Dissimilar metal welds - how good are they?
TWI Bulletin, November - December 2004
Even with large defects, the tearing resistances are such that they would require loads well in excess of design to propagate the defect.
John Wintle is a consulting engineer for structural integrity and is a leader in the development of reliability engineering at TWI. He takes a keen interest in failure investigations, particularly of pressure equipment, and is a strong advocate of TWI's multi-disciplinary approach.
John Wintle explains how the integrity of a defective dissimilar metal weld between ferritic and austenitic pipe has been tested and assessed. The work took place within the major European collaborative project ADIMEW involving eight organisations from five countries. Although the weld in question was related to nuclear plant, the results have relevance for dissimilar metal welds of this kind, which are found in many other applications.
Dissimilar metal welds are joints between materials of different microstructures, as in the case of different types of steel (eg ferritic and austenitic) or different metals altogether (eg aluminium and steel). They are a challenge to the welding engineer because the optimum conditions for welding to one material may not be optimum for the other. Consequently, dissimilar metal welds frequently have a complex fabrication history, with the use of butterings and multiple weld metals to ensure compatibility.
ADIMEW
The dissimilar metal weld studied in the ADIMEW project joined an A508 ferritic pipe to a 316L austenitic pipe and was representative of a joint within the primary circuit of a French N4 pressurised water reactor nuclear power plant. The pipes had a diameter of 500mm and a thickness of 50mm. The weldment comprised a 308L austenitic weld metal manual metal arc (MMA) buttering the A508 weld preparation and then a second MMA buttering layer using 309L weld metal. The joint between the buttered A508 to the 316L pipe was made using more 308L weld metal deposited using an automated MIG process. Heat treatment and NDT completed the welding procedure. The whole assembly is shown in Fig.1.
It is not surprising that there are different microstructures and mixes of materials in welds of this kind and a variation of properties can be expected, see Fig.2. In this particular weld, the region of interest was the first layer of buttering close to the fusion line with the ferritic parent pipe, having a potentially low tearing resistance in the event of cracking from the outside surface.
Fig.1a) ADIMEW test pipe wall section showing dissimilar metal weld
Fig.1b) Transverse section across test pipe showing defect
Fig.2. Macro of fusion line with the ferritic pipe showing the buttering layers and the coarse grain heat affected zone
Consequently, the objective of the project was to test the behaviour of an external surface breaking defect in this weld region when subjected to high load, and to assess the ability of methods for predicting the behaviour should such a defect arise in practice.
Electricité de France (EdF) led the project and designed and undertook the testing. Framatome (now part of the Areva Group) fabricated the test weld and the Joint Research Centre at Petten inserted and measured the defect before and after the testing. The Finnish Research Centre VTT co-ordinated the materials testing, and other partners included CEA, Serco Assurance, and the Bay Zoltan Institute Hungary. TWI's role was to co-ordinate the predictive analysis, and to undertake fracture testing, finite element analysis and residual stress measurements and was undertaken with the support of the Health and Safety Executive.
Procurement
The project began in November 2001. The pipes were procured and the test weld fabricated, heat treated and inspected at Framatome's welding centre at Chalon-sur-Saône. At the same time another dissimilar metal weld between sections of these pipes was made to exactly the same procedure. This second weld was intended to be sectioned for the materials testing programme and residual stress measurements.
The test assembly was then sent to the JRC where the defect was inserted. This was an electro discharge machined slot, parallel to the fusion face and 2mm from it in the buttering, with a flat front extending a maximum of 17mm in depth from the outside surface, see Fig.1b. Preliminary analysis had indicated that this defect would be sufficient for the objectives of the test. These were to achieve limited growth by stable tearing, but avoiding unstable propagation around the circumference.
Materials testing
A comprehensive programme of materials testing was undertaken on the second weld. The aim was to determine the tensile and fracture properties of the different materials within the weldment - the A308 and 316 parent pipes, the heat affected zone of the ferric pipe, the weld buttering, and the bulk weld metal. The yield strength of the ferritic pipe was of course much higher than the austenitic pipe and weld metals, and it was of interest to see how this would affect the testing and the fracture behaviour of the defect, where the plastic zone would be constrained. The tensile testing was challenging in itself, since it required miniature specimens and the determination of true stress/true strain properties for use in elastic plastic finite element analysis.
The greatest interest was in the fracture toughness properties of the buttering material that were being obtained by TWI. A series of tests on different sized specimens indicated that the buttering material had the lowest toughness and the greatest susceptibility to fracture of all the materials in the weldment. In terms of crack growth resistance, J R , the buttering had a toughness at initiation of tearing of around 55 MN/m and a flattish R curve, Fig.3. Would the defect in weld metal buttering behave like a hot knife through butter in the test?
