fatigue of welded st st

This Research Report is for the exclusive use ofIndustrial Members of The Welding Institute, andits content should not be communicated to other

individuals or organisations without writtenconsent. It is in the interest of all members to

respect this confidence.

August 2006 857/2006

Improving thefatigue performance

of welded stainless steels

By S J Maddox

No embargo

Electronic copyright in this document as follows:Copyright 2006, TWI Ltd

The Welding Institute, Granta Park, Great AbingtonCambridge CB1 6AL, United Kingdom

Telephone: +44 (0)1223 899000Telefax: +44 (0)1223 892588

TWI Ltd 2006

TWI

1363

1.01

/200

5/12

57.3

13631.01/2005/1257.3 Copyright 2006, TWI Ltd

CONTENTS TECHNOLOGY BRIEFING i Background i Objectives i Experimental Approach i

1. INTRODUCTION 1

2. OBJECTIVES 2

3. PROGRAMME ORGANISATION 3

4. EXPERIMENTAL DETAILS 3 4.1. MATERIALS 3 4.2. SPECIMEN DESIGNS 3 4.3. MANUFACTURE OF SPECIMENS 4 4.3.1. Welding 4 4.3.2. Spot Heating 4 4.4. APPLICATION OF IMPROVEMENT TECHNIQUES 4 4.4.1. Toe Grinding 4 4.4.2. TIG or Plasma Dressing of Duplex 5 4.4.3. Ultrasonic Impact Treatment (UIT) 6 4.5. CHARACTERISATION OF TEST SPECIMENS 6 4.5.1. Weld Profiles 6 4.5.2. Residual Stress Measurements 7

5. FATIGUE TESTS 7 5.1. FATIGUE TESTING CONDITIONS 7 5.2. FATIGUE TEST RESULTS 8

6. BASIC FATIGUE STRENGTH OF MAG WELDED SPECIMENS 9 6.1. BACKGROUND 9 6.2. TRANSVERSE FILLET WELDED TYPE S31803 DUPLEX STEEL 9 6.3. TRANSVERSE FILLET WELDED TYPE 304L AUSTENITIC STEEL 10 6.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL 11 6.5. LONGITUDINAL FILLET WELDED TYPE S31803 DUPLEX STEEL 11 6.6. SIGNIFICANCE OF STEEL TENSILE STRENGTH 12

7. INFLUENCE OF WELDING PROCESS ON FATIGUE STRENGTH 12 7.1. BACKGROUND 12 7.2. FILLET WELDED TYPE S31803 DUPLEX STEEL 13 7.2.1. Duplex Plates with Transverse Fillet Welded Attachments 13 7.2.2. Duplex Plates with Longitudinal Fillet Welded Attachments 14


7.3. FILLET WELDED TYPE 304L AUSTENITIC STEEL 14 7.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL 15 7.5. SIGNIFICANCE OF STEEL TENSILE STRENGTH 15

8. APPLICATION OF WELD TOE IMPROVEMENT TECHNIQUES 16 8.1. BACKGROUND 16 8.2. TRANSVERSE FILLET WELDED TYPE S31803 DUPLEX STEEL 17 8.3. TRANSVERSE FILLET WELDED TYPE 304L AUSTENITIC STEEL 17 8.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL 18 8.5. LONGITUDINAL FILLET WELDED JOINTS IN TYPE S31803 DUPLEX STEELS AND TYPE 304L AUSTENITIC 19 8.6. SIGNIFICANCE OF STEEL TENSILE STRENGTH 20 9. INVESTIGATION OF FACTORS THAT COULD REDUCE THE BENEFIT OF WELD TOE IMPROVEMENT TECHNIQUES 21 9.1. FACTORS CONSIDERED 21 9.2. TIG DRESSING OF TYPE 304L AUSTENITIC FILLET WELDS WITH FLAWS 21 9.3. INFLUENCE OF A CORROSIVE ENVIRONMENT 22 9.4. EFFECT OF HIGH TENSILE MEAN STRESS 24 9.4.1. Background 24 9.4.2. TIG Dressed Fillet Welds 24 9.4.3. UIT Treated Fillet Welds 26 9.5. BENEFIT OF IMPROVEMENT TECHNIQUES UNDER SPECTRUM LOADING 27 9.5.1. Background 27 9.5.2. As-Welded Joints 27 9.5.3. TIG Dressed Fillet Welds 28 9.5.4. UIT Treated Fillet Welds 30

10. GENERAL DISCUSSION 31 10.1. INFLUENCE OF WELDING PROCESS 31 10.2. BENEFIT FROM IMPROVEMENT TECHNIQUES 31 10.2.1. Fatigue Failure in the Plate 31 10.2.2. Toe Grinding 32 10.2.3. TIG and Plasma Dressing 33 10.2.4. Ultrasonic Impact Treatment (UIT) 34 10.2.5. Influence of Steel Type 35

11. CONCLUSIONS AND RECOMMENDATIONS 35

12. ACKNOWLEDGEMENTS 39

13. REFERENCES 39


TABLES 1-54 FIGURES 1-59 Annex 1: Fatigue Testing Programme Annex 2: Spot Heating Trials Annex 3: Investigation of the Effect of Weld Toe TIG- or Plasma Dressing on the

Ferrite Level in Duplex Stainless Steel Annex 4: Investigation of Plain Plate Failure in Type 304L Austenitic Stainless Steel

Specimens


i

TECHNOLOGY BRIEFING Background A previous project showed that the fatigue performance of welded austenitic stainless steel is similar to that of conventional C-Mn steels, and that higher strength duplex is no better. Thus, ways are needed of improving the fatigue lives of the lower fatigue strength details, notably fillet welds. This report presents the results of an investigation of possible ways, by choice of welding process or the application of post-weld improvement techniques, of improving the fatigue performance of fillet welded stainless steels. Objectives

To establish methods for achieving improved fatigue resistance from fillet welded joints in austenitic and duplex stainless steels with a target increase of 60% by:

a) Suitable choice and control of welding process; b) Application of a post-weld improvement technique.

To produce practical guidance on the application of the techniques and their benefits in real welded stainless steel structures.To quantify the extra benefit achievable from high-strength duplex or Cr-Mn austenitic stainless steel by the use of improvement measures.

Experimental Approach Three types of fillet welded specimen were manufactured using three types of stainless steel and fatigue tested. The specimens were: 10mm thick S31803 duplex stainless steel plates with either transverse or

longitudinal fillet welded attachments. 10mm thick Type 304L austenitic stainless steel plates with either transverse or

longitudinal fillet welded attachments. 3mm thick Cr-Mn austenitic stainless steel plates with transverse fillet welded pad

attachments. The basic test series were MAG welded, but further specimens were either TIG or powder plasma arc welded (PPAW) to investigate possible improved fatigue performance from the use of these welding processes. In addition, the improvement in fatigue performance attainable using four weld toe improvement techniques, namely grinding, TIG dressing, plasma dressing and ultrasonic impact treatment (UIT), was investigated using MAG welded specimens. The specimens were fatigue tested under axial loading at a stress ratio of R = 0.1 or under high tensile mean stress conditions, usually achieved by cycling down from a fixed maximum stress level close to material yield strength. Most tests were performed in air under constant amplitude loading, but toe ground and plasma dressed specimens were also tested in 3% NaCl solution. Further tests were performed to determine the effect of variable amplitude loading on the benefit of TIG dressing and UIT. Many of the testing conditions were selected specifically to investigate features of service operation that might reduce the benefit of the improvement techniques.


ii

Apart from the fatigue testing, detailed measurements of the weld profiles, metallurgical examinations and residual stress measurements, as appropriate, were carried out in order to assist in the evaluation of the test results. Conclusions and Recommendations On the basis of a programme of fatigue tests on three types of fillet welded joint in three stainless steels, 10mm thick S31803 duplex and Type 304L austenitic and 3mm thick Cr-Mn austenitic, which aimed to establish the improvement in fatigue performance attainable by the use of either a particular welding process or a weld toe improvement technique, the following conclusions were drawn. Measures that improve fatigue performance in air - No support was provided for previous indications that TIG transverse fillet welds

produce better fatigue performance than MAG welds. Similarly, PPA welding showed no particular benefit for any of the specimen types. A key feature seemed to be weld toe radius, with the highest fatigue performance being coupled with a generous toe radius regardless of welding process. However, this link was not established conclusively and there is scope for further investigation.

- All four of the weld toe improvement techniques investigated, grinding, TIG dressing, plasma dressing and ultrasonic impact treatment, improved the fatigue performance of the fillet welds. The extent of the improvement generally increased with decrease in applied stress range, the greatest benefit being an increase in the fatigue limit.

Comparison of weld toe improvement techniques Grinding - The improvement in fatigue strength for welds in 10mm thick steel was up to 60%,

but the possibility of failure in the plate could limit the improvement to 30% making this a more realistic assumption for design. Nevertheless, the increase in the fatigue limit was greater.

- Toe grinding was less successful when applied to thin (3mm) Cr-Mn austenitic steel, partly because of the significant increase in nett section stress due to the loss of thickness. The weld toe re-melting techniques are preferable for such thin material since they do not cause any loss of thickness.

