Residual Stresses in Thick Bi-metallic Fusion Welds

Abstract

Welding is applied in many industrial sectors to join components, and has become an important manufacturing process because it enables the fabrication of structures that could not otherwise be constructed. Weld regions have inhomogeneous microstructures and are more susceptible to crack initiation and crack propagation than the surrounding base material regions. Residual stresses are also formed, which superimpose with applied loads, resulting in a reduction of the maximum applied load a component can sustain. In particular for nuclear installations, the limited failure tolerance and the relative abundance of rather large welds require a reliable assessment of component integrity for lifetime management. Residual stresses need to be considered in such assessments because they can contribute to initiation and propagation of defects. Commonly, residual stresses are more difficult to determine experimentally or to predict by numerical methods than stresses resulting from applied loads; hence residual stress assessment in welded nuclear components is an important area of research. The present work concerns the experimental determination of residual stresses by neutron diffraction in three full-scale mock-ups of components found in nuclear power installations. Two of these mock-ups represent dissimilar metal girth welds joining ferritic steel pressure vessel nozzles to austenitic stainless steel primary piping sections. The third represents a welded clad layer on a section of a reactor pressure vessel wall. In this work neutron diffraction has been used as the technique for residual stress determination. This technique is based on the principle of Bragg diffraction and measures changes in lattice spacing; i.e., strain. Residual stresses can be determined in three directions in the bulk of a component at a spatial resolution of typically 1-5 mm. Such a resolution is appropriate in view of the distances over which the residual stresses normally vary in welded metallic components. There are a number of challenges associated with neutron diffraction measurements addressed in the current study. One of these is that component dimensions and geometries necessitate machining to facilitate access of the neutron beams to the measurement locations. Neutron diffraction measurements are also known to be sensitive to the inhomogeneity in the microstructure and to local variations in chemical composition in the weld region; therefore dedicated reference specimens are needed in order to calibrate the strain determinations. Prior to the description of the experimental work undertaken, background information is provided on the main aspects of the work. Definitions are given for residual stresses and strains and their main characteristics are explained (chapter 2). A short overview of the most important techniques for strain and stress measurement being used today is given in chapter 3. These are the relaxation techniques, where the relaxation of strains due to material removal is measured; and the diffraction techniques, where lattice deformations caused by stresses are observed through Bragg diffraction. A detailed account of residual stress determination based on neutron diffraction is followed by a description of the facility at the Joint Research Centre that was used for the present investigations. An overview of the industrially relevant welding techniques is given in chapter 4, where additional detail is provided on the arc welding methods used for the manufacture of the components investigated in this study. The mechanism responsible for residual stress formation in welds, namely local plastic deformation caused by steep temperature gradients, and the dendritic microstructure of stainless steel fusion welds similar to those investigated here are discussed based on a few examples. The experimental work and the results obtained are described in chapters 5, 6 and 7 with each chapter covering one of the three components. The first component is a “thin" walled (25 mm wall thickness) bi-metallic girth welded pipe. The second component is a ferritic steel block with a 10 mm thick welded stainless steel clad layer applied to one of its surfaces. This component represents a nuclear reactor pressure vessel wall with a thickness of 146 mm. The third component is a thick walled (51 mm wall thickness) bi-metallic girth welded pipe. Both, the thin and the thick walled bi-metallic pipes, represent joints between ferritic steel pressure vessel nozzles and austenitic stainless steel pipes. For each component, details are provided concerning the manufacturing techniques employed, including the welding processes, the associated heat treatments and the final machining applied to the thick walled pipe. For the access of the neutron beams in three measurement orientations windows and access slots were cut into each of the bi-metallic piping weld specimens. The clad layer specimen thickness was locally reduced from 146 mm to 25 mm. This significant reduction was necessary to achieve a high spatial resolution in the measurements. The alterations described lead to stress relaxation and stress redistribution within the components. For the thick bi-metallic pipe the relaxation of strains was monitored by strain gauges during cutting showing negligible impact at the location of the neutron diffraction measurements. All component modifications are described in detail as well as the design and manufacture of the reference specimens needed for the calibration of the strain measurements. The measurement procedures and the data analyses are explained, and for each specimen the neutron diffraction results are presented in terms of residual strains and the derived stresses. In the thin walled bi-metallic pipe, tensile residual stresses have been found in the welding direction within the fusion zone. These tensile stresses reach values not far below the yield level of the material near the outer surface; and they decrease to almost 0 MPa toward the inner surface of the pipe. In the ferritic part of the pipe compressive residual stresses have been found near the austenitic-ferritic material interface. Here, the highest compression is observed close to the inner surface, decreasing toward the outer surface. In the welding transverse direction, tensile stresses have been obtained near the outer surface changing gradually to compressive stresses near the inner surface. The maximum stress levels attained in