Christopher M. Gill

Finite Element Modelling and Measurements of Residual Stress and Phase Composition in Ferritic Welds

This paper describes finite element (FE) modelling and neutron diffraction (ND) measurements to investigate the development of residual stresses in two different geometries of ferritic weld. All specimens were produced using SA508 Grade 3 steel plates, depositing a low carbon SD3 weld filler by mechanised TIG welding. The FE analyses were carried out using Abaqus/VFT and the behaviour of the SA508 steel was modelled using a simplified (Leblond) phase transformation model with isotropic hardening using VFT’s UMAT-WELD subroutine, which includes the change in volume due to phase transformation. Single bead-on-plate specimens were manufactured using a range of mechanised TIG welding parameters. One pass and three pass groove welds were also produced, in order to investigate the cyclic hardening behaviour of the materials, as well as phase transformation effects in a multi-pass weld. FE analyses were then performed to determine how accurately these effects could be modelled. During manufacture, a number of thermocouples were attached to each of the specimens in order to calibrate the heat input to the FE models. The residual stresses in each of the bead on plate welds, as well as the groove weld after the first and the third passes, were then measured using ND at the middle of the plate. The ND measurements for the three pass weld showed no significant cyclic hardening behaviour although some was predicted by the FE analysis. Another key finding of the FE modelling that was seen in all of the models was that the phase transformation acts to reduce the stress levels in the deposited weld metal leaving the largest tensile stresses in a ring at the outer edge of the full heat affected zone (HAZ). There are plans to refine the FE studies using improved material properties when material testing of SA508 and SD3 are completed in the near future.Copyright

Development of Residual Stress Profiles for Defect Tolerance Assessments of Thick Section Electron Beam Welds

This paper describes the results of weld model analysis and deep hole-drilling measurements undertaken to evaluate residual stress distributions in austenitic and ferritic steel thick section electron beam welds. The work was undertaken in support of a Rolls-Royce and TWI development programme in the UK, for a Reduced Pressure Electron Beam (RPEB, 0.1 to 1mbar) welding process using a mobile local vacuum seal for the manufacture of thick section pressure vessel and pipe welds for nuclear power plant applications. Measurements were undertaken on representative mock-ups including a 160mm thick SA508-3 forging circumferential seam weld, in both the as-welded and furnace post weld heat treated condition. A 316L stainless steel plate butt weld and a 304L pipe girth weld of 80mm and 36mm thickness respectively were also analysed.There is now considered to be sufficient understanding of the residual stress fields generated by thick Electron Beam (EB) welds, to propose through thickness ‘upper bound’ R6 Level 2 stress profiles for use in defect tolerance assessments. The intention is to incorporate residual stress profiles of this type into the R6 structural integrity assessment procedure, following peer review. This would represent a significant step forward in demonstrating technology readiness for plant applications. It is also anticipated that an ASME Code Case will be drafted and proposed for the RPEB welding process.EB welding is a relatively low heat input process, compared with a multi-pass arc weld, such that the fusion zone and heat affected zone are narrow. The centre of an EB weld is the last region to solidify and cool-down, so typically there is a high degree of restraint to weld metal contraction, thereby generating a highly tri-axial yield magnitude tensile stress state at the mid-thickness location. The stress components acting in the longitudinal welding direction and through-thickness orientation tend to be large in the centre of EB welds of high aspect ratio (depth / width). By contrast, lower stress levels are produced on the surfaces acting transverse to the weld plane compared to conventional multi-pass metal arc welds. The transverse stress component is most likely to be required for the assessment of any postulated EB welding defects. The residual stress field decays rapidly with distance from the EB joint into the adjacent parent metal. Symmetrical stress distributions are typically generated in a 1-pass EB plate weld and stress fields are characteristically sinusoidal of wavelength between 1 and 4 times the section thickness.Copyright

Design Optimisation of a Ferritic Ring Weld Specimen Using FE Modelling

This paper describes the design optimisation of an SA508 ferritic steel ring weld specimen using FE modelling techniques. The aim was to experimentally and analytically study the effect of post weld heat treatment upon a triaxial residual stress field. Welding highly constrained geometries, such as those found in some pressure vessel joints, can lead to the formation of highly triaxial stress fields. It is thought that application of post weld heat treatments will not fully relax hydrostatic stress fields. Therefore a ferritic multi-pass ring weld specimen was designed and optimised, using 2D finite element modelling, to generate a high magnitude triaxial stress field. The specimen thickness and weld-prep geometry was optimised to produce a large hydrostatic stress field and still allow efficient use of neutron diffraction to measure the residual stress. This paper reports the development of the test specimen geometry and compares the results of welding FE analysis and neutron diffraction measurements. Welding residual stresses were experimentally determined using neutron diffraction; both before post weld heat treatment. Three dimensional moving heat source weld finite element modelling has been used to predict the residual stresses generated by the welding process used. Finite element modelling examined the effect of phase transformation upon the residual stress field produced by welding. The relaxation of welding stresses by creep during post weld heat treatment has also been modelled. Comparisons between the modelled and measured as-welded residual stress profiles are presented. This work allows discussion of the effect of post weld heat treatment of triaxial stress fields and determines if finite element modelling is capable of correctly predicting the stress relaxation.Copyright