TIG or plasma dressing - TIG dressing was generally more successful than plasma dressing, which was

consistent with the fact that the weld profiles produced were more uniform. - The benefit of TIG dressing was greatest in the 3mm austenitic steel, resulting in a

60% increase in fatigue strength, more in the region of the fatigue limit. - Around 30% increase in fatigue strength was obtained from TIG dressed welds in

the 10mm thick steels and from plasma dressed welds in the duplex.


iii

- However, plasma dressing of the 10mm austenitic steel was of little benefit, due to the small radius of the dressed weld toe.

Ultrasonic impact treatment (UIT) - UIT produced a 30% increase in the fatigue strength of the fillet welds in the 10mm

thick stainless steels, with around three-fold increase in the fatigue limit. Factors that might limit the benefit from improvement techniques - Immersion of unprotected specimens in 3% NaCl solution reduced the fatigue

performance of welds treated using the weld profile improvement methods. However, the tests were limited in terms of the temperature (20C) and cycling frequency (2 to 12Hz). Therefore, further testing is recommended to ensure, in particular, that the effect is not even worse at lower frequencies, as would be relevant to wave loading.

- The fatigue strengths of toe ground and plasma dressed welds in the 10mm thick duplex stainless steel were reduced in 3% NaCl, but still generally 30% higher than that of the as-welded joints in 3% NaCl. Thus, weld toe dressing effectively restored the in-air performance.

- The fatigue performance of plasma dressed welds in 10mm austenitic steel was reduced in 3% NaCl, but there was little benefit from the plasma dressing in air. More effective dressing may experience a greater influence of the corrosive environment.

- TIG dressing was applied successfully to austenitic welds containing embedded solidification crack. However, the flaws themselves acted as fatigue crack initiation sites and limited the fatigue performance of the joints. Thus, they should be avoided by proper control of the welding.

- The benefit of TIG or plasma dressing was not reduced significantly by the application of high tensile mean stress loading or a Gaussian variable amplitude load spectrum.

- The benefit of UIT was effectively lost under either high tensile mean stress or spectrum loading containing stresses above yield.

Influence of type of stainless steel The only observed influences of the steel type were that the higher strength duplex and Cr-Mn austenitic steels could sustain applied stresses above the yield strength of the Type 304L austenitic, and ground or dressed welds in duplex performed better than the austenitic steel in 3% NaCl. Practical issues concerned with the use of improvement techniques - The benefit of an improvement technique was often limited by the occurrence of

fatigue failure in the plate. The Type 304L austenitic steel seemed to be particularly vulnerable, either as a result of the presence of surface notches from mechanical damage or if the maximum applied stress exceeded yield.


iv

- The weld toe re-melting techniques are preferable to grinding for very thin material since they do not cause any loss of thickness.

- Investigation of the effect of TIG dressing on the austenite-ferrite content of duplex stainless steel showed that it was prudent to add 1 to 2% nitrogen to the shielding gas to ensure an acceptable phase balance.

- Care should be taken to avoid spatter from TIG or plasma dressing since it can provide the site for fatigue crack initiation. If it does occur, it should be ground smooth to avoid premature fatigue failure.

- In both TIG and plasma dressing, it seemed that a finished radius of at least 6mm was required in order to ensure 30% improvement in fatigue strength.

- If any part of the weld toe is missed during TIG or plasma dressing that region will be as-welded and there will be no benefit from the dressing treatment. Thus, care is needed to avoid missing the toe, which should be corrected if detected during post-treatment visual inspection of the dressed weld toe.

The effect of the various improvement techniques and the influence of the type of loading and environment on that improvement are summarised in the following table.

Fatigue strength improvement *

Method Weld detail and steel Air 3% NaCl

High tensile mean stress Spectrum loading

Toe grinding

Transverse duplex Better than x1.3, x2.5 near fatigue limit

Reduced benefit, but close to x 1.3 in high-cycle regime

N/I N/I

Transverse 304L x1.3, or x2 near fatigue limit, but limited by potential plate failure

N/I N/I N/I

Transverse Cr-Mn

Limited benefit due to reduced cross-section; x 1.3 if nett stress < yield

N/I N/I N/I

TIG dressing

Transverse duplex x1.3 N/I x 1.3 As constant amplitude in air

Transverse 304L x1.3, or x2 near fatigue limit, but limited by potential plate failure

N/I x1.3 or more in high-cycle regime provided Smax < yield

x1.3

Transverse 304L with flaws

None; failure from flaws

N/I N/I N/I

Transverse Cr-Mn

Better than x1.3 N/I x 1.3 Reduced benefit, but still x1.3

Plasma dressing

Transverse duplex Better than x1.3 Reduced benefit, but generally around x 1.3

N/I N/I

Transverse 304L None due to sharp plasma dressed weld toe profile.

Below as-welded in air

N/I N/I

Longitudinal duplex

x1.3 but limited by potential failure from weld root

N/I N/I N/I


v

Fatigue strength improvement *

Method Weld detail and steel Air 3% NaCl

High tensile mean stress Spectrum loading

Longitudinal 304L

x1.3 if Smax < yield N/I N/I N/I

Transverse Cr-Mn

Barely x1.3 ; may need condition eg minimum radius of dressed toe > 6mm.

N/I N/I N/I

UIT Transverse duplex x1.3 N/I Reduced or even lost at low stresses if high R

x 1.3

Transverse 304L x1.3 but limited by potential plate failure.

N/I Benefit lost if Smax > yield

Benefit lost if Smax > yield

Longitudinal duplex

Better than x1.3, up to x3 near fatigue limit.

N/I N/I N/I

Longitudinal 304L

Better than x1.3, up to x3 near fatigue limit

N/I N/I N/I

Note: * Compared with as-welded specimens tested with R=0.1 in same environment; N/I Not investigated

1


1. INTRODUCTION Growing awareness of the need for more durable products makes stainless steels increasingly attractive alternatives to conventional ferritic C-Mn steels. As a result, they are finding their place in structural applications, such as vehicles and construction equipment, as well as the more obvious chemical processing and power generation industries. Since in many cases avoidance of fatigue failure is a major design criterion, this emphasises the need for sound design guidance, particularly for welded joints since these are usually the most critical details from the fatigue viewpoint. A previous project (1) provided fatigue design guidance for as-welded stainless steels. In the present project, attention is turned to ways of improving their fatigue performance.

Fatigue is the major cause of failure in welded structures, the weld toe being the primary source of crack initiation. This reflects the severe stress concentration associated with the weld toe, due partly to the sharp change in section but compounded by the presence of high tensile residual stresses and microscopic crack-like imperfections, which are an inherent feature of arc welds in steels (2,3). Recognition of the importance of the weld toe led to the development of a number of techniques for reducing its severity and hence improving the fatigue life of the welded joint (2-5). Weld toe improvement techniques rely on two main principles:

a) Reduction of stress concentration - the main techniques for doing this are grinding or

re-melting the weld toe with the arc of a TIG or plasma-welding torch. Apart from smoothing the abrupt change of section, these techniques are capable of removing any crack-like flaws. It is believed that the resulting improvement in fatigue life is due to the introduction of a significant fatigue crack initiation period, which is negligible in the case of the as-welded joint.

b) Introduction of compressive residual stresses - the main techniques for doing this are

shot, needle and hammer peening. In this case the improvement technique does not necessarily reduce the stress concentration associated with the weld toe or remove the flaws. Instead, it has the effect of clamping the weld toe region in compression, so that under fatigue loading the effective cyclic stress range is partly, or even completely, in compression and hence less damaging. Thus, the benefit is believed to be due to a reduction in the rate of growth of a fatigue crack from the weld toe discontinuity.

