Sunteți pe pagina 1din 12

International Journal of Pressure Vessels and Piping 82 (2005) 258269 www.elsevier.

com/locate/ijpvp

Analysis of residual stresses at weld repairs


P. Donga,*, J.K. Honga, P.J. Bouchardb
a

Center for Welded Structures Research, Battelle Memorial Institute, 505 King Avenue, Columbus, OH 43201, USA b British Energy Generation Ltd, Barnwood, Gloucester, UK

Abstract In contrast to initial fabrication welds, residual stresses associated with nite length weld repairs tend to exhibit some important invariant features, regardless of actual component congurations, materials, and to some degree, welding procedures. Such invariant features are associated with the severe restraint conditions present in typical repair welding situations. In this paper, residual stress results from several weld repair case studies, using both advanced computational modelling procedures and experimental measurement techniques, are presented and reviewed. From these results, it is evident that weld repairs typically increase the magnitude of transverse residual stresses along the repair compared with the initial weld and that the shorter the repair length the greater the increase in the transverse stress. Also, beyond the ends of the repair the transverse stress sharply falls into compression. For selected cases, predicted stresses are compared with detailed residual stress measurements and the adequacy of nite element simulation procedures is assessed. Welding procedure related parameters (pass lumping, heat input and inter-pass temperature) appear to be more important in analysing weld repairs than in initial fabrication welds. Also great care must be taken when employing simplied two-dimensional cross-section nite element models with applied restraint conditions to simulate the residual stress eld at a specic point along the length of a repair. q 2004 Elsevier Ltd. All rights reserved.
Keywords: Finite element; Residual stress; Pressure vessel and piping

1. Introduction Over the last decade or so, welding-induced residual stresses have received increasing attention in the pressure vessel and piping research community. The driving force for this interest can be attributed to the fact that application of modern structural integrity assessment procedures for defective welded components (e.g., BS7910:1999 [1], R6 [2], and API RP-579, 2000 [3]) require more accurate information on the weld residual stress state to give a more realistic assessment. The conventional approach for characterising a weld residual stress prole is to adopt an upper bound solution. However, as reviewed recently by Bradford [4], Dong et al. [5], Bouchard and Bradford [6], this approach not only lacks consistency for the same type of joints and welding parameters [6], but can either signicantly over-estimate the residual stress level in some cases [5,6], or under-estimate it in others [7,8].
* Corresponding author. Tel.: C1-614-424-6424; fax: C1-614-4245263. E-mail address: dongp@battelle.org (P. Dong). 0308-0161/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijpvp.2004.08.004

As weld repairs have increasingly become a structural integrity concern for aging pressure vessel and piping components, the need for better characterisation of residual stresses at repairs has become more evident. Both repair procedure development and the subsequent safety assessment require a better understanding of the repair welding effects on structural components [7,9,10]. This is because weld repair residual stress distributions can be drastically different from those in original fabrication welds, typically with the presence of higher tensile residual stresses than in original fabrication welds [11]. As advanced computational modelling techniques [12] as well as new and improved experimental methods have become available over recent years [13], more accurate residual stress information can now be obtained for various structural integrity assessment applications. As discussed by Dong et al. [10], residual stresses in weld repairs typically exhibit strong three-dimensional (3D) features, depending on both component and repair geometries. As a result, 3D effects should be taken into account in both experimental measurements and numerical modelling. However, even with todays fast computer

