strains at the mid-thickness plane than near the surface. The profiles of the transverse component of the residual stress from the two locations (1.5 mm and 4.5 mm) are similar. An important difference between the cross-weld residual stress profiles at mid-thickness and at 1.5 mm depth is that, the distance between the tensile peaks is narrower at mid-thickness (7 1.5 mm from the weld centre-line compared with 72 mm). This is because the width of the fusion zone plus HAZ is smaller at the midthickness plane than at 1.5 mm below the top surface, again noting
500
Residual Stress (MPa) Microhardness (HVN0.5)
500 400 300 200
Longitudinal Transverse Normal Microhardness Fusion Line HAZ Boundary
400 300 200
Residual Stress (MPa)
600
600
Longitudinal Transverse Normal Fusion Line HAZ Boundary
100
100
0
0
-100
-100 -15
-10
-5
0
5
10
15
Distance from the weld centreline (mm)
Fig. 7. Cross-weld residual stress and microhardness profile in Plate A (8 kW, 1.5 m/min) at 1.5 mm below top surface.
-200 -15
-10
-5
0
5
10
15
Distance from the Weld Centreline (mm)
Fig. 8. Cross-weld stress profile in Plate A (8 kW, 1.5 m/min) at 4.5 mm below surface.
166
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
that it is the martensitic transformation in the fusion zone and HAZ that governs the width of the residual stress profile. 3.6. Through-thickness stress profiles Through-thickness profiles of the longitudinal, transverse and normal components of residual stress at the weld centre-line in Plate A are shown in Fig. 9. It can be observed that the uncertainties associated with these weld metal measurements are much higher ( 730 MPa) than those in the parent metal region ( 710 MPa (Figs. 7 and 8)). This is on account of higher microstresses in the fusion zone and the HAZ as compared to the same in the parent metal. The transverse component of residual stress is most tensile, the normal component most compressive and the longitudinal component lies in between. The final residual stress state in the weld metal is the result of competition between martensitic transformation induced dilation and thermal contraction. In the normal direction, thermal contraction will be smaller than that in the longitudinal direction; therefore, the normal component of residual stress is more compressive than the longitudinal component. Residual stress below the mid thickness plane is more compressive than that above the mid plane. This is because, the heat source is applied from the top side, leading to less heat input below the mid-thickness plane and therefore, the dilative martensitic transformation commences first in this region, producing a more compressive region than that above. The variation in residual stresses through the thickness at the weld centre-line is most likely to be associated with differential cooling between the bottom and top of the weld. Greater heat is input by the laser welding process at the top surface; it can be noted from Fig. 1 that the fusion zone and the HAZ are slightly narrower towards the bottom side of the plate. This means that the bottom of the welded plate will cool faster than the top and lead to a resultant residual stress distribution controlled by the combined influences of differential thermal contraction, different degrees of martensitic transformation and self-constraint. Numerical weld computational mechanics is probably the best way to understand the complex mechanical behaviour involved. This sort of residual stress profile with a compressive zone below the midthickness plane and tensile zone above contrasts from that observed in the case of multi-pass arc-welded joints in this material [12] where significantly higher tensile residual stress develops below the mid-thickness plane relative to above it. This is on account of the thermal cycles associated with the succeeding passes introducing martensitic dilation and more compressive stresses in final weld capping passes.
Residual stress maps for longitudinal and normal components on the mid-length cross-section of the low heat input welded plate were generated on one side of the weld centre-line; using the through thickness measurements along the weld centre-line and along the measurement lines at distances of 2 mm, 4 mm and 10 mm from it. For distances of 15 mm and 30 mm the weld centre-line, values from cross-weld measurements at 1.5 mm and 4.5 mm below the top surface were used to generate these plots. The plots were generated using MATLAB. Figs. 10 and 11 show maps of the longitudinal and normal components respectively of residual stress on one side of the welded joint. These maps show that the cross-weld profiles for the longitudinal and the normal components represented in Figs. 7 and 8 exist through the entire thickness of the plate. It can also be seen from these maps that the cross-weld profiles of the longitudinal and normal components of stress are very similar in nature, i.e. both components show a low tensile or compressive trough in the weld centre and two tensile peaks on the either side of the joint in the parent metal adjacent to the PM–HAZ boundary. 3.7. Effect of welding speed on residual stress profile Figs. 12–14 compare cross-weld profiles of the longitudinal, normal and transverse components of residual stress respectively in Plates A and B, which were welded at the same laser power (8 kW) but with different welding speeds (1.5 m/min and 0.75 m/min, respectively) giving low and high heat input conditions. A cross-weld
Fig. 10. Map of longitudinal component of residual stress in Plate A at the midlength cross-section in the one side of the weld centre-line.