In both cases, there is reason to expect that the improvement in fatigue life will be greater in a high strength than low strength steel (5-9). This contrasts with the behaviour of the as-welded joint, for which the fatigue strength is independent of material tensile strength. In the case of the weld toe dressing techniques, the introduction of a significant fatigue crack initiation period is relevant since the crack initiation endurance will increase with increase in material tensile strength. In the case of the compressive residual stress techniques, to some extent the magnitude of the beneficial residual stress will be higher in a high strength material. Although improvement techniques have been well known for many years, the vast majority of experience of their benefit is confined to their use on welded ferritic structural steels (2-4). In particular, no evidence could be found that they will be equally effective when applied to welded stainless steels. Thus, the main aim of the present project was to confirm that they

2


would also be effective when applied to welded stainless steels. The profile improvement techniques selected were weld toe grinding and weld toe re-melting by TIG or plasma dressing. The residual stress technique was ultrasonic impact treatment (UIT). At the same time it was recognised that there is a need to ensure that the benefit of a weld toe improvement technique seen in a laboratory test will be realised in a real structure. In this respect, it is known that differences in residual stress levels, fatigue loading conditions and environment can be significant. With regard to the first two issues, there is evidence from tests on welded steel beams that improvement techniques, notably those relying on the introduction of compressive residual stresses, are less effective in large welded structures, containing high tensile residual stresses from welding, than when they are used to treat small-scale laboratory specimens (10,11). This is undoubtedly linked with the fact that the increase in fatigue life resulting from an improvement technique, particularly one relying on compressive residual stresses, tends to decrease with increase in applied tensile mean stress (12). Similarly, doubt exists about the effectiveness of improvement techniques, particularly those relying on compressive residual stresses, under variable amplitude loading, as normally occurs in real structures (9,13) With regard to the third issue, experience with welded C-Mn steels has clearly shown that a corrosive environment can eliminate the benefit of weld toe grinding (14). There is the need to check the extent to which the same may be true for stainless steels. Another matter that could be relevant to the use of weld toe improvement techniques on austenitic stainless steel is the presence of near-surface micro-fissures. These can arise when fully austenitic weld metal is used. Although they do not appear to be significant in relation to the fatigue performance of the as-welded joints (15), their presence may inhibit the application of an improvement technique. Certainly this has been found to be the case in C-Mn steel welds containing shallow cold laps at the weld toe (16). Finally, there are positive indications, including from previous projects (1,17), that it may be possible to improve the fatigue performance of some types of welded joint by the choice and control of the welding process. In particular, TIG and plasma welding appear to offer similar benefit to TIG or plasma dressing (17), which suggests that it is the production of a favourable weld toe shape that leads to an improvement. Similar claims have been made for a relatively new process, powder plasma arc welding (PPAW) (18). This project set out to investigate all these issues on the basis of fatigue tests performed on fillet welded joints in three types of stainless steel, basic Ni-Cr austenitic, higher strength Cr-Mn austenitic and high-strength ferrite-austenite (duplex). 2. OBJECTIVES

To establish methods for achieving improved fatigue resistance from fillet welded joints in austenitic and duplex stainless steels with a target increase of 60% by:

a) Suitable choice and control of welding process; b) Application of a post-weld improvement technique.

3


To produce practical guidance on the application of the techniques and their benefits in real welded stainless steel structures.To quantify the extra benefit achievable from high-strength duplex or Cr-Mn austenitic stainless steel by the use of improvement measures.

3. PROGRAMME ORGANISATION This was a collaborative project undertaken by seven organisations from five European countries, as detailed in Annex 1, with TWI acting as co-ordinator. The main experimental work involved the fabrication of the test specimens, the application of post-weld improvement methods, fatigue testing and metallographic examination of the test specimens. The project was in three phases, running concurrently, with input from each partner as summarised in Annex 1. The phases were: Phase 1 - Fatigue testing of untreated welded specimens. Phase 2 - Fatigue testing of toe treated MAG welded specimens Phase 3 - Fatigue testing of toe treated MAG welds under conditions that could limit the

beneficial effect of the treatment. 4. EXPERIMENTAL DETAILS 4.1. MATERIALS The fatigue test specimens were fabricated from three types of stainless steel plate: 10mm thick type 1.4301 (referred to as 304L austenitic); 10mm thick type 1.4462 (referred to as S31803 duplex); 3mm thick type 1.4376 (referred to as austenitic Cr-Mn). The dimensions are nominal values and, in practice, the plate thicknesses were approximately 9.5 or 2.5mm, respectively. The material properties obtained from tensile tests and chemical analyses performed on actual plate samples are given in Table 1. In later references to the strengths of the steels, the 0.2% proof strength is assumed to represent yield. 4.2. SPECIMEN DESIGNS Three designs of specimen consisting of plates with fillet welded attachments were selected, as shown in Fig.1 to 3. Types 1 and 2 were produced in the 10mm thick steels, while Type 3 specimens were made from the 3mm Cr-Mn steel. In all cases the most likely mode of fatigue failure is by fatigue crack growth from the weld toe through the main (loaded) plate. Thus, the fatigue performance of such details should be amenable to improvement by the use of welding conditions that produce favourable weld profiles or the application of post-weld improvement techniques. If the fatigue performance of the test specimens proves to be particularly good, there is a high risk of fatigue failure initiating in the parent plate itself, especially if it is gripped in wedge jaws for testing. Therefore, in some cases the specimens were waisted to smaller widths, down to 50mm in some cases, to ensure that the nominal stress in the vicinity of the weld details was significantly higher than that elsewhere.

4


4.3. MANUFACTURE OF SPECIMENS 4.3.1. Welding Most of the specimens were MAG welded but the potential benefit in terms of fatigue resistance of using TIG or PPAW was also investigated. PPAW is particularly suitable for the welding of thin materials. It was necessary to make the present fillet welds in three passes when PPA welding the 10mm thick specimens. The resulting high heat input, higher than normally recommended for the welding of duplex stainless steels, caused excessive distortion, suggesting that the process is not well suited for this thickness. In the case of the Series 1.1 TIG welded duplex specimens, that were intended to reproduce the high fatigue strength seen in the previous project (1), care was taken to use identical welding conditions as before. The final welding conditions for all the specimens fabricated are summarised in Table 2. Figure 4 shows examples of some of the welds produced. A particular requirement for Series 3.1 was to produce MAG fillet welds in Type 304L austenitic steel containing micro-fissures. An attempt was made to do this by making the welds with special type 347 flux-cored filler wire, balanced to produce a fully austenitic microstructure. However, it was not successful and, as an alternative, the procedure was modified to produce sub-surface solidification cracking in the same region. Full details of the resulting welding conditions are included in Table 2, while Fig.4 shows examples of the flaws produced. It was considered that these welds were still suitable for checking the ability of TIG dressing in the presence of weld defects. 4.3.2. Spot Heating One of the objectives of the proposed tests on welds treated by UIT, to introduce compressive residual stress, was to ensure that such treatment was still beneficial in the presence of high tensile residual stresses. Such residual stresses can be expected to be present in virtually any real as-welded structure. However, relatively narrow specimens with transverse fillet welds of the types shown in Fig.1 and 3 are not expected to contain high levels of residual stress. Spot heating is a technique that has been used to induce tensile residual stresses in C-Mn steel specimens but, as far as is known, the technique has not been applied to stainless steels. One issue of concern in relation to the duplex steel was the metallurgical effect of such spot heating, particularly the possibility that it might precipitate undesirable intermetallic phases. These could influence the corrosion resistance and fracture toughness of the steel, although such factors would not be relevant in their fatigue testing in air. Therefore, trials were performed on specimens of the type shown in Fig.1, in both austenitic and duplex steels, to see if spot heating could induce high tensile residual stresses. The trials included examination of the microstructure of the spot-heated duplex steel to check for the presence of third phases. Details of the study are presented in Annex 2. The outcome was that spot heating did induce relatively high tensile residual stresses in both steels without having an adverse effect on the microstructure of the duplex steel. Consequently, it was decided to spot heat the test specimens used to investigate the beneficial effect of UIT. 4.4. APPLICATION OF IMPROVEMENT TECHNIQUES 4.4.1. Toe Grinding The primary aim of weld toe grinding is to remove or reduce the size of any local sharp imperfections introduced by the welding, such as undercut, cold laps and even crack-like

5


flaws, from which fatigue cracks can readily propagate. At the same time, it aims to reduce the local stress concentration effect of the weld profile by smoothly blending the transition between the plate and the weld face. The draft IIW recommendations for treating welds in normal structural steels (19) were the basis of the procedure used to treat the present welds. In general, the procedure requires complete removal of material at the actual weld toe to a sufficient depth, usually between 0.5 and 1mm, to remove all traces of undercut or other weld toe discontinuities. The final ground surface must be smooth with any grinding marks oriented at right angles to the weld toe, and have a radius no less than 0.25 x plate thickness and 4 x grinding depth. The last two conditions are to reduce the stress concentration due to the groove produced by grinding, to ensure that it does not itself act as a notch. Transverse fillet welded specimens of the types shown in Fig.1 and 3 were treated. Some details of the grinding equipment and the method of application are given in Fig.5. Referring to Fig.5(a), the toes of the fillet welds in the 10mm thick specimens were treated in two passes, starting with a 12mm diameter tungsten carbide burr with a 2mm radius tip, and finishing with a 19mm diameter abrasive burr with a 2.5mm radius tip. Figure 5(b) and (c) show photographs of sections of 304L austenitic steel specimens before and after burr grinding. A smaller burr, 6mm diameter tapering to 1.5mm radius at the tip, was used to grind the weld toes in the thinner transverse fillet welded pad specimens (Fig.3), as indicated in Fig.5(d). In this case the grinding operation was done in a single pass. Fig.5(e) shows examples of toe ground specimens. Based on the above two criteria for the shape and depth of the groove produced by toe grinding, the tool tip radius determined the maximum grinding depth to be 0.5mm in the 10mm thick specimens and 0.375mm in the 3mm thick specimens. In the event the mean grinding depths were 0.5 and 0.1mm, respectively. However, as will be evident from Fig.5(e), the depth was up to 0.3mm in some of the thin ones. Even so, these depths are less than the 0.5mm that has been found to be required for treating C-Mn steel welds. However, visual inspection, supported by the available weld cross-sections, indicated that the grinding was sufficient to remove the original toe and produce a smooth transition from the plate to the weld surface. 4.4.2. TIG or Plasma Dressing of Duplex The aims of TIG or plasma dressing are the same as those for grinding, namely to remove any weld toe imperfections and to reduce the stress concentration effect of the local weld toe profile by providing a smooth transition between the plate and the weld face. However, these aims are achieved by re-melting the weld toe region using a TIG or plasma welding torch, without the addition of any filler metal. Recommendations for the application of TIG dressing have been made by the IIW (19) and these again were the basis of the procedures used to treat the present stainless steel specimens. As in the case of spot heating, an issue of concern with regard to the re-melting of the weld toe in the duplex stainless steel was that the high local temperature reached would change the austenite-ferrite balance, which could be detrimental to the steels properties. In the same way, autogenous welds in duplex stainless steels may contain unacceptably high levels of ferrite. However, this can be avoided by the addition of nitrogen, which is an austenite stabiliser, to the shielding gas. Therefore, trials were carried out to check the possibility that the same approach would ensure that TIG dressing did not produce unacceptable levels of ferrite. Full details of the study are given in Annex 3. It was concluded that it would be prudent to add 1 to 2% nitrogen to the shielding gas when TIG, or indeed plasma, dressing