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

259

speed and advanced numerical procedures, 3D nite element (FE) solid element models used in simulations for multi-pass repairs to realistic component congurations are very challenging. To date, most residual stress analysis results for repair welds are based on 2D or axi-symmetric models with applied restraint conditions. These conditions may be either assumed [12], or equivalent conditions derived from a 3D-shell weld model [10], or with timedependent generalized plane-strain conditions obtained from a 3D-shell weld model [14]. An alternative cost effective approach is to use a composite shell element model, such as developed by Zhang et al. [15], to capture some of the 3D global residual stress features in repair welds. The success of this approach in providing an adequate resolution scale for through-thickness distributions has been demonstrated for a repair weld in a 35 mm thick pipe, by comparing with deep hole drilling measurements [10]. With the development of improved residual stress measurement techniques [13] such as neutron diffraction and deep hole drilling, more comprehensive and reliable experimental data are becoming available, for example [11]. These data allow a more discriminative assessment of weld residual stress simulations including the adequacy of simplied models and the relative importance of different weld modelling parameters. Before modelling or measuring a weld residual stress eld, it is of fundamental importance to understand what kind of residual stress eld to expect so that an effective numerical modelling scheme or experimental method can be devised to capture it. This paper reviews residual stress results from several weld repair case studies using both advanced computational modelling procedures and experimental measurement techniques. From these results general characteristics of repair weld residual stress elds are then inferred. This is followed by a detailed discussion of a simplied 2D cross-section analysis for a weld repair using equivalent applied restraint conditions to account for 3D effects. For selected cases, predicted stresses are compared with detailed residual stress measurements and from this the adequacy of FE simulation procedures used is assessed.

joining two 610!152!5 mm3 panels. The two-pass weld was modelled with a moving heat source model [15]. Fig. 1b shows that the predicted distribution of longitudinal residual stress was reasonably uniform along the weld length, except near the welding torch start and stop positions. The peak value of tensile stress exceeded the material yield strength (414 MPa). In contrast, the transverse residual stress component showed a signicant variation along the weld, see Fig. 1c, ranging from a tensile value of 138 MPa around 50100 mm from the stop position to a compressive value of about K276 MPa at the stop position. X-ray residual stress measurements in the heat-affected zone (HAZ) conrmed such a variation along the weld direction, as shown in Fig. 1d. These results illustrate how in initial fabrication welds, the longitudinal component of residual stress (i.e. that parallel to the direction of welding) is the dominant component, often reaching and exceeding the magnitude of the tensile yield strength of the material. The transverse component of residual stress typically exhibits lower residual stress magnitudes, with a distribution that is strongly dependent upon welding procedures and restraint conditions specic to the component of concern. When a short length weld repair was introduced at the mid-length of the initial butt-weld in the same specimen (Fig. 1a), this resulted in a strongly varying pattern of transverse residual stress surrounding the repair, as seen in Fig. 2a. The magnitude of transverse tensile residual stresses in the region of the repair was high and approached that of the longitudinal stress component (Fig. 2b). Moreover, the transverse residual stress eld had a long sphere of inuence in the transverse direction. Immediately beyond the repair start and stop positions, two distinct compressive zones of transverse stress were predicted. Less marked compressive zones are also evident in the predicted distribution of longitudinal stresses, but the overall distribution was broadly similar to that of the initial butt weld, compare Fig. 2b with Fig. 1b. 2.2. Repaired carbon steel vessel A recent tness-for-service assessment for a series of high level radioactive waste tanks (radius to thickness ratio of about 800) containing weld repairs employed nite element methods to estimate residual stresses for various repair weld conditions. The storage tanks were made of ASME Div. 2 A285 carbon steel. The girth welds and local weld repairs were made by shielded metal arc welding (SMAW) processes with E6010 electrodes. A majority of stress corrosion cracks found in these tanks by remote nondestructive inspection techniques were seen around weld repairs. However, detailed repair weld information such as the pass sequence, repair depths, and repair lengths were not available at the time of the evaluation. Several approximations in the weld simulation procedure were adopted in order to provide conservative estimates of overall residual stress distributions for tness-for-service assessments of

2. Case studies 2.1. Repaired aluminium alloy butt weld This case study illustrates important differences between the residual stress eld associated with an initial fabrication weld and that subsequently introduced by nite length weld repairs. Dong et al. [14] have studied the inuence of introducing a repair into a butt-welded aluminium alloy (AlLi) cryogenic space shuttle tank mock-up, see Fig. 1a. In this investigation, a special 3D shell element model [15] was used to simulate the welding process for the initial butt weld

260

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

Fig. 1. Residual stresses in an initial longitudinal butt weld of a long aluminium alloy test panel: (a) specimen geometry and dimensions (b) predicted transverse residual stress, (c) predicted longitudinal residual stress, and (d) comparison of predicted with measured residual stresses in the HAZ adjacent to the initial butt weld.