Fig.3. Fracture toughness R curve tearing resistance of buttering material
Residual stresses
In welds between austenitic and ferritic materials there are likely to be residual stresses. These arise because of the different coefficients of expansion (the austentic is 50% greater than the ferritic material), the differential rates of heating and cooling as the weld is fabricated, and volumetric changes as the molten materials solidify and cool. Residual stresses can be important in initiating certain types of defect, and they can play a role in promoting fracture, particularly if the material is of low toughness.
It was therefore relevant to measure and calculate the residual stresses in this complex dissimilar metal weld. The work was shared between Framatome, the JRC and TWI. The hole drilling technique was used to measure the stresses at the inside and outer surfaces, and neutron radiography used to determine the through thickness stress distribution. The Swiss Martech laboratory later confirmed these measurements using a block sectioning method. Calculations were made using finite element analysis of the weld fabrication, firstly making simplifying assumptions about the cooldown temperature, and then by sophisticated modelling of the weld fabrication step-by-step throughout its history.
As expected, the residual stress distribution across and through the weld was complex. Surface residual stresses were predominately compressive, probably as a result of the surface machining that had taken place. Stresses below the surface and through the bulk of the weld were predominately tensile, tending to drive the defect through. Whether this would have a noticeable effect in material where extensive yielding was expected to take place was uncertain until the test took place.
Test configuration
The test configuration was to load the pipe assembly in four point bending as shown the Fig.4. Two fixed outer supports counteracted forces applied by two large displacement controlled hydraulic rams located between the outer supports. The bending moment generated across the section produced a predominately tensile axial stress on the defect. Since the defect was inclined to the pipe axis, the stress field acting on it was of mixed tensile/shear mode.
Fig.4. Test configuration
The test was to be undertaken at 300°C, the operating temperature of the reactor primary circuit. Instrumentation was applied to measure the pipe deflection, applied forces, strains, crack opening displacement. Initiation and growth of the defect was measured using electrical potential drop.
Analysis
The key questions for the analysis were to predict the load at which the defect would start to grow (initiation), whether the growth would be stable or unstable, and the direction that growth would take. A variety of analysis methods were used to make predictions. These included engineering approaches based around standards such as BS 7910 and R6, and detailed finite element analyses modelling the defect. The inputs to the analyses were the small-scale materials test data (tensile and R curve), the geometry and dimensions of the pipe and the defect, and the ram forces, applied as displacements.
The analyses predicted that the defect would initiate at a moment of between 1.4 MNm and 1.8 MNm, the lower figure being obtained by the conservative engineering approaches and the higher value from the more detailed finite element analysis. The defect was expected to grow by stable tearing as the system was displacement controlled, and extensive growth could be expected at virtually constant load once a threshold moment of around 2MNm was reached. Growth was predicted to be predominately through the thickness of the buttering with very little growth circumferentially.
Results
The test took place at EdF's testing centre near Fontainebleau in July 2003. Altogether it took about 12 hours as the load was slowly increased. It was apparent that when the moment reached 2MNm a plateau had been arrived at and the defect was growing significantly at constant load. However, it was not until the test was finished and the results analysed and the defect was opened, Fig.5, that the full story became clear. The defect had grown substantially, a total of 28mm through the thickness towards the inside surface of the pipe. However, no growth had occurred circumferentially and the defect length was unchanged. The defect had grown straight and parallel to the fusion line with the ferritic pipe, but did not cross the line. This was probably due to the plastic zone around the defect front being constrained within the lower strength buttering by the higher yield strength ferritic material, Fig.6. The moment at which growth initiated was estimated from electric potential drop and changes in compliance to be about 1.85MNm, compared with predictions in the range 1.4 to 1.8MNm. It must be said that this moment far exceeds that which would ever be present in a pressure system designed to modern codes. It compares with the moment required for plastic collapse of the pipe in bending of about 2.5MNm.
Fig.5. Section of the pipe containing the defect showing starter EDM notch, tearing during the test, cleavage of remaining ligament on opening
Fig.6. Stress contour around defect showing plastic zone constrained to buttering
Verdict
The ADIMEW project has demonstrated a number of important things about dissimilar metal welds between ferritic and austenitic steel pipes. Firstly, that there is the possibility of low tearing resistance in the buttering layer adjacent to the ferritic material in which defects may grow substantially if subjected to sufficiently high loads. However, even with a large defect, the tearing resistance is such that it would require loads well in excess of design to propagate the defect. The test showed that defects would grow through the wall and not lengthways. This is good news in a safety sense, as it means that defective pipes would tend to leak before they break.
The project also showed that we may have confidence in the analytical procedures such as BS 7910, that are used to assess defects in even these complex welds. The conservatism of engineering solutions and the greater accuracy of finite element analysis were clearly demonstrated. The greatest uncertainty is probably the scatter in the small scale fracture toughness data.
Inevitably, the project has raised new technical questions: why the buttering tearing resistance is low, and how this should be measured, the nature and role of residual stress, and the effect of mixed mode loading.