Fig. 11. Map of normal component of residual stress in Plate A at the mid-length cross-section in the one side of the weld centre-line.
300 600 Longitudinal Transverse Normal
HAZ Boundary_Plate A
500
Residual Stress (MPa)
400 100 300 0 0
1
2
3
4
5
6
7
8
200
Fusion Line_Plate B
Longitudinal Stress (MPa) Microhardness (HVN0.5)
200
HAZ Boundary_Plate B Longitudinal_Plate A Longitudinal_Plate B Microhardness_Plate A Microhardness_Plate B Fusion Line_Plate A
Distance from the top surface (mm) -100
100 0
-200 -100 -15 -300
Fig. 9. Through-thickness profiles of the longitudinal, transverse and normal components of residual stress in Plate A at the weld centre-line.
-10
-5
0
5
10
15
Distance from the weld centreline (mm)
Fig. 12. Cross-weld profile of the microhardness and longitudinal component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
200
100
Residual Stress (MPa)
300
Normal_Plate A Normal_Plate B Fusion Line_Plate A HAZ Boundary_Plate A Fusion Line_Plate B HAZ Boundary_Plate B
0
-100 -15
-10
-5
0
5
10
15
Distance from the weld centre line (mm)
Fig. 13. Cross-weld profile of the normal component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
250 Transverse_Plate A 200
Transverse_Plate B
Re 150 sid ua l 100 Str es 50 s
Fusion Line_Plate A HAZ Boundary_Plate A Fusion Line_Plate B HAZ Boundary_Plate B
3. There is little variation in residual stress through the plate thickness except in the fusion zone, where longitudinal and normal stresses are more compressive in the bottom half of the weld zone. This contrasts with conventional multi-pass welds in P91 steel where the residual stress is more tensile in the bottom side than in the top side. 4. The widths of the fusion zone (1.2 mm for the low heat input and 1.8 mm for high heat input) and the low tensile/compressive troughs in the residual stress profiles (2 mm wide for low heat input and 3 mm for high heat input) were wider and marginally deeper for the high heat input weld compared with the low heat input weld, but the peak magnitudes of stress were similar. 5. There is clear evidence of a very strong influence of the strains associated with the martensitic transformation on the magnitude and spatial profile of all three components of residual stress in both welded plates. 6. The most significant components of residual stress in the laser welds are in the longitudinal and normal directions. This contrasts with conventional multi-pass welds where longitudinal and transverse components are the most significant. This is mainly owing to the high aspect ratio (depth to width ratio) of the fusion zone in laser welds. 7. The width (FWHM) of the diffraction peaks shows one to one correlation with as transformed region on account of increased microstresses.
Acknowledgements
0 -50 -100 -15
167
-10
-5
0
5
10
15
The research was supported by the Indian Department of Atomic Energy and by grant EP/101215X/1 from the UK Engineering and Physical Research Council. The authors gratefully acknowledge the award of neutron beam time by the ILL Grenoble, France.
Distance from the weld centerline (mm)
Fig. 14. Cross-weld profile of the transverse component of residual stress in Plate A and Plate B at 1.5 mm below top surface.
microhardness profile is also superimposed. The shapes and peak magnitudes of the profiles from the two plates are almost identical. However, it can be seen that the lower welding speed has led to both broadening and deepening of the trough in the cross-weld profiles of residual stress in both the longitudinal and normal directions. The increased width of the trough correlates with the wider fusion zone in Plate B which experienced nearly double the heat input of A; compare the fusion zones in Fig. 1(a) and (b). These results further illustrate how the martensitic transformation in the fusion zone and HAZ of a laser welded joint in P91 steel leads to reduced levels of residual stress in the weld region.
4. Conclusions 1. Neutron diffraction has been used to measure as-welded residual stresses in 9 mm thick Grade 91 steel plates that were joined using constant power (8 kW) low (320 J/mm) and high (640 J/mm) heat input laser welding procedures. 2. Cross-weld profiles of the longitudinal and normal components of residual stress in both of the laser welds show a low tensile or compressive trough (−250 MPa to 100 MPa) in the welded fusion zone and high tensile peaks (500 MPa to 600 MPa for longitudinal and 200 to 350 for normal) in the parent metal near the HAZ–PM interface on the either side of the welded joint.