Phase Transformation Properties Sensitivity Study in a Ferritic Groove Weld

A previous paper (PVP2010-25649) presented work carried out to model a three pass groove weld in an SA508 plate using Abaqus with the VFT material model. The model used the SA508 physical and mechanical properties for both parent and weld metal, with phase transformation properties for a slightly different material as these were the only properties available at the time. A material properties testing programme has since been completed allowing the analysis to be rerun using a complete set of properties for both the parent metal (SA508) and the weld metal (SD3). Properties to describe the phase transformations on cooling from austenite were derived for a range of austenite grain sizes. This paper presents a sensitivity study comparing the predicted stresses and phase proportions when using the different properties, and the effects of using different properties for the two materials. With the updated material properties and by using separate properties for the parent and weld, the results have been improved significantly. The results show that, while changing the phase properties affected the predicted phase proportions and the stresses in the weld metal, the residual stress distribution in the parent metal, where the peak tensile stresses occur, did not change significantly.Copyright

Design and manufacture of welded plate specimens for residual stress experiments

A long-term UK research programme has been established in order to improve the understanding of thermo-mechanical behaviour and residual stresses generated in pressure vessel steel welds as well as developing finite element (FE) welding simulation methods. The production of representative test specimens is an important element of this research project, since quality measurement data are needed to validate FE models. This paper describes the design, development and manufacture of welded plate specimens used for residual stress (RS) experiments. To date, research has focused mainly on developing the understanding of SA508 pressure vessel steel welds. Specimen dimensions were selected to facilitate stress measurements using a range of techniques. The philosophy adopted was to start with relatively simple 1-pass weld specimens and gradually increase the complexity to multi-pass groove welds in plates. Simple 1-pass weld specimens were generally designed to investigate the effect of welding parameters on thermo-mechanical behaviour, such as heat-affected zone (HAZ) microstructures and phase transformations. Later specimens are more representative of multi-pass power plant welds. They are being used to study material thermal cyclic hardening/softening behaviour. Other issues of concern are also being investigated, such as the effect of restraint during welding on RS and the effectiveness of post weld heat treatment (PWHT). Specimens were also designed to study peak stresses arising at bead stop/start positions and whether such peak stresses are annealed in overlaying additional weld metal. These investigations were performed on multi-pass groove welds in both austenitic and ferritic steel plates. Practical issues encountered during welding trials are discussed, including plate distortion and magnetisation of the ferritic steel plates. Information is also provided about welding temperature measurements and metallurgical examinations.Copyright

On the Thermo-Mechanical Behaviour of SA508 Grade 4 Ferritic Steel

The development of residual stresses in the Heat Affected Zone (HAZ) during welding of a ferritic steel can be critical to weld structural integrity. The Prior Austenite Grain Size (PAGS), the thermo-mechanical properties of the phases that develop during phase transformation, and the transformation strains are some of the key parameters that can alter residual stress development during welding. Understanding the trend in variation of these parameters is crucial for Finite Element (FE) modelling of residual stress development in weld. In this study, the effect of PAGS on the phase transformation in SA508 grade 4 was determined. For this purpose, samples were heated up to 900, 1050, 1250, and 1350°C and held for various time intervals to produce different austenite grain sizes. The measured austenite grain sizes were then used to fit parameters in an exponential equation implemented in an FE User MATerial subroutine (UMAT) for the modelling of welds. With performing various free dilatometry experiments, it is shown that the only phase that austenite transforms to upon cooling is martensite. In addition, the mechanical properties of as-received material, austenite, and martensite as a function of temperature were measured. Also, various uni-axial loads were applied during cooling cycles, and before the onset of phase transformations, to measure the evolution of transformation strain to generate an empirical formulation for numerical modelling.Copyright

Residual Stress Modelling in a Thick-Sectioned Electron Beam Weld

Previous work presented residual stress measurements in an electron beam weld in a thick section ferritic forging [1]; this weld was also modelled using finite element analysis. Due to the tool used to model the heat source, the mesh density in the region of the weld was limited. This work improves on the previous work by using a DFLUX subroutine to provide a mesh-independent heat source input, allowing a better mesh in the region of the weld. The modelling was carried out in Abaqus[2] using the VFT[3] user material model to allow phase transformation effects to be included. This however does not include creep properties and so the as-welded stresses were seeded on to a model that used Abaqus built-in material properties in order to model the heat treatment. The results of this analysis have been compared with analyses run using just the VFT material model (with no creep) and using just the Abaqus properties (with no phase transformation) in order to investigate the sensitivity of the stresses predicted to the material model used. The results of all three analyses have also been compared to the results of the original analysis and with the deep hole drilling residual stress measurements.Copyright

Development of Material Model Parameters Suitable for the Finite Element Simulation of Ferritic Welds