6


duplex. However, at this stage of the project plasma dressed duplex specimens had already been produced without the addition of nitrogen to the shielding gas. Therefore, samples from those specimens were also examined to check the austenite-ferrite balance. As noted in Annex 3, fortunately the level of ferrite was found to be only slightly outside the acceptable range and therefore the treated specimens were considered to be suitable for fatigue testing. However, the remaining TIG dressing of duplex specimens was performed with nitrogen in the shielding gas. The TIG and plasma dressing conditions finally adopted for the various specimen types are summarized in Table 3. In the case of the transverse fillet welded specimens in 10mm thick steel (Fig.1), the dressing was carried out on the welded panels, which were then cut into specimen widths. It was hoped that this would ensure reasonably uniform dressing and so minimize scatter in fatigue performance. However, it was necessary to apply the dressing manually in the case of the short welds in the specimens incorporating longitudinal fillet welds and the transverse fillet welded pad specimens in thin sheet (Fig.2 and 3, respectively). Examples of treated welds are shown in Fig.6. These include TIG dressed Series 3.1 welds with solidification cracks. The original welds had rather peaky profiles but TIG dressing still produced reasonable weld toe profiles. 4.4.3. Ultrasonic Impact Treatment (UIT) Ultrasonic impact treatment is a technique that was developed in the former Soviet Union some decades ago but has only emerged as a commercial proposition in the past 10 years or so (20). Now the equipment is being manufactured in the USA, Canada and The Netherlands and is available to industry, usually on a contract basis. Figure 7 shows a weld toe being treated by UIT. As will be evident, UIT has some similarity with needle peening, although claims are made that it has a more far reaching effect on the material treated (21). However, in the context of the use of these techniques for fatigue life improvement, in both cases the benefit seems to derive mainly from the introduction of beneficial compressive residual stresses resulting from plastic deformation of the surface in the region of the weld toe. UIT is applied manually, in much the same way as needle peening. Thus, the UIT tool, with needle dimensions chosen to allow access to the weld toe itself, is moved along the weld toe, typically at 300-1500mm/minute, in order to plastically-deform the surface and leave a uniform groove with a smooth surface. More than one pass is usually needed to ensure full coverage, which is checked by visual inspection. In the present case, the treatment rate was around 400mm/min. and 4 or 5 passes were required to complete the treatment. 4.5. CHARACTERISATION OF TEST SPECIMENS 4.5.1. Weld Profiles Macro-sections of typical weld profiles in the test specimens are shown in Fig.4-6. It was anticipated that the weld profile, especially the weld toe radius, would influence the weld toe stress concentration and hence the fatigue performance of the welds tested. This was expected to be especially true in the case of specimens designed to include favourable weld profiles, either from the choice of welding process or from the use of weld toe grinding, TIG dressing or plasma dressing. Indeed, differences in weld profile between such test series might explain

7


any variation in fatigue performance. Therefore, measurements were made of weld profiles on selected specimen types. The weld toe radius and weld face angle with respect to the plate surface were measured using a MAXTASCAN X-Y table measuring facility, fitted with a video monitor optical system capable of accuracy of 0.01mm and 0.5 for the radius and angle respectively. The radius was determined by locating five points near the weld toe, moving down the weld face across the toe to the plate surface. The systems software then fitted a circle to these points from which the weld toe radius was deduced. Trials showed that five points were sufficient to produce the best estimate of the radius, since the use of more points gave the same result. The weld face angle at the weld toe was also measured in some cases by locating one point on the plate surface near the weld toe and another at the beginning of weld toe curvature. In both cases, several measurements were made along the weld toe to establish the mean values and the variations in each parameter. Typical results are shown in Fig.8, while Table 4 summarises the results for the range of test series in which weld profiles were measured. Less detailed measurements were also made on some weld profiles, making measurements on photographs of macro-sections. Usually just one section was available, containing four weld toes, and so these measurements did not provide any significant information about the variation in weld profile in the test series as a whole. The results are included in Table 4. Reference will be made to all these measurements when the fatigue test results are discussed. 4.5.2. Residual Stress Measurements Some residual stress measurements were made using X-ray diffraction equipment. The results of measurements made at the weld toe, the region of primary interest since it is the usual site for fatigue crack initiation, are presented in Table 5. It will be observed that the residual stresses in all the as-welded specimens considered were very low or even compressive. This is not surprising in view of the narrow widths of the test specimens. As anticipated, UIT introduced significant beneficial compressive residual stress, up to around 60% of yield in the case considered. The other interesting result is that burr grinding the weld toes in both the austenitic and duplex steels also induced very high compressive residual stresses. These can be expected to act in conjunction with the improved weld profile to increase the fatigue strength of the welded joint. 5. FATIGUE TESTS 5.1. FATIGUE TESTING CONDITIONS The specimens were fatigue tested under axial loading in hydraulic fatigue testing machines. Tests were performed at stress ratios of R = 0.1 or 0.5, or cycling down from a fixed maximum tensile stress, Smax, close to material proof strength in some cases. The stress ratio varied from test to test in the latter case, increasing with decrease in applied stress range. This cycling down from yield type of loading was adopted to simulate the presence of high tensile residual stress, as would be expected to be present in real welded structures, but not necessarily in test specimens. Most of the tests were performed under constant amplitude loading, in laboratory air at ambient temperature and cycling frequencies in the range 3-20 Hz. However, the beneficial

8


effects of some of the improvement techniques were also checked under variable amplitude loading. The loading sequence used was a Gaussian spectrum with a sequence, or block, length of 5 x 104 cycles, as detailed in Fig.9. The applied stress ratio was R = 0.1. The load cycles in a block were applied in a random order. This spectrum was chosen as being a reasonable representation of a wide range of service load spectra, as well as being similar to the one used previously to test as-welded stainless steel specimens (1). The fatigue test results obtained under variable amplitude loading will be presented in terms of the maximum stress range in the spectrum, as an unambiguous indicator of the test loading. However, an alternative method for comparing the results obtained under constant and variable amplitude loading is to consider the latter in terms of an equivalent constant amplitude stress range, calculated assuming that Miners linear cumulative damage rule is correct. This approach was used in the previous project (1). On this basis, the equivalent constant amplitude stress range eq is:

m

i

im

ieq n

n/1

)(

= [1]

where the exponent m is the slope of the constant amplitude S-N curve for the same type of specimen and ni is the number of cycles applied at stress range i (i = 1,2 3. . . .). In the present case, ni was the number of cycles in each block, for which ni = 5 x 104 cycles, and the summation m.ni was performed for one block. With regard to the assumption that Miners rule is correct, tests under the same spectrum in the previous project (1) confirmed that this was approximately the case. Thus, again it is considered to be reasonable to apply the equivalent constant amplitude stress in the analysis of the present spectrum loading fatigue test results. As before, single slope S-N curves, neglecting a fatigue limit, were fitted to the relevant constant amplitude data to establish the appropriate value of m. Some fatigue tests were also performed with the specimens freely corroding in 3% NaCl solution, to check if any benefit seen in air is still obtained in a corrosive environment. In these tests, the specimen was immersed completely in the water, which was aerated and its temperature maintained at 20 2oC. The tests were performed at frequencies ranging from 2 to 12Hz, the frequency used being chosen to ensure that the test duration was sufficient to enable the corrosive environment to have its full effect. However, in retrospect the frequency was too high in some cases to achieve this objective and, ideally, further tests at around 1-2Hz are needed to establish the potential effect of a salt-water environment. In some test series, specimens that were unbroken after endurances well beyond the expected fatigue life were re-tested at a higher stress, to increase the database. If such specimens eventually failed, the fracture surface was examined for evidence of crack growth from the first test. The result obtained in the re-test was only accepted as valid if there was no such evidence. 5.2. FATIGUE TEST RESULTS The fatigue test results are given in Tables 6-54. The specimens are identified in terms of the phase number (see Section 3) followed by series number followed by specimen number in that series. Thus, specimen number 2.6-4 refers to the fourth specimen in Phase 2 Series 6.