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

261

Fig. 2. Predicted residual stresses in aluminium test panel (Fig. 1a) after introducing a repair weld: (a) transverse residual stress, (b) longitudinal residual stress.

potential crack growth at repair locations. A lumped-pass weld simulation technique was used, that is without considering moving-arc weld metal deposition effects. The heat transfer analysis procedures were tuned to achieve the same size of HAZ as a moving arc weld. As discussed in [16], such a simplication tends to over-estimate the overall residual stresses in weld repairs but ignores localized stress concentration features associated with a moving-arc [10,16]. Solid state phase transformation effects were not modelled since they were judged to contribute to higher

order perturbations of the underlying residual stress distribution [12] that are of less structural signicance. These simplications were adopted to avoid employing a much more rened 3D nite element mesh, given that the intent of the investigation was to provide a conservative estimate of the effects of repair depth, length, and width on the overall residual stress distributions. Predicted distributions of transverse residual stresses introduced by a low heat input repair are shown in Fig. 3 and illustrate some of the important residual stress features

Fig. 3. Predicted residual stresses transverse to the welding direction in a 16 mm thick and 5000 mm diameter carbon steel vessel, for repair widths w and 2w, repair depths 1/4t and 1/2t and repair lengths 152 mm and 305 mm (only 1/2-repair weld length with stop position shown).

262

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

associated with repair welds. The contour plots show a through-wall tension-compression-tension distribution along the length of the repair, but with signicantly increased tension on the repair side of the surface (i.e. the outer surface) compared with the initial weld when the repair length is relatively small (Cases 13 in Fig. 3). As the repair depth increases from 1/4t (t is the base plate thickness) to 1/2t (Case 1 versus Case 3), the difference in the transverse residual stresses is mostly seen in the throughthickness distributions. As the repair weld width increases from 1 to 2w (Case 2 versus Case 3), the area subjected to tensile residual stresses area is signicantly increased, where w represents the initial weld width. It is of interest to note that the compressive zone beneath the repair starts to disappear when the repair width, w, or repair depth is increased. This is largely due to fact that an overall throughthickness structural reaction induced by the increased repair weld metal shrinkage force starts to dominate the residual stress distribution. If the repair length is doubled from 152 to 305 mm (Case 2 versus Case 4), a marked change in the transverse residual stress distributions can be seen, with signicantly reduced levels of tensile stress along the repair. Further discussion on repair length effects is given in a later section of this paper. 2.3. Repaired stainless steel pipe girth weld (35 mm thick) Residual stresses introduced by nite length repairs to an AISI Type 316H stainless steel pipe girth weld (35 mm thick and 541 mm outer diameter) have been studied using

the composite shell modelling procedures described in [15,16]. Three repairs of varying lengths, all symmetrically positioned on the original girth weld centre-line, with a repair depth 75% of the wall thickness from the outer surface were analyzed, see Fig. 4. A moving-arc weld metal deposition analysis for each of four lumped-bead passes was employed with an assumed inter-pass temperature in the range 170200 8C. The direction of welding was alternated between passes. The predicted variations in axial (i.e. transverse) residual stress around one-half of the circumference at the weld centre-line on the inner and outer surfaces are shown in Fig. 5a and b. The repair weld lengths (angular arc) are also marked on these graphs. The magnitudes of the original girth weld residual stresses are evident at circumferential positions far from the ends of the 20 and 558 arc-length repairs, that is about 230 MPa at the OD and about 200 MPa at the ID surface. All the repairs increase the axial residual stress levels within the repair length. Beyond the ends of each repair, the stress falls rapidly into compression before approaching the stress state corresponding to the initial weld. Fig. 5a shows that the shorter the repair the higher are the tensile residual stresses at the outer surface along the length of the repair. Fig. 6 shows the predicted through-thickness axial residual stress distributions at the weld centre-line and near the fusion line of the repair. The distributions are highly non-linear due to multi-pass deposition effects and very sensitive to axial position relative to the weld. The proles at both locations exhibit a membrane stress (i.e. the uniformly distributed stress giving the same

