Fatigue of Friction Welds Manufactured in Air or Underwater

Fatigue Performance of Friction Welds Manufactured Both in Air and Underwater

Siak Manteghi
BP Exploration Operating Company Ltd Sunbury, United Kingdom

Dave Gibson
Proserv Westhill, Aberdeenshire, United Kingdom

Carol Johnston 
TWI Ltd, Granta Park Great Abington, Cambridge, CB21 6AL, UK

Paper presented at the 36th International Conference on Ocean, Offshore & Arctic Engineering OMAE2017 June 25-30, 2017, Trondheim, Norway

Abstract

Friction welding is being performed offshore in environments where arc welding may be difficult and where fatigue performance is critical. Friction welding underwater with Remotely Operated Vehicles (ROVs) can greatly reduce the cost of a project compared with using divers and arc welding because the support vessel, which is the major cost component in such an operation, is smaller. This paper describes two different programs of experimental work in which the fatigue endurance of friction welds were found to be better than that which could be expected from arc welded joints of similar geometry. The first program involved experimental work done with 25mm diameter steel bars. It found that, in the as-welded condition, friction welds have high fatigue strength. Residual stress measurements showed that this was due to a beneficial residual stress distribution in which compressive stresses are present at the surface adjacent to the failure site. Further evidence of this was obtained by subjecting some specimens to thermal stress relief. The fatigue strength of the stress relieved specimens was reduced compared with the as-welded joints but nevertheless the fatigue strength of these specimens was still high. The second program involved fatigue tests on friction stud welds in which the friction welding equipment was deployed offshore by divers or ROVs. The test specimens were made up of 19mm diameter studs friction welded onto structural steel plate. As with the first program, the specimens showed high fatigue endurance with results approximating to a DNV Class C1 curve. In some of the tests, the studs were preloaded in tension and results from specimens that were preloaded to the correct value specified for the joint were all stopped as run-outs, with specimens remaining unbroken.

1. Introduction

Friction welding is a solid phase welding process where there is no electric arc or liquid weld pool. One component, typically a bar or tube, is rotated at high speed and forced against a static component. Friction at the surface generates heat which raises the temperature at the interface to a level where the metals can flow plastically under pressure, typically 1,200°C for steel. Metal is extruded to the periphery of the weld forming an external “flash” and removing impurities from the weld interface. Rotation is then stopped but the force is maintained for short time to produce a solid phase forged weld.

Portable friction welding systems have been used offshore in situations where arc welding may be difficult such as welding underwater, welding on pipes containing fluids, through coatings or in potentially explosive atmospheres (Gibson, 2015). Knowledge of the fatigue performance of friction welds is important for many of these applications.

This paper first presents details of experimental work carried out at TWI (Manteghi, 1994) to investigate the fatigue strength of friction welds. It then compares this data to fatigue test data from welds made by underwater friction welding offshore (Hsu et al, 2005).

2. Experimental work

2.1 Material

The test specimens were fabricated from 25mm diameter solid bars conforming to BS 970 Grade 080 M40, a medium strength plain carbon steel which corresponds closely to the AISI, 1039 and 1040 steel grades. This material is used in general engineering applications requiring reasonable strength and resistance to wear. The material was quenched and tempered, and subsequently straightened by reeling through rolls prior to receipt.

2.2 Specimen Fabrication

All the specimens were produced at TWI using a continuous drive rotary friction welding machine, with the following specifications:

All welds were made in air.

The frictional and forge forces were set by a programmable hydraulic ram system. An initial ram force was set from the initial “touch point” to a desired “burn-off” distance. Once this distance was achieved, the chuck rotation was ceased and an increased forging ram force was applied and held to promote joint consolidation.

In order to ensure acceptable quality in the test joints, preliminary welding trials were carried out to determine the welding conditions. During these studies the rotational speed was maintained at 1,420 rev/min. However, the frictional force and forge force were varied. The quality of the welds produced by each given combination of welding parameters was assessed by visual examination, bend test, hardness survey, and (in the latter stages of the study) tensile strength. The studies led to the following conditions being selected:

Figure 1 shows a schematic of the specimen layout.