9


Some series numbers are missing since the results were subsequently found to be invalid and are therefore omitted from this report. The original series numbers are retained to avoid confusing the laboratory records. Unless stated otherwise, fatigue failure was by fatigue crack growth from a weld toe through the main plate, as illustrated in Fig.10. Other failure modes included crack growth from the weld root through the plate (included in Fig.10), crack growth across the plate where it was gripped, crack growth through the plate from weld spatter and simply crack growth in the plate remote from the weld detail. The test results are expressed in terms of the nominal stress range in the loaded plate, applied force/cross-sectional area of plate, the stress commonly used to express design S-N curves for welded joints. However, some specimens were misaligned, due to distortion from welding, with the result that secondary bending stresses were induced when they were loaded. In some such cases the magnitude of the resulting total nominal stress near the weld toe was measured using electrical resistance strain gauges and this local stress is included in the table of results. With regard to the fatigue lives quoted, the criterion of failure was that a fatigue crack had propagated through the plate thickness, but in most cases the specimens were tested to complete rupture. Cases where a result was obtained from a re-test are indicated in the tables. In addition, in some cases it was of interest to determine the weld profile in a specific specimen and relevant measurements are included in the appropriate table. In order to address all the objectives of the project, the test results were evaluated from four viewpoints. First, relevant test series from Phase 1 were used to establish S-N curves representing the basic fatigue performance of the as-welded specimens for comparison with the results obtained from specimens with improved welds. Second, these basic data were compared with the results obtained from those series, from Phase 1 but also from the previous project (1) in some cases, used to investigate the possible improvement in fatigue performance due to the choice of welding process. Similarly, they were then compared with results obtained in Phase 2 from specimens that had been treated with one of the post-weld improvement techniques to determine the benefit obtained. Finally, the results obtained in Phase 3, in circumstances that were expected to limit the benefits of the improvement techniques, were evaluated. Where appropriate, the fatigue test results obtained under specific conditions from more than one steel type were compared in order to identify any advantage offered by the choice of steel, particularly its yield strength. 6. BASIC FATIGUE STRENGTH OF MAG WELDED SPECIMENS 6.1. BACKGROUND The objective in this part of the project was to establish reference S-N curves against which to assess the effect on fatigue performance of the choice of welding process or the application of a post-weld improvement technique. In general, results were obtained from MAG welded specimens of all three types under R=0.1 and under high tensile mean stress conditions (ie R>0.1 or constant maximum stress Smax close to the proof or yield strength). 6.2. TRANSVERSE FILLET WELDED TYPE S31803 DUPLEX STEEL The relevant results, Series 1.2 at R = 0.1 and Series 1.3 tested with Smax = 400MPa, close to yield, are given in Tables 7 and 8 respectively, and plotted in Fig.11. No other results were found in a literature search for this detail in duplex welded by MAG. Therefore, the present results are compared with the scatter-band enclosing a large database for the same weld detail in ferritic C-Mn steels (22). The results for the two series are seen to be in very good

10


agreement despite the difference in loading conditions. Consequently, a single S-N curve was fitted, as shown. This proved to be shallower than that for the C-Mn steel data. Residual stress measurements on similar specimens (Table 5) indicated relatively low tensile or even negligible residual stress near the weld toe, which could partially explain the shallow S-N curve. However, it was still surprising to find that applying loading that cycled down from a fixed high tensile stress did not overcome the influence of the residual stress and give lower lives, with a steeper S-N curve, than loading with R = 0.1. 6.3. TRANSVERSE FILLET WELDED TYPE 304L AUSTENITIC STEEL The relevant results from Series 1.7 tested with R = 0.1 and Series 1.8 tested with Smax = 250MPa (ie yield), are detailed in Tables 12 and 13 respectively. They are also plotted together with the results obtained from nominally similar specimens in the previous project (1) in Fig.12(a). There is some consistency between them but both sets of data are very widely scattered and do not follow well-defined S-N curves. A problem with the results from Series 1.7 is that only four of them were actually valid, in the sense that fatigue failure was from a weld toe. The remaining specimens failed in the parent plate, usually where the specimen was gripped. The results from Series 1.8, obtained under the theoretically more severe Smax = 250MPa loading conditions, are still very widely scattered and provide even less indication that they lie on a single S-N curve. No explanation could be found for these results from an examination of the test specimens. The wide scatter in the present results was surprising in the light of other data for fillet welded austenitic steels. Figure 12(b) shows such data, obtained from plates with transverse fillet welded attachments (23), cruciform joints between plates (24) and cruciform joints between rectangular section tubes (25). In all cases, fatigue failure was from the fillet weld toe. Regression analysis of these data gave a well-defined S-N curve, which agreed very closely with the 95% confidence intervals enclosing C-Mn steel data (22), as shown previously in Fig.11. Figure 12(b) shows those data in comparison with both the mean and 95% confidence intervals (mean 2 standard deviations of log N, or mean 2SD) obtained by regression analysis of the austenitic steel data. The 95% confidence intervals enclosing the published data for fillet welds in austenitic steels are included in Fig.12(a). As will be seen, although many of the present results lie within the scatter-band enclosing the published data, they and those from the previous project lie towards the upper limit, or even above it, and suggest a shallower S-N curve. This might have been due to the presence of favourable residual stresses. As seen in Table 5, a compressive residual stress of 71MPa was measured in a similar specimen. If this is typical it could be one factor to explain the generally favourable results obtained from these test series, as well as the shallow S-N curve. However, as noted earlier, it had been expected that cycling down from a high tensile stress close to yield would eliminate any influence of the residual stress state in the specimens and bias the results towards lower-bound fatigue performance, with a steeper S-N curve than that obtained at low R values. Clearly this was not the case with the present Series 1.8 austenitic stainless steel specimens. In a further attempt to understand the differences between the data, it was decided to extend the present database, particularly that for Series 1.8, by performing extra tests. Some of these were performed using spare Series 1.7 specimens and some that had failed in the grips, after

11


removing the cracked part and reducing their widths. In addition, another panel was produced, using exactly the same welding procedure as before, and specimens extracted for testing. Some of these were instrumented with strain gauges to check for any evidence of secondary bending stress due to misalignment of the specimens. Such misalignment could explain the wide scatter in the fatigue test results. In the event, only one of the specimens showed such evidence, indicating that misalignment of the specimens was not a major issue. The extra data are included in Tables 12 and 13 and Fig.12(a). Considering the new larger database from the present and previous projects, the overall picture is little changed in that the results still suggest a significantly shallower S-N curve than the published data. This is evident from the mean curve fitted only to the results that refer to failure from the weld toe, which is included in Fig.12(a). However, it is noticeable that all the results from failed specimens of Series 1.7 and the extra tests lie within the published data scatter-band. They are still widely scattered, but clearly could be regarded as belonging to the same population. This draws attention to Series 1.8 as perhaps the main source of the problem of correlating the results from the various series. In particular, recalling that the extra Series 1.8 results were obtained from Series 1.7 specimens or the new panel, it is possible that there was some aspect of the welded panel from which the original Series 1.8 specimens were extracted, such as widely varying residual stresses or weld profiles, that contributed to the wide scatter in fatigue lives. This being the case, it would be legitimate to neglect the original Series 1.8 results in order to establish a reference S-N curve to represent as-welded Type 304L austenitic steel transverse fillet welds for comparison with the results obtained from improved welds. This issue is re-considered in Section 6.6. 6.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL The results obtained from the two series of as-welded MAG Cr-Mn steel specimens, Series 1.11 and 1.12, are given in Tables 16 and 17 respectively and plotted in Fig.13. Also shown for comparison is the scatter-band enclosing the published data from austenitic steel specimens (from Fig.12(b)). In contrast to the results obtained from the Type 304L austenitic steel specimens these data provide well-defined reference S-N curves for comparison with test results obtained from improved welds. That obtained for Series 1.12 under loading that cycled down from a constant maximum tensile stress, 420MPa in the case of this high tensile strength austenitic steel, is lower and steeper than that obtained at R = 0.1, as anticipated. The Series 1.12 results are also consistent with the published data for austenitic steels. In contrast, those obtained from Series 1.11 at R = 0.1 lie on an S-N curve that is significantly shallower than that for the published data. However, there is good agreement between these results and those obtained from the transverse fillet welds in 304L austenitic steel, which also suggest a shallow S-N curve. This was rather surprising from one point of view. These specimens were only 3mm thick, and yet their fatigue performance was no better than that obtained from thicker austenitic steel specimens. In general it is found that the fatigue strengths of welded joints failing from the weld toe tend to decrease with increase in plate thickness, leading to the well-known thickness effect penalty found in most fatigue design rules. Such a thickness effect is not apparent from the present results from 3 and 10mm thick specimens. 6.5. LONGITUDINAL FILLET WELDED TYPE S31803 DUPLEX STEEL The results for MAG welded longitudinal fillet welded joints, Series 1.15, are presented in Table 20. They are plotted in Fig.14, together with published data for the same weld detail (27,28). It is assumed that these data were obtained from MAG welded specimens but the