Fig. 4. Composite shell nite element model for multi-pass repair welding simulation in a 35 mm thick, 541 mm OD stainless steel pipe.

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

263

Fig. 5. Comparison of predicted axial residual stress distributions along the original/repair weld centre-line for three repair lengths (short208, medium558, and long1108 arc-lengths) in a 35 mm thick, 541 mm OD stainless steel pipe (Fig. 4).

integrated force) that decreases as the repair length becomes longer, and an overall bending character that is tensile at the outer surface. Residual stress measurements have been performed on a mock-up of the repaired pipe [17] using the deep hole drilling technique. The measured results for the through-thickness distribution at Point B in Fig. 6 are shown in Fig. 7. The overall agreement between the measured and predicted stresses is good despite the limitations of the composite shell element model with lumped-bead weld passes. 2.4. Repaired stainless steel pipe girth weld (19 mm thick) Recently, high quality residual stress measurement data for a 19.6 mm thick, 432 mm OD pipe girth weld mock-up containing offset weld repairs have become available [11]. Prior to fabrication of this mock-up and the measurements programme, residual stresses in this type of repaired component were investigated using a 3D shell FE weld simulation model by Dong et al. [10,18] using assumed repair welding conditions. A model for a 19 mm thick, 541 mm OD pipe with a girth weld was constructed using special composite shell elements having four layers (Fig. 8). This allowed three layers to be used to simulate the deposition of lumped-bead repair weld passes to a depth of 75% of the shell thickness. Structural symmetry along the weld centre-line was assumed to simplify the analysis (i.e. the repair offset was ignored). The pipe original girth weld residual stress eld, from a 2D FE simulation model,

Fig. 6. Comparison of predicted through-thickness axial residual stress distributions at repair mid-length for three repair lengths (short208, medium558, and long1108) in a 35 mm thick, 541 mm OD stainless steel pipe (Fig. 4); (a) at Point A (HAZ), and (b) at Point B (weld centreline).

was rst mapped onto the 3D shell model giving an initial axi-symmetric distribution of as-welded residual stress. Local excavation of the repair groove process was then represented using a layer activation/deactivation scheme; this resulted in a re-distribution of the original weld residual stresses. The repair weld simulation involved two steps; an analytical thermal analysis to predict moving weld-arc transient temperatures for the three consecutively deposited lumped-beads, followed by an ABAQUS mechanical analysis based on the temperature history. The direction of welding was alternated between the three lumped-bead repair passes. The analysis employed temperature dependent material properties, annealing of historical plastic strains on melting and isotropic hardening behaviour.

264

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

Fig. 7. Predicted through-wall residual stress distributions in the HAZ (Point B in Fig. 6) at mid-length of a short (z208 arc-length) weld repair to a 35 mm thick, 541 mm OD stainless steel pipe girth weld, compared with measured stresses for centrally embedded 75% wall-thickness weld repair to a 37 mm thick, 432 mm OD stainless steel pipe mock-up; (a) axial residual stress, and (b) hoop residual stress.