2.3 Flash Removal

The external flash produced during friction welding was subsequently machined-off in four of the specimens. Flash removal was performed in a lathe using a carbide tip cutting tool.

Ideally, with perfect alignment of the steel bars after welding, flash removal can result in a continuous cylindrical bar with a bonded area which is virtually indistinguishable from the other regions. This is achieved by removing all the flash so that the diameter of the bar in the bonded area is reduced to that of the plain unwelded bar.

Clearly, if significant misalignment exists, then a seemingly continuous bar cannot be achieved by flash removal alone. In particular, with axial misalignment a step-like discontinuity or “ridge” will become apparent as soon as all the flash is removed.

It is noteworthy that the four specimens studied in this programme were essentially continuous after flash removal, indicating that almost no misalignment (axial or angular) was present.

2.4 Visual and Macro-Section Examination

All specimens were examined visually prior to testing. The specimens with their flash removed were further examined by dye penetrant to ensure that they did not contain any surface breaking defects in the weld region.

A typical specimen, in the as-welded condition, was sectioned longitudinally. The surfaces were then polished, cleaned and etched in natal to reveal the profile of the flash, bond line and the HAZ; see Fig. 2.

2.5 Hardness Survey

A hardness survey of the regions of interest around the joint interface in an as-welded specimen was carried out using a Vickers diamond indenter with a 5kg applied load. The locations included in the hardness survey are defined, and shown schematically in Fig.3.

2.6 Post-Weld Heat Treatment

A group of eight specimens was stress relieved thermally, by heating in a furnace to approximately 600 °C +20 °C for one hour, and finally cooling in the furnace under a controlled cooling rate of less than 100 °C /hour.

One of these specimens was used later to study the residual stresses. The remainder were used for fatigue testing.

2.7 Residual Stress Measurements

Two specimens, one in the as-welded condition and the other after post-weld heat treatment to relieve the residual stresses, were instrumented with centre hole strain gauge rosettes and the residual stresses were determined as described by Beaney (1976).

The measurement locations, coinciding with the centres of the rosettes, are indicated schematically in Fig. 4. At each point the maximum and minimum principal stresses were calculated. The axial and circumferential components of the residual stress were also derived. In each specimen, the residual stresses were measured first at positions 1 to 5 while the flash was still in place. The flash was then machined off, following the procedure indicated above, and the residual stresses were measured at positions 6 to 10. Measurement 6 was on the weld centerline. Measurement 7 was the same distance from the centerline as measurements 1 and 2, which were made as close as possible to the weld. Measurements 8 to 10 were symmetrically opposed to measurements 3 to 5, and were made in order to determine the extent to which flash removal can relax the residual stresses remote from the joint line, both in the as-received and stress-relieved conditions.

2.8 Tensile Testing

Two specimens were tested to failure under static tensile loading in a Baldwin servohydraulic universal testing machine, with maximum capacity of 1750kN. The variation of load vs crosshead displacement was recorded.

2.9 Fatigue Testing

Fatigue tests were performed in air at ambient temperature (approximately 20°C) under constant amplitude sinusoidal axial loading. Each test was continued until specimen separation. All the tests were carried out in a Losenhausen hydraulic machine, with a maximum cycle loading capacity of +380kN. The tests were conducted under load control, at frequencies in the range 5 to 17Hz. The test machine was equipped with suitable wedge jaws which enabled the circular bars to be gripped at both ends. With the exception of the first few tests, all the specimens were wrapped in coarse emery paper at both ends before being gripped in the jaws. This had a marked effect in reducing the earlier tendency for frequent failures in the gripped area.

Twenty five specimens were tested in total. A datum S-N curve was generated first using ten specimens tested in the as-welded condition. For this series the applied stress ratio, R, i.e. the ration of the minimum to maximum applied cyclic stress, was kept unchanged at approximately 0.1.