12


welding process was not always identified. In contrast to the results for transverse fillet welds discussed above, the results for this type of specimen are seen to be in good agreement with other data. In view of this, all the data were combined for regression analysis and the resulting 95% confidence intervals are included in Fig.14. These results are used later as a reference against which to examine the effect of TIG or PPA welding. It is of interest to note that one of the useful characteristics of this type of longitudinal fillet welded specimen is that it displays relatively little scatter in fatigue lives (2). This is thought to reflect the good reproducibility of the weld toe geometry over the short length of weld at the end of the attachment but chiefly the fact that the specimen retains a relatively high level of residual stress after welding, generally close to tensile yield. This adds weight to the theory that the fatigue performance of the present transverse fillet welded specimens in 10mm thick steel was strongly influenced by the presence of favourably low tensile or compressive residual stresses. 6.6. SIGNIFICANCE OF STEEL TENSILE STRENGTH It is well established that the fatigue lives of welded joints in ferritic steels are independent of the tensile strength of the steel, providing a severe limitation on the use of high-strength steels for welded fabrications. There is some evidence to indicate that the same is true for stainless steels (eg 1, 27, 28), and the present results from MAG welded transverse fillet welded joints in Type 304L austenitic and S31803 duplex steel add further evidence of this. The relevant results obtained with R = 0.1, from Series 1.2 and 1.7, are compared in Fig.15(a). It will be evident that there is no significant difference between the two steels, and certainly no evidence that the higher strength duplex has performed any better than the austenitic steel. The same conclusion can be drawn from the results obtained under higher stress ratios, Series 1.3 and 1.8, as seen in Fig.15(b). No attempt has been made to fit an S-N curve to these data in view of the very wide scatter obtained from Series 1.8. However, recalling the doubt expressed earlier about the validity of the results obtained from the original Series 1.8, it is interesting to see that there is reasonable correlation between the results for the two duplex Series, 1.2 and 1.3, and those from austenitic Series 1.7 and the extra Series 1.8. These data are shown in Fig.16, together with the mean and mean 2SD curves obtained by regression analysis of the combined data. As will be seen, the scatter is still quite wide but otherwise there is good correlation between the data. On this basis, the mean and mean 2SD are used later as reference S-N curves for transverse fillet welds in both steels, covering a wide range of stress ratios, in the evaluation of the improvement methods. 7. INFLUENCE OF WELDING PROCESS ON FATIGUE STRENGTH 7.1. BACKGROUND One of the aims of this project was to show that an improvement in the fatigue performance of MAG welded stainless steels can be achieved by using a different welding process. The two processes that were expected to show this were TIG and PPA welding. Both of these welding processes were compared with MAG in the case of welded joints made in the 10mm thick austenitic and duplex steels, while only PPAW and MAG were compared in the case of the 3mm austenitic Cr-Mn steel.

13


7.2. FILLET WELDED TYPE S31803 DUPLEX STEEL 7.2.1. Duplex Plates with Transverse Fillet Welded Attachments The relevant test Series are 1.1 to 1.5 for MAG, TIG and PPA welds tested with R = 0.1 or under high tensile mean stress conditions. The results are presented in Tables 6-10 and those for TIG and PPA welds are plotted in Fig.17(a). Instead of showing the individual test results for MAG welds, the scatter-band enclosing the reference data for MAG welds, from Fig.16, is used to represent their performance. The mean S-N curve fitted to the Series 1.1 TIG welded specimens is also shown. In spite of the relatively low residual stresses measured in similar specimens, close to zero (see Table 5), there was no evident influence of applied mean stress on the fatigue lives of the PPA welded specimens. Therefore, a single S-N curve was fitted to the results from both series, as shown in Fig.17(a). Figure 17(a) also includes the results for TIG welds from the previous project (1). These are very high, with few specimens actually failing from the weld toe, and were one of the reasons for supposing that the fatigue performance of welded stainless steels could be improved by choice of welding process. However, it will be evident from Fig.17(a) that the exceptionally high fatigue strength seen previously is not repeated here and there is no indication that the fatigue performance of this type of welded joint can be improved by the use of TIG welding. However, this is not altogether surprising when considering the weld toe profiles. As seen in Table 4, the weld toe radius in the present TIG welds (Series 1.1) was only 0.53mm. On the basis of available macro-sections, the corresponding radius in the TIG welds tested in the previous project was around 10 times larger, leading to a much-reduced local stress concentration. Thus, further work is needed to confirm the potential benefit of TIG welding, and to establish the welding conditions needed to achieve that improvement. The PPA welds also offered no advantage over the MAG welds, except perhaps at very high applied stresses. However, in contrast to the TIG welds, some improvement would be expected on the basis of their profiles. As noted in Table 4, limited measurements indicated a weld toe radius of around 6mm, which is similar to that of the TIG welds in the previous project (1). One feature of the PPAW specimens that could have confused the results was that they had suffered angular distortion. Consequently, axial loading introduced secondary bending in the region of the welded attachment, increasing the stress on one surface and decreasing it on the other. Series 1.5 specimens were instrumented with strain gauges to measure the resulting stresses and they are included in Table 10. Strain measurements were not made on Series 1.4 specimens, but it was assumed that they were also misaligned in the same way as Series 1.5. Thus, the 34% average increase in local stress due to misalignment-induced bending in Series 1.5 was assumed to apply to Series 1.4. When considered in terms of this local stress, the results for the PPAW specimens are more in line with those from the MAG welds, and even in agreement with some of the high results obtained previously from TIG welds, as seen in Fig.17(b). It is possible that the single PPA weld macro-section used to measure the profile was not representative of all the specimens and in practice the weld toe radius was smaller in those that gave the lower fatigue strengths. Whatever the explanation, again there is no overwhelming evidence that the use of a different welding process, PPAW in this case, produces fillet welds with better fatigue strengths than MAG welds.

14


7.2.2. Duplex Plates with Longitudinal Fillet Welded Attachments The effect of welding process on the fatigue performance of longitudinal fillet welded duplex was also investigated, giving the results for MAG (Series 1.15, Table 20) and PPAW (Series 1.16, Table 21), as well as those from the TIG welds tested in the previous project (1). As in the case of the transverse PPA welds, it was noticed that some of the present type of PPA welded specimen were slightly distorted. Therefore, a check was made on the potential increase in stress due to misalignment-induced secondary bending by instrumenting one specimen (number 1.16-3) with strain gauges. This revealed a local increase in stress at the weld toe of around 14%, as indicated in Table 21. The results are all shown together in Fig.18(a), where it will be evident that there is no distinct influence of the welding process. In contrast to the results from transverse welds discussed above, there is excellent agreement between the various test series. Furthermore, the agreement extends to available published data (27,28) and regression analysis of all the data leads to a well-defined S-N curve. The mean 2SD curves enclosing the data are included in Fig.18(a). The picture is not changed significantly if allowance is made for the small misalignment-induced bending in the PPAW specimens, as seen in Fig.18(b). The PPAW results are plotted on the basis that the 14% increase in stress due to misalignment in specimen number 1.16-3 applies to every specimen. They are now higher than the MAG results, but still in agreement with the database as a whole. 7.3. FILLET WELDED TYPE 304L AUSTENITIC STEEL The test results obtained from the Type 304L austenitic steel are presented in Tables 11 -15. These were from TIG welded Series 1.6 (Table 11), MAG welded Series 1.7 and 1.8 (Tables 12 and 13) and PPA welded Series 1.9 and 1.10 (Tables 14 and 15). All three processes can be compared on the basis of tests carried out with Smax = 250MPa. Representing the MAG weld results with the scatter-band enclosing the reference data in Fig.16, the relevant results from TIG and PPA welds, from Series 1.6 and 1.10, respectively, are shown in Fig.19. It will be evident that there is no significant difference in fatigue performance between the MAG and PPA welded joints. This is consistent with the fact that their weld toe radii were virtually identical, as seen in Table 4. In contrast, it appears that the TIG welds did not perform as well as either the MAG or PPA welds. It is likely that this was due to weld profile differences. Detailed measurements were not made of the austenitic TIG weld profiles. However, measurements made on the macro-section shown in Fig.4 indicated that they had similar poor profiles to those produced in duplex Series 1.1 using essentially the same welding procedure, with a mean weld toe radius of only 0.5mm (Table 4). There had been reason to suppose that simply the use of TIG welding would offer some advantage in fatigue, due to the lower likelihood of introducing sharp weld toe discontinuities (5,17). However, the present results imply that a favourable profile is also required. Further tests are required to confirm this. At this stage, on the basis of the present data, there is no consistent indication that the use of PPA or TIG welding improves the fatigue performance of transverse fillet welds compared with the MAG process. This conclusion is supported by comparison of the present results with the scatter-band enclosing published data for fillet weld toe failure in austenitic stainless steels, from Fig.12(b), which is included in Fig.19. Although the present results for MAG and PPA welds tend to lie towards the upper bound of this scatter-band, virtually all of them are within it,