concentration end effects have not been observed in residual stress measurements from the long (z608 angular span) weld repair in the 19.6 mm thick, 432 mm OD pipe mockup [11]. It is open to debate as to whether the predicted end effects are artefacts of the analysis procedure or whether a more complete mapping of measured residual stress on a circumferential-radial cross-section of the repair weld might reveal a more complex pattern of stress. Comparisons between the measured [11] and predicted through-thickness distributions of hoop and axial stress in the HAZ at mid-length of a short (z208 angular span) repair are shown in Fig. 10. It is evident that both deep hole drilling and neutron diffraction techniques provided remarkably consistent measurements for the throughthickness stress distributions. As for the pre-measurement 3D special shell model predictions, the agreement with the measured data seems to be poor, at rst glance. However, the predicted membrane stress levels are similar and within the repair depth (i.e. outer 2/3rd of the pipe thickness), the overall trends between the predicted and measured can be viewed as consistent, even though the magnitudes differ signicantly. A careful examination of the mock-up fabrication conditions, measurement data, and the earlier 3D shell element study assumptions [10,18] was carried out. The following are believed to contribute the discrepancies shown in Fig. 10: (a) The actual weld repair was axially offset from the original girth weld centre-line by 12 mm (see Fig. 10 inset), whereas the modelled repair was assumed to be symmetrically aligned with the girth weld centre-line. (b) The repair weld passes were performed in an essentially continuous manner, only leaving enough time for removing slag in between passes. This would have resulted in the development of inter-pass temperatures well above room temperature. In the 3D shell element model, each lumped-bead pass was deposited when the previous pass had cooled down to room temperature. Here it is worth noting that in the 35 mm thick pipe repair case discussed earlier, an inter-pass temperature of 170200 8C was simulated and this resulted in a bending dominant type of through-wall residual stress distribution near the fusion line (see Figs. 6 and 7). (c) The actual welding conditions were not simulated owing to the use of three lumped-bead weld passes in the 3D shell FE model (instead of 12 passes actually used). In general, lumped-bead weld simulation procedures inherently over-estimate the actual heat input per unit time to a weldment. To investigate observations (b) and (c), a sensitivity FE analysis was performed by dening a 6-pass conguration (see Fig. 8) in the same 3D shell element model [10,18] and depositing each of the passes simultaneously, that is using a lumped-pass procedure without moving-arc effects.

Predicted axial residual stresses on the outer surface of the pipe are shown in Fig. 9 for three repair weld lengths. The plots illustrate the general pattern of stress associated with repair welds which is similar to that described above for the repaired aluminium plate butt weld. For the repair case with a z208 angular span, the overall distribution is essentially the same as the one shown in Fig. 2a. As the repair length increases, the high tensile axial (i.e. transverse) residual stresses are predicted to become more concentrated towards the repair ends with peak values occurring in the area near the stop position. However, such stress

Fig. 8. Details of composite shell nite element models (3-pass and 6-pass) used for simulating multi-pass repairs in a 19 mm thick, 541 mm OD stainless steel pipe girth weld.

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

265

Fig. 9. Predicted repair weld axial residual stresses (from 3-pass model) on the outer surface of a 19 mm thick, 541 mm OD stainless steel pipe girth weld for three repair lengths (short208, medium578, and long1148 arc-lengths).

Fig. 10. Comparison of predicted through-wall residual stresses in the HAZ at mid-length of a short (208 arc-length) weld repair to a 19 mm thick 541 mm OD pipe (see Figs. 8 and 9) with measured residual stresses in the HAZ (see inset) at mid-length of a short (z208 arc-length) weld repair in a 19.6 mm thick, 432 mm OD stainless steel pipe mock-up; (a) axial residual stress, and (b) hoop residual stress.

266

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

This was achieved by depositing each of the 6 lumped-passes at the melting temperature with a hold time corresponding to the time for the weld torch to travel from end to end of the short repair. This approximate procedure did not accurately model the effective heat input of the repair passes, but was adopted as a pragmatic approach. Each of the deposited passes was allowed to cool down to 200 8C before the deposition of the next pass. The new 3D shell model results show a much closer correlation with the experimental measurements, see Fig. 10. This nding demonstrates how the predicted through-wall residual stress distribution at a repair weld is highly sensitive to the details of the repair weld simulation procedure, that is the number of passes, the welding heat input and the assumed inter-pass temperature.