Similarly, a group of seven specimens was used to generate an S-N curve for the stress relieved condition at R=0.1. The remaining eight specimens were used to study the influence of flash removal and applied R. The overall fatigue test matrix is summarized in Table 1.

3. Results of the experimental work

3.1 Static Strength

Two specimens tested under static tensile loading both failed in the material away from the weld area. This confirmed the success of the earlier trials aimed at achieving welding conditions necessary for acceptable weld quality. Ultimate failure loads of 399kN and 405kN were measured, which corresponded to ultimate tensile strengths of 788 and 799N/ mm² respectively, based on the measured cross-sectional area of the fractured bars. These values are within the range of plain material strength associated with this particular grade and condition of steel.

In summary, the static tensile tests are in line with the expectation that with proper attention to the welding parameters friction welds can be match the tensile strength of the plain unwelded material.

3.2 Hardness

All of the measurements were much below 300 HV5. This is true of the bond area and heat affected zone (HAZ), as well as the plain material, where the peak recorded measurements were 241, 199, and 246 HV5, respectively. Therefore the results of the hardness survey do not show any excessive increase in hardness arising from friction welding.

3.3 Residual Stress Measurements

The residual stress measurements in the as-welded and stress-relieved specimens are reported in Tables 2 and 3. They are plotted in Fig.5 and Fig.6, for the axial and circumferential components, respectively.

As the centre of the weld (i.e. at the centre of the bar) cools last it might reasonably be expected that that region would contain tensile residual stresses. Hence, the surface residual stresses associated with the friction weld would be expected to be compressive.

It is noteworthy that all the stress data reported in Fig.5 and Fig.6 were compressive. Of particular interest here is the magnitude of the axial residual stress near the weld, which are perpendicular to the plane of the fatigue cracks (see Fig.7c) thus influence their behavior.

Not surprisingly, the largest axial residual stresses were measured near the weld in the as-welded condition, with a maximum of approximately -200N/ mm². Elsewhere on the surface of the parent bar away from the weld, compressive residual stresses were smaller, but still significant.

Stress relief by PWHT did not eliminate the residual stresses completely: small compressive residual stresses remained in the welded region and on the surface of the parent bar, near the weld. This was true of both the axial and circumferential components.

Flash removal did not appear to have a consistent effect on the distribution of residual stresses. In any case flash removal is not expected to alter significantly the residual stresses away from the welded region. The asymmetry of the “flash removed” and “with flash” data near the weld in Fig.5 and Fig.6 can be explained in terms of the expected scatter in measurements of this kind.

It is also interesting that compressive residual stresses of the order of -200N/ mm² in the circumferential direction and ‑80N/ mm² in the axial direction were measured some 30mm from the weld centerline on the side with the flash intact, representing the as welded condition. Bearing in mind that the welding residual stresses are expected to decline fairly rapidly away from the weld so the measured compressive stresses in this specimen may well be, at least partly, due to other reasons, e.g. reeling.
The influence of the compressive surface residual stress on the fatigue strength of the friction welded joints is considered in the next section.

3.4 Fatigue Performance

The results of the fatigue tests are given in Tables 4, 5 and 6. They are also plotted in the form of an S-N diagram in Fig.8 where the mean and design lines corresponding to BS7608 Class B and C are also shown for comparison.

In all the specimens with flash intact (whether as-welded or stress-relieved) which failed in the welded area the fatigue cracks initiated on the surface at the edge of the weld, as indicated in Fig.7. In the specimens tested after flash removal it is not easy to determine the precise location of the fatigue initiation site, with respect to the original weld.

The stress range plotted in Fig. 8 is the nominal stress obtained by dividing the applied load by the measured cross-sectional area of the bar away from the welded joint. In practice some misalignment may exist between the two bars which would give rise to secondary bending stresses in the vicinity of the joint, even if the external load were applied axially. In the work reported here secondary effects might be expected to show as scatter in the fatigue results. Therefore it is encouraging that the specimens exhibited relatively little scatter. This is consistent with relatively small axial misalignment measured in the specimens with flash removed, and suggests that it may be feasible to achieve almost perfect alignment in friction welded components.

