You are on page 1of 5

Education & Training

www.nafems.org

FEA and Failur e Assessment of Bi-Metallic Welded Joints


A. Cornec, W. Schnfeld, W. Brocks Institute of Materials Research, GKSS Research Centre, Geesthacht, Germany
Realistic prediction of the failure of welded connections is an area of strong interest for engineers in many industrial sectors. This article presents some of the work done under the EU project BIMET, to explore the characteristics of Bimetallic joints.

Introduction
Welded structures such as bridges, offshore platforms, pipelines and nuclear components may represent potential risks to human lives, environment, and capital investment. The assessment of their integrity has to include the effects of crack-like defects, which are potentially present in welded joints, either generated by the welding process or evolving from surface imperfections during service. Such defects may extend under service loading and lead to final failure before the design life has been reached. Integrity analyses based on fracture mechanics can help reduce the risk as well as reduce maintenance costs by setting up appropriate inspection intervals. These analyses face the problem of the heterogeneity of material properties, which is due to the dissimilar microstructure. It is still controversial to what extent fracture toughness data obtained from homogeneous specimens may be applied to heterogeneous components. Guidance on how to treat such complex multi-material structures with crack-like defects is highly sought after. Minimising the efforts but nevertheless achieving realistic and reliable predictions is the general requirement for advanced engineering assessment methods (EAM), which are increasingly based on finite element analyses. Their reliability has been validated by a benchmark test on a welded joint of two dissimilar tube segments under bending load.
Table 1: Overview of topics and participants of the BIMET Project, financed by the European Commission, DG XII, Euratome Programme

1) Task Leader GKSS 2) Local Approach with Beremin Model 3) Residual stress simulation 4) Elastic-plastic, with weld gradients, CTOD/5 fracture approach 5) Elastic-plastic, J-integral fracture approach The BIMET project aimed at contributing to the development and validation of failure-assessment procedures with respect to crack-like defects in bi-metallic welds. The analysis (TG.5) included engineering flaw assessment methods (EAM) and finite element analyses. Extensive material characterisation, residual stress measurements and fractographic analyses complemented the two component tests. Only the FEA contribution of the GKSS Research Centre will be presented and discussed in the following.

EU Project BIMET
Welded joints of low-alloy and austenitic steels, occurring in PWR and BWR nuclear power plant pipings, represent a challenging task with respect to both the complexity of design and the potential consequences of unexpected failure. This problem has been investigated in a multinational 3-year benchmark project BIMET (Structural Integrity of Bi-Metallic Components), performed within the 4th Euratom Research Framework Programme Nuclear Fission Safety. Table 1 gives an overview of the project structure. Page 2

Test Set Up
The test configuration is shown in Figure 1. Two thickwalled pipe segments (length L = 200 mm, outer diameter 2R0 = 168 mm, wall thickness t = 25 mm) made of a ferritic steel SA 508 and an austenitic steel 304L, respectively, are welded together, representing the actual test component. The tubes are extended by two

July 2004

www.nafems.org

Education & Training


realistically predict the deformation and failure. Flat miniature specimens of 1 x 2 mm cross section were taken across the whole weldment to determine the local variation of Figure 3: BIMET weldment stress-strain curves, but tested up to the onset of necking only. Additionally, round tensile bars were tested to obtain true strain-strain curves even beyond necking. The specimens were oriented longitudinally to the pipe-axis and tangentially to the circumference of the pipe. All stress-strain curves are

Figure 1: BIMET test configuration

bending arms on either side, made of a conventional construction steel DIN St 52. The component is subject to four-point bending by specially designed loading equipment. The central part of the joint is shown in Figure 2. The weldment consists of various zones, a ferritic heat affected zone (HAZ-F), two buttering layers, and the weld metal, as represented in Figure 3, each of the zones showing a varying microstructure inside. A 0.25 mm wide slot was inserted by spark erosion as a defect in the interface of the two buttering layers parallel to the HAZ-F. Two component tests were performed at Elctricit de France (EDF) with two different crack depths, a, namely a/t = 1/3 and 1/2, respectively. Only the analysis of the first test with a/t = 1/3 will be presented here. The measured global displacement of the RAM, and the crack Figure 2: mouth opening BIMET weld sections with surface defect displacement, CMOD, can be used for the validation of the FEA. In addition, data from potential drop measurements are used for detecting crack initiation.