3.1. Case study:-repaired aluminium alloy butt weld In the repaired aluminium alloy butt weld example discussed in Section 2.1 above, a 3D special shell element was used to capture the overall in-plane residual stress characteristics (see Fig. 2). The corresponding local residual stress distributions were analyzed using a 2D cross-section model under generalized plane strain conditions, as discussed in [16]. The 2D cross-section model is shown in Fig. 11 for the repair weld. In this analysis, the cross-section model was intended to approximate the residual stress state corresponding to a unit slice from the mid-length of the weld repair in Fig. 2. The following assumptions were introduced. The line corresponding to the point at the right end of the 2D model in Fig. 11a remains as a line during welding, allowing translation in both x and y directions. In this case, it was convenient to use one element in the through thickness direction and to introduce a triangular element for imposing the displacement boundary conditions (see Fig. 11a) in the form ux(t), uy(t), where t is the time from start of welding. Secondly, a plane corresponding to the xy plane of the 2D model remains as plane, i.e. uz tZ atxC btyC ct; where a(t), b(t), and c(t) signify the time-dependency of these coefcients during welding and measure rotations of the plane with respect to the x and y axes, and translation in z, respectively. To obtain ux(t), uy(t), and uz(t), the special shell element model [15], as used for generating the 3D residual stress results shown in Fig. 2 [14] is particularly effective. For the case shown in Fig. 2, ux(t), uy(t) were obtained at a boundary node of the shell model at the mid-length of the specimen. Note that in this process

3. 2D cross-sectional FE models for repairs The examples presented above illustrate that repair welds introduce complex three-dimensional distributions of residual stress. Nonetheless, if the important features of the residual stress eld for a specic repair scenario can be identied then it is possible, and desirable, to tailor the nite element analysis approach that is adopted to quantify the stress conditions of interest. For example, if the overall residual stress distribution is of interest, then simplied analysis procedures using 2D cross-section or axi-symmetric assumptions can be used with equivalent restraint conditions applied to account for three-dimensional effects. However, such 2D repair weld models only provide a residual stress solution at a given location along the repair weld length. Some of the issues associated with such simplied approaches are claried in the following case study.

Fig. 11. 2D cross-section nite element model under generalized plane strain conditions and its relationship to a 3D shell element model of the aluminium alloy test panel shown in Fig. 1.

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

267

the rotational term from the shell element model at the node can be ignored for most applications. Otherwise, signicant mesh renements in 2D cross-section models are typically required in order to capture such rotation-induced shear within the vicinity of the boundary. The function uz(t) was obtained as relative displacements between two adjacent planes (two parallel lines with a unit distance in between) transverse to the weld direction in the shell model. After a series of parametric studies was performed [19], it was found that ux(t), uy(t), and uz(t) could be treated in the 2D cross-section model as constants that correspond to the nal values from the welding simulation of the shell model. It is worth noting that axi-symmetric conditions can be recovered as one specic case in such a generalized cross-section model by imposing an appropriate uz(t). With the above approach, both the initial weld (Fig. 1a) and the repair weld (Fig. 2) were analyzed using the 2D cross-section model depicted in Fig. 11b. The residual stress results are compared with X-ray measurements in Fig. 12 along the same cross section that was modelled. The measured and predicted residual stresses on the top surface of the specimen are plotted with respect to the distance measured from the weld centre-line. The measured and predicted longitudinal residual stresses compare reasonably well with each other (Fig. 12a). The rapid reduction in the predicted longitudinal residual stresses across the fusion