References [1] A. Shibli, F. Starr, Int. J. Press. Vessels Pip. 84 (2007) 114–122. [2] K. Laha, K.S. Chandravathi, P. Parameshwaran, K.Bhanu Sankara Rao, S. L. Mannan, Metall. Mater. Trans. A 38 (2007) 58–68. [3] S. Pillot, Z. Zhao, S. Corre, C. Chauvy, L. Coudreuse, P. Toussaint. Proc. of ASME 2010 Press. Vessel. and Pip. Div. Conf. (PVP 2010) 1–14. [4] R. Mythili, P.V. Thomas, S. Saroja, M. Vijaylakshmi, V.S. Raghunathan, J. Nucl. Mater. 312 (2003) 199–206. [5] R.K. Shiue, K.C. Lan, C. Chen, Mater. Sci. Eng., A 287 (2000) 10–16. [6] P.K. Ghosh, U. Singh, Sci. Technol. Weld. Joining 9 (3) (2004) 229–236. [7] H.O. Andrén, G. Cai, L.E. Svensson, Appl. Surf. Sci. 87–88 (1995) 200–206. [8] S.K. Albert, M. Tabuchi, H. Hongo, T. Watanabe, K. Kubo, M. Matsui, Sci. Technol. Weld. Joining 10 (2) (2005) 149–157. [9] Z. Xu, Report No. ANL/NE-12/12, Argonne National Laboratory, Chicago, Illinois, 2008, September. [10] K. Seok-Hoon, K. Jong-Bum, L. Won-Jae, J. Mater. Process. Technol. 209 (2009) 3905–3913. [11] D. Dean, M. Hidekazu, Comput. Mater. Sci. 37 (2006) 209–219. [12] S. Paddea, J.A. Francis, A.M. Paradowska, P.J. Bouchard, I.A. Shibli, Mater. Sci. Eng., A 534 (2012) 663–672. [13] N.S. Rossini, M. Dassisti, K.Y. Benyounis, A.G. Olabi, Mater. Des. 35 (2012) 572–588. [14] P.J. Withers, C. R. Phys. 8 (2007) 806–820. [15] G.A. Webster, R.C. Wempory, J. Mater. Process. Technol. 117 (2001) 395–399. [16] W. Wanchuck, E.M. Vyacheslav, M. Pavel, A. Gyu-Baek, S. Baek-Seok, Mater. Sci. Eng., A 528 (2011) 4120–4124. [17] M.J. Park, H.N. Yang, D.Y. Jang, J.S. Kim, T.E. Jin, J. Mater. Process. Technol. 155– 156 (2004) 1171–1177. [18] A. Paradowska, J.W.H. Price, R. Ibrahim, T. Finlayson, J. Mater. Process. Technol. 164–165 (2005) 1099–1105. [19] J.W.H. Price, A. Paradowska, S. Joshi, T. Finlayson, Int. J. Press. Vessels Pip. 83 (2006) 381–387. [20] M. Hofmann, R.C. Wimpory, Int. J. Press. Vessels Pip. 86 (2009) 122–125. [21] C. Ohms, R.C. Wimpory, D.E. Katsareas, A.G. Youtsos, Int. J. Press. Vessels Pip. 86 (2009) 63–72. [22] S. Pratihar, M. Turski, L. Edwards, P.J. Bouchard, Int. J. Press. Vessels Pip. 86 (2009) 13–19.
168
S. Kumar et al. / Materials Science & Engineering A 575 (2013) 160–168
[23] D. Thibault, P. Bocher, M. Thomas, M. Gharghouri, M. Côté, Mater. Sci. Eng., A 527 (2010) 6205–6210. [24] W. Wanchuck, E.M. Vyacheslav, C.R. Hubbard, H.J. Lee, P.Kwang Soo, Mater. Sci. Eng., A 528 (2011) 8021–8027. [25] M. Rogante, P. Battistella, F. Rustichelli, J. Alloy. Compd. 378 (2004) 335–338. [26] W. Suder, S. Ganguly, S. Williams, A.M. Paradowska, P. Colegrove, Sci. Technol. Weld. Join. 16 (3) (2011) 244–248.
[27] D.W. Brown, T.M. Holden, B. Clausen, M.B. Prime, T.A. Sisneros, H. Swenson, J. Vaja, Acta Mater. 59 (2011) 864–873. [28] S. Ganguly, L. Edwards, M.E. Fitzpatrick, Mater. Sci. Eng., A 528 (3) (2010) 1226–1232. [29] B.J. Ganesh, S. Raju, K.R. Arun, E. Mohandas, M. Vijayalakshmi, K.Bhanu Sankara Rao, R. Baldev, Mater. Sci. Technol. 27 (2) (2011) 500–512. [30] A.F. Mark, J.A. Francis, H. Dai, M. Turski, P.R. Hurrell, S.K. Bate, J.R. Kornmeier, P. J. Withers, Acta Mater. 60 (2012) 3268–3278.