Figure 4: Stress-strain curves

Material Characterisation
Much attention was focused on a careful material characterisation across the various zones, which was performed by VTT Manufacturing Technology, providing stress-strain curves and fracture toughness data. As the experimental effort for heterogeneous problems like this is enormous, such extensive testing has to be minimised for practical use. But, any analysis which does not account for the variations in the material properties will not be able to

plotted in Figure 4, the whole strain range in the left diagram and the initial region up to 0.8%, for determining the 0.2% proof stresses, Rp0.2, in the right one. There is a significant variation over all zones across the weldment and bulk material segments, but relatively little variation across the HAZ-F and buttering zones. The property gradients of the weldment are discretised as 15 sections, denoted as S1 to S15 (see Figure 5) for which the local representative stress-strain curves are displayed in Figure 4. The determination of toughness data raises the question: which fracture parameter is supposed to appropriately characterise a multimaterial component? This is a physical problem rather than a problem of testing and analysis. For homogenous cases with limited stable crack extension, fracture assessment concepts are based on the J-integral. Due to path dependence the application of J is Page 3

July 2004

Education & Training


questionable for the present p r o b l e m , however, though its respective data was acquired within the project as well. Damage mechanics concepts are still quite far from p r a c t i c a l application and would have required nonstandardised methods for parameter identification. An alternative parameter, Figure 5: which represents local Meshing of the weldment properties much better than J, is the crack tip opening displacement, CTOD. A particular technique developed and validated at GKSS, namely 5, being the opening displacement measured over a 5 mm gauge length at the crack tip, has been adopted. It is easy to apply, and resistance curves can be obtained from representative bend specimens, taken directly from the welded component. Residual stresses in the weldment were experimentally determined by neutron scattering at JRC Petten. As they release under plastic deformation, they have not been considered in the FEA of the component test.

www.nafems.org

global coarse mesh. Figure 5 shows the mesh of the weldment region including the embedded box and the close-up range of the latter, separately. The numerical calculations were performed by ABAQUS/Standard. The loading was applied through the RAM displacement (Figure 1). The material behaviour was described by incremental plasticity and the von Mises yield criterion with isotropic hardening. A large strain formulation was used. A number of parametric studies varying the material properties of the different sections from a homogeneous pipe and perfectly plastic materials of various yield strengths to full consideration of the material gradients have been performed to clarify the respective effects on local and global deformations. The plastic yielding in the heterogeneous weldment is displayed in Figure 6 in terms of plastic strains at final loading, RAM = 40 mm, corresponding to CMOD = 1.21 mm. For comparison, Figure 6a presents the plastic strain distribution in a homogeneous pipe without a crack for the bulk material 304L, while Figure 6b shows the BIMET pipe with all material gradients. The weakest section is the austenitic segment 304L (Ls = 200 mm) with a yield strength of Rp0.2 = 322 MPa (Figure 4) and not the cracked section, due to its higher yield strength of Rp0.2 = 364 MPa. Yielding of the weldment starts from two sites, see Figure 6c, namely at the crack tip, growing through the pipe wall, and independently also in the austenitic segment spreading over its full length, see Figure 6b. As the ferritic segment SA 508 has a significantly higher yield strength, pure elastic deformation takes place. In a homogeneous pipe, the plastic strains spread out nonsymmetrically from the crack tip, due to the inclined crack, quite similar to the behaviour shown in Figure 6c, but remain in the centre of the weld metal. Plastic strains in the cracked section can be activated only if the hardening of the austenitic segment 304L is sufficiently high. In the limiting case of a perfectly plastic material, the stresses in the cracked segment would be limited, the crack would not extend, and failure would occur by plastic collapse of the austenitic segment.