boundary and inside the weld metal is due to the use of under-matched ller metal (about 50% lower in yield strength [14]). The discrepancies within this region can be attributed to the resolution scale in the X-ray measurement techniques and surface conditions due to the presence of the weld bead prole. The effects of repair on the longitudinal residual stress magnitude are not signicant. As discussed earlier, restraint conditions for the longitudinal residual stress development are already high even under initial welding conditions. The predicted and measured transverse residual stresses in the initial weld (see Fig. 12b) are of low magnitude relative to the peak longitudinal stresses. The overall agreement between the predicted and measured results is evident. The cross-section model predicted that the introduction of a repair signicantly increased the transverse residual stresses. However, it is worth noting here that the measured transverse residual stresses tend to be higher than the predicted values, but with an overall trend being consistent with the predicted one. This suggests that the 2D cross-section model tends to under-estimate the transverse residual stresses in this case due to the inherent assumptions at the boundary associated with the denitions of ux(t), uy(t), and uz(t) discussed earlier. Examination of the aluminium alloy butt weld specimens showed noticeable out-of-plane deformations that were not captured by the linear deformation assumptions described above. However, for thicker sections such discrepancies are expected to be less signicant.

4. Discussion Regardless of the overall component geometry (e.g. plate structures versus pipes or vessels) and materials (aluminum alloys versus stainless steels or carbon steels), the overall distribution of residual stress distribution associated with nite length weld repairs share the following invariant features: (a) weld repairs increase the magnitude and importance of transverse residual stresses along the repair compared with the initial weld, (b) the shorter the repair length the greater the increase in the transverse stresses, but for very long repairs the transverse residual stresses within the central region of the repair length approaches that of the initial weld, (c) tensile transverse residual stresses along the length of the repair sharply fall into compression beyond the ends of the repair, and (d) the through-thickness variation within the repair is inuenced by the repair width, repair length, heat input per pass, and inter-pass temperature. The above observations are based on either numerical predictions or systematic measurements from a range of

Fig. 12. Comparison of X-ray residual stress measurements and nite element predictions using a 2D cross-section model (Fig. 11) with applied restraint conditions derived from a 3D model: (a) longitudinal residual stresses, and (b) transverse residual stresses.

268

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269

different applications. These important common features are governed by the high restraint conditions typically associated with repair welding. Any deviations from these general features can be attributed to the change in restraint conditions associated with specic weld repair applications. From the case studies presented in Sections 2 and 3, it has been inferred that the welding thermal conditions (e.g. number of passes, heat input and inter-pass temperature) are important parameters in repair weld residual stress analyses. This is because the mechanical restraint conditions between passes can be altered noticeably and this should be accounted for in performing residual stress analysis. For example, Fig. 10 shows how the through-wall bending component of stress is signicantly changed by the assumed welding conditions. Such sensitivity is typically not seen in analyzing initial fabrication welds. Further investigations are needed to establish appropriate pass lumping procedures and their implications on an equivalent heat input for repair applications.

Acknowledgements The authors acknowledge funding by British Energy for some of the repair case studies on Type 316 stainless steel girth welds. This paper is published with the permission of British Energy Generation Ltd.