Regression analysis on the results obtained from the as-welded specimens at R=0.1 which failed results in a slope (m) of 6.46, an intercept (Log C) of 22.13 and a standard deviation of 0.143. This is considerably shallower than the S-N curve for Class B. Some of the  specimens remained unbroken after 10⁷ cycles at stress ranges up to 170N/ mm². If a sufficient number of specimens could be tested to define the fatigue limit (defined here as that corresponding to 10⁷ cycles), it may well be expected to exceed 170N/ mm² for the particular welds tested here.

The test results were considerably better than could be reasonably expected of a similar joint produced by arc-welding. For instance, if a bar had been produced by full-penetration butt welding it would be assigned to Class D or E in BS 5400 Part 10 (the pre-cursor to BS7608), depending on the degree of control exercised during welding, welding position and weld process. By contrast, the data from the friction welds lie above the mean BS7608 Class B curve. With the shallow slope associated with the friction weld data, their superior performance compared with Class D increases as the applied stress range is reduced.

The shallow slope of the S-N curve for the friction welds is consistent with the presence of compressive residual stresses at the surface of the specimen in the weld region, where the fatigue cracks initiated. Additionally, it could be explained in terms of a significant initiation phase in the total fatigue life of the specimens. This would be in contrast with typical arc-welded joints where, even in welds of acceptable quality, it is reasonable to expect crack-like discontinuities such as slag intrusions, undercut, etc. at the toe (Signes et al., 1967). The presence of these initial flaws in arc-welded joints essentially removes any significant crack initiation phase and the majority of fatigue life is occupied in crack propagation.

Further studies, including micro-sectioning and examination under high magnification are required to investigate the presence, or otherwise, of non-metallic initial flaws in friction welded bars. This might determine whether the shallow slope of the S-N curve can be attributed, at least in part, to the presence of a significant initiation phase.

To demonstrate the effect of residual stresses more clearly, a group of seven specimens was fatigue tested after stress relief by PWHT. As expected, by reducing the beneficial i.e. compressive residual stress, the fatigue strength of the specimen was adversely affected. The reduction in fatigue strength was particularly noticeable at the lower stress ranges. For instance at 300N/ mm², an as-welded specimen had remained unbroken after nearly 1.2 x 10⁶ cycles, whereas a stress relieved specimen failed after less than 300,000 cycles. The results suggest that Class C might be the relevant design curve for stress relieved joints, but further tests are required to confirm that.

The presence of beneficial residual stresses, some of which may be due to reeling, and the favourable weld quality featuring in the test programme highlighted the need to exercise caution when using data reported here as a basis for the design of friction welds. Further data, particularly including less favourable, although realistic friction welds are needed before any firm design guidance can be formulated.

The effect of stress ratio is also shown in Fig. 8. Although the number of tests is limited, a clear pattern is demonstrated, with results obtained under fully reversed loading appearing significantly stronger than those obtained at R= 0.1. This is consistent with the fact that, with all other factors such as residual stress, etc. being the same, a smaller proportion of the applied loading is tensile (and thus damaging) at R= -1.

The flash produced during friction welding is not always removed in practice. However, there are cases where flash is removed despite the extra effort, and cost, involved. For instance, in some applications this is done in an attempt to reduce the risk of corrosion by eliminating the crevice-like geometry associated with the flash. In others flash removal is undertaken prior to post weld heat treatment, or merely for aesthetic reasons. Regarding the effect of flash removal on fatigue, the data so far suggests that it is beneficial giving improvement to fatigue strength. This is not surprising, as the discontinuity at the edge of the flash in an as-welded specimen acts as a source of stress concentration. The extent of improvement appears to increase at lower stress ranges. However, more data are required to establish whether the fatigue enhancement can be achieved consistently. In particular, if the as-welded specimens include large axial misalignments, the “ridge” effect, described above will be magnified after flash removal and may well act as a fatigue initiation sight. More work is required to clarify this issue.