FEA: Meshing, Results and Validation


A full 3D FE mesh of the BIMET component with 20-node hexahedral solid elements was generated using the programme IDEAS. The mesh consists of 4376 elements and 65903 nodes. At the crack region very small elements are required to record CTOD and capture concentrations of stresses and strains. A meshing technique developed at GKSS has been applied. The refined mesh is confined in a box along the crack front, which can be g e n e r a t e d independently from the global mesh. Figure 6: Plastic yielding: Homogeneous and BIMET pipe a) Homogeneous pipe, base material 304L, without defect This box is placed as b) BIMET pipe, with all gradients including defect a block into the c) Close-up of plastic strain distribution at the weld Page 4

Figure 7: Equivalent plastic strain distribution ahead of the defect

Figure 7 presents the concentration of plastic strains near the crack front. Plastic strains not higher than 50% occur in a small region ahead of the crack front. Note that the plastic strains do not govern the crack path. Actually, the

July 2004

www.nafems.org

Education & Training


value of CMODi,FEA = 0.41 mm compared to a test value of CMODi,test and a moment of Mi,FEA = 162 kNm compared to Mi,test = 154 kNm. This means an overestimation of the initiation load by 5%, which is well within the range of uncertainty for detecting crack initiation by the potential drop technique in the test. The effect of ductile crack extension, which was not considered in the FEA, was analytically estimated using the measured 5R curve. It reduces the bending moment, M, slightly from 187 kNm to 182 kNm at the end of the test, see curve #3 in Figure 9.

latter follows the hydrostatic stresses, h, as plotted in Figure 8. High hydrostatic stresses occur outside the region of plastic strain concentration towards the s t r o n g e r Figure 8: Hydrostatic stresses ahead of the defect and material zone superpostion of the crack path in the test HAZ-F, which obviously imposes a local constraint and a high stress triaxiality. A coaction of both, plastic deformations (characterised by CTOD) and hydrostatic stresses govern ductile tearing, as described by models of ductile damage. Conventional global fracture mechanics concepts would need two parameters, CTOD and stress triaxiality. The parametric studies have revealed that local deformations and stresses, and hence crack initiation and extension in a heterogeneous component, are strongly affected by the interaction of the stress-strain curves of the various constituents, besides the defect size and shape, the geometry of the component and respective loading type, of course. A careful material characterisation is therefore essential for a realistic failure prediction.

Figure 10: Influence of the weld gradients

Figure 9: Prediction of load-CMOD Curve with load at initiation and end of test

In order to estimate the influence of the gradients in the weld metal, two cases are compared in Figure 10. Case A is the reference one from the BIMET with all material gradients, and case B considers a homogeneous weld with the property weld-inside, see Figure 4, sections S7, S10, S13. As the strength of the weld-inside material is higher than that of the surrounding weld metal volume elements, the global behaviour of model B must be stiffer than the reference curve A. The reduction of local CTOD is 13% at crack initiation and about 20% at the end of the test, while the load increases by only 3% at the crack initiation level. The effects from the other weld-metal sections are even smaller. No significant error would hence be made by neglecting the material gradients over the pipe thickness. Most important are the different weld properties in the direction of the pipe axis. Figure 11 presents the crack extension along the crack front predicted from the FEA by the 5R curve in comparison to the actual crack front obtained in the test. Maximum crack extension is predicted in the centre where CTOD is highest. The FEA, though based on a stationary crack, provides a quite realistic prediction of ductile crack extension. The experimental crack profile exhibits two local maxima outside the centre, which may be due to variations of triaxiality along the crack front. As only a single 5R curve has been measured and applied, triaxiality Page 5