The prediction of residual stresses in ferritic welds using finite element techniques requires materials properties to describe the thermal, tensile, cyclic and phase transformation behaviour that the material undergoes during welding and also during creep as the effect of post weld heat treatment is also of interest. Ferritic steels will transform at a temperature above about 850°C to austenite. As the steel is cooled, a further phase transformation in the structure occurs. The precise structure formed depends on the detailed chemical make-up of the steel and on the rate at which it is cooled. On slow cooling from above 850°C a pearlite-ferrite microstructure is formed. On more rapid cooling, other microstructures, particularly bainite at intermediate cooling rates and martensite at the highest cooling rates are formed. Predicting the phase on cooling requires a Continuous Cooling Transformation diagram that is suitable for welding thermal cycles and reflects the time spent above the austenitisation threshold which influences the austenitic grain size formed and subsequently the phase of material on cooling. Material properties for a SA508 Grade 3 steel and a low carbon SD3 filler metal have been generated and fitted to constitutive models that are available in the finite element codes ABAQUS and SYSWELD. The choice of hardening model and its associated parameters have been evaluated on the basis of the observed cyclic behaviour in materials testing. Validation of these models has then been carried out by finite element simulations of welded mock-ups which have been measured using neutron diffraction. These include an autogenous weld beam and groove weld specimens containing up to eight weld passes. The rationale for using these simple specimens has been to:• Validate the capability of the model to predict the correct phase transformation behaviour and resulting stresses.• Account for the different behaviour of the parent and filler material.• To develop the capability for representing the material cyclic behaviour.On the basis of these simulations recommendations have been made on the material models (and their parameters) that may be used for the finite element simulation of the welding process.Copyright

Finite Element Modelling of Reduced Pressure Electron Beam Stainless Steel Plate and Pipe Welds

Rolls-Royce plc is conducting work to investigate the feasibility of using Reduced Pressure Electron Beam Welding (RPEB) for thick section welded joints in power plant construction. As part of the work, simple specimens have been manufactured at TWI ltd in order to develop welding parameters and conditions and to examine the achievable weld quality. Previous work in this project has shown good correlations between measured and predicted stresses in RPEB welds in ferritic components [5,6].This paper describes Finite Element (FE) modelling that was carried out to try to predict the residual stress field generated by the welding process in three of the specimens. The first specimen that was modelled was a full penetration butt weld in 80 mm thick Type 316L plate (W17). The other two models were of circumferential butt welds in 14 inch nominal diameter Type 304L pipe. The first pipe model (W20) was a single pass, 360° weld, while the second (W22) featured a slope-up and slope-down each lasting for 16° either side of a 360° full penetration weld, giving a total weld of 392°. The modelling was carried out in Abaqus [1] using a DFLUX user subroutine to model the welding heat input as a cylindrical heat source, due to the reduced pressure during specimen manufacture, only radiation heat losses were considered. The built-in Chaboche mixed hardening model was used for both materials during the structural analysis.The residual stresses predicted by the FE modelling have been compared with the results of Deep Hole Drilling (DHD) that was carried out on the equivalent specimens. Full details of the measurements are reported in [4].Copyright

Finite Element Modelling of a Tube to Vessel Attachment Weld and Local Post-Weld Heat Treatment

This paper describes Finite Element (FE) modelling of a weld between a tube and a machined feature on a curved pressure vessel surface. The components were manufactured from a ferritic steel with a matched weld metal deposited by a mechanised TIG process. The weld region then underwent a local Post-Weld Heat Treatment (PWHT) which used heating bands and cooling air flows to control the temperature distribution. The PWHT’s aim was to provide stress relief and HAZ tempering, while minimising the stresses due to thermal gradients in the component. Trial welds on representative test pieces had predicted significant welding-induced distortions. Therefore, during the weld and PWHT, restraints were applied to the tube to prevent excessive deformation. The material behaviour was represented using Abaqus’ built-in material options, with the same properties for both the base metal and the filler. Isotropic hardening was assumed and the stress relaxation during the PWHT was modelled by applying a Norton creep law only during the hold time. Phase transformation effects in the ferritic material were not included. Initial modelling used a 2D axisymmetric model to allow sensitivity studies to inform the development of the PWHT process. These showed that the degree of stress relief was much more sensitive to the soak temperature than the hold time. Subsequent runs analysed a 3D model using a segmented block-dumping technique, with the deposition modelled by introducing the weld elements in 90° segments. The 3D modelling was undertaken in order to more accurately model potentially asymmetric welding distortions and residual stresses. The torch was represented by a body flux into each segment after its introduction. This model was also run without restraint to provide validation by comparing the predicted distortion with measurements from the welding trials; a good match was demonstrated. Further comparisons were made between the predicted stresses and results of Incremental Centre Hole-Drilling (ICHD) stress measurements made on the trial specimens both in the as-welded condition and after PWHT. The measured stresses were close to those predicted by the FE analysis and the key features of the predicted stress field were apparent in the measurement data. Due to the location of the tube’s attachment to the pressure vessel, thermal expansion of the vessel during the PWHT caused the tube to bend. The induced bending stresses were then relaxed during the soak and re-introduced in the opposite sense as the system cooled. This effect was captured by running the analysis as a submodel of a global FE model with displacements read across at nodes in the pressure vessel shell immediately below the weld.Copyright