3.5 Concluding remarks on the experimental work

The fatigue strength of friction welded steel bars was investigated. The weld parameters were selected in order to ensure quality on the basis of visual examination, bend tests, hardness surveys and tensile testing. The work has led to the following conclusions:

1. The fatigue test results obtained from as-welded specimens at R=0.1 exhibited little scatter, and specimens remained unbroken after 10⁷ cycles at stress ranges up to 170 N/ mm². This is well in excess of the corresponding value for mean Class B in BS7608, which is 100 N/ mm².
2. The mean S-N curve representing the as-welded specimens tested at R=0.1 had a slope of 6.46, much shallower than that of Class B. The shallow S-N curve and high fatigue limit are consistent with the beneficial (i.e. compressive) residual stresses measured at the surface of the specimens near the weld.
3. Stress relief by PWHT had a detrimental influence on fatigue strength. This is as expected and arises from the relaxation of beneficial compressive residual stress. The results obtained to date suggest that Class C might be the appropriate design curve.
4. As-welded specimens tested under reversed loading (R=-1) had significantly higher fatigue strengths (stress range) than those tested at R=0.1.
5. Limited results obtained to date suggest that flash removal increases fatigue strength. The extent of improvement increases with reduction in applied stress.

4. Friction welds made underwater

4.1 Offshore Friction Welding Systems

Friction welding equipment designed for use offshore is lighter and more compact than the static machines used in onshore manufacturing. The equipment used underwater is remotely operated by technicians on the surface and relies on divers or Remotely Operated Vehicles (ROVs) to position it (Blakemore, 2000). A clamp, consisting of either magnets, suction cups or a mechanical clamp is required to react against the axial force produced during friction welding (Fig. 9 and Fig. 10). The hydraulic power to the tool is either provided by the ROV or for diver operation from a hydraulic power pack on the surface via an umbilical.

4.2 Offshore Friction Weld configurations

Offshore friction weld configurations are typically friction stud welds (Fig. 11) onto flat or curved surfaces and are typically required for retrofitting structural attachments or cathodic protection. The weld is performed with a small shroud covering the weld area which reduces the cooling of the weld by the surrounding water. The flash is normally left in place and there is no post weld heat treatment.

4.3 Comparison of laboratory results with underwater friction welds

A test programme was designed to qualify the fatigue performance of 19mm studs friction welded onto 20mm thick EH36 plate, underwater, as described in detail in Hsu et al (2005). The studs were required to install bumpers on the hull of a SPAR to minimize riser movement in the moon pool, in situ offshore.  After friction welding the studs were tensioned to hold the bumpers in place. The objective was to verify that the friction welds had at least BS7608 Class F2 performance.

Three types of specimens were manufactured by underwater friction welding and subjected to fatigue testing. The first series of specimens represented the worst case scenario, with no preload on the stud. In the second series of specimens, most of the studs were correctly pre-tensioned. Two of these specimens were pre-tensioned to 46% and 80% of the required value. The third series was a loading scenario in which the plate experienced global bending. The results are summarized in Fig 12.

The tests were carried out at a constant mean stress and various stress ranges. In order to aid comparison with the experimental work, the stress ranges have been corrected using the Goodman correction to convert them to an R ratio of 0.1. They are plotted with the results from the laboratory test programme in Fig 13. It can be seen that the underwater friction welds have performance that is similar to those results from the laboratory tests. The studs which were correctly preloaded were all runouts, and all of the tests were stopped above DNV Class C1 design. The beneficial effect of pre-tensioning was shown by two specimens that were tensioned to 46% and 86% of the specified installation value. These were not run outs but nevertheless failed at values above a DNV C1 curve.