References
[1] British Standards Institute. Guide on methods for assessing the acceptability of laws in metallic structures. BS7910:1999, Amendment No. 1, BSI: London; 2000. [2] British Energy. Assessment of the Integrity of Structures Containing Defects. Procedure R6 revision 4: British Energy Generation Ltd, Gloucester, 2; 2003. [3] API RP-579. First edition, Recommended practices for tness-forservice; 2000. [4] Bradford RAW. Through-thickness distributions of welding residual stresses in austenitic steel cylindrical butt welds. Proceedings of sixth international conference on residual stresses (ICRS-6). London: IOM Communications Ltd: 2000, p. 1373381. [5] Dong, P., Osage, D., Prager, M., Development of weld residual stress distributions for tness for service assessment. Proceedings of ASME pressure vessels and piping. New York: ASME; 2000; PVP-411:5364. [6] Bouchard PJ, Bradford R. Validated axial residual stress proles for fracture assessment of austenitic stainless steel pipe girth welds, pressure vessel and piping. Proceedings of ASME conference, New York : ASME;2001;PVP-423: 9399. [7] Dong P, Brust FW. Welding residual stresses and effects on fracture in pressure vessel and piping components: a millennium review and beyond, ASME Trans J Press Vessel Technol 2000;122(3):32838. [8] Dong P, Rahman S, Wilkowski G, Brickstad B, Bergman M, Bouchard PJ, Chivers T. Effect of weld residual stresses on crackopening area analysis of pipes for LBB applications. Proceedings of ASME pressure vessel and piping conference. New York: ASME;1996;PVP-324:4764. [9] Zhang J, Dong P, Hong JK, Bell W, McDonald EJ. Analytical and experimental study of residual stresses in a multi-pass repair weld. Proceedings of ASME pressure vessel and piping conference. New York:ASME;2000;PVP-410-1:5364. [10] Dong P, Zhang J, Bouchard PJ. Effects of repair weld length on residual stress distribution. Trans ASME J Press Vessel Technol 2002; 124(1):7480. [11] Bouchard PJ, George D, Santisteban JR, Bruno G, Dutta M, Edwards LE, Kingston E, Smith DJ. Measurement of the residual stresses in a stainless steel pipe girth weld containing long and short repairs. Int J Press Vessel Piping 2004;11 special issue on residual stresses at repair welds. [12] Dong P. Modeling of weld residual stresses and distortions: advanced computational procedures and practical applications. Proceedings of the sixth international conference on residual stresses (ICRS-6) London: IOM Communications Ltd; 2000, pp. 122335. [13] Withers PJ, Bhadeishia HKDH. Residual stress part 1-measurement techniques. Mater Sci Technol 2001;17:35565. [14] Dong P, Hong JK, Rogers P. Analysis of residual stresses in AlLi repair welds and mitigation techniques. Welding J 1998;77(11): 439s445s. [15] Zhang J, Dong P, Brust FW, A 3-D composite shell element model for residual stress analysis of multi-pass welds. Transactions of the 14th international conference on structural mechanics in reactor technology (SMiRT 14), 1997;1:33544.

5. Concluding remarks Residual stress results from several weld repair case studies have been presented and the important general features of the repair residual stress elds identied. For selected cases, predicted stresses have been compared with detailed residual stress measurements and the adequacy of the nite element simulation procedures assessed. The following general conclusions can be drawn. 1. Finite length weld repairs increase the magnitude and importance of transverse residual stresses along the repair compared with the initial weld. Beyond the ends of the repair the transverse stress sharply falls into compression. The shorter the repair length the greater the increase in the transverse stresses, but for very long repairs the transverse residual stress within the central region of the repair length approaches that of the initial weld. 2. Welding procedure related parameters appear to be more important in analysing repair welds than in initial fabrication welds. These include bead and pass lumping, heat content of lumped-beads, inter-pass temperature, as well as pass sequencing. This is not surprising since the high restraint conditions typical to repair welds can be altered by some of these parameters, resulting in signicant changes to the through-thickness distribution of residual stress. 3. Simplied 2D cross-section nite element models with applied restraint conditions can be used to capture the general stress eld at a specic point along the length of a repair, but great care must be taken with assigning the boundary conditions.

P. Dong et al. / International Journal of Pressure Vessels and Piping 82 (2005) 258269 [16] Dong P. Residual stress analyses of a multi-pass girth weld: 3D special shell versus axi-symmetric models. Trans ASME J Press Vessel Technol 2001;123(2):20713. [17] George D, Smith DJ, Bouchard PJ. Evaluation of through wall residual stresses in stainless steel repair welds. Proceedings on fth European conference on residual stresses (ECRS5) Delft-Noordwijkerhout, The Netherlands; 1999.

269

[18] Zhang J, Dong P, Cao Z. Repair length effects on residual stress distributions in a pipe girth weld. Battelle project to Nuclear Electric Ltd, Report No. N0059182-01, September; 1997. [19] Dong P. Modeling and analysis of alloy 2195 welds in super lightweight tank. Battelle project report to Marshall Space Flight Center, Huntsville, Alabama, July; 1995.

S-ar putea să vă placă și