The bending moment, M, as calculated from the measured load, CMOD and RAM from the test data are used for the validation of the FEA. Figure 9 shows the MCMOD curves from FEA and experiment revealing a nearly perfect comparison. The point of crack initiation in the FEA was predicted from the numerical CTOD versus CMOD curve taking 5i = 0.2 mm, as initiation from the resistance curve, 5R = 5(a), measured on a single edge-cracked bend specimen. This yielded a predicted

July 2004

Education & Training


induced variations of tearing resistance are not accounted for in the present analysis. For larger crack extension, the coupling effect of the new crack front shape must be taken into account.

www.nafems.org

by local constraint due to mismatch and will generally not coincide with regions of high plastic strain. A CTOD based criterion has been successfully applied to predict crack initiation and extension. It is particularly useful for the fracture prediction of heterogeneous components. CTOD is easy to measure in a test and evaluate from the FEA. A CTOD criterion is better suited for welded joints than the Jintegral, which is a far field quantity suffering from the path dependence through material gradients.
Figure 11: Prediction of crack extension along the crack front

Conclusions and Recommendations


A comprehensive understanding of the deformation and fracture behaviour of a multimaterial system has been established by the FEA presented. The following conclusions can be drawn for related analyses:

In order to achieve realistic results a multimaterial component must be analysed completely threedimensionally. Besides the overall geometry and the defect dimensions, all boundary conditions have to be realistically modelled. Accurate stress-strain curves of the various materials are highly important for the quality of the model. In the vicinity of defects, they have to cover large plastic strains. The scatterband of material properties should be determined by a sufficient number of specimens. For heterogeneous joints, it is necessary to receive information about the local distribution of material properties. Small flat test pieces must be supplied for which test standards are not yet available. As the effect of material gradients can often not be estimated in advance, simplifying assumptions should not be introduced in advance without respective parametric studies. Full 3D finite element meshing with hexahedra may become a problem for cracked components, as very small elements around the defect have to be combined with a much coarser global mesh. The embedded.crack.box technique used in the present analysis is advantageous, as the box can be generated with small hexahedra independently from an existing global mesh. Free meshing with triangular type elements is not recommended for elastic-plastic analyses of crack problems. Plastic deformations spread from the weakest section, where deformation arises by the interaction of strength mismatch, geometry and loading. Plastic hinges can develop in low-strength regions outside the cracked section, and an existing crack may only initiate and extend, if weaker sections outside have enough hardening capacity to increase the stresses in the cracked section. Plastic strain contours give no indication about the crack path, as hydrostatic stress affects ductile damage evolution as well. High hydrostatic stresses can be induced

CTOD allows the prediction of ductile crack initiation and extension, but not its orientation in multimaterial systems. Transitions to other fracture modes, such as cleavage fracture or to shear mode must be considered separately. Although the results have been obtained for a specifically designed component, they provide a general insight into the problem and can be generalised for other multimaterial systems. Applications in practical engineering will require further benchmark validations.

Acknowledgement
The authors acknowledge the contributions of all participants in the project, which made this article possible. The financial support from the Euratom Fourth Framework Programme is greatly appreciated.

R eferences
[1] W. Brocks, A. Cornec, W. Schnfeld, Fracture mechanics assessment of a bi-metallic welded joint. Proceedings of the 14 Bienniel Conference on Fracture, 8-13 Sept. 2002, Cracow, Poland, Eds. A. Neimitz et al., EMA Publishing, Sheffield, UK, 2002. [2] A. Cornec, W. Schnfeld, W. Brocks, K.-H. Schwalbe, Finite element analysis including weldment gradients. EUProject BIMET, Task Group 5: Analysis, Report Technical Note GKSS/WMS/2000/10, GKSS Research Centre, Geesthacht, Germany. [3] K.-H. Schwalbe, A. Cornec, D. Lidbury, Fracture mechanics analysis of the BIMET welded pipe tests. Int. Journal of Pressure Vessels and Piping, 81 (2004), 251277.

Contact
Alfred Cornec E e-mail: cornec@gkss.de

Page 6

July 2004

You might also like