The data from these tests reinforce the conclusions from the experimental work, and show that friction welds in the as-welded condition can have much better fatigue strength than that which would be expected from equivalent arc welds.

5. Discussion

Concern is sometimes expressed that the profile of friction welds in the as-welded condition with the flash present might result in poor fatigue strength compared with arc welds. The experimental work reported here indicates that this is not the case and that this is due mainly to the presence of beneficial compressive residual stresses at the weld.

Fatigue data from underwater friction stud welds show similar fatigue performance, so both datasets suggest that friction welds can have much better fatigue strength than the equivalent arc welds.

The extension of the life of offshore assets often requires repairs and retrofit operations underwater. The major cost element in such an operation is usually the provision of a support vessel for the work offshore.  Support vessels for ROV operations are usually smaller, much less costly and often more readily available than those for operations with divers. This is because the vessel does not require a diving system and the requirements for the dynamic positioning system are to a lower specification for safety reasons.

Friction welding is a mechanised process that readily lends itself to remote operation, particularly from ROV’s were the use of underwater arc welding processes such as wet shielded metal arc welding or dry habitat welding would be impractical.  Furthermore, detrimental effects occur with arc welding processes due increased water depth (hydrostatic pressure) on changes in arc voltage and weld pool chemistry. These effects are not present with friction welding. The results show that there is a potential cost saving from the use of subsea friction welded offshore. and a number of subsea friction welding projects have now been successfully performed offshore with ROVs (Gibson 2015).

6. Conclusions

• Results from fatigue tests on friction welded bars carried out at R = 0.1 in the as-welded condition showed performance above BS7608 Class B mean. The slope was shallower than Class B (at 6.46).
• Residual stress measurements showed that the high fatigue strength was due to the presence of compressive residual stresses at the surface in the vicinity of the weld.
• Tests carried out on friction welded bars with the flash removed indicate that flash removal does improve fatigue strength slightly.
• Published data on friction welds made underwater offshore were compared to the results obtained in the laboratory experiments. The underwater welds had similar fatigue strength to the laboratory-made welds.

7. Acknowledgments

The experimental work described here was supported by the Industrial Members of TWI.

8. References

1. Beaney E.M. “Accurate measurement of residual stress on any steel using centre hole method”, Strain, Vol.12, No.3, July 1976, pp. 99-106.
2. Blakemore Gordon, “Back to the Future – Underwater Repair by Friction Welding”, Underwater Intervention 2000, Conf., Houston Texas 24-26 January 2000.
3. BS 5400: “Steel, Concrete and composite bridges – Part 10. Code of practice for fatigue”. British Standards Institution 1980.
4. BS 7608: “Fatigue design and assessment of steel structures”, code of practice. British Standards Institution 2014.
5. DNVGL, 2016: ‘DNVGL-RP-C203, Fatigue design of offshore steel structures’ Det Norske Veritas, Norway.
6. Gibson Dave, Institute of Corrosion Aberdeen Branch. Presentation at the January 2015 Branch Meeting. ICorr Presentation Rev02 207012015.pdf  ICorrAbz/resources-center
7. Hsu T M, Buitrago Jamie, Herman Arthur and Mc Keighan Peter C.,. “Fatigue performance of friction welded studs”, OMAE2005-67209, June 12-17 2005, Halkidi Greece.
8. Manteghi S., “Some fatigue tests on friction welded steel bars”, TWI Report 485/1994, July 1994.

Table 1 Test matrix showing the number of fatigue tests performed at each combination of specimen and load conditions.

Table 2 Residual stresses measured in an as-welded specimen

Table 3 Residual stresses measured in a stress relieved specimen

Notes on tables 2 and 3

1 For gauge locations, see Figure 4.
2. Gauges 1 to 5 are on bar in the as-received condition. Gauges 6 and 10 were attached after flash was machined off.
3. x direction = axial

Table 4 Fatigue test results from the as-welded specimens

Table 5 Fatigue test results from the specimens with flash removed

Table 6 Fatigue test results from the stress relieved specimens