tension and compression are slightly lower than those in the welding direction. The piping radial direction exhibits overall the lowest stress levels in this component with values varying between +100 MPa and -100 MPa. The stainless steel clad layer of the clad component exhibits high tensile stresses in the welding longitudinal direction as well as in the welding transverse direction. The stress level ranges between 250 and 500 MPa with a considerable scatter of the data. The ferritic steel substrate is found to be in compression in both directions with minimum stress levels between -150 and -200 MPa near the austenitic-ferritic material interface. The stresses in the interface normal direction observed in this component scatter about 0 MPa. In the fusion zone of the thick walled bi-metallic piping component, tensile residual stresses have been found in the welding direction. In the ferritic part compressive residual stresses have been found in this direction near the austenitic-ferritic interface. Both of these observations are similar to those made for the thin walled component. The maximum stress levels observed are slightly higher than in the thinner component; the tensile stresses in the fusion zone reach the nominal yield level. No clear trend is observed for the welding longitudinal stresses between the outer and the inner surface. For the other two measurement directions the scatter of the data is so high that the magnitude of the stresses cannot be determined with sufficient accuracy. For all three components the data recorded from the fusion zones exhibit higher scatter than those from the base materials. The neutron diffraction results have been compared to third party numerical predictions of the residual stresses and to stress measurements by strain relaxation techniques. The numerical predictions have been performed by finite element analyses. Simplified and more detailed approaches have been applied. In the simplified approaches the welding process itself has been neglected unlike in the detailed models. In all cases the detailed numerical assessments produced a better agreement with the neutron diffraction results than the simplified ones. In particular, in the fusion zones the simplified numerical approaches under predict the residual stresses found by neutron diffraction. The applied third party strain relaxation techniques were the ring core method and deep hole drilling for the clad layer component, and surface hole drilling and the crack compliance method for the thick walled bi-metallic pipe. Most comparisons with the neutron diffraction data show a qualitative, but not a close quantitative agreement. The discrepancies are attributed to the use of differently extracted test pieces from the original components and to the differences in the measurement geometries that apply to the different methods used. Subsequent to the presentation of the measurements and their results, an assessment of the experimental methods and the analysis of the results are performed. In particular attention is given to the method of obtaining the measurement uncertainties. In stress determination by neutron diffraction it is common practice to calculate the uncertainties solely from the fitting uncertainties of the neutron data; a practice that has been applied in this work as well. The detailed assessment of the results indicates that this approach can provide appropriate uncertainty values only for materials that are sufficiently homogenous. This is the case, for example, for the ferritic steel substrate of the clad layer component. For the less homogeneous regions, like the welds in these components, it is observed that the experimental data exhibit larger scatter than one would expect on the basis of the counting statistics. The analysis of the effect suggested that the uncertainty is underestimated by a factor of up to 10 in the worst case presented. The second uncertainty contributor analysed in more detail is the uncertainty in the detector position. . The analysis shows that, for the highest strains measured, an uncertainty in detector position as small as 1° or 2° results in an additional strain uncertainty comparable to that stemming from the fitting uncertainties. Other sources of uncertainty, such as possible errors in specimen positioning or variations of the specimen temperature, are also briefly analysed. It is found that these do not produce significant additional uncertainty contributions in these investigations. Based on the observations and subsequent analyses of the findings and comparisons several conclusions are derived. The most significant conclusions can be summarized as follows: • The applied cutting schemes for the specimens and the use of the dedicated reference specimens have made the neutron diffraction measurements possible. The necessary alterations to the specimens have an impact on the stresses under investigation. It is demonstrated that the impact should be quantified through experiments or modelling, in order to relate the stresses measured to the original stresses present in the test piece. • It is shown that for a material like the welds studied in this work, the impact of the material inhomogeneity on the neutron diffraction measurements must be considered in a complete uncertainty analysis. It is found that this uncertainty contribution can be larger than the contribution from the fitting uncertainty of the neutron data. • The neutron diffraction stress measurements can be used for the validation of numerical stress prediction methods. It is demonstrated that the simplified numerical approaches for these specimens are not sufficient. Following from the above, a number of recommendations are formulated for further improvements in the application of neutron diffraction for future residual stress measurements in large welded components. In similar cases it is recommended to apply experimental techniques capable of mitigating the problem of the high scatter in the results obtained for the fusion zones. The time-of-flight technique is presented as one option, or where possible, rocking of the specimen during measurements could be applied to increase the number of grains sampled during a diffraction measurement. Furthermore, the importance of quantifying the stress relaxation due to modifications of the specimens is pointed out. This quantification could be achieved by numerical simulation, but preferably by measurement, in particular when the recommendations just mentioned are followed.