15


with the indication from the reference scatter-band that some MAG weld results lie just above it. As will be seen later, the relatively good fatigue performance of the present MAG welds has repercussions with regard to the beneficial effects of weld toe improvement techniques applied to them. The fatigue performance of MAG and Series 1.9 PPA welds can also be compared for the less severe loading of R = 0.1, as seen in Fig.20. In this case the extremely wide scatter in the PPA weld results hinders the comparison, although they are still reasonably consistent with the published data. However it is clear that they have not performed any better than the MAG welds. Indeed, the wide scatter could indicate a lack of consistency in the PPA welds, with variations in either profile or residual stress, 7.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL In the case of the thin austenitic Cr-Mn steel the comparison is between MAG and PPA welding. The relevant Series are 1.11 and 1.13, tested with R=0.1 and 1.12 and 1.14, tested with Smax held constant. The results are presented in Tables 16-19, and compared in Fig.21 and 22. The PPAW results in Fig.21, obtained with R = 0.1, are more widely scattered than those in Fig.22, obtained with Smax held constant, perhaps reflecting a wider variation in residual stress with this process. It will also be noted from Table 4 that there was a wide variation in local weld toe profile, the radius varying from around 0.6mm (sharp) to 3.5mm, and this was probably significant too. In contrast, the MAG weld profiles were very consistent (around 0.6 to 0.7mm radius) and the test results display little scatter. In spite of the generally poorer weld profiles in the MAG welds, there is good agreement between the results for the two welding processes, such that it was legitimate to combine them when fitting an S-N curve to those obtained with Smax constant. Thus, overall it may be concluded that there is no indication that the use of PPAW instead of MAG has improved the fatigue performance of this type of welded joint. 7.5. SIGNIFICANCE OF STEEL TENSILE STRENGTH The lack of influence of steel tensile strength on the fatigue lives of welded joints stems from the fact that their lives are dominated by fatigue crack growth and fatigue crack growth rate is independent of steel strength. However, the fatigue crack initiation process may benefit from steel tensile strength. In this respect, use of a welding process that reduces the severity of the weld toe stress concentration, from an improved profile, could increase the proportion of life spent initiating a crack. Then there may be some benefit from the use of high strength steel. This was examined on the basis of the present results obtained from TIG and PPA welded transverse fillet welds in the 10mm thick austenitic and duplex steels. Not surprisingly, since neither of these processes showed any improvement in fatigue performance compared with MAG welds, there was no significant difference between the results for the two steels. This is evident in Fig.23 for the TIG welds. The mean and mean 2SD lines from regression analysis of the combined data are included; analysis of just Series 1.1, which displayed less scatter than Series 1.6, gave virtually the same mean S-N curve. The results for the PPA welds are compared in Fig.24(a). They are too widely scattered to make it worth fitting an S-N curve and therefore the scatter-band enclosing the TIG results is included to provide a reference. It might be concluded that the performance of the austenitic steel specimens was better than that of the duplex specimens. However, it will be recalled

16


that the stress local to the weld toes in the duplex specimens was increased as a result of misalignment-induced secondary bending. Considering those test results in terms of the local stress, as in Fig.24(b), brings the results for the two steels together. Thus, again it can be concluded that the high strength duplex steel has shown no benefit over the lower strength austenitic steel. Finally, although this project only included tests on duplex longitudinal welded specimens in the as-welded condition, they can be compared with the results obtained from similar specimens in Type 304L austenitic steel from the previous project (1). All the results are shown in Fig.25. There is a tendency for the austenitic steel data to lie in the lower part of the scatter-band. However, consideration of available published data for both duplex and austenitic steel in the previous project showed that there was no significant difference between them. Thus, the variation in Fig.25 is more likely to reflect differences in the various weld geometries. On this basis, regression analysis was performed on the combined data in Fig.25, from specimens that failed from the weld toe, and the resulting mean and mean 2SD S-N curves are included. These will be used later as reference data for as-welded specimens of this type in either steel, for comparison with the results obtained from improved joints. 8. APPLICATION OF WELD TOE IMPROVEMENT TECHNIQUES 8.1. BACKGROUND Four weld toe improvement techniques were investigated in this project, namely burr grinding, re-melting by TIG or plasma (using PPAW equipment) and ultrasonic impact treatment. The aim in Phase 2 was to establish the basic improvement obtained under an applied stress ratio of R = 0.1 for the three types of weld detail included in the project. Apart from fatigue testing, the weld toe geometry and nature of the residual stress were examined in both as-welded and treated specimens for better understanding of the influence of the improvement techniques. The effect of each improvement technique was assessed in relation to the fatigue performance of the relevant as-welded specimen, as given by the mean S-N curve fitted to those results. However, in view of the good correlation between the results obtained at R = 0.1 0.63 from the Series 1.2, 1.3 and 1.7 transverse fillet welds in 10mm duplex and austenitic steels (Fig.16), the mean S-N curve fitted to the combined results and the mean 2SD scatter-band were used to represent the fatigue performance of as-welded specimens in both steels. Similarly, the reference data for longitudinal fillet welded specimens in either duplex or austenitic steel were those in Fig.25. An objective was to establish techniques that could achieve a 60% improvement in fatigue strength. The commonly accepted improvement, for example in some design rules that allow weld toe grinding (19), is 30%. As the present fatigue testing of improved specimens progressed it became apparent that this level of improvement was a more realistic target. This was primarily because there was a pronounced tendency for the specimens to fail in the parent plate, or even from the weld root, rather than from the improved toe, in lives that were below the 60% target. Another significant factor was the fatigue performance of the as-welded specimens. As noted earlier, this tended to be towards the upper bound of the scatter-band enclosing published data, or even above it in some of the transverse fillet welded specimens. This is significant because experience indicates that the benefit of a weld toe

17


improvement technique is greatest when it is applied to the lower fatigue performance welds (2-4). In view of this situation, the fatigue test results obtained from the improved specimens are compared with a target 30% improvement, corresponding to a factor of 1.3 applied to stress range. 8.2. TRANSVERSE FILLET WELDED TYPE S31803 DUPLEX STEEL The test results for the relevant Series, 2.1 to 2.4, are presented in Tables 22-25 and plotted in Fig.26 to 28. In each case, as noted above, the results are compared with the reference data for as-welded specimens in both duplex and austenitic steel (from Fig.16), represented by the mean 2SD to indicate the scatter in the lives of the as-welded specimens. The target S-N curve, based on achieving a fatigue strength 30% higher than that obtained from the mean S-N curve for as-welded specimens, is also included. The results obtained from Series 2.1, plotted in Fig.26, show that weld toe grinding improved the fatigue strength significantly. In fact, an improvement closer to 60% was achieved in this case. As is usually found, the improvement was greatest in the high-cycle regime with an indication that the fatigue endurance limit, assumed to correspond to N = 107 cycles on the S-N curve, could well have been increased from around 100MPa to more than 250MPa. The results obtained for the other two improvement methods that aim to improve the weld profile, TIG and plasma dressing, from Series 2.2 and 2.4 respectively, gave very similar results and therefore they are presented together in Fig.27. However, there is very little overlap of the two sets of data and combining them may not be valid. Indeed, considered together they appear to lie on an S-N curve that is steeper than that for the as-welded joint, which is the opposite of the usual effect of an improvement technique. Thus, it seems more reasonable to assume that if combined they should lie on a curve that is no steeper than that for the as-welded joint. On this basis, the S-N curve shown is the mean fitted assuming that the slope is the same as that for the as-welded joint. As will be seen, this lies well above the target 30% increase S-N curve, but some individual results are very close to the target. Thus, it can be concluded that TIG and plasma dressing achieve the target improvement in fatigue strength but with little margin. The results obtained from Series 2.3 UIT treated welds are plotted in Fig.28. In this case they follow an S-N curve that is essentially parallel to those for the as-welded joint. The level of improvement in fatigue performance is very similar to that seen for TIG and plasma dressing, with the mean S-N curve lying well above the 30% increase target, but some individual results close to the target curve. Thus, again it seems that the 30% target is achieved with little margin. 8.3. TRANSVERSE FILLET WELDED TYPE 304L AUSTENITIC STEEL The results obtained from the relevant test Series, 2.5 to 2.8, are presented in Tables 26 to 29, respectively. They are plotted in Fig.29 to 32 in comparison with the reference data for the as-welded specimens in both austenitic and duplex steel (from Fig.16). Each figure also includes the target S-N curve corresponding to a 30% increase in fatigue strength. Referring to Fig.29, the results for the toe ground Series 2.5 specimens are very widely scattered, some indicating an improvement in fatigue performance that far exceeds the 30%

18


level. This is especially true in the high-cycle regime, where the indication is that the fatigue limit could have been more than doubled. The same was found in the duplex welds discussed above. However, others are close to the 30% level of improvement, perhaps reflecting the fact that the weld toe radius produced by grinding these specimens was less than that produced in the duplex specimens (see Table 4). It will be evident that potential fatigue failure in the plate, in this case where it was gripped for test, provides a limit on the extent of improvement achievable, which was less than 30% in one case. Wedge grips indent the plate surface and thus provide notches from which fatigue cracks can propagate. It seems from these results that the fatigue strength of Type 304L austenitic steel plate with such notches is unlikely to be higher than that corresponding to the 30% target sought for the improvement techniques. A similar situation arose with Series 2.6 TIG dressed weld toes, as seen in Fig.30. However, in this case low fatigue lives were obtained even when plate failure was not due to gripping. Thus it seems that Type 304L austenitic plate might be vulnerable to plate failure even in the absence of notches as severe as those introduced by gripping. The results from Series 2.7 in Fig.31 show the effect of UIT treatment of the weld toe. Only four of the specimens failed from the weld toe, with a fifth failing in the parent plate. Apart from the plate failure result, all the others were above the target S-N curve, but only just in the case of two of the weld toe failures. Examination of these specimens did not reveal anything to explain their relatively low results. Clearly, the database is small and further tests may increase confidence in meeting the target improvement. The specimen that failed in the plate gave a low fatigue life, within the scatter-band enclosing the results for the as-welded specimens. Thus again there is an indication that there is limited scope for improving the fatigue lives of welded joints in Type 304L austenitic steel because of vulnerability to fatigue failure in the plate. Finally, the results obtained from Series 2.8, plotted in Fig.32, show that plasma dressing was less successful than TIG dressing. The results for TIG dressed Series 2.6 are included for ease of comparison. Although the results are widely scattered, it will be evident that there is no reason to conclude that plasma dressing has produced any significant improvement. The most likely reason for this was that the treatment did not improve the weld profile significantly. The weld toe radii were measured in some of these specimens (see Table 29). As indicated in Table 4, the average was found to be 3.6mm, around 60% of those for other plasma dressed weld toes and the toes of TIG dressed specimens that had achieved the target fatigue strength improvement. The two specimens that gave lives below the mean of the reference scatter-band for as-welded specimens had radii of only 2.5 and 3.2mm. Thus, it is not surprising that the treatment was less successful than expected. It is assumed that the plasma dressing procedure was in some way deficient when this series was treated. These results imply that plasma dressing must produce a minimum weld toe radius of the order of 6mm in order to achieve the target 30% improvement in fatigue strength. 8.4. TRANSVERSE FILLET WELDED AUSTENITIC CR-MN STEEL The relevant test results, Series 2.9 to 2.12, are presented in Tables 30 to 33 and plotted in Fig.33 and 34. The fatigue test results obtained from the appropriate as-welded specimens, Series 1.11 tested at R = 0.1, gave such a well-defined S-N curve with so little scatter that no clarity is lost by including them in each figure for direct comparison with the results from

19


improved specimens. As before, the fitted mean S-N curve and another drawn parallel to it but 30% higher on stress are included in the figures. Referring to Fig.33 and 34, it will be seen that only TIG dressing achieved the 30% improvement in fatigue strength from all the specimens tested. However, bearing in mind the relatively high fatigue strength of the as-welded joint at R = 0.1, past experience would suggest that this is to be expected. As noted earlier, the highest improvements are seen when weld toe improvement techniques are applied to low fatigue strength weld details. In the case of weld toe ground Series 2.9 (Fig.33), only a third of the specimens gave lives exceeding the target 30% improvement. It will be noted from Table 4 that the average weld toe radius after grinding, 1.64mm compared with 0.67mm for the as-welded joint, was still rather sharp compared with the profiles achieved with the thicker specimens. It is also possible that the depth of grinding (0.1mm mean but up to 0.3mm based on the macro-section in Fig.5(e)), which was less than recommended by the IIW(19), was insufficient. However, with such thin material it is clear that care is needed to ensure that the grinding is not so deep that the increase in nett section stress far outweighs the benefit of grinding. Nevertheless, it will be noted that, although the available database is small, there is a hint that the fatigue limit for the toe ground joint could be around 200MPa, in which case the effect of toe grinding of joints subjected to higher applied stresses is probably of academic interest only. As already noted, TIG dressing proved to be the most effective of the improvement techniques applied to welds in this steel, with all the Series 2.10 specimens tested exceeding the 30% improvement. This reflects the fact that the application of TIG dressing produced very favourable weld profiles, as seen in Fig.6. The radius of the weld profile in the region of the weld toe was typically 6mm, but effectively much higher at the toe itself. It will be noted from Table 31 or Fig.34 that none of the treated specimens failed from the weld toe. Instead, fatigue cracking was in the main plate, initiating at the plate edge or at the weld root. The occurrence of the latter mode of failure suggests that no further improvement in fatigue performance can be obtained from the present type of joint. However, the improvement is still very significant, especially in the high-cycle regime where the data indicate a possible three-fold increase in the fatigue limit. Figure 34 also includes the results obtained from Series 2.12 plasma dressed specimens. This technique was less effective than TIG dressing with most failures initiating at the plasma dressed weld toe and some tests not achieving the 30% improvement in fatigue strength. According to the weld profile measurements the average radius near the weld toe of 6.46mm was similar to that obtained by TIG dressing, but more variable (down to around 3mm). It is likely that the lower test results were associated with the sharper profiles. Even so, as with toe grinding, the results suggest that the fatigue limit could have been increased to around 250MPa. 8.5. LONGITUDINAL FILLET WELDED JOINTS IN TYPE S31803 DUPLEX STEELS AND

TYPE 304L AUSTENITIC Just two of the improvement techniques, plasma dressing and UIT, were investigated using 10mm thick longitudinal fillet welded specimens, in both S31803 duplex and 304L austenitic steels. The results obtained for duplex, Series 2.13 and 2.14, are presented in Tables 33 and 34. It will be noted that there were several cases of fatigue failure by crack growth from the weld root though the plate thickness, as illustrated in Fig.10(c), in both these series. All

20


the results for duplex are plotted in Fig.35, together with those for Series 1.15, as-welded duplex specimens also tested with R = 0.1. To put the results into a broader perspective, also shown is the scatter-band corresponding to the mean 2SD lines enclosing all the results obtained from duplex and austenitic steel plate specimens with longitudinal attachments obtained in the present and previous projects, as presented originally in Fig.25. The same reference data are compared with the results obtained from Series 2.15 and 2.16 in austenitic steel, as shown in Fig.36. The test results for those series are presented in Tables 35 and 36. In each figure, the target S-N curve, corresponding to a factor of 1.3 on stress range applied to the mean curve for the as-welded joint, is included. Referring to Fig.35 for duplex, both techniques have improved the fatigue strength by a factor of 1.3, more in most cases. There is also a hint that the fatigue limit has been increased by two to three times. Similarly, they were also effective when applied to the austenitic steel specimens, as seen in Fig.36. UIT was particularly effective in both steels, with only one specimen in each failing from the treated weld toe. However, plate failure from spatter has limited the improvement from plasma dressing and there is one case that did not achieve the 30% target improvement. It seems very likely that it was significant that the maximum applied stress in this case was equal to the yield strength of the steel. Noting this, it can be concluded that both techniques met the target 30% improvement in fatigue strength as long as the maximum stress was below yield. The specimens in both steels were vulnerable to failure from sites other than the weld toe, particularly the weld root or spatter on the plate surface, suggesting that further improvement is unattainable unless measures are taken to avoid such failures. One way to avoid crack initiation at the weld root is to use a full penetration weld to join the attachment, as recommended in (19). In practice it is usually sufficient for just the first 50mm or so of the attachment to be joined in this way. The spatter originated from the plasma dressing and clearly every effort should be made to avoid this, or to grind away any that does occur. 8.6. SIGNIFICANCE OF STEEL TENSILE STRENGTH Recalling the previous discussions of the possible effects of steel tensile strength on the fatigue behaviour of welds, it was expected that the effect was most likely to be seen in welds that had been treated with one of the post-weld improvement techniques. This possibility can be explored on the basis of the present results obtained from the 10mm thick transverse and longitudinal fillet welded specimens in types 304L austenitic and S 31803 duplex steel. The fatigue test results for the transverse welds are presented in Fig.37 to 39, in all cases in comparison with the reference S-N curves for as-welded specimens in either austenitic or duplex steel from Fig.16. Similarly, the results for the longitudinal welds are presented in Fig.40 and 41, again in comparison with the reference data for as-welded specimens (from Fig.25). Neglecting Fig.38 at this stage, the general indication is that there is no significant difference between the fatigue strengths of the improved specimens in the two steels. The comparison is hindered in some cases because there is virtually no overlap of the results, those for the austenitic steel having been obtained at lower applied stress ranges than the higher strength duplex. This reflects one difference between the results that does depend on the steel strength, in that any benefit from an improvement technique tends to decrease as the maximum applied

21


stress approaches yield. Consequently, the higher strength duplex still exhibits improved fatigue performance at stress levels above the yield strength of the austenitic steel. A possible exception to the above is seen in Fig.40, the results obtained from TIG and plasma dressed transverse fillet welds. I

fatigue of welded st st

Documents

austenitic steel

austenitic fillet welds

austenitic crmn steel

type s31803 duplex steels

tig dressing of type

duplex plates

plasma dressing of duplex